Word Formation

PR1
Word-Formation
HSK 40.l
PR2
HandbUcher zur
Sprach- und Kommunikationswissenschaft
Handbooks of Linguistics
and Communication Science
Manuels de linguistique et
des sciences de communication
MitbegrOndet von Gerold Ungeheuer
Mitherausgegeben (1985-2001) von Hugo Steger
H erausgegeben von I Ed ited by

Herbert Ernst Wiegand
Band 40.1
De Gruyter Mouton
Edites par
PR3
Word-Formation
An International Handbook
of the Languages of Europe
Volume 1
Edited by
Peter 0. MUiier
Ingeborg Ohnheiser
Susan Olsen
Franz Rainer
De Gruyter Mouton
PR4
ISBN 978-3-11-024624-7
e-ISBN (PDF) 978-3- 11-024625-4
e-ISBN (EPUB) 978-3-11-039320-0
ISSN 1861-5090
Library of Congress Cataloging-in-Publication Data

A CTP catalog record for this book has been applied for at the L ibrary of Congress.
Bibliographic information published by the Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationailbibliografie;
detailed bibliographic data are available on the Internet at http://dnb.dnb.de.
2015 Walter de Gruyter GmbH, Berlin/Boston
Typesetting: Meta Systems Publishing & Printservices GmbH, Wustermark
Printing and binding: Hubert & Co. GmbH & Co. KG, Gottingen
Cover design: Martin Zech, Bremen
@Printed on acid-free paper
Printed in Germany
www.degruyter.com
PR16
Contents
XVI
11 1. Units and processes in wo rd-f o rm atio n II: Specia l cases

28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
Affective palatalization in Basque Jose Ignacio Hualde

Parasynthesis in Romance David Serrano-Dolader
Affix pleonasm Francesco Gardani . . . . . . . .
Interfixes in Romance Michel Roche . . . . . . .
Linking elements in Germanic Nanna Fuhrhop and Sebastian Kurschner
Synthetic compounds in German Martin Neef . . . . . . .
Verbal pseudo-compounds in German Christian Fortmann
Particle verbs in Germanic Nicole Dehe
Particle verbs in Romance Claudio lacobini
Particle verbs in Hungarian Maria Ladanyi .
Noun-noun compounds in French Pierre J. L. Arnaud
Verb-noun compounds in Romance Davide Ricca
Co-compounds Bernhard Walchli . . . . . . . . .
Multi-word units in French Salah Mejri . . . . . .
Multi-word expressions and univerbation in Slavic Olga Martincova
Compounds and multi-word expressions in Slavic Ingeborg Ohnheiser
Paradigmatically determined allomorphy: the "participial stem" from Latin
to Italian Anna M. Thornton . . . . . . . . . . . . . . . . . . . . . .. . . . .
Volu me 2
IV. Rules and restrictio ns
Ge nera l aspects
in
word-formation I:
45.
46.
47.
48.
Rules, patterns and schemata in word-formation Heike Baeskow

Word-formation and analogy Sabi ne Arndt-Lappe
Productivity Livia Gaeta and Davide Ricca
Restrictions in word-formation Livio Gaeta
V.
Rules and rest rictions in word-formation 11 :

Spec ial cases
49.
50.
5 1.
52.
Argument-structural restrictions on word-formation patterns Holden Hartl

Phonological restrictions on English word-formation Renate Raffelsiefen
Morphological restrictions on English word-formation Lothar Peter
Semantic restrictions on word-formation: the English suffix -ee Heike
Baeskow
53 . Dissimilatory phenomena in French word-formation Marc P lenat
54. Closing suffixes Stela Manova
55. Closing suffix patterns in Russian Dmitri Sitchinava
517
524
537
551
568
582
594
611
627
660
673
688
707
727
742
757
780
PR17
Contents
VI. Semantics and pragmatics 1n word-formation I:

General aspects
56.
57.
58.
59.
60.
61.
62.
Motivation, compositionality, idiomatization Daniela Marzo

Word-formation and folk etymology Sascha Michel
Categories of word-formation Volkmar Lehmann
Schemata and semantic roles in word-formation Hanspeter Ortner and
Lorelies Ortner
Word-formation and argument structure Manfred Bierwisch
Word-formation and metonymy Manfred Bierwisch
The pragmatics of word-formation Lavinia Merlini Barbaresi
VII. Semantics and pragmatics in word-formation II:

Special cases
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
75.
76.
77.
78.
79.
80.
Noun-noun compounds Christina L. Gagne and Thomas L. Spalding

Gender marking Ursula Doleschal
Singulatives Paolo Acquaviva
Collectives Wiltrud Mihatsch
Action nouns Maria Koptjevskaja-Tamm
Action nouns in Romance Livio Gaeta
Verbal nouns in Celtic Paul Russell
Nominalization in Hungarian Tibor Laczk6
Result nouns Chiara Melloni
Quality nouns Franz Rainer
Status nouns Hans Christian Luschiitzky
Agent and instrument nouns Franz Rainer
Patient nouns Susanne Miihleisen
Place nouns Bogdan Szymanek
Intensification Franz Rainer
Negation Marisa Montero Curiel
Negation in the Slavic and Germanic languages Jozef Pavlovic
Spatial and temporal relations in German word-formation Ludwig M.
Eichinger
81. Adverbial categories Davide Ricca
82. Denominal verbs Andrew Mcintyre
83. Valency-changing word-formation Dieter Wunderlich
84. Word-formation and lexical aspect: deverbal verbs in Italian N icola
Grandi
85. Word-formation and aspect in Samoyedic Beata Wagner-Nagy
86. Verbal prefixation in Slavic: a minimalist approach Petr Biskup and
Gerhild Zybatow
87. Denumeral categories Bernhard Fradin
88. The semantics and pragmatics of Romance evaluative suffixes Martin
Hummel
89. Morphopragmatics in Slavic Alicja Nag6rko
XVII
PR18
Contents
XV Ill
Volume 3
VIII. Foreign word-formation, language planning and p urism I:
General aspects
90. Types of foreign word-formation Wieland Eins
91. Word-formation in Neo-Latin Thomas Lindner and Franz Rainer
92. Foreign word-formation, language planning and purism Wolfgang Pockl
IX. Foreign word-formation, language planning and purism II:

Special cases
93.
94.
95.
96.
97.
98.
99.
100.
101.
Foreign word-formation in German Peter 0. Muller

Foreign word-formation in English Klaus Dietz
Foreign word-formation in Italian Claudio Iacobini
Foreign word-formation in Polish Krystyna Waszakowa
Word-formation and purism in German Mechthild Habermann
Word-formation and purism in French Petra Braselmann
Word-formation and purism in Croatian Branko Tofovic
Word-formation and language planning in Estonian Virve Raag
Individual initiatives and concepts for expanding the lexicon in Russian
Wolfgang Eismann
X.
Historical word-formation I: General aspects
102. Mechanisms and motives of change in word-formation Franz Rainer

I 03. C hange in productivity Carmen Scherer
XI. Historica l word-formation 11: Special cases

104. Grammaticalization in German word-formation Mechthild Habermann
105. The grammaticalization of prepositions in French word-format ion Dany
Amiot
106. The Romance adverbs in -mente: a case study in grammatical ization U lrich
Detges
107. Grammaticalization in Slavic word-formation Krystyna Kleszczowa
108. The origin of suffixes in Romance David Pharies
XI I. H istorica I word-formation 111: Language sketches

I 09.
110.
111 .
112.
Historical word-formation in German Peter 0. Muller

Historical word-formation in English Klaus Dietz
From Latin to Romance Eva Buchi and Jean-Paul Chauveau
From Latin to Romanian Marina Radu lescu Sala
PR19
Contents
113.
114.
115.
116.
117.
118.
119.
XIX
From Old French to Modern French Franz Rainer and Claude Buridant
From Old Irish to Modern Irish David Stifter
Historical word-formation in Slavic Swetlana Mengel
From Ancient Greek to Modern Greek lo Manolessou and Angela Ralli
The history of word-formation in Uralic Johanna Laakso
From O ld Hungarian to Modern Hungarian Tamas Forgacs
Historical word-formation in Turkish Claus Schonig
XI 11. Word-formation in language acquisition and aphasia

120. Word-formation in first language acquisition Hilke Elsen and Karin
Schlipphak
12 1. Word-formation in second language acquisition Cornelia Tschichold and
Pius ten Hacken
122. Word-formation in aphasia Carlo Semenza and Sara Mondini
XIV. Word-formation and language use

I 23.
124.
I 25.
126.
127.
128.
129.
I 30.
Word-formation
Word-formation
Word-formation
Word-formation
Word-formation
Word-formation
Word-formation
Word-formation
and
and
and
and
and
and
and
and
text Anja Seiffert

brand names Elke Ronneberger-Sibold
planned languages Klaus Schube1t
sign languages Ronnie B. Wilbur
technical languages Ivana Bozdechova
literature Peter Handler
orthography Hannelore Poethe
visuality Lore lies Ortner
XV. Tools in word-formation research

131. Dictionaries Renate Belentschikow
132. Corpora Ulrich Heid
133. Internet Georgette Dal and F iammetta Namer
Volume 4
XVI. Word-formation 1n the individual European languages
lndo-European
Germanic
134. German Irmhild Barz

135. English lngo Plag
PR20
Contents
xx
136.
137.
138.
139.
140.
141.
142.
143.
Dutch Geert Booij

Frisian Jarich F. Hoekstra
Yiddish Simon Neuberg
Farnese Hjalmar P. Petersen
Danish Hans Gotzsche
Norwegian John Ole Askedal
Swedish Kristina Kotcheva
Icelandic Porsteinn G. Indrioason
Romance
144.
145.
146.
147.
148.
149.
150.
151.
Portuguese Bernhard Poll

Spanish Franz Rainer
Catalan Maria Teresa Cabre Castellvi
French Franck Floricic
Ladin Heidi Siller-Runggaldier
Sardinian Immacolata Pinto
Italian Franz Rainer
Romanian Maria Grossmann
Celtic
152. Breton Elmar Ternes

153. Welsh Paul Russell
154. Irish Brian 6 Curnain
Slavic
155.
156.
157.
158.
159.
160.
161.
162.
163.
164.
165.
t 66.
167.
168.
Upper Sorbian Anja Pohontsch

Polish Alicja Nag6rko
Kashubian Edward Breza
Czech Ivana Bozdechova
Slovak Ma1tina Ivanova and Martin Olostiak
Ukrainian Ievgeniia Karpilovska
Belarusian Ataksandr Lukasanec
Russian Igor' S. Uluhanov
Slovene Irena Stramljic Breznik
Croatian Mario Grcevi6
Serbian Bofo Coric
Bosnian Branko Tofovi6
Bulgarian Cvetanka Avramova and Julia Baltova
Macedonian Lidija Arizankovska
Map of Languages
PR21
Contents
Volume 5
Baltic
169. Lithuanian Bonifacas Stundzia

170. Latvian Agne Navickaite-Klisauskiene
Albanian
17 1. Albanian Monica Genesin and Joachim Matzinger
Greek
172. Greek Angela Ralli
Inda-Iranian
173. Ossetic David Erschler

174. Tat Gilles Authier
Ura lie
175.
176.
177.
178.
179.
180.
181.
Nenets Beata Wagner-Nagy

Finnish Kaarina Pitkiinen-Heikki lii
Estonian Krista Kerge
Permic Laszlo Fejes
Mari Timothy R iese
Mordvinic Sandor Maticsak
Hungarian Ferenc Kiefer
Basque
182. Basque Xabier Artiagoitia, Jose Ignacio Hualde and Jon Ortiz de Urbina
Semitic
183 . Maltese Joseph Brincat and Manwel M ifsud
Turkic
184. Turkish Jens Wilkens

185. Bashkir Gu lnara Iskandarova
186. Tatar Llszl6 Karoly
XX!
PR22
Contents
xx11
187.
188.
189.
190.
Crimean Tatar Henryk Jankowski

Gagauz Astrid Menz
Karaim Eva A. Csat6
Chuvash Galina N. Semenova and Alena M. Ivanova
Mongolic
191. K almylk D anara Suseeva
Northwest Caucasian
192. Abkhaz Viacheslav A. Chirikba

193. Adyghe Yury Lander
194. Kabardlian Ranko Matasovic
Northeast Caucasian
195.
196.
197.
198.
199.
200.
201 .
202.
203 .
204.
205.
206.
207.
Rutul Mikhail Alekseyev

Budugh Gilles Authier and Adigozel Haciyev
Udi Wolfgang Schulze
Aghul Timur Maisak and Dmitry Ganenkov
Arc hi Marina Chumakina
Khinahug Wolfgang Schulze
Lak Wolfgang Schulze
D argwa Nina Sumbatova
Bezhta Madzhid Kbali lov and Zaira Khalilova
Botlikh Mikhail Alekseyev
Akhwakh Denis Creissels
Avar Madzhid Khalilov and Zaira Khalilova
Khwarshi Zaira Khalilova
Subject index
Map of languages
PA1
I. Word-formation as a linguistic discipline

1. The scope of word-formation research
1.
2.
3.
4.
5.
6.
7.
8.
9.
Introduction
Morphological building blocks and the internal structures of complex lexemes
Word-formation patterns
Approaches to word-formation research
Levels of analysis and description in word-formation research
Theoretical models of word-formation
Modelling dynamic aspects of word-formation: productivity and lexicalization
Conclusion
References
Abstract
The first article of this volume presents an introductory survey of the scope of wordformation research. It defines and demarcates the subject-matter of word-formation and
explains the basic notions related to the internal structures of complex lexemes and the
cross-linguistically important word-formation patterns. Major approaches, analytical
and descriptive levels and models in the field of word-formation research are outlined
from a bird's eye view. The final section deals with productivity and lexicalization.
1. Introduction
Word-formation research investigates the patterns and regularities underlying the formation of complex lexemes by means of existing building blocks with the aim of formulating rules and other types of generalizations. Complex lexemes (e.g., E. headteacher or
trivialize) are characterized by the fact that they consist of two or more constituents.
Unlike most simple lexemes such as head, teach and trivial, complex lexemes are not
entirely arbitrary signs, but instead are morphologically motivated by their constituents
and by the semantic links shared with other stmcturally identical formations. A precise
understanding of the nature of this motivation forms the main interest of word-formation
research.
The scope of word-formation research in linguistics can be defined by demarcating
word-formation from neighbouring fields. The adjacent domain of inflectional morphology deals with elements and operations which produce word-forms of lexemes (e.g.,
teaches, teaching, taught) rather than new lexemes (teacher, headteacher, to teamteach,
etc.), as is the case in word-formation. Word-formation and inflectional morphology are
not separated by a clear boundary, however (Bybee 1985; Scalise 1988; Plank 1994;
Booij 2000; Stump 2005; see article 14 on the delimitation of derivation and inflection).
A classic example of a delimitation problem is the dispute over whether the English
PA2
2
adverb-forming suffix -ly as in really or elegantly should be treated as a derivational,

i.e. lexical, or inflectional and thus grammatical morpheme (cf. Giegerich 2012). Syntax,
while having emerged as a prominent source of inspiration for theory-building in wordformation, differs from word-formation in that the output of syntactic operations is phrasal and clausal rather than lexical in nature. Needless to say, boundary issues exist as
well, e.g., in the distinction of certain types of nominal compounds from noun phrases
(e.g., Benveniste 1967; Bauer 1988a; Olsen 2000; see articles 20 on composition, 38 on
noun-noun compounds in French and 135 on English). Demarcation problems are also
very common at the porous boundary to phraseology, for instance when it comes to
classifying semi-idiomatic phrases such as black market or particle verbs of the type get
up and make up for (see articles 23 on particle-verb formation and 24 on multi-word
expressions). Many practitioners of word-formation research distinguish word-formation
(in a narrow sense) from ways of extending the lexical resources which do not involve
changes in the forms of linguistic signs, mainly metaphorical or metonymic transfers
and other forms of lexical change resulting in purely semantic extensions or shifts (see
article 61 on word-formation and metonymy). Finally, as suggested by the definition
given above, word-formation can be, but is not always, distinguished from what is referred to as coinage (Lieber 2010: 51), word-creation or word manufacture (Bauer 1983:
239), which does not rely on existing building blocks (see article 26 on word-creation).
Frequently quoted examples include product and brand names such as Kodak or Google.
The subject-matter of word -formation research is also demarcated by the definition
provided above. Essentially, four aspects define the rem it of word-formation research.
Firstly, word-formation research analyses and describes the internal structures and
constituents of complex lexemes and identifies and classifies the forms and meanings of
the lexical and morphological building blocks of a given language (see article 15 on
units of word-formation). As a part of this segmentation, identification and classification
procedure, the lexical bui lding blocks involved in word-formation rnus.t be distinguished
from inflectional morphemes (see article 14 on the delimitation of derivation and inflection). The results of analytical and classificatory efforts feed into models of word-formation processes.
Secondly, word-formation research identifies, classifies and models the processes underlying the formation of existing and new complex lexemes (see chapters II and III of
the handbook). This is typically accomplished by segmenting established complex lexemes and describing their grammatical, morphological, semantic and phonological properties as well as those of their constituents. While most researchers in the field agree on
a set of basic types of word-formation processes, there has been considerable controversy
over the precise way in which they should be modelled.
Thirdly, because of the multifaceted nature of complex lexemes, word-formation research tends to be a multi-level (Lipka 1983) or multi -perspectival endeavour. Traditionally, morphological, syntactic, semantic and phonological aspects have taken centre stage
in word-formation research, as these perspectives provide the basis for systematic and
parsimonious generalizations regarding word-formation rules and patterns. More recently, sociopragrnatic, psycholinguistic, cognitive and textual aspects have been attracting
increasing interest (see article 13 on word-formation in psycholinguistics and neurocognitive research and chapter XIV of the handbook).
Fourthly,. word-formation research tries to provide adequate models of the creative
and dynamic aspects of word-formation. On the level of word-formation rules and pat-
PA3
terns, this relates to the changing productivity of word-formation processes and the elements involved in them (see articles 4 7 on productivity and 103 on change in productivity). On the level of individual complex lexemes, an explanation must be found for how
new creations are motivated, how they find their way into the lexicon of a language and
how their forms and meanings change in the course of time (see article 56 on motivation,
compositionality, idiomatization).
As regards the terminology used to refer to the core interest of word-formation research, the terms word-formation process, word-formation type, word-formation model,
word-formation rule and word-formation pattern will be used interchangeably here, even
though they highlight different aspects of the phenomena at hand and have been defined
in more specific ways by some authors (e.g., Fleischer and Barz 2012: 67-69; Hansen
et al. 1990: 28).
Section 2 of this article will be devoted to the morphological building blocks and
internal structures of complex lexemes. Section 3 will provide a survey of cross-linguistically important word-formation patterns. The next three sections wi ll discuss different
approaches to word -formation research (section 4), survey the linguistic perspectives
included in word-formation research (section 5) and provide a sketch of the major types
of theoretical models attempting to capture the nature of word-formation m ies and patterns (section 6). Section 7 will focus on research into the temporal and dynamic aspects
of word-formation, i.e. the productivity of word-formation patterns and the types of
changes that take p lace as complex lexemes are coined, spread and become part of the
lexicon.
2. Morphological building blocks and the internal structures

of complex lexemes
Three basic approaches to describing the constituents of complex lexemes can be distinguished: a word -based, a root-based and a morpheme-based approach (see article 15 on
units of word-formation).
The first type proceeds from the assumption that words constitute the cores of complex lexemes. Word-formation rules combine several words in the case of compounding,
and words and affixes in derivation. This assumption is known as the word-based hypothesis (Aronoff 1976: 21; Scalise 1986: 40-42, 71 - 78).
In the second approach, roots or stems, rather than full-fledged words, are considered
the key constituents of complex lexemes. Terminology is far from uniform in this area,
especially regarding the term stem (see Chelliah and de Reuse 20 11 : 314 for a survey).
This notion is sometimes used to refer to the base of a word-formation process, i.e. the
e lement to which further morphological material is added, and sometimes to the part
which remains constant before inflectional endings are added (Bauer l 988b: 11). According to the first interpretation, national would be the stem of (the company was) nationalized, and according to the second, nationalize. To keep these readings apart, some authors
(e.g., Hansen et al. 1990: 41) distinguish the word stem of a lexeme, qua word-form
minus inflectional affix, from the word-formation stem, i.e. the base of the (final) wordformation process. Whether words or stems are more useful as basic units of wordformation research may well depend on the language being investigated (Bloomfield
1933: 224- 226; Kastovsky 1999).
PA4
4
The third approach relies on the notion of morpheme (Baudouin de Courtenay 1895:
10), usually defined as the smallest meaning-bearing units of words (Bloomfield 1926:
155). Morphemes are classified with regard to their distributional properties into (potentially) free morphemes and (obligatorily) bound morphemes, and with regard to their
function into grammatical morphemes and lexical morphemes. In English word-formation research (see article 15 on units of word-formation for traditions in other languages),
free lexical morphemes, also known as root morphemes, are considered to correspond
to simple lexemes (e.g., hand, great, eat); bound grammatical morphemes correspond to
inflectional endings, e.g., {genitive -s}, {past tense -ed}, marking word-forms for case,
number, tense, etc. Bound lexical morphemes are derivational affixes used for the purpose of word-formation (e.g., un-, -ment); free grammatical morphemes are function
words such as of, to or the, which are orthographically autonomous, serve grammatical
functions and are synsemantic rather than autosemantic. Morphemes are theoretical constructs abstracting commonalities over morphs, which are often regarded as perceptible
physical events realizing morphological building blocks in actual speech or writing (see
article 15 for a critique of this view).
While the notion of morpheme has proved useful for the description of basic constituents of complex lexemes, demarcation problems are rampant, with regard to both the
identification of morphemes as such and the classification of morphemes into lexical vs.
grammatical and free vs. bound morphemes. The general properties of morphemes are
reviewed by Mugdan (1986) and Luschtitzky (2000), see also article 15 on units of wordforrnation. Prototypical lexical morphemes, which lie at the heart of word-formation
research, have to meet the following morpheme- and lexeme-related criteria:
a) Identifiability: It must be possible to describe each of the potential morphemes precisely in terms of their extent and form and the corresponding meanings.
b) Exhaustive segmentability of complex lexemes into morphemes: It must be possible
to segment a given complex lexeme exhaustively into morphemes and other clearly
identifiable non-meaning-bearing elements such as linking elements (as in G. Notenstander 'music stand' +--- {note} + n + {stander}).
c) Autonomy of at least one constituent: Every complex lexeme must consist of at least
one free lexical morpheme.
d) Compositionality of complex lexemes: It must be possible to trace the meaning of
the complex lexeme back to the meanings of the morphemes.
In practical applications problems tend to arise with regard to all four criteria, resulting
in the postulation of different types of pseudo-morphemic, quasi-morphemic or submorphemic units (cf. Kubrjakova 2000 and article 15). One notorious challenge are
English verbs of Romance origin such as insist, p ersist and resist or ascribe, prescribe
and subscribe which suggest an analysis in terms of a prefix (a-, in-, p er-, pre-, re- and
sub-) and a stem (-sist or -scribe). Such an analysis runs into difficulties with criteria a),
c) and d), because, at least from a synchronic point of view, it is neither possible to
ascribe a meaning to the potential stems -sist and -scribe nor are these stems free morphemes; as a result, compositionality is also violated. The term bound root has been
introduced to capture such meaning-bearing units. Neoclassical compounds such as biology or bibliography, which are found in most European languages, cause partly similar
problems. While their parts - bio-, biblio-, -logy and -graphy - seem to carry identifiable
lexical meanings, none of them are free forms, thus violating criterion c ). In English and
PA5
5

Romance linguistics, the terms combining form and affixoid have become established to
describe this phenomenon; in German linguistics the term Konjix (Schmidt 1987: 50) is
commonly used. Typical lexical blends also defy descriptions in terms of morphemes,
as they cause problems for all four criteria. The constituent units of blends, for example
br- and -unch in the case of the classic brunch, are known as splinters (Lehrer 1996).
Furthermore, the particles of English phrasal verbs (give up, pass out), German prefix
and particle verbs (beraten, anbahnen) and similar multi -word lexical items also fall
within the scope of morpheme-like building blocks of complex lexemes. The semantic
contribution of these elements to the meaning of the multi-word unit is often opaque,
however, so that criteria a) and d) are not met. The term formative has been suggested
either to refer to such minimal units which lack an identifiable meaning, also including
linking elements (G. Fugenelemente) (Kastovsky 1982: 70; Lipka 2002: 87), or as a
superordinate term comprising both morphemes proper and semantically empty morphological building blocks (Bauer 1988b: 24; Bauer, Lieber and Plag 2013: 16). Finally,
phonaesthemes (e.g.,jl- injlicker,jlip,jlap,jlurry) are non-arbitrary pairings of phoneme
clusters and meanings (Firth 1964: 184- 185). Their semantic significance is usually
restricted to a limited range of words and their meanings do not go beyond soundsymbolic allusions and associations which are shared by a set of words but are quite
difficult to pinpoint. The portions of words that remain when the phonaestheme is segmented do not carry meanings; as a result, the words are non-compositional.
The internal structures of complex lexemes are usually described not in terms of
shallow, chain-like sequences of morphemes and morpheme-like elements but as hierarchical structures which follow the principle of binary branching of immediate constituents (cf., e.g., Booij 1977: 32; Lieber 1990: 80; Scalise 1984: 146-151) also prominent
in syntactic theorizing. In addition, the right-hand head rule (Williams 1981) states that
the two sister constituents on one level are not equipotent but related to each other in a
detenninans-determinatum (Marchand 1969) or modifier-head relation (cf. also Lieber
1992: 26- 76). The determinatum or head is the semantically and, more importantly,
grammatically dominant constituent, which is specified by the determinans or modifier.
To conclude this section, Figure 1.1 illustrates the most important concepts introduced
here.
nationalized
nation
al
free lexical
bound lexical
morpheme
morpheme
root
derivational suffix
base (word-formation stem)
(word) stem
modifier
modifier
head
ize
bound lexical
morpheme
derivational suffix
suffix
ed
bound grammatical
morpheme
inflectional suffix/
inflectional ending
head
Fig. 1.1 : Terms used for the description of the internal structures of complex lexemes
PA6
3. Word-formation patterns
Figure 1.2 represents one of the most common ways of classifying the major types of
word-formation patterns found in the languages of the world. While many other classifications are of course possible, depending, among other things, on the type of language
and the individual researcher's aims and convictions, the key categories found in Figure
1.2 have proved adequate for the descriptions of the wide range of different languages
provided in articles 134 to 207 of this handbook.
word-formation patterns
----
---~
morphemic patterns
non-morphemic patterns
~~~
composition
convernion
zero-
derivation
~---::::.------
reduplication blending clipping
auon:ym-
fonnariou
derivation
prefi..'\'.ation suffLxation
Fig. 1.2: Survey of word-formation patterns
As all these types of patterns are discussed in greater detail and with reference to a large
number of fanguages in parts II, III and XVI of this handbook, it will suffice here to
provide a general survey.
Composition, or compounding, is defined as the combination of (at least) two lexemes
or words (Bauer 1988b: 33; article 20 on composition), or stems (Hansen et al. 1990:
43; Fleischer 2000: 889), or bases (Lieber 2010: 43; see also article 135 on English), or
free lexical morphemes (Schmid 2011). Many authors explicitly include bound roots or
confixes - in addition to free lexical morphemes proper - as potential bases of compounds (Fleischer and Barz 2012: 84; Plag 2003: 135; article 135 on English). Compounds are sub-classified in terms of their semantic structure into endocentric and exocentric compounds, with the former being further divided into determinative and coordinative ones. As regards their forms, root compounds, consisting of only two free lexical
morphemes, are distinguished from synthetic (verbal, verbal nexus) compounds, which
include bound lexical morphemes (see articles 20 on composition and 33 on synthetic
compounds in German). Rules detailing the formation of compounds have to take the
word-classes of their constituents into account, granting a prominent role to the head
constituent, so that blackbird would be referred to as a nominal compound of the form
Adj+ N.
Derivation (see article 16 on derivation) is the process of adding an affix to a stem,
or a bound 1exical morpheme to a free one, in order to create a new lexeme. The main
forms of derivation are prefixation or prefix-derivation, where the bound morpheme
precedes the free one, and suffixation or suffix-derivation where the order is reversed.
Rarer types not rendered in Figure 1.2 are infixation and circumfixation. In certain
schools of word-formation research mainly in English studies (Marchand 1969; Kastovsky 1982; Hansen et al. 1990), zero-derivation or derivation by zero-morpheme is postu-
PA7

lated as a further type of derivation producing new lexemes by the addition of a formally
empty zero-morpheme. As indicated by its place in Figure 1.2, the process of backfonnation (see article 18) straddles the boundary between morphemic and non -morphemic
word-formation patterns. In contrast to derivation by means of the addition of an affix,
lexemes formed by backformation are the product of the deletion of a bound morpheme
or morpheme-like element.
Roughly speaking, conversion (see article 17) is a word-formation process which
transposes a lexeme to a new word class without the addition of an overtly marked
suffix. As Valera (see article 17) puts it, "the form of the converted item does not change,
while its inflectional potential, its syntactic function and its meaning do, such that the
item displays inflectional, syntactic and semantic properties of a new word-class". Conversion and zero-derivation - as well as the concept of paradigmatic derivation used in
work on word-formation in Polish; see article 156 on Polish - are essentially competing
ways of making sense of the same phenomena. Where the converted item does not
acquire the full range of inflectional possibilities (e.g., the rich), this is termed partial
conversion by proponents of the conversion approach and conversion by those of the
zero-derivation approach.
Morphemic word-formation processes make up the core of word-formation in the
sense that they are to a large extent regular and predictable (in hindsight) and therefore
amenable to generalizations couched in the format of rules or schemas (see section 7).
Equipped with knowledge of such rules, no competent speaker of English who knows
the verb to tweet will be surprised when confronted with a derived verb such as to detweet, as this is clearly a potential or possible word (Aronoff 1976: 17- 19). The same
kind of hindsight predictability does not apply to the group of non-morphemic processes
listed in Figure 1.2, which are also known as minor word-formation types. In addition,
these processes are more flexible and more creative, which causes difficulties when it
comes to attributing a shared meaning to formations of a similar type.
Reduplication (see article 25) is a word-formation process which involves the repetition of a word, word-like element or part of a word either in unchanged form (e.g., hushhush), with a different vowel (e.g., hip-hop) or a different consonant (e.g., boogie-woogie). Blending (see a1iicle 21) is a cover term for a range of processes which, like composition, combine (at least) two lexemes, but, unlike composition, also fuse their forms by
either shortening one or both of the input lexemes or by telescoping them into each other
at portions where their forms overlap. Unlike all other word-formation processes, clipping and acronym-formation (see article 19) both preserve the denotative meanings of
the source lexemes. Both are form-shortening processes only. Clipping, the deletion of
initial and/or final portions of words, can be applied to single words, while acronyms,
with some variation, are created by deleting everything but initial letters of two-word or
longer expressions. Some researchers distinguish between acronyms in a narrow sense,
which can be pronounced like normal words, and initialisms, pronounced as series of
individual letters.
Needless to say, the description of different word-formation models and the analysis
of individual complex lexemes do not stop at this level of granularity. More specific
descriptions from a structural perspective tend to consist of at least six types of information. These are illustrated by English deverbal nominalizations in Table 1.1.
Textual, pragmatic and sociolinguistic aspects related to characteristics of the use and
frequency of word-formation patterns in different text types and with different functions
can complement this infonnation.
PA8
Tab. J. l: Parameters for the description of complex lexemes (inspired by Fleischer and Barz 2012:
73-74)
parameter
exemp lified for English deverbal nominalization in

-(a)tion
a) general morphological properties of

the constituents
base:
- word-class: transitive verbs, e.g., explain, combine, describe
- morpho logical status: free lexical morpheme
affix:
- morphological status: bound lexical morpheme,
suffix
- form: -(a)tion
b) order of constituents
- base - suffix
c) formal and semantic characteristics

of the process
- formal description: [V + -(a)tion ]N

- semantic description: 'process or product of
V-ing'
d) various other characteristics and effects of the complex lexemes generated by the process, especially phonological and graphological ones, sometimes depending on the specific
characteristics of the input
- stress movement and vowel change, e.g.,

hks'plem/ > /ekspl1:i' ne1Jn/
- possible insertion of <a> /e1/ before <tion> !Jn!
e) restrictions on the nature of the input

and on the applicability of the process (see articles 48 to 52)
- generally not applicable to English words of

Germanic origin, cf. hear> *hearation, eat>
*eatation
f) degree of productivity (see article 47)
- fu lly productive, especially in formal contexts
4. Approaches to word -formation research

The study of word-formation has been approached from a variety of angles. Some of
these are shared by other linguistic disciplines, while others are specific to word-formation research or manifested in specific forms.
As in other fields of linguistics, we can distinguish between historical approaches (see
parts X, XI, X II), and those which investigate the present-day language in its current state.
Historical investigations can be carried out in a diachronic manner (cf. Kastovsky 2009) or
in a synchronic one. What is specific to word-formation research is that there is always a
latent diachronic element in synchronic descriptions, because the very idea that complex
lexemes are the products of a formation process actually entails a dynamic perspective (see
article 16 0111 derivation). While a synchronic description can in principle rely solely on observing paradigmatic relations between words sharing the same apparently productive elements, complex lexemes, much more obtrusively than unmotivated simple lexemes, evoke
the impression that they are the effects of a process (cf. Hansen et al. 1990: 31- 32).
As in all fields of linguistic inquiry, typological word-formation research compares
different languages with the aim of identifying similarities and differences or producing
a universally valid description of word-formation in language as such.
PA9
Word-formation research has been carried out from a semasiological perspective,
which has been the dominant one, but also from an onomasiological one (e.g., Stekauer
1998, 2005; see article 6 on word-formation in onomasiology). Semasiological investigations aim to describe the structures and meanings of complex lexemes and wordformation patterns. In doing so, they start out from an examination of existing complex
lexemes - e.g., explanation, combination, description - which share the same morphological structure and are therefore hypothesized, at least metaphorically speaking, to
share a common formation history (roughly [V + (a)tion]N) Therefore, it is the analytical
aspect of a hearer or reader confronted with a complex lexeme - or indeed the linguist
trying to discover structural generalizations - that comes to the fore in semasiological
approaches. The complementary onomasiological perspective takes into consideration
the fact that competent speakers are not only able to segment complex words into their
constituents, but also use the results of such analyses for their own generative potential
to create new words. The onomasiological perspective reflects speakers' states of mind
while trying to encode a given conceptualization by means of applying a word -formation
model. What, for instance, are the speaker's choices when aiming to encode the result
of an action by deriving a noun from a verb: -ation as in transformation +----- transform;
-ment as in achievement +----- achieve; -ence as in existence +----- exist; -ing as. in painting
+----- paint; -al as in denial +----- deny; -age as in blockage +----- block? The onornasiological
approach thus corresponds to the synthetic aspect of word-formation. The sernasiological
and the onomasiological approaches diffe r in terms of their potential to be applied for
descriptive purposes. Due to the virtually unlimi ted range of meanings that can be expressed by means of compounds, especially root compounds, the semasiological perspective has been dominant in this field. Systematic descriptions of prefixation and suffixation can, in principle, be arranged from a semasiological angle, by providing the meanings associated with the different prefixes and suffixes depending on the types of bases,
or from an onomasiological one, by listing, for example, the prefixes that encode negation, location or time, or the suffixes that turn verbs into person-denoting nouns.
5. Levels of analysis and description in word-formation resea rch

Due to its position at the interface or crossroads of many different aspects of language,
word-formation research potentially includes the full range of linguistic levels, from
phonetics and phonology to syntax, semantics, pragmatics and sociolinguistics.
Morphological considerations related to the constituents of complex lexemes and the
formation rules and schemas naturally form the core of word-formation research, since
a solid understanding of the composition of complex lexemes and their internal structures
is required for all other levels of investigation (see article 15 on units of word -formation).
Semantic aspects are of key importance not on ly for the internal segmentation and structural analysis of complex lexemes, but also for describing the semantic links between
constituents and the semantic characteristics of word-formation processes and patterns
such as compounding or derivation (see articles 5 on word-formation in inhaltbezogene
Grammatik, 49 on argument-structural restrictions on word-formation patterns and 58
on categories of word-formation). Phonetic, phonological and morphonological aspects
are equally important from a descriptive, an analytical and a heuristic perspective. In the
PA10
10
field of derivation, for example, it is crucial to understand and model regularities regarding stem allomorphy, e.g., systematic changes in the stress patterns and vowel qualities
of bases and derivatives, as, for example, in English explain ---+ explanation or sane ---+
sanity (see articles 16 on derivation and 50 on phonological restrictions on English wordformation). Stress is also considered a diagnostic for compound status in English and
other languages (cf. Bauer l 988a; Giegerich 2009). Syntactic aspects of word-formation
came into focus with attempts to transfer insights from syntactic structures and rules to
the internal grammar of words in early generative grammar (see article 7 on wordformation in generative grammar). They also provide important insights into word-classspecific restrictions on word-formation rules and their productivity (cf. Plag 1999).
The remaining levels to be mentioned here hold a less trad itional position in wordformation research. This does not mean that they are less important, however. Over the
past two or three decades, psycholinguistic and neurolinguistic research has greatly enriched the field of word-formation research by probing the extent to which models of
word-formation are realistic and plausible from a psychological and neuronal perspective
(see article 13). Questions that have been addressed from this perspective include the
way in which novel and lexicalized complex lexemes, especially compounds, derivations
and blends, are represented in the mind and the brain, and how they are processed in
actual usage (e.g., Libben and Jarema 2006; Schmid 2008). More recently, the cognitivelinguistic perspective focusing on the way in which knowledge of word-formation
models and schemas as well as of individual complex lexemes becomes entrenched and
is influenced by general cognitive abilities such as categorization and figure-ground
segregation has gained in importance (cf., e.g., Ungerer 2002 and 2007; Heyvaerts 2003
and 2009; Onysko and Michel 2010; Schmid 201 1). The sociolinguistic perspective looking at word-formation in diverse regional and social varieties (e.g., Biermeier 2008;
Braun 2009) and the pragmatic perspective (Downing 1977; Bauer 1979; Clark and Clark
1979; Schmid 2011) highlighting interactional contexts and communicative functions in
actual usage-events have also been gaining momentum in the field of word-formation
research. This includes the study of lexical creativity in various registers and text-types
(Munath 2007) involving a text-related and discourse-related perspective (see chapter
XIV).
6. Theoretical models of wo rd-formation

The first chapter of this handbook (see articles 3 to 12) presents models of word-formation proposed in different theoretical frameworks, ranging from the historical-comparative tradition and structuralism to categorial grammar, generative grammar, natural morphology, opt imality theory, cognitive grammar and construction grammar. As one would
expect, in each case word-formation is modelled in line with the aims and assumptions
typical of the corresponding approaches to linguistics in general. While categorial grammar and generative grammar focus on formal aspects of word-formation, natural morphology, optimality theory, cognitive grammar and construction grammar take a functional stance. As generative grammar favours a modular architecture of language (see
article 7), a key issue has been to clarify whether word-formation belongs to syntax or
to the lexicon or is an interface or even a module in its own right. Cognitive grammar
PA11
(see article 10) and construction grammar (see article 12), on the other hand, favour a
holistic conception of word-formation processes which unites morphological, syntactic,
semantic, cognitive and even pragmatic aspects. Natural morphology (see article 9) and
optimality theory (see article 11) attempt to develop explanatory frameworks which are
of cross-linguistic or even universal validity and relevance rather than describing individual processes.
Supporters of the various theoretical models advocate very different conceptions of
how word-formation processes are to be modelled. A frequently quoted early classification of these conceptions was proposed by Hockett (1954), who distinguished between
three types of models: item and arrangement, item and process, and word and paradigm.
The first approach aims to describe the patterns of word-formation starting out from
listing morphemes and describing "the arrangements in which they occur relative to each
other in utterances" (Hockett 1954: 212). Secondly, item and process models, as suggested by the term, highlight the procedural aspects of word-formation; they regard the root,
rather than the morpheme, as the basic input to morphological processes. Thirdly, word
and paradigm models, which tend to focus on inflectional rather than derivational morphology, consider unsegmented words as the basic unit of word-formation and try to
disclose paradigmatic similarities by comparing them.
Moving to the present state of the art, the range of current theories which aim to
model word-formation patterns and processes can be broadly divided into four types of
approaches:
-
rule-based models
schema-based models
exemplar-based models
exemplar-cum-schema-based models
The most prominent representatives of rule-based models are the different variants of
generative approaches. Much of the research carried out in this framework is concerned
with devising general rules applying across different word-formation types as well as
type-specific rules or rule schemas, as they are called, in such a way that they can
account for the empirical fac ts largely gleaned from introspection. The precise way in
which these rules have been formulated has changed in line with the different stages of
generative grammar: transformations and rewriting rules expressed in analogy to the
phrase-structure rules of early generative grammar (Chomsky and Halle 1968; Selkirk
1982) were followed by applications of X-bar-theoretical principles (e.g., Scalise 1986)
and sets of projection rules compatible with the government-and-binding and the principles-and-parameters approaches (see article 7 on word-formation in generative grammar
and Lieber and Mugdan 2000 for more details). What rule-based approaches share is
their focus on structural and formal rather than semantic or functional aspects, their
commitment to the modularity assumption, giving rise to the need to identify the place
where word-formation is situated in the architecture of language and linguistic knowledge, and their goal to formulate maximally generalizable insights and predictions.
Well-known general hypotheses that have been postulated for the fie ld of word-formation include, next to the binary-branching hypothesis and the right-hand head rule already
mentioned, the unitary base hypothesis (Aronoff 1976: 47- 48; Scalise 1984: 137-146)
and the uni tary output hypothesis (Scalise 1984: 137), both of which have turned out to
be problematic, however. The former states that the syntactic and semantic properties of
11
PA12
12
the bases of derivational rules are clearly specified and unique, which implies that affixes
cannot be attached to words of different word-classes. The latter states that affixes must
be functionally and semantically unitary in the sense that they cannot be attached to
different word-classes and bring about different changes in meaning. In both cases, potential empirical counter-evidence has been dealt with by maximizing the homonymy of
affixes. This means, for instance, that the form -able in the deverbal adjective manageable and the denominal adjective marriageable (Bauer, Lieber and Pla,g 2013: 635- 636)
and the form -ese in the noun Japanese and the adjective Japanese would each h ave to
be considered as two different but homonymic affixes.
As a concrete example of the rule-based approach to describing individual wordformation models, the rule schema proposed by Aronoff (1976: 63) for English adjectival
prefixation with unis provided in (1).
(1)
"Rule of negative un#

a.
[X]Adj ~ [un#[X]Adj]Adj
semantics (roughly) un#X = not X

b.
Forms of the base

1. Xven (where en is the marker for the past participle)
2. Xv#ing
3. Xvable
4. X+y (worthy)
5. X+ly (seemly)
6. X#ful (mindful)
7. X#like (warlike)"
(Aronoff 1976: 63; original italics)
The endeavours of schema-based models are also directed towards reaching generalizations, but th ese are not couched in the form of rules, but are instead formulated in terms
of (constructional) schemas (see article 45 on rules, patterns and schemata in wordfonnation). These are defined as schematic form-meaning pairings representing lexical
and phrasal knowledge and sanctioning concrete uses of complex lexemes. While rules
are essentially variable procedural instructions, schemas are unit-like elements containing
variable slots (Booij 2010: 41-43). In this way, schemas account for productivity and
analyzability as well as creativity in word-formation. Schemas are considered to be connected by formally and semantically motivated hierarchical relations, yielding multidimensional networks of schemas and sub-schemas arranged on several levels of specificity (Ryder 1994; Tuggy 2005: 248- 264). Schemas on lower levels inherit information
from superordinate schemas. In contrast to rule-based approaches, schema-based models
subscribe to a holistic, non-modular conception of linguistic knowledge and therefore
unite and integrate formal structural, semantic and functional aspects in their accounts
of schemas. While rule-based models tend to keep up the strict separation of grammar
and usage, schema-based models are compatible with usage-based approaches, which
assume that linguistic knowledge emerges from the experience of concrete usage-events
in social situations and is subject to the frequencies of occurrence of certain elements
(cf., e.g., Kemmer 2003; Bybee 2010). Figure 1.3 provides an idealized representation
of a schema-based network for un-adjective formation inspired by Tuggy (2005) and
Booij (2010):
PA13
prefix
'con1ra.st'
...
un
' neg'
...
l/11
x .'i.dJ
' neg'
QUALITY
-~
13
un
V-able."-'IJ
UJI
V-edAdJ
1111
N-fulAi;
rm
'neg'
QUALITY
'neg'
QUALITY
'neg'
QUALITY
' neg
r;;-i
~
beliemble
believable'
~
~
I
predictable' I
predictable
l.
~lemploJed
~!employed'
~I expected'
QUALITY
I
~ Isuccessjitl
r,;;;-i Ihappy I
~ lsltccessful'
~ ! happy'
~!expected
X-yAdJ
I
I
~ l lm9ful
~l lawfur
I
I
~ l likeli1
~ l tikely'
I
I
Fig. 1.3: Idealized illustration of a schema-based network for adjectival un-prefixation in English
In Figure 1.3, schemas are represented by boxes detailing formal properties above the
dividing horizontal lines and meanings below them. Boxes indicate the status of symbolic
units, with bolder boxes marking hypothetically more entrenched schemas. The top-level
schema in Figure 1.3 represents the most general prefixation schema which is not specified with regard to the form and the meaning of the base. In line with Schmid (2011:
160- 162), the meaning of the pattern of prefixation as such, i.e. across different prefixes
and different word-classes of the base, is glossed as expressing a contrast to the base.
The sub-schemas represent increasingly specific information, from un-prefixation applicable not only to adjectives but also verbs and nouns to adjectival un-prefixation, different
types of un-adjectives and, finally, individual words sanctioned by these schemas.
Schema-based and exemplar-based models have more in common than rule-based
and schema-based models. Like schema-based models, exemplar-based models (Krott,
Schreuder and Baayen 2002; Eddington 2004: 71-98; Bybee 2010: 14-33; 165-193;
Bybee and Beckner 2010; Arndt-Lappe 201 1) conceive of lexical knowledge and knowledge about word-formation processes in terms of associative networks that are continuously reorganized under the influence of usage and exposure. The main difference between the two approaches is that strong versions of exemplar-based approaches, especially connectionist ones (e.g., Skousen 1992; Skousen, Lonsdale and Parkinson 2002), deny
the existence of symbolic representations such as schemas and assume that linguistic
knowledge is only avai lable in the form of representations of individual exemplars stored
under the impression of specific usage events. In theory at least, the structures of exem-
PA14
14
plar-based networks are much more complex than those of schema-based ones, since the
nodes are individual exemplars related by means of similarity relations, ideally on all
possible levels of description, i.e. meaning, morphological form and phonological form.
The structure of networks - in terms of the strengths of nodes and the links between
them - is considered to be influenced by probabilistic information extracted from the
frequencies of exemplars. Instead of rules or schemas, analogies based on similarities
are considered to play the key role of motor and motivation for productive and creative
processes.
Some network models (e.g., Bybee 2010) try to reconcile characteristics of schemabased and exemplar-based models. They postulate the existence of schemas representing
experience that has been abstracted from usage-events without denying the importance
of exemplar-based knowledge and thus argue that multiple and redundant representations
on different levels of schematicity co-exist, from general schemas all the way down to
individual lexemes. This way, they accommodate the potential to form new complex
lexemes both on the basis of productive schemas and also by means of analogical formations based on similarities to stored exemplars (Eddington 2006). Like pure exemplarbased models, such exemplar-cum-schema-based models keep track of usage frequencies
of exemplars but have a more differentiated view of the effects of frequency of exposure.
While token repetition is considered to reinforce the representation of exemplars, type
repetition strengthens schemas sanctioning novel uses (Bybee 2010). Figure 1.4 takes up
English adjectival un-prefixation once more and provides an exemplary and idealized
fragment of an exemplar-cum-schema network indicating form-based clusters of exemplars which are candidates for schemas.
-----
u~~<1)
6134
unpopmar
8
111weccessmy
1814
11nforu111are
JSSL
unaware
1133
vnsatisfoctory
760
Fig. 1.4: Idealized schema-cum-exemplar-based network representation of English adjectival unprefixation

(Clusters: (1) un-Adj; (2) un-X -able (various morphological structures); (3) un-V -able;
( 4) un-Adj(-y) (various morphological structures); (5) un-V-D 2 (various allomorphs of
D 2); un-V-D 2 , only /rd/ allomorph; (7) un-N(-al); (8) un-X-al (various morphological
structures); (9) un-N-ful; (10) un-Adj(monosyllabic); (l l) un-Adj(bisyllabic, '--)
PA16
16
2005; Trips 2009; Schroder 2008, 201 1) or of several patterns wh ich are in onomasiological competition, e.g., -ity and -ness (cf. Gaeta and Ricca 2006; Baeskow 2012). The
second major branch of research into productivity tries to provide maximally detailed
descriptions of the general or pattern-specific limits or restrictions on productivity (see
article 48). For example, a wide-ranging and in fact somewhat trivial restriction on
productivity is that there must be a communicative need for a potential formation. Pattern-specific restrictions can be morphological, semantic, phonological, syntactic and
etymological in nature or derive from semasiological or onomasiological competition
(blocking) of existing lexemes (Kastovsky 1982: 159-164; Rainer 2005). Recent webbased research (e.g., Schroder and Mi.ihleisen 2010) has indicated that many productivity
restrictions are tendencies rather than hard and fast rules.
Research in the field of the lexicalization of individual complex lexemes has been
characterized by considerable variation in the choice of terms. Among the terms causing
confusion are lexicalization, institutionalization, idiomatization, conventionalization, diffusion, spread, propagation, listing and establishment (cf. Brinton and Traugott 2005:
32- 61 ). The term lexicalization itself, for example, has been conceptualized as "the
process by which complex words come to have meanings that are not compositional"
(Lieber 2010: 201 ), i.e. more or less synonymous with idiomatization; as "the strength
of the lexical representation of a particular lexeme and its forms" (Bauer, Lieber and
Plag 2013: 117), i.e. more or less synonymous w ith entrenchment; and by stating that
"[w]hen a possible word has become an established word, we say that it has lexicalized"
(Booij 2005: 17), i.e. in sociopragmatic tenns. As this terminological diversity is partly
caused by lumping together different levels of description, it seems desirable to separate
these levels both conceptually and terminologically. Schmid (2011: 71-83) therefore
proposes that the term lexicalization be reserved for the structure-oriented description of
formal processes (fusion, reduction, erosion) and semantic processes (idiomatization)
noticeable in the word itself. The sociopragmatic perspective focusing on how a given
word spreads in the speech community is captured by the terms institutionalization and,
more generally, conventionalization, while the cognitive perspective describing changes
taking place in the mind, such as the strength of a unitary holistic representation or
the density of the link to "neighbouring" words, is described in terms of degrees of
entrenchment.
8. Conclusion
In keeping with the nature of the opening contribution to any handbook, this article can
supply no more than a broad-brush survey of the main issues in word-formation research.
In doing so, it aims to chart the terrain for the much more fine-grained explorations
offered in the subsequent articles. As has been shown, one of the beauties and challenges
of word-formation research is that its ramifications branch out onto all levels of linguistic
analysis, description and theorizing. Recently, the field has been marked by an increasing
awareness that speakers and writers use the productive and creative resources of their
language in a much more flexible and variable manner than had been predicted by rigid
and overly reductionist models. Doing justice to this flexibility while upholding the aim
to produce valid generalizations could be one of the major challenges to be faced in
future research into word-formation.
PA17
9. References
Aronoff, Mark
1976
Word Formation in Generative Grammar. Cambridge, MA: MIT Press.
Arndt-Lappe, Sabine
2011
Towards an exemplar-based model of stress in English noun-noun compounds. Journal
of Linguistics 11: 549- 585.
Baayen, R. Harald
1989
A corpus-based approach to morphological productivity: Statistical a nalysis and psycholinguistic interpretation. Ph.D. dissertation, Vrije Universiteit, Amsterdam.
Baayen, R. Harald and Rochelle Lieber
199 1 Productiv ity a nd English word-forma tion. A corpus-based study. Linguistics 29: 80 1843.
Baeskow, He ike
20 12
-Ness and -ity: Phono logical exponents of n or meaningfu l nomina lizers of diffe re nt
adjectiva l domains? Journal of English Linguistics 40 : 6-40.
Ba udouin de Courte nay, Jan N .
Versuch einer Theorie phonetischer Alternationen. Straf3bu rg: Triibner.
1895
Bauer, Laurie
On the need for pragmatics in the study of nominal compounding. Journal ofPragmatics
1979
3: 45- 50.
Bauer, Laurie
English Word-Formation. Cambridge: Cambridge University Press.
1983
Bauer, Laurie
1988a When is a sequence of two nouns a compound in English? English L anguage and
Linguistics 2: 65- 86.
Bauer, Laurie
1988b Introducing Linguistic Morphology. Edinburgh: Edinburgh University Press.
Ba uer, Laurie, Rochelle L ieber and Ingo Plag
20 13
The Oxford Ref erence Guide to English Morphology. Oxford: Oxford University Press.
Benveniste, Emile
1967
Fondeme nts syntax iques de la composition nominale. Bulletin de la Societe de Linguistique de P aris 62: 15-3 1.
Biermeier, Thomas
2008
Wordjormation in New Englishes. A corpus-based analysis. Be rlin: LIT.
Bloomfield, Leonard
1926
A set of postulates for the science of language. Language 2 : 153- 164.
Bloomfield, Leonard
1933
Language . London: A llen & Unw in.
Booij, Geert
Dutch Morphology. A Study of Word Formation in Generative Grammar. Dordrecht:
1977
For is.
Booij, Geert
2000
Inflection and derivation. Jn: Geert Booij , Christian Lehmann and Joachim Mugdan
(eds), Morphology. An International Handbook on Inflection and Word-Formation.
Vol. I, 360- 369. Berli n/New York: de Gruyter.
Booij, Geert
The Grammar of Words. An Introduction to Morphology. Oxford: Oxford University
2005
Press.
Booij, Geert
20 10
Construction Morphology. Oxford: Oxford University Press .
17
PA18
18

Braun, Maria
2009
Word-Formation and Creolisation. Th e Case of Early Sranan. TU bingen: Niemeyer.
Brinton, Laurel J. and Elizabeth Closs Traugott
2005
Lexicalization and Language Change. Cambridge/New York: Cambridge University
Press.
Bybee, Joan L .
1985
Morphology. A Study of the Relation between Meaning and Form. Amsterdam/Philadelphia: Benjamins.
Bybee, Joan
2010
Language, Usage and Cognition. Cambridge/New York: Cambridge University Press.
Bybee, Joan L. and C lay Beckner
2010
Usage-based theory. In: Bernd Heine and Heiko Narrog (eds.), The Oxford Handbook
of Linguistic Analysis, 826- 855. Oxford: Oxford University Press.
Chelliah, Shobhana L. and Willem J. de Reuse
2011
Handbook of Descriptive Linguistic Fieldwork. Dordrecht: Springer.
Chomsky, Noam and Morris Halle
1968
Th e Sound Patterns of English. New York: Harper and Row.
Clark, Eve and Herbert H. C lark
1979
When nouns surface as verbs. Language 55: 7 67-8 11.
Downing, Pamela
1977
On the creation and use of English compound nouns. Language 53: 810- 842.
Eddington, David
2004
Spanish Phonology and Morphology. Experimental and Quantitative P erspectives. Amsterdam/Philadelphia: Benjamins.
Eddington, David
2006
Spanish diminutive formation without rules or constraints. Linguistics 40: 395-419.
Firth, John R.
1964
The Tongues of Men & Speech. Oxford: Oxford U nivers ity Press.
F leischer, Wolfgang
2000
Die K lassifikation von Wortb ildungsprozessen. In: Geert Booij , Christian Lehmann and
Joachim Mugdan (eds.), Morphology. An International Handbook on Inflection and
Word-Formation. Vol. 1, 886-897. Berlin/New York: de Gruyter.
Fleischer, Wolfgang and Irmhild Barz
20 12
Wortbildung der deutschen Gegenwartssprache. 4 111 ed. Berlin/Boston: de Gruyter.
Gaeta, Livio and Davide Ricca
2006
Productivity in Italian word formation. A variable-corpus approach. Linguistics 44: 5789.
Giegerich, Heinz J.
2009
Compounding and lexicalism. In: Rochelle L ieber and Pavo! Stekauer (eds.), The Oxford
Handbook of Compounding, 178- 200. Oxford: Oxford University Press.
Giegerich, Heinz J.
2012
The morphology of -ly and the categorial status of 'adverbs' in English. English Language and Linguistics 16: 341 -359.
Hansen, Barbara, Klaus Hansen, Albrecht Neubert and Manfred Schentke
1990
Englische Lexikologie. 2"d ed. Leipzig: Enzyklopadie.
Heyvaert, Liesbeth
2003
A Cognitive-Functional Approach to Nominalization in English. Bedin/New York: Mouton de Gruyter.
Heyvaert, Liesbeth
2009
Compounding in cognitive linguistics. In: Rochelle Lieber and Pavo! Stekauer (eds.),
The Oxford Handbook of Compounding, 233- 254. Oxford: Oxford Univerity Press.
PA19
H ockett, Charles F.
1954
Two models of grammatical description. Word 10: 210-231.
Kastovsky, Dieter
1982
Wortbildung und Semantik. Dusseldo rf: Bagel & Francke.
Kastovsky, Dieter
1999
English and German morphology. A typological co mparison. In: Wolfgang Falkner and
Hans-forg Schmid (eds.), Words, Lexemes, Concepts - Approaches to the Lexicon. Studies in Honour of Leonhard Lipka, 39-52. Tubingen: Narr.
Kastovsky, Dieter
2009
Diachronic perspectives . In: Rochelle Lieber and Pavel Stekauer (eds .), The Oxford
Handbook of Compounding, 323-340. Oxford: Oxford Uni verity Press.
Kemmer, Suzanne
2003
Schemas and lexical blends. In: Hubert Cuyckens, Thomas Berg, Rene Dirven and
K laus-Uwe Panther (eds.), Motivation in Language. Studies in honor of Gunter Radden,
69-97. Amsterdam/Philadelphia: Benjamins.
Krott, Andrea, Robert Schreuder and R. Harald Baayen
2002
Analogical hierarchy : Exemplar-based modeling of linkers in Dutch noun-noun compounds. fo: Royal Skousen, Deryle Lonsdale and Dilworth B. Parkinson (eds.), Analogical Modeling, 181 - 206. Amsterdam/Philadelphia: Benjamins .
Kubrjakova, Elena S.
2000
Submorphemische Einheiten. In: Geert Boo ij, Christian Lehmann and Joachim Mugdan
(eds .), Morphology. An International Handbook on Inflection and Word-Formation. Vol.
I, 4 17- 426. Berlin/New York: de Gruyter.
Lehrer, Adrienne
1996
Identifying and interpreting blends: A n experimenta l approach. Cognitive Linguistics 7:
359- 390.
Libben, Gary and Gonia Jarema (eds.)
2006
The Representation and Processing of Compound Words. Oxford: Oxford University
Press.
Lieber, Roche lle
1990
On the Organization of the Lexicon. New York/London: Garland.
L ieber, Rochelle
1992
Deconstructing Morphology. Word Formation in Syntactic Theo1y. C hicago: Chicago
University Press.
Lieber, Rochelle
20 IO
Introducing Morphology. Cambridge: Cambridge University Press.
Lieber, Roc he lle and Joachim Mugdan
2000
Internal structure of words. In: Geert Booij, Christian Lehmann and Joachim Mugdan
(eds.), Morphology. An International Handbook on Inflection and Word-Formation.
Vol. I, 404-4 16. Berlin/New York: de Gruy ter.
Lipka, Leonhard
1983
A multi-level approach to word-formation. Complex lexemes and word semantics. In:
Shiro Hattori and Kazuko Inoue (eds.), Proceedings of the XJJI11' International Congress
of Linguists, Tokyo 1982, 926- 928. Tokyo: The Committee.
Lipka, Leonhard
2002
English L exico/ogy. Lexical Structure, Word Semantics and Word Formation. Tubingen:
Narr.
Luschutz ky, Hans C hristian
2000
Morphem, Morph und Allo morph. In: Geert Booij, Christian Lehmann a nd Joachim
Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-Formation. Vol. 1, 451-462. Berlin/New York: de Gruyter.
19
PA20
20

Marchand, Hans
1969
Th e Categories and Types of Present-Day English Word-Formation.. A Synchronic-Diachronic Approach. 2nd ed. Miinchen: Beck.
Mugdan, Joach im
1986
Was ist eigentlich ein Morphem? Zeitschriftfar Phonetik, Sprachwissenschaft und Kommunikationsforschung 39: 29-43.
Munat, Judith (ed.)
2007
Lexical Creativity, Texts and Contexts. Amsterdam/Philadelphia: Benjamins.
0 )sen, Susan
2000
Compounding and stress in English: A closer look at the boundary between morphology
and syntax. Linguistische Berichte 181: 55-69.
Onysko, Alexander and Sascha Michel (eds.)
2010
Cognitive Perspectives on Word-Formation. Berlin/New York: Mouton de Gruyter.
Plag, Ingo
1999
Morphological Productivity. Structural Constraints in English Derivation. Berlin/New
York: Mouton de Gruyter.
Plag, Ingo
2003
Word-Formation in English. Cambridge: Cambridge University Press.
Plag, Ingo
2006
Productivity. In: Bas Aarts and April McMahon (eds.), Th e Handbook of English Linguistics, 537- 556. Oxford: Blackwell.
Plag, Ingo, Christiane Dalton-Puffer and Harald Baayen
1999
Morphological productivity across speech and w riting. English Language and Linguistics 3: 209- 228.
Plank, Frans
1994
Inflection and derivation. In: Ronald E. Asher (ed.), Encyclopaedia of Language and
Linguistics. Vol. 3, 1671- 1678. Oxford: Pergamon.
Rainer, Franz
2005
Constraints on productivity. In: Pavo! Stekauer and Rochelle Lieber (eds.), Handbook
of Word-Formation, 335- 352. Dordrecht: Springer.
Ryder, Mary Ellen
1994
Ordered Chaos. The Interpretation of English Noun-Noun Compounds. Berkeley: University of California Press.
Scalise, Sergio
1984
Generative Morphology. Dordrecht: Foris.
Scalise, Sergio
1986
Generative Morphology. Berlin/New York: de Gruyter.
Scalise, Sergio
1988
Inflection and derivation. Linguistics 26: 56 1- 581.
Scherer, Carmen
2005
Wortbildungswandel und Produktivitat. Eine empirische Studie zur nominalen -er-Derivation im Deutschen . Tubingen: Niemeyer.
Schmid, Hans-JOrg
2008
New words in the mind: Concept-formation and entrenchment of neologisms. Anglia
126: 1-36.
Schmid, Hans-JOrg
2011
English Morphology and Word-formation. An Introduction. Berlin: Schmidt.
Schmidt, Gunter Dietrich
1987
Das Affixoid. Zur Notwendigkeit und Brauchbarkeit eines beliebten Zwischenbegriffes
der Wortbildung. In: Gabriele Hoppe, Alan Kirkness, Elisabeth Link, Isolde Nortmeyer,
Wolfgang Rettig and Gunter Schmidt (eds.), Deutsche Lehnwortbildung. Beitrage zur
PA21
21
Erforschung der Wortbildung mil entlehnten WB-Einheiten im Deutschen, 53- 10 l. Tiibingen: Niemeyer.
Schroder, Anne
2008
Investigating the morphological productivity of verbal prefixation in the history of English. Arbeiten aus Anglistik und Amerikanistik 33: 47-69.
Schroder, Anne
2011
On the Productivity of Verbal Prefixation in English. Synchronic and Diachronic Perspectives . Tu bingen: Narr.
Schroder, Anne and Susanne Miihleisen
2010
New ways of investigating morphological productivity. Arbeiten aus Anglistik und Amerikanistik 3 5: 43-58.
Selkirk, Elisabeth
The Syntax of Words. Cambridge, MA: MIT Press.
1982
Skousen, Royal
1992 Analogy and Structure. Dordrecht: Kluwer.
Skousen, Royal, Deryle Lonsdale and Dilworth B. Parkinson (eds.)
2002
Analogical Modeling. Amsterdam/Philadelphia: Benjamins.
Stekauer, Pavol
An Onomasiological Theoty of English Word Formation. Amsterdam/Philadelphia: Ben1998
Jamms.
Stekauer, Pavol
2005
Onomasiological approach to word-formation. In: Pavol Stekauer and Rochelle Lieber
(eds.), Handbook of Word-Formation , 207- 232. Dordrecht: Springer.
Stump, Gregory T.
2005
Word-formation and inflectional morphology. In: Pavo! Stekauer and Rochelle Lieber
(eds.), Handbook of Word-Formation , 49- 7 1. Dordrecht: Springer.
Trips, Carola
2009
Lexical Semantics and Diachronic Morphology. The development of -hood, -dom and
-ship in the history of English. Tiibingen: Niemeyer.
Tuggy, David
2005
Cognitive approach to word-formation. In: Pavo! Stekauer and Rochelle Lieber (eds.),
Handbook of Word-Formation , 233-265. Dordrecht: Springer.
Ungerer, Friedrich
2002
The conceptual function of derivational word-formation in English. Anglia 120: 534567.
Ungerer, Friedrich
2007
Derivational morphology and word-formation. In: Dirk Geeraerts and Hubert Cuyckens
(eds.), The Oxford Handbook of Cognitive Linguistics, 991 - 1025. Oxford: Oxford University Press.
Williams, Edwin
1981
On the notions of 'lexically related' and 'head of a word'. Linguistic Inquiry 12: 245274.
Hans-Jorg Schmid, Munich (Germany)
PA22
22
2. Word-formation research from its beginnings

to the 19th century
I.
2.
3.
4.
5.
Introduction
Greek and Roman antiqu ity
The Germa n tradition (l 6 1h- 191h century)
The French tradition (l 61h- 191h century)
References
Abstract
In Greco-Latin grammatical theory as well as in the first grammars ofEuropean vernacular languages, word-formation was treated within the presentation ofword classes. This
article aims to show how specific theories of derivation and composition were progressively developed by grammarians of German and French.
1. Int roduction
This article starts with a brief survey of research on word-formation in Greek and Roman
antiquity (Aristarchus, Dionysius Thrax, Quintilian, Varro). The following sections treat
theories developed in modern times by major grammarians of German (Schottelius, Adelung, Becker, Grimm, Paul) and French (Meigret, Arnauld and Lancelot, Beauzee, Butet
de la Sarthe, Darmesteter). Our overall focus lies on salient aspects, such as the distinction between derivation and composition and the place of word-formation in grammatical
theory.
2. Greek and Roman antiquity

The assumption that there are two basic processes of word-formation, derivation and
composition, still widely held in contemporary linguistics (Lallot 2008: 63; McLelland
20 10: 13), may be traced back to ancient Greek linguistic thought (Vaahtera 1998: 6076). Aristarchus's analysis of complex words in Greek is based upon the distinction
between derivatives and compounds. The latter are formed by combining two or more
autonomous words, the first by adding derivational suffixes (considered to be non-autonomous) to a word (Matthaios 1999: 254- 272; summarized in Matthaios 2004: 16- 20).
This idea was later developed more fully by Dionysius Thrax (TXV11, ed. Lallot 1998).
It must be stressed that anc ient Greek grammarians did not view word-formation as a
separate domain; rather, they treated it within the framework of word class theory, on
the basis of categories describing properties ("accidents", e.g., number and gender) of
word classes (Matthaios 2004: 8- 9). Of these acc idents, two are specific to word-formation, "Eioo~" [species] and "crx~a." [figure]. Both were attributed to various word classes; the most detailed description was made for the nominal class (Matthaios 2004: 11-
PA23
2. Word-formation research from its beginnings to the 19th century
16). The category species allows for a distinction between "primary" and "derived"
lexical items, while figure is the basis for distinguishing "simple", "composed" and
"decompound" words, the latter term designating nominal derivatives from compounds.
(Greek grammarians used the term "parasyntheta'', in Latin grammar this latter term was
rendered as "decomposita"; cf. article 29 on parasynthesis in Romance.) This view of
word-formation as part of word class theory, including the fundamental dichotomy composition - derivation, was then adopted by grammarians of classical Latin, who latinized
the Greek terms pertaining to word-formation (e.g., "species primitiva"/"derivativa'',
"figura simplex"/"composita"/"decomposita"). A discussion of complex words in classical Latin is to be found in writings by Quintilian, Priscian, and Varro, among others
(Fogen 2008; Lindner 2002). Classical and medieval Latin grammaticography (entitled
" De partibus orationis") is characterized by a canonized system of "parts. of speech"
(word classes), word-formation is considered strictly within this framework, with a focus
on derivation. In early modern times, this word class system in turn constituted the
theoretical framework for the process of "grammatisation" (Auroux 1992: 11 - 64) of
vernacular languages (Kaltz 2000: 693-698).
3. The German tradition (16th_19th century)

The development of word-formation theory in Gennan grammaticography from the 16th
to the 19th century may be briefly summarized as follows. In the first stage of the
"grammatisation" process (l61h-early l7 1h century), grammarians deal with derivation
and composition strictly in terms of the categories "species" and "figura", defined as
accidents of word classes of German (see section 3.1). 17th-century grammarians generally continue to present German word-formation within the section on word classes ("etymologia'', "Wortforschung") though some (Ratke, Schottelius) begin to conceive of
word-formation as being a grammaticographical topic in its own right (see section 3.2).
In the 18th century, theories of word-formation in German gradually gain more independence from the Latin tradition, particularly with Mazke, Heynatz, and Adelung (see section 3.3). 191h-century linguists (Becker, Grimm, Paul, among others) contribute to further development and refinement of word-formation theory as a domain of its own (see
section 3.4).
3.1. 16th_early 17th century

Early grammars of German, mostly written in Latin (e.g., Albertus, Olinger, Clajus)
closely adhere to the conventions of Latin grammar. The focus is on derivation by "terminationes" ('endings') while authors of grammar books pay little attention to composition.
Complex words formed by the addition of "praepositiones'', i.e. either separable or inseparable prefixes, are analyzed as compounds, as are "decomposita'', resulting from both
prefixation and suffixation: "gerecht" ("figura simplex"), "ungerecht" ("figura composita"), "Ungerechtigkeit" ("figura decomposita"; Clajus 1894 [ 1578]: 43 ). Albertus occasionally refers to the notion of "radix" ("root"; 1573 [ 1895]: 66, 74); originally elaborated by grammarians of Hebrew, unknown to the tradition of Greek and Roman grammar
23
PA24
24
(Kaltz 2005: 107- 111), thi s notion was to become a key concept in Schottelius's theory
(see section 3.2).
3.2. 1rh century

While the latinizing tradition is still present in early l 7t11 -century German grammaticography (Padley 1985: 98; note 44), more grammarians now make use of the vernacular,
for pedagogical reasons (e.g., Brucker 1620: 5; Ratke 1959 [1612- 1615], 1959 [1630])
as well as ideological motives (legitimation of the German language; e.g., Schottelius
1641). In his Wortbedeutungslehr [Treatise on the meaning of words), Ratke redefines
the traditional categories "Art" [species] and "Gestalt" [figura], arguing that derivation
and composition are essential means by which the proper and necessary meaning of
discourse is grasped ("wesentliche Mittel [...] daraul3 die Bedei.itung der Rede eigentlich
vnd nothwendig erkennet wird"; 1959 [ 1630]: 277). This work, "the first attempt to provide a semantically based theory of word derivation applied to the vernacular" (Padley
1985: 11 2), is also the first to deal with word-formation not just in the framework of
"Wortforschung", but rather as a topic in its own right (Kaltz 2005: 113-116).
The central figure among 17th-century Gennan grammarians is undoubtedly Schottelius (cf. Gi.itzlaff 1989; McLelland 2010, 2011 ); he had a major impact on late 17111 and
early 18t11-century grammaticography and lexicography (Kaltz 2005: 117- 125). While
Schottelius (1663) follows tradition by discussing derivation and composition within the
section on word classes (" Wortforschung"), he also addresses the topic of word-formation separately. In three Lobreden [discourses of praise], he praises the German language
for its abundance of " Stammworter" or "Wurzelen" [root words], which are monosyllabic, he claims. This section of his grammar may be seen as the first attempt to describe
the constituents of complex words in a somewhat systematic manner. Schottelius analyses both derivatives and compounds as being formed on the basis of root words (1663:
49- 103); the first result from the addition of "Haubtendungen'', i.e. derivative suffixes,
which are now differentiated from inflectional ones ("zufallige Endungen"). The binary
structure of compounds, formed by combining "Stammworter" with suffixes, prefixes or
other root words, is described by the terms "Beygefi.igtes/vorderstes Glied des Wortes"
([attributive/ first element of the word]; i.e. determiner) and "Grund/Haubtglied" ([basis],
[main element]; i.e. primary word; 1663: 75). According to Schottelius, there are four
"Verdoppelungs=arten" [types of compounds] in German. The first class, formed by
nouns only ("welche aus lauter Nennworteren entstehet"; 1663: 77), includes determinative compounds as well as copulative ones (e.g., Freudenpein; 1663: 79); compounds of
the structure "Nennwort + Zeitnennwort" (i.e. a deverbal noun) form the second (e.g.,
Mordbrenner, Taglohner; 1663: 84). The third class assembles compounds formed by
adding a preposition to verbs or nouns ("wan <lurch die Vorworter (Praepositiones) die
Verdoppelung geschiehet"; primarily verbs such as erheirathen, einbrokken, vertiejfen,
but nouns as well, e.g., Gezisch, Gesausel; 1663: 88 f.) while combinations of one or
two Haubtendungen with one or two root words constitute the last ("wan mit einem
oder zweyen Stammwortern eine oder zwo Haubtendungen der abgele iteten verdoppelt
werden", e.g., weibisch, Jungling, Hofnung, Gottseligkeit; 1663: 90). Contrary to
present-day linguistics, "Ableitung" [derivation] refers to suffixation only here (1663:
PA25
318) whereas the term "Verdoppelung" [composition] is used for composition proper
and prefixation, and includes the formation of complex verbs with inseparable prefixes.
3.3. 18th century

Among early 18111 -century grammarians, Longolius is worth mentioning as he maintains
the distinction between " Primitiva" [root words], "Derivativa", and "Composita" (also
labelled "zusammengestuckte Worter"; 1715: 63) but rejects the idea of the monosyllabic
nature of root words (1715: 617 f.). Following Schottelius, he considers the last word
("das letzte Wort") to be the main element ("Grund"), which is merely restricted and
determined by the [word] added ("welchen das beygefiigte nur gewisser Maaf3en restringiret und determiniret"; 1715: 623 ). His analysis of nominal compounds, however,
is somewhat more elaborate than that of Schottelius; in particular, Longolius notes that
verb stems may be first constituents of nominal compounds: "Ein Teutsches Substantivum, das mit einem Verbo componiret ist/bedeutet eine Sache von einem Vermogen zu
derjenigen Verrichtung/so besagtes Verbum anzeigen" [a German noun that is combined
with a verb denotes the ability to perform the action referred to by the said verb], e.g.,
Lockvogel, Heuchelchrist, Hakkebret (1715: 623 f.). The term Decomposita is used here
for complex compounds such as Erzpfajfenfreund and stockpechdickfinster (1715: 627).
Like his predecessors, Aichinger (1754) differentiates "Stammworter" [root words]
and "abstammende" [derivatives], "einfache" [simple words] and "zusammengesetzte"
[compounds] (Kaltz 2005: 130); he also discusses the "Gattung" and "Gestalt" of complex words in the traditional manner as accidents of word classes (1754: 136f., 157).
Aichinger points out an essential difference between derivation and composition in German: "niemand [ darff] leicht sich selber deriuativa schmieden, ausser den Dich tern"
([with the exception of poets, no one has the right to fonn derivatives]; 1754: 139); on
the other hand, he notes that speakers of German have the liberty to create new compounds every day ("die Freyheit, alle Tage neue zusammengesetzte Worter zu machen";
1754: 104).
In the late l 81h century, significant changes in the theory of word-formation occur as
grammarians progressively turn away from the traditional perspective. Heynatz (1777
[1770]: 121) deserves special mention for his analysis of monosyllabic, non-suffixed
deverbal derivatives such as Druck, Schlag, Trieb, while Miizke (1776: 5 f.) criticizes the
use of the traditional term "praepositiones inseparabiles" for inseparable German prefixes
such as be, er, ent (Kaltz 2005: 132 f.). Adelung, the most interesting and most influential
18111 -century grammarian of German, no longer argues in terms of "figura" and "species".
Just as many linguists nowadays, he considers prefixation to be a type of derivation: "Die
Ableitung der Worter geschiehet entweder <lurch Vorsylben oder <lurch Nachsylben, oder
<lurch beide zugleich" ([derivation of words occurs either by prefixes or suffixes, or both];
1784: 97 f.). Derivatives are described in detail within the context of word class theory
while composition is dealt with in a separate lengthy section (Adelung 1782: II, 209274). Adelung stresses the special importance of compounds in German, noting that this
has been a much neglected topic due to the excessive weight accorded to the Latin
tradition (Kaltz 2002; 2005: 131). As some of his predecessors, he insists upon the binary
structure of (determinative) compounds, including complex ones such as Fastnachtspiel,
25
PA26
26
which he describes w ith the terms "Grundwort" [basic/primary word] and "Bestimmungswort" ([determiner]; 1782: II, 215), still commonly used in modem studies of wordformation in German. Copulative compounds (e.g., Fiirst=Bischof), on the other hand,
are analyzed as a type of apposition (Adelung 1782: II, 229).
3.4. 19th century

Throughout the 19th century, word-fonnation was a much discussedl topic in general
grammar as well as in pedagogical (e.g., Becker 1831) and historical-comparative grammar (e.g., Grimm 1878 and article 3 on word-formation in historical-comparative grammar). Becker, who represents both general and pedagogical grammar (Jankowsky 2004;
Forsgren 2008: 134-135), is the author of the first comprehensive study dealing exclusively with word-formation in German (Deutsche Wortbildung, 1824; Kaltz 2005: 136).
The author establishes a di stinction between "Begriffsworter" and "Formworter" ([notional words] vs. [form words]) on the one hand, " Wurzelwort" [root word] and "abgeleitetes Wort" [derivative] on the other. There are two types of derivatives, "Stamm"
([primary form], e.g., Bund) and "Sprof3form" ([secondary form], e.g., bundig). Composition, Becker argues, must be seen as "wahrhafter Ableitungsvorgang" ([true process of
derivation]; Becker 1824: 369 f.) since compounds are new words for new concepts
formed on the basis of existing linguistic material ("Stoff" ), the same way as derivatives
stricto sensu are formed by "Ablautung" [ablauting] and "Umendung" ([change of ending], [suffixation]). In other words, Becker uses "Ableitung" as the comprehensive term
for both derivation and composition. The terms "Bestimmungswort" and "Grundwort",
introduced by Adelung for the constituents of compounds, are reinterpreted by Becker:
the first constituent of compounds is the principal one: "Hauptwort" [main word] whereas the second one is defined as "Beziehungswort" [relational word]. He does follow
Adelung, however, as far as the principle of the binary structure of compounds is concerned, noting that it appl ies to complex compounds as well (e.g., Nuflbaum-holz,
Schneider-handwerk, Herzbeutel-wassersucht; Becker 1831: 45).
J. Grimm had a major impact on 19th_ as well as early 20th-century grammar; in fact,
he was perceived as the initiator of scientific word-formation theory: "Die wissenschaftliche Wortbildungslehre ist [ ... ] eine Schopfung J. Grimms" (Paul 1896: 17). Grimm
starts out by observing that word-formation theory in particular has been unduly neglected in traditional grammar ("zumahl die wortbildungslehre [ist] ungebiihrlich verabsaumt
worden"; 1878: VI). Word-formation, he argues, occurs "<lurch innere anderung oder
<lurch auf3ere mehrung der wurzel" [through internal change or external addition to the
root]. "Zusammensetzung" and "Ableitung" differ as follows: "Zusammensetzung kann
vorne oder hinten an der wurzel eintreten, ableitung nur hinten" ([composition may occur
by combining elements preceding or following the root, derivation by elements following
the root only]; 1878: 1). ln contrast to "innere wortbildung" ([internal word-formation];
1878: 1), derivation (i.e. suffixation only) and composition are dealt with in great detail
(Kaltz 2005: 144-146). Grimm's classification of compounds is based upon the distinction between "eigentliche" and "uneigentliche composition" ([proper]/[improper composition]; e.g., wein-stock, gras-griin vs. tages-Licht; 1878: 386). In spite of his sharp criticism of traditional grammar, he does retain the term "decomposita" for both complex
PA27
compounds (e.g., schlajka.mmer-thiirhiiter) and complex derivatives such as un-absehlich
( 1878: 383, 912). Similarly to Adelung, Grimm stresses the binary structure of determinative compounds (including "decomposita"; 1878: 912), but does not give much consideration to copulative compounds such as christ-kind, tier-mensch and furst-bischof,
which result from "appositionelle verhaltnisse" ([appositional relation s]; 1878: 416).
Grimm's theory of word-formation was later refined by Paul and other Neogrammarians ("Junggrammatiker"; Fleischer 1983). As his 18th-century predecessors Mazke and
Adelung, Paul insists that it is impossible to draw a sharp line between inflection, derivation, and composition: "Die Scheidelinie zwischen Kompositionsglied und Suffix kann
nur nach dem Sprachgefiihl bestimmt werden" ([linguistic intuition is the only way to
decide between a compound constituent and a suffix]; 1880: 348); "auf die gleiche Weise
wie die Ableitungssuffixe entstehen Flexionssuffixe. Zwischen beiden gibt es ja uberhaupt keine scharfe Grenze" ([inflectional suffixes arise the same way as derivational
suffi xes. In fact, there is no sharp line separating both]; 1880: 349). While the focus is
on morphological aspects here, Paul later argues that more consideration should be given
to semantics in w ord-formation theory; when analyzing complex words, "Funktionen"
[functions] representing meaning, not " Bildungsweisen" [word-formation types] ought
to be the primary criterion (1896: 18). In his Deutsche Grammatik (1920), Paul further
distances himself from Grimm by rejecti ng the categories of "proper" and " improper"
composition for nominal compounds as they are motivated by morphological considerations. Instead, referring explicitly to the classification of compounds in Sanskrit grammaticography (Brocquet 2008: 19), he now argues that the logical relation between constituents as well as that between the compound as a whole and its constituents must
be decisive when classifying compounds ("nach dem logischen Verhaltnis der Glieder
zueinander und des Ganzen zu den Gliedern"; 1920: 6). This principle leads him to
differentiate three types of nominal compounds (Kaltz 2005: 150): a) "kopulative" or
copulative compounds (e.g., Prinzregent, Fiirstbischof; = "dvandva" in Indian grammar;
1920: 7), b) "solche, in denen das zweite Glied <lurch das erste bestimmt wird" [compounds in which the second constituent is determined by the first], such as Senfsauce,
Mitbiirger (1920: 9, 23), and c) "possessive Zusammensetzungen" or possessive compounds (e.g., Liigenmaul, Rotkehlchen; = the Indian "bahuvrihi"; 1920: 30). With respect
to the place of word-formation, Paul notes that 19t1i_century grammarians have chosen
various options: following " Flexionslehre" [theory of inflection] and preceding syntax
(e.g., Grimm), or preceding the former. Paul argues that the status of inflectional suffixes
may change as a result of processes of isolation and that their syntactic function changes
as they become "Wortbildungssuffixe" [word-formation suffixes]. Compounds, on the
other hand, result from syntactic structures. Therefore, he suggests that the most appropriate way is to discuss word-formation in a separate section, following syntax (1920:
3 f.). In 20 1h-century German linguistics, the separate treatment of word-formation became predominant with the gradual shift from the historical perspective to the synchronic, though some grammarians still choose the haditional position by integrating wordformation into the description of the word class system (Kaltz 2005: 152).
4. The French tradition (l 61h-1 9th century)

The history of word-formation theories in France from the l 61h to the 19th century may
be summarized as follows. In the first grammars of French, written in Latin or French,
27
PA28
28
word-formation is studied using the conceptual framework of Latin grammar. The richest
developments on lexical morphology are undoubtedly to be read in the first grammar of
French written in French, Le Trr;tte de la gramm?re Jram;o?ze (Meigret 1980 [1 550])
(see section 4.1 ). 17th-century grammarians of French no longer refer to the Latin categories of "espece" [species] and "figure" [figure]. Even if the contrasts between primitive
and derivative or between simple and compound are sometimes mentioned, general
grammar assigns the task of dealing with derivation and composition to the dictionary
(see section 4.2). Word-formation is then reintroduced into the field of grammar with
Beauzee's theses. The encyclopaedist's approach to derivation and composition - based
on the theory of the sign specific to general grammar - had a long-standing influence
(see section 4.3). In the l 9 1h century, word-formation theories are mostly developed
outside the field of grammar, first and foremost in Darmesteter's works (see section 4.4).
4.1. 16th century

From Sylvius's Grammatica latino-gallica (1531) to Ramus's Grammaire (1572), derivation and composition were regarded as word accidents, in compliance with the GrecoRoman tradition, with the consequence that word-formation was integrated into grammar
but had no real autonomy: it was included in the presentation of word classes. The
species implies an opposition between "primitifs" and "derivatifs" and corresponds to
suffixal derivation whereas the figure opposes "simples" and "composes" and describes
combinations of at least two words considered as autonomous words. Both the species
and the figure are discussed along with the other word accidents - essentially morphosyntactic variations. The parts devoted to noun formation are by far the most developed
and some elements pertaining to lexical morphology can also be read into the description
of the category of "diminutif' [diminutive], traditionally dealt with separately from other
suffixed derivatives. 16111-century grammarians pay variable attention to the species and
the figure, most of them simply asserting the two dichotomies and exemplifying them.
The processes of composition are sometimes studied more precisely. Estienne, for example, distinguishes three types of formations taken up from Donat (Colombat 1999: 243):
two "whole words" ("mots entiers"; e.g., malheur), a ''whole word" and a "corrupted
word" ("mot corrompu"; e.g., ennemi), a "corrupted word" and a "whole word" (e.g.,
chascun; 1557: 17). Meigret, known for his interest in morphology (Glatigny 1985),
added a fourth pattern based on the association of two corrupted words (e.g., benivole;
1980 [ 1550]: 49). But Le Tr~tte, which uses French as metalanguage for the first time,
deals above all with derivation in long and detailed discussions in chapter 8 ("Des
noms") and chapter 12 ("Des denominatifs", that is to say proper nouns derived from
nouns but nouns derived from verbs, for example, are addressed in the same chapter as
well). Meigret attempts to identify different classes built, in particular, on semanticoreferential categories: French derivatives can result from names of "affections et qualites" ([affections and qualities]; e.g.,fievreux), "arts et sciences" ([arts and sciences]; e.g.,
mathematicien), "chefs de disciplines et sectes" ([heads of disciplines and sects]; e.g.,
platonique), etc. But the classification is primarily organized around suffixes, which are
extensively listed. It appears that suffixal derivation is one of the privileged features
enabling the differentiation between Latin and the vernaculars. The linguistic filiation is
PA29
indeed systematically studied in the chapter "Des noms": it must also be observed that
all the Latin derivatives ending in cus, or Greek derivatives ending in cos become qe or
9ien ("11 faut aussi entendre que nous toumons en qe ou 9ien tous les derivatifs que nous
tirons de la langU1e latine termines en cus, ou de la grecque en cos"; 1980 [1550]: 27).
However, it is also important to highlight the specificities of the French language: the
Latin ending -arius becomes -(jre in French, but in the case of Censorius, it becomes
<;(fnsorin and not <;(fnSo(jre: because it does not sound right ("au regard de Censorius,
nous le tournons en <;(fnsorin et non pas <;(fnSO(jre: parce qu 'il sonne mal a l 'oreille";
1980 [1550]: 28). Meigret tries to shed light on general principles whenever possible
(for example, the relation between the grammatical classes of derivative and primitive
or between the suffix and the gender of the derivative), but he is also very attentive to
restrictions on usage. While the form and the meaning of derivatives are often associated
in an attempt at generalisation, the grammarian also remarks that the same suffix does
not always have the same meaning and that one meaning can be expressed by different
suffixes.
4.2. 17th centu ry

After Ramus's Grammaire, no more references to the categories of species and figure
can be found in the grammars of French. The contrast between primitive and derivative
or between simple and compound may be mentioned from time to time in re lation to the
class of the nouns (Chifflet 1659: 8 and lrson 1662 [ 1656]: 18- 19), but a specific section
is no longer devoted to them in the presentation of word classes. However, Irson 's
Nouvelle methode (1662 [ 1656]) contains a final section entitled "Les etymologies",
which is a list of entries arranged alphabetically. These entries correspond to a collection
of comments on both the origin of the words and their derivatives in compliance with
etymology in its traditional sense (Delesalle and Maziere 2002). A parallel can be drawn
between thi s insertion of a lexical section at the end of a grammar and 17th-century
distribution of the roles of grammar and the dictionary in the treatment of word-formation. This distribution is explicitly asserted by Arnauld and Lancelot's Grammaire generate et raisonnee in which we find the following passage (1 660: 105): "On n'a point
parle, dans cette G rammaire, des mots derives ni des composes, dont il y aurait encore
beaucoup de choses tres-curieuses a dire, parce que cela regarde plutot l 'ouvrage d 'un
Dictionnaire general, que de la Grammaire Generate" [derived and compound words
are not mentioned in this grammar even though a great many fascinating things remain
to be said on the subject but they belong in a General Dictionary rather than in a General
Grammar]. The authors of the Grammaire generate, however, assume a certai n form of
derivation - between substantives and adjectives - by contrasting "signification" and
"maniere de signifier" [manner of signifying]. All nouns which appear independently in
discourse ("subsistent par eux-memes dans le discours") - which represents their "manner of signifying'' - are called nouns, even if they signify "accidents" (e.g., rougeur).
Conversely, if the element expressing "connotation" - which implies that an adjective
cannot be self-sufficient in discourse - is added to a noun which signifies a substance,
the process results in the creation of an adj ective (e .g., humain; Delesalle 1990). As
shown, morphological regularity depends above all on the mental conceptions of things,
29
PA30
30
which is consistent with the programme of French general grammar: Arnau ld and Lancelot show that derivation is a linguistic fact contributing to the expression of thought.
Furthermore, only the transformation leading to the creation of an adjective noun from
a substantive noun is regarded as enabling generalization. When the link between form
and meaning in the lexicon can no longer be systematized, the reader is invited to turn
from grammar to the dictionary, or in other words not expect a rule but to consult a list
of items.
The Port-Royal logic and linguistic theory were taken up in the Dictionnaire de
l' Academie fram;aise (1694) with a view to applying them to a particular language. This
dictionary, which aims at showing the creativity of the French language, is important for
lexical morphology for two reasons. Firstly, at a micro-structural level, it enables the
emergence of morphosemantic definitions (Maziere 1996; Leclercq 2002). The syntactic
forms and semantic categories used at the beginning of the definitions attest to the
remarkable stability characterizing the definitions of derivative words. These recurring
syntactico-semantic patterns allow an organization of the definition which renders the
link between the base form and the derivative explicit. Secondly, at a macro-structural
level, not only does the dictionary set up a morphological arrangement of entries but more importantly - it imposes constraints (explained in the preface) on this classification,
which shows that derivation and composition are perceived in synchrony, regardless of
etymology (Leclercq 2002). Essentially, the arrangement is used only if a French primitive word is attested: construire and destruire, for example, are regarded as independent
since Latin struere has not come down to French ("n 'a point passe en frarn;:ais"). To be
taken into account in the nomenclature, the formal structure of the derivation must be
observable in the French language from a synchronic point of view. It should be noted
that the academician Regnier-Desmarais, in his grammar published in the early l 8 1h
century, distinguished between the etymon and the base form in synchrony: Simple or
primitive nouns [ . .. ] are those which do not derive their origin from another noun of the
same language, but owe their signification to the first institution of this language ("Les
noms simples ou primitifs [ ... ] sont ceux qui ne tirent point leur origine d'un autre nom
de la meme langue, mais qui doivent leur signification a la premiere institution de cette
langue"; 1705: 179).
4.3. 1gth_ea rly 19th cent ury

Regnier-Desmarais is one of the few 181h-century grammarians of French who devoted
a substantial discussion to derivatives and compounds (both are studied in the section
dealing with the noun). As in the l 71h century, word-formation was very rare in particular
grammars. However, Beauzee reintroduced it into the grammarian's field of studies,
integrating it into general grammar. He didn't address the question in his Grammaire
generate (1767), but in several articles of the Encyclopedie (1751-1765). Through a
series of subdivisions (article "Grammaire", cf. Beauzee l 75 l- 65b), word-formation was
made a part of "etymology", which was itself a part of "lexicology" (a term which,
according to Beauzee, was introduced by Girard), lexicology in its turn being included
in grammar. Word-formation groups together composition, derivation and inflection
(derivation and inflection are both derivational processes that use "inflexions'', the first
PA31
being called "philosophical derivation" and the second "grammatical derivation" ). But in
the a11icle "Formation" (cf. Beauzee 175 l - 65a; probably written with Douchet; Bourquin
l 980a: 25), Beauzee blames the grammarians who preceded him for having preferred
inflection at the expense of derivation. According to him, the explanation lies in the
widespread position that grammar may legitimately restrict its focus to ready-made
words ("les mots tout faits"). The encyclopaedist justifies the reintegration of the lexical
aspects of word-fonnation into grammar using an argument compatible with the principles of general grammar: derivation and composition not only suppose a uniformity in
processes ("uniformite de precedes") within a language, but also some feature common
to all languages (article "Formation"). More precisely, the link between form and meaning is interpreted as a regular phenomenon: "Nous disons en premier lieu, que ces terminaisons sont soumises a des lois genera/es, parce que telle tenninaison indique invariablement une meme idee accessoire" [First of all we believe that these endings are subject
to general laws, because an ending invariably indicates the same accessory idea]. The
theory of word-formation is incorporated into the theory of the sign that characterizes
general grammar: both derivation and composition allow a "primitive" or "fundamental"
idea to be modified due to an accessory idea. However, despite the grammarian's asserting the principle of generality, it should be noted that most of the examples given by
Beauzee are in Latin and the author explicitly points out that French has fewer regularities in this field than ancient languages.
In the early 19111 century, influenced by the work of etymologists (especially de Brosses and Le Bel) and probably inspired by the intense activity of lexical creation that
characterized the revolutionary period (Dougnac 1982; Steuckardt 2008), Butet de la
Sarthe developed and above all implemented Beauzee 's rationalist theory. He produced
a new "science of words" (1801: 2), as he called it in his Abrege d'un cours comp/et de
lexicologie. We can consider this book as the first systematic description of formation
processes in French (le "premier traitement systematique des precedes de formation en
fran9ais"; Schlieben-Lange 2000: 31). This book isolates "lexicology" from grammar
and gives it autonomy. However this autonomy is relative in fact since suffixal derivation
and inflection are considered jointly, as was the case in Beauzee's work. The systematic
approach inaugurated by Beauzee and Douchet is reinforced. Butet, who taught physics,
exposes not simply a metaphor but an analogy between the combinations that give rise
to words and the molecules of matter assembled into bodies: both obey laws and are
reducible to formulae. His theses on word-formation aim at implementing a semantic
structuring of the lexicon and are fundamentally based on the compositionality of meaning. He takes up his predecessor 's theory of the sign: "prepositions" and "endings" are
"accessory ideas" which change the main idea expressed by the "root". However, the
classification of the resulting words that is proposed is tripartite and no longer bipartite.
Indeed Butet distinguishes between "radical constructions" ("constructions radicales" ),
which only consist of "roots'', that is to say elements used to represent the "main idea"
("primitive" or "simple"), "prepositive constructions" ("constructions prepositives") and
"postpositive constructions" ("constructions postpositives'', which comprise suffixal
derivation and inflection). So composition is divided and organized with a view to singling out "compound words'', as they are called nowadays. The effort of completeness
and systematicity is otherwise remarkable, even though the classification of suffixes,
which is based on meaning rather than form, does not take into account the diachronysynchrony opposition: the formation of French words is put next to that of Latin words
31
PA32
32
and French words resisting analysis are placed on the same footing as French derivatives
(words like instant or proteger constitute examples of prepositive constructions).
4.4. 19 1h century
In the 19th century, grammarians dealing with word-formation did not innovate and
generally adopted a pedagogical perspective: suffixal derivation, related to etymology,
is mainly seen as a way of focusing on the meaning of words and above all their spelling
from a didactical point of view (Jullien 1849: 150). Word-formation is still not clearly
separated from etymology and priority is always given to meaning, in compliance with
the tradition of general grammar. During this period, progress is made above all in books
which are 011.1tside the field of grammar. One example is the case of synonym dictionaries
that offer a systematic description of suffixes through the study of morphologically related synonyms. It is exemplified by the work of Lafaye (1858 [1841]). Already addressed
by Roubaud (1785) and taken up by Guizot ( 1809), this systematic description is taken
one step forward by Lafaye, the author of the Dictionnaire des synonymes de la langue
fran<;aise ( 1858 [1841]). He introduces a distinction between synonyms formed from
different stems and grammatical synonyms. With the latter, whose "prefixes" or "endings" constitute the main difference (e.g., renunciation and renoncement), it is possible
to extend the difference found in a particular example to al l others sharing the same
modification (" faire servir la difference trouvee dans un exemple particulier a la distinction de tous les autres qui presentent la meme modification"; 1858 [ 1841 ]: 26). Thus
synonym dictionaries create lists of suffixes that were taken up during the 19th century.
But the pivotal work of the century is of course Darmesteter's. While associated
with historical grammar (his Cours de grammaire historique de la langue franr;aise was
posthumously published from 1891 on), along with Raynouard and Diez, the author
breaks new ground by dealing with "today's language", "living language" and by defining a field of study centered on the creativity of the French lexicon. His aim is to separate
it from the historical study of word-formation. In his book De la creation actuelle des
mots nouveaux dans la langue franr;aise, he proposes to shift the point of view on the
language: he concentrates on word -formation seen as a creative process in progress and
not on the historical development of French vocabulary (1877: 1-2). Darmesteter takes
into consideration three strategies of lexicon enrichment: a) "French formation", lexical
constructions created from French "stems" (the term "popular formation" is used but
rejected by Darmesteter), b) "learned formation", corresponding to borrowings from
Greek or Latin and derivatives and compounds derived from Greek and Latin stems and
c) bon-owing from modern languages. Concern for the observation of the "real language"
is also noticeable in the expansion of the notion of demotivation: a derived word can be
considered as a single word if it does not express a "double-idea" (1877: 70), that is to
say if compositionality of meaning is no longer obvious. Darmesteter's theories are also
characterized by a detailed classification of the different modes of formation. Derivation
is split into "derivation propre" (using a suffix) and "derivation impropre" (covering the
processes now called "conversion" and "back formation"). Composition (to which he
devotes a separate work entitled Traite de la formation des mots composes dans la
langue franqaise, 1894 [ 1875]), which is defined by its ability to create a conceptual
PA33
unity which obliterates the specific meanings of the elements, implies a division between
compounds formed by juxtaposition ("les composes formes par voie de juxtaposition";
e.g., plafond, arc-en-ciel), compounds formed with particles, viz. adverbs or prepositions
("Jes composes formes a 1'aide de particules"; e.g., surprendre, malaise) and compounds
formed by compo sition as su ch ("les composes formes par composition proprement
<lite"; e.g., chou-fleur, timbre-paste). Although isolated, prefixation is therefore still regarded as being part of composition and the term of "parasynthesis'', taken over from
the Greek grammatical tradition, is introduced with its meaning of "resultat d'une composition et d ' une derivation agissant ensemble sur un meme radical" [result of a composition and a derivation acting together on the same stem] (1894 [1875]: 96). The opposition between the two remaining sets is founded on the nature of the syntactic relationships between the various elements of the compounds, with the use of the notion of
"ellipsis": the composition as such is indeed elliptical, the compound is in this case a
" proposition en raccourci" - "timbre-paste ne veut pas dire simplement timbre et paste,
mais timbre de la paste, timbre pour la paste" (1890: 72) [postage-stamp doesn't mean
just stamp and postage, but stamp for postage] - while composition by juxtaposition
follows the usual syntactic construction. This innovative syntactic theory of composition
was taken up widely in the 20th century, especially in a transformational perspective
(particulary in Guilbert's and Dubois's works). More generally, Darmester's theses and
categories came down with minor changes through the first half of the 20th century and
are still used today, despite Saussure's reflections on synchrony and analogy (Kerleroux
2000).
Acknowledgements
We are most grateful to Helmut Puff (University of Michigan, Ann Arbor) and Wilfrid
Andrieu (Universite d e Provence) for their valuable help with the English version.
5. Refe rences
Adelung, Johann Christoph
1782
Umstiindliches Lehrgebiiude der Deutschen Sprache. Zur Erliiuterung der Deutschen
Sprachlehre far Schulen. Vol. 2 . Leipzig: Breitko pf. [Repr. Hildesheim: 0 lms, 1971 ).
1784
Gebrauch und MiI3brauch der Etymologie. Magazin fur die D eutsche Sprache II(2): 9612 1.
Aichinger, Carl Friedrich
1754
Versuch einer teutschen Sprachlehre, anfanglich nur zu eignem Gebrauche unternommen, endlich aber, um den Gelehrten zu fernerer Untersuchung Anlaj3 zu geben, ans
Liecht gestellt von C.F.A. Frankfurt/Leipzig: Kraus. [Repr. Hildesheim: ODms, 1972].
Albertus, Laurentius
1573 [1895]
Teutsch Grammatick oder Sprach-Kunst. Augsburg: Manger [Die deutsche
Grammatik des Laurentius Albertus. Ed. by Carl Miiller-Fraureuth, 1895, StraI3burg:
Triibner].
Arnauld, Antoine and Claude Lancelot
1660
Grammaire generate et raisonnee. Paris: Le Petit.
33
PA34
34

Auroux , Sylvain
1992
Histoire des idees linguistiques. Vol. 2: l e developpement de la grammaire occidentale.
Liege: Mardaga.
Beauzee, Nico las
l 75 l- 65a
Formation. In: Denis Diderot and Jean Le Rood d'Alembert (eds.), Encyclopedie
ou Dictionnaire raisonne des sciences, des arts et des metiers. Vol. 7, 172- 176. Paris/
Neuchatel: Le Breton.
Beauzee, Nicolas
1751-65b
Grammaire. In: Denis Diderot and Jean Le Rond d'Alembert (eds.), Encyclopedie
ou Dictionnaire raisonne des sciences, des arts et des metiers. Vol. 7, 841-847. Paris/
Neuchatel: Le Breton.
Beauzee, Nicolas
1767
Grammaire generate. Paris: Barbou.
Becker, Ka rl Ferdinand
1824 Die D eutsche Wortbildung oder die organische Entwickelung der deutschen Sprache in
der Ableitung. Frankfurt/M.: Hermann'sche Buchhandlung. [Repr. Hildesheim: Olms,
1990].
Becker, Karl Ferdinand
183 1 Schulgrammatik der deutschen Sprache. Frankfurt/M.: Hermann'sche Buc hhandlung.
Bourquin, Jacques
1979
La place et la fonction de la morphologie derivationnelle dans la grammaire scolaire au
x1xes iecle. langue fran<;aise 4 1: 60- 76.
Bourquin, Jacques
1980a La derivation suffixale (theorisation et enseignement) au XlXe siecle. These presentee
devant l'Universite de Besarn;:on ( 1977). Lille: Universite de Lille TJI.
Bourquin, Jacques
1980b La terminologie du lexique construit (derivation suffixale et prefixation). Langue fran<;aise 47: 33- 47.
Brocquet, Sylvain
2008
Les mots non simples dans la trad ition indienne panineenne. In: Barbara Kaltz (ed.),
Regards croises sur !es mots non simples, 11-33. Lyon: ENS Ed itio ns.
Brucker, Jacob
1620
Teutsche Grammatic I das ist I Kurtzer Vnterricht I wie eyner etlicher massen recht
reden und schreiben lehrnen so/le. Frankfurt/M.: Jennis.
Butet de la Sarthe, Pierre Roland Fran9ois
1801
Abrege d'un cours comp/et de lexicologie a !'usage des eleves de la quatrieme classe
de I' Ecole polymathique. Paris: Renouard.
Chiftlet, Laurent
1659
Essai d' une parfaite grammaire fran<;aise. Anvers: van Meurs.
Clajus, Johannes
1894 [1578]
Die deutsche Grammatik des Johann es Clajus. Nach dem iiltesten Druck von
15 78 mit den Varianten der iibrigen A usgaben. Ed. by Friedrich We idling. Freiburg :
Triibner.
Colombat, Bernard
1999
La grammaire latine en France a la Renaissance et al' Age classique. Grenob le: ELLUG
(U niversite Stendhal).
Corpus de textes linguistiques fondamentaux: http://ctlf.ens-lyon.fr
Darmesteter, Arsene
1877
De la creation actuelle de mots nouveaux dans la langue fran<;aise. Paris: Vieweg.
Darmesteter, Arsene
1894 [ 1875]
Traite de la formation des mots composes de la langue fran<;aise. Paris: Vieweg.
PA35
Darmesteter, Arsene
1895
Cours de grammaire historique de la langue franqaise. Vol. 3: Formation des mots et
vie des mots. Paris: Delagrave.
Delesalle, Simone
1990
De la definition du nom et du verbe dans la Logique et la Grammaire de Port-Royal.
In: Jacques Chaurand and Francine Maziere (eds.), La definition, 72-77. Paris: Larousse.
Delesalle, Simone and Francine Maziere
2002
La liste dans le developpement des grammaires. Histoire Epistemologie Langage 24( 1):
65-92.
Dictionnaire de I'Academie franr;aise
1694
Paris: Coignard.
Dougnac, Franc;oise
La neologie. Histoire Epistemologie Langage 4(1): 67- 72 .
1982
Estienne, Robert
1557
Traite de la grammaire franqaise. Geneve: Estienne.
Fleischer, Wolfgang
Zur Geschichte der germanistischen Wortbildungsforschung im 19. Jahrhundert: Jacob
1983
Grimm und die Junggrammatiker. In: Wolfgang Fleischer (ed.), Entwick/ungen in Wortbildung und Wortschatz der deutschen Gegenwartssprache, 74- 100. Berlin: Akademie
der Wissenschaften der DDR - Zentralinstitut fi.ir Sprachwissenschaft.
Fogen, Thorsten
2008
La formation des mots et l'enrichissement de la langue vus par quelques auteurs latins.
In: Barbara Kaltz (ed.), Regards croises sur !es mots non simples, 65-84. Lyon: ENS
Editions.
Forsgren, Kjell-Ake
2008
La conception de la formation des mots selon Karl Ferdinand Becker. In: Barbara Kaltz
(ed.) , Regards croises sur /es mots non simples, 131- 149. Lyon: ENS Editions.
G latigny, Michel
1985
L'exception dans le systeme morphologique de L. Meigret. Languefranqaise 66: 9-19 .
Grimm, Jacob
1878
Deutsche Grammatik. Zweiter Th eil. Besorgt <lurch Wilhelm Scherer. 2nd ed. Berlin:
Diimmler. [Repr. Hildesheim: Olms, 1989].
Gi.itzlaff, Kathrin
Von der Fugung Teutscher Stammworter. Die Wortbi/dung in J. G. Schottelius' 'Ausfuhr1989
/icher Arbeit von der Teutschen HaubtSprache'. Hildesheim: Olms.
G uizot, Franc;ois
Nouveau dictionnaire universe/ des synonymes. Paris: Payen.
1809
Heynatz, Johann Friedric h
D eutsche Sprachlehre zum Gebrauch der Schulen. 3rd ed. Berlin: Mylius.
1777 [1770]
Irson, C laude
Nouvelle methode pour apprendre facilement /es principes et la purete de la
1662 [ 1656]
langue franqaise contenant p/usieurs traites . 2nd ed. Paris: Beaudouin.
Jankowsky, Kurt R.
2004
Karl Ferdinand Becker's ( 1775- 1849) concept of word formation within the framework
of hi s general linguistic theory. In: Kjell-Ake Forsgren and Barbara Kaltz (eds.), Studien
zur Geschichte der Wortbi/dungstheorien, 89-106. Munster: Nodus.
Jullien, Bernard
1849
Cours superieur de grammaire. Paris: Hachette.
Kaltz, Barbara
Zur Entwicklung der Wortbildungstheorie in der deutschen Grammatikographie 17502002
1800. Beitriige zur Geschichte der Sprachwissenschaft 12(1): 27- 47.
35
PA36
36

Kaltz, Barbara
2005
Zur Herausbildung der Wortbildungs lehre in der deutschen Grammatikographie: Von
den Anfiingen bis zum Ende des 19. Jahrhunderts. In: Peter Schmitter (ed.), Sprachtheorien der Neuzeit IIIIJ, 105-162. Tiibingen: Narr.
Kerleroux, Fran9oise
2000
France and Switzerland. In: Geert Booij, Christian Lehmann and Joachim Mugdan
(eds.), Morphology. An International Handbook on Inflection and Word-Formation. Vol.
1, 138-145. Berlin/New York: de Gruyter.
Lafaye, Pierre Benjamin
1858 [1841]
Dictionnaire des synonymes de la languefranc;aise. 2nd ed . Paris: Hachette.
Lallot, Jean
1998
La grammaire de Denys le Thrace. Traduite et annotee. 2nd ed. Paris: Editions du
CNRS.
Lallot, Jean
2008
De Platon aux grammamens: Regards grecs sur la structure des mots non simples.
In: Barbara Kaltz (ed.), Regards croises sur Les mots non simples, 51-63. Lyon: ENS
Editions.
Leclercq, Odile
2002
Aspects grammaticaux d ' un dictionnaire de langue: Deux traitements de la morphologie
par le Dictionnaire de l'Academie ( 1694). Histoire Epistemologie Langage 24(1): 107118.
L indner, Thomas
2002
Lateinische Komposita. Morphologische, historische und lexikalische Studien. Innsbruck: Institut fiir Sprachwissenschaft.
Longolius, Johann Daniel
1715
Einleitung zu griindlicher Erkiintnift einer ieden I insonderheit aber Der Teutschen Sprache I Welcher man sich Zu accurater Untersuchung j eder Sprache I und Besitzung
einer untadelhafften Beredsamkeit in gebundenen und ungebundenen Reden I Wie auch
besonders In Teutschen far allerley Condition, Alter und Geschlechte I Zu einem deutlichen und niitzlichen Begriff der Mutter=Sprache I bedienen kan. Budiss in [Bautzen]:
Richter.
Mazke, Abraham Gotthe lf
1776
Grammatische Abhandlungen iiber die Deutsche Sprache. Breslau: Meyer.
M atthaios, Stephanos
1999
Untersuchungen zur Grammatik Aristarchs. Texte und Interpretation zur Wortartenlehre.
Gottingen: Vandenhoeck & Ruprecht.
Matthaios, Stephanos
2004
Die Wortbildungstheorie in der alexandrinischen Grammatik. In: Kjell-Ake Forsgren
and Barbara Kaltz (eds.), Studien zur Geschichte der Wortbildungstheorien , 5-22. Munster: Nodus.
Maziere, Francine
1996
Un evenement linguistique: La definition des noms abstraits dans la premiere edition du
Dictionnaire de l'Academie ( 1694). In: Nelly Flaux, Michel Glatigny and Didier Samain
(eds.), Les noms abstraits. Histoire et theories. Actes du collogue international "Les
no ms abstraits" (Dunkerque, sept. 1992), 16 1- 174. Lille: Presses Universitaires du Septentrion.
McLelland, Nicola
2010
Justus Georgius Schottelius ( 1612- 1676) and European linguistic thought. Historiographia Linguistica 37(1): 1-30.
McLelland, Nicola
2011
J. G. Schotteliuss Ausfahrliche Arbeit von der Teutschen Haubtsprache (J 663) and its
place in early modern European vernacular language study. Oxford: Blackwell.
PA37
Meigret, Louis
1980 [1550]
L e traite de la grammairefranr;aise. Ed. by Franz Josef Hausmann. Tiibingen:
Narr.
Padley, George Arthur
1985
Grammatical Theory in Western Europe 1500-1700. Trends in Vernacular Grammar 1.
Cambridge: Cambridge University Press.
Paul, Hermann
1880
Prinzipien der Sprachgeschichte. Halle/S.: Niemeyer. [Repr. Tiibingen: Niemeyer, 10 1"
ed. 1995].
Paul, Hermann
Ueber die Aufgaben der Wortbildungslehre. In: Sitzungsberichte der philosophisch-phi1896
lologisclren und der historischen Classe der k.b. Akademie der Wissenschaften zu
Miinchen, 692-713. [Repr. in: Leonhard Lipka and Hartmut Gunther (eds.), Wortbildung,
17-35. Darmstadt: Wissenschaftliche Buchgesellschaft, 1981].
Paul, Hermann
1920
Deutsche Grammatik. Vol. 5: Wortbildungslehre. Halle/S.: Niemeyer.
Ramus, Petrus
Grammaire. Paris: Wechel.
1572
Ratke, Wolfgang
Sprachkunst. In: Erika Ising (ed .), Wolfgang Ratkes Schriften zur deut1959 [1612- 1615]
schen Grammatik. Vol. 2, 7-22. Berlin: Akademie Verlag.
Ratke, Wolfgang
1959 [1 630]
Die WortbedeutungsLehr der Christlichen Schute[ ... ]. In: Erika Ising (ed.), Wolfgang Raikes Schriften zur deutschen Grammatik. Vol. 2, 269-318. Berlin: Akademie
Verlag.
Regnier-Desmarais, Frarn;:ois Seraphin
1705
Traite de la grammaire franr;aise. Paris: Coignard.
Rey, Alain
La lexicologie. Paris: Klincksieck.
1970
Roubaud, Pierre-Joseph
1785
Nouveaux synonymes franr;ois. Paris: Moutard.
Schlieben-Lange, Brigitte
2000
La revolution fran~a ise. In: Sylvain Auroux (ed.), Histoire des idees linguistiques. Vol. 3,
23-34. Liege: Mardaga.
Schottelius, Justus Georg
1641
Teutsche Sprachkunst [.. .]. Braunschweig: Gruber.
Schottelius, Justus Georg
Ausfahrliche Arbeit Von der Teutschen HaubtSprache. Braunschweig: Zilliger. [Repr.
1663
Tiibingen: N iemeyer 1967; online].
Steuckardt, Agnes
2008
Presentation. In: Pierre-Nicolas Chantreau, Dictionnaire national et anecdotique. 1790.
Reprint ed. by Agnes Steuckardt, 9- 91. Limoges: Lambert- Lucas.
Sylvius, Jacobus Ambianus
153 l
Grammatica Latino-Gallica. Paris: Estienne.
Vaahtera, Jaana
Derivation. Greek and Roman Views on Word Formation. Turku: Turun Yliopisto.
1998
Barbara Kaltz, Freiburg (Germany)

Odile Leclercq, Aix-en-Provence (France)
37
PA38
38
3. Word-formation in historical-comparative grammar

I.
2.
3.
4.
5.
6.
Terminological preliminaries
Word-formation immediately before the rise of the historical-comparative method
Historical-comparative descriptions
Descriptions based on semantic criteria
Compounding in comparative grammar
References
A bstract
Word-fo rmation in historical-comparative linguistics emerged on the one hand from
Classical and German studies at the turn of the J 81h and 19'11 centuries and on the other
hand from Indian grammar which had become known in Europe early in 1800. The
present article tries to delineate the main developments from the 19th century to the
present.
1. Term i nologica l pre limina ries

1.1. Basic term inology
Morphological terms such as root, stem or affix as well as the segmentation of words
and the consciousness of word-formational processes which they imply represented fundamental linguistic insights that provided concepts of morphological analysis which
lndo-European studies, but also other grammatical traditions, could no longer do without.
Nevertheless, in comparison with terms relating to syntax, parts of speech or case they
are not really old: they do not, as one might be inclined to think, reach back to antiquity,
nor were they well-established elements of traditional terminology at the outset of lndoEuropean studies. Morphological analysis was neglected in ancient grammar: the procedures of analyzing words which are now so familiar to linguists and which Bopp (18243 1) referred to as dissection [Zergliederung] were unknown to the ancient grammarians,
concepts such as ' root', 'stem' or 'affix' were also completely unfamiliar to them.
In the wake of the Indian grammatical tradition, Franz Bopp was the first to recognize
that Indo-European words could generally be broken down into the structure root +
derivational affix + inflectional affix. The identification of what was indistinctly called
Grundform 'basic form ', Stammform ' stem form', Stamm 'stem' or Thema ' theme', was
so new and groundbreaking at the time that Bopp felt obliged to provide the following
clarifications:
Die lndischen Grammatiker fassen die Nomina (sowohl Substantive, als Adjektive, Pronomina und Zahlworter), in ihrem absoluten, von allen Casusverhiiltnissen unabhiingigen, und
von alien Casuszeichen entblo13ten Zustande auf, und nehmen daher eine Grund- oder
Stammform an, zu welcher der N ominativ und die obliquen Casus der drei Zahlen sich als
PA39
3. Word-formation in hi storical-comparative grammar
abgeleitet verhalten. Diese Grundform kommt hiiufig in zusammengesetzten Wortern vor,
indem die ersten Glieder eines Compositums aller Casusendungen beraubt, und somit identisch mit der Grundform sind. (Bopp 1827: 23)
[The Indian grammarians conceived of nouns (substantives, adjectives, pronouns and numerals) in their absolute state, independent of all relations and markers of case, assuming the
existence of a basic or stem form, from which the nominative and the oblique cases of all
three numbers were derived. This basic form often appears in compound words, the first
members of compounds being deprived of all case endings and therefore identical with the
basic form.]
These stem forms of the old Indo-European languages thus have only been recognized
in the Western world since Bopp and his disciples (cf. Lindner 2012: 121). However,
Jacob Grimm in his Deutsche Grammatik (18 19 ff.) did not yet recognize the stem principle: In the analysis of the first members of compounds, where it should have been most
obvious, he resorted to the concept of a "composition vowel" (Compositionsvocal) which
would haunt Indo-European studies until the early Neo-Grammarians (cf. Lindner 2012:
115). German linguists soon adopted Bopp's analysis, first and foremost among them
was Eberhard Gottlieb Graff (cf. Lindner 2012: 122), but the classical handbooks continued to perpetuate the old view codified over the centuries in school grammars of Latin
and Greek. Philipp Buttmann was the first to use and differentiate the concepts of root
(Wurzel), stem (Stamm) and theme (Th ema), but he did not recognize that the first members of compounds were stems, calling them declensional or nominal endings, or linking
vowels (cf. Lindner 2012: 124). Nevertheless, he was the first classical phifologist who
had recognized, at least theoretically, the role of the stem in the segmentation of Greek
nominal inflection, especially of the third declension. However, he did not yet recognize
clearly the amalgamation of stem vowels and inflectional desinences, which is why
segmentation did not find its way into school grammars for still quite some time. We
will have to wait until well into the second half of the 19th century to see Bopp's insight
establish itself in grammars of the classical languages, essentially thanks to the indefatigable endeavors of his disciple Georg Curtius. It was especially the precise conceptual
distinctions in Cmtius' Griechische Schulgrammatik (1852) which eventually enhanced
general consciousness regarding morphological structure.
As we have seen, only comparative linguistics and, hesitantly, classical grammar
started differentiating on the one hand root (Germ. Wurzel) and stem (Stamm), and on
the other ending (Endung) and termination (Ausgang). The latter distinction has become
established in school grammars and has lived on in scientific discourse until the present
day. As a consequence Ausgang still refers to the amalgamated combination of a stemforming element and an inflectional morpheme and probably constitutes the ending par
excellence for the unsophisticated speaker (cf. Lindner 2012: 138). Alongside the terms
root, stem, (inflectional) ending, and termination, which became established as late as
around 1870 in scientific morphology, Hermann Perthes introduced Wortstock 'root or
stem, respectively' exclusively for pedagogical purposes from 1875 onwards (cf. L indner
2012: 138). As this terminology was becoming popular in school grammars of the classical languages in the last quarter of the 19th century, an umbrella term for morphological
elements was introduced around 1880 in entirely different scientific contexts, viz. morpheme. It had been coined by Jan Baudouin de Courtenay in the Kazan school of linguistics towards the end of the decade of 1870, taking as a model the term phoneme which
was already in use (cf. Lindner 2012: 140). Thema, originally a synonym of Stamm
39
PA40
40
'stem ' in morphology, also shows an interesting conceptual evolution: In the course of
the 20th century in Old German studies it has narrowed its meaning and become equivalent to Stammbildungsvokal 'stem-forming vowel' or Formans (cf. Lindner 2012: 140143).
1.2. The ambiguity of Wortbildung 'word-formation'

In the second half of the 19th century there was also some confusion concerning the
notion Wortbildung 'word-formation ' itself. The early Neogrammarians used Wortbildung in the sense of 'inflection ', while what is called Wortbildung today went under the
term Stammbildung 'stem-formation':
Es zerfii llt demnach d ie wortbildungslehre (formen- , flexionslehre) in die lehre von der
bildung der nomina und in die lehre von der bildung der verba. Jenen liegen nominal-, diesen
verbalstamme zu grunde. Die lehre von der bildung der nom ina nennt man declination,
die lehre von der bildung der verba conjugation. (Miklosich 1876: l ; vgl. Miklosich 1875)
[Word- formation (the study of forms, of inflection) therefore comprises the formation of
nouns and verbs. The fo rmer are derived from noun stems, the latter from verb stems. The
study of the formation of nouns is called declension, the study of the formation of verbs
conjugation.]
Die Suffixe theilt man ein in Wortbildungssuffixe oder Flexionssuffixe im engern Sinne,
wozu einerseits die Casusendungen [ ... ], anderse its die Personalendungen [ ... ] gehoren, und
Stammbildungssuffixe [ ... ]. Eine scharfe G renze zwischen beiden Suffixgattungen ist nicht
zu ziehen, da manches E lement, das urspri.inglich nur ableitend (stammbildend) war, mit
den wortbildenden Suffixen [= Flexionsendungen, T h. L.] auf g leiche Linie gekommen ist.
(Brugmann 1886: 15)
[Suffixes are divided into word-formational or inflectional suffixes in the narrow sense,
which on the one hand comprise case endings [ ... ] and on the other personal endings, and
stem-forming suffixes [ ... ] . It is impossible to draw a sharp line between the two kinds of
suffixes, because some elements that originally were derivational (stem-forming) have
turned into word-forming suffixes [i.e. inflectio nal endings; Th. L.].]
This interpretation as well as the arrangement of word-formation before inflection in

grammars goes back to Schleicher 's Compendium, who subsumed both "roots and stems"
(stem-formatio n) as well as word-formation (inflection) under the headi ng Morphologie
'morphology' (cf. Lindner 20 12: 94). Brugmann in turn, in the first ed it ion of his GrundrijJ (1892), subsumed both stem-formation and inflection under word-formation. This
arrangement met with criticism on the part of Hennann Paul (1896: 692), which is why
it was later on abandoned again (also by Brugmann in his final edition of the Grundrifi
in 1906). At the latest in the first decades of the 201h century the trad itional terminology
had again become standard, cf. Debrunner 's preface to his Griechische Wortbildungslehre:
Noch e in Wort iiber die Begrenzung des Stoffs: Das Biichlein soil die Wortbildung, nicht
d ie Formenb ildung behandeln. Ausgeschlossen ist also die gesamte Deklination und Konjugation [ ... ]. (Debrunner 1917: IX)
PA41
[Just one more word concerning the demarcation of the subject: the book is intended to treat
word-formation, not form-formation. All declensions and conjugations are therefore excluded[ ... ).]
The very fact that Debrunner felt obliged to make such a remark, however, shows that
the term Wortbildung could still provoke uncertainties, otherwise he would not have
made it. We have come full circle, returning to the usage of early historical-comparative
grammar (and earlier traditions), when word-formation was used in the sense it has today
and was located after inflection (cf. Lindner 20 12: 94-95, with references).
2. Word-formation immediately before the ri se

of the historical-comparative method
At the beginning of the 19rh century important insights had been gained with respect to
the approaches of the ancient grammarians which had been in use in grammars and
textbooks until around 1800. Before Buttmann, nobody talked of word-internal structure,
only of "endings" (Endungen, Endigungen). Especially first elements of compounds
were still treated in the traditional way by regarding them as case-forms or mutilations
thereof; Christian August Lobeck, for example, continued this tradition of analysis well
into the 19th century (cf. Lindner 2012: 125). The systematization of derivational, especially suffixal, word-formation and composition was still in its infancy. Until the l 81"
century the poorly developed word-formational analysis of the ancient grammarians was
treated in the etymological parts of grammars, i.e. those dealing with words and parts of
speech, under the headings of species andfigura (etymologia had the meaning of 'word
analysis', cf. Lindner 201 1: 29 and article 2 on word-formation research from its beginnings to the 19th century). Only towards the end of the 181h century could a new orientation be observed that manifested itself in a separate treatment of word-fonnation, especially composition, by Adelung in his Umstandliches Lehrgebaude (1782a: 216-236,
1782b: 209- 274). In classical grammar that was considered as a revolution, which 1s
why Trendelenburg could write in the second edition of his Griechische Grammatik:
Ueberhaupt, hoffe ich, wird man nicht leicht einen Theil der Sprachlehre an vorhin unbekannten Bemerkungen ganz leer finden [ ... ] Als Beispiel nenne ich nur das eilfte Kapitel
[ ... ], in welchem ich einen Versuch gemacht babe, die Wortbildung genauer, wie bisher,
auseinander z u setzen. Wenigstens glaube ich mir den zweiten Abschnitt [ ... ] von der Zusammensetzung der W orter mit Recht zueignen zu konnen. Denn ich erinnere mi ch nicht,
in irgend einem grammatischen Werke iiber diesen Theil der Worter, an welchen das Griechische so auf3erordentlich reich ist, etwas gelesen zu haben, was die Grundsiitze und die
Analogie, welche die Sprache in Zusammensetzung der Worter befolgt, our einigermaf3en
auseinandersetzte. Alie Sprachlehrer haben sich blof3 auf den mechanischen Theil, auf d ie
Art und Weise, wie zwei Worter an einander gefiigt werden, eingeschrankt [ ... ]. Ich erwehne
dieses Versuchs, besonders deswegen, um Freunde von dergleichen Untersuchungen zu bitten, so wot die Grundsiitze, welche ich hier angenommen habe, zu priifen und, woes nothig
ist, zu berichtigen, als auch gelegentl ich das Ihrige dazu beizutragen, diese Lehre mit neuen
Bemerkungen zu bereichern, welche sich in diesem Theil der Sprache besonders bei der
Vergleichung des Griechischen mit dem Deutschen darbieten werden . (Trendelenburg 1790:
VI f.)
41
PA42
42

[I hope that one will hardly find any part of this grammar devoid of hitherto unknown
observations [... ] I would just like to mention Chapter XI [ ... ], where I have tried to deal
with word-formation in more detail than is usual. I think that at least the second section
[ ... ] about composition is mostly original work. I do not remember having read in any
grammatical treatise about this kind of words which abound in Greek anything that would
explain the principles and the analogy which the language follows in compounding. All
grammarians have limited themselves to the purely mechanical part of the question, the way
in which two words are put together [... ]. I mention this essay particularly because I would
like to invite friends of this kind of research to assess and, where necessary, correct the
principles I have assumed and to enrich this ana lysis with new observations that one will
not fail to make in this area of language especially by comparing Greek and German.]
One should also point out the progress made in the treatment of Greek word-formation
by Buttmann, from the short chapter of the first, second and third editions (1805: 266268, Von der Zusammensetzung) of his Griechische Grammatik to the extensive chapter
in the fourth and fifth editions (1808, 1810: 399-419, Wortbildung <lurch Endungen,
420-428, Wortbildung <lurch Zusammensetzung):
Der bedeutendste gro/3ere Zusatz [ ... ] ist der Abschnitt von der Wortbildung, der mich zwar
noch keineswegs befriedigt, von welchem ich aber <loch hoffe, da/3 er auch so schon seinen
Zweck in der Hauptsache erreichen wird. (Buttmann 18 10: X)
[The most important substantial addition [... ] is the section on word-formation, which is
still far from satisfying me completely, but which hopefully will also attain its main aims
as it stands.]
Die Wortbildung im vollen Verstande des Wortes liegt au/3erhalb der Grenzen der gewohnlichen Sprachlehre. Denn da die Analogien in dem alteren Theile des Wortvorrathes [ ... ]
vielfaltig zerrissen und verdunkelt sind [ ... ], [wird daher] eine gewisse Masse von Wortern
lexikalisch voraus[gesetzt] [ ... ]. Gewisse Arten der Ableitung jedoch, von welchen man
eben deswegen annehmen kann, da/3 sie neuer sind, haben sich so vollstandig und innerhalb
gewisser Grenzen durchgehend erhalten, da/3 sie mit Sicherheit zusammen gestellt werden
konnen; und diese Vereinigung derselben unter einem Gesichtspunkt erleichtert und beschleunigt die Kenntnis der Sprache [ .. . ]. (Buttmann 1810: 399)
[Word-formation in the full sense of the word lies outside ordinary grammar. Since in the
older p art of vocabulary the analogies [ ... ] have often been broken and are now opaque
[ ... ], a certain number of words must be considered as lexically primitive [ ... ]. However,
some kinds of derivation, which for that very reason may be considered as younger, have
been preserved so fully and completely within certain limits, that they can be put together
without question; and assembling them from one specific v iewpoint makes knowledge of
the language easier and more rapid [ ... ] .]
Formal analogies were also decisive for early comparative grammar which built on these
forerunners. At the beginning, derivational word-formation was formally oriented exclusively according to the "endings", i.e. stem-forming suffixes:
Bei der Anhangung der Endungen [sci/. Wortbildungssuffixe] walteten zwei Prinzipe vor,
das Bestreben gleichartige Bedeutungen <lurch einerlei Endung auszudriicken, und das Bestreben, die Endung der Form des Stammworts moglichst anzupassen. Allein <lurch die
Koll is ion dieser Prinzipe entstand zweierlei Verwirrung der Analogie: I) ist diesel be Art der
Bedeutung haufig unter verschiedene Formen vertheilt; 2) Endungen, die urspriingl ich nur
PA43
von gewissen Formen des Stammworts gebildet wurden [ ... ], gingen, wenn eine gewisse
Bedeutung bei melu-eren W ortern gleichen Ausgangs fiihlbar geworden war, auch auf andere
Stammworter iiber, deren Form nicht dazu pal3te [ ... ]. (Buttmann 1810: 400)
[Two principles preside over the attachment of endings [i.e. derivational suffixes], the endeavor to express the same meaning with the same ending, and the endeavor to adapt the
ending to the form of the stem as far as possible. Through the collision of these two principles analogy became confused in two ways: 1) often the same meaning is distributed among
different forms; 2) endings that had originally been attached only to stems of a certain form
came to be attached to other stems whose form was inappropriate as soon as more words
with the same ending were felt to form a semantically coherent group[ ... ].]
It still took quite some time before word-formation would also be analysed according to
"sameness of meaning", i.e. according to semantic derivational categories. At the beginning, all treatments of word-formation were organized according to forman parameters
concerning the suffixes.
Shortly before the publication of Grimm's monumental German word-formation
(1826), Becker published his Deutsche Wortbildung ( 1824). In this first monographic
treatment of German derivation and composition Becker analyzed the "formation of
verbs" (71-261 ), "derivation by suffixes" (262-368) and "composition" (369-451 ).
Concerning derivation, Becker speaks of Kernformen ('nuclear forms', i.e. bases) and
SprojJformen ('offshoot forms', i. e. derivatives); the process of derivation is called
Umendung 'change of ending'. Ableitung 'derivation' for him is the hyperonym, which
leads him to distinguish derivation by change of ending from derivation by composition.
In the realm of composition he builds on Adelung's criteria (1782b: 215 ff.) distinguishing Verschmelzung 'fusion', the amalgamation of two words into a conceptual unit, and
Zusammenfugung 'putting together', the syntactic union of two words. These di stinctions
can also be found in hi s Organism (1827) and in his Deutsche Grammatik (1829), placed
in the context of a wider grammatical system.
3.
H istorica I-comparative descriptions
3.1. 19th century

Jacob Grimm (1826) represents the historical-comparative turn in German word-formation: "The scientific study of word-formation is [ ... ] a creation of J. Grimm" (cf. Paul
1896: 692). He takes into account older stages of the language and provides abundant
comparisons with other Germanic and non-Germanic material. First, he distinguishes
derivation in the more restricted sense and composition:
ableitung hei/3t die zwischen wurzel und flexion eingeschaltete, an sich selbst dun.Ide mehrung des wortes, kraft welcher der begriff der wurzel weiter geleitet und bestimmt wird.
[... ] die able itung unterscheidet sich von der zusammensetzung [ ... ] letztere verbindet zwei
lebendige oder doch deutliche wurzeln miteinander [ ... ]. (Grimm 1826: 89- 90)
[Derivation :refers to the in itself opaque material that comes between the root and the
inflection and which develops and determines the concept expressed by the root. [ ... ] Derivation must be distinguished from composition [ ... ] the latter combines two roots that are
still distinctly fe lt to be such[ ... ].]
43
PA44
44
Grimm then goes on to differentiate formal patterns of "vocalic" and "consonantic"

derivations (92-386), the semantics of which is briefly touched upon (395-398). The
latter part is constituted by composition (405-985). Here Grimm introduces the concept
of Compositionsvocal 'composition vowel', which he did not identify as a stem vowel
and which would live on for a long time in German school grammars. Only around 1840
does Grimm abandon this doctrine after harsh criticism from comparatists (cf. Lindner
20 12: 117). His distinction of eigentliche (echte) and uneigentliche (unechte) Komposition (407f.; cf. Lindner 2011: 15-20), compounding proper and univerbation, on the
other hand, has stood the test of time.
Franz Bopp treats Old-Indian word-formation for the first time in his Ausfiihrliches
Lehrgebaude, where he introduces the Indian terms for types of compounds which are
used up to the present day (1 827: 268ff., 310ff.; cf. section 5). In a comparative context
he does the same in the fifth volume of his Vergleichende Grammatik ( 1833-52: 1072 ff.,
1410 ff.), classifying derivation according to the form of suffixes and composition according to the Indian categories. Already before Bopp, August Friedrich Pott in his
Etymologische Forschungen (1836: 351 ff.) had treated word-formation from a comparative perspective, but in a rather casuistic and unsystematic manner.
In the meantime Friedrich Diez had published a treatise of Romance word-formation
in the second volume of his Romanische Grammatik (1838: 219ff.). The treatment of
derivation follows the formal criterion of the form of suffixes (244 ff.), that of composition is by parts of speech following the example of Grimm (334 ff.). A book that was to
become important for the analysis of word-formation, especially of Ancient Greek, by
the early Neogrammarians was Curtius' De nominum Graecorum formatione linguarum
cognatarum ratione habita (1842), dedicated to his teacher Bopp.
In Neogrammarian times the theoretical manifesto, also for word-formation, was Hermann Paul's Prinzipien (1880: 161ff., 5 1920: 325 ff.). In turn, the model up to the present
day for all descriptively oriented treatises of the word-formation of individual lndoEuropean languages is constituted by the monumental volume on word-formation in
Brugmann's GrundrifJ. The second edition of 1906 set the standard and continues to be
indispensable due to the wealth of its comparative material and its comprehensive description. The introduction discussing the structure of Indo-European word-forms as well
as the motives and kinds of word-formation processes (1 - 49, Allgemeines) is followed
by a description of composition (49- 120) and thereupon the listing of Stammbildungsformantien based on formal criteria (120-582; the relevant literature on word-formation
is summarized on pp. 49 ff. and 120 ff.). In the concluding section the material is arranged according to semantic groups (582-685, Bedeutung der Nominalstamme). What
is worth highlighting is Brugmann's rejection of the affix-terminology and his introduction of the term Formans 'formative' to refer to a derivational affix (]906: 8ff.).
3.2. 20th century

The reference works on the word-formation of old Indo-European languages published
during the 201" century can only be enumerated here (further references can be found in
the bibliographies of the works cited as well as in Heidermanns 2005, Luhr and Balles
2008 and Liihr and Matzinger 2008): from a comparative perspective Brugmann (1906)
PA45
(esp. Delbruck 1900: 139ff. and Brugmann 1904: 297ff., 1906: 49ff.); for Old Indian:
Wackemagel (1905) (Wackemagel and Debrunner 1957), Debrunner (1954); for Avestan:
Duchesne-Guillemin (1936), Kellens (1974); for Greek: Debrunner (1917), Chantraine
(1933), Schwyzer (1939) as well as Risch (1974) for the language of Homer, for Latin:
Leumann (1977), Kircher-Durand (2002) and Lindner (1996, 2002a), for Germanic: Paul
(1920), Meid (1967) and Carr (1939). I would also like to draw attention to the part
dedicated to word-formation in the on-going project of the Indogermanische Grammatik
published by Winter in Heidelberg: Lindner (2011 ff.) on compounding, Sadovski
(forthc.) on derivation, both with exhaustive bibliographies.
4. Descriptions based on semantic criteria

The first scholar to take a semantic category, viz. verbal abstracts, as point of departure
was Karl von Bahder (1880), subordinating the formal side to the semantic perspective:
Ich hoffe, dass es mir gelungen ist, einige neue gesichtspuncte fur die wortbi ldung aufzudecken [ ... ] Abschliessende resultate wird kein billig denkender van einem versuche fordern, der als der erste einer zusammenhangenden betrachtung einer wortkategorie [ ... ] wol
bezeichnet werden darf. (van Bahder 1880: If.)
[I hope to have been able to discover some new viewpoints for word-formation [... ] It
would be unfair to expect definitive results from an essay that may probably be called the
first one to treat a word-category from a coherent perspective.]
In his Nominate Stammbildungslehre (1886) Friedrich Kluge followed this model for all
categories. He was the first to abandon the prevailing fonnal approach in a comprehensive treatise of word-formation and to proceed according to the meanings of the formatives; those belonging together from a semantic point of view were treated! together in
special chapters (suffixes for personal nouns, diminutive suffixes, collective suffixes,
suffixes for inanimate concrete nouns, abstract nouns, etc.). Cf.:
Freil ich schlie/3e ich mich in der gruppierung des stoffes nicht an linguistische vorbilder an;
ich habe nicht die lautform, wie es bisher iiblich war, zum ausgangspunkte fiir die anordnung
gemacht. (Kluge 1886: VIII)
[In my arrangement of the material I do not fol low existing models; I did not take as the
point of departure the phonic form, as has been done up to now.]
This way of proceeding was to become widely accepted. Wilhelm Meyer-Lubke, for
example, followed Kluge 's arrangement in his Italienische Grammatik (1890):
Wenn so van Seiten einer streng historischen Grammatik die italienischen Fachgenossen,
die an der Quelle s itzen, mein Buch werden vielfach erganzen konnen, so ist das in noch
hoherm Grade der Fall bei der Wortbildungslehre. Die Darstellung derselben, wie ich sie im
III. Kapitel gegeben habe, hat die Unkommlichkeit, <lass sie einmal viele Suffixe nicht
bespricht, und sodann, dass manche Erscheinungen, wie die Verkniipfung verschiedener
Suffixe, die Priipositionalbildungen, die Suffixvertauschungen nicht zur Sprache kommen.
Allein ich habe nicht den historischen Entwicklungsgang geben wollen, m ir lag hauptsiichlich daran, eine Darstellung zu bieten, wie sie meines Wissens noch fiir keine romanische
45
PA46
46

Sprache geboten ist. Die begriftliche Seite ist fur diesen Theil der sprachlichen Biologie
ebenso wichtig wie die formale, und um dies hervorzuheben, habe ich die letztere fast ganz
beiseite gelassen, um so eher, als bei der einseitig aufs Formale gerichteten Aufinerksamkeit
der romanischen Forschung eine Erganzung nach dieser Seite hin nicht schwer fiillt. (MeyerLiibke 1890: ix-x)
[Colleagues working on Italian from a strictly historical perspective on the basis of more
abundant material will certainly be able to complete the present book, especially regarding
word-formation. My description in chapter III has the disadvantage of not treating many
suffixes and that some phenomena such as the combination of different suffixes, prepositional formations, or suffix exchange are not mentioned. However, my intention was not to
present the historical development, I wanted to provide a description which, to the best of
my knowledge, does not yet exist for any Romance language. The conceptual side is as
important for this part of linguistic biology as the formal side. In order to highlight this fact,
I have almost completely left aside the latter, all the more so as the one-sided attention of
research in Romance on the formal side easily allows completing it in that respect.]
Brugmann's Grundrij3 (1892, 1906) eventually combined both a formal and functional
perspective. Furthermore, Paul (1896) is an important contribution to the semantic aspects of word-formation categories and their diachronic developments.
5. Compounding in compa rative grammar

The formally-oriented doctrine of compounding of the ancient grammarians (cf. Lindner
2002a: 181) remained unchallenged unti l the 18th century. But when Sanskrit appeared
on the scene of European scholars in the last decades of the 18th century, the Western
world became acquainted with the approach and terminology of Indian grammarians,
who had been led early on by the high productivity of compounding in Sanskrit to
develop a syntactico-semantic, functional typology of compounds which mutatis mutandis is still in use today (cf. Lindner 2011: 20- 21 ).
The first interface between the Indian tradition and its transmission in Europe can be
found in the first Sanksrit grammars printed in Europe at the end of the 181h and the
beginning of the 19th century (Lindner 2012: 148). Bopp eventually coined the latinized
versions of the Indian terms that are still in use: 1. Copulative compounds, called dvandva; 2. possessive compounds, called bahuvrfhi ("the compounds of this class are adjectives or common nouns denoting the possessor of what the parts of the compound mean,
so that the notion of possessor always has to be supplemented. That is why I call them
'possessive compounds"'); 3. Determinative compounds, called karmadharaya ("The
last member of this class of compounds is a noun or adjective, which is detennined or
described more precisely by the first member"); 4. Dependency compounds (Abhangigkeits-Composita), called tatpuru$a ("this class comprises compounds whose first member
depends on or is governed by the second one, always realizing some oblique case relation"); 5. Collective compounds, called dvigu; 6. Adverbial compounds, called avyay fbhava (Bopp 1827: 311 ff., 1833-52: 1427ff.).
This classification remains practically unchanged over the next decades, excepting
order and denomination. The insight that the determinative compounds comprised both
karmadharaya and tatpuru$a was formulated somewhat later, by Theodor Benfey in his
Vollstandige Grammatik der Sanskritsprache (1852), who did not follow the six-class-
PA47
system of the Indian grammarian Vopadeva directly (as most Europeans before him), but
another tradition going back directly to Pao.ini, who calls karmadharaya and dvigu subclasses of tatpuru.$a; this system, however, has not always been adopted by later research
on compounding (cf. Benfey 1852: 245 ff.: l51 class: Copulative compounds (dvandva);
2nd class: Determinative compounds (tatpuru$a), subdivided into: l51 species: case-determined compounds (tatpuru$a stricto sensu), 2nd species: apposition-determined compounds (karmadhiiraya), Yd species: numeral-determined compounds (dvigu), which he
eventually sees as a special case of the second species, in which the appositional determination is realized by a numeral; Yd class: relative compounds (bahuvrfhi). Benfey goes
as far as to consider the bahuvrfhi as a special kind of compounding of the second class).
At the beginning of the decade of 1860 an important study was published that would
set standards for the later research on compounds in general and lndo-European linguistics, Ferdinand Justi 's Uber die Zusammensetzung der Nomina in den indogermanischen
Sprachen (1861 ). What is at issue here is genetic and typological comparison based on
material also from non-lndo-European languages, with a wealth of examples that would
form the basis of Neogrammarian research some fiftee n or twenty years afterwards. I
would like to draw attention particularly to the first in-depth analysis of the bahuvrfhi
type, interpreted as "a higher kind of compounding" condensing a relative clause into a
word, a stance that gave rise later on to a controversial discussion about the origin
of possessive compounds leading to different theories (cf. L indner 2002b: 269- 273).
Chronologically the first one was Justi 's relative-clause theory, which was further developed and expanded by Jacobi (1897):
Es g ibt nun e ine art wortzusammensetzung, welche e inen ganzen bezi.iglichen satz zu einem
wort vereinigt, das aber wie der ganze satz ebenfalls bezi.igliche, relative bedeutung hat.
Statt zu sagen Eq>cXVl] 'Hill<; nnvt oi MK'tuAOt fficm: p6oa dcriv, zieht man den ganzen relativsatz zusammen und bringt ihn in numerale, casuelle und geschlechtliche congruenz mit dem
nomen, auf d as er sich bezieht, und sagt also ecpav11 'Hill<; poooMK'tuA.o<;, welches aber genau
aufgelost bedeutet ,Eos, welcher finger wie rosen sind'. - ibid. Der bahuvrlhi (sic!) ist nun
die bildung, in welcher die wortzusammensetzung den g ipfel ihrer vollendung erreicht hat;
sie ist ebenso schon w ie kurz und bi.indig [ . .. ]. (Justi 1861: 117)
[There is one kind of composition that condenses a whole relative clause into one word,
which however also has a relative meaning like the whole clause. Instead of saying ecpavl]
1-Jcb<; TI'tlVt Ot OcXKTUAOt cO<H p6oa d criv, the whole relative clause is pulled together and
made agree with the noun it is predicated of in number, case and gender, saying Eq>cXVl] 'Heb<;
poboOaK'tUAo<;, w hose explicit meaning however is: 'Eos, whose fingers are like roses'. [ ... ]
The bahuvrihi is that kind of formation w here compounding has reached its epitome; it is
as beautiful as it is conc ise[ ... ].]
The question of the nature, history, and origin of stem composition has also become the
subject of controversy (cf. Lindner 201 2: 88-12 1, 2013: 149-154). Furthermore, the
general and ph ilosophical aspects of compound ing first addressed by Justi have been
elaborated in Ludw ig Tobler's book Uber die Wortzusammensetzung nebst einem Anhang
uber die verstarkenden Zusammensetzungen (1868). Moreover, Justi 's book stimulated
further research on Greek and Latin compounds published in shorter essays (cf. L indner
and Oniga 2005; Lindner 2011: 12, 20 12: 131 - 134).
Apart from such sma ller contributions, one has to mention the comprehensive descriptions of syntactic compounds by Francis Meunier (1872) - we usually call them Juxta-
47
PA48
48
posita in Inda-European studies (cf. Lindner 20 11: 18) -, as well as Leopold Schroder's
great monograph Uber die formelle Unterscheidung der Redetheile im Griechischen und
Lateinischen mit besonderer Berucksichtigung der Nominalcomposita (1874). In this first
near-exhaustive description of the dispersed material Schroder rightly criticizes that one
defect of his predecessors was "that they were far from complete in their empirical
coverage, being content with single examples where complete enumerations would be
highly desirable" (cf. Schroder 1874: 194). What is more, he coined the terms "synthetic
compound" (1874: 206; synthetisches Kompositum) as well as composita immutata and
mutata (1874: 208) to replace Justi 's "lower" and "higher" formations. Some fifteen
years later the present term " exocentric compounds" was introduced. It is first attested
in Aleksandrow (1888: 110), going back probably to Baudouin de Courtenay (Lindner
2009: 190-192): exocentric compounds ("compounds whose semantic center is not located in its members") and esocentric compounds ("composition whose semantic center
corresponds to one of the members"), the latter, nowadays called endocentric, further
subdivided into bicentric (= dvandva) and monocentric compounds (with the subtypes
primocentralia and alterocentralia, depending on whether the center coincides with the
first or the second member). On the history of this tenninology cf. Lindner (2009) and
(2011: 27- 28); on reference works of the 20 1h and 2l51 century concerning Inda-European composition as well as further literature see section 3.2 as well as Lindner (2002a:
322-323, 2003: 134), and (2011: 51-53).
6. References
l 782a Umstandliches Lehrgebaude der Deutschen Sprache, zur Erlauterung der Deutschen
Sprachlehre far Schulen. Vol. 1. Leipzig: Breitkopf.
I 782b Umstandliches Lehrgebaude der Deutschen Sprache, zur Erlauterung der Deutschen
Sprachlehre far Schulen. Vol. 2. Leipzig: Breitkopf.
Aleksandrow, Aleksander
1888
Litauische Studien I. Nominalzusammensetzungen. Dorpat: Hermann.
Bahder, Karl von
1880
Die Verbalabstracta in den germanischen Sprachen ihrer Bi/dung nach dargestellt.
Halle/S.: Niemeyer.
1824
Die Deutsche Wortbildung oder die organische Entwickelung der deutschen Sprache in
der Ableitung. Frankfurt/M.: Hermann.
1827
Organism der Sprache als Einleitung zur deutschen Grammatik. (Deutsche Sprachlehre,
Vol. 1.) Frankfurt/M.: Reinherz. [2d ed. 1841].
1829
Deutsche Grammatik (Deutsche Sprachlehre, Vol. 2). Frankfurt/M ..: Hermann/Kettembeil. [2 11d ed. 1836, 3'd ed. 1842].
Benfey, Theodor
1852
Vollstandige Grammatik der Sanskritsprache (Handbuch der Sanskritsprache, Vol. 1).
Leipzig: Brockhaus.
Bopp, Franz
1824- 31
Vergleichende Zergliederung des Sanskrits und der mit ihm verwandten Sprachen. 5
Abhandlungen. ln: Abhandlungen der historisch-philologischen Klasse der Koniglichen
Akademie der Wissenschaften zu Berlin, Jg. 1824- J83 l.
PA49
Bopp, Franz
Ausfahrliches Lehrgebi:iude der Sanskrita-Sprache. Berlin: Dummler.
1827
Bopp, Franz
1833-52
Vergleichende Grammatik des Sanskrit, lend, Griechischen, Lateinischen, Litthauischen, [Altslawischen,} Gothischen und D eutschen . 6 Vol. Berlin: Dummler.
Brugmann, Karl
1886
Grundrij3 der vergleichenden Grammatik der indogermanischen Sprachen:. Vol. 1: Einleitung und Lautlehre. StraJ3burg: Triibner.
Brugmann, Karl
1892
Grundrij3 der vergleichenden Grammatik der indogermanischen Sprachen. Vol. 2: Wortbildungslehre (Stammbildungs- und Flexionslehre). StraJ3burg: Trubner.
Brugmann, Karl
1904
Kurze vergleichende Grammatik der indogermanischen Sprachen. Stral3burg: Triibner.
Brugmann, Karl
1906
Vergleichende Laut-, Stammbildungs- und Flexionslehre nebst Lehre vom Gebrauch der
Wortformen der indogermanischen Sprachen. Vol. 1: Lehre von den Wortformen und
ihrem Gebrauch. Part 1: Allgemeines, Zusammensetzung (Komposita), Nominalstiimme.
2nd ed. StraJ3burg: Trubner.
Buttrnann, Philipp
3
1805, 5 1810
Griechische Grammatik. Berlin: Myl ius .
Carr, Charles T.
Nominal Compounds in Germanic. London: Milford.
1939
Chantraine, Pierre
1933
La formation des noms en grec ancien. Paris: K lincksieck [Reprint 1979].
Curtius, Georg
1842
De nominum Graecorum formatione linguarum cognatarum ratione habita. Berlin:
Dummler.
Curtius, Georg
1852
Griechische Schulgrammatik. Prag: Tempsky.
Debrunner, Albert
19 17
Griechische Wortbildungslehre. Heidelberg: Winter.
Debrunner, Albert [and Jacob Wackernagel]
Altindische Grammatik. Vol. 2,2: Die Nominalsujfvce. Gottingen: Vandenhoeck & Ru1954
precht.
Delbriick, Berthold
1900
Ve1gleichende Syntax der indogermanischen Sprache. Vol. 3. StraJ3burg: Triibner.
Diez, Friedrich
1838
Grammatik der romanischen Sprachen. Vol. 2. Bonn: Weber. [SCh ed. 1882].
Duchesne-Gu illemin, Jacques
1936
Les composes de /'Avesta. L iege/Paris: Droz.
Grimm, Jacob
1819
Deutsche Grammatik. Vol. 1. Gottingen: Dieterich. [2nd ed. 1822].
Grimm, Jacob
1826
Deutsche Grammatik. Vol. 2. Gottingen: Dieterich. [= 3. Buch: Von der Wortbildung].
[2nd ed. 1878].
Heidermanns, Frank
Bibliographie zur indogermanischen Wortforschung. Wortbildung, Ety mologie, Onoma2005
siologie und Lehnwortschichten der a/ten und modern.en indogermanischen Sprachen in
systematischen Publikationen ab 1800. 3 Vo l. Tiibingen: Niemeyer (also available as
CD-ROM).
49
PA50
50

Jacobi, Hermann
1897
Compositum und Nebensatz. Studien iiber die indogermanische Sprachentwicklung.
Bonn: Cohen.
Justi, Ferdinand
1861
Ueber die zusammensetzung der nomina in den indogermanischen sprachen. Gottingen:
Dieterich.
Kellens, Jean
Les noms-racines de /'Avesta. Wiesbaden: Reichert.
1974
Kircher-Durand, Chantal (ed.)
2002
Grammaire Fondamentale du latin. Tome IX: Creation lexicale: la Formation des noms
par derivation suffixale. Louvain: Peeters.
Kluge, Friedrich
1886
Nominate Stammbildungslehre der altgermanischen Dialekte. Halle/S.: Niemeyer. [3'd
ed. 1926].
Leumann, Manu
1977
Lateinische Laut- und Formenlehre. Vol. I of Manu Leumann, Johann Baptist Hofmann
and Anton Szantyr Lateinische Grammatik. Munchen: Beck. [I st ed. 1926-28].
Lindner, Thomas
1996
Lateinische Komposita. Ein Glossm; vornehmlich zum Wortschatz der Dichtersprache.
Innsbruck: lnstitut fiir Sprachwissenschaft.
L indner, Thomas
2002a Lateinische Komposita. Morphologische, historische und lexikalische Studien. Innsbruck: Institut fiir Sprachen und Literaturen.
Lindner, Thomas
2002b Nominalkomposition und Syntax im Indogermanischen. In: Heinrich Hettrich (ed.), lndogermanische Syntax. Fragen und Perspektiven, 263- 279. Wiesbaden: Reichert.
L indner, Thomas
2003
Aspekte der latein isch-romanischen Kompositaforschung. Moderne Sprachen 47: 115141.
L indner, Thomas
2009
A Note on 'endocentric'. Historiographia Linguistica 36(1): 190- 192.
L indner, Thomas
2011
Komposition (Indogermanische Grammatik 411). Lieferung I . Heidelberg: Winter.
L indner, Thomas
2012
Komposition (I ndogermanische Grammatik 411). Lieferung 2. Heidelberg: Winter.
L indner, Thomas
2013
Komposition (lndogermanische Grammatik 4/ 1). Lieferung 3. Heidelberg: Winter.
Lindner, Thomas and Renato Oniga
2005
Zur Forschungsgeschichte der lateinischen Nominalkomposition. In: Gualtiero Calboli
(ed.), Lingua Latina! Proceedings of the Twelfth International Colloquium on Latin
Linguistics, Bologna 2003, 149- 160. Roma: Herder.
Luhr, Rosemarie and Irene Balles
2008
Nominate Wortbildung des lndogermanischen in Grundziigen. Die Wortbildungsmuster
ausgewiihlter indogermanischer Einzelsprachen. Vol. I: Latein, Altgriechisch. Hamburg:
Kovac.
Luhr, Rosemarie and Joachim Matzinger
2008
Nominate Wortbildung des lndogermanischen in Grundziigen. Die Wortbildungsmuster
ausgewiihlter indogermanischer Einzelsprachen. Vol. 2: Hethitisch, Altindisch, Altarmenisch. Hamburg: K ovac.
Meid, Wolfgang
1967
Wortbildungslehre. Vol. 3 of Hans Krahe and Wolfgang Meid Germanische Sprachwissenschaft. Berlin: de Gruyter.
PA51
Meun ier, L.-Francis
Les composes syntactiques en grec, en latin, en franr;ais et subsidiairement en zend et
1872
en indien. Paris: Durand et Pedone-Laurie!.
Meyer-Lubke, Wilhelm
1890
ltalienische Grammatik. Leipzig: Reisland.
Miklosich, Franz
1875
Vergleichende Grammatik der slavischen Sprachen. Vol. 2: Stammbildungslehre. Wien:
Braumuller.
Miklosich, Franz
1876
Vergleichende Grammatik der slavischen Sprachen. Vol. 3: Wortbildungslehre. 2nd ed.
Wien: Braumuller (1 st ed. 1856 as Vol. 3: Formenlehre].
Paul, Hermann
1880
Prinzipien der Sprachgeschichte. Halle/S.: Niemeyer. [5 111 ed. 1920].
Paul, Hermann
1896
Ober die Aufgaben der Wortbildungslehre. In: Sitzungsberichte der philosophisch-philologischen und historischen Classe der k. b. Akademie der Wissenschaften zu Munchen.
Jahrgang 1896, Heft 4, 692-713. Munchen: Verlag der K. Akademie. [Reprinted in:
Leonhard Lipka and Hartmut Gunther (eds.) 1981: Wortbildung, 17- 35. Darmstadt: Wissenscha ftl iche Buchgesellschaft].
Paul, Hermann
1920
Deutsche Grammatik. Vol. 5: Wortbildungslehre. Tubingen: Niemeyer.
Pott, August Friedrich
Etymologische Forschungen auf dem Gebiete der /ndo-Germanischen Sprachen, insbe1836
sondere des Sanskrit, Griechischen, Lateinischen, Littauischen und Gothischen. Vol. 2:
Grammatischer Lautwechsel und Wortbildung. Lemgo: Meyer.
Risch, Ernst
Wortbildung der homerischen Sprache. 2nd ed. Berlin/New York: de Gruyter.
1974
Sadovski , Velizar
forthc.
Derivation (lndogermanische Grammatik 4/2). Heidelberg: Winter.
Schroder, Leopold [von]
Uber die formelle Unterscheidung der Redetheile im Griechischen und Lateinischen mit
1874
besonderer Berucksichtigung der Nominalcomposita. Leipzig: Kohler.
Schwyzer, Eduard
1939
Griechische Grammatik. Vol. l: A llgemeiner Tei!, Lautlehre, Wortbildung, Flexion.
Munchen: Beck.
Tobler, Ludwig
1868
Uber die Wortzusammensetzung nebst einem Anhang iiber die verstarkenden Zusammensetzungen. Ein Beitrag zur philosophischen und vergleichenden Sprachwissenschaft.
Berlin: Dummler.
Trendelenburg, Johann Georg
1790
Anfangsgriinde der griechischen Sprache. 3rd ed. Leipzig: Barth. [JS' ed. H82].
Wacke rnagel, Jakob
1905
Altindische Grammatik. Vol. 2, I: Einleitung zur Wortlehre. Nominalkomposition. Gottingen: Vandenhoeck & Ruprecht [New impression 2 1957 with additions by Albert Debrunner].
Wackernagel and Debrunner
1954
see Debrunner 1954.
Wackernagel and Debrunner
1957
see Wackernagel 1905.
Thomas Lindner, Salzburg (Austria)
51
PA52
52
4. Word-formation in structuralism
I.
2.
3.
4.
5.
Overview
Ferdinand de Saussure
Directions in structuralism
Word-formation in structuralist schools
References
Abstract
The subsumption under the designation structuralism of directions in linguistic research
from the first half of the 2ot1 century which were very different in detail refers to efforts
to found a linguistic theory which meets the demands of mathematical requirements. The
guiding impulses for this development stem from F de Saussure with his establishment
of the language system as an autonomous subject of linguistics which is independent of
and a precondition to all other aspects of the study of language. The language system
was seen as a hierarchy of levels, the units of which are connected to one another
through specific relations. The scientific ideal demanded a strict formal, inherent differentiation of levels, units and classes of the language system. The analysis was applied
primarily to phonemes and morphemes. The suggested theoretical approaches established guiding foundations for phonology, morphology, and for syntactic dependency or
constituent structure grammars. Word-formation played a rather marginal role. This can
be explained by the strict methodological postulates.
1
1. Overview
At the beginning of an, in a narrow sense, scientific study of natural languages stood
language comparison and language history. The systematic comparison of languages lead
to insights regarding the family relationships among natural languages. There was also
a central interest in cultural similarities and differences. The investigation of the history
of individual languages was also originally motivated by ethnological interest. With the
neogrammar ians, this phase of linguistics reached its zenith. Both language comparison
as well as the study of the history of individual languages were, however, only based on
individual linguistic forms, in particular in the context of the word. The characteristics
which constitute the essence of natural language were not systematically investigated.
The historical-comparative linguistics of the 19th century could carry out its program
with a rather undifferentiated concept of grammar. Linguistic research was oriented towards models which had been developed for classical languages.
As a fundamental criticism of this one-sided direction of research, the idea emerged
at the beginning of the 20th century that language comparison and language history are
only possible on a scientific basis when a theory is proposed which describes the actual
core of natural language. Since human languages are by their nature sign systems, i.e.
systems which link sounds with their meanings and thus make possible communication
PA53
between people, a linguistic theory must describe fundamental characteristics of linguistic sign systems. All other phenomena linked with language presuppose such a theory.
The results of preceding linguistic research were, thus, not rejected on principle, but
merely seen as provisional and inadequate.
The goal of the directions in linguistics collectively designated as structuralism was
to lift linguistic research to the level of natural sciences such as physics, chemistry and
biology. The requirements for theoretical languages for these precise empirical sciences
were also discussed in contemporaneous language philosophy. The Vienna Circle, in
particular Rudolf Carnap, Hans Reichenbach, and Herbert Feigl, played a decisive role
in the formulation of a program which rationally reconstructed the theories of empirical
sciences with the help of mathematical logic. The model was Gottlob Frege's attempt to
reconstruct mathematics within the framework of logic. Logical empiricism demanded
that all meaningful statements be either directly reduced to observation sentences, or that
they are at least able to be brought into a logical relation to observation sentences, so
that they can be verified and confirmed by accepted observation sentences.
Structuralism is the designation for a phase in the development of linguistics which
is quite varied in its details. The common features which led to this designation are,
above all, of methodological nature. Similar stages of development can also be found in
the history of the humanities, in psychology, economics, and in other social sciences. The
debates on methodology in linguistics had a strong influence on anthropology, ethnology,
psychology and literary studies, cf. Levi-Strauss (1958).
The main motivator behind the new direction in linguistics was the Genevese linguist
Ferdinand de Saussure (1857-19 13). His posthumously published work Cours de linguistique generate decisively influenced the development of all structuralist lines of
research. The questions he raised led directly to those of the European schools which
have gone down in history as the Geneva school, the Copenhagen school and the Prague
school. But also the directions in American linguistics which fall under the designation
of structuralism were decisively characterized by Saussure's ideas and formulations of
the problem.
The term structure is only rarely used by Saussure. His central term is system. Especially in the Prague school and in American linguistics, the term structure is found more
often. Structuralism can be seen as the true onset of scientifically founded investigation
of human language if one considers the introduction of strict, i.e. verifiable, methodological postulates as characteristic of scientific activity. In any case, the linguistic research
of the following period would be inconceivable without structuralist approaches. The
current widely accepted organization of possible topics in the scientific study of human
languages was already outlined by Saussure. Even the methodological foundations of
linguistic research developed by Noam Chomsky can be seen as an extension and a more
precise rendering of Saussure's ideas, cf. Motsch (2006).
2. Ferdinand de Sa ussure
Ferdinand de Saussure initially started his career with historical linguistic research in
neogrammarian circles, but then sharply criticized the research of his time. He proposed
the outline of a theory of language which differed substantially in its fundamental as-
53
PA54
54
sumptions from historical-comparative linguistics. In this theory, he demands a fundamental departure from the atomistic study of language details and emphasizes the importance of an examination of the relationships into which the individual units enter with
one another within the framework of a system. Only in this way can one arrive at a
complex system which consists of units of different levels and their mutual relationships,
strictly speaking, a system of subsystems. Language is, according to Saussure, by its
very nature a system of values, of units of different levels which mutually condition
each other and are only determined through their position in the system. This basic idea
was taken up and developed further in, to some extent, very different variations in the
linguistic research of the following period.
Beginning with a basic model of linguistic signs, Saussure attempts to differentiate
various levels of observation which detennine separate objects of research. Central to
this endeavor is the distinction between langue and parole. The langue corresponds to a
sign system which is used by the speakers or writers of a social community to transmit
messages to members of this community. The processes which take place during the
transmission and the reception of messages compose the parole. They are individual in
nature in contrast to the social character of the langue. The langue is the sign system
that determines a particular individual language. A further generalization comprises the
assumption of a sign system which underl ies all human languages. For this, Saussure
introduces the term langage, the human language faculty. Although the langue has social
and psychological facets - it must be accepted by a language community and it is situated
in the human brain - according to Saussure it can and must be described completely
independently of these perspectives. Even meanings may only be included from the
formal point of view of the system. The thoughts or ideas which stand behind the meanings, the substance of the content, form an amorphous mass which does not belong to
the subject matter of the description of the system of a language. This means: the essential properties of the sign system which underlies a language are independent both of
analyses of the substance of the content, the psychological existence and the social use
of this system, as well as of the processes of the formation and comprehension of utterances. One can of course investigate the relationships between the sign system of a language and questions of language philosophy or its existence in the human brain, just as
one can, e.g., investigate the influence of social factors on the structuring of a language
into a standard language and regional and social variants, or on the vocabulary of a
language. But in any case, such studies presume a description of the sign system which
is completely independent of these kinds of questions. Thus an investigation of the
processes of the formation and understanding of utterances, i.e. the parole, demands a
knowledge of the system of a language and additionally the inclusion of psychological
questions and theories as well as the consideration of the given context.
According to Saussure, if one sees the sign system as a fundamental prerequisite for
communication among the members of a speech community, a separation of the description of this system from an investigation of the history of this system follows automatically. Linguistic signs are valeurs, values in a system in which they coexist w ith other
signs. The coexistence is only possible on the axis of simultaneity (axe des simultaneites). Change in the sign system is irrelevant for the user of the sign system. It arises
only from the perspective of the linguist. Saussure views every natural language as very
much simultaneously a contemporary institution and a product of the past (Saussure
1916: 24). But studying the relationships is on ly possible if the inner organism of the
PA55
language is initially described without reference to society and history (Saussure 1916:
40, 124). These considerations led Saussure to the dichotomy of synchrony and diachrony.
The observations thus outlined strengthen Saussure's assumption that the langue must
be the cenhal subject matter of linguistics. It is the only autonomous topic within the
scope of the overall phenomenon of language which is not investigated by any other
discipline and whjch is always assumed in the investigation of other aspects of language.
This train of thought decisively influenced the further development of linguistics. It
forms the methodological basis for almost all directions in modern linguistic research.
Saussure provided stimulating guidelines with respect to the internal structure of linguistic sign systems. Especially his distinction between syntagmatic and paradigmatic
relations was taken up and expanded upon by other structuralist schools. Saussure differentiates between relations into which units of the system enter in praesentia, i.e. with
other linguistic units at the same level, e.g., in syntactic chains, as well as associative
relations (Saussure 1916: 170); these are units which can occupy the same position in a
syntagm. This differentiation forms the basis for the assumption of syntactic categories
and levels of sentence structure. Syntagm and paradigm are complementary concepts.
The language system thus consists of sets of units (phonemes, words, word groups) and
a small quantity of very general patterns in which phonemes and words can be inserted
under certain conditions. Every linguistic unit can thus also be characterized through the
relationships into wh ich it can enter with other units at the same level, cf. Wells ( 194 7)
and Jager (2006).
3. Directi ons in stru cturalism

The Geneva school:
Immediately following Saussure are his students and colleagues Charles Bally, Albert
Sechehaye, Henri Frei, Sergej Karcevski and Robert Godel. Their work was discussed
in the Genevan Cercle F. de Saussure and was published primarily in the series Cahiers
F de Saussure. In these papers, Saussure's thoughts and arguments are expanded upon
and further developed, and misunderstandings are cleared up.
Special attention is due Lucien Tesniere, also considered part of the Geneva school,
who in his book "Elements de syntaxe structurale" (Tesniere 1959), sketched in 1934
and only released posthumously in 1959, published a model of syntactic analysis which
was later extended as dependency grammar.
An overview of Tesniere and the further development of his ideas can be found in
Godel (1961 ), Kunze (1975), Baumgartner (1976), Baum (1976), Heringer, Strecker and
Wimmer (1980) and Hudson (1980).
The Copenhagen school:
Head of the Copenhagen school was Louis Hjelmslev, who in 1943 published a programmatic work in Danish Omkring sprogteoriens grundlceggelse. An English translation
appeared in 1961 under the title Prolegomena to a Theory of Language. To the Copenhagen school belong further Hans forg U ldall, Viggo Brnndal, Berta Siertsema, Henning
Spang-Hanssen, Knut Togeby, and Paul Diderichsen.
55
PA56
56
This approach refers to itself as glossematics. The goal of this approach is to create
an immanent algebra of language, a general calculus for the description of texts in a
natural language. Hjelmslev initially only proposes prolegomena of such a linguistic
theory, i.e. methodological principles and a few basic assumptions about fundamental
properties of the system which underlies natural languages. He emphasizes that a linguistic theory cannot be a sum of hypotheses; it must rather be a system of premises and
definitions which is arbitrary, yet suitable for the description of generally accepted and
empirically verified data. When establishing the calculus, direct reference to either phonetic or ontological concepts is not allowed.
To the premises which determine the central properties of the theory belong the distinction between expression and content as well as between form and substance. This
corresponds to distinctions which had already been made by Saussure: signifie [signified]
and signijiant [signifier] as well asforme [form] and substance [substance]. The calculus
must take into consideration the character of natural languages as sign systems as well
as the distinction between syntagmatic and paradigmatic relations. It must also permit
hierarchical relations.
The glossematic algebra is, as are calculuses of formal logic, a sys.tern of dependencies (functions) between terms, which are only determined through the ir mutual dependencies. Units must be the result of an analytic procedure, a deduction. Hjelmslev thus
does not accept the discovery procedures which were later suggested in American approaches.
Hjelmslev's analysis does not lead to a separation of syntax and morphology. To a
degree, the distinction between paradigmatic and syntagmatic relations corresponds to
these components in other linguistic theories. Most applications of glossematics relate
to areas which belong to phonology or morphology in other theories. Products of wordformation are only used as examples for general morphological questions. Glossematic
grammar corresponds rather more to a dependency grammar than to a constituent structure grammar.
An overview of glossematics is provided by i.a. Martinet (1946), Siertsema (1955),
and Spang-Hanssen (1961). A comparison of glossematics with American linguistic
theories can be found in Haugen (1951) and Garvin (1954).
The Prague school:
In 1926, Vilem Mathesius, Bohuslav Havranek, Jan Mukafovsky, Bohumil Trnka and
Josef Vachek founded the Cercle linguistique de Prague. At the first International Linguistic Congress, which took place in 1928 in The Hague, Roman Jakobson, Sergej
Karcevski and Nikolai S. Trubetzkoy presented theses in the name of this circle, cf.
Jakobson, Karcevski and Trubetzkoy (1929). In these theses, theoretical positions on
phonology, morphology including word-formation, and on the functional view of linguistic phenomena are presented. In these theses as well, the basis is Saussure's requirement
of investigating language as an autonomous system, i.e. independently of philosophy,
psychology, acoustics, sociology and other disciplines. But members of the Prague
school also concern themselves with topics that go beyond the language system. In
contrast to other structuralist schools, the Prague school stipulates a .functional approach
to the language system, taking the path from function to fonn, cf. Mathesius (1929).
Typical for the Prague school is a broadly based interest in language phenomena.
Besides phonology, morphology and syntax, i.e. the areas which belong to the language
PA57
system, problems of literary texts, text analysis, language acquisition, and a structurally
oriented historical linguistics are also discussed.
American approaches:
The actual father of American structuralism is considered to be Leonard Bloomfield,
whose groundbreaking work Language was published in 1933. His most important successors are Zellig S. Harris, Bernard Bloch, Rulon S. Wells, Charles F. Hockett and
Charles C. Fries.
As did the European schools originating with Saussure, the American structuralists
tried to find methods which permit the identification of linguistic units and their organization into classes by purely formal means. Bloomfield, who stands at the beginning of
this movement, refers principally to Saussure. Through Jakobson and Martinet, ideas of
the Prague school reached the Linguistic Circle of New York in the 30s. The focal points
were initially phonology and morphology. The distributionalists, especially Harris, also
systematically include syntax in their linguistic analysis. Characteristic for American
structuralism is the search for discovery procedures, i.e. for procedures which permit the
determination of the grammatical structure of sentences after a finite number of steps.
With Leonard Bloomfield, theoretical linguistics in the United States reached a new
level. All following language researchers are in Bloomfield's debt. Bloomfidd 's methodological premises are influenced by the dominating physicalistic scientific ideal of his
time. He only recognizes such data that can be discovered through direct observation
and immediate experience. Mentalistic procedures which in principle permit reference
to the linguistic knowledge of speakers/hearers of a language are excluded. Of course
Bloomfield doesn't deny the character of natural languages as sign systems, he only
excludes the explicit reference to meanings in the definition of grammatical concepts.
However, he states that the ability of native speakers to decide whether two forms are
the same or not must be permitted (Bloomfield 1933: 77). The meanings assoc iated with
linguistic forms are, in Bloomfield's view, not accessible to scientific description, at least
not for the current state of linguistic research. They must be investigated according to
principles which the behaviorist psychology of his time applied in the study of the
behavior of animals. He gives an example of how one should proceed according to these
principles in order to make scientifically founded statements about mental phenomena
such as linguistic meaning. With this example, he would like to demonstrate that even the
application of behavioristic methods to the analysis of linguistic meanings is practically
impossible.
We owe a significant further development of Bloomfield's approaches to Zellig
S. Harris. Harris viewed linguistics, similarly to Hjelmslev and later Chomsky, as applied
mathematics (Harris 1968: 1). His first great work was published in 1951 under the title
Methods in Structural Linguistics. He is the founder of distributionalism. The task of
linguistics is, according to his conception, "to describe the distribution or arrangement
within the flow of speech of some parts or features relatively to each other". There are
three tasks to be solved: 1. The units of different levels must be segmented, 2. The
resulting segments must be classified by means of an observation of their distribution.
The appearance of segments in the same context means membership in the same class.
3. The relations between the resulting classes must be determined (Harris 1951 : 20).
HaiTis assumes that these tasks can be completed via an effective procedure which
after a finite number of steps detennines the grammatical structure of linguistic utteran-
57
PA58
58
ces. This procedure must be possible without explicitly taking into account the meaning
or the sounds of linguistic units. "In exact descriptive linguistic work [ . .. ] considerations
of meaning can only be used heuristically, as a source of hints, and the determining
criteria will always have to be stated in distributional terms" (Harris 1951: 365 fn. 6).
In fact Harris must also allow for a native speaker who only answers the questions: Is
that the same? and Is that possible? Harris is however of the opinion that this only
makes short cuts possible which otherwise abbreviate very elaborate procedures. "It may
be presumed that any two morphemes having different meanings also differ somewhere
in distribution" (Harris 1951: 365 fn. 4). The assumption of effective discovery procedures generated great interest, but was also seriously questioned. In particular two points
are controversial: are the assumed procedures really effective, i.e. do they lead to an
acceptable conclusion? And is the justification for the acceptability of short cuts acceptable? The tenor of the criticism is: it is practically impossible to investigate all segments
in every possible context. In phonology this is in any case simpler than in morphology
or syntax. For the substitution-in-frames-procedures suggested by Harris, it cannot be
theoretically proven that it delivers the desired definitions after a finite number of steps
(Lees 1957). Problems which arise for the definition of the concept of morpheme are
also pointed! out by Bierwisch ( 1961 ).
The background of distributionalistic thinking is immediate constituent grammar. This
type of grammar assumes a very clear and manageable schema involving a strictly hierarchical structuring of the sentence. A sentence consists of phonemes. The combination
of phonemes yields morphemes. The combination of morphemes yields words and word
groups. The combination of word groups yields sentences (Postal 1964).
As most structuralists, Harris also follows Saussure's view that only an immanent
linguistic description of language can be the foundation for "historical linguistics, dialect
geography, for relations of language to culture and personality, to phonetics and semantics and for the comparison of language structure with systems of logic" (Harris 1951:
3).
4. Word -form ation in structuralist schools

In a description of the history of research on English word-formation, Valerie Adams
notes that American structuralism was only interested in questions of word-formation in
the context of morphological and syntactic problems, since its adherents focused their
main interest on units which are smaller or larger than words. Furthermore, Saussure's
distinction between synchrony and diachrony discredited the primarily historically oriented research on word-formation of the 19th century (Adams 1973: 5). It also applies to
the European schools that typical phenomena of word-formation were not systematically
followed up on by the founders of structuralist schools. Only to the extent that means
of word-formation are associated with the analysis of morphemes or with syntactic operations were they taken into consideration.
Hockett discusses, for example, models according to which words can be segmented
into morphemes (Hockett 1958: 393). He prefers an item-and-arrangement model, i.e.
the segmentation of expressions in units and the determination of types of arrangements
of these uni ts. The German verb form sucht ' searches' can, for example, be segmented in
PA59
such + t. The unit such also combines with other units: Such+e 'search', Ver+such+ung
'temptation', such+te 'searched' . All of these units are morphemes, i.e. the smallest
grammatical units which are combined with a meaning. Problems are caused by words
such as Engl. cran+berry, which contain units with which no meaning is associated,
words with ablaut (Engl. took : take, Ger. zieh : zag ' pull : pulled' and other problematic
cases in which words can be segmented into grammatical units which do not correspond
to a segmentation of the word into morphemes (Engl. worse, bad + comparative, Ger.
Lau/ 'walk', lauf + nominalization). Problems of this type are associated both with
inflectional forms of words as well as with products of word-formation, cf. zieh 'to pull' :
Zug 'pull, train', seh 'to see' : Sicht 'sight'. To describe these phenomena, Bloomfield
suggests an item-and-process model. Bloomfield also discusses relationships between
compounds and syntactic expressions (Bloomfield 1933: 227-32). He differentiates semisyntactic compounds from asyntactic compounds. Syntactic compounds have syntactic
expressions as direct parallels, cf. house keeper: keep house, blue-eyed: blue eyes. For
asyntactic compounds, no syntactic constructions exist in which the members of the
compound are constituents, cf. door knob. This distinction makes clear Bloomfield's
physicalistic scientific ideal. The members of a compound must be able to be immediately adjacent to each other in syntactic constructions. Without syntactic argumentation, he
rejects parallels such as knob of a door or door has a knob.
Tesniere includes products of word-formation which function as translatives in his
syntactic theory. Characteristic of sentence structure in Tesniere's model is the possibility
to transfer words with a specified grammatical category into words with another grammatical category. In the French construction le livre rouge 'the red book', rouge ' red' is
a word categorized as an adjective. But also in le livre de Pierre 'the book of Peter', de
transfers the noun Pierre to an adjective. De is a translative and Pierre a translate; de
Pierre and rouge are syntactically equivalent. The noun Pierre becomes an adjective via
the preposition de in de Pierre. Adjectives can be transferred to verbs w ith the help of
etre 'to be', e.g., La maison est neuve ' The house is new'. Bound morphemes can also
function as translatives. In Latin Venit Romam ' (He/she) came to Rome' the accusative
morpheme -am causes the transfer of a noun to an adverb. The use of German gut 'good '
in the syntactic position Er sch/aft gut ' he sleeps well' makes the adjective an adverb
without a translative. Verbs are seen as the basic category. But they can also be transferred to other categories. Thus, the subordination of relative clauses is a translation
from a structure with verb into an adjective. An object clause makes a structure with
verb into a noun. Means of word-formation can also be means of recategorization. The
German verb erobem ' to conquer' becomes a noun via the suffix -ung, as does the
adjectivefrei 'free' via the suffix -heit.
The reserved attitude toward a systematic search for rules for new wordls is, strictly
speaking, an expected consequence of the methodological principles assumed by the
individual schools:
1.
2.
3.
4.
Strict synchrony;
Mathematical regulari ty of the combination of linguistic expressions;
Meanings may only be taken into consideration via purely formal methods;
A mentalistic capturing of data, i.e. reference to the linguistic knowledge of native
speakers, is not permitted, with the exception of strictly regulated exceptions;
5. Language use does not influence statements about the language system;
59
PA60
60
6. These requirements cause concepts such as affixoid, unproductive rule, degree of

productivity of a rule, lexicalized word, ad-hoc formation, or acceptability of a new
formation to be excluded from the analysis of the system. They require a reference
to historical contexts or to language use.
Saussure already distinguishes suffixes from other morphemes. He assumes semantic
and grammatical functions for suffixes. This is permissible, because the question of
whether a morpheme is connected with a meaning or not is acceptable, just as is the
question of the function of a form within a syntagm. The suffixes of word-formation
(derivational suffixes) are associated with meanings, other suffixes (inflectional suffixes)
are markers for grammatical relations. But problems are caused by products of wordformation which entered the lexicon of a language in historic times, cf. Ger. Brom+beere
'blackberry', Dick-icht 'thicket', Fahr-t 'ride', Klang 'sound' (nominalization of kling
via ablaut). In this case, neither the question of whether the suffixes have their own
meaning, nor the question of why there are no comparable new formations can be decided on the basis of the principles considered to be permissible. It should generally be
noted that the syntactic aspect preferred in the structuralist theories. only concerns a
portion of word-formation patterns. Except for the syntactic categorization of complex
words, recategorization with the help of patterns of word -formation, and the few cases
of syntactic restructuring, the principles for the formation of new words are of a semantic
nature, cf. Fanselow (1987) and Motsch (2000).
A systematic investigation of the regularities of word-formation within the framework
of the methodological principles of most structuralist schools would have led to relatively
uninteresting, strongly overgeneralized statements. For example: all words occurring in
texts can be divided into derivatives and compounds. Derivatives are combinations of
words and affixes. Affixes are suffixes or prefixes. In contrast to words, which can occur
alone, affixes are bound to words. Affixoid is strictly speaking not a permissible concept,
since it assumes historical or semantic concepts. Affixes are signs, and thus carry meaning. Some suffixes have an innergrammatical function; they indicate the change in category of a word, cf. Ger. pruf 'to test' : Pruf+ung 'test',frei 'free' : Frei+heit 'freedom'.
Suffixes and prefixes are bound to word categories which are to be syntactically defined.
We must however note that, for example, a rule "N+ig yields a possible adjective in
German" is much too undifferentiated. Although there are many words for which this
rule applies,. cf. sandig 'sandy', eckig 'angular; lit. edgy', steinig 'stony', staubig 'dusty',
wolkig 'cloudy', for many other words, this statement is problematic, cf. *tischig 'tabley', *stuhlig 'chairy', *zimmerig 'roomy', *tagig 'day-y', *hausig 'housey', *dorfig
'villagey', *knopfig 'buttony'. Even if more specific statements about the meanings of
affixes were permitted and subclassifications of base words were possible, the problem
of overgeneralized statements would remain, cf. Motsch (2011). Compounds are combinations of words which belong to specific word categories. Whether formations with
syntactic constructions as a first member - cf. Ger. warm-herz+ig 'warm hearted; lit.
warm hearty', kurz-atm +ig 'short winded; lit. short breathy', Zwei-zu-eins-Sieg 'two-toone-win' - can be satisfactorily described in the framework of the principles of structuralist theories would need to be clarified, see article 33 on synthetic compounds in German. Compounds can be differentiated through the category membership of their constituents. Beyond that, only parallels to syntactic constructions can be noted. Probably
even the classic, semantically based differentiation between determinative, copulative
and exocentric compounds could not be reproduced.
PA61
It is not surprising that word-formation in European languages was primarily investigated by scholars who placed more moderate conditions on methodological principles.
This includes in particular the work of Marchand ( 1969), Dokulil ( 1968), and Fleischer
(1969). Synchronic analysis is the focus of this work as well. Many suggestions of
structuralist approaches are adopted, although in a weakened form. The greatest impact
was from Saussure's requirement of separating synchrony and diachrony. Apart from a
few historical references, one doesn't find systematic historical investigations in the
large body of works containing descriptions of word-formation. Thus, for example, an
investigation of derivational affixes following the assumptions of Jacob Grimm that they
trace back to independent words was almost completely lost. Grimm's distinction between proper compounds (echte/eigentliche Komposita) and improper compounds (unechte/uneigentliche Komposita) no longer plays a systematic role either, see article 3 on
word-formation in historical-comparative grammar.
Particularly worth mentioning in this context is also Coseriu's (1970) attempt to determine the structure of the vocabulary of a language. He assumes thereby jdeas of the
classical structuralists, in particular the postulate that all linguistic units, and therefore
also the meanings of linguistic expressions, enter into syntagmatic and paradigmatic
relations. He follows up on ideas of Hjelmslev, who, parallel to the segmentation of
morphemes into phonemes, suggested a decomposition of the meanings of the content
plane in figurae. According to Coseriu, the phenomena of the expression plane correspond to semes, with the help of which semantic relations between lexemes in the lexicon
can be described. Word-formation patterns are seen as special means for the establishment of semantic relations in the lexicon of a language.
Generative grammar also directed attention to structural connections which appear
to point out new directions for research on word-formation. Thus Lees (1960) uses
transformations in the first version of a generative grammar for the analysis of compounds formed by two nouns. He establishes a connection between syntactic constructions and compounds which makes it possible to highlight differences in meaning. The
problem nevertheless remains whether all prominent phenomena of word -formation can
be described in the context of a stricter theory of grammar, cf. Motsch (2011). The
mentalistic perspective taken by Chomsky and the idea of a modular organization of
systems of knowledge adopted from cognitive psychology freed the way for the systematic integration of semantic and pragmatic aspects in the analysis of word-formation.
However, for central questions of the theory of grammar, word-formation still remains
as a whole of marginal interest. Chomsky (1982: 96) took the view in a discussion with
Henk van Riemsdijk that "where we have options to get an infinite vocabulary, it appears
to be through pretty trivial mechanisms".
In any case, it is fair to say that structuralist schools have made great contributions
to the analysis of the elementary structure of words. Fundamental concepts were worked
out for analyzing the internal structure of words and their role in the combination of
words to phrases and sentences. We thank structuralist approaches for concepts such as
stem morpheme, derivational and inflectional morpheme, discontinuous morpheme, and
free and bound morpheme. Bloomfield, Harris and Hockett contributed greatly to these
results. It would be impossible to imagine modern research in morphology and wordformation without the preliminary work of the structuralists.
As an example for the influence of classical structuralist schools on research in wordformation, the influential theory of Hans Marchand will be briefly illustrated here.
61
PA62
62
Marchand's The Categories and Types of Present-Day English Word-Formation was

first published in 1960 and appeared in a second edition in 1969. As one of his most
well-known students, Dieter Kastovsky, writes, this book is even today "the most comprehensive synchronic description of Modem English word-formation" (Kastovsky 1999:
29). Marchand had already developed the theoretical basis of his work in the 1940s of
the last century. He bases his important points on European structuralist thought, especially on the basic ideas of Saussure and extensions of these ideas by Charles Bally. But
he also takes American stmcturalists into account. The confrontation with the wordformation theory of Robert Lees led to an extensive reworking of his book, which appeared in a second edition in 1969.
In contrast to the largely neogramrnarian orientation of the linguistics of his time,
Marchand emphasizes the priority of a synchronic description. References to historical
contexts do not have a systematic character. To the foundation of his theory of wordformation belong the concepts syntagm and motivation, which can be traced back to
Saussure and Bally, cf. Bally (1944). The subject matter of word-formation are regular
formations which can be analyzed as a syntagm which consists of two morphemes.
Morphemes are seen as linguistic units with the character of a sign. According to Saussure, the bond between the two components of the sign, the form and the meaning, is
fundamentally arbitrary and unmotivated. Complex linguistic expressions, syntagms, are
on the other hand principally motivated, since they can be semantica lly interpreted on
the basis of their constituents and a grammatical pattern. Marchand draws from this the
conclusion that meaning is just as important as form, since the smallest linguistic signs,
the morphemes, must be interpreted as signs (Marchand 1969: 1). He thus does not
accept the trend to a purely distributional definition of linguistic units in American linguistics and makes, as did later Chomsky, the mentalistic assumption that the speakers of
a language know the meanings of linguistic expressions and that this aspect of linguistic
competence must be taken into consideration in an adequate theory, even though there
were at the time no fully-developed proposals as to how linguistic meanings should be
described in an appropriate theoretical form.
This approach results in important consequences for determining the subject matter
of a theory of word-formation. There are words that form a syntagm, i.e. which are
formed from morphemes. Among these are, according to Marchand, compounds, prefixations, suffixations, zero-derivatives and backformations. Other analyzable words cannot
be reduced to syntagms. This includes onomatopoeia (Ger. Schnick-Schnack 'tittle-tattle '), blends (Ger. Kurlaub ' medical treatment holiday', from Kur 'cure' and Urlaub
'vacation'), words with ablaut (Ger. Zug 'train' : zieh(en) 'to pull' ). Words with parts
which do not have a meaning or for which there are no recognizable patterns also do
not belong to synchronic descriptions of word-formation (Engl. cran+berry). In contrast
to the distributionalists, Marchand rejects analyses such as Engl. re-ceive, de-ceive, conceive, etc., since neither the prefixes nor the bases have a recognizable meaning. He
regards such words as monomorphemic. Word-formation is based on bimorphemic syntagms, which must be productive to a certain degree, i.e. are the basis for new formations. In contrast to Harris, who completely excludes productivity from grammatical
analysis, Marchand uses this phenomenon for the definition of word-formation patterns.
Harris ( 1951: 255) writes with theoretical consequence from his methodological premises, "The methods of descriptive linguistics cannot treat the degree of productivity of
elements". The tenn, to be defined by language use, offers, together with the concept of
PA64
64

Baum, R ichard
1976 Dependenzgrammatik. Tesnieres Model! der Sprachbeschreibung zn wissenschaftsgeschichtlicher und kritischer Sicht. Ti.ibingen: Niemeyer.
Baumgartner, Klaus
1976
Konstituenz und Dependenz: Zur Integration der beiden grammatischen Prinzipien. In:
Hugo Steger (ed.), Vorschliige far eine strukturale Grammatik des Deutschen, 52- 77.
Darmstadt: W issenschaftliche Buchgesellschaft.
Bierwisch, Manfred
1961
Uber den theoretischen Status des Morphems. Studia grammatic a 1: 51 - 89. Berlin:
Akademie Verlag.
Bloomfield, Leonard
1933
Language. New York: Holt.
Chomsky, Noam
1982
On the Generative Enterprise. A Discussion with Riny Huybregts and Henk van Riemsdijk. Dordrecht: Foris.
Coseriu, Eugenio
1970
Einfuhrung in die strukturelle Betrachtung des Wortschatzes. Ti.ibingen: Niemeyer.
Dokulil, Milos
1968
Zur Theorie der Wortbildungslehre. Wissenschaftliche Zeitschrift der Karl Marx Universitat Leipzig. Gesellschafts- und Sprachwissenschaftliche Reihe 17: 203-211.
Fanselow, Gisbert
1987
Gemeinsame Prinzipien von Wort- und Phrasensemantik . In: Brigitte Ansbach-Schnitker, Herbert E. Brekle and Johannes Roggenhofer (eds.), Neuere Forschungen zur Wortbildung und Historiographie der Linguistik, 177- 194. Tiibingen: Niemeyer.
Fleischer, Wolfgang
1969
Wortbildung der deutschen Gegenwartssprache. Tiibingen: Niemeyer.
Garvin, Paul L.
1954
Review of Prolegomena to a Theory of Language by Louis Hjelmslev. Language 30(1):
69-96.
Godel, Robert
1961
L'Ecole saussurienne de Geneve. In: Christine Mohrmann, Alf Sommerfelt and Joshua
Whatmough (eds .), Trends in European and American Linguistics 1930- 1960, 294299. Utrecht/Antwerpen: Spectrum.
Harris, Zellig S .
1951
Methods in Structural Linguistics. Chicago: Chicago University Press.
Harris, Zellig S .
1968 Mathematical Structures of Language. New York: Wiley lnterscience.
Haugen, E inar
1951
Directions in modern lingu istics. Language 27: 2 11- 222.
Heringer, Hans Ji.irgen, Bruno Strecker and Rainer Wimmer
1980
Syntax: Fragen - Losungen - Alternativen. Mi.inchen: F ink.
Hjelmslev, Louis
1943
Omkring sprogteoriens grundlreggelse. Copenhagen: Munksgaard.
Hjelmslev, Louis
1961
Prolegommena to a Theoty of Language. Madison: University of W isconsin Press.
Hockett, Charles F.
1958 A Course in Modern Linguistics. New York: Macmillan.
Hudson, R.A.
1980
Constituency and Dependency. Linguistics 18: 179- 198.
Jakobson, Roman, Sergej Karcevski and Nikolai S. Trubetzkoy
1929
Proposition au Premier Congres International de Linguistes a La Haye, 10- 15 avril
1928, 33- 36. Leiden: S ijthoff.
PA65
Jager, Ludwig
Ferdinand de Saussure zur Einfiihrung. Hamburg: Junius.
2006
Jespersen, Otto
A Modern English Grammar on Historical Principles. Vol. 6: Morphology. London/
1942
Copenhagen: George Allen & Unwin.
Kastovsky, Dieter
Hans Marchand's theory of wordjormation : Genesis and development. In: Uwe Clark
1999
and Peter Lucko (eds.), Form, Function and Variation in English. Studies in Honor of
Klaus Hensen, 19-39. Frankfurt/M.: Lang.
Kunze, Jiirgen
Abhiingigkeitsgrammatikfiir das D eutsche. Berlin: Akademie Verlag.
1975
Lees, Robert
Review o f Syntactic Structures. Language 33: 375-408.
1957
Lees, Robert
The Grammar of English Nominalizations. The Hague: Mouton.
1960
Levi-Strauss, Claude
1958
Anthropologie Structurale. Paris: Pion.
Marchand, Hans
1969
The Categories and Types of Present-Day English Word-Formation. A Synchronic-Diachronic Approach. 2nd ed. Mi.inchen: Beck.
Martinet, Andre
1946
Au sujet des fondaments de la theorie lingu istique de L. Hjelmslev. Bulletin de la Societe
Linguistique de Paris 124: 19- 43.
Mathesius, Vilem
1929
Zur Satzperspektive im modernen E nglisch. Archiv far das Studium der neueren Sprachen und Literaturen 155: 200- 210 .
Motsch, Wolfgang
2000
Syntaktische Konsequenzen van Wortbildungsmustern. In: Josef Bayer and Christine
Romer (eds.), Von der Philologie zur Grammatiktheorie: Peter Suehs/and zum 65. Geburtstag, 289- 302. Ti.ibingen: Niemeyer.
Motsch, Wolfgang
Methodologische Aspekte der neueren Sprachforschung. In: Kristel Proost a nd Edeltraud
2006
Winkler (eds.), Von der lntentionalitiit zur Bedeutung konventionalisierter Zeichen.
Festschriftfar Gisela Harras zum 65. Geburtstag, 327- 336. Ti.ibingen: Narr.
Motsch,Wolfgang
2011
Grammatische und sprachpsychologische Aspekte der Wortbildung. In: Hilke Elsen and
Sascha Michel (eds.), Wortbildung im Deutschen zwischen Sprachsystem und Sprachgebrauch. Perspektiven - Analysen - Anwendungen, 43- 72. Stuttgart: ibidem.
Postal, Paul
Constituent Structure. T he Hague: Mouton.
1964
Saussure, Ferdinand de
19 16
Cours de linguistique generate. Ed. by Charles Bally and Albert Sechehay. Paris/Lausanne: Payot.
Siertsema, Berta
A Study of Glossematics. The Hague: Mouton.
1955
Spang-Hansen, Henning
1961
Glossematics. In: Christine Mohrmann, Alf Sommerfelt and Josua Whatmough (eds.),
Trends in European and American Linguistics 1930- 1960, 128-164. Utrecht/Antwerpen: Spectrum.
Tesniere, Lucien
Elements de syntaxe structurale. Paris: Klincksieck.
1959
65
PA66
66

Trubetzkoy, N ikolai S.
Grundzuge der Phon.ologie. Prag: [Harrassowitz].
1939
Wells, Ru lon S.
De Saussure's system of linguistics. Word 3: 1-3.
1947
Wolfgang Motsch, Berlin (Germany)
5. Word-formation in inhaltbezogene Grammatik

1.
2.
3.
4.
Preliminary remarks on principles

On the specifics of an inhaltbezog ene theory of word-formation
Conclusion and prospects
References
Abstract
Word-formation in inhaltbezogene Grammatik [content-based grammar} directs its interest at the activity of the word-formation faculty and thus at the creative potential developed in a language community, to make an organized reality accessible through the
formation of words and concepts.
1. Preliminary remarks on principles

Every standard language (written, standardized language of culture, further developed
through modernization) is the manifestation of the creativity of many generations, groups
and creative personalities. In confrontation with their respectively experienced reality,
they have expanded their linguistic talents and - in extending the language's means of
access - created a many-membered, much usable instrument of mind that goes far further
than the purposes of elementary understanding. On the one hand, through its vocabulary,
it makes a network of semantic categories available, with the help of which classes of
entities can be differentiated and named in a nuanced manner from various points of
view, by means of partially synonymous words from lexical paradigms (word fields).
On the other hand, during language acquisition, during the process of becoming familiar
with the possibilities of symbolic action (of communicative practice) in a community of
language and interaction, grammatical forms for the organization of speech are also
conveyed and thus syntactic categories for the construction of complex fonns which can
categorize situations and intentions. How many and which categories are made distinct
in a language and how they are applied depends obviously on the cognitive and communicative expressive needs of the speakers, the former and present participants in a lan-
PA67
guage community. In order to gain new means of expression, they especially use the
efficacy of their word-forming faculty (Humboldt 1973 [ 1836]: 99), their ability "sich
iiber Zeichenkonstruktionen eine gegliederte Wirklichkeit zuganglich zu machen" [to
make an organized reality accessible for themselves via symbolic constructions] (Schlogl
2004: 14). Thereby the orientation necessary for any communal life becomes possible within the scope of the respective socio-cultural circumstances - agreement on "Weltwissen" [world knowledge] and the "Welt des gesellschaftlichen Handelns" [world of social
action] (Soeffner 2004: 47). Considering the creativity of the individual language communities which have become selectively effective under specific historical conditions,
their different character and often also formal as well as functional alteration of categories, it is not surprising that the discipline of comparative linguistics has ascertained
"unterschiedlichste Verteilungen (... ] iiber die Welt hin" (the most variable distributions
across the world] (Hartmann 1958: 44) and come to the finding " <lass sich eine Parallelitat zwischen Kategorien und Realitat kaum von nur einer Sprache aus biindig entscheiden
!asst" [that a parallelity between categories and reality can hardly be conclusively decided from the perspective of only one language] (Hartmann 1958: 45). For example, the
color spectrum is not subdivided into the same number and value of color terms in all
languages, also, not all individual languages correspond in their rendering of space and
time reference, for example in the temporal gradation of a verbal tense system, that is,
the development of the human language faculty has not generally led to the development
of language systems which completely correspond with respect to the available semantic
and syntactic categories: "The empirical fact is that different cultural groups of the world
have conventionalized very different sets of linguistic conventions and cons.tructions, in
some cases based on very different grammatical principles" (Tomasello 2008: 309). This
is not only shown by the examination of non-European languages, for instance the
"American aboriginal languages" (Boas 1948: 199 and 207). Even the genetically related
languages of the lndo-European language family do not show "in allen Sprachen gleichermaf3en vorhandene Kategorien ohne Defektivitiit und mit eindeutigen, unterschiedenen Signifikanten" [in all languages equally available categories without defectiveness
and with unequivocal signifiers] (Adrados 1985: 9). From this follows the task for linguistics to also work out what goes beyond those universals assumed as probable: the
specifics of individual languages with regard to their "obligatory aspects" (Deutscher
2010: 151 ). Making conscious points of view (Sichtweisen) with a partially differing
categorial network for grasping reality is a special concern of the "inhaltbezogene Grammatik" [content-based grammar] (Weisgerber 1962) which is able to resume ideas of
Wilhelm von Humboldt "iiber die Verschiedenheit des menschlichen Sprachbaues und
ihren Einfluss auf die geistige Entwickelung des Menschengeschlechts" [on the diversity
of human language construction and its influence on the mental development of the
human species] (1830-35, first printing 1836).
"Die Vorstellung, daf3 Denken und Sprache miteinander korrelieren, ist charakteristisch for den Rationalismus der Aufkliirungszeit" [The notion that thinking and language
correlate w ith one another is characteristic for the rationalism of the Age of Enlightenment] (Gardt 2000: 177). Gottfried Wilhelm Leibniz has already ascertained that language is "kein blof3er Spiegel des Verstandes mehr, der nur passiv das Denken abbildet,
sondern zugleich <lessen Instrument und Stimulans" [no longer a simple mirror of the
intellect, which just passively reflects thinking, but also at the same time its instrument
and stimulant] (Gardt 2000: 179). Thus language has a cogni tive function alongside its
67
PA68
68
communicative function. Leibniz still had the hope of being able to overcome the perspectivity of perception. "Fur Humboldt dagegen scheint die sprachbedingte Perspektivitat fast zur conditio humana zu gehoren" [For Humboldt, on the other hand, perspectivity
conditioned by language appears to almost belong to the human condition] (Gardt 1999:
23). "Die Zuspitzung des Problems zur These vom sprachlichen Relativitatsprinzip geht
auf den Amerikanisten B. L. Whorf zuri.ick" [The extreme version of the problem of the
thesis of the principle of linguistic relativity dates back to the Americanist B. L. Whorf]
(Gipper 1972: 5; see also Gumperz and Levinson 1996 as well as Wildgen 2008: 12- 14
and 42-44). "However, a weaker version of the Sapir-Whorf hypothesis is generally
accepted" (Crystal 1991: 15). For all constructivist theories it is common knowledge
"dass sie der Sprache ein erkenntnistheoretisches Apriori zuerkennen, Sprache als ganz
entscheidend fiir die Wahrnehmung und intellektuelle Verarbeitung der Wirklichkeit betrachten" [that they judge language to possess an epistemological apriority, regard language as fully decisive for the perception and intellectual processing of reality] (Gardt
2009: 1199).
Incorporating modern insights into the nature and role of linguistic signs (Ferdinand
de Saussure, Walter Porzig, Jost Trier, Ernst Cassirer), inhaltbezogene Grammatik was
designed in its objectives and its systematics above all by Leo Weisgerber (from 1942
to 1967 Professor of General Linguistics and Celtic Studies in Bonn) and founded as a
linguistic line of thought which endeavors to go beyond the sound-based view of language. According to this conception, it is essential to convert sound-based observations
into content-based insights (Weisgerber 1962: 119) and to make the content side, which
structures reality, the point of reference of the description. This holds for the description
of patterns of sentential semantics which are available to conceptualize situations (as
actions or events, as existence or essence). This is also true for the examination of the
vocabulary, which is not only sound-based, entered in individual entries of an alphabetically ordered dictionary - with occasional glances at semantically related words - but
rather studied and described as a semantic network. In this way, the conceptual grasping
of reality may become clear, the respectively reached particularity of potential for differentiation which a linguistic convention offers and suggests to the participants in a language community. With the attempt to place the mental side of language and its effects
for the individual and the respective society in the focus of research, linguistics increases
its value as a science of culture. It would become a "Basiswissenschaft innerhalb der
Kulturwissenschaften" [basic disc ipline within the sciences of culture] (Kamper and Eichinger 2008: 370), and could contribute to a "Kulturanthropologie in der heutigen Konstellation einer [ ... ] zunehmend multikulturell begriffenen Welt" [cultural anthropology
in today 's constellation of an [ ... ] increasingly multiculturally understood world] (Fuchs
2001: 21).
"The lexicon of a language is perhaps the most direct link with cultural conceptualizations in the sense that lexical items largely act as labels, and hence 'memory banks' for
conceptualizations that are culturally constructed[ ... ]. In short, the lexical items of human languages need to be viewed as capturing and storing cultural conceptualizations
such as cultural schemas and categories" (Sharifian 2009: 168). That is why the cognitive
anthropology developed in the USA sees as necessary an "ethnosemantischen Zugriff'
[ethnosemantic access] (Maeder and Brosziewski 2007: 269) to "den in einer Kultur
angelegten Sprachkategorien" [the language categories inherent in a culture] and points
out "<lass Dinge, Ereignisse und Verhaltensweisen Einheiten sind [ ... ],die i.iberhaupt erst
PA69
im Netz sprachlicher Differenzierungen, Abstraktionen und Verweisungen konstituiert
werden" [that things, events and behaviors are entities [ ... ] which are anyway only
constituted in a network of linguistic differentiations, abstractions and references]
(Maeder and Brosziewski 2007: 273).
"Dass Weisgerber in der (heute wieder sehr aktuellen) Debatte um sprachlichen Universalismus oder Relativismus eine berilcksichtigungswerte Position eingenommen hatte,
die es verdienen wilrde, rezipiert und diskutiert zu werden" [That Weisgerber has taken
a position worthy of consideration in the (today once again very current) debate about
linguistic universalism or relativism which would deserve being received and discussed]
(Sylla 2008: 702) is the result of a detailed critical examination of the theses of research
on language content. It is thus not surprising "<lass die Sprachinhaltsforschung [ ... ] in
der ehemaligen UdS SR, in Japan und in Sudkorea auf Interesse stoJ3t" [that research on
language content [ ... ] meets with interest in the former USSR, in Japan and in South
Korea] (Kaltz 2006: 2174). It is without doubt that it - although modified and more
strongly operationalized - has become especially important for the further development
of semantics and likewise for research on word-formation.
For this, Hermann Paul already demanded a "Erganzung <lurch die Bedeutungslehre"
[supplementation through the study of meaning] (Paul 1981 [1896]: 18 and 19). One
must also begin with functions and investigate "welche Ausdrucksformen dafilr neben
einander zur Verfiigung stehen" [which forms of expression are simultaneously available
in order] to get a clear view of the "Konkurrenz gleichbedeutender Ausdrucksfonnen"
[competition of synonymous expressions] (Paul 1981[1896]:19). Inhaltbezogene Grammatik then seeks to fulfill the demand of reaching "eine zusammenfassende Behandlung
der verschiedenen gleicher Funktion dienenden Bildungsweisen" [a summarizing treatment of the various kinds of formations which serve the same functions] and paying
attention to functional "Schattierungen" [shadings] ( 19).
2. On the specifics of an inhaltbezogene theory of word-formation

2.1. General remarks
The creativity which was mentioned initially, the productive process of the development
of linguistic energeia in a language community is especially obvious in the area of wordformation. Here, the objective is to gain insight how, i.e. with which methods, objects
(that which is experienced, felt, thought) which have become relevant for the text are
made nameable and communicable. It is a question of preferred methods of operation for
conceptualization and nomination, also of clarifying how patterns of word-formation in the service of particular speaker strategies - "sich auf die Erfassung und Darstellung
sprachlich vermittelter Wirklichkeit auswirken" [affect the capturing and depiction of
reality conveyed by language] (Erben 2003: 253 7). It is thus necessary to not be content
with taking inventory of individual means of word -formation and with working out
possible patterns of construction (structural schemata), that is with the form-oriented
descriptions of a morphological theory of word-formation. In such a theory, individual
forms and combinations are indeed assigned functional values, but the syste matic, paradigmatically graded interaction of means of word -formation, which differ in form but
69
PA70
70
are similar in function , in the realization of patterns typical in language are not worked
out. As a result, the task does not become apparent to also investigate the contentual
directions of expansion of the vocabulary (Weisgerber 1971: 114) especially the development of cognitively and communicatively important derivational categories, for each of
which a number of word-forming morphemes have become common. If one would like
to avoid an all too simple conception of word-formation as a mechanism for the creation
of words and to better perceive the creativity of the word-formation faculty in the development of speech (texts) and in the expansion of language (of the lexicon), then the
perspectives and findings of the inhaltbezogene theory of word-formation must additionally be gained. Since the German language possesses a particular wealth of options in
word-formation and extensive descriptions of this are available, especially the multivolume inventory of the Institut fur deutsche Sprache in Mannheim: "D eutsche Wortbildung. Typen und Tendenzen in der Gegenwartssprache" (1973-1992, cf. Ortner and Ortner 1984; Ortner, Ortner and Wellmann 2007), the opportunity presents itself to examine
the available results from the perspective of the theory of content-based word -formation
and to point out much of fundamental significance. It is important "dass die in ihrer
Leistung sich nahe kommenden Ableitungsmittel (wortbildenden Morpheme, Affixe)
auch in ihrer inhaltlichen Nahe und Zusammengehorigkeit erkennbar gemacht [werden]"
[that the means of derivation (morphemes of word-formation, affixes) which are similar
to one another in terms of performance also [be] made distinguishable in terms of their
contentual nearness and coherence] (Erben 1979: 158) as members of a word class
(Wortstand) or a functiona l class (Funktionsstand).
A simple example is for instance the regular interaction of -chen and -lein in the
diminution of base nouns, in order to express a perspective of the speaker regarding
subj ective assessment, sympathy (Wellmann 1975: 123) with respect to the denoted entities.
In a content-based theory of word-formation, the word-formation paradigms (word
or functional classes - Wort- oder Funktionsstiinde) are thus to be described, which are
available to the participants in a language community for particular contentual directions
of expansion of the vocabulary. This availability relates not only to the possibilities of
semantic modification of base words (base morphemes) such as in the case of diminution
(Hund-chen 'doggy') or of gender transposition, gender differentiation (Hund-in ' bitch').
Since word-formation occupies a central position between the lexicon and syntax, and
the word is almost to be seen as a "Vermittlungseinheit zwischen semantischer und
syntaktischer Struktur" [unit of mediation between semantic and syntactic structure]
(Heidolph, Flamig and Motsch 1981: 159), the fact is still of fundamental importance
that word-formation can make areas important to communication accessible in various
semantic-syntactic subdivisions (Aufgliederungen). The word-formation faculty helps the
expansion of the open classes of lexematic words, i.e. especially the nouns, adjectives
(adverbs) and verbs, which in regular interaction take particular roles in the sentence as
well as in the development of speech. They specify through their categorial meaning at
the same time " [<las] Wie der Erfassung der auf3ersprachlichen Welt" [the how of the
comprehension of the world outside of language] (Coseriu 1972: 82).
This is variable, for example in the case of the words of the series warm - Wiirme erwiirmen 'warm - warmth - to warm', which grasp and portray a phenomenon as a
property, material (indeed measurable) substance, or as an action. Word-formation can
therefore make possible a change in word class (transposition) and thereby in categorial
PA71
meaning. The methods of operation of derivation can create transitions and thus increase
the flexibility of statements via the remodelling of base words. Thus a base adjective
such as warm 'warm' can be verbalized or nominalized: er-warm-en 'to warm', Warm-e
' warmth' . In this way, various representations of a state of affairs are possible: as something that is there/thus (Das Wasser ist warm, hat eine Warme van 38 Grad ' The water
is warm, has a warmth of 38 degrees') or rather happens/causes/acts (Die Sonne erwarmt
das Wasser ' The sun warms the water' ), so that a situation can be presented to the hearer
(reader) as static or dynamic, for instance as in the "Denkmodell der Handlung" [paradigm of action] (Buhler 1934: 25 1).
2.2. The diversity of the formation of nouns

For the formation of nouns, a particularly large number of means and patterns of wordformation for compound ing and derivation have been developed in German so that constantly new or newly conceived entities can be seen and named as bearer, goal or circumstance of an occurrence (being). This occurs via semantic modification of base nouns or
via transposition of verbs or adjectives to the word class of nouns (reclassification).
What is captured verbally as occurrence (doing) can be detached from the individual
event, e.g., x warms y, and at the same time grasped on a level of abstraction, hypostasizing and typifying, as a discussible entity via the derivation of action nouns (nomina
actionis) such as Erwarm-ung 'warming'. The necessity o f often being obligated to speak
about actions and actors has caused the derivational categories action noun and agent
noun to be particularly markedly developed, whereby even a series of foreign suffixes
has been taken into service. Thus, beside the particularly frequent -ung, the suffix
-(at)ion has also become available, beside -er, we also find -ant/-ent, -(at)or, -eur and
-ist. Equally important is the development of quality nouns (nomina qualitatis), which
make it possible to speak about human or animal qualities or the nature of envisaged
things. Here, the adjective is used - on the level of simple existential pred ication but when one wishes to make general statements about qualities, then abstract nouns
(Nominalabstrakta) must be activated or formed, e.g., Warm-e 'warmth ', Streng-e 'strictness', Schlau-heit 'cleverness', Naiv-itat 'naivety'.
For the development of important derivational categories, a large number of affixes
are available which correspond in their functional direction, that is, they belong to the
same functional class, and, where applicable, can be deployed in a paradigmatically
graded manner, depending among other factors on the phono logical-morphological nature of the base word: Biill-chen 'small ball ', Biich-lein 'small book'. The range of
choices and the conditions of use of converging or competing affixes must each be
described, also with respect to their frequency of occurrence and the share that they each
have as "Indikatoren for semantische Muster" [indicators for semantic patterns] (Motsch
2004: 12).
Thus in contemporary German, it is for example the primary function of -schaft
(-schaft') to indicate collective terms for persons (Arbeiter-schaft 'employees', .1frzteschaft 'medical profession'). Collective formations with -tum, the most frequent function
of which (-tum 1) is the assignment-of-roles type Auj3enseitertum ' outsiderdom ' ('the fact
of being an outsider'), are less frequent. They only follow from a secondary function of
71
PA72
72
-tum (-tum 2 : Bauem-tum 'peasantry; lit. farmer-dorn', Burger-tum 'bourgeoisie; li t. citizen-dom') and are derivatives which not only designate a multiplicity grasped as a unity
(of farmers or citizens) but rather "immer noch die Komponente der geistigen Richtung
und des Verhaltens enthalten" [still contain the components of mental direction and of
behavior] (Wellmann 197 5: 178).
For the functional class of collective formations as well as for other paradigms of
word-formation as contentual directions of expansion of the vocabulary, "das Zusamrnenwirken inhaltlich benachbarter Wortnischen aus verschiedenen Ableitungstypen"
[the interaction of contentually neighboring word niches from various derivational types]
(Weisgerber 1981 [1964]: 42) is of particular importance, whereby in principle the possibility of semantic differentiation holds, which accentuates particular points of view. In
this way, Burger-schaft 'the entirety of the citizens of a community' can be differentiated
from Burger-tum 'social class of citizens' . In the present case, especially the lemma-rich
semantic niches (Baldinger 1950: 117) of the formations with -schaft' and -tum 2 serve
as a model for the formation of add itional collective nouns from personal nouns. The
more semantic niches an affix helps to develop, the less explicit its signal value appears
to be, but it is not least the number and usualness of the formations which belong to a
semantic niche which are decisive for its degree of explicitness and systemic importance.
For this reason statistical findings are just as indispensable as analyses substantiated by
corpus linguistics.
They show, for example, that the agent type Lehr-er 'teacher' (-er' ) accounts for
approximately two thirds of the total inventory of formations w ith -er, whereas formations such as Versprech-er 'slip of the tongue; lit. misspeaker' (-er 5 ), only constitute a
very small niche in the paradigm of action nouns, at most 1 % of all deverbal nouns. In
the case of designations for personal roles, a combination with the gender differentiating
suffix -in is possible, that is the coupling -er-in. As a result, the unmarked form which
abstracts from natural gender (cf. the occupational title Lehr-er ' someone who teaches')
becomes a designation which is unambiguously gender determined, which refers exclusively to female actors (Lehr-er-in ' female teacher; lit. teacheress' ). In contrast, according to a feminist approach, the generic, that is the common noun used for both genders,
grammatically masculine forms (e.g., j eder!alle Lehr-er 'every/all teacher/s', Benutz-er
'user', Les-er 'reader', Mitarbeit-er ' co-worker') should be avoided, because " l'oscillation entre masculin specifique et masculin generique" [the oscillation between the specific masculine and the generic masculine] suggests a specific masculine reading (Elmiger
2008: 112) and thereby "une vision masculine du monde" [a masculine view of the
world] (Elmiger 2008: 35). This linguistic effect cannot be completely excluded, but is
probably no>! generally intended.
"Mit der inhaltbezogenen Wortbildungslehre riicken auch die Wortfamilien in ein neues Licht[ ... ] , alles <las, was aus demselben Wortstamm <lurch Ableitung und Zusammensetzung an Wortgut gewonnen ist" [With the content-based theory of word-formation,
the word families are also shown in a new light [ ... ], all that which is won from the
same word stern in terms of vocabulary through derivation and compounding] (Weisgerber 1962: 2B), the "Ga nzheit der dem 'Sprachgefiihl' gege nwiirtigen Bildungen gleichen
Stammes" [entirety of the formations of the same stem which are avaifable to ' linguistic
intuition' ] (Weisgerber 1962: 233). Which degree is respectively reached by the diversification of the derivationally and compositionally active base morphemes depends on their
relevance for the discourse of an age, and from the ir semantic appropriateness as an
PA73
onomasiological base for the conceptual or metaphorical classification of phenomena.
This has an effect not only in derivation, but also in compounding. In the case of the
particularly frequent detenninative compounds, the first element appended for differentiation states "unter welchem Gesichtspunkt eine Erscheinung dieses Feldes gesehen und
gesondert ist" [under which point of view a phenomenon of this field is seen and separated] (Brinkmann 1981 [1956]: 196), cf. Wach-hund 'watch dog', Hute-hund 'herding
dog ', Haus-hund 'house dog', Jagd-hund 'hunting dog', Schdfer-hund 'shepherd dog ',
SchojJ-hund 'lap dog', Wild-hund 'wild dog'. The conventionally preferred means of
naming, which for instance highlight use or affil iation, con-espond in the norm for designation characteristic of individual languages to a series of relational or constructional
meanings (Erben 2006: 73), which faci litate the formation and understanding of additional compounds.
In principle it is of course correct, that, e.g., a compound such as Papier-korb 'wastepaper basket; lit. paper basket' only says in a very abstract manner " Korb, der etwas mit
Papier zu tun hat" [basket that has something to do with paper] (Coseriu 1977: 50). The
semantic relation between the elements of a compound is not always more precisely
characterized. But the existence in the mental lexicon of compounds which have long
since become established, such as Brot-korb 'bread box', Flaschen-korb 'bottle basket',
Salat-korb 'salad basket', Wdsche-korb 'laundry basket', together with the corresponding
specialized and world knowledge (not acquired and formed without communicative, that
is, linguistic influence) as well as the contexts typical in language (wiif die Zeitung in
den Papierkorb ' throw the newspaper in the wastepaper basket') leave little doubt that
Papier-korb cannot be a 'basket made out of paper ', but rather a 'basket which is used
to collect paper', a 'basket for paper', that is an expression of purpose. Beyond that, it
can be established that for formations with this semantic pattern it is necessary "in vielen
Beispielen [ ... ] das Pradikat AUFBEWAHREN zu interpolieren" [in many cases[ ... ] to
interpolate the predicate AUFBEWAHREN 'keep, store'] (Motsch 2004: 415), when the
second element designates a vessel or container. According to its constructional meaning,
Papier-korb is in accordance with a considerable series of additional formations such as
Akten-schrank ' filing cabinet; lit. file cabinet', Blumen-vase 'flower vase', Milch-kanne
' milk can', Sand-kasten 'sand box'.
In the domain of compounding as well the task presents itself to go beyond the level
of a theory of form-based word-formation, to recognize formations related to constructions and to grasp them as realizations of semantic patterns, and thus also the particularity
of the concept and mental-linguistic classification of that which is designated. A characteristic example for this are the dialectal designations of the potato, which as a bulbshaped fruit is associated with its place of occurrence: Erd-apfe l lit. 'earth-apple' or
Grund-birne lit. 'ground-pear'.
In the lesser used type of the coordinative compounds "wird das Relatum unter zwei
verschiedenen Gesichtspunkten im Hinblick auf zwei verschiedene Seins-/Existenzweisen mit zwei kohyponymen Lexemen angesprochen" [the relatum is referred to with two
co-hyponymic lexemes from two different perspectives with respect to two different
modes of being/existing] (Ortner and Ortner 1984: 53): Gott-mensch 'God-man; lit. godhuman being' or Hass-liebe ' love-hate; lit. hate-love'. Similar bundlings of different
aspects are shown by adjective compounds such as nass-kalt 'wet and cold; lit. wetcold' or schaurig-schon 'scary and beautiful; lit. scary-beautiful ' .
73
PA74
74
2.3. On the formation of verbs and adjectives

For the formation of verbs, above all a multitude of prefixes and prefix-like particles are
available to semantically or syntactically modify base verbs or to verbalize nouns with
a particular marking. Here as well many semantic niches are marked respectively, mostly
with ver- and er-, the fewest with ent- and zer-, which are thus more explicit. And there
are functional classes - "mit starkerer Haltekraft und groJ3erer Dichte, aber ( ... ] auch
auJ3erst lose" [with stronger cohesion and greater density, but( ... ] also very loose] (Henzen 1981 (1958]: 80) - which more or less indicate a graduated interaction of different
prefixes.
Thus mainly formations with an- and be- are found when the concern is to indicate
the spatial orientation of a convergence on an object or its grasping through the process
of the action of the verb. But an- often signals more "einen punktuellen Kontakt, behingegen die voile Erfassung des Objekts" (a punctual contact, be- on the other hand
the full grasping of the object] (Kiihnhold and Wellmann 1973: 202), cf. etwas an- und
be-strahlen 'to beam at and illuminate something'. If the objective is to show that the
action of the verb has led to a conclusion, then formations with er- typically show the
successful conclusion, with veron the other hand often the missing of a target, or the
impairment or the elimination of an object, cf. etwas er-wirtschaften 'to gain something'
and ver-wirtschaften 'to squander away something'. The semantic modification of a base
verb can be linked to a change in syntactic valency. In many cases, a transitivizing of
intransitive verbs is brought about, that is the establishing of a relation to an object:
weinen 'to cry' ~ jemanden/etwas be-weinen 'to cry about something/someone; lit. to
cry something/someone'. Even more important is the possibility of reversing common
relations to objects, that is of changing the syntactic valency, which at the same time
changes the perspective of the representation of a state of affairs: jemanden um etwas
bitten 'to ask someone for something' ~ etwas van j emandem er-bitten 'to ask something of someone'. Means of word-formation can extend the manner of construction of
base verbs; they make possible via introduction of arguments or alteration of argument
structure a presentation of states of affairs which is adequate to the situation and effective
on the level of sentential semantics.
Here as well it becomes clear that word-formation offers different modalities of
presentation, and thus contributes significantly to the constitution of texts. Important
thereby are also forms of reclassification of base verbs to deverbal adjectives. This
concerns especially the formation of adjectives which evaluate verb processes as behavior (riihr-ig 'active') or as suitability (ess-bar 'edible').
Formations with -bar can be formed from every transitive verb with which an action
can be described. They have then a passive-modal meaning, i.e. ess-bare Friichte 'edible
fruits' are those which can be eaten (according to human experience without harm).
The sorting of phenomena, whether they are suitable for human purposes, finds a clear
formulation in this type of word-formation, which spares laborious syntactic construction.
2.4. The diversity of verba l or nom inal structures of comparison

If one asks what kinds of contentual relations appear particularly important, which are
therefore expressed in the expansion of the vocabulary in all three open word classes,
PA75
then one encounters the endeavor to grasp new perceptions through a comparin g referral
back to known classes, that is to express relationships of comparison: pendel-artig ' like
a pendulum', pendel-n 'to move back and forth like a pendulum' , Pendl-er 'commuter '.
Often the nomination forms of compounding and derivation compete: Gold-kind lit.
'gold-child ', gold-ig-es Kind lit. 'golden child ' for the meaning 'welcome, cherished like
gold'.
It is not surprising that especially the adjective, which also features comparing forms
in its inflectional morphology (comparative and superlative), exhibits compounds with
second elements such as -ahnlich 'similar', -artig 'in the manner of; lit. mannerly',
-formig ' in the form of; lit. formly', and -gleich 'same'. Further, there are a multitude
of suffixes with similar functional direction (cf. Ktihnhold, Putzer and Wellmann 1978:
330), not only -haft, which is above all used to join something likened to another or an
equivalent (comparable) thing. Beside tolpel-haft 'booby-like' can be found tolp-isch,
with a shortened base before a vowel-initial suffix, beside balladen-haft 'ballade-like'
also ballad-esk ' balladesque' (after a fore ign base). In the case of specialized meanings,
adjectival and nominal comparative formations can also be found w ith homo-, homoo-,
iso- and -aid.
Since judgments of comparison evidently belong to the essential mental-linguistic
performances of communication, it is not surprising that such a "Vielfalt von Vergleichssignalen und Vergleichsstrukturen entwickelt [worden ist]" [diversity of signals and
structures of comparison [has been] developed] (Erben 1988: 329), whereby the interaction of syntax, semantics and word-formation deserves particular interest. The central
position of word-formation between lexicon and syntax becomes particularly apparent
here and likewise the necessity of also making its contentual relations recognizable.
3. Conclu sion and prospects

The task of the theory of inhaltbezogene word-formation is to present the word-formation
paradigms of a language and to give insight into mental-linguistic methods of operation:
how that which is experienced, felt or thought is made nameable and communicable by
means of particular patterns of word -formation. In derivation, it is necessary to describe
the functional classes of affixes, which differ in form but are partially similar in function
(converging or competing with each other), and which can serve as indicators of particular categories (word-formation meanings). In compounding, it is the semantic patterns
(constructional meanings) which characterize the general structural schema of compounding on the level of a norm for designation of an individual language and constitute
conventionally preferred means of naming, highlighting particular aspects and features
of the denotatum. Ultimately the question arises how the availability of pa1iicular patterns of word -formation and derivational categories, which make possible and suggest
a particular mental-linguistic grasping of reality, affect the participants in a language
community.
Attempts to find a ''wirkungbezogene Wortbildungslehre" [effect-based theory of
word-formation] (Weisgerber 198 1: 51 - 54) have not thus far been convincing. As to the
controversy surrounding the accusativization of man, cf. Helbig (1974: 156- 159) and
Werlen (2002: 291- 293). But it is probably uncontroversial "that memory and perception
75
PA76
76
are affected by the availability of appropriate words and expressions" (Lyons 198 1: 305),
just as that neologisms can serve to influence opinions and to steer the behavior of
particular groups. Specialized terms also open "Hingst nicht ausnahmslos die Moglichkeit, wertneutral iiber fachliche Denotate zu sprechen, die bereits vor jeder sprachlichen
Gliederung klar umrissen sind. Oft klassifizieren sie die Wirklichkeit, indem sie unterschiedliche Inhaltsmerkmale eines Begriffs verbalisieren und so die Aufmerksamkeit auf
jeweils spezifische Aspekte lenken" [by far not without exception the possibility to neutrally speak about specialized denotata which are already clearly outlined before any
linguistic classification. They often classify reality in that they verbalize different contentual features of a concept and thus draw attention to specific respective aspects] (Reinart
2009: 498- 499). Regarding laboratory studies, cf. Knorr Cetina (2007 : 330- 331 ).
A convincing answer to the fundamental question, "Was machen wir mit der Sprache
und was macht sie m it uns?" [What do we do with language and what does language do
with us?] (Bieri 20 10: 215) is evidently only to be gained in an interdisciplinary manner,
in interaction between linguistics, psychology, cognitive science and translation studies.
At present it must already be noted on the basis of available studies that "the evidence
from research indicates that language does influence thought and perception of reality,
but language does not govern thought or reality" (Tohidian 2009: 7; further cf. Deutscher
20 10: 233-236).
4. References
Adrados, Francisco
1985
Der Ursprung der grammatischen Kategorien des lndoeuropiiischen. In: Bernfried
Schlerath (ed.), Grammatische Kategorien. Funktion und Geschichte, 1- 46. Wiesbaden:
Harrassowitz.
Baldinger, Kurt
1950
Kollektivsuffixe und Kollektivbegriff. Ein Beitrag zur Bedeutungslehre im Franzosischen
mit Beriicksichtigung der Mundarten. Berlin: Akademie Verlag.
Bieri, Peter
2010
Die Vielfalt des Verstehens. In: Christian Starck (ed.), Jahrbuch der Akademie der Wissenschaften zu Gottingen 2009, 215 - 231. Berlin/New York: de Gruyter.
Boas, Franz
1948
Race, Language and Culture. New York: MacMillan.
Brinkmann, Hennig
Die Zusammensetzung im Deutschen. In: Leonhard Lipka and Hartmut
1981 [1 956- 1957]
Gunther (eds.), Wortbildung, 187- 199. Darmstadt: Wissenschaftl iche Buchgesellschaft.
Buhler, Karl
1934
Sprachtheorie. Die D arste/lungsfanktion der Sprache. Jena: Fischer.
Coseriu, Eugenio
1972
Semantik und Grammatik. In: Hugo Moser (ed.), Neue Grammatiktheorien und ihre
Anwendung auf das heutige Deutsch, 77-89. Dtisseldorf: Schwann.
Coseriu, Eugenio
1977
Inhaltliche Wortbildungslehre. In: Herbert E. Brekle and Dieter Kastovsky (eds.), Perspektiven der Wortbildungsforschung. Beitri:ige zum Wuppertaler Wortbildungsko/loquium vom 9. - 10. Juli 1976. Anli:ij3lich des 70. Geburtstags von Hans M archand am 1.
Oktober 1977, 48- 61. Bonn: Bouvier.
Crystal, David
1991
The Cambridge Encyclopedia of Language. Cambridge: Cambridge University Press.
PA78
78

Kaltz, Barbara
2006
Die Entwicklung der inhaltbezogenen Grammatik in Deutschland. In: Sylvain Auroux,
Ernst Frideryk Konrad Koerner, Hans-Josef Niederehe and Kees Versteegh (eds.), Ge-
schichte der Sprachwissenschaften. Ein intern.ationales Handbuch zur Entwicklung der

Sprachforschung van den Anfangen bis zur Gegenwart. Vol. 3, 2166- 2178. Berlin/New
York: de Gruyter.
Kamper, Heidrun and Ludwig M. Eichinger (eds.)
2008
Sprache - Kognition - Kultur. Sprache zwischen mentaler Struktur und kultureller
Priigung. Berlin/New York: de Gruyter.
Knorr Cetina, Karin
2007
Neue Ansatze zur Wissenschafts- und Techniksoziologie. In: Rainer Schi.itzeichel (ed.),
Handbuch Wissenssoziologie und Wissensforschung, 328- 342. Konstanz: UVK Verlagsgesellschaft.
Ki.ihnhold, Ingeburg and Hans Wellmann
1973
Deutsche Wortbildung. Typen und Tendenzen in der Gegenwartssprache. Eine Bestandsaufnahme des lnstituts fiir deutsche Sprache. Forschungsstelle Innsbruck. Erster Hauptteil: Das Verb. Mit einer Einfahrung van Johannes Erben. Di.isseldorf: Schwann.
Ki.ihnhold, lngeburg, Oskar Putzer and Hans Wellmann
1978 Deutsche Wortbildung. Typen und Tendenzen in der Gegenwartssprache. Eine Bestandsaufnahme des Jnstituts far deutsche Sprache. Forschungsstelle Innsbruck. Drittter
Hauptteil: Das Adjektiv. Di.isseldorf: Schwann.
Lyons, John
1981
Language and Linguistics. An Introduction. Cambridge: Cambridge U niversity Press.
Maeder, Christoph and Achim Brosziewski
2007
Kognitive Anthropologie. In: Rainer Schi.itzeichel (ed.), Handbuch Wissenssoziologie
und Wissensforschung, 268-275. Konstanz: UVK Verlagsgesellschaft.
Motsch, Wolfgang
2004
Deutsche Wortbildung in Grundziigen. 2nd ed. Berlin/New York: de Gruyter.
Ortner, Hanspeter and Lorelies Ortner
1984
Zur Theorie und Praxis der Kompositaforschung. Mit einer ausfiihrlz'chen Bibliographie.
Tiilbingen: Narr.
Ortner, Lorelies, Hanspeter Ortner and Hans Wellmann
2007
Das Projekt "Deutsche Wortbildung". In: Heidrun Kamper and Ludwig M. Eichinger
(eds.), Sprach-Perspektiven. Germanistische Linguistik und das Ins titut far Deutsche
Sprache, 9 1-131. Tiibingen: Narr.
Paul, Hermann
1981 [1896]
Ober die Aufgabe der Wortbildungslehre. In: Leonhart Lipka and Hartmut Gunther (eds.), Wortbildung, 17- 35. Darmstadt: Wissenschaftliche Buchgesellschaft.
Reinart, Sylvia
2009
Kulturspezifik in der Fachiibersetzung. Die Bedeutung der Kulturkompetenz bei der
Translation fachsprachlicher undfachbezogener Texte. Berlin: Frank & Timme.
Schlogl , Rudolf
2004
Symbole in der Kommunikation. In: Rudolf Schlogl, Bernhard Giesen and Jiirgen Osterhammel (eds.), Die Wirklichkeit der Sy mbole. Grundlagen der Kommunikation in historischen und gegenwiirtigen Gesellschaften, 9- 38. Konstanz: UVK Verlagsgesellschaft.
Sharifian, Farzad
2009
On collective cognition and language. In: Hanna Pishwa (ed.), Language and Social
Cognition. Expression of the Social Mind, 163- 180. Berlin/New York: de Gruyter.
Soeffner, Hans-Georg
2004
Protosoziologische Oberlegungen zur Soziologie des Symbols und Ritua ls . In: Rudolf
Schlogl (ed.), Die Wirklichkeit der Symbole. Grundlagen der Kommunikation in historischen und gegenwartigen Gesellschaften, 41 - 72. Konstanz: UVK Verlagsgesellschaft.
PA79
79
6. Word-formation in onomasiology
Sylla, Bernhard Josef
2008
Leo Weisgerbers Sprachinhaltsforschung, ihre philosophischen Implikationen und ihr
Bezug zu Heidegger. Ph.D. dissertation, Universidade do Minho, Braga.
Tohidian, Iman
2009
Examining linguistic relativity hypothesis as one of the main views on the relationship
between language and thought. Journal of Psycholinguistic R esearch 38: 65- 74.
Tomasello, Michael
2008
Origins of Human Communication. Cambridge, MA: MIT Press.
Weisgerber, Leo
1962
Grundzuge der inhaltbezogenen Grammatik. 3rd ed. Diisseldorf: Schwann.
Weisgerber, Leo
1971
Die geistige Seite der Sprache und ihre Etforschung. Diisseldorf: Schwann.
Weisgerber, Leo
1981 [1964]
Vierstufige Wortbildungslehre. In: Leonhard Lipka and Hartmut Gunther (eds.) ,
Wortbildung, 36-54. Darmstadt: Wissenschaftliche Buchgesellschaft.
Wellmann, Hans
1975
Deutsche Wortbildung. Typ en und Tendenzen in der Gegenwartssprache. Eine Bestandsaufnahme des lnstituts far deutsche Sprache. Forschungsstelle Innsbruck. Zweiter
Hauptteil: Das Substantiv. Diisseldorf: Schwann.
Werlen, Iwar
2002
Sprachliche Relativiti:it. Eine problemorientierte Einfi.ihrung. Tiibingen/Basel: Francke.
Wildgen, Wolfgang
Kognitive Grammatik. Klassische Paradigmen und neue Persp ektiven. Berlin/New York:
2008
de Gruyter.
Johannes Erben, Bonn (Germany)
l.
2.
3.
4.
Introduction
Onomasio logists and their models
Outlook: European word-formation issues from an onomasiological perspective
References
Abstract
This article is structured in a historical order. Each section is dedicated to the onomasiological model of one linguist: Milos Dokulil, Jan Horeckj, Pavo! Stekauer, Andreas
Blank, Peter Koch, Joachim Grzega. Some models refer particularly to word-formation,
some are general models of the name-giving process. The article illustrates the various
theoretical works with examples from the original works and additional examples. Finally, the article will briefly discuss a number of onomasiological word-formation issues
with respect to European languages.
PA80
80
1. Introduction
In word-formation studies, approaches have always been rather analytical. An onomasiological perspective, in contrast, takes as its starting point an extralinguistic entity (a
concept) and the need of a speaker to denote an extralinguistic entity and looks for forms
that denote or may denote this concept. The term onomasiology was coined by Zauner
in 1902, but the onomasiological approach was already carried out earlier, notably by
D iez (187 5). The onomasiological perspective may be purely synchronic, or static, or it
may include diachronic, procedural aspects. If procedural aspects are included, the coinages of designations can be grouped into a) already existing names used in a new way
(i.e. semantic change), b) borrowings, c) new coinages based on indigenous linguistic
material. This latter is referred to as word-formation.
2. Onomasiologists and their models

2.1. M ilos Dokulil
Dokulil ( 1962, 1968, 1997) elaborated the first comprehensive onomasiological theory
of word-formation. Its influence reached mainly linguists from Central and Eastern European countries: Horecky (e.g., 1983, 1994, 1999), Buzassyova (e.g., 1974), Furdik (e.g.,
1993), Stekauer (e.g., 1996, 1997, 1998, 2001, 2005a, 2005b), Puzynina (1969), Grzegorczykowa (1979), Szymanek (1988), Waszakowa (1994), Nescimenko (1963, 1968),
Kubrjakova (e.g., 1977a, 1977b), Fleischer (1969), Polenz (1973) and Huke (1977).
Dokulil 's essential idea is the "onomasiological category", that is the different ways
of structuring the concept with respect to its expression in a language, in other words:
the basic conceptual structures that initiate the naming process in humans' mind. These
structures consist of two elements: As in traditional logic, the concept to be named is
first classified as a member of a certain conceptual group in an "onomasiological base"
or "onomasiological basis" (Cz. onomaziologicka baze), and this basis is then specified
by an "onomasiological mark", or "onomasiological feature" (Cz. onomaziologickf pfiznak); the terms basis and feature are the translations used by Horecky (cf. section 2.2),
base and mark are the translations used in the works by Stekauer and Grzega (cf. sections
2.3 and 2.6); Dokulil himself uses basis and mark in his English summary. For example,
in the words Cz. cernice, Cz. cernozem, Engl. black earth the onomasiological bases are
-ice 'something', -zem 'earth' and earth respectively. The onomasiological marks are
cern- 'black' and black. The base is always simple, the mark may be simple or complex.
A simple mark in the conceptual category of SUBSTANCE is quality (e.g., Cz. cernice
'black earth' or Engl. blackberry) or action without regard to its object (e.g., Cz. soudce
'judge' or Engl. singer). An example w ith a complex mark is Cz. knihvazac 'book
binder', where the object of action (books) is specified, too. The two components of the
mark, i.e. the determining and the determined one, may or may not be visible in the
expression.
In Dokuhl 's system, the basic types of onomasiological structure can be determined
according to the categorial nature (SUBSTANCE, ACTION, QUALITY, CIRCUMSTANCE) of the base and the determining part of the mark, called motive. For example,
PA82
82
triggered by an individual speaker 's concrete intent for creating a new naming unit. The
process includes five levels:
1. the conceptual level: The concept to be named is analyzed and conceptually categorized in the most general way (i.e. "SUBSTANCE, ACTION (with internal subdivision into ACTION PROPER, PROCESS, and STATE), QUALITY, and CONCOMITANT CIRCUM STANCE (for example, that of Place, Time, Manner, etc.)" (Stekauer
2001: 11)).
2. the semantic level: The semantic components are structured.
3. the onomasiological level: This is the level of naming in a more abstract sense. One
prominent semantic component is selected as the onomasiological base (representing
a class like agent, object, instrument, etc.) and another as the onomasiological mark
of thi s base (the mark, in turn, is divisible into a determining constituent - sometimes
showing a specifying and a specified element - and a determined constituent).
4. the onomatological level: This is the level of naming in a more concrete sense, featured by the morpheme-to-seme-assignment principle (MSAP). Concrete morphemes
are selected, allowing the speaker creativity within productivity constraints (cf. Stekauer et al. 2005).
5. the phonological level: The forms are combined according to morphological, phonotactic and suprasegmental rules.
Stekauer defines the following onomasiological types:
a)
Onomasiological Type 1, where all three constituents are present in the form onomasiological base, determining constituent, determined constituent - , e.g.,
[[p iano] [play]] -[er].
b)
c)
d)
e)
Onomasiological Type 2, where the determining constituent is not present on the

formal level, e.g., [lock] [pin], [play] -[er].
Onomasiological Type 3, where the determined (actional) constituent is not present
on the formal level, e.g., [pian(o)] [ist], [piano][man].
Onomasiological Type 4, where the onomasiological mark cannot be split into a
determining and a determined part, e.g., [un][happy] and [extra][large].
Onomasiological Type 5, traditionally known as "conversion".
In Stekauer's model, the classical juxtaposition of compounds and derivations is considered dispensable. Speakers simply transform the salient aspects and relations selected on
the onomasiological level into words by combining morphemes (either free or bound )
on the onomatological level. The relation of agent of an action, for instance, can be
verbalized in English by the morphemes -man, -er, -ist, -ant and others.
2.4. Andreas Blank

Blank obtained fame for his cognitive approach to historical semantics (cf. Blank l 997a),
but he also applied his theory to word-formation (Blank l 997b). Although this extension
did not draw very much attention, its basic elements shall be mentioned for the sake of
completeness. In Blank 's model, too, speakers' first step is the perception of salient
subconcepts of the concept to be named. One subconcept for which there already is a
PA83
name is then taken as the semantic basis for a coinage. The semantic difference between
the basic concept and the concept to be named is bridged by a "co-basis" in the form of
an affix or a second subconcept. These relations between basis, co-basis and the new
concept rest on the same associative principles as are semantic changes, namely contiguity (i.e. the cooccurrence of things or concepts in a specific context), contrast, and similarity - principles first formulated by Roudet ( 1921 ). Blank distinguishes various types of
compounding and affixation. In section 5 of his paper he mentions, among others, the
type "similarity/contrast within a category + conceptual contiguity'', a type which is
"characterized by the similarity between a prototype and a peripheral member as well
as by conceptual contiguity", such as in Fr. wagon-lit 'sleeping car; lit. car-bed', It.
autostrada 'freeway; lit. auto-street', Sp. maquina de escribir 'type-writer; lit. machine
of writing', and the type "similarity/contrast within a category plus metaphorical similarity'', similar to the preceding one, but in which the determinant is based on a metaphor,
such as in ModE frogman.
2.5. Peter Koch

Koch has proposed an overall model of word -finding processes based on an interplay of
three dimensions, or axes (cf. Koch 2001: 19, 2002: 1159 ff.): a) the morphological axis
(including "zero" [= semantic change], conversion, suffixation, prefixation, composition
and others), b) the cognitive-associative, or semantic, axis (including various forms of
contiguity, similarity and contrast), and c) the stratification axis (differentiating between
foreign models and native models). Koch's definitions of the word-fonnation types are
apparently traditional ones. Like semantic shifts, word-formations, too, can be activated
by the same cognitive-associative relations, e.g., cab driver only shows contiguity,
screwdriver also shows similarity (the instrument is seen like a person who does something).
Such an overall model avoids the neglect of compounds triggered by foreign influences: loan translations, loan renditions and pseudo-loans (including "neoclassical compounds"). A loan translation is created on the onomatological level: the single elements
of a foreign coinage are translated, e.g., Ger. Ubermensch > Engl. super man; Engl.
skyscraper > Fr. gratte-ciel. Loan renditions are created on the onomasiological level:
the name-giver seems to pass the semantic level, looking at a foreign language on the
way to the onomatological level, and returns to the native language on the onomatological level, where the image, or motive, of a foreign coinage is now rendered with native
material, e.g., Engl. skyscraper > Ger. Wolken-kratzer lit. 'cloud-scraper'. The third
crossbreed of word-formation and borrowing are lexical pseudo-loans, i.e. words that
are made of foreign material, but never existed as such in the assumed donor language.
So-called neoclassical compounds, such as Engl. microphone, are one form of such
pseudo-loans, which have become an essential process in the creation of technical terms
all over Europe and the Americas. Today, the prestige of English attracts many Europeans to form pseudo-Anglicisms. Here, it seems as if the name-giver person reaches the
onomatological level and there takes material from a foreign language, which then undergoes the usual adaptation changes on the morphonological level. However, with calques
it is not always easy to decide whether the coinage was really modelled on a foreign
term or whether it represents an independent, albeit parallel construction.
83
PA84
84
2.6. Joachi m Grzega

In Grzega (2004), an overall model of both processes and forces of lexemic change was
established (cf. also the preliminary study in Grzega 2002a): the Cognitive and Social
Model for Onomasiological Studies (CoSMOS) (cf. also Grzega 2007, 2009; 2012b).
CoSMOS incorporates and further elaborates the models of Stekauer, Blank and Koch.
Moreover, it takes into account psycholinguistic observations, too (e.g., Mangold-Allwinn et al. 1995; Dietrich 2002). It is presented as a linguistic sign model (a shorter
textbook version of the model is given in Grzega and Schoner 2007: 53 ff.).
lfeaturc anaJ~ sis
I local rca1urcsl
onomasiological
le\CI
abstract
concrete
pltorretic re11/iv1tio11
Imorpl1onolog11::1I lc\'cl J
linguistic
(languege-speclflc)
extralinguistic
Fig. 6.1: Visualization of CoSMOS
The following paragraphs describe how to read the model, illustrating the single phases
with the help of terms for the concept BENGAL TIGER.
1. The starting-point is either a thought, or the concept, or a specific referent in context,
or a type of referent. The term context is to cover the speaker-hearer situation, the
type of discourse, the communicative goal, the syntactical context (this comes close
to what Nerlich and Clarke 1992 call "meta-semantic expert system").
Example: I have an animal in front of me and I need to give it a neutral-sounding,
but catchy name.
2. The speaker classifies the referent through processing its more basic, "global" and its
more specific, "local" traits. The speaker then categorizes the referent by using some
kind of mental checklist for the absence and presence of specific features and by
comparing the overall image of the referent with other images already in the mind.
This is the perceptual level.
Example: The animal is a subspecies of the tiger family. It is particularly large and
some exemplars show a completely white background of their coat. In France, they
call it tigre royal.
3. If the (concrete) referent can be identified as the member of a familiar (abstract)
concept, the speaker may resort to an already existing designation or create a new
designation, sometimes more, sometimes less consciously. The decision wi ll be based
PA85
on a cost-benefit-analysis in which the speaker weighs the goals of the designation
and the utterance it is embedded in: do I want to sound like the hearer, do I want to
sound different from others, do I want to sound primitive, elaborate, polite, impolite,
do I want to simply take the first word that comes to my mind? The cost-benefit
analysis can be described as "linguistic economy". If the speaker does not want to
coin an innovation consciously, but doesn't have the best word at hand, a word whose
semantics is close might even be used on the spur of the moment, in the hope that
the hearer may infer the suitable meaning from the context of the interaction. In the
case of intentional, conscious innovation the speaker subsequently goes through several levels of a word-finding, or name-giving, process.
Example: What can I call this animal (to achieve my communicative goals)?
Actually, there seems to be a supra-force, or super-force, that could be tenned "pressure of acceptance" or rather a supra-goal, or super-goal, of "Get accepted", either
because the group you want to be accepted by recognizes your style as similar to
theirs and different from others or because the group you want to be accepted by
admires your being different from them. The less infonnal and the less private a
communicative situation is, the more important becomes this supra-goal. Maxims of
differentiation may lead to systemic language change. Enduring linguistic change can
either be planned (cf., e.g., the contributions in Coulmas 1989), or it simply occurs,
as a by-product.
The coinage of a new designation can be incited by various, potentially concurrent,
forces. Early attempts of lists of forces were unsystematic and sometimes did not
neatly separate processes from forces (cf. the overview in Grzega 2004: 163-165).
Blank was the first to establish a comprehensive empirically based, cognitive-linguistic catalog for forces triggering semantic change (1997a: 345ff., 1999: 70ff.). Grzega's works (cf. 2002a and 2004) are the first that strive for an empirically based,
comprehensive and systematic catalog of forces for all types of lexical change. These
factors are positioned on a conscious-subconscious continuum using the "word death"
metaphor, where the gradual subconscious loss of a word is compared to "natural
(designation) death" and where the conscious avoidance of a word is compared to
"(designation) murder" (these two extremes embrace several intermediate degrees)
(cf. Grzega 2004: 272, 2007: 22). Among the forces of the CoSMOS catalog are
cognitive, sociocultural and linguistic reasons that can also be linked with Grice 's
conversational maxims (cf. Grzega 2004: 268, 2007: 21 f.).
4. The next phase in the name-giving process is again an analysis of the specific traits
of the concept (= feature analysis) - with a focus on the local traits. This step is
skipped if the speaker simply borrows a word from a foreign language or variety that
is used for the concept in question; it can also be skipped if the speaker simply picks
an already existing designation and reduces its form.
Example: This species of the tiger family typically lives in India, Bengal, and neighboring countries. In contrast to other species of the wider area, it is particularly large
and some exemplars showing a completely white background of their coat. In France
they call it tigre royal.
5. The speaker will then highlight one or two features as a basis for the designation.
Stekauer refers to this as "naming in a more abstract sense". The designation motives
(or iconemes or, as Alinei 1995 said, iconyms) are commonly based on similarity,
contrast, partiality and contiguity/contact relations. These relations can affect the linguistic side as well as the extralinguistic side and the abstract as well as the concrete
85
PA86
86
side. In Stekauer's model this is called "onomasiological leve l". Here again, the speaker keeps in mind the extralinguistic context. The concrete associations can or cannot
be incited by a model, which may be of a speaker's own idiom or a foreign idiom.
Example: Variant 1: This tiger typically lives in Bengal. Variant 2: This tiger typically
lives in India. Variant 3: This member of the tiger family is comparatively large that
it seems like the "king" of the tiger family. Variant 4: Exemplars of this tiger can be
white. Variant 5: The corresponding French term tigre royal literally means 'royal
tiger'.
6. The next step is what Stekauer calls "onomatological level" ("naming in a more
concrete sense"). Here, concrete morphemes are selected. The speaker either reduces
the form of an already existing lexeme for the concept or has decided to create a new
one, which can be achieved through several types of processes, founded either on the
speaker's own idiom or on a model from a foreign idiom or on no mode l at all . Due
to the already existing words in a language, the perceptual level, feature analysis
and onomasiological level may lead to the preference of a certain process on the
onomatological level (e.g., a preference of using -er with agent nouns in English),
but such model-like concept-to-process matching is not a general, let alone universal
or diachronically rigid rule.
Example: Variant 1 (English): {Bengal} + {tiger}. Variant 2 (Spanish): {tigre} +
{India} + {relatedness} . Variant 3 (German): {Konig 'king'} + {Tiger}. Variant 4
(Russian): {bel- 'white'} +{quality morpheme/adjectival ending} + {tigr}. Variant 5
(English): {royal} + {tiger}
The catalog of processes embraces these:
1.
adoption of an already existing lexeme

a) of the speaker's own idiom (semantic change)
(N.B.: This also includes the phenomenon traditionally known as "semantic
Joan", the copy of a certain polysemy found in a donor language.)
(i)
(ii)
(iii)
(iv)
(v)
(vi)
(vi i)
metaphor ("similar-to" relation)

metonymy ("neighbor-of" relation)
synecdoche ("part-of" relation)
generalization and specialization ("kind-of" relation)
cohyponymic transfer ("sibling-of" relation)
antiphrasis and auto-antonymy ("contrast-to" relation)
conceptual recategorization
b) from a foreign idiom, including dialects and jargons (loanword)

(i)
(ii)
(iii)
"true loan"
"incomplete loan" (traditionally called "morphological pseudo-loan")
"mis-loan" (i.e. fo lk-etymological formal change of a loan, fo lk-etymological semantic extension due to an only phonetically similar loan)
(iv) "creative loan" (traditionally called " lexical pseudo-loan")
2. syntactical recategorization (traditionally called "zero-derivation" and "conversion")
3. composition (lato sensu, i.e. the combination of existing morphemes) (N.B.: This
includes what is traditionally called "compounds" and "derivations". This also includes w hat is traditionally called " loan translations" and " loan renditions".)
PA87
(i)
"complete complex structure" (complex composites, i.e. complete determinative composites with a base and a mark, consisting of a determining
component and a determined component; e.g., Engl. pianoplayer, sky-
(ii)
"incomplete complex structure 1" (composites with absence of determining component of the mark; e.g., Engl. pianist, pianoman)
"incomplete complex structure 2" (composites with absence of determined
component of the mark; e.g., Engl. player)
"incomplete complex structure B" (composites with absence of the base;
e.g., Engl. redbreast)
"simplex structure" (simplex composites, no action-determinated relationship between the elements; e.g., Engl. blackberry, vice chair)
"copulative structure" (copulatives composites; e.g., Engl. actor-singer,
scraper)
(iii)
(iv)
(v)
(vi)
deaf-mute)
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
(N .B.: The variety of associations and relations that can be expressed by

just combining two word(stem)s was already underlined by Whitney 1875:
121 - but in a way that rather reminds one of a generative approach.)
blendings (i.e. overlapping of already existing lexemes, sometimes folk-etymologically; e.g., Engl. smog +-- smoke +fog 'fog consisting of smoke', Ger. blass
erstaunt ' highly astonished; lit. pale-astonished' +-- blass 'pale' + bass erstaunt
'hi ghly astoni shed; lit. we ll astonished', with an otherwise obsolete word for
'good ')
back-derivation (e.g., Engl. to edit+-- editor)
reduplication (full or partial) (e.g., Engl. bye-bye, wishy-washy)
morpholog ical alteration (e.g., number change as in peoplesg 'several human beings' <p eopleP1 'nation', gender change as in It. melo 'apple-tree'< mela 'apple')
wordplaying
phonetic-prosodic alteration (e.g., stress shift in Engl. import vs. import)
graphic alteration (e.g., Engl. discrete vs. discreet)
phraseologism (e.g., Engl. the red thread 'the recurrent theme')
root creation (including onomatopoetic words, i. e. words that verbalize the
sounds the concept makes, e.g., Engl. cuckoo = Ger. Kuckuck = Sp. cuco =
Hung. kakukk, etc., and expressive words, i.e. words that verbalize the sounds
that a concept is metaphorically related with, e.g., Engl. to peck = Ger. picken =
Sp. picar = Hung. csipegetni)
clarifying compounds/composites (i.e. tautological compounds/composites; e.g.,
Engl. hound dog instead of just hound)
formal shortening of already existing designations
a) morpheme deletion (ellipsis; e.g., Engl. turkey +-- turkey-hen, Fr. dinde +geline d'Inde 'hen from India')
b) morpheme shortening (clipping; e.g., Engl. Ger. Du. Fr. It. prof 'professor')
c) morpheme symbolization (acronyms, incl. alphabetisms, and short-forms;
e .g., Engl. laser +-- ight ~mplijication by _limulated ~mission of r_adiation)
Processes may be combined.
7. Then, the word is provided with a fixed form-content relation and ce11ain grammatical
traits - the sign is completed.
87
PA89
that are merged in the model presented in Grzega (2004). In the first approach the main
point is the differentiation between a step in the name-finding process that can be termed
naming in a more abstract sense, where an onomasiological base and mark are selected,
and a step that can be termed naming in a more concrete sense, where the concrete
morphemes are selected. In the second approach the associations related to extralinguistic, formal and mental types of similarity, contrast and contiguity are focussed on.
Dokulil, Horecky, Blank and Koch understand "compounding" - a bit unfortunately in its traditional formal definition and not in a definition that follows from the overall
model; therefore, they view this process differently from affixation. Stekauer 's wordformation model abstains from making this distinction at all (the model only knows
" onomasiological types", excluding any sort of non-regular name-giving processes, such
as b lends, semantic changes and loans). The overall model in Grzega (2004) does not
distinguish between the traditional forms of compounding, affixation and phrases either,
but it uses the term "composites" in order to distinguish the concatenation of morphemes
from other types of name-giving processes (such as blending, clipping, semantic change,
borrowing).
In comparison to analytical models, all onomasiological models offer the advantage
of carrying out contrastive studies not only on the level of the sign, but also on previous
cognitive steps.
3. Outlook: Eu ropean word-format ion issues from an

onomasiological perspective
As already demonstrated with the designations for the Bengal tiger in several European
languages, an onomasiological approach enables one to overcome a purely form-driven
word/phrase-distinction, which in many languages would exclude many words that functionally work not as two units, but as one (especially in Romance languages, which
favor a construction "base + specified element of the mark in the form of a preposition
+ specifying element of the mark", and in Slavic languages, which favor a pattern in the
order "base-morpheme + specifying element of the mark + specified element of the mark
in the form of an adjectival suffix" or in reverse order). Second, comparisons before the
level of the sign, particularly the onomasiological and the onomatological levels, are
possible as well. Only some examples can be given here for illustration.
A culturally interesting topic may be the study of the frequency of suppletive vs.
morphologically consociated patterns. Such opposites were first studied thoroughly by
Osthoff (1899). Among other conceptual fields, he analyzed kinship terms and pointed
out the replacement of the pair /rate 'brother ' and sora 'sister ' in Northern Italy by /rate
and frata. Likewise, we can mention the Standard Spanish pair hermano 'brother ' and
hermana 'sister '. Similar cases are It. jiglio 'son' and jiglia 'daughter ', Sp. hijo 'son'
and hija 'daughter', It. zio, Sp. tio 'uncle' and It. zia, Sp. tia 'aunt ' . A contrastive analysis
of the word-pairs for 'father : mother ', 'brother : sister', 'son : daughter ', 'uncle : aunt'
and ' nephew : niece' in a selection of European languages shows that Spanish has 5 and
Italian and Greek 4 consociative pairs, while Polish and Rumanian have 2, English,
Swedish, French, Finnish and Russian 1 and German, Dutch and Czech 0 consociative
pairs. With such comparisons a consociation index may be established.
89
PA90
90
Another topic of contrastive word-formation research from an onomasiological perspective can be the degree of technical terms formed with Neo-Latin morphemes in
comparison to formation with indigenous material. (Neo-)Latin morphemes that have
adapted equivalents all over Europe including Russia are inter-, intra-, trans-, multi-,
maxi-, mini-, auto-, -ismus, -logia, -graphia, etc. Equivalents of Engl. international are
found in all European languages but Icelandic, Irish and Greek. In some languages, the
internationalism has to share the stylistic ranges with a synonym formed with indigenous
components, e.g., Ru. meidunarodnyj, Dan. mellemfolkelig, Cz. mezinarodni, Lith. starptautisk, and Hung. nemzetkozi. Usually, the internationalism is the stylistically more
marked item. The prominence of internationalisms in contrast to formations with indigenous morphemes varies from country to country. Internationalisms seem most prominent in France. Despite manifold diastratic variation, it seems as if the trend to form
words with indigenous material is, on the whole, weakest in the Romance and in the
English-speaking countries and strongest in Iceland, Finland, Eastern Central Europe and
South Eastern Europe (cf. Grzega 2012a: 90 f.). Onomasiological word-formation studies
may be especially interesting for hierarchically shuctured conceptual fields.
Contrastive word-formation onomasiology need not be restricted to nominal categories. Comrie (2006) illustrates that in causative/anti-causative pairs (e.g., 'to bring to an
end'/ 'to come to an end') the anti -causative is often derived from the causative in European languages, while globally it is mostly the other way around. A prominent exception
is English with its many labile pairs (i.e. forms that serve both as causative and anticausative).
A final topic could be the investigation of less frequent word-formation processes
such as acronyms, blends, and clippings. What is their real prominence in specific registers (such as expert language and youth language)? Some minor word-formation processes seem to be restricted to a certain substandard - interestingly, in several languages (e.g.,
back slang, known as verlan in France and fotrovacki in the South Slavic languages,
e.g., 'head': Engl. deeache for head, Fr. teute for tete and Croat. vugla for glava).
The forces influencing the decision when a new word is coined and what type of
process is used are manifold. A complete catalog of such forces has been suggested by
Grzega (2002b, 2004). An experiment on the preference of word-formation patterns with
new words was carried out for English and Slovak by Stekauer et al. (2005) and Kortvelyessy (2009).
4. References
Alinei, Mario
1995
Theoretical aspects of lexical motivation. Svenska Landsmal och Svenskt Folk/iv
118(32 1): 1-10.
Bauer, Laurie
ls there a class of neoclassical compounds, and if so is it productive? Linguistics 36:
1998
403-422.
Blank, Andreas
1997a Prinzipien des lexikalischen Bedeutungswandels am Beispie/ der romanischen Sprachen.
Tubingen: Niemeyer.
PA91
Bla nk, Andreas
1997b Outlines of a cognitive approach to word-formation. In: Proceedings of the 16 1" i nternational Congress of Linguists, Paper No. 029 1 [s.pag .]. Oxford: Pergamon.
Blank, Andreas
1999
Why do new meanings occur? A cognitive typology of the motivations for lexical semantic change. In: Andreas Blank and Peter Koch (eds.), Historical Semantics and
Cognition, 6 1-90. Berlin/New York: Mouton de Gruyter.
Buzassyova, Klara
1974
Semanticka struktura slovenskjch deverbativ. Bratislava: Veda.
Comrie, Bernard
2006
Transitivity pairs, markedness, and diachronic stability. Linguistics 44: 302- 318.
Coulmas, Florian (ed.)
1989
Language Adaptation. Cambridge: Cambridge University Press.
Dietrich, Rainer
2002
Psycholinguistik. Stuttgart/Weimar: Metzler.
Diez, Friedrich
1875
Romanische Wortschopfung. Bonn: Weber.
Dokulil, Milos
1962
Tvo feni slov v cdtine. Vol. I: Teorie odvozovani slov. Prague: Nakladatelstv i Ceskoslovenske Akademie Ved.
Dokulil, Milos
1968
Zur Theorie der Wortbildungslehre. Wissenschaftliche Zeitschrift der Karl Marx-Universitiit Leipzig. Gesellschafts- und sprachw issenschaftliche Reihe 17: 203- 211.
Dokulil, Milos
1997
The Prague School's theoretical and methodological contribution to "word-formation"
(Derivology) . In: Jarmila Panevova and Zdena Skoumalova (eds.), Obsah - vyraz ryznam. Milosi Dokulilovi k 85. narozeninam, 179- 2 10. Praha: FF UK.
Fleischer, Wolfgang
1969
Wortbildung der deutschen Gegenwartssprache. Leipzig: Bibliograph isches Tnstitut.
Furdik, Juraj
1993
Slovotvorna motivacia a j ej jazy kove funkcie. Levoca: Modr)r Peter.
Grzega, Joachim
2002a Some thoughts on a cognitive onomasiological approach to word-formation w ith special
reference to English. Onomasiology Online 3.4. http://www.onomasiology.de [last access
9 Sept 20 14].
Grzega, Joachim
2002b Some aspects of modern diachronic onomasiology. Linguistics 40: 1021- 1045 .
Grzega, Joachim
2004
Bezeichnungswandel: Wie, Warum, Wozu? Ein Beitrag zur englischen und allgemeinen
Onomasiologie. Heidelberg: Winter.
Grzega, Joachim
2007
Summary, supplement and index for Grzega, Bezeichnungswandel, 2004. Onomasiology
Online 8 : 18- 196. http://www.onomasiology.de [last access 9 Sept 2014].
Grzega, Joachim
Compounding from an onomasiolog ical perspective. In: Rochelle Lieber and Pavo ! Stek2009
auer (eds.), The Oxford Handbook of Compounding, 217-232. Oxford: Oxford U niversity Press .
Grzega, Joachim
20 12a Europas Sprachen und Kulturen im Wandel der Zeit. Eine Entdeckungsreise. Tiibingen:
Stauffenburg.
91
PA92
92

Grzega, Joaclhim
2012b Lexical semantic variables. In: Juan M. Hernandez-Campoy and J. Camilo Conde-Silvestre (eds.), Handbook of Historical Sociolinguistics, 271-292. Oxford: Wiley-Blackwell.
Grzega, Joaclhim and Marion Schaner
2007
English and General Historical Lexicology. Materials for Onomasiology Seminars.
Eichstatt: Katholische Universitat. http://www.onomasiology.de [last access 9 Sept
2014].
Grzegorczykowa, Renata
Za1ys slowotw6rstwa polskiego. Slowotw6rstwo opisowe. Warszawa: PWN.
1979
Horeccy, Jan
1983
Vyvin a te6ria jazyka. Bratislava: SPN.
Horecky, Jan
1994
Semantics of Derived Words . Prefov: Acta Facultatis Philosophicae Universitatis Safarikanae.
Horecky, Jan
1999
Onomaziologicka interpretacia tvorenia slov. Slovo a slovesnost 60: 6- 12.
Huke, Ivana
Die Wortbildungstheorie von Milos Dokulil. Ph.D. dissertation, Universitat Gief3en.
1977
Koch, Peter
2001
Bedeutungswandel und Bezeichnungswandel: Von der kognitiven Semasiologie zur kognitiven Onomasiologie. Zeitschriftfar Literaturwissenschaft und Linguistik 121: 7- 36.
Koch, Peter
Lexical typology from a cognitive and linguistic point of view. In: D. Alan Cruse, Franz
2002
Hundsnurscher, Michael Job and Peter Rolf Lutzeier (eds.), Lexikologie. Ein internationales Handbuch zur Natur und Struktur von Wortern und Wortschiitzen . Vol. 1, 1142Ll 78. Berlin/New York: de Gruyter.
Kortvelyessy, Livia
2009
Productivity and creativity in word-formation: A sociolinguistic perspective. Onomasiology Online 10: 1- 22. http://www.onomasiology.de [last access 9 Sept 2014].
Kubrjakova, E lena S.
1977a S lovoobrazovatel'naja nominacija. In: Boris A. Serebrennikov and Anna A. Ufimceva
(eds.), Jazykovaja nominacija: ObsCie voprosy, 55-98. Moskva: Nauka.
Kubrjakova, E lena S.
1977b Teorija nominacii i slovoobrazovanie. In: Boris A. Serebrennikov and Anna A. Ufimceva (eds.), Jazykovaja nominacija: Vidy naimenovanij, 222-303. Moskva: Nauka.
Mangold-Allwinn, Roland, Stefan Baratelli, Markus Kiefer and Hans-Gerhard Koelbing
1995
Worter far Dinge. Von jlexiblen Konzepten zu Benennungen. Opladen: Westdeutscher
Verlag.
Nerlich, Brigitte and David D. Clarke
1992
Outline of a model for semantic change. In: Gunter Kellermann and Michael D. Morrissey (eds.), Diachrony with Synchrony, 125- 141. Frankfurt/M.: Lang.
Nescimenko, Galina
1963
Zakonomernosti slovoobrazovanija, semantiki i upotreblenija suscestvitel'nyx s suffiksami sub"ektivnoj ocenki v sovremennom ceskom jazyke. In: Aleksandra G. Sirokova
(ed.), Jssledovanija po cciskomu jazyku. Voprosy slovoobrazovanija i grammatiki, 105158. Moskva: Nauka.
Nescimenko, Galina
lstorija imennogo slovoobrazovanija v cciskom literaturnom jazyke konca XVIII-XX
1968
vekov. Moskva: Nauka.
Osthoff, Hermann
1899
Vom Suppletivwesen der indogermanischen Sprachen . Heidelberg: Horning.
PA93
6. Word-formation in onomasi ology
Polenz, Peter von
1973
Synpleremik I. Wortbildung. In: Hans P. Althaus, Helmut Henne and Herbert E. Wiegand
(eds.), Lexikon der germanistischen Linguistik, 145-163. Ti.ibingen: Niemeyer.
Puzynina, Jadwiga
1969
Nazwy czynnofri we wsp 6lczesnym j fzyku po/skim (slowotw6rstwo, semantyka, skladnia). Warszawa: Wydawnictwo UW.
Roudet, Leonce
192 1
Sur la classification psychologique des changements semantiques. Journal de Psychologie 18: 676- 692 .
Stekauer, Pavo!
1996
A Th eory of Conversion in English. Frankfurt/M.: Lang.
Stekauer, Pavo!
On some issues of blending in English word formation. Linguistica Pragensia 7: 261997
35.
Stekauer, Pavo!
1998
A n Onomasiological Theot)I of Word-Formation in English. Amsterdam/Philadelphia:
Benj amins.
Stekauer, Pavo!
English Word-Formation. A Histo1y of Research (1960- 1995) . Ti.ibingen: Narr.
2000
Stekauer, Pavo)
200 1 Fundame ntal principles of an onomasiological theory of English word-formation. Onomasiology Online 2.4. http://www.onomasiology.de [last access 9 Sept 20 14].
Stekauer, Pavol
2005a Onomasiological approach to word-formation. In: Pavo! Stekauer and Rochelle Lieber
(eds.), Handbook of Word-Formation , 207- 232 . Dordrecht: Springer.
Stekauer, Pavo!
2005 b Compounding and affixation: Any diffe rence? In: Wolfgang U . Dressler, Dieter Kastovsky, Oskar E. Pfe iffer and Franz Rainer (eds.), Morphology and Its Demarcations. Selected Papers from the 11 111 Morphology Meeting, 15 1- 159. Amsterdam/Philadelphia:
Benjamins.
Stekauer, Pavo !, Do n C hapman, Slavka Tomascikova and Stefan Franko
2005
Word-Formation as creativ ity w ithin productivity constraints: Sociolinguistic evidence.
Onomasiology Online 6: 1- 55. http://www.onomasiology.de [last access 9 Sept 20 14].
Szymanek , Bogdan
1988
Categories and Categorization in Morphology. Lublin : Katolicki Uniwersytet Lubelski.
Waszakowa, Krystyna
1994
Slowotw6rstwo wsp 6lczesnego } fZyka p olskiego - rzeczowniki sufiksalne obce. Warszawa: Wydawnictwa Uniwersytetu Warszawskiego.
W hitney, William D .
Th e Life and Growth of Language. 61h ed. London: Kegan Paul, Trenc h and Co.
1875
Zauner, Ado lf
1902
Die romanischen Namen der Korperteile. Eine onomasiologische Studie. Erla ngen:
Junge.
Joachim Grzega, Eichstiitt and Pappenheim (Germany)
93
PA94
94
7. Word-formation in generative grammar

l.
2.
3.
4.
5.
6.
7.
Introductio n
Early generative grammar
The beginnings of generative morphology as a separate discipline
The heyday
Reactions to classic generative morphology
Conclusion
References
Abstract
This article traces the treatment of word-formation in generative grammar from the early
days of the theory in which word-formation, if discussed at all, is treated either as a
matter of syntax or phonology to the present. We look at the beginnings of the theory in
the work of Halle (1973) and Jackendoff (1975) and its heyday in the 1970's and 1980's
with the work of Aronoff (1976), Lieber (1980), Williams (1981), and Selkirk (1982). Of
special interest is the placement of rules of word-formation with respect to other components of the grammar as well as the issue of lexical integrity. Later developments, such
as the return to a syntactic theory of word-formation and the fragm enting of generative
morphology into a variety of different approaches are covered as well.
1. Introduction
We use the term generative grammar to refer broadly to the various theoretical frameworks that have arisen from the work of Noam Chomsky, Morris Ha lle, and their students in the 1950's and 1960's (Chomsky 1957, 1968, 1981 , 1995; Chomsky and Halle
1968). Generative grammar has gone through many phases in its more than six decades
of existence. Here we w ill not provide a general overview of the framework, as this can
be found elsewhere (e.g., Newmeyer 1986). What is critical for the purposes of this
article are the philosophical and psycho logical foundations of generative grammar which
have remained constant over the history of the theory. Those foundational assumptions
are: first, that linguistics is a fundamentally mentalist enterprise in the sense that the
linguist mus t attempt to model the native speaker's mental representation of the grammar; second, that the grammar must be able to generate all and only acceptable utterances
in the language under study; and third, that all humans are born with some set of linguistic universals, by which we mean mental constructs that allow them to acquire language,
and indeed that determine the form of the grammars that they acquire. The formalization
of rules, the organization of the grammar, and the assumed nature of linguistic universals
have all changed in various ways over the years, but these foundational assumptions
have not.
In its early years, generative grammar was largely concerned w ith syntax and phonology, which can to some extent be interpreted as a natural reaction to the American
Structuralist tradition that preceded it. The latter tradition was at least partly built on the
PA95
95
attempt in the early part of the twentieth century to document what we would now
call understudied and endangered languages, and was quite heavily concerned with the
description of morphology and phonology, and much less so with syntax. In contrast, in
its early years generative grammar had very little to say directly about morphology, and
especially about word-formation, by which we mean here derivation, compounding, and
conversion.
Here we will focus on the treatment of word-fonnation in generative grammar. We
will approach our subject from two directions. Our overall treatment of generative morphology will be historical, looking at the treatment of word-formation in the earliest
works on generative grammar, at the rise of generative morphology as a discipline in its
own right, and at the many ways in which generative morphology has subsequently
developed. But we will also treat several themes that have been prominent since the
inception of generative grammar, among them the status of the morpheme as a linguistic
construct, the nature of rules and the organization of the mental lexicon.
2. Early generative grammar

Although early work in generative grammar did not recognize word-formation as a topic
of study per se, word-formation did figure obliquely in both syntactic and phonological
theory. We will illustrate this here by looking briefly at two representative works that
deal with di fferent aspects of derivation and compounding. On the syntactic side, we
take up Lees's (1960) The grammar of English nominalizations and on the phonological
side Chomsky and Halle's (1968) Sound Pattern of English (henceforth, SPE).
2.1. Syntactic approaches

Lees's The grammar of English nominalizations is typical of the treatment of derivation
and compound ing in early generative grammar. In the earliest versions of generative
theory, the gramm ar consisted of a set of phrase-structure rules that determine basic
constituent structure and a set of transformational rules that modify initial phrase structure trees in various ways, including the addition, deletion, and reordering of morphemes
and indeed whole constituents. Lees used this basic framework to explore the relationship between nominal compounds like pronghorn (a kind of sheep) and the sentential
underlying structure from which they were presumed to derive, in this case the sentences
The sheep has a horn. The horn is like a prong. In his derivation, a generalized transformation first combines the two base sentences and then a series of simple transformations
reduces that structure to the compound, as follows (Lees 1960: 156):
(1)
a. The sheep has a horn. The horn is like a prong.

b. The sheep has a horn which is like a prong. ~
c. The sheep has a horn like a prong. ~
d. . .. sheep with a horn like a prong ... ~
e. . .. sheep with a pronghorn ... ~
f. ... pronghorn ...
PA96
96
The actual compound formation transformation which Lees calls NPN effects the change
from (1 d) to (1 e), deleting a central preposition or verb, reordering the nouns on either
side, and inducing the distinctive left stress pattern typical of English compounds; NPN
in its simplest form is stated as in (2):
A further process of ellipsis takes us from ( 1e) to the compound in (1 f).

Lees's analysis of compounds obviously uses the theoretical devices at the disposal
of early generative theory, but it also reflects the major question with which generative
theory was preoccupied at that moment, specifically the derivational relationship between semantically "equivalent" sentences or complex words. It is important to keep in
mind that in the first blush of theoretical exuberance, the power of transformational rules
was not yet an issue for generative grammar.
2.2. Phonological approaches

Chomsky and Halle's SPE is a comprehensive effort to apply the principles of generative
grammar to phonology and has little explicit to say about morphology. But as a great
deal of English phonology is specific to particular derivational patterns, SPE of necessity
requires some implicit assumptions about complex words. Several sorts of derivational
complexity are recognized in SPE and are coded in lexical representations with three
different sorts of boundaries. The # symbol is used to represent word boundaries, but it
is also deployed to represent the boundary between a root and a so-called "neutral" affix,
by which Chomsky and Halle mean an affix that does not affect the stress placement of
its base (-like, -able, -ish, -ness, etc.). Non-neutral affixes, that is, those which affect
stress like -ity, -ic, -al, and -ive are signalled by a + boundary. Finally, Latinate fonnatives like ceive, pel, mil, scribe, and so on are separated from the prefixes with which
they occur (as in, for example, deceive, repel, transmit, prescribe) by a special boundary
represented as =. Chomsky and Halle are not generally concerned with how affixes and
bases are put together, but passing remarks give an idea of how complex words might
be constructed: neutral affixes are "syntactically distinguished" (1968: 86), being "assigned to a word by a grammatical transformation [ ... ]". Non-neutral affixes, on the
other hand, are "largely, internal to the lexicon."
3. The beginnings of generative morphology as a separate

discipline
The treatment of word-formation as a special subdiscipline of generative morphology is
generally said to have its origins in Chomsky's ( 1970) Remarks on nominalization. In
this seminal article, Chomsky notes that whereas a gerundive form exists for any given
verb (3), derived nominals are far more idiosyncratic in terms both of the form of the
derived noun and the productivity of the process with which the noun is formed (4).
PA97
(3)
a.
b.
John's refusing the offer

John 's criticizing the book
(4)
a.
b.
John's refusal of the offer

John's criticism of the book
97
Given the formal and semantic regularity of the gerundive construction, a transformational derivation is conceivable, but the derived nominals, Chomsky argues, are a matter for
the lexicon. The subject of what exactly the lexicon looks like is not broached in that
article, but Chomsky's proposal that complex words might be a matter for the lexicon
clearly suggested to researchers of the time that the internal structure of the lexicon
merited further thought.
Halle (1973) takes up the structure of the lexicon in his Prolegomena to a the01y of
word-formation. In that article, Halle sketches a separate morphological component of
the grammar which consists of a list of bases and affixes, a set of word-formation rules,
and a dictionary which contains idiosyncratic information concerning semantics and phonology as well as a feature [-lexical insertion] that applies to words produced by the
word-formation rules that are not "actual" words. The dictionary thus serves not only to
store item-familiar words, but to filter out items that do not exist. The actual form of
word-formation rules is left vague, although Halle mentions the possibility of stating
them in the form of templates (1973: 10).
Jackendoff's (1975) Morphological and semantic regularities in the lexicon is another
early effort to flesh out the internal structure of the lexicon. Jackendoff proposes that
the lexicon should consist of lexical entries fully specified for syntactic, phonological,
and semantic information about words. Lexical entries are related to one another by
"lexical redundancy rules", essentially formal devices that evaluate the degree to which
those entries overlap: for example, the lexical entry for the derived word decision will
be "cheaper" to the grammar to the extent that it overlaps with the entry for decide
(1975: 642- 643):
(5)
/decTd/
I deeId + ion/
+V
+N
+ [NP 1 's _
+ [NP 1 _ on NP2]
NP, DECIDE ON NP2
on NP2]
ABSTRACT RESULT OF ACT

OF NP1 's DECIDING ON NP2
Lexical redundancy rules are conceived as part of the evaluation metric of a grammar,
that is, the mechanism by which the simplicity of the grammar is assessed. The more
redundancy, the simpler the grammar, simplicity, of course, being highly valued in a
generative grammar. The creative aspect of word-formation is not a central concern in
Jackendoff's article, although he does note that to the extent that new forms may be
derived, the lexical redundancy rules are permitted to function as generative devices.
Both proposals are very much products of their times. Both display an emphasis on
formalism that has characterized generative grammar throughout its existence. And both
also exhibit a somewhat prescriptive stance towards the lexicon, starting from the assumption that there is something we might characterize as "the existing words of a
PA98
98
language". Perhaps the most impo11ant point that they have in common, however, is
simply the fact that they identify morphology as a component of the grammar distinct
from both phonology and syntax, and worthy of study in its own right.
4. The heyday
The mid-1970's saw the publication of the first monograph on generative morphology,
Aronoff's (] 976) Word-formation in generative grammar, and this was. quickly followed
by a number of developments that set the stage for discussions of word-formation within
generative grammar for the next two decades. We have in mind here the work of Siegel
(1974, 1977), Allen (1978), Lieber (1980), Williams (1981), Kiparsky (1982), Selkirk
(1982), Scalise (1984) and Di Sciullo and Williams (1987), among others. Rather than
look chronologically at key works of generative morphology within this period, we will
approach our review thematically in this section, looking at the overall architecture of
the theories, including the status of the morpheme as a theoretical construct, the formal
nature of rules of word-formation, and the relationship among components of the grammar, and then at a number of specific proposals that form some of the foundational
claims of generative morphology.
4.1. Morphology and the architecture of the grammar

As the first full-length monograph treating morphology in the generative tradition, Aronoff's work set the stage for the set of issues that dominated generative morphology in
its first decade. Prominent among these was the status of the morpheme and its role in
the theory of word-formation. The morpheme is typically defined as the minimal meaningful unit of which words are constituted. Aronoff (1976) begins by pointing out the
major issue with this classic definition, namely that there are bits of words that can be
formally isolated but are not meaningful (so-called cran -morphs, Latinate formatives
like pel, mit, and ceive). He therefore rejects the traditional definition of morpheme, and
this move in turn leads him to the proposal that, "All regular word-formation processes
are word-based. A new word is formed by applying a regular rule to a single already
existing word" (1976: 21 ). Word-formation rules (WFRs) are conceived of formally as
phonological operations on a designated set of bases. In addition to changing the phonological form of the base, WFRs specify a syntactic category and subcategorization, a
semantic representation for their output, and a set of appropriate bases. (6) is Aronoff's
WFR for negative un- (1976: 63):
(6)
Rule of negative un#

a) [X] Adj ~ [un#[X]Adj]Adj
semantics (roughly) un#X
not X
b) Forms of the base

1. Xven
(where en is the marker for the past participle)
2. Xv #ing
3. Xvable
PA99
4.
5.
6.
7.
X+y
X+ly
X#ful
X#like
(worthy)
(seemly)
(mindful)
(warlike)
For cases in which the base to which an affix attaches is not apparently a surface word
(for example, nomin in nominee or proscrip in proscription), Aronoff proposes rules of
truncation or readjustment. The form nominee, for example, starts out as nominate + ee,
with the -ate affix being removed by truncation. As we will see shortly it is important
that WFRs in Aronoff's theory look formally very much like phonological rules, albeit
with a wider range of restrictions.
Subsequent generative models of this classic period can be divided into those that
embrace Aronoff's word-based framework and those that do not. Scalise (1984), for
example, falls into the former camp, although in a slightly different fonn: he rightly
points out that the notion of 'word' needs to be modified to 'stem' for highly inflected
languages like Latin where the bases to which WFRs apply are never themselves free
forms. On the other hand, the theories of Lieber ( 1980), Williams (1981) and Selkirk
(1982) explicitly reject the word-based model and assume the trad itional morpheme as
a legitimate unit of analysis. In all three theories affixes are treated as lexical items on
a par with bases as units having lexical entries. Affixes and bases are assembled by rules
formally analogous to the phrase-structure rules of generative grammar; the three proposals differ in the technical characterization of the word-structure rules that create complex
words; Selkirk's and Williams's rules, for example, look like the phrase-structure rules
in the standard theory of generative grammar (7a). Lieber's rules are more like the rules
of the later principles and parameters model, with lexical entries like (7b) for morphemes
that are then inserted into unlabeled binary-branching word trees to which percolation
conventions (see below) apply.
(7)
a.
A ---. N AAf
(Selkirk 1982: 66)
b.
-1ze (phonological representation)

semantic representation: causative
category/subcategorization: JN _ Jv
insertion frame: NP _ (NP)
diacritics: Level II
These theories make slightly different predictions, but in many ways they are comparable.
A second issue that forms part of the landscape of generative morphology concerns
the relationship among the components of the grammar. The vast majority of generative
morphologists professed some form of what has come to be called the strong lexicalist
hypothesis or the lexical integrity hypothesis. Lapointe (1980: 8), for example, proposes
what he calls the generalized lexicalist hypothesis, which states that "[nJo syntactic rule
can refer to elements of morphological structure". This version of the hypothesis in
effect rules out early generative analyses like Lees's (1960) as well as Levi's (1978)
transformational analysis of complex nominals. Selkirk's version, the word structure
99
PA100
100
autonomy condition (1982: 70), is similar. Di Sciullo and Williams' version is the atomicity thesis, which is a bit more explicit: "Words are 'atomic ' at the level of phrasal
syntax and phrasal semantics. The words have 'features', or properties, but these features
have no structure, and the relation of these features to the internal composition of the
word cannot be relevant in syntax[ ... ]" (1987: 49). A corollary of the strong lexicalist
hypothesis is Botha 's ( 1983) no phrase constraint, which prevents syntactic phrases from
being the input to rules of derivation and compounding.
An important consequence of the lexicalist hypothesis is that generative morphologists like Lieber, Selkirk, and Williams treat both inflection and word-formation as part
of a single morphological component of the grammar. Lieber 's argument in favor of this
arrangement of the grammar hinges on the behavior of inflectional allomorphy, specifically that in highly inflected languages stem allomorphs that figure in inflectional paradigms also form the input to rules of derivation. Anderson ( 1982), in contrast, advocates
a division between inflection and word-formation; arguing on the basis of case marking
and agreement in Breton, for example, he suggests that inflectional features must be
generated as part of the syntactic component, with features being associated with phonological forms only after syntactic operations have taken place. This separation between
word-formation, which takes place in the morphological component, and inflection,
which is part of the syntax, has come to be known as the weak lexicalist hypothesis.
Classical generative morphology also saw the formulation of a number of theoretical
restrictions on word-formation or word-structure rules, among them the righthand head
rule, feature percolation conventions, the unitary base hypothesis, the unitary output
hypothesis, the atom condition, and the adjacency condition.
One of the most influential proposals to come out of classical generative morphology
is the righthand head rule. Borrowing the concept of "head" from syntactic theory, the
head of a word was defined as that morpheme that determines the category and morphosyntactic features of the word as a whole. According to Williams (1981), the rightmost
morpheme in a complex word is the head, a principle that is known as the righthand
head rule (RHHR). While the RHHR is largely correct for English word-formation, it is
problematic in at least two ways. Lieber (1980) points out that is not necessarily true in
other languages: for example, Vietnamese is a language that has left-headed compounds.
Moreover, there are clear exceptions in English as well, where, for example, the prefix
de- can be category-changing and therefore are at least apparently heads, and where
several affixes like -dom, -hood, and -ship do not change category, and therefore are
righthand elements that cannot be considered heads.
An alternative to the RHHR is what Lieber (1980) refers to as percolation conventions. In Lieber 's version of generative morphology, features belonging to an affix percolate (in other words are attributed to) the derived word as a whole whether the affix is
a prefix or a suffix. Most, but not all, prefixes in English lack categorial and other
morphosyntactic features, which accounts for why English is mostly but not entirely
right headed. While feature percolation is clearly a weaker proposal than the RHHR, it
is cross-linguistically more supportable.
Another important theoretical proposal is Aronoff's (1976) unitary base hypothesis
which require that WFRs be stated to apply to a single sort of base. According to the
unitary base hypothesis, WFRs cannot be formulated, for example, to apply to both
nouns and verbs. Clear counterexamples can be found in English and in many other
languages, however. For example, the suffix -er in English can attach to verbs to form
PA101
agent or instrument nouns like writer or computer, but it also attaches to nouns to form
personal or instrument nouns like villager or freighter. To the extent that the semantic
contribution of -er seems to be the same in both instances, this looks like evidence
against the unitary base hypothesis. Cases like this can be circumvented only by claiming
that what appears superficially to be a single affix is actually two homophonous affixes.
Aronoff's example is the suffix -able, which he argues is in fact two suffixes, one that
attaches to nouns (marriageable) and another to verbs (washable). Scalise (1984) points
out an obvious problem with this fix: since nearly every prefix in English attaches to
multiple syntactic catgories, we would therefore be forced into the untenable conclusion
that we have as two or more homophonous affixes for the vast majority of prefixes.
Scalise also proposes what he calls the unitary output hypothesis, which states that
WFRs must have as their output words of a single morphosyntactic category. Like the
unitary base hypothesis, the unitary output hypothesis has potential counterexamples; for
example, the prefix un- in English attaches to adjectives (unhappy) to form adjectives
and to verbs (unwind) to form verbs, but has nevertheless been argued to be a single
negative prefix (Horn 1989; Lieber 2004).
Several proposals in generative morphology are concerned with the locality of wordformation rules. The atom condition proposed by Williams (1981) and the adjacency
condition proposed by both Siegel (1977) and Allen (1978) all prohibit the statement of
morphological rules from seeing any but the outermost layer of structure. In other words,
in a structure like [[x]y], a rule can refer to information contained in y but not x. As in
the case of the other restrictions we have reviewed here, there are possible counterexamples. Aronoff (1976: 52-53) mentions the case of the suffix -ment. Nouns in -ment can
generally only be made into adjectives by suffixing -al if -ment is itself attached to a
bound base (ornamental, regimental). If -ment is attached to a free verb base, the adjective in -al is generally not possible (*discernmental, *containmental). If we were to state
a restriction to this effect on suffixation of -al, we would need to refer to x in the
configuration [[x]y], which is ruled out by the atom condition and the adjacency condition.
Although it is true that the majority of theorists that contributed to the early period
of generative morphology were based in North America, we should not forget that there
were theorists in Europe, South Africa, and Asia that early adopted a generative stance,
among them Booij (1977), Scalise (1984), Botha (1983), and Kageyama (1982).
4.2. Morphology and phono logy

At the same time that classical generative morphologists were interrogating the relationship between morphology and syntax, the relationship between morphology and phonology was also of great interest, giving rise to both the framework of lexical phonology
and morphology and to prosodic morphology.
Lexical phonology and morphology, developed in work by Siegel (1974, 1978), Allen
( 1978), Kiparsky ( 1982), Halle and Mohanan (1985), Mohanan (1986), among many
others, begins from the premise that morphology and phonology are not distinct components, with word-formation strictly preceding the operation of phonological rules. Instead, rules of morphology and phonology are organized into a series of strictly ordered
levels with rules of word-formation and phonological rules interspersed within each
101
PA102
102
level. Affixes are introduced by rules that specify the phonological form of the affix and
the environment into which they are inserted. They do not themselves have lexical entries. In Kiparsky's (1982) classic version of the theory, English morphology, for example, consists of three levels. Level 1 includes morphological rules involving what SPE
called "non-neutral" derivational affixes, as well as rules for irregular inflectional forms.
Level 2 contains rules that add # boundary affixes (what SPE calls "neutral" affixes)
and rules that create compound words. Level 3 includes the rules for regular inflection.
Each level is associated with its own set of phonological rules. These three levels are
distinct from the syntax, which is subject to its own set of phonological rules referred
to as postlexical rules. Rules that apply to an earlier level cannot apply to the output of
a later level. It is this strict ordering of levels that accounts for the generalizations that
non-neutral affixes cannot occur outside neutral ones, that phonological rules that apply
to non-neutral affixes do not affect words formed with neutral affixes, that inflection is
outside of both types of derivational affix, and so on. Kiparsky's illustration of the model
(1982: 5) is given in (8):
(8)
underived lexical entri es
"+ boundary" inflection and
----+
derivation
stress, shortening
"# boundary" derivation and

compounding
compound stress
"# boundary" inflection
laxing
.__s_y_n_t_ax_ _ _ _ _ _ _ __,:
----+
level 1
level 2
level 3
postlexical phonology
Lexical phonology and morphology was an enormously influential theory, but it was
beset by problems from the start, prominent among them the need to introduce loops
allowing words formed on later levels to be cycled back to earlier levels; for example,
Giegerich (I 999: 3) cites examples like systems analyst, where a regularly inflected form
appears as the first element in a compound.
PA103
103

Prosodic morphology, also known as autosegmental morphology, was another theory
that developed from generative roots. Originating in the work of McCarthy ( 1979), prosodic morphology essentially sought to apply the principles of autosegmental phonology
to the treatment of non-concatenative processes of word-formation. Autosegrnental phonology postulates that phonological segments are not unstructured lists of distinctive
features, but rather are organized into a number of tiers that are themselves composed
into hierarchically organized prosodic structures (syllables, feet, prosodic words). Prosodic morphology capitalizes on this hierarchical organization by postulating that morphemes can consist of segmental tiers or prosodic constituents, and not just of sequences
of whole segments as is typical of generative morphology. This sort of organization
makes it possible to treat non-concatenative word-formation processes more on a par
with conventional affixation and compounding. McCarthy's seminal work was on the
root and pattern morphology of the Semitic languages, but the theory was soon applied
to reduplication (Marantz 1982) and processes of vowel and consonant mutation (Lieber
1984, 1987). McCarthy's now-classic analysis of the Arabic conjugation partitions each
verb into three tiers. One tier consists of a pattern of organization of consonants and
vowels that corresponds to what Hebrew grammarians call "binyanim". A second tier
contains the consonantal root, which carries the central lexical meaning of the complex
word. Finally a third tier contains vowels that typically signal aspect and voice:
(9)
""
k t b
c / c/ v c v c
'write'
binyan 9
\_,,/
a
perfective active
A major result of the work of this period on prosodic morphology was a substantial
deepening of our knowledge of the scope and lim its of non-concatenative word-formation cross-linguistically.
5. Reactions to classic generative morphology

By the mid l 980's the time was ripe for reconsidering the assumptions of classic generative morphology. We begin with a consideration of the anti -lexicalist reaction and then
consider a number of post-generative frameworks that can be said to have their roots in
the generative movement.
5.1. The resyntacticization of morphology

By the mid 1980's morpho logy was clearly established in the generative framework as a
legitimate area of study. This newly established legitimacy allowed theorists of the period
to reconsider the organization of the grammar and revive the possibility that some or all
of morphology could be treated as a matter of syntax; that is, one could claim morphology as a legitimate area of research interest without committing oneself to a separate
PA104
104
component of morphology in the grammar. The rejection of morphology as a separate

component did not, however, mean a return to the sort of analysis proposed by Lees in
the early days of generative grammar. By the 1980s the theoretical landscape of generative syntax was very different from Chomsky's early theory. The move back to syntactic
morphology took place in the context of the theory that has come to be known alternatively as government-binding theory or the principles and parameters model (Chomsky
1981 ).
It is in the context of this theory that Fabb (1984), Sproat (1985), Roeper (1988),
Baker (1988), and Lieber (1992) adduce several sorts of evidence to call even the weak
form of the lexicalist hypothesis into question. By the early l 980's the nature of argument structure had become a topic of great interest in generative syntax, giving rise
within the principles and parameters model to case theory on the one hand and theta
theory on the other. Since both derivational affixes and synthetic compounds exhibit
argument structure effects - for example, the propensity for a derived word in -er to
denote the external argument of its base verb or the interpretation of the first element of
a synthetic compound as the internal argument of the second element's base - it was
natural that these types of word-formation come under special scrutiny and that attempts
be made to put them in line with the syntactic theory of the time.
Fabb (1984), for example, splits derivational affixes into two sorts, what he calls
syntactic affixes and lexical affixes. The former are productive, regular, and subject to
case theory. They generally (but not always) attach to verbs and either allow the projection of verbal arguments (for example, gerundive -ing) or affect the verbal projection in
some regular way (for example, the suffix -er which is associated with the external theta
role of the verb in the sense that derived -er words denote whatever the thematic role of
the subject of a given verb would be). The lexical affixes are less productive and the
forms derived with them are associated with idiosyncratic properties. What is problematic about Fabb's proposal is that there are lexicalized forms derived from most affixes,
which leads to the conclusion that most affixes must have both syntactic and lexical
versions. Sproat's (1985) proposal is much stronger than Fabb 's, abandloning the lexicalist hypothesis entirely. He uses the existence of so-called bracketing paradoxes like unhappier to argue for postulating two levels of analysis for derived words, a syntactic
level and a phonological level; there is no separate morphological component. He argues
that bracketing paradoxes can be resolved by assuming that on the syntactic level, -er is
attached to the negative adjective unhappy, from which we can derive the appropriate
semantic representation. On the phonological level, negative unis attached to the comparative form happier; with the latter bracketing, we are able to maintain the phonological restriction that comparative -er attaches to words of no more than two syllables. The
two levels are associated to each other via mapping principles.
Baker (1985, 1988) adduces a different sort of evidence for the syntactic treatment
of word-formation. He formulates what he calls the the mirror principle, which says that
"[m]orphological derivations must directly reflect syntactic derivations (and vice versa)"
(1985: 375). This principle allows him to account for facts of languages like Chichewa
in which it is possible either to apply a process of applicativization to a passive sentence,
or alternatively to passivize an applicative sentence. In the former case, the passive
morpheme appears inside the applicative and in the latter the applicative appears inside
the passive, the order of the morphology reflecting the ordering in which the operations
are performed (example from Baker 1988: 14; abbreviations used in examples (10) and
PA105
105

(11) are as follows: SP= subject prefix, PASS = passive, APPL = applicative, ASP = aspect,
3M = third person masculine, 3N = third person neuter, DET = determiner, suF = nominal
inflection suffix):
( l 0)
a. Mpiringidzo un-na-perek-edw-a
kwa mtsikana ndi mbidzi
Crowbar
SP-PAST-hand-PASS-APPL-ASP to girl
'The crowbar was handed to the girl by the zebras.'
by zebras
b. Mtsikana a-na-perek-er-edw-a
mpiringidzo ndi mbidzi
girl
SP-PAST-hand-APPL-PASS-ASP crowbar
by zebras
'The girl was handed the crowbar by the zebras.'
Baker (1988) also brings to the discussion the process of noun-incorporation in which it
appears that a syntactic object can become part of a complex verb. Examples like (11)
from Mohawk are prima facie persuasive, as they appear to illustrate that a noun stem
can be extracted from an NP leaving behind its possessor (1988: 20):
(11)
a.
Ka-rakv ne sawatis hrao-nuhs-a?

3N-be.white DET John 3M-house-suF
' John's house is white.'
b.
Hrao-nuhs-rakv ne sawatis
3M-house-be.white DET John
' John's house is white.'
Both the mirror principle and the syntactic analysis of noun incorporation have not been
uncontroversial, however. Grimshaw ( 1986), for example, argues that the facts accounted
for by the mirror principle can be better explained without assuming a syntactic analysis.
Similarly, Rosen (1989) argues against a syntactic treatment of noun-incorporation.
In the context of this discussion, synthetic compounds also were taken to suggest the
syntactic treatment of morphology. Arguing on the basis of facts concerning control and
aspect, Roeper (1988) proposes that compounds like rock-throwing in English should be
analyzed as the output of a syntactic operation of head movement that starts out with a
sentential structure. Roeper 's derivation of the compound rock-throwing is illustrated in
(12).
The N rock is incorporated into the verb and the -ing suffix is lowered onto the
complex verb from an Infl node. As part of the derivation Roeper assumes that the IP
undergoes a category-change to an NP, with a PRO subject under SPEC, thus explaining
why in a sentence like John enjoyed rock-throwing/or hours, the rock-thrower is typically interpreted as John. The Infl node in the underlying structure of the same sentence
accounts for the progressive interpretation of the synthetic compound.
Lieber (1992) continues in the effort to apply principles of syntax to word-formation.
Lieber starts from the premise that the existence of phrasal compounding and derivation
suggests that morphology and syntax cannot be separated into distinct components. She
then attempts, on the grounds of simplicity, to develop a theory in which the principles
of morphology and syntax are in fact the same. As part of this program she tries to
derive the ordering of affixes from the normal settings of the X-bar parameters, an
PA106
106
(12)
IP
~
SPEC
/\--V
throw
attempt that is in the end undermined by the difficulty of applying terms like complement
and specifier to morphemes in complex words.
All of these proposals have come under criticism and the re lative merits of counting
word-formation as a part of syntax or as a distinct component are still being debated
(see, for example, Lieber and Scalise 2006). Di Sciullo (2005) is notable in this regard:
although she seeks to apply the principles of the minimalist program to word-formation,
Di Sciullo nevertheless continues to maintain that the mechanisms that apply to wordformation are distinct in key ways from syntactic principles.
5.2. The fragmentation of generative morphology

Up to the early 1990's it is possible to think of the theory of generative morphology as
a single framework united not so much by specific principles or a set of formal mechanisms espoused by any specific theorist, but by the set of foundational assumptions
mentioned at the outset. What has characterized the study of morphology in generative
grammar over the last two decades has been an increasing specialization of theory. In
this section we look at some of the outgrowths of generative morphology, giving most
attention to those that have focused on word-formation as opposed to inflection.
5.2.1. A digression on inflection

Although the subject of thi s article is word-formation in generative grammar and not
inflection, it is difficult to continue a discussion of more recent developments without
at least passing mention of frameworks whose primary focus is inflection.
PA107
Anderson's (1992) a-morphous morphology and Stump's (2001) paradigm function
morphology both have their roots in Matthews' ( 1972) word-and-paradigm framework
which begins from the observation that the relationship between meaning/morphosyntactic features and phonological representation is rare ly one to one. When looking at complex inflectional systems, we often find on the one hand that a single morphosyntactic
feature is expressed by more than one morpheme (so-called multiple exponence), and
on the other that a single morpheme can express more than one morphosyntactic feature.
Anderson (1992) takes this observation as a point of departure for a theory consisting
of rules that assoc iate bund les of features (sometimes with internal structure) with phonological forms; like Aronoff's classic theory, Anderson's framework is word-based, but
his rules do not introduce inflectional features, as Aronovian-style rules would, but rather
associate phonological form with sets of features. Stump 's (2001) paradigm function
framework differs in formal details from Anderson's, but is based on a similar premise.
What is important for our purposes is that both Anderson's and Stump's frameworks are
conspicuous in that they have almost nothing to say about word-formation (each contains
only a brief chapter or part of a chapter devoted to the subject).
Another highly influential framework that was proposed in the early l 990's is distributed morphology (OM) (Halle and Marantz 1993; Harley and Noyer 1999). Like amorphous morphology and paradigm function morphology, the main emphasis in this
framework has been on the analysis of inflection, with relatively little attention devoted
to word-formation. OM is conceived broadly as a part of the larger m inimalist program
(Chomsky 1995). Morphology proper is part of the syntax. Nodes in a syntactic tree are
associated with bundles of features, and bundles of features can undergo movement rules
from one structural position to another. Only after the syntactic derivation is complete
are bundles of features associated with phonological forms and subjected to certain sorts
of readjustment rules, an arrangement that is referred to as "late insertion"; OM thus
embraces Beard 's separation hypothesis. Roots - the lexical bases to which affixes attach - are inherently categoryless, and are associated with a syntactic category as part
of the syntactic derivation; versions of the theory differ on whether roots as well as
affixes are subject to late insertion. Attempts have been made within OM to extend the
discussion at least to compounding (Harley 2009), but it remains to be seen how the
theory might be adapted to account for derivational word-formation.
5.2.2. Lexeme morpheme base morphology
Most of the syntactic models discussed in section 5.1 take for granted that the morpheme
should be taken as the standard unit of analysis. Beard (1995), however, takes as a point
of departure the same lack of a one to one co!Tespondence between meaning and form
that motivates Anderson's and Stump's theories and explores the consequences of this
observation for derivation. The result is what he calls lexeme morpheme base morphology. One of Beard 's most important achievements is to bring the issue of semantics
into the discussion of word-formation by observing that derivation, like inflection, often
involves mismatches between form and meaning. Beard's solution to these mismatches
is to treat the semantics of derivation on a par with that of inflection. For Beard, wordformation consists in the addition of derivational features like [Agent] to a lexemic base.
The agentive nouns corresponding to the verbs write and account would both be the
107
PA108
108
same at this stage of derivation. However, these two representations are spelled out with
different phonological forms, resulting in the one case in the form writer and in the other
the form accountant. The disjunction between the semantic representation of a word and
its ultimate phonological fonn is what Beard calls the separation hypothesis. A major
question that arises from Beard 's theory, however, is whether the semantics of derivation
can really be accounted for with a featural system analogous to that needed for inflection.
5.2.3. The lexical semantics of word-formation
Beard's solution to the mismatch between morphology and semantics is essentially an

architectural one: by designing a theory that separates the semantic part of word-formation from its phonological realization he is able to address the important issue of meaning-form mismatches. Lieber (2004) agrees that such mismatches must be accounted for,
but argues that the success of any treatment depends on having an explicit theory of
lexical semantics that accounts for derivation, compounding, and conversion, as well as
the lexical semantics of free morphemes. Beginning from the well known observation
that the affix -er not only forms agent nouns like writer, but also instrument nouns
(printer), location nouns (diner), and even patient nouns (loaner), she proposes a framework of lexical semantic representation that allows her to explain why derivational affixes are so often wildly polysemous. Beard relies on inflectional categories like [agent],
[means] (the feature Beard uses for instruments), or [location] to characterize the meanings of derivational morphemes, thus making it seem accidental that a suffix like -er can
express all these meanings. Lieber's lexical semantic representations consist of hierarchical arrangements of features and their arguments. Features cover broad aspects of lexical
meaning, such as for example the characteristic of being concrete or abstract, which
Lieber encodes in the feature [+/-material] and eventive or non-eventive, encoded in the
feature [+/-dynamic]; features are deployed across syntactic categories, so that nouns
may be characterized as eventive or non-eventive, just as verbs are. Both affixes and
bases in her theory have lexical semantic representations, indeed the same sorts of representations. When these representations are combined, arguments of the affix may be
linked to arguments of the base via what Lieber calls a principle of co-indexation:
(13) a.
write
[+dynamic [( ), ( )]]
b.
-er
[+material, dynamic ([ ], <base>)]
c.
writer
[+material, dynamic
(L ], [+dynamic(L ], [ ])])]
The polysemy of an affix like -er falls out in her theory not so much from the architecture
of the grammar but from the extent to which the semantic representation of the affix is
underspecified.
6. Conclusion
Active study of word-formation continues in the generative tradition. However, the emphasis in recent years has to some extent changed. For many years, generative morpholo-
PA109
gy focused on a broad-brush characterization of the architecture of the theory and relied
on formalism and principles modeled on the syntactic or phonological theory of the
moment. In recent years we have seen a gradual movement towards a line of thinking
that does not essentially piggyback on syntactic or phonological theory. One reason for
this change in emphasis is the welcome convergence of theoretical approaches to morphology in the generative tradition with an independent thread of psycholinguistic study
of word-formation that has gone on almost in parallel. Represented in the work of Taft
(1988), Marslen-Wilson et al. (1994), Baayen, Dijkstra and Schreuder (1997), De Vaan,
Schreuder and Baayen (2008), Kuperman, Be11ram and Baayen (2010), among many
others, this line of research has attempted on the basis of psycholinguistic experiments
to model the extent to which complex words are stored as wholes in the mental lexicon
as opposed to generated on line. This line of research as well as the increased attention
to the issue of productivity in word-formation (for example, Plag 1999; Bauer 2001 ),
has brought home to many current morphologists that words are different from sentences.
Words persist in the mental lexicon in a way that sentences typically don 't. Further, the
increasing availability of searchable corpora like the British National Corpus or the
Corpus of Contemporary American English gives theorists today the means to track
productivity and the semantics of complex words in context, and this w ill certainly lead
to new insight into the organization of the mental lexicon.
7. References
Allen, Margaret
1978
Morphological investigations . Ph.D. dissertation, University of Connecticut, Storrs.
Anderson, Stephen
1982
Where's morphology? Linguistic Inquiry 13(4): 57 1- 6 12 .
Anderson, Stephen
A-morphous Mo1phology. Cambridge: Cambridge University Press.
1992
Aronoff, Mark
Word-formation in Generative Grammar. Cambridge, MA: MIT Press.
1976
Baayen, Harald, Ton Dijkstra and Robert Schreuder
1997
Singulars and plurals in Dutch: Evidence for a parallel dual-route model. Journal of
Memory and Language 17(1): 94- 117.
Baker, Mark
1985
The mirror principle and morphosyntactic explanation. Linguistic Inquiry 16(3): 373415.
Baker, Mark
1988
Incorporation. Chicago: University of Chicago Press.
Bauer, Laurie
200 1 Morphological Productivity. Cambridge: Cambridge University Press.
Beard, Robert
1995
Lexeme Morpheme Base Morphology. Albany: State University of New York Press.
Booij, Geert
1977
Dutch Morphology. Lisse: Peter De R idder Press .
Botha, Rudolph
Morphological Mechanisms. Oxford: Pergamon Press .
1983
C homsky, Noam
1957
Syntactic Structures. The Hague: Mouton.
109
PA110
110
Chomsky,
1968
Chomsky,
1970
Noam
Aspects of the Theot)' of Syntax. Cambridge, MA: MIT Press.
Noam
Remarks on nominalization. In: Roderick A. Jacobs and Peter S. Rosenbaum (eds.),
Readings in English Transformational Grammar, 184- 221. Waltham, MA: Ginn.
Chomsky, Noam
198 L Lectures on Government and Binding. Dordrecht: Faris.
Chomsky, Noam
1995
The Minimalist Program. Cambridge, MA: MIT Press.
Chomsky, Noam and Morris Halle
The Sound Pattern of English. New York: Harper and Row.
1968
De Vaan, Laura, Robert Schreuder and Harald Baayen
2008
Regular morphologically complex neologisms leave detectable traces in the mental lexicon. Th e Mental Lexicon 2(1): 1-24.
Di Sciullo, Anna Maria
2005
Asymmetry in Morphology. Cambridge, MA: MIT Press.
Di Sciullo, Anna Maria and Edwin Williams
1987
On the Definition of Word. Cambridge, MA: MIT Press.
Fabb, Nigel
1984
Syntactic affixation. Ph.D. dissertation, MIT Cambridge, MA.
Giegerich, Heinz
Lexical Strata in English . Cambridge: Cambridge University Press.
1999
Grimshaw, Jane
1986
A morphosyntactic explanation for the mirror principle. Linguistic lnquity 17(4): 745749.
Halle, Morris
1973
Prolegomena to a theory of word formation . Linguistic Inquiry 4(1): 3-16.
Halle, Morris and K. P. Mohanan
1985
Segmental phono logy of Modern E nglish. Linguistic Inquiry 16(1): 57-116.
Halle, Morris a nd Alec Marantz
Distributed morphology and the pieces of inflection. In: Keneth Ha le and Samuel Jay
1993
Keyser (eds.), The View from Building 20, 111- 176. Cambridge, MA: MIT Press.
Harley, Heidi
2009
Compounding in distributed morphology. In: Rochelle Lieber and Pavo! Stekauer (eds.),
The Oxford Handbook of Compounding, 129-144. Oxford: Oxford University Press.
Harley, Heidi and Rolf Noyer
Distributed morphology. Glot International 4(4): 3- 9 .
1999
Horn, Laurence
1989
A Natural History of Negation. Chicago: University of Ch icago Press.
Jackendoff, Ray
1975
Morphological and semantic regularities in the lexicon. Language 51: 639- 671.
Kageyama, Taro
1982
Word formation in Japanese. Lingua 57(2- 4): 215- 258.
Kiparsky, Paul
1982
Lexical phono logy and morphology. In: In-Seok Yang (ed.), Linguistics in the Morning
Calm, 3- 91. Seoul: Hanshin Publishing Company.
Kuperman, Victor, Raymond Bertram and Harald Baayen
2010
Processing trade-offs in the reading of Dutch derived words. Journal of Memory and
Language 62(2): 83- 97.
Lapointe, Steven
1980
A theory of grammatical agreement. Ph.D. dissertation, University of Massachusetts,
Amherst.
PA111
Lees, Robert
1960
The Grammar of English Nominalizations. Bloomington, IN: Indiana University Press.
Levi, Judith
The Syntax and Semantics of Complex Nominals. New York: Academic Press.
1978
L ieber, Roche lle
1980
On the organization of the lexicon. Ph.D. dissertation, MIT Cambridge, MA. [Published
1981 Bloomington, IN: Indiana University Linguistics Club and 1990 New York: Garland Publications.]
Lieber, Roche lle
1984
Consonant gradation in Fu la: An autosegmental approach. In: Mark Aronoff and Richard
Oehrle (eds.), Language Sound Structure, 329-345. Cambridge, MA: MIT Press.
Lieber, Roche lle
An Integrated Themy of Autosegmental Processes. A lbany: State University of New
1987
York Press.
Lieber, Rochelle
Deconstructing Morphology. Chicago: University of Chicago Press.
1992
Lieber, Rochelle
Morphology and Lexical Semantics. Cambridge: Cambridge University Press.
2004
L ieber, Roche lle and Sergio Scalise
The lexical integrity hypothesis in a new theoretical universe. Lingue e Linguaggio 5(1):
2006
7-32.
Marantz, Alec
Re reduplication. Linguistic lnquify 13(3): 435-482 .
1982
Marslen-Wilson, William, Lorraine Tyler, Rachelle Waksler and Liane Olde
Morphology and meaning in the English mental lexicon. Psychological Review 101(1):
1994
3- 33.
Matthews, Peter H.
Inflectional Morphology. Cambridge: Cambridge University Press.
1972
McCarthy, John
1979
Formal problems in Semitic phonology and morphology. Ph.D. dissertation, MIT Cambridge, MA.
Mohanan, K.P.
The Theory of Lexical Phonology. Dordrecht: Reidel.
1986
Newmeyer, Frederick
1986
Linguistic Theory in America. New York: Academic Press.
Plag, Ingo
Morphological Productivity. Structural Constraints in English Derivation . Berlin: Mou1999
ton de Gruyter.
Roeper, Thomas
Compound syntax and head movement. In: Geert Booij and Jaap Van Marie (eds.),
1988
Yearbook of Morphology 1988, 187-228. Dordrecht: Foris.
Rosen, Sara
1989
Two types of noun incorporation: A lex ical analysis. Language 65(2): 294-317.
Scalise, Sergio
1984
Selkirk, Elisabeth
1982
Siegel, Dorothy
1974
Topics in English morphology. Ph.D . dissertation, MIT Cambridge, MA.
Siegel, Dorothy
The adjacency condition and the theory of morphology. In: Mark J. Stein (ed.), Proceed1977
ings of the Eighth Annual Meeting of the North East Linguistic Society, 189- 197. Amherst, MA: GSLA.
111
PA112
112
Sproat, Richard
1985
On deriving the lexicon. Ph.D. dissertation, MIT Cambridge, MA.
Stump, Gregory
2001
Inflectional Morphology. Cambridge: Cambridge University Press.
Taft, Marcus
1988
A morphological-decomposition model of lexical representation. Linguistics 26: 65767.
Williams, Edwin
1981
On the notions 'lexically related' and 'head of a word'. Linguistic lnqui1y 12(2): 245274.
Rochelle Lieber, Durham, NH (USA)
8. Word-formation in categorial grammar

1.
2.
3.
4.
5.
Introduction
Categorial grammar
Derivation
Compounding
References
Abstract
So far, morphology has rather been neglected in categorial grammar which was primarily developed for syntax. Categorial grammar presupposes that there are basically two
types of linguistic categories, namely functors/operators and arguments/complements.
This notation can be applied to morphology, covering both inflection and word-formation. Word-formation can be perceived of as an interface between lexicon and syntax, or
as an integration of syntax into vocabulary. The descriptive power of categorial grammar becomes evident when applied to the field of word-formation and morphology in
general; it manifests itself in discovering and handling the most detailed morphosyntactic and morphosemantic differentiations.
1. Int roduction
Up to now, morphology has been rather neglected in categorial grammar, which was
primarily developed for syntax. This is true as far as inflection is concerned but also for
word-formation as well. In fact, morphology and word-formation are as good as missing
in the most comprehensive descriptive categorial grammar so far by Zifonun, Hoffmann
and Strecker (1997, 3 volumes, approx. 2500 pages). Categorial Grammars by Mary
PA113
McGee Wood (1993) comprises a short overview in which several theoretica~ approaches
are treated only briefly and in addition some problematic cases are sketched. Moreover,
she points out that (obviously) both inflectional and derivational suffixes are considered
by most authors as functors to be applied to lexical bases. Karl Reich! 's work Categorial
Grammar and Word-Formation. The De-adjectival Abstract Noun in English (1982) is a
more comprehensive and detailed monograph, quoted frequently as such. The main interest here lies in the linguistic material treated in detail but less in general theoretic questions. A rather theoretic focus is given in Jack Hoeksema's Categorial Morphology
( 1985) in which categorial approaches are compared to others (cf. also Hoeksema and
Janda 1988). This is why it does not offer either a comprehensive presentation of categorial grammar as a theoretical framework nor a discussion of the different patterns of
word-formation within thi s framework.
Inflection and word-formation is discussed more extensively in Ulrich Wandruszka
(1976, 1997 and 2007). In these contributions, an effort is made, on the one hand, to
consider in a formally consistent way the fundamental difference between syntax and
morphology, and, on the other, to model the interfaces and transitions between morphology and syntax amongst others by means of the operation of functional composition. In
this context, the author also tackles the much discussed phenomenon of government
inheritance from a derivational base to the derived word in cases of the type German
die Sehnsucht nach Liebe 'the longing for love' or die Emporung iiber den Preis 'the
indignation about the price' and so on (cf. Moortgat 1988). Furthermore, the question as
to whether forms like Korkenzieher 'corkscrew; lit. cork puller' should be considered as
compounds or derivations is discussed, and how constructions in which the first part is
an adjective or a numeral, as, e.g., Dickhauter 'pachyderm; lit. thick skin-er' or Dreimaster 'three-master', Tausendfiij3ler 'millipede; lit. thousand feet-er', etc., should be analyzed. As far as inflection is concerned, constructions of the type ((Wein trinkende)NiN
MannerN)N 'wine drinking men ', in which a complex participial attribute is generated
by means of functional composition, are treated in principle analogously to Korkenzieher.
Also considered in this theoretical framework is the question of whether such bound
morphemes are mentally stored as isolated units or always as parts of complex words or
word forms.
2. Categorial grammar
Categorial grammar presupposes that there are basically two types of linguistic categories, namely functors/operators and arguments/complements. Functors are so-to-speak
the active elements, taking an active part in building up a complex expression. This
activity arises from the valence of functors, resulting negatively from a blank, i.e. the
lack of a particular structural partner and the necessity of closing or filling this blank.
In this context, there are two kinds of functors: governing or head functors, operating
as the governing head of a combination and dependent functors, constituting the adjunct
or attribute of a combination. Together with a complement, head functors form a complex
expression whose category they define and represent. Thus, a transitive verb such as
drinks takes a noun phrase, such as a beer, form ing with it the predicate drinks a beer,
which now together w ith a subject NP like Brigitte forms the sentence Brigitte drinks a
beer. Such functors are represented in categorial grammar by a fraction of the type Y I
113
PA114
114
X, i.e. an expression that takes an X and returns a Y. Our predicate drinks a beer thus
represents the type S/NP and the verb form drinks the type (S/NP)/NP in that it forms
together with an NP the expression S/NP, which then together with an NP forms a
sentence. Such a verb form is thus bivalent, as opposed to the monovalent sleeps of the
type S/NP, as in (the catNP sleepss!NP)s.
Dependent functors, on the other hand, attach to a head X, forming with it an (extended) expression of the category X. Since dependent functors are dependent attributes they
cannot define or change the category of the resulting expression. Because they maintain
the respective category, they belong to the formal type X IX. Thus, e.g., an adjective is
a functor that combines with a noun and forms a (complex) noun, such as (littleNIN
lightN)N Accordingly, an adverb is an expression that combines w ith a verb and forms
an (extended) verb, cf. German (falschv!V singenv)v 'to sing out of tune'.
The other large group of linguistic expressions embraces arguments or complements
that by virtue of their role as complements to head functors do not have empty slots and
thus are zero-valent, such as the NP a beer which, when it combines with a transitive
verb, is transformed into a direct object as in (the phrase) to drink a beer. NPs can also
be the complement of a preposition forming a prepositional phrase with an NP, as in
French (danspp/NP (la ville)Np)pp, 'in the city' which is of the type PP/NP. Complements
of attributes on the other hand can be of any valence, since the latter do not have any
influence on the category of the resulting expression. This is why, e.g., an adverb together with a monovalent verb can produce again the same type: (sleepss!NP long(StNP)l(SI
NP))S/NP
Now this notation can be applied to morphology, covering both inflection and wordformation which is of particular interest in the present context. It is generally suitable
for all complex linguistic expressions. Thus, the German suffix -er in the noun Maler
' painter ' is of the type N/ Stv, in that it forms together with a verb stern (Stv) a noun:
(Mal-st , -erN/stJN Semantically speaking, the function of -er is to produce with the help
of an action verb the denomination of the respective doer. In a form like German kirchlich 'churchly', the suffix -Zich forms with a noun or a noun stem (Stn) an adjective
(Adj), thus being of the type Adj/Stn: (kirch- 510 -lichAdj!St,)Actj The suffix/flexive -en
in turn belongs to the category V / Stn, generating with a noun or noun stern a verb:
(haus-s 1" -env1s1 Jv '(to) live' . The French suffix -ier, as in poirier 'pear-tree' forms
nouns with the help of noun sterns, which can be analyzed as follows: (poir-s 10
-ierN! stJN As we will discuss more in detail, defining derivational suffixes as functors,
i.e. as instruments for forming new words, is quite suitable for depicting linguistic reality.
The applicability of this model to inflectional morphology, the central grammatical
part of rnorphosyntax, can be touched upon only briefly in the present context. Here
again, in the realm of conjugation and declension, the focus lies on generating morphologically complex lexical units by means of bound morphemes. This is the case, e.g., in
forms like German (sie) bell-t '(she) bark-s' which result from a functional element,
hence a functor, combining with a categorially as yet "unformatted" verb stern to form
with it, e.g., a predicate (S/NP). The ending -t thus belongs to the type (S/NP)/ Stv
generating by means of a bound morpheme a morphologically closed form of a certain
syntactic category: (bell-s 1, -t(sJNP)/St. )sJNP The fact that symbols of the form (S/NP)/ Stv
show a morphologically bound element in the denominator and a syntactic category in
the numerator makes it clear that this is the domain of rnorphosyntax, the overlap between lexicon and syntax, connected by the respective inflectional ending. As for declension, one might consider the Latin genitive ending -is, which generates together with a
PA115
noun stem an attribute, as in (domusN (patr-st" -is(NIN)lst)NIN)N 'father 's house', 1.e.
patris is a noun functioning as an attribute.
3.
Derivation
3.1. Suffixation
Word-formation as well can be perceived of as an interface between lexicon and syntax,
or as the integration of syntax into vocabulary - a process also known as lexicalization or
morphologization (cf. Morrill 201 1). As opposed to inflection, which is about generating
different forms of the same word in the sense of a lexeme, here the task is the formation
of new words from bound morphemes and /or (free) words, thus extending the lexicon
of a language with its own means and own material. The central morphologically bound
functor in this field is, along with the prefix, the derivational suffix, a head functor wh ich
is different from an inflectional element in that it does not generate word forms but new
words and thus - at least in inflectional languages - forms stems. To these products of
word-formation, then, the several inflectional elements with syntactic functions and other
gramm atical features can be applied, as, e.g., in the deverbal Italian noun lava-tor-el
lava-tor-i 'washer/washers' or the adjective lava-bil-e/lava-bil-i 'washable', and so on
(cf. Werner 1990).
Whereas in the realm of inflection the head status of the ending is precarious, in the
case of derivation this designation can generally be taken unequivocally, since the suffix
always represents the part of speech and also the type of meaning or denotation of the
new word. Thus, suffixes like -tor(-e) and -bil(-e) are morphologically and semantically
the head of their respective derivations: lava-tore denotes a person or an instrument and
lava-bile a proper ty, whereas the nouns lava-ggio/ lava-tura 'washing' describe an activity, as does the verbal derivational base lav-are '(to) wash'. Moreover, even if the suffix
does not change the part of speech of the derivational base, as in automobil-ista 'motorist' or automobil-ismo 'motoring', the morphosemantic dependency structure remains
clear (formations like German Ismen ' -isms' exemplify that the suffix is in fact perceived
of as a head constituent).
Inflectional elements, on the other hand, form together with the stem of a given word
class a special form of this very word, such as, e.g., parl-ate 'you talk', and thus belong
to the type word x/stem x, or Wx/Stx, where W stands for a special function of the word.
In the case of suffixation however there are generally two stems involved: Sty/ Stx, i.e.
stemylstemx, which can belong to the same word class (Stn/ StnJ, as in automobil-ist(a). In other words, inflection happens within the borders of a word, whereas suffixation
as a type of word-formation generally transcends these borders, relating at least two
different words as a matter of principle.
The head status of suffixes is thus even more pronounced than that of inflectional
elements, in that the former, representing the word class and the respective denotation
type, defines the basis for word class specific determination and functional inflectional
elements. This is true, e .g., in the above mentioned adjective ((lava- s 1v -bil-svsiJs1 -e
Adj/ stJActj> in which the adj ective ending - eAdj/ St. together with an adjective stem forms
an adjective, or in a complex noun like Italian formalizzazione 'formalization':
115
PA116
116
((((fio rm-st n -al-sta

1stn)st
)stv -zion-stn1stv)stn -eNlst n) N
a -izza-st v1sta
N
N/St0
form-
-al-
-izza-
-zion-
-e
On the left in the tree diagram, it is quite visible how adding a suffix to the nominal
stem form- converts it into an adjective stem, then into a verb stem and finally into a
noun stem again. Moreover, the difference between various stem forming suffixes and
the word form building the inflectional element -e of the type Wx/Stx also becomes
apparent. As opposed to the left hand side of the tree, where the derivational bases, i.e.
the complements of the suffixes ascend level by level, on the right hand side the suffixes
themselves and the ending, i.e. the functors of the right-headed constructions are aligned.
From a formal point of view, there is another peculiarity regarding derivation: although we are dealing with head functors and complements, similar to the case of attribution the resulting structure is iterative because applying further suffixes each time will
produce new words or word stems. Since suffixes form more complex words or word
stems with words or stems, the "products" along with the suffix always contain a word
or rather a word stem as an immediate constituent. This is a form of recursiveness that
does not occur in syntax, where, e.g., a simple sentence is never an immediate part of a
complex sentence, but commonly of an adverbial, as in (before (you came)s)Adv' or of
a verb phrase, as in (hopev [that] (Elvira comes)s)vp However, as opposed to attributes,
suffixes, as heads, are not optional structurally and thus cannot be simply deleted in a
syntactic structure. This means, that on the level of lexical categories suffixes are head
functors of the type X IX, whereas head functors in "free" syntax are of type Y/X. This
is due to the fact that in syntax the common category word is missing and words instead
are combined to a variety of syntactic constructions (syntagmas).
Processes of word-formation are so-called relations in, i.e. relations within the set of
words of a language. This is why the products of word-formation show the structure of
an onion: deleting the head of a complex word again reveals a word, such as in German
Nation-al-itiit 'nationality', whereas Italian represents the generation of complex stems
out of simpler ones: nazion-al-ita. Thus, word-formation simply entails producing a new
word with a given word or a new stem with a given stem. In the case of category change,
that is to say by applying inflectional elements of the type word xfstemx, or Wx/Stx, i.e.
inflectional elements forming together with a stem a morphologically closed word, no
further iteration is possible, cf. iformalizzazion-s 10 -ew s1Jw Likewise, in derivations
01
PA117
like German Kind-heit 'childhood', Kind 'child' does not represent the nominative singular (das Kind), but might be perceived of as a stem with a free slot for a suffix.
This constancy of categories in word-formation is true only for the value word, but
not for a special part of speech which can be changed by means of word-formation
processes. In this domain, suffixes behave like normal head functors of the type Y/X,
making with a word of category X one of category Y, allowing also the change of
syntactic functions. For example, the German suffix -Lich combines with nouns to form
adjectives likefreund-lich 'friendly', abend-lich 'vespertine ' or vater-lich 'fatherly', thus
belonging to the functional category adjective/noun, or adj/n, if morphological categories, i.e. parts of words, are represented by lower case letters. Hence, together with an
expression of the syntactic category noun (N), a new type of the category nominal
attribute is created, i.e. N IN, as in (vaterlichesNIN HausN)N (see above for the analogous
function of the genitive ending in Lat. (patr- 51" -is(Nt N)t siJN1N) Conversely, the suffix
-heit of the type n/adj combines with an adjective to form a noun as in (dieseNPIN
FrechheitN)NP 'this impertinence' . A suffix like -er, in turn, combines with verb stems
to give nouns, hence N/ St v, as in (derNPIN FahrerN) NP 'the driver', whereas the suffix
-ier(en) combines with a noun to produce a verb or a verbal stem, St) n, as in tackier- 'lacquer', which together with an inflectional element becomes a verb: (lackier-s 1v
-env1s1Jv 'to lacquer ' or a specific verb form as in
(Maria NP ((lackier-stv -te((s/NP)/NP)/st)(S/NP)/NP das Auto NP)SINP)s

In this manner, the syntactic-functional implications of word-formation become quite
evident.
It is, by the way, an essential feature of natural language that the potential syntactic
functions of a sign are predetermined by its part of speech to be used only in special
structural contexts. This is why even in functionalistic categorial grammar not only the
function but also the part of speech of a syntactic unit has to be named. In cases where
the part of speech does not change, as in the aforementioned Automobil-ist 'motorist',
the argument and the value of the functions have to be differentiated by other means in
order to label the suffix formally as the head functor, as, e.g., (Automobil-n, -istnytnJN
It is essential that the linguistic expression here, as opposed to attribution (see below) is
assigned a new head. Semantically, this suffix could be defined as follows: -ist is a
functor that - amongst others - combines with the name of an instrument to form the
term for the user of this instrument, for example, as in Maschinist ' machinist', especially
with musical instruments as in Flotist ' flutist' , Saxophonist 'saxophonist', Hornist 'hornist', etc.
Categorial grammar is particularly suitable for representing the semantic structure
and, in general, the semiotic status of bound morphemes such as suffixes and flexives.
The loss of semiotic autonomy influences the meaning and the reference of such signs
and, thus, their symbolicalness. The bound morphemes of a complex sign are both morphologically and semantically interdependent, which has interesting consequences for
the relation between form and content. The meaning of functional morphemes must
always be formulated with respect to the respective stem complement; it is impossible
to assign them a meaning or denotation in isolation. For example, the German suffix -er
which forms agent nouns, as in Fahrer 'driver' or Lehrer 'teacher', does. not simply
mean '(male) person' or functionally speaking 'doer' or 'agent', but rather '(male) per-
117
PA118
11 8
son, carrying out the activity denoted by the stem'. This means that a suffix only takes
on a denotation together with a stem and only together with a stem does it function as
a sign. Hence, suffixes can be viewed as synsemantica. As opposed to the nominal head
of the German compound Apfelbaum 'apple tree' the suffix in Fr. pomm-ier does not
mean simply 'tree', but 'tree carrying the fruit denoted by the stem' based on the general
morphosemantic pattern 'object which has something to do with the referent of the
derivational base'.
Representing a suffix or an ending by means of a functor that together with a stem
forms a morphologically closed term of a certain category avoids the need for assigning
meaning to isolated morphemes like the suffix -ier or German -ung in Wohnung 'flat'
or Italian -oso in sabbioso 'sandy'. Semantically, these suffixes are functors that together
with the meaning of a stern produce the complex meaning of a noun. The meaning of a
functor is generally defined with respect to its function, just like its rnorphosyntactic
category. In the compound Apfelbaurn, however, the head constituent -baurn does not
have the same function as the suffix -ier. Rather, it is a complement of the attributive
functor Apfel- of the type N IN (see section 4); hence, the "active" operator in the case
of the compound is the attributive determinans and in the case of suffixational derivation
the head, i.e. the determinatum. The discussion of a possible formal representation for
the semantic structure of morphemes is not part of the present article.
3.2. Prefixation
Whereas suffixes are normally unequivocal cases of synsemantica that only function as
a sign in their morphological context, prefixes in comparison show a more complex
behaviour. Above all, prefixes are not heads but attributes, in fact attributes of entire
words, as morphological analysis makes clear. As opposed to suffixes and inflectional
elements, prefixes do not combine with stems but attach to morphologically complete
words (forms), as in German (ein-v!V (Jahr-st. -envistJv)v 'to drive in' . In other words,
prefixations behave like compounds. Thus, in contradistinction to suffixes, prefixes are
not morphologically necessary constituents of the whole expression. They do not represent the expression, but constitute, just like syntactic attributes, an ornissible category
that extends the expression and, hence is of the type X IX. This analysis implies that a
prefix like ein- would have to be classified as an adverb, having the same function as
the adverb hinein 'into' in (hineinv!V fahrenv)v or even as schnell 'fast' in (schnellv!V
fahrenv)v 'to drive fast'.
In fact, the prefix ein-, like most of these elements, originates from an adverbial
attribute of this verb which, however, has developed into a quasi bound morpheme, since
ein- in German is hardly ever used as an adverb. Furthermore, especially in combination
with -fahren, ein- has undergone a clear semantic lexicalization. If, by means of a histori cal process, ein- has lost its status as a free morpheme, it can in principle no longer
combine with a free morpheme or a word; ein- combines only with a bound stem and
forms with it an extended stem. In thi s respect, it is a stem-forming expression like a
suffix, with the difference that it does not appear as a head functor, but as a dependent
functor of the stem, representing type Stxf Stx, leaving open the slot to the right of the
stem to be filled with suffixes or inflectional elements.
PA120
120
Here again complements can be found on the left side of the tree and the functors on
the right, with the exception of the attributive functor de-. Derivations, both suffixations
and prefixations, are thus generally predetermining and right-headed. The tree diagram
shows that prefixations can always be analyzed as prefix+word, de+formalizzazione,
even if the prefix in the present case, as already mentioned, is not an immediate constituent of the whole derivation morphosemantically but a structural partner of the verb stem
formalizza- or the verb formalizzare.
Prefixes are, like suffixes, sensitive to parts of speech. For the most part, they only
determine representatives of certain parts of speech. The Romance privative prefix in-I
im-, for instance, combines with adj ectives, but - for semantic reasons - not with verbs.
This is why the derivational base of Italian in-dipendente 'independent' is not a verb,
hence not an element of the paradigm of dipendere '(to) depend', but the adj ective
dipendente. Likew ise, the German adjective abhangig [compare English dependent] has
its privative counterpart unabhiingig [independent], whereas the verbal form abhiingend
[depending] cannot form *unabhiingend [* independing], since there is no verb *unabhiingen. The form passend [compare fitting] on the other hand has. mutated into an
adjective and thus allows the derivation of unpassend [unfitting].
As opposed to attributes in free syntax, prefixes can change the valency of their base,
as in the following case (in simplified notation):
(dringtv?!PP (durchpp/NP das HolzNp)pp)vp -
(durchdringtv?!NP das HolzNp)vp
in phrases like Wasserdampf dringt durch das Holz - Wasserdampf durchdringt das
Holz 'Water vapour penetrates the wood '. This morphologization can be represented by
a so-called functional composition in which the functor dringtvp/pp is applied directly
to the immediately following functor durchpp/NP resulting in the function value durchdringtv?!NP> which then together w ith a NP forms a verb phrase, formally: VP/PP PP/
NP= VP/NP. Restructuring a sequence a) into a sequence b) as in:
a)
b)
(head functor 1 - (head functor 2 - complement)compl) ((head functor 1 +head functor2 )runct - complement)
in which both head func tors merge into one, which now makes the original complement
of the headfunctor 2 its own, can be analyzed and represented as functional composition.
In this article, however, this important operation in the field of categorial grammar cannot be discu ssed in further detail (cf. Dowty 1988). Similar examples are, e.g., fiihrt um
den Baum - umfiihrt den Baum 'drives around the tree' or jliegt iiber das Haus uberjliegt das Haus 'flies over the house '. Such an impact of an attributive functor on
its head is only possible because the result is not a syntactic structure, but a new word
or a new stem with a special meaning which can include a change in syntactic valency.
4. Compounding
On the background of the affinity between prefixation and compounding, the question
arises of how to di stinguish these two patterns of word-formation in our formal model.
PA121
121
The first constituent of a German compound noun like Gartenzaun 'garden fence' is
without doubt an attributive functor and, thus, could on first glance be assigned the same
category as the adjective in kleiner Zaun 'little fence', namely NIN. However, this again
would not make it clear that the value of the adjective functor is a nominal syntagma,
whereas the value of the adnominal noun is a compound, hence a word. In order to
clarify that we are dealing with innerlexical processes within the scope of a word, the
symbol Wx (Word,J for morphologically closed forms of the pa1i of speech x can be
used especially for parts of compounds in analogy to the minuscules in St 11 for verbal or
St 11 for nominal stems. For the sake of convenience, this meaning has to be assigned to
the minuscule x itself, so that the determinans of our compound of the type W 0 / W0 is
categorized n /n. If the part of speech of this nominal attribute should be expressed as
well, it has to be assigned the categorial symbol 11.11111 , representing a noun making a
(complex) noun together with a noun - in general notation: Xx/x Thus, the compound
has to be assigned the following morpholexical structure: (Garten 0010 Zaun 0 ) 0 'garden
fence', whereby all expressions of the category n in the sense of a subset are also
elements of the category N.
Moreover, nouns behaving as word forming functors can also combine with representatives of other parts of speech, in which case they then belong to the type Yxlx This
way, we have in German messerscharf 'razor-sharp', aalglatt 'slippery as an eel' or
(schnee"djtadj weij3a dj) adj 'snow-white' a noun forming together with an adjective an adjective (as distinguished from the nominal compound (Schnee""'" Flocke 0 ) 0 'snowflake' in
which Schnee specifies a noun), and in (riicken 11 vi v schwimmenv) v 'to backstroke' a noun
combines with a verb to form a verb. The combination of nouns with adjectives with a
specific inherent valency, as, e.g., armlreich (an) 'poor/rich (in)' or freilvoll (van) 'free/
full (from/of)' in compounds like schneearm 'snow-poor, i.e. snowless' or bleifrei 'lead
free' has to be distinguished from the pattern of schneeweij3. In these cases the adjective
has to be defined as a suffix-like head functor taking a noun as a complement in order
to form a complex adjective and is thus of the type (y/ x)y: (schnee0 arm(adj /n)ad) adj . In
this case we put the specification of the function in the first position, followed by a
subscript that specifies the part of speech, because an adjective like arm or frei is - in
principle and not only in this context - a head functor of the type adj/n or in free syntax
Adj/PP (compare also Lat. patrisNINN as a nominal noun attribute).
Whereas in these compounds the relation between noun and adjective is determined
by the valency of the adjective, in the case of schneeweij3 it is the nominal attribute in
its function as an object of comparison which establishes the syntactico-semantic relation
between the constituents. Semantically speaking, this is equivalent to the fact that
schneeweij3 describes a kind of weij3, while schneearm however is not a kind of arm nor
is bleifrei a kind of.frei. In schneearm, schnee- does not describe a property of arm, in
schneeweij3 however it does describe a property of weij3. Graphically, the two compound
adjective types behave as follows:
adj
~
nadj/adj
adj
vs.
adj
sch nee-
-weifJ
(adj/n)adj
I
blut-
-arm
PA122
122
Finally, the much discussed type French compte-gouttes, Italian conta-gocce or Spanish
cuentagotas 'dropper, drop counter' based on a verb-object-relation, remains to be mentioned. The combination of a morphologically bound head functor with a word instead
of another morphologically bound element is rather particular. Such a combination is
common at best with attributes in the form of a prefix, as, e.g., in Fr. re-voir 'to see
again', It. s-comodo 'incovenient, uncomfortable' or Germ. ent-lassen 'to dismiss'. A
similar morphological pattern can be found in German forms like Schlag-arm ' punch
arm', Wett-biiro 'betting office', Kauf-haus 'department store; lit. buy house' or Lachtaube ' laughing dove' and Priigel-knabe 'whipping boy', in wh ich the verbal base functions as a predetermining attribute. In any case, all these constructions do not meet the
classic definition of the compound as a complex word consisting of several words.
In the case of Fr. compte-gouttes or compte-tours ' revolution counter', although parallel in their surface structure, the verb stem, if this constituent is to be interpreted as a
stem, is a functor head governing the direct object gouttes just like garde- in gardemagasin 'quartermaster', garde-malade 'home nurse', etc. - and the structure of the
word should be represented accordingly. Now the noun gouttes 'drops' (used without
the article) at the level of word-formation can be interpreted as a noun phrase in the
function of a direct object. We thus are dealing with a kind of morphological NP which
according to our convention has to be represented by minuscules: -gouttes0 P; the whole
construction then is as follows: (compte-s 1" ... -gouttes 0 p) 0 . In this perspective, the interpretation as a stem is preferable to that of a finite verb form as, e.g., 3rct ps. sg., since
the latter is not a typical element of word-formation but combines at the syntax level
with an entire NP les!des gouttes. Especially in the field of word-formation and generally
in morphology, the descriptive and analytical power of categorial grammar becomes
evident in its ability to discover and to handle the most detailed morphosyntactic and
morphosemantic differentiations (cf. Wandruszka 1999).
5. References
Dowty, David
1988
Type raising, functional composition, and non-constituent conjunction. In: Richard T.
Oehrle, Emmon Bach and Deirdre Wheeler (eds.), Categorial Grammars and Natural
Language Structures, 15 3- 197. Dordrecht: Reidel.
Hoeksema, Jack
1985
Categorial Morphology. New York, NY: Garland.
Hoeksema, Jack and Richard D. Janda
1988
Implications of process-morphology for categorial grammar. In: Richard T. Oehrle, Emrnon Bach and Deirdre Wheeler (eds.), Categorial Grammars and Natural Language
Structures, 199-247. Dordrecht: Reidel.
Moortgat, Michael
1988
Mixed-composition and discontinuous dependencies. In: Richard T. Oehrle, Ernrnon
Baich and Deirdre Wheeler (eds.), Categorial Grammars and Natural Language Structures, 319- 348. Dordrecht: Reidel.
Morrill, Glyn V.
2011
Categorial Grammat: Logical Sy ntax, Semantics, and Processing. Oxford: Oxford University Press.
PA123
123
9. Word-formation in natural morphology

Reich!, Karl
Categorial Grammar and Word-Formation. The De-adjectival Abstract Noun in English.
1982
Tiibingen: N iemeyer.
Wandruszka, Ulrich
Probleme der neufranzosischen Wortbildung. Tiibingen: Niemeyer.
1976
Wandruszka, Ulrich
1997
Syntax und Morphosy ntax. Eine kategorialgrammatische Darstellung anhand romanischer und deutscher Fakten. Tiibingen: Narr.
Wandruszka, Ulrich
Uber den Nutzen formaler Modelle der natiirlichen Sprache. Papiere zur Linguistik
1999
60(1): 31 - 50.
Wandruszka, Ulrich
2007
Grammatik: Form - Funktion - Darstellung. Tiibingen: Narr.
Werner, Heinz
1990
Kann die Montague-Grammatik einen Beitrag zur diachronischen Linguistik leisten?
Folia Linguistica Historica 11(1-2): 147- 182.
Wood, Mary McGee
1993
Categorial Grammars. London/New York: Routledge.
Zifonun, Gisela, Ludger Hoffmann and Bruno Strecker
Grammatik der deutschen Sprache. 3 Vol. Berlin/New York: de Gruyter.
1997
Ulrich Wandruszka, Klagenfurt (Austria)
9. Word-formation 1n natural morphology

1.
2.
3.
4.
5.
6.
7.
Introduction
From natural phonology to natural morphology
Functionalism and natural morphology
Naturalness and the lexicon
Parameters of naturalness in word-formation
Conc luding remarks
References
Abstract
Natural morphology is not a descriptive model of word structure but an explanatory
framework designed to evaluate morphological techniques from a fanctionalist p oint of
view along principles deduced from Peircean semiotics. Word-formation is seen as largely determined by extralinguistic factors such as communicative needs, cognitive exigencies and the conditions on speech performance and processing. Universal pref erences
are formulated that are reflected in cross-linguistic distribution, order of acquisition,
diachronic change and other fields of evidence, but predictions are relativized by typological and language-specific criteria.
PA124
124
1. Introduction
Natural morphology has developed in the late 1970's and early 1980's from mainly two
different sources. The first of these was a transposition of basic assumptions from natural
phonology onto morphology, whence the epithet "natural". The second is the functional ist approach established by the Prague school, with its concept of markedness, which
can be seen as the converse of naturalness, although with the important difference that
markedness is conceived as absolute and binary, while naturalness is seen as a relative
and gradual quality by the proponents of naturalness theory. For the sake of characterizing the theoretical stance of natural morphology properly, paying regard to its initial
motivation, it is necessary first to very briefly recall the basic tenets of natural phonology, inasmuch as they embrace a good deal of the idea of naturalness as a technical
notion in linguistics (section 2). In a second step, the functionalist foundations of natural
morphology will be sketched (section 3), leading to considerations on lexical enrichmen t
as the primary function of word-formation (section 4). From this outset it will be possible
to delineate the specific implementation of the principles of naturalness as applied to the
phenomena of word-formation (section 5).
2. From natural phonology to natural morphology

In natural phonology, as first conceived by Stampe ( 1969 and 1979 [ 1973 ]), the claim
was made that phonology can be reduced to a set of universal processes reflecting the
phonetic capacities of language users. In spite of their phonetic motivation, these processes are in themselves not physiological, but mental operations. Still, they are not to be
confused with the so-called production rules in formal models of grammar and in artificial intelligence, nor with spellout rules generating target forms. Natural phonological
processes are not generative, but are conceived as mental substitutions relating the intention of a speaker to the "actual rumble" of his speech (to use a Sapirean form of wording). The substitutions work both ways, relating not only the intention to the perceived
output but also the incoming signal to its perception in terms of distinctive units. These
processes are abstract in the sense that they operate at the level of representations specifiable in terms of distinctive features. An aspect of this particular notion of process that
has always been emphasized by the proponents of natural phonology is their role as
constraints in terms of production, as well as perception, of utterances: "These automatic
processes not only govern automatic alternations or add allophonic features. Together
with the prosody, they account for the native 'accents' of speakers, they constrain the
perception and pronunciation of second languages, and they are the basis of systematic
variation and change" (Donegan and Stampe 2009: 6; see also Donegan 2001).
What makes phonological processes "natural" (in spite of them being called "phonological") is that their motivation is entirely phonetic, which means that they are based
on extralinguistic factors and settings (i.e. human nature, tout court). This makes them
universal at the same time, on the premise that the anatomical and (neuro)physiological
endowment of humans is roughly the same for all exemplars of the species, at least as
far as the handling of what is usually encoded in terms of phonological features is
concerned. Notice that for phonology (and for morphology, as will be seen) the theory
PA125
of naturalness circumvents the innateness problem by outsourcing what is universal from
the abstract system, i.e. from the grammar of a language, to the apparatus of its execution. No innate explicit system of abstract entities is assumed except the natural processes, but these are unordered at first, there being no such thing as universal grammar in a
Chomskian spirit that would keep them in a definite configuration. All the structuredness
and patterning making up the phonological system of a language is the result of language
acquisition, triggered and directed by communicative experience. What makes phonological systems differ across languages is not a difference in the processes as such but a
different pattern of selection from the universal pool of processes, with weighting mechanisms ranging from complete suppression over restriction to the subtleties of ordering
of particular processes (for an overview of process application see Donegan and Stampe
1979: 145- 158).
Since the set of universal processes has never been explicitly defined by the proponents of natural phonology, nor have they ever supplied a technical apparatus that would
make the theory applicable as a descriptive model, they can easily be criticized for trying
to build a house from the roof downwards instead of having erected firm walls at first.
It is hard to overlook, however, that much of the technical affordances of optimality
theory could count as an operationalization of the principles of natural phonology, although the proponents of both approaches are far from lying in each other's arms and
the differences with respect to the foundation of constraints are considerable (for a comparison see Tobin 2009).
Inasmuch as natural phonology emphasized those aspects of the "sound shape of
language" (to use a Jakobsonian form of wording) that are ultimately rooted in phonetics,
i.e. outside Saussure's chessboard, the contrast to the more arbitrary properties of phonological systems was made very explicit in terms of the dichotomy of processes versus
rules (cf. Dressler 1985b: 11 - 24; Donegan 2002). Rules, as opposed to natural processes,
were characterized as conventional, language-specific, extrinsic operations that have to
be specifically learned (Dressler 1984a: 30; cf. the similar distinction between automatic
and nonautomatic rules made in elaborations of generative phonology, e.g., Kenstowicz
and Kisseberth 1979). In short, the parameter of rule naturalness could have been assigned the labels "phonetically motivated" on the positive side and "morphological/lexical" on the negative side. Diachronic change might then make a process slide away from
the " natural" end of the scale towards the "unnatural" domain, potentially ending up as
a morpho(no)logical rule or an ossified alternation in the lexicon. The paralEel to lexical
phonology (and morphology, cf. Kiparsky 1983) with its layers of cyclic rules and postlexical component is very close in this respect, but, as in the case of optimality theory,
the main difference lies in the foundation of the principles rather than in the principles
themselves.
From what has been briefly sketched in this section it could be expected that morphology should remain outside the scope of naturalness theory, being concerned with everything about the shape of words that was not in some way stimulated and at the same
time constrained by the circumstances of linguistic performance. In fact, the founders of
natural phonology had always concentrated on core issues in phonology and on the
phonetics-phonology interface; they never considered extending their approach to other
components of language (cf. Hurch and Rhodes 1996: XI). In the European version of
natural phonology, however, there has been a strong tendency to account for interfaces
and transition phenomena, in particular morphonology, and to apply the model to dia-
125
PA126
126
chronic change (cf. Dressler 1982). The interaction between phonology and morphology
was thus something to be taken up in a more general approach, built on principles that
should hold for all components of the language system. At this point the second line of
thought joins in that stems from functionalist approaches to language structure (most
notably Jakobson, but also Benveniste and Seiler, cf. Dressler et al. 1987: 3), according
to which all material components of the language system are determined by performance
factors, even though to a disputable degree, under the rule of the principle "form follows
function". Whether this principle is a "sound premise" or a "problematic" one in the
sense of Friedrich (1975: 199-200) is still likely to remain a matter of dispute for some
time to come.
3. Functiona lism and natural morphology

Performance factors, if that notion is taken in its widest sense, comprise all kinds of
variables, from the neurophysiologically-based mechanisms of signal processing at the
inward-directed parts of the communication network to the socially and situationally
governed realm of pragmatics at its outward-directed branches. Naturalness theory can
be seen as a theoretical approach that tries to unite these very different dimensions under
a common heading, with "naturalness" as an evaluation metric for assigning benchmark
values to the structural properties of languages according to their contribution to the
overall functioning of the communication process. For such a theory, in order to be noncircular on the one hand and avoid over-determinism on the other, it is essential to be
non-reductive and to allow for multiple interactions of particular forces. The subtitle of
one of the earliest monographs in this tradition (given here in translation from the German original) is illustrative of the breadth of the approach, with its enumeration of
phenomena to be accounted for by what was at this time still natural morphology in
statu nascendi: "Reduplication, echo words, morphological naturalness, haplology, productivity, rule telescoping, paradigmatic levelling" (Mayerthaler 1977). While Mayerthaler joined the naturalness enterprise with a pronounced consideration for universal
markedness theory and a tendency towards axiomaticity and formalization (cf. Mayerthaler 1981: 40- 41, 60- 64; in English: Mayerthaler 1988), Dressler had been developing
what he at that time called a "polycentristic" theory of language, by which he meant
"a lexicon and an item-and-process grammar which contains several semi-autonomous
components" (Dressler 1977a: 13), one of them being word-formation. " For each corresponding component of language man is endowed with universal, 'natural tendencies'"
(Dressler 1977a: 13), from which preferences directing linguistic behaviour can be deduced. Before reviewing the tendencies affecting word-formation in the following sections, some further remarks on the general character of the "naturalist program" seem
appropriate.
Since natural morphology is an explicitly non-formalist framework (pace Mayerthaler's aspirations), its claims and proposals can also be seen as a counterbalance with
respect to the treatment of morphology in the generative tradition. After a period of
relative neglect, word-structure had become an issue in generative theory in the form of
the lexicalist hypothesis (Chomsky 1970; Halle 1973 ), and since then there have been
constant endeavours to accommodate derivation and compounding in the various models
PA127
of universal grammar over the last decades, more or less by mapping principles of
syntax onto word structure, but in any case maintaining the fundamental assumptions
that grammar is an autonomous mental technique for handling verbal signs and that the
potential of this technique as well as the constraints on it are of a strictly formal nature
that can be accoU1nted for in terms of computational algorithms (see article 7 on wordformation in generative grammar).
Natural morphology not only denies the autonomy of the language system as such
but also conceives the system itself as epiphenomena!, being the result of a compromise
between a multitude of partly antagonistic functions that cannot all be fulfilled at the
same time (cf. Dressler 2002). Moreover, as mentioned already, these functions are assumed to be motivated or determined to a large extent by extra-linguistic factors, but the
range of possible morphological phenomena is underdetermined by the entirety of these
factors. For instance, the processing of utterances is subject to neurophysiological capacity, from which it follows that for the chunks to be processed in real-time verbal communication there are certain preferred states in terms of size and complexity. All other
things being equal, a word form should be long enough as to be perceivable as a separate
unit of speech distinguishable from others, and short enough as to allow for concatenation with other word forms, at least up to the phrase-level, within a single breath-group
or intonational phrase (cf. Dressler 1985c: 53). Since in reality no other things are ever
equal, though, variation within these margins is driven by factors that may challenge the
limits set by one particular factor. For example, in the cases of multiple compounding
and freezing, which usually result in constructions exceeding the preferred length of a
word form (see article 24 on multi-word expressions), it is the function of conceptualization which overrides the processing-based preference for well-portioned chunks, while
in the opposite case it is the function of phonetic parsimony which may reduce a word
form to the size of a single phoneme or reduce a marker to zero, thus exceeding the
limits of preferred size of a signans at the other extreme. Notice that both of these factors
are non-morphological, yet morphology is strongly affected by them.
By taking the multitude of communicative functions as its basis of reasoning, natural
morphology is not a theory that can make strong predictions about the possible structure
of languages but rather a theory that explains why the range of variation among language
structures is not as perspicuous as the proponents of an autonomous innate universal
grammar would like to see, and why unstoppable diachronic change keeps transforming
these structures all the time and everywhere. The key aspect of natural morphology as
compared to other approaches (except optimality theory, but with important reservations,
see article 11 on word-formation in optimality theory) is that it aims at providing an
evaluation scheme for morphological processes according to which asymmetries in their
distribution can be accounted for. However, a significant restriction on the liability of
predictions based on the principles of natural morphology consists in the overall architecture of the approach. Universal naturalness is the overarching layer, based on systemindependent and extra-linguistic factors that can be captured in terms of markedness.
The preferences directed at this level can be thwarted, though, at the level of typological
adequacy (cf. Dressler 1985c), where the intercomponential interactions of preferences
are situated. Finally, at the level of individual languages, system-dependent naturalness
may prevail (cf. Wurzel 1985), which is directed mainly by such factors as congruence
in patterning, implicational relationships within paradigms and the overall frequency of
127
PA128
128
types and tokens (and is therefore more relevant to inflectional morphology than to
word-formation, cf. Dressler 1987: 121).
4. Naturalness and the lexicon

According to its explicitly functionalist approach in terms of a means-end model of
language, natural morphology embarks on the analysis of word structure by first asking
for the function of the phenomena under scrutiny. In the case of word-formation this is
lexical enrichment in the first place, i.e. the supply of lexical items expressing either
new content (e.g., in English, the coining of the compound dreadlocks upon the introduction of rasta hairstyle in popular culture) or differentiating existing content (e.g., the
coining of deathhawk, denoting a particular type of mohawk hairstyle). It follows from
the modularity of the language system that the basic function of expressing content can
be fulfilled by more than one of its components. For example, instead of by means of a
new word (in this case a compound), the content expressed by such items as dreadlocks
or deathhawk could in principle as well be expressed by syntax in terms of circumscription with existing lexemes (as in a dictionary-type explication), or by the lexicon itself,
without intervening morphology, through items borrowed from another language. However, both alternatives have serious shortcomings (longwindedness in the first case, possible unavailability in the second, above all), so that the preference for fulfilling the
onomasiological function by the creation of a naming unit (a term taken up from Mathesius and established by Stekauer 1998; see article 6 on word-formation in onomasiology)
with the hdp of morphology turns out to be the most natural solution. However, for a
serious theoretical assessment of such a conclusion it would be necessary to explicitly
specify its premises, i.e. to assure that what here has been sloppily referred to as longwindedness is indeed a property of linguistic expressions that can be gauged in terms of
enhancement or impediment with respect to the success of a speech act in communication. The factors of frequency and entrenchment are of main importance here, and natural
morphology is well disposed to link in with usage-based models (e.g., Bybee 2013) and
quantitative approaches to complexity and economy across components (e.g., Fenk and
Fenk-Oczlon 2008).
As far as methodological rigour is concerned, it is important to mention that natural
morphology does not have the tendency to exclude the occurrence of types of constructions that are considered unnatural or dysfunctional by virtue of its own principles. By
postulating preferences instead of absolute universals it avoids the problem of immediate
empirical fallibility that would arise once an instance of something that has been excluded on theoretical grounds turns up (cf. Dressler 1988). Thus natural morphology does
not exclude the use of circumscription as an onomasiological technique, but just maintains that this is not likely to be the main solution for the task of naming concepts in a
given language system (cf. Seiler's 1975 dichotomy of "description" versus " labeling").
Examples of descriptive expressions can easily be found in English, also in the semantic
sphere of hairstyles, e.g., high and tight or short back and sides, but this is not the
prevailing pattern, apart from the fact that such descriptions are never exhaustive. The
same holds for borrowing: it is obvious that borrowing occurs, especially in semantic
fields like the one adduced here for illustration (consider, e.g., odango, from Japanese,
PA129
or chignon, from French), but it is highly improbable that a language community will
resort actively to borrowing as the main source for new expressions (there may be exceptions to this for technical terminology of languages for special purposes, though, see
article 92 on foreign word-formation, language planning and purism and article 127 on
word-formation and technical languages, and probably also for language death, cf. Dressler l 977b).
The example of lexical enrichment in the semantic field of hairstyles may appear
simplistic, but it has been chosen deliberately in order to make the typical line of reasoning in natural morphology as clear as possible. The argumentation is strictly deductive,
departing from the definition of a goal in communication and then trying to evaluate the
various possibilities of its attainment along the parameters of naturalness. Once a solution is assessed as more natural with respect to others, the reasoning can move on stepwise by further differentiating the spectrum of possibilities for attaining the immediate
goal, once the alternatives to it have been discarded. So far it has become clear that
circumscription does not enrich the lexicon (by definition, since it is syntactic; but notice
that our examples above are not full phrases but rather a sort of complex labels, ready
for becoming lexicalized), while borrowing does enrich the lexicon, but not in an optimal
way, since it introduces unmotivated items that may require adaptation (and it overthrows
the identity of a language when it prevails).
For the creation of a naming unit that does not exceed the limits of wordhood and is
not borrowed from another language, there are basically two possibilities: either a new
formal expression is coined by means of morphology, as in the case of dreadlocks or
deathhawk, or a new meaning is attached to an existing lexeme by means of metaphor
or metonymy (see article 61 on word-formation and metonymy). Also for these solutions
examples can be found in the coiffure-related vocabulary of English: ponytail is a metaphoric transposition on the basis of shape, while iry league is a metonymic transposition
based on the social group of people displaying a preference for the respective haircut
(also eponymous hairstyles like pompadour or rachel belong to this type). As in the case
of circumscription and borrowing, it can be said that metaphor and metonymy, although
pervasive in the lexicons of all languages (and essential to cognition, cf. Danesi 1995),
are not natural techniques for lexical enrichment, because they produce polysemy and
thus violate the preference for biuniqueness based on the very general semiotic principle
of "one meaning one form" (but see Marzo 2013: 60- 73 for a refined account). Moreover, the notion "lexical" in the term lexical enrichment cannot be understood to refer to
meaning only, otherwise it would be more adequate to speak of "semantic enrichment".
If a language would resort to nothing other than metaphor and metonymy for naming
new content, it would become more and more cumbersome for speakers to encode meaning properly and recover intended meaning from form in real-time communication, although "language users seem to have much fewer problems with polysemy than many
lexicographers and linguists" (Ungerer 1999: 309). An implicit principle underlying this
line of argumentation in natural morphology is akin to Kant's categorical imperative: for
judging the relative naturalness of a linguistic operation, imagine what would happen if
it were made to apply across the board. This monopoly principle, as it may be called, is
also useful as an antidote against precipitate typological generalizations, as has been
shown by Dressler (1985c: 53). As a matter of fact, the monopoly principle was already
envisaged by Saussure (2011 [1916]: 133) in the chapter on the "mechanism of lan-
129
PA130
130
guage'', where it is stated that unrestricted arbitrariness "would lead to the worst sort of
complication".
Semantic transposition, if applied across the board, would inflate the lexicon with
polysemous items instead of enriching it. However, a moderate degree of application of
this onomasiological technique seems to be common, and there are types of regular
polysemy that can be found with considerable cross-linguistic frequency and uniformity,
like the relation between designations for plants and their respective fruits, or between
containers and their contents (Murphy 2010: 89- 90). This kind of regularity seems to
be more typical for metonymic transpositions than for metaphorical ones (Cruse 2006:
15 1), but the boundary between lexical meaning and pragmatic inference is fluid in such
cases (cf. Nun berg 1979), so that nothing conclusive can be said about how much of the
task of word-formation can be taken care of by semantics alone.
The remaining candidates to be judged for their aptness to supply a language with
items for fulfilling the most basic of its functions, viz., the referential function, are wordcreation, composition and derivation. The first of these falls outside the scope of wordformation, ff this is defined as the creation of new words from existing ones. Still,
leaving aside most obvious presuppositions, one could ask why lexical items are hardly
ever created from scratch, like the English word blurb, which is reported to have been
invented by the American humorist Gelett Burgess (cf. Bloomfield 1933: 424) and eventually gained general acceptance. It could be claimed that a new form for a new concept
would represent a state of affairs that should not be considered dispreferred in terms of
isomorphism and biuniqueness. However, this would lead a long way back, one step
behind Saussure, who had called attention to the importance of lexical motivation (cf.
the quotation above). Furthermore, new concepts are hardly ever new in all respects:
there are relations of hyponymy, meronymy, etc., accommodating any new content in an
existing network of semantic specifications. Unmotivated coinage can be at work, of
course, in well-defined functional niches like the creation of brand names, but a closer
look at such items usually reveals that they are hardly ever unmotivated in all respects
(cf. Piller 1999; see article 26 on word-creation and article 124 on word -formation and
brand names). Special standards hold, of course, for poetic language (cf. Dressler 1981,
1994).
In spite of Marchand 's extensive use of the notion "motivation" (cf. Kastovsky 2005),
this term has not gained much currency in English linguistic metalanguage. Definitions
are lacking in Bauer's (2004) and Cruse's (2006) glossaries; also Haspelmath and Sims
(2010) do n.ot include it in their otherwise quite comprehensive list of definitions (but
see Nanny and Fischer 2001; Radden and Panther 2004; Panther 2008 and Lehmann
2007 for recent appraisals; see also article 10 on word-formation in cognitive grammar
and article 56 on motivation, compositionality, idiomatization). Haiman (1985: 2) uses
the term in a different sense (''ways in which the linguistic form is a diagram of conceptual structure, and homologous with it"). In natural morphology, motivation has the
prominent role of a main function of word-formation, viz., to create complex items
whose meaning can be recovered from the meanings of their respective parts, according
to the principle of compositionality (on which see Cruse 2011: 44; cf. also Klos 2011:
35- 59). By establishing regular formal relationships between semantically related items,
motivation a lso facilitates storage and retrieval in the declarative compartment of longterm memory (on the distinction of declarative versus procedural memory see Ullman
2004: 244- 248). From this follows a general preference for transparency as opposed to
PA131
opacity of complex word forms that is further differentiated in natural morphology in
terms of semiotically-based naturalness parameters, which will be discussed in the following section.
5. Parameters of naturalness in word-formation

Natural morphology assumes that each (sub-)component of the language system harbours
system-specific tendencies that are not only potentially in conflict with the tendencies of
other (sub-)components but also partly antagonistic within each (sub-)component itself.
For the lexicon, a number of tendencies have been identified that cannot be all fulfilled
at the same time (cf. Waugh and Newfield 1995): motivation, which warrants the semantic
interpretability of complex words, stands against a tendency towards autonomy of complex words. This tendency, which manifests itself in idiomatization and lexicalization,
can in turn not be fulfilled without detriment to another tendency, that of enhancing the
solidarity between semantically related words in lexical fields. The multiple antagonisms
inherent to the (sub-)components of the system is a main source of variation and makes
language resemble a perpetuum mobile, with the impulse for diachronic change already
built into the mechanism. This is in accordance with the conclusions to be drawn from
the historical study of languages, which boil down to the insight that it is not change as
such which has to be explained but rather its absence (cf. Shapiro 1991: 11- 12).
Since semiotics is at the basis of all evaluative considerations in natural morphology,
the parameters or scales that have been set up in this approach are all directly or indirectly derived from semiotic principles, of which iconicity is the dominant and most general
one. In the following subsections only those parameters will be discussed that are of
immediate relevance to word-formation. In the interest of space, the use of examples
will be kept to a minimum.
5.1. lconicity
Our look at the semantic field of hairstyles would be incomplete if we neglected formations that do not fit to any of the types of expression considered thus far. For example,
p erm or afro are clippings (of p ermanent wave and afro-american, respectively) and as
such represent the process-type of subtractive derivation (see article 19 on clipping), as
opposed to incremental derivation. Providing that no change in meaning is expressed by
the clipping, as in the case of p erm, this kind of process can be considered "extragrammatical'', as has in fact been proposed (Dressler and Merlini Barbaresi 1994: 36- 41; cf.
also Dressler 2000b and Dressler 2005: 269-270). On the other hand, clipped forms
easily become lexicalized and acquire a meaning that differs from that of the full form.
This seems to be the case for afro, which has a specific meaning as a hairstyle, while
the full form may refer to anything being afro-american. Furthermore, afro is a noun,
while afro-american is an adjective (potentially convertible to a noun), so the clipping
here involves a change in word-class too, which is quite typical for grammatical but not
for extragrammatical derivation. These examples show that formations that belong to the
same type in one respect can behave very differently in other respects, which means that
131
PA132
132
the criteria for defining what is extragrammatical need to be refined beyond what has
been brought forward thus far in the literature (cf. Ronneberger-Sibold 20 10; for a summary of the problems and proposals see Mattiello 2013: 1-6). The fact remains, however,
that clipping, as popular as it may be in the creation of all kinds of expressions enriching
the lexicon, is hardly ever used regularly and productively as an exponent of a particular
semantic category (cf. Dressler 1984b; 2000a).
From the fact that lexical enrichment is the primary function of word-formation it
follows that this function is best fu lfilled if procedures are made available that can be
applied productively and multiply, and it is fulfilled even better if the productive and
multiple application of these procedures does not impair redundancy. Subtraction is definitely not a good candidate for such a job description, since it dimini shes redundancy in
direct proportion to the multiplicity of its application. If the monopoly principle mentioned in section 4 is applied to subtraction, the result is a language whose lexicon
implodes instead of expanding. Only in view of the criterion of optimal word-size may
subtractive derivation be rated as relatively natural (cf. Ronneberger-Sibold 1995), responding to the preference for bi- and trisyllabicity that seems to hold cross-linguistically: according to Dressler (1985c: 53), the optimal size of word forms, i.e. with bound
morphemes already included (but excluding compounds), is about two to three syllables
(for general considerations and a comparative survey of word length cf. Popescu et al.
20 13). However, while subtraction is preferred from the point of view of economy of
expression, at least as long as the need for redundancy in the lexicon is not compromised,
other semiot ically-based preferences stand against it, with the preference for constructional iconicity at the top.
5.1.1. Constructional iconicity
In section 4, isomorphism was invoked as a most natural state of the relation between
meaning and form, but without paying regard to the relations between the meanings of
particular items. This missing aspect is brought in by the parameter of constructional
iconicity (or diagrammaticity, in Peircean semiotic terms, cf. Johansen 1993: 98), whi ch
affords that an addition to the semantic specification of an item be accompanied by an
addition of phonic substance. In the case of word-formation this leads to the apparent
paradox that even diminutive meaning is most commonly expressed incrementally, i.e.
by means of affixes (also termed "additive coding", cf. Carstairs-McCarthy 1992: 219).
This is because the isomorphy of meaning and form as conceived in the notion of constructional iconicity does not refer to the semantic content as such, which is to be dealt
with under the Peircean rnbric of "image" (see section 5. 1.2), but to the abstract relation
between what is basic and what is derived. All this rests, of course, on the prerequisite
that meanings like 'smallness' or ' lessness' are cognitively represented in such a way as
to be encoded in terms of a semantic category; otherwise the lexicon of a language
would be full of unmotivated words like dwarf or puppy, where smallness is semantically
implied but not encoded categorially the way it is in derivational diminutives like booklet, starlet or state/et (English is, however, exceptionally parsimonious with productive
diminutive formations). Formations like diminutives and augmentatives have received
much attention in natural morphology (cf. Dressler and Merlini Barbaresi 1994 for a
PA133
thorough treatment), because they are related to a wider array of functional dimensions
(especially pragmatics, see article 62 on the pragmatics of word-formation) than category-changing derivation, like the formation of deverbal action nouns or deadjectival quality nouns, so that the interplay and trade-off between a variety of partially conflicting
preferences can be studied (cf. Zacarias 2006).
5.1.2. Imaging iconicity
In the functional evaluation of word-formation processes as envisaged by natural morphology, the parameter of iconicity is compartmentalized according to Peircean semiotics
into three types of iconic relations, diagrammaticity being agreed upon to be the most
important for word -formation (cf. Dressler 2005: 269). A second type of relation is
constituted by the most immediate kind of similarity between form and content, manifesting itself most obviously in onomatopoeia, but also, to a lesser degree, in phonaesthetic features of morphological markers, e.g., the preference for palatality in expressing
diminutivity or attenuativity (cf. Dressler and Merlini Barbaresi 2001, 1994: 93; see also
article 28 on affective palatalization in Basque). The parameter of imaging iconicity is
of course only applicable to a highly restricted set of semantic features. It works well in
the formati on of hypocoristics from personal names and pet names, in forms of endearment and the like, whereas it is highly improbable for a category like agentivity or
instrumentality to be ever expressed by means of phonic substance that bears a resemblance to its semantic content, since no speaker is likely to come up with an idea of
what agentivity or instrumentality "sound like". The imaging type of iconicity is also
very limited in its applicability because of intervening preferences of a different kind.
For example, the most iconic way of expressing duality would be full reduplication (see
article 25 on reduplication). However, duali ty ranks far below plurality in terms of semantic relevance and generality, so reduplication is much more often used for expressing
plurality than duality cross-linguistically (cf. Rubino 2005; Kouwenberg 2003). In a
study of reduplication in Bikol (Central Philippine), Mattes (2006) further elevates the
semantics of the reduplication process to a higher level of abstraction, arguing that
"change of quantity" is the semantic value covering all the different senses of fully
reduplicated forms in that language (cf. also Kouwenberg and LaCharite 2001 ). Furthermore, from the point of view of economy of expression, full reduplication should be
disprefen-ed, since it produces forms that easily exceed the optimal size of word forms.
Apart from the fact that the category of number is not a prototypical derivational category (cf. Dressler 1989: 9), this shows that the image-type, although the most iconic of all
types of iconic signs, can only be employed as a supplementary device in word-formation
and is not an optimal candidate for the task of lexical enrichment.
5.1.3. Metaphoric iconicity

The third type of iconic relation between content and form is more abstract and at the
same time more problematic than the other two. Under the assumption that derivation is
semantically additive, in that it introduces an extra specification to the meaning of a
133
PA134
134
base form, affixation is iconic, mirroring the semantic increment through the attachmen t
of phonic substance, while subtraction is clearly anti -iconic. This leaves the morphological process of modification, as in verbal believe versus nominal belief, somewhere in the
middle of the scale of diagrammaticity, because a change from a vo iced obstruent to a
voiceless one is hardly interpretable as an addition of phonic substance (but rather as a
loss of the feature [+ voiced] , which might even suggest a classification of this type of
modification as "mildly subtractive"). While it can be said that the modification of the
sound shape of a lexeme, as in apophony and metaphony (also termed "modulatory
coding", cf. Carstairs-McCarthy 1992: 2 19), is still iconic to the degree of signaling a
change in meaning (e.g., sing : song, full : fill), it is obvious that the way of signaling
this change inflicts a loss of formal identity on the respective lexeme. A morphological
operation that, in order to fulfill the task of producing semantically closely related items
from one another, resorts to the very technique by which semantically unrelated lexemes
are kept apart (e.g., string : strong, pull : pill) cannot be ranked high in terms of semiotic
convenience.
5.1.4. Scales of iconicity
The parameterization approach in natural morphology consists in setting up scales along

which formal operations on lexical items are ordered in proportion to their relative naturalness with respect to the parameters set up according to the Peircean model of semiotics. The gamut of possible formal operations on lexical items comprises the following :
they can be invented, substituted, combined, mixed, multiplied, enlarged, reduced or
reshaped. In terms of word-formation this corresponds to coinage, suppletion, compounding, blending, reduplication, affixation, subtraction and mod ification, respectively.
Signs can also be deleted, of course, but this is an operation of a different order and
certainly not of interest to the study of processes serving the function of lexical enrichment.
Even without recourse to any of the possible formal operations, the lexicon can still
be enriched by means of an abstract operation. This is the case of conversion, as in
verbal to host from nominal host (see a1iicle 17 on conversion). Since the absence of
any change in the shape of a word cannot seriously be interpreted as the signaling of an
addition or change of meaning, conversion should be ranked lowest on the parameter of
diagrammaticity. Instead, it usually appears somewhere in the middle of the scales that
have been proposed in the literature. For example, it has been argued by Crocco Galeas
(1990) that conversion is a kind of metaphor, although at a first glance this is not in
accordance with the Peircean definition of metaphor, which affords some sort of parallelism in representation between the content and the expression of the respective signs.
However, in conversion this parallelism is only invisible when word forms are analyzed
in isolation. Crocco Galeas proposes to extend the scope of formal properties of a word
form to its (morpho)syntactic behaviour, which leads to the conclusion that the semantic
difference between nominal and verbal host is paralleled by a difference in syntagmatic
collocation (cf. Crocco Galeas 1998: 84- 86). Consequently, in her scale of diagrammaticity (Crocco Galeas 1998: 32), conversion (under the name of "metaphoricity") is
placed below modification and above total suppletion, while in the equally expl icit and
PA135
even more detailed scale of diagrammaticity proposed by Kilani-Schoch (1988: 166117) conversion is ranked below strong suppletion, but together with total suppletion
under the heading of "adiagrammaticite". An ex-aequo solution for the arrangement of
the scale with respect to this type of process is envisaged also by Dressler (2005: 269),
who ranks modification as more diagrammatic than conversion, but does not exclude at
the same time that both of them could occupy an equal rank on the scale, referring to
the fact that conversion is very productive in English whereas modification is unproductive. If this means that universal scales of naturalness can be rearranged according to
what is frequent and productive in particular languages, the often-repeated cfaim that the
framework is deductive and that the preference degrees on the parameters, which are
claimed to be universal, are based on extralinguistic factors (Dressler 2005: 268) loses
much of its credibility (but cf. Dressler and Manova 2005 for a more deliberate account).
5.2. Problems with scaling

It comes as no surprise that a theory that explicitly rejects formalism runs into problems
as soon as it starts to pin down any of its components in a formal guise, even in as
simple a form as a one-dimensional scale. Problems like the ranking of process-types
are indicative of the fact that semiotics alone is perhaps not sufficient for the purpose
of setting up generalizations concerning particular types of construction. The general
impression is that the naturalness scales that can be found in the literature appear so
plausible to the authors who contrive them that not enough care is taken in the arguments
leading to the desired result. For example, it is not clear why the modification of the
stem vowel in a derived agent noun like German Laufe r 'runner' (from laufen 'to run')
should downgrade the diagrammaticity of this derivational type. One could! as well assume that diagrammaticity is even enhanced by the modification, since the addition of
the agentive suffix -er could be seen as mirroring the "addition of meaning" (whatever
that may look like in terms of cognition), while the modification of the diphthong in the
stem would mirror the overall semantic change brought about by this addition. Instead,
the modification is qualified as indexical (see section 5.3 below), but it is not made clear
in what way the presence of an indexical element subtracts from the diagrammaticity of
a construction. By founding its main principles on semiotics, which is basically sensible,
natural morphology does not always succeed in keeping the adopted semiotic concepts
apart and in specifying what counts as what in a given type of construction. This is a
conundrum for which Anttila (1975: 10) has set the stage with the precept that linguistic
signs "are not simple, but intricate blends of symbolic, indexical, and iconic elements"
(emphasis in the original; cf. also Noth 2008: 84). Furthermore, natural morphology
relies heavily on semantics but has never made explicit what kind of semantics it is that
underlies its assumptions. In the case of word-formation this is crucial, since derivational
morphemes, unlike inflectional ones, are semantically complex and especially prone to
interact with lexical meaning.
Another problem consists in the fact (which is never stated explicitly by proponents
of natural morphology), that the naturalness scales regard only a subset of the formal
operations mentioned above (section 5.1.1 ): suppletion, affixation, subtraction, modification and conversion. For example, Crocco Galeas (1998: 32) ranks agglutinative affixa-
135
PA136
136
tion highest on her scale of diagrammaticity. However, if the definition of diagrammaticity rests on compositionality ("diagrammaticity entails a relation of biuniqueness between
segmentability of signans and compositionality of signatum'', Crocco Galeas 1998: 30,
italics in the original), one may ask why the top position on this scale is not assigned to
compounding (cf. Stolz 1990, who asks the same question - and many more - in his
review of Dressler et al. 1987). From the design of the naturalness scales it follows that
compounding and derivation are implicitly kept apart as different compartments of wordformation, but this neglects the gradualness of the passage from one to the other, with
intermediate phenomena like affixoids and combining forms.
Reduplication, which would also deserve a place in the upper ranks of diagrammaticity, is likewise missing for no obvious reason. The same holds for the scale in KilaniSchoch (1988: 116- 117), who assigns the highest rank of diagrammaticity to suffixation,
with other kinds of affixes unaccounted for. Dressler (2005: 269-269), who discusses
the diagrammaticity of different types of compounds but does not explicitly place them
on a scale, also omits reduplication, while Thornton (2005: 164) places reduplication as
the maximally iconic type of construction at the top of a scale of constructional iconicity
(which is restricted to plural formations, however). As far as subtypes are concerned,
Crocco Galeas (1998: 32) considers only one type of synaffixation, viz., the combination
of an affix with modification of the base, while other variants of synaffixation, like
circumfixation, are left out, and so forth.
The main problem with naturalness parameters, for which divergent assignments like
the one concerning conversion are only symptomatic, lies in fact that the extra-linguistic
foundation of the tendencies or preferences postulated in natural morphology lacks a
solid empirical basis. This is in the nature of things, of course. While natural phonology,
dealing with meaningless form, could walk its way safely on the relatively firm ground
of physiology and phonetics, natural morphology has to cope with meaningful units,
which affords explicit reference to semantics, the extralinguistic basis of which is cognition. It seems that by relying directly on semiotics, the founders of natural morphology
have tried to avoid a commitment to any explicit theory of semantics, which is understandable in view of the labyrinthic nature of this branch of linguistics and its extensions
into similarly entangled fields like cognitive science and philosophy of language. However, that shortcut takes its toll when it comes to being concrete about relations between
meaning and form. Only if the way in which meaning is represented is made explicit
can judgements be made on how iconically it is expressed by form. Mayerthaler seems
to have been aware of the problem and provided what he called "semantic markedness
relations" (Mayerthaler 1987: 39-40; cf. also Mayerthaler 1980: 27-29), but these are
rather conceived for inflection than for word-formation, and no further steps have followed this first one (cf. the criticism in Marzo 2008); cf. also Andersen's (2008: 14)
critical assessment: "N[atural]ness theory's odd idea that all linguistic relations are scales
makes it impossible for the linguist who applies this theory to attain descriptive adequacy
in the face of the actual variety of, say, semantic relations."
There is nothing that could compensate for the lack of a semantic component in a
theory dealing with the expression of meaning, nor for the lack of a theory of the lexicon
in a theory dealing with lexical enrichment. The wealth of external evidence adduced by
natural morphology in support of its claims is impressive, but cannot be used as an
explanans for naturalness. Early acquisition, cross-linguistic frequency, diachronic stability and the like can only be explananda in a deductive framework. It is indicative for
PA137
this dilemma that proponents of natural morphology often skip semantics and resort
directly to cognition as a justification for their assertions, as if cognition were something
obvious or at least well-defined. For example, Dressler (1987: 99), delineating the functions of word-formation, states that "the cognitive function is best served by labelling
the concepts needed as precisely as possible", without adding any hint on what evidence
from cognitive science in support of this claim should look like (nor making explicit
what it means for labeling to be "precise"). In spite of great advances in the recent past,
very little is known as yet about the nature of cognition, so any assumption about what
may be acting in its service remains intuitive.
5.3. Other parameters

The parameter of diagrammaticity has been discussed in some detail here since the
proponents of natural morphology consider it to be of central relevance to word-formation (Crocco Galeas 1998: 125; Dressler 2005: 268-269), and because it best reveals the
basic merits and problems of this approach. A less perspicuous parameter cllerived from
Peircean semiotics is indexicality. "The relation between sign and object is indexical
when it is defined by a spatio-temporal (factual or existential) contiguity between them"
(Shapiro 1983: 39). In natural morphology this is interpreted as a matter of distance of
an operator from its operand, from which it follows that an affix is maximaHy indexical
if it is adjacent to the lexeme it modifies. This is said to explain why inflectional morphemes, in that they refer to external constituents, occupy the peripheral positions in a
complex word, while derivational morphemes are placed closer to the lexical base
(Dressler 1987: 111).
The fact that proponents of natural morphology are not always at ease with their own
claims can be illustrated with the in terpretation of indexicality in terms of word structure:
While Dressler (1987: 111) has insisted that multiple modification, as in introflecting
languages, constitutes the highest degree of indexicality (which is plausible, there being
no way of being closer to each other than being interwoven), the same author later holds
that "afftxation is more natural than modification, because the indexical relation between
an affix and its base is clearer" (Dressler 2005: 270). Carstairs-McCarthy (1992: 226),
relying on the earlier conception, has applied the monopoly principle (without naming
it so) to indexicality, arriving at the conclusion that "perfect indexicality would presumably require in every word form the total fusion of the lexical stem with the realisations
of all its associated derivational processes and morphosyntactic properties, yielding inflectional paradigms consisting entirely of unsegmentable suppletive 'stems"'. This provides a more reasonable solution to the problem than Dressler's appeal to "clarity'', since
"even with only the few inflectional categories of English, such a language would impose
a totally implausible burden on the memory" (Carstairs-McCarthy 1992: 226). Thus, in
spite of the fact that "all of morphology is indexical insofar a morphological marker
refers to the base of the rule that introduces it" (Dressler 2005: 270), unlike other parameters, indexicality does not lend itself easily to operationalization, nor does it seem to
be understood as a major factor by proponents of natural morphology. If indexicality
were a driving force in the structuring of word forms, one should expect modification
to be preferred for derivational morphology and dispreferred for inflection, in view of
137
PA138
138
the explanation for the relative order of derivational and inflectional morphemes with
respect to their base mentioned above, but this does not seem to be borne out by the
facts. Kilani-Schoch (1988: 126), who explicitly affirms that the contiguity between the
grammatical morpheme and the lexical base is particularly strong in derivations like
Arabic katab-a ' he wrote' ~ katib 'writer', makes no claim in this direction.
Crocco Galeas (1998: 22) is right in pointing out that "within natural morphology,
there is no general agreement on the number of the universal parameters, nor is there a
fixed list of semiotic principles (or factors) from which parameters can be deduced".
The present survey is not intended as a complete overview of all the proposals that have
been made, repeated and modified in the literature. The parameters based on iconicity
have been reviewed more thoroughly in the preceding subsections in order to expose the
ambitious and at the same time tentative nature of the approach. Other important topics,
like transparency (in terms of compositionality and segmentability, thus closely related
to diagrammaticity and often treated under the same heading), biuniqueness and optimal
shape of units have been mentioned only in passing. This imbalance is justifiable, however, in view of the fact that the universal preferences postulated in natural morphology
hold for word structure as such and not for word-formation in particular. Therefore, and
with regard to the explicitly functionalist design of the framework, it has been considered
appropriate to highlight those aspects that are of most direct concern to the principal
function of word-formation, viz., lexical enrichment.
6. Concluding remarks
Natural morphology can be characterized as an ambitious research program designed to
explain asymmetries in the occurrence and distribution of structural aspects of complex
word forms . The epistemological design of the program is strictly deductive, without
recourse to ad-hoc conventions and circular argumentation, which guarantees that predictions are empirically testable (cf. Dressler l 985a), according to the Popperian criterion
of fallibility. It has stimulated research in many areas previously neglected and even
opened up new sub-fields of morphology, like morphopragmatics (Dressler and Merlini
Barbaresi 1994; see article 62 on the pragmatics of word-formation) .
As far as word-formation is concerned, an impressive amount of empirical research
has been inspired by natural morphology over the last three decades,. especially in the
field of child language studies (cf. Korecky-Kroll 2011 for a recent account). The success
of the framework in this respect is not paralleled in other domains, however. The predictive potential of naturalness theory concerning the directionality of morphological change
has been overshadowed by the enormous amount of attention devoted to grammaticalization theory (cf. Giacalone Ramat 1995: 134), and the proponents of both sides have
refrained from integrating their respective approaches (cf. also Dahl 2004: 11 5-117 for
criticism in another vein). Such drawbacks notwithstanding, the idea of multidimensional
scaling underlying the approach of natural morphology remains attractive as a model for
a multifunctional phenomenon like word-formation, especially in view of recent increase
of typological research: " In our view, the universal principles of Natural Morphology and
its three sub-theories (universal preferences, typological adequacy and system-dependent
naturalness), are at the heart of typological explanations" (Stekauer, Valera and Kortvely-
PA139
essy 201 2: 9). On the other hand, Bauer 's (2003: 265) request for "greater detail in some
of the many areas which are currently rather obscure" has relinquished none of its topicality in the decade since it was made.
7. References
Andersen, Henning
2008
Naturalness and markedness. In: Klaas Willems and Ludovic De Cuypere (eds.), Naturalness and Iconicity in Language, 101-119. Amsterdam/Philadelphia: Benjamins.
Anttila, Ra imo
The lndexical Element in Morphology. Innsbruck: Institut fur S prachw issenschaft der
1975
Universitat Innsbruck.
Bauer, Laurie
2003
Introducing Linguistic Morphology. 2nd ed. Edinburgh: Edinburgh Univers ity Press.
Bauer, Laurie
2004
A Glossary of Morphology. Edinburgh: Edinburgh University Press.
B loomfie ld, Leona rd
1933
Language . New York: Holt.
Bybee, Joan L.
20 13
Usage-based theory and exemplar representations of constructions . In: Thomas Hoffmann and Graeme Trousdale (eds.), The Oxford Handbook of Construction Grammar,
49- 69. O xford: Oxford Univers ity Press.
Carsta irs-M cCarthy, Andrew
1992
Current Morphology. London/New York : Routledge.
Chomsky, Noam
1970
Remarks o n nomina lization. In: Roderick A. Jacobs and Peter S. Rosenbaum (eds.),
Readings in English Transformational Grammar, 184- 22 1. Waltham, MA: G inn. [Reprinted 1972 in: Noam Chomsky, Studies on Semantics in Generative Grammar, 11-6 1.
The Hague: Mouton].
Crocco Galeas, G razia
1990
Conversion as morphological metaphor. In : Julian Mendez Dosuna and Carmen Pensado
(eds.), Naturalists at Krems. Papers from the Workshop on Natural Phonology and
Natural M01phology (Krems, J- 7 July 1988), 23- 32. Salamanca: Ediciones Universidad
de Salamanca .
Crocco Galeas, Grazia
1998
The Parameters of Natural Morphology. Padova: Unipress .
Cruse, Alan D.
2006
A Glossary of Semantics and Pragmatics. Edinburgh: Edinbu rgh University Press .
Cruse, A la n D.
20 11
Meaning in Language. An Introduction to Semantics and Pragmatics. 3rd ed. Oxford/
New York: Oxford University Press.
Dahl, Osten
2004
The Growth and Maintenance of Linguistic Complexity. Amsterdam/Philade lphia: Benjamms .
Danesi, Marcel
1995
The iconic ity of metaphor. In: Marge E . Landsberg (ed.), Sy ntactic Iconicity and Linguistic Freezes. Th e Human Dimension, 265- 284. Berlin/New York: Mouto n de Gruyter.
Donegan, Patric ia
200 1 Constraints and processes in phonologica l pe rception. In: Kata rzyna Dziubalska-Kolaczyk (ed.), Constraints and Pref erences, 43- 68. Berlin/New York: Mouton de Gruyter.
139
PA141
Dressler, Wolfgang U.
2000a Subtraction. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.) , Morphology. An International Handbook on Inflection and Word-Formation. Vol. 1, 581 - 587.
Berlin/New York: de Gruyter.
2000b Extragrammatical vs. marginal morphology. In: Ursula Doleschal and Anna M. Thornton
(eds .), Extragrammatical and Marginal Morphology, 1- 10. Miinchen: LTNCOM Europa.
2002
Naturalness and functionalism. In: Katarzyna Dziubalska-Kolaczyk and Jaroslaw Weckwerth (eds.), Future Challenges for Natural Linguistics, 83- I02. Munchen: LTNCOM
Europa.
2005
Word-formation in natural morphology. In: Pavol Stekauer and Rochelle Lieber (eds.),
Handbook of Word-Formation, 267-284. Dordrecht: Springer.
Dressler, Wolfgang U. and Stela Manova
2005
The morphological technique of conversion in the intlecting-fusional type. In: Laurie
Bauer and Salvador Valera (eds.), Approaches to Conversion/Zero-Derivation, 67-101.
Munster: Waxmann.
Dressler, Wolfgang U., Willi Mayerthaler, Oswald Panagl and Wolfgang U. Wurzel
1987
Introduction. In: Wolfgang U. Dressler, Willi Mayerthaler, Oswald Panagl and Wolfgang
U. Wurzel (eds.), Leitmotifs in Natural Morphology, 3-22. Amsterdam/Philadelphia:
Benjamins.
Dressler, Wolfgang U. and Lavinia Merlini Barbares i
1994
Morphopragmatics. Diminutives and Intensifiers in Italian, German and other Languages. Berlin/New York: Mouton de Gruyter.
Dressler, Wolfgang U. and Lavinia Merlini Barbaresi
200 1 Morphopragmatics of diminutives and augmentatives: On the priority of pragmatics over
semantic s. In: Istvan Kenesei and Robert Harnish (eds.), Perspectives on Semantics,
Pragmatics, and Discourse. A Festschrift for Ferenc Kiefer, 43 - 58. Amsterdam/Philadelphia: Benjamins.
Fenk, August and Gertraud Fenk-Oczlon
2008
Complexity trade-offs between the subsystems of language. In: Matti Miestamo, Kaius
Sinnemaki and Fred Karlsson (eds.), Language Complexity: Typology, Contact, Change,
43- 65. Amsterdam/Philadelphia: Benjamins.
Friedrich, Paul
1975
The lexical symbol and its relative non-arbitrariness. In: Dale M. Kinkade, Kenneth L.
Hale and Otmar Werner (eds.), Linguistics and Anthropology. Jn Honor ofC. F. Voegelin ,
199- 247. Lisse: Peter de Ridder Press.
Giacalone Ramat, Anna
1995
lconicity in grammaticalization processes. In: Raffaele Simone (ed.), Jconicity in Language, 119- 139 . Amsterdam/Philadelphia: Benjamins.
Haiman, John
1985
Introduction. In: John Haiman (ed.), l conicity in Syntax, 1-7. Amsterdam/Philadelphia:
Benjamins.
Halle, Morris
1973
Prolegomena to a theory of word formation. Linguistic inquiry 4(1): 3- 16.
Haspelmath, Martin and Andrea D. S ims
2010
Understanding Morphology. London: Hodder Education.
Hiraga, Masako K.
1994
Diagrams and metaphors: Iconic aspects in language. Journal of Pragmatics 22: 5- 21.
141
PA142
142

Hurch, Bernhard and Richard Rhodes
1996
Introduction. ln: Bernhard Hurch and Richard Rhodes (eds.), Natura l Phonology. The
State of the Art, VII- XII. Berlin/New York: Mouton de Gruyter.
Johansen, J0rgen Dines
1993
Dialogic Semiosis. An Essay on Sign and Meaning. Bloomington: Indiana University
Press.
Kastovsky, Dieter
2005
Hans Marchand and the Marchandeans. In: Pavol Stekauer and Rochelle L ieber (eds.),
Handbook of Word-Formation, 99- 124. Dordrecht: Springer.
Kenstowicz, M ichael J. and Charles W. Kisseberth
1979
Generative Phonology. Description and Theory. New York: Academic Press.
Kilani-Schoch, Marianne
1988
Introduction a la mo1phologie naturelle. Berne: Lang.
Kiparsky, Paul
1983
Word-formation and the lexicon. In: Frances A. Ingemann (ed.), Proceedings of the
19.82 Mid-America Linguistics Conference, 3- 29 . Lawrence (Kansas): De partment of
Linguistics, University of Kansas.
Klos, Verena
2011
Komposition und Kompositionalitiit. Moglichkeiten und Grenzen der semantischen Dekodierung von Substantivkomposita. Berlin/New York: de Gruyter.
Korecky-Kroll, Katharina
2011
De r Erwerb der Nominalmorphologie bei zwei Wiener Kindern: E ine Untersuchung im
Rahmen der Natiirlichkeitstheorie. Ph.D . dissertation, University of Vienna.
Kouwenberg, Silvia
2003
Twice as Meaningful. Reduplication in Pidgins, Creoles and other Contact Languages.
London: Battlebridge Publications.
Kouwenberg, Silvia and Darlene LaCharite
2001
The iconic interpretations of reduplication: Issues in the study of reduplication in Caribbean creole languages. European J ournal of English Studies 5( I): 5 9-80.
Lehmann, Christian
2007
Motivation in language: Attempt at a systematization. In: Peter Gallmann, Christian
Lehmann and Rosemarie Liihr (eds.), Sprachliche Variation. Zur lnterdependenz von
lnhalt und Ausdruck, 100-135. Tiibingen: Narr.
Marzo, Daniela
2008
What is iconic about po lysemy? A contribution to research on diagrammatic transparency. In: Klaas Willems and Ludovic De Cuypere (eds.), Naturalness and lconicity in
Language, 10 1- 119. Amsterdam/Philadelphia: Benjamins.
Marzo, Daniela
2013
Polysemie als Verfahren lexikalischer Motivation. Theorie und Empirie am Beispiel von
Metonymie und Metapher im Franzosischen und l talienischen. Tiiblngen: Narr Francke
Attempto.
Mattes, Veronika
2006
One form - opposite meanings? Diminutive and augmentative interpretation of full
reduplication in Bikol. In: Tenth International Conference on Austronesian Linguistics,
17-20 January, Palawan, Philippines . L inguistic Society of the Philippines I SIL International. Available online at: http://www.sil.org/asia/philippines/ical/papers.htm [last access 2 Feb 2013].
Mattiello, E lisa
2013
Ex tra-grammatical Morphology in English. Abbreviations, blends, reduplicatives, and
related phenomena. Berlin/New York: Mouton de Gruyter.
PA143
Mayertha le r, W illi
1977
Studien zur theoretischen und zur franzosisch en Morphologie. Reduplikation, Echoworter, morphologische Natiirlichkeit, Haplologie, Produktivitat, Regeltelescoping, paradigmatischer Ausgleich. Tiibingen: Niemeyer.
Mayertha ler, W illi
1980
Ikonismus in der Morphologie. Zeitschriftfur Semiotik 2: 19- 37.
Mayerthaler, W illi
198 1 Morphologische Naturlichkeit. Wiesbaden: Athe na ion.
Mayertha ler, Willi
1987
System-indepe ndent morpho log ical naturalness. I n: Wo lfga ng U. D ressler, Will i Mayerthaler, Oswald Panagl and Wolfgang U. Wurzel (eds .), Leitmotifs in Natural Morphology, 25- 58 . Amsterdam/Philadelphia: Benjamins.
Mayerthaler, Willi
1988
Morphological Naturalness. With a preface by Rupert Riedl. Ann Arbor: Karoma Press.
Murphy, M. Lynne
2010
Lexical Meaning. Cam bridge: Cambridge University Press.
Nanny, Max and O lga Fischer
200 1 Introduction: Veni vidi v ici. In: Olga Fischer and Max Nanny (eds.), The Motivated
Sign, I - l 4. Amsterdam/Philade lphia: Benjam ins .
N oth, Winfried
2008
Semiotic foundations of natural linguistics and diagrammatic iconicity. In : Klaas Wille ms and Ludovic De Cuypere (eds .), Naturalness and lconicity in language, 73- 100.
Amsterda m/Philadelphia: Benja mins .
Nunberg, Geoffrey
1979
The no n-uniqueness of semantic solutio ns: Polysemy. linguistics and Philosophy 3:
143- 184.
Panther, Klaus-Uwe
2008
Conceptual a nd pragmatic motivation as an explanatory concept in lingu istics. Journal
of Foreign l anguages 3 1(5): 2- 19.
Pi lier, Ingrid
1999
Iconic ity in brand names. In: Max Nanny and Olga Fischer (eds.), Form Miming Meaning, 325-341 . Amsterdam/Philadelphia: Benjamins.
Popescu, Ioan-Iovitz, Sven Naumann, Emmerich Ke lih, Andrij Rovenchak, Haruko Sanada, Anja
Overbeck, Reginald Smith, Radek Cech, Panchana n Mo hanty, Andrew Wilson and Gabrie l Altmann
2013
Word length: Aspects and languages. In: Re inhard Kohler and Gabriel Ahmann (eds.),
Issues in Quantitative linguistics 3. Dedicated to Karl-Heinz Best on the occasion of
his 70 1" birthday, 224- 28 1. Liide nsche id: RAM Verlag.
Radden, G unter and Klaus-Uwe Panther
2004
Introduction: Reflections on motivation. In: Gunter Radden and K laus-Uwe Panther
(eds.), Studies in Linguistic Motivation , 1- 46. Berlin/New York : Mouton de Gruyter.
Ronne berger-Sibo ld, E lke
1995
Different ways of optimiz ing the sound shape of words. In: Henning Andersen (ed.),
Historical Linguistics 1993. Selected Papers from the I 11" International Confe rence on
Historical linguistics, Los Angeles, 16-20August1993, 42 1- 432. Amsterdam/Philadelphia: Be nj am ins.
Ronneberger-Sibold, E lke
20 10
Word creatio n: Definition - function - typology. Tn: Franz Raine r, Wolfgang U . D ressler,
Dieter Kastovsky and Hans C hristian Luschiitzky (eds.), Variation and Change in Morphology. Selected Papers from the 13 1" International Morphology Meeting, Vienna, February 2008, 20 1-2 16. Amste rdam/Philadelphia: Benj amins.
143
PA144
144

Rubino, Carl
2005
Reduplication: Form, fu nction and distribution. In: Bernhard Hu rch (ed.), Studies on
Reduplication, 11-29. Berlin/New York: Mouton de Gruyter.
2011 [1916]
Course in General Linguistics. Translated by Wade Baskin. Edited by Perry
Me isel and Haun Saussy. New York/Chichester: Columbia University Press.
Seiler, Hansjakob
1975
Die Prinzipien der deskriptiven und der etikettierenden Benennung. In: Hansjakob Seiler
(ed.), Linguistic workshop III. Arbeiten des Koiner Universalienprojekts 1974, 2-57.
Miinchen: Fink.
Shapiro, Michael D.
1983
The Sense of Grammar. Language as Semeiotic. Bloomington: Indiana U niversity Press.
Shapiro, Michael D .
1991
The Sense of Change. Language as History. Bloomington/ Indianapolis: Indiana University Press.
Stampe, David
1969
The acquisition of phonetic representation. In: Anthony Bruck, Robert A. Fox and Michael W. La Galy (eds.), Papers from the Parasession on Natural Phonology, 443- 453.
Chicago: Chicago Linguistic Society.
Stampe, David
1979 [1973]
A Dissertation on Natural Phonology. New York: Garland.
Stekauer, Pavo!
1998 An Onomasiological Theory of English Word-Formation. Amsterdam/Philadelphia: Benjamins.
Stekauer, Pavo!, Salvador Valera and Livia Kortvelyessy
2012
Word-Formation in the World's Languages. A Typological Survey. Cambridge: Cambridge U niversity Press.
Stolz, Thom as
1990
Review of Dressler, Mayerthaler, Panagl and Wurzel ( 1987). Linguistics 28( 1): 162167.
Thornton, Anna M.
2005
Morfologia . Roma: Carocci.
Tobin, Yishai
2009
Comparing and contrasting natural phonology, optimality theory and the theory of phono logy as human behavior. Poznan Studies in Contemporary Linguistics 45( 1): 169189.
Ullman, M ichael T.
2004 Contribut ions of memory circu its to language: The dec larative/procedural model. Cognition 92: 23 1-270.
Ungerer, Friedrich
1999
Diagrammatic iconicity in word-formation. In: Max Nanny and Olga Fischer (eds.),
Form Miming Meaning, 307- 324. Amsterdam/Philadelphia: Benjamins.
Waugh, L inda and Madele ine Newfield
1995
lconicity in the lexicon and its relevance for a theory of morphology. In: Marge E.
Landsberg (ed.), Syntactic l conicity and Linguistic Freezes. The Human Dimension,
189- 221. Berlin/New York: Mouton de Gruyter.
Wurzel, Wolfgang Ullrich
1985
On morphological naturalness. Nordic Journal of Linguistics 7: 165 - 183.
Zacarias, Ramon
2006
Formaci6n de deminutivos con el sufijo /-it-/: Una propuesta desde la morfologia natural. Anuario de Letras 44: 77- 103.
Hans Christian Luschiitzky, Vienna (Austria)
PA145
10. Word-formation in cognitive grammar

I.
2.
3.
4.
5.
6.
7.
Basic principles of cognitive grammar

Word structure: analyzability
Entrenchment and drift; countering analyzability
Characterizing the parts
Composition
Conclusion
References
Abstract
This article summarizes some of the basic assumptions ofcognitive grammar with an eye
on their relevance to the study of word-formation. Issues addressed include analyzability,
entrenchment, and productivity; the different processes of word-formation (affixation,
compounding, and blending); the rationale for recognizing words as distinctive units of
linguistic structure; and the similarities and differences between word structure and
phrase structure.
1. Ba sic principles of cognitive grammar

Cognitive grammar is a theory of language which Ronald Langacker began working on
in the late 1970s. The standard exposition of the theory is the two-volume Foundations
of Cognitive Grammar (Langacker 1987, 1992); an updated account is Langacker (2008).
Taylor (2002) is a text-book introduction. (For the position of cognitive grammar within
a more broadly characterized cognitive linguistics movement, see Taylor 2002, Ch. 1.)
Each of these publications has extended discussions of words and their structure. Additional resources - both those which explicitly adopt Langacker's theory and those which
are broadly compatible with its goals - are cited in this a1iicle.
The starting point of cognitive grammar is uncontroversial: a language is a means for
relating sound and meaning. More specifically, a language enables speakers to represent
their thoughts and intentions by making available to them an inventory of symbolic
associations between units of form (phonological structures) and units of meaning (semantic structures). Hearers familiar with the symbolic associations are able to recover,
or to intimate, the speaker's semantic intentions.
A notable feature of the theory, in contrast with competing accounts, is its minimalist
approach to the form-meaning relation. There are, namely, only three objects of study in
cognitive grammar. The first comprises language in its perceptible form, typically as
sound, but also as writing, or (in the case of signed languages) as gesture. Using the
term rather broadly, this aspect of language is referred to as phonological structure. The
second is the symbolized content, referred to as semantic structure. Semantic structure
comprises a person's conceptualizations, in a very general understanding of the term; it
includes not only referential intent and propositional content, but also affect, evaluations,
145
PA146
146
and attitudes towards possible hearers. The third object of study are symbolic associations between phonological and semantic structures. The latter are taken, following Saussure ( 1964 [ 1916]), to be conventional in character; their properties and indeed their
very existence need to be specifically learned by members of a speech community.
Cognitive grammar proposes a direct relation between phonological and semantic
structure. In many alternative theories the relation, on the contrary, is mediated by intervening levels of organization. These are autonomous in the sense that they are composed
of elements and relations which are unique to these levels and which cannot be reduced
to matters of phonology or semantics. One such intervening level (or, in some theories,
cluster of intervening levels) is the syntax. Cognitive grammar does not deny the existence of syntactic phenomena. The claim, rather, is that syntactic elements - notions of
lexical and phrasal categories (noun, noun phrase, etc.); categories such as complement
and adjunct; syntactic relations (clausal subject, phrasal head, etc.), and even the very
notions of word and phrase - can themselves be fully described in terms of the minimalist ontology of the theory. An aim of this article is to indicate the validity of this approach
to the description of words and their internal structure.
1.1. A structured inventory of units

In cognitive grammar, knowledge of a language is equated with knowledge of an inventory of units, a unit being defined as any phonological, semantic, or symbolic structure
that has been established, or entrenched, in the speaker's mind through frequency of
previous use. Importantly, the units do not constitute a random, unordered list. Rather,
the inventory is structured by virtue of the ways in which the units are related to each
other. Three kinds of relation are of special importance. These are the part-whole relation, the schema-instance relation, and the relation of similarity (or, in the limiting case,
identity).
a) The part-whole relation. This is the relation whereby a linguistic structure may be
analysed into its component parts. The part-whole relation applies to phonological,
semantic , and symbolic structures alike. Thus, the phonological form [sIIJ] breaks
down into its component sound units, namely, the sounds [s], [1], and [IJ], while the
symbolic unit [singer] can be analysed into its constituent symbolic units [sing] and
[-er]. The counterpart to analysis is composition. This is the process whereby speakers
recruit already established units in order to create structures which are not already
entrenched in their mental grammar.
b) The schema-instance relation. A linguistic structure - whether phonological, semantic, or symbolic - may count as an instance of a more abstractly characterized unit.
The phonological structure [sitJ] may be seen as an instance of a schematic phonological unit [Syllable], which in turn can be analysed into its schematically characterized
parts [Onset] and [Rhyme]. While speakers encounter only instances, the sanctioning
schemas are immanent in the instances to the extent that the instances may be recognized as matching the specifications of the schema.
These two relations are recursive. [A] may be a part of [B], which in turn is a part of
[C]. Likewise, [X] may be an instance of [Y], which in turn is an instance of [Z]. The
PA147
two relations interact in numerous ways. For example, the very possibility of analysing
a structure into its parts is constrained by the existence of a schema which sanctions the
analysis; at the same time, the analysis strengthens the sanctioning schema. The interplay
of the two relations is also evident in the process of composition. The possibility of
creatively combining already established units rests not only on the availability of the
combining units themselves, but also on the entrenchment of a schema which sanctions
the combination.
c) The similarity relation. Although difficult to define formally, similarity - or more
accurately, speakers' subjective perception of similarity - plays an important role in
the emergence of schemas. The fact that [A] and [B] come to be regarded as instances
of [C] rests on the prior recognition that [A] and [B] are similar in some respect(s).
Moreover, the similarity between [A] and [B] may be perceived to be similar to the
way in which [D] and [E] are similar. The commonality between the two cases may
give rise to a h igher order schema (Nesset 2008). For example, the perceived similarity between word pairs such as urbane/urbanity, insane/insanity, profane/profanity
may give rise to a schema representing the alternating phonological forms.
1.2. The autonomy of phonological and semantic structures

In v iew of the role of language as a means for giving overt expression to a speaker 's
conceptualizations, it is only to be expected that the study of symbolic relations will
constitute the heart of any cognitive grammar investigation. However, the theory also
allows for the possibility that phonological and semantic structures exhibit a degree of
autonomy vis-a-vis symbolic relations, in the sense that components of these structures
need not participate in symbolic associations. Thus, with respect to phonological structure, units such as vowels and consonants, syllables and their parts, do not of themselves
have symbolic value. Hickory-dickory-dock, of the nursery rhyme, is perfectly wellformed phonologically, in that the expression is composed of established units and is
sanctioned by well-entrenched phonological schemas; the expression, however, corresponds to nothing at all at the semantic level. Similarly, there are many elements of
semantic structure which are not overtly symbolized by any phonological material. This
is very obviously the case with noun-noun compounds (Benczes 2007). There is noth ing
in the phonological structures of the expressions to indicate that while a cheese shop
sells cheese, a barber shop does not sell barbers (rather, it sells the service which barbers
provide) (Jackendoff 2010: 446).
1.3. A usage-based grammar

Cognitive grammar is a strictly bottom-up, or usage-based model of linguistic knowledge
and its acquisition (Tomasello 2003). Abstractions (as encapsulated in schemas) are abducted on the basis of acquaintance with their instances. Repeated exposure leads to the
progressive entrenchment of instances and of the sanctioning schemas. Entrenchment is,
to be sure, a matte r of degree; its extent also varies from language user to language user.
147
PA148
148
It is therefore taken as rather evident that different speakers of the "same" language may
have acquired rather different mental representations.
The usage-basis of cognitive grammar entails that a speaker's mental grammar may
contain a good deal of redundancy. Expressions which are fully consistent with a sanctioning schema and which could, in principle, be created compositionally, may nevertheless have unit status, provided that they have been entrenched through previous linguistic experience. For example, high frequency plurals may well be stored in the mental
grammar alongside the singular forms.
1.4. Motivation
Symbolic relations are conventional, in the sense that they have to be specifically learned
(on the basis of experience). This wording is noncommittal w ith respect to the Saussurian
notions of arbitrariness and motivation. There is, to be sure, much in a language which
is arbitrary, the fact that the phonological form [Sil)] is associated with the semantic
concept 'sing' being but one example. Equally, a very great deal in a language is motivated. Motivation is often taken to refer to the role of semantic structure, or conceptualization more generally, in the shaping of phonological and symbolic structures (Panther and
Radden 201 1). While not discounting this aspect, the proposal that linguistic knowledge
constitutes a structured inventory of units offers an alternative perspective, where each
unit stands at the hub of a network of relations (part-whole, schema-instance, and similarity) to other units which are already entrenched (to a greater or lesser degree) in the
mental grammar (Taylor 2004). Even a phonological form such as [Sil)] - whose association with the concept 'sing' may indeed be arbitrary - is motivated by the fact that it
conforms with a schema for syllable structure and is made up of units which are well
entrenched elsewhere in the language. Language-internal motivation undoubtedly facilitates the learning, storage, and accessing of linguistic units, and testifies, again, to the
redundancy inherent in the language system.
2. Word structure: analyzabi lity

Knowledge of a language consists of a network of units, structured by the part-whole,
schema-instance, and similari ty relations. The focus in this section is on the part-whole
relation, with specific reference to words. On what basis it is possible to identify a part
of a word?
Observe that the question, as posed, refers to words, that is, to symbolic associations
of form and meaning. The fact that [s] is part of the phonological pole of the symbolic
unit [sing] is not strictly speaking relevant to word structure. In speaking of word parts
we are primarily interested in parts which themselves have symbolic status, whereby a
part of the phonological structure of a word can be associated with a semantic value
which contributes to the word's semantic structure.
In order for a word-part to be recognized, a number of conditions need to be satisfied,
at least in the canonical cases (some non-canonical cases are discussed below). First, we
need to be able to establish an identity relation (or, at least, a relation of similarity)
PA149
between a piece of a word's phonology and (a piece of ) the phonology of at least one
other symbolic unit. Second, we need to be able to associate the phonological segments
with an identical (or similar) semantics. Third, the postulated semantic value of the
word-part needs to feature in the semantic structure of the word itself. Grounds for
identifying the word-part will be strengthened in proportion to the number of symbolic
units exhibiting the postulated word-part; in the limiting case, the correspondence may
be perceived to be accidental or fortuitous, or indeed not noticed at all. Below, I discuss
a range of examples, selected so as to illustrate some of the issues which arise when
considering the internal structure of words.
2.1. farmer
The analyzability of this word into its parts [farm] and [-er] is rather evident. It is
instructive to reflect on why this intuition should be so clear. Essentially, we accept this
analysis because it fits in with so many other facts about the English language; the
analysis, in other words, is very strongly motivated by language-internal relations. First,
farmer is obviously related to the well-entrenched verb form farm , both phonologically
(the phonological structure of the verb is fully incorporated into the derived form) and
semantically (the derived word invokes the same kind of activity as the base verb).
Second, there exists a large number of words for which a comparable analysis is indicated: singer, walker, driver, and countless more. These words characterize a person in
terms of what they do, the activity in question being supplied by the verbal element
while the agent notion can plausibly be attributed to the suffix. The existence of this
cohort of examples, with comparable semantics, enables the abstraction of a schema
which captures their commonality, viz. the agentive noun schema [NV-er]. The schema
is supported by its compatibility with a very general pattern of word-formation in English, namely suffixation, whereby the suffix determines the kind of enti ty that a word
refers to while the word-specific semantic content is supplied by the stem. In this sense,
the agentive noun schema is itself an instance of a higher order schema for derived
words.
2.2. philosopher and barber

In the canonical case, a word can be exhaustively divided up into constituent symbolic
units. The "building block" metaphor (Langacker 1992: 186) does not always apply.
Given the high degree of entrenchment of the agentive noun schema (this being a function of the large number of examples on which it is based), we are likely to detect the
[-er] suffix in words such as philosopher and grocer, and perhaps even in barber. These
words correspond only partially with the agentive noun schema. While the words undoubtedly characterize a person in terms of what they do, English lacks the putative base
verbs (to) philosoph [f1'lusgf], (to) grace, and (to) barb. In the case of philosopher we
can appeal to the phonological similarity (and even orthographic identity) with the underlined portions of philosophy and philosophical. Not only this, but the phonological and
semantic relatedness of philosophy and philosopher matches the phonological and se-
149
PA150
150
mantic relatedness of other word pairs, such as geography and geographer, biography
and biographer. For barber, the situation is less clear, not only because there is no
verb (to) barb, but also because there are no semantically related words containing the
phonological sequence [ba:b], or phonologically similar sequences.
2.3. Thursday
The case of barber is not unusual. We often encounter words, a part of which can be
identified as a symbolic unit, the remainder of the word, however, lacks this status. Take,
as a simple case, the names of the days of the week. These all terminate in [-day], a unit
whose semantic contribution to the names is evident. Yet no meaning at all can be
attributed to [thurs-], except in the trivi al sense that [thurs-] contrasts with [wednes-]
and the dangling parts of the other weekday names.
2.4. perform
Sometimes the case for analysis may be quite compelling, even in the absence of semantic relatedness. Take the case of (mostly) bisyllabic words of usually abstract meaning.
Initial segments include ad-, con-, in-, p er-, pro-, trans-; final segments include -ceive,
-fer, -form, -late, -fain. To be sure, not all combinations are possible, though a good
many are. The semantic contribution of the constituent units is likely to be opaque to
English speakers, even to those who have an inkling as to their etymology. One would
be hard pressed to explain the meaning of p erform on the basis of the putative meanings
of per- and -form.
2.5. meat
In spite of the evident semantic relation between meat and eat - meat is someth ing that
you eat - we would probably reject outright an analysis of meat in terms of prefixation
of the verb eat with [m-]. This is because the analysis fails to fit in with other facts
about the language. Other words terminating in -eat have nothing to do with eating:
heat, bleat, pleat, etc. Neither are there any other examples of nouns being formed by
prefixing a verb with m-. There is therefore no basis for proposing the word-formation
schema [Nm-V]. Moreover, such a schema would conflict with general patterns of Eng1ish word-formation. While English is certainly not lacking in derivational prefixes (un -,
over-, out-, etc.), prefixation by a single consonant is rare. One of the few examples is
the negative import of initial n-, as in one/none, either/neither, or/nor, the now archaic
aye/nay, and dialectal owt ('something')/nowt ('nothing').
Even in the case of meat, however, we may need to tread with caution. In light of
the vagaries of English spelling, English speakers sometimes have recourse of mnemonic
aids for distinguishing the different spellings of homophones. Readers may be familiar
with the school-room rule for distinguishing the spellings <principle> and <principal>:
PA151
<principal>, with an <a>, is an t;!:_djective. Likewise, one can imagine poor spellers appealing to the semantic relation in order to distinguish <meat> from <meet>. These
observations serve as a reminder, not only of the speaker-specific nature of the mental
grammar, but also of the role of orthography in the mental representation of words.
2.6. more
Even though we would probably not want to recognize m- as a meaningful part of meat,
there are numerous cases where a phonological constituent may be felt to be associated
with a semantic value (Rhodes and Lawler 1981). Several words commencing in m- are
associated with a large number or quantity: many, more, most, multiple, magnitude, million, myriad, as well as, perhaps, magnificence, magnanimous, mighty. There are, of
course, many words in m- which do not share this semantic value, and some, even,
which convey the notion of smallness: minuscule, minimal, microscopic. Nevertheless,
the association may contribute to the feeling that the phonological form of seemingly
unanalysable words is not entirely arbitrary. Take the case of droop, discussed next.
2.7. droop
This word would probably be regarded as internally simple, yet correspondences and
resonances with other words can be detected. Consider words which terminate in -oop,
such as hoop, loop, coop, swoop, stoop, and scoop. These words invoke the notion of
curvature. (Imagine your posture if you are 'cooped up' in an enclosed space.) Now
consider words commencing in dr-. These include dry, drip, drop (of water), draft (of
beer), drain, dredge, drench , drought, draw (water from a well), drool, and dribble.
These have to do with water (or its absence). Remarkably, the meaning of droop is
almost compositional: curvature (of a plant) through ( lack of) moisture.
Often, the proposed semantic value appears to be based in facts of articulation or
perception, or in some more vaguely characterized synaesthesic association between
sound and meaning (Firth 1930). High front vowels - where there is a small aperture
between tongue and palate - tend to be associated with ideas of smallness, whi le consonant clusters with voiceless plosives may be associated with sudden or jerky movement
(split, spike, strip, etc.). The above discussed examples, however, lack this synaesthesic
component. By no stretch of the imagination can [dr] be considered a ''watery", or even
as a "dry" sound. Whatever their status, these conventionalized sound-symbolic relations
do, however, set up resonances within the language network and no doubt contribute to
speakers' intuitions that words somehow match their meanings.
2.8. motel
This word is a blend of motor and hotel. The initial part of the one word is dovetailed
with the final part of the other: [m:;iu[t]d], the blend having a semantic value akin to
151
PA152
152
that of a compound (motor hotel). Blend ing is very different from canonical cases of
analyzabi lity, since the sequences [ m~ut] and [ ti:J] do not in themselves have symbolic
status (with respect to the blend in question). Nevertheless, the viability of a blend
depends on the possibility of hearers being able to recover the input words, a crucial
factor in this regard being the cohort of words containing the phonological strings in
question (Gries 2006). It would be inconceivable to propose [l'ot] as a blend of land and
yacht, since initial [l] is hardly likely to activate the word land.
3. Entrenchment and drift; countering analyzability

Wheeler and Schumsky (1980) report a remarkable finding. When speakers of English
were invited to divide words up into their component parts, about half failed to mark
any internal division in the word baker. We can imagine speakers being reluctant to
mark a boundary in barber. But baker?
A likely explanation lies in the entrenchment of the word (Hay 2001; Taylor 2012:
131). A feature of entrenchment is that a unit does not need to be assembled (compositionally) from its parts on each occasion of its use, nor do language users need to refer
to its parts in order to understand it (Langacker 1987: 59). It already exists as a preformed unit and is stored and accessed as such. As a consequence, the unit's internal
structure tends to be obscured. Monitoring tasks show that speakers are slower to recognize a word (such as of) when it occurs as part of a frequently occurring string (such as
sort of) than when it is part of a less entrenched sequence (such as example of) (Sosa
and MacFarlane 2002).
One symptom of the enhenchment of (potentially analysable) units is the phenomenon of drift, whereby an expression takes on phonological and semantic properties in
addition to (or even at variance with) its compositional value; the expression, as it were,
acquires a life of its own and drifts away from its source. Inflected forms of nouns and
verbs typically have the ir own characteristic distributions. Plural eyes distributes in the
language quite differently from singular eye: the singular tends to be used in idiomatic
locutions (keep an eye on, etc.), whereas reference to the organ of sight dominates for
the plural form (Sinclair 2004). Derived forms often acquire semantic nuances additional
to their compositional values. Bybee (1985) cites the example of dirty, an adjective
whose use need not invoke the notion of 'dirt' . Similarly, baker, as an entrenched unit,
is not restricted to referring to 'one who bakes'; more commonly, the word is used as
the name of a kind of retailer, or retail outlet, which sells certain kinds of baked goods,
typically breads, pastries, and cakes. If you buy something 'from the baker', you are not
necessarily buying from 'the person who baked' the stuff.
Phonological drift - evidenced by such processes as assimilation, elision, palatalization , consonant lenition, and vowel reduction, sometimes operating cumulatively over
lengthy time periods - is liable to obscure the internal make-up of words. Although still
evident from the more conservative spelling, the relation of cupboards to cups and boards
is probably lost on most English speakers, as is the relation between state and station.
Informal enquiries suggest that many speakers fail to perceive the relatedness of preside
and president, and even of horizon and horizontal.
PA153
4. Characterizing the parts

Derived words - the example of farmer is typical - exhibit a marked asymmetry with
respect to the properties of their parts (Taylor 2002; Tuggy 1992). Some o:f the parameters are applicable to the analysis of phrases; others are specific to the domain of word
structure.
a) Autonomy vs. dependence. A unit is autonomous to the extent that it can be conceptualized in and of itself, without reference to other units. Although initially proposed
with reference to semantic units, in particular to verbs and their complements (Langacker 1992: 286), the notion also applies to phonology. Stop consonants are phonologically dependent in that we cannot conceive of a stop consonant without invoking
some notion of an adjacent vowel-like segment.
Suffixes, whether derivational or inflectional, are overwhelmingly phonologically dependent. The suffix [-er], being unstressed, absolutely requires as a host a stressed
syllable. Prefixes, on the other hand, tend to be phonologically autonomous. Forms
such as over- (as in overeat), out- (as in outperform), and anti- (as in anticommunist)
in fact exist as phonologically autonomous words. The same, of course, goes for
parts of a compound; both air and port - parts of airport - are phonologically
autonomous.
b) Schematic vs. contentful. Some conceptual units, such as those symbolized by do,
be, and it, have a highly schematic semantics; others, such as those symbolized by
basic level terms such as walk, sing, and table, are more contentful. Prefixes and
suffixes are notable for their semantic schematicity while the stems to which they
attach supply the bulk of the semantic content. Prefixed out- merely refers to some
schematically characterized process (whose content is supplied by the stem to which
it attaches) which exceeds some expected norm.
Some affixes are phonologically schematic. The (regular) plural morpheme of English is an alveolar fricative. Whether the sound emerges as voiced or voiceless, and,
if voiced, whether preceded by an epenthetic schwa, is determined by properties of
the host.
c) Valence. The specification of a unit often requires reference to the kinds of items
with which it can combine. (The phenomenon is studied in syntax under the rubric
of subcategorization.) A transitive verb (by definition) is one which requires a direct
object complement. Another way of putting this is to say that the characterization of
a unit may make reference to the (schematically characterized) construction in which
it needs to occur, in the case in point, the transitive clause construction. Noun-forming
-ness needs to attach to an adjective, while agentive -er attaches to verb stems. The
case of -er is, to be sure, more complex. Like many symbolic units, the suffix is
polysemous (Panther and Thornburg 2001; Ungerer 2007). Thus, on one of its meanings, the suffix can characterize a person in terms of their location, as in villager,
islander, Londoner, and back-bencher.
d) Selection (or choosiness). Syntacticians distinguish between subcategorization and
selectional restrictions. While drink subcategorizes for a nominal object, it semantically selects the name of a liquid. The "choosiness" of word-parts goes well beyond
matters of semantic plausibility. There is no semantic reason why apt and correct
should prefer the periphrastic comparatives more apt and more correct over the in-
153
PA154
154
fleeted forms apter and correcter, while poor and simple exhibit the opposite tendency, preferring the inflected forms poorer and simpler. One factor is the phonological
properties of the adj ectives - those terminating in a consonant cluster, more specifically an obstruent cluster, tend to shun the inflection (Hilpert 2008; Mondorf 2003).
Amongst suffi xes with roughly comparable semantic values we likewise find large
differences in their range of uses. While noun-forming -ness is fairly promiscuous,
in that it can attach to a very large and possibly open-ended set of adj ectival stems,
noun-forming -th, -ity, and -al are much more choosy, being virtually restricted to
occurring with a closed set of hosts. Factors which motivate these differences are an
important site of research (Baayen and Renouf 1996; Bauer 2001 ).
e) Profile determinant. This term refers to that constituent of an expression which determines the kind of entity that the expression refers to (Langacker 1987: 235). (The
notion thus corresponds to the traditional notion of head.) Suffixes are profile determinants. The fac t that farmer refers to a person, or that farmed refers to a past-time
event, is due to the affix. The case of prefixes is more diverse. Un -, n-, and anti- do
not impinge of the semantic type of the referent. Out-, on the other hand, imposes a
verbal conceptualization; (to) out-Nixon (Nixon).
f) Coercion. One unit may influence the phonological shape of a neighbouring unit. The
shape of the (regular) plural morpheme in English is determined by the phonological
properties of (the final segment of) its host. More complex are the effects of an affix
on its host. For example, a feature of adjective-forming -ic is that stress is placed
on the immediately preceding syllable; this, in combination with other phonological
schemas, may trigger changes in vowel quality. Compare photograph and photographic.
The above parameters allow us to characterize some of the basic categories in morphological theory (Taylor 2002; Tuggy 2005; Ungerer 2007; van Huyssteen 2010). Suffixes
(in English) are phonologically dependent, are semantically schematic, are profile determinants, they can be quite choosy with respect to the hosts to which they attach, and
can coerce the phonological structure of their hosts. Clitics, such as possessive s in
English, are phonologically dependent, are non-coercive, and can attach to practically
any kind of host (think of examples like The person I met's new car, where the clitic
attaches to a verb). Freedom of combination is also a defining feature. of words, along
with such features as phonological autonomy and (in most cases) semantic contentfulness. The approach also accounts for - indeed, leads us to expect - that the boundaries
of these categories will be fuzzy, and that some items will be able to "migrate" from
one category to another. When unstressed, prepositions have clitic-like properties, while
prefixes such as anti- and ex- can function as independent words, as befits their phonological autonomy. Parts of a blend can also achieve symbolic status. Kemmer (2003)
discusses the case of forms in -erati (chatterati, culturati, and, more recently, twitterati),
based, according to her, on the blend glitterati (glitter + literati).
5. Composition
The counterpart of analyzability is composition, the process of creating new forms from
existing resources. One approach - we might call it the Lego model - views composition
PA155
in terms of the "fit" between the parts. The model works tolerably well for syntax (at
least, for basic phrase structure configurations). Tea matches the subcategorization and
selectional requirements of drink; hence drink tea is an acceptable (grammatical) formation. The Lego model does not always work, even for syntax. By all accounts, lie 'to
speak untruthfully' is an intransitive verb. Yet in He lied his way through the interview
the verb takes a direct object. The transitive use of lie is imposed by the constructional
schema in which the verb occurs, whereby the specifications of the schema override the
properties of a constituent unit (Jackendoff 2010). The Lego model is even less appropriate for word-formation. For one thing, some word-formation processes, such as blending,
do not consist at all in the fitting together of already entrenched parts. But even when
parts can be identified, their combination is associated with all manner of restrictions,
idiosyncrasies, and overrides. One cannot, willy-nilly, attach -al to a verb stem to form
a noun designating an instance of the event. Cooker does not refer to a person who
cooks (the concept is pre-empted by the noun cook). A person who practises geology is
not a geologer (geologist is the term). Nor, for that matter, can one be sure that attaching
Isl to a singular noun will always generate the acceptable plural form.
5.1. Productivity
It is common in the literature to speak of the productivity of an affix. By this is meant,
not so much the number of forms which exhibit the affix, but the potential of the affix
for the creation of new forms. By this measure, -ness turns out to be a highly productive
noun-forming affix, -ity, -th, and -al much less so (Baayen and Lieber 1991). It would
be more accurate, perhaps, to speak of the productivity of the schema which sanctions
the use of the affix. Productivity is a function of entrenchment, which itself is a function
of frequency of occurrence. The relation to frequency is not a simple one, however (Bybee 1995, 2001; DClbrowska 2004). At issue is not the frequency with which the schema
per se is instantiated, but the number of different instances on which it is instantiated (in
comparison with the number of counter-indications). Word-formation processes which are
not particularly frequent in the language at large may nevertheless be productive, provided
that they are instantiated on a sufficiently wide range of examples. Nominals in -ee are
not all that frequent, yet the word-formation process appears to be quite productive, as a
consequence of the large number of different types. Equally, processes which are frequently attested may not be productive. The pattern of past tense formation exhibited by the
high frequency verbs do, say, and have is not able to be extended to other verbs.
5.2. The power of the schema

In view of its evident inadequacies, for both syntax and word -formation, the Lego model
is being abandoned by an increasing number of theorists in favour of an approach which
focuses on the sanctioning schemas (or constructions) (cf. article 12 on word-formation
in construction grammar; Taylor 2012; Tomasello 2003). As an illustration of the power
of schemas in word-formation, consider the example of hamburger (Taylor 2002: 293).
According to the Oxford English Dictionary, the word was used in the late 19th century
155
PA156
156
to refer to a minced meat patty, a so-called 'Hamburger steak', supposedly a speciality

of the city of Hamburg. At this time, the word was presumably analysed as [Hamburg]
+ [-er]. At some point, however, the word was re-analysed as [ham] + [burger]. We can
be confident of this because of the existence of forms such as cheeseburger, and even
the emergence of burger as an autonomous unit.
What could have triggered the reanalysis? Hamburgers, after all, are made of beef,
not ham. The crucial factor is not the semantics, but the phonology, more precisely, a
phonological schema. The stress pattern of hamburger ['cr ,cr cr] matches the stress pattern
of compound nouns; compare dog-whistle, clothes cupboard, soup-kitchen, and countless
more. When applied to hamburger, the schema indicates a comparable analysis, with [burger] emerging as the profile determinant and [ham-] as a kind of nominal modifier.
The way was then open for burger to emerge as an independent word and for it to be
modified by other words in a compound structure.
6. Conclusion
The minimalist ontology of cognitive grammar, in association with its rigid adherence
to a usage-based account of language acquisition, offers the possibility of a rigorous and
constrained description of the internal structure of words and of the processes which
sanction creative word-formation. As will be apparent from the brief overview presented
in this article, the cognitive grammar approach relies heavily on the notions of entrenchment and of schematic representations; the latter arise through familiarity w ith instances
and are able to sanction both the analyzability of encountered expressions and the creative construction of new expressions. Moreover, the issues raised in this article are by
no means unique to the analysis of words and the creation of new words, being applicable both to "pure" phonology and to syntactic organization alike. To this extent, wordformation takes its p lace in a broad, unified conception of the symbolic system which
constitutes a human language.
7. References
Baayen, Harald and Rochelle Lieber
1991
Productivity and English derivation: A corpus-based study. Linguistics 29: 801 - 843.
Baayen, Harald and Antoinette Renouf
1996
Chronicling the Times: Productive lexical innovations in an English newspaper. Language 72: 69-96.
Bauer, Laurie
2001 Morphological Productivity. Cambridge: Cambridge University Press.
Benczes, Reka
2007
Creative Compounding in English. The Semantics of Metaphorical and Metonymical
Noun-Noun Combinations. Amsterdam/Philadelphia: Benjamins.
Bybee, Joan
1985
Bybee, Joan
1995
Regular morphology and the lexicon. Language and Cognitive Processes 10: 425- 455.
PA157
Bybee, Joan
200 1
Phonology and Language Use. Cambridge: Cambridge University Press.
D~browska,
2004
Ewa
Rules or schemas? Evidence from Polish. Language and Cognitive Processes 19: 225-
271.
Firth, John
1930
Speech. London: Benn's Sixpenny Library.
Gries, Stefan Thomas
2006
Cognitive determinants of subtractive word formation: A corpus-based perspective. Cognitive Linguistics 17: 535- 558.
Hay, Jennifer
200 L Lexical frequency in morphology: ls everything relative? Linguistics 39: 1041-1070.
Hilpert, Martin
2008
The Engl ish comparative - language structure and languag e use. English Language and
Linguistics 12: 395-417.

Jackendoff, Ray
2010
Meaning and the Lexicon. The Parallel Architecture 1975-2010. Oxford: Oxford Uni-
versity Press.
Kemmer, Suzanne
2003
Schemas and lexica l blends. In: Hubert Cuyckens, Thomas Berg, Rene Dirven and
Klaus-Uwe Panther (eds.), Motivation in Language, 69-97. Amsterdam/Philadelphia:
Benjamins.
Langacker, R o nald W.
1987
Foundations of Cognitive Grammar. Vol. 1: Theoretical Prerequisites. Stanford: Stanford University Press.
Langacker, Ronald W.
1991
Foundations of Cognitive Grammar. Vol. 2: Descriptive Application. Stanford: Stanford
University Press.
Langacker, Ronald W.
2008
Cognitive Grammar. A Basic Introduction. Oxford: Oxford University Press .
Mondorf, Britta
2003
Support for more-support. In: Gunter Rohdenburg and Britta Mondorf (eds .), Determinants of Grammatical Variation in English, 251-304. Berlin: Mouton de Gruyter.
Nesset, Tore
2008
Abstract Phonology in a Concrete Model. Cognitive Linguistics and the MorphologyPhonology Interface. Berlin: Mouton de Gruyter.
Panther, Klaus-Uwe and Gunter Radden (eds.)
20 11
Motivation in Grammar and the Lexicon. Amsterdam/Philadelphia: Benjam ins.
Panther, Klaus-Uwe and Linda Thornburg

200 I
A conceptual analysis of English -er no minals. In: Martin Putz, Suzanne Niemeier and
Rene Dirven (eds.), Applied Cognitive Linguistics. Vol. 2, 149- 200. Berlin: Mouton de
Gruyter.
Rhodes, Richard and John Lawler
1981
Athematic metaphors. In: Roberta Hendrick, Carrie Masek and Mary Frances Miller
(ed s .), Papers from the Seventeenth Regional Meeting of the Chicago Linguistic Society,
318- 342. Berkeley, CA: Chicago Linguistic Society.
1964 [ 1916]
Cours de linguistique generale. Paris: Payot.
Sinclair, John
2004
Trust the Text. Language, Corpus and Discourse. London: Routledge.
157
PA158
158
Sosa, Anna Vogel and James MacFarlane

2002
Evidence for frequency-based constituents in the mental lexicon: Collocations involving
of Brain and Language 83: 227- 236.
Taylor, John R .
Cognitive Grammar. Oxford: Oxford University Press.
2002
Taylor, John R .
2004
The ecology of constructions. In: Gunter Radden and Klaus-Uwe Panther (eds.), Studies
in Linguistic Motivation, 49-73. Berlin: Mouton de Gruyter.
Taylor, John R.
2012
The Mental Corpus. How Language is Represented in the Mind. Oxford: Oxford University Press.
Tomasello, Michael
Constructing a Language. A Usage-Based Th eory of Language Acquisition. Cambridge,
2003
MA: Harvard University Press.
Tuggy, David
1992
The affix-stem distinction: A Cognitive Grammar analysis of Data from Orizaba Nahuatl. Cognitive Linguistics 3: 237-300.
Tuggy, David
2005
Cognitive approach to word formation. In: Pavo l Stekauer and Rochelle Lieber (eds.),
Handbook of Word-Formation , 233 - 265. Dordrecht: Springer.
Ungerer, Friederich
2007
Word-formation. In: Dirk Geeraerts and Hubert Cuyckens (eds.), Th e Oxford Handbook
of Cognitive Linguistics, 650- 675. Oxford: Oxford University Press.
van Huyssteen, Gerhard B.
2010
(Re)defining component structures in morphological constructions: A Cognitive Grammar perspective. In: Alexander Onysko and Sascha Michel (eds.), Cognitive Perspectives on Word Formation, 97- 126. Berlin/New York: de Gruyter.
Wheeler, Cathy J. and Donald A. Schumsky
1980
The morpheme boundaries of some Engl ish derivational suffixes. Glossa 14: 3- 34.
John R. Taylor, Christchurch (New Zealand)
11 . Word-formation 1n optimality theory

I . Introduction
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
Markedness vs. faithfulness in phonology

Constraint interaction and the prosodic organization of affixes
Affix placement
N eutralization in affixes
Faithfulness in base-derivative relations
U nmarkedness in hypocoristics
A llomorphy and gaps
Anti-faithfulness or anti-homophony constraints
Criticism
Summary
References
PA159
11. Word-formation in optimality theory
Abstract
Optimality theory (henceforth OT) models natural language competence in terms of
interactions of universal constraints, notably markedness and faithfulness constraints.
This article illustrates some of the major advances in the understanding of word-formation phenomena originating from this theory, including the prosodic organization of
morphologically complex words, neutralization patterns in derivational affixes, allomorphy, and in.fixation.
1. Introduction
The goal of OT, shared with all models of generative grammar, is to account for linguistic
well-fonnedness by teasing apart the rule-governed from the idiosyncratic and the universal from the language-specific. This task is approached in traditional generative grammar by mapping underlying forms, which are designed to encode the idiosyncratic, to
observable surface forms through a successive application of structure-changing rules.
OT addresses this task by mapping (hypothetical) inputs to observable surface forms by
selecting optimal forms among output candidates. The selection process involves the
evaluation of the candidates with respect to ordered, often inherently conflicting constraints. While the constraints themselves are assumed to be universal, the order among
them is largely language-specific (Prince and Smolensky 2004).
To clarify some of the major new ideas developed within OT, the model will first be
introduced with phonological data. This is because the pioneering work in OT concerns
phonology and OT analyses of word-formation phenomena have been largely focused
on the morphology-phonology interface (cf. McCarthy 2006). Articles with other foci
typically concern inflectional rather than derivational morphology (cf. Ackema and
Neeleman 2005; Wunderlich 2006; Xu 2011). OT-based approaches to word-formation
focusing on semantics are somewhat tentative at this point (Lieber 2010).
An emphasis on the phonology-morphology interface is motivated not only by the
work hitherto carried out in OT but also by the significance of this interface to one of
the central controversies in word-formation theory. This controversy concerns the question of whether word-formation phenomena can be adequately captured in a simple IA
(item-and -atTangement) model. Such a model presupposes inputs consisting of linearly
ordered morphemes, including both stems and affixes, where output forms may be subject to phonological modification. Noting that such modifications can be drastic and
"unnatural'', some have found such a model inadequate. Instead, it has been argued that
both inputs and outputs in word-formation consist of words to be related by a seri es of
processes, including both affixation and so-called "allomorphy rules" (Aronoff 1976).
This view raises the questions of how to capture shared properties between affixes and
stems and whether or not such IP (item-and-process) models are subject to any restrictions. OT is most relevant to this discussion as it allows for analyses which are consistent
with a simple IA model w ithout assuming phonological modifications of input forms .
159
PA160
160
2. Markedness vs. faithfulness in phonology

To illustrate the main differences between OT and previous generative grammar consider
the patterns of loanword adaptation in (1) (the transcriptions of the German words are
adopted from Duden 2005).
(1)
a. English
[bet] <bet>
[pep] < pep>
[step] < step>
[web] <web >
>
>
>
>
German
[bet]
[pep]
[step]
[vep]
b.
English
[breg]
[sprem]
[fren]
[ dred]
< bag >

<spam >
< fan >
<dad>
>
>
>
>
German
[bek]
[spem]
[fon]
[<let]
The data show that both English [] and [re] are uniformly adapted as [] in German.
Also English voiced and voiceless obstruents in syllable-final position are uniformly
adapted as voice less obstruents in German. In trad itional generative grammar the mapping of input to output forms involves a series of rules, which successively modify
structure. By contrast, in OT this mapping rests on the comparison of potential input
and output forms. G iven for instance the form / breg/ as an input, a device called GEN(erator) supplies output candidates, including the form identical to the input but also a
potentially infinite number o f forms deviating from the input in variou s ways. A few
example candidates are given in the lefthand column in (2). (It is assumed here that the
constraints of a phonological grammar concern the phonemic level of representation.)
breg
(2)
*Oas[+VOICE]coo
*[-back] /[+low]
*
a.
/ breg/
b.
/ bEg/
c.
/ brek/
d.
e.
llSf'
/ bEk/
/ pEk/
F AITH
*
*
**
***
The core component of the grammar, called EVAL(uation metric), is responsible for
ensuring that the actually attested form, in this case the form /bi;:k/, emerges as the
winner. (The symbol i& indicates the winner). This task is solved by ordering constraints
such that all rival candidates are eliminated from the competition and only the winner
remains. The relevant procedure involves the successive evaluation of the candidates
with respect to the ordered constraints, starting with the highest constraint, and eliminating candidates whenever they fare worse than at least one of the remaining competitors.
The task of the linguist is then to identify and order the constraints relevant for e liminating non-win.ning competitors.
Returning to the tableau in (2), a key role for selecting the winner is played by the
constraint F A ITH ("faithfulness"), a novel type of constraint introduced in OT which
militates against deviations between output and input structures. The assumption of such
a constraint implies that all candidates which are more similar to the input than the
winner must be eliminated due to intrinsic flaws (cf. (2a, b, c)). Such flaws concern
PA161
violations of markedness constraints, which express universal preferences pertaining to
the structural well-formedness of output forms. The two markedness constraints referred
to in (2) are stated in (3). The winner in (2d) differs from the more faithful candidates
in (2a, b, c) in that it satisfies those two constraints.
(3)
a.
b.
*[+low]/[- back]
*OBs[+VOICE] c oo
Low front vowels are prohibited

Voiced obstruents in coda position are prohibited
Whereas all candidates exhibiting greater similarity to the input than the winner are
eliminated due to violations of specific markedness constraints, those exhibiting less
similarity can be eliminated as a result of violating FAITH. The tableau in (2) illustrates
the elimination of the candidate /pek/, which, unlike the winner, does not contain any
segment with an unaltered correspondent in the input. Consideration of additional candidates such as /bak/, which like the winner also differs in only two segments, ind icates the
need to split FAITH into several subconstraints referring to individual features. Ordering
FAITH(back) above FAITH(+low) would account for the preference of the candidate
/bek/ over / bak/ . Additional possibilities for differentiating between faithfulness constraints are listed in (4). An adequate ordering of these constraints ensures the elimination
of other promising candidates such as /bregi /, /bre/, or /grep/ (for an overview over
constraints commonly used in OT, cf. Kager 1999):
(4)
a.
FAITH(back)
b.
FAITH(+low)
c.
MAXIMALITY
d.
DEPENDENCE
e.
LINEARITY
f.
CONTIGUITY
Correspondent segments have identical values for feature

[back] (No fronting or backing!)
Correspondent segments have identical values for feature
[+low] (No raising!)
Every segment in the input has a correspondent in the output (No deletion!)
Every segment in the output has a correspondent in the
input (No epenthesis!)
The input string reflects the precedence structure of the output string and vice versa (No reversal!)
Correspondent segments in inputs and outputs form contiguous strings (No morpheme-internal insertion or deletion!)
The description of the data in (1) is complete when all markedness and faithfulness
constraints have been ordered such that, for any given input, the relevant winner is
selected as the output. More generally, an OT grammar is complete, when the optimal
output is determined for any hypothetical input. This ban on any restrictions on inputs
is known as "richness of the base" and, together with the constraint-based parallel evaluation of output forms, constitute major innovations originating with OT.
Conceiving of grammar in terms of interacting markedness and faithfulness constraints sheds new light on the notion of function in language. For instance, the constraint
ranking in (2) indicates an advantage of German grammar compared to English in that
markedness constraints prevail, which facilitate the pronunciation, perception, and perhaps also the learning and storage of words. English, a language where FAlTH dominates
both markedness constraints, is on the other hand "better" in more fully exploiting the
contrastive potential by allowing more feature combinations in more positions. Hence,
161
PA162
162
the phonology of English, but not that of German, allows for the meaning differences in
(5) to be formally distinguished as follows:
(5)
/ breg / < bag > vs. /bt:g / < beg > vs. / brek/ <back> vs. / bt:k/ < beck>
This, in a nutshell, is the basic conflict to be resolved by a phonological grammar: the

need to optimize form by favoring maximally unmarked structures versus the need to
optimize the contrastive potential by allowing maximal freedom of choice for combining
features or phonemes. The observation that the resolution of this conflict can differ for
various positions can be captured by linking constraints to specific contexts. One approach is known as "positional faithfulness" (Beckman 1999), the other as "positional
markedness" (Zoll 2004). On the latter approach, illustrated in the tableau in (2), the
absence of voiced obstruents in coda position is captured by ordering the restricted
markedness constraint as specified in (3 b) above a general faithfulness constraint
FATTH(vorcE) (Corresponding segments have identical values for the feature [voice]),
cf. (6a)). By contrast, on the "positional faithfulness" approach, the absence of voiced
obstruents in coda position is captured by ordering a faithfulness constraint
FAITH(vmcE)oNs linked to the syllable onset above the constraint *Oss[+vo1cE], a
general markedness constraint against voiced obstruents (cf. (6b)), which then in tum
dominates FATTH(vo 1c E). As a result, voiced and voiceless obstruents contrast in the
syllable onset, but not in the coda. The notion of "positional faithfulness" hence captures
the well-known generalization that prominent positions tend to exhibit contrasts which
are neutralized in nonprominent positions (Trubetzkoy 1958: 2 12).
(6)
a.
b.
*Oss[ +vo1cE] coo >> FAITtt(vo1cE)

FAITH(vo1cE)oNs >> *Oss[+v01cE] >> FAITH(vmc E)
The notion of (positional) markedness has a close analogue in formal constructs such as
"surface structure constraints" or "filters" introduced in generative grammar (cf. Shibatani 1973). However, such devices have an awkward status in that model, in contrast to
the key role played by markedness constraints in OT. The innovation in OT without any
formal analogue in previous frameworks concerns the concept of faithfulness, that is,
constraints which actively favor sameness of structure between input and output. Given
the premises of OT, all permutations of universal markedness and faithfulness constraints
constitute possible grammars and all grammars consist of specific rankings of universal
markedness and faithfulness constraints. There are no constraints on permissible inputs.
3. Constraint interacti on and t he prosod ic orga nizatio n of affixes

Prosodic organization plays an important role in word-formation, to be demonstrated
throughout this article. Before addressing the conditions concerning the prosodic organization of morphemes in complex words, I will briefly present some constraints determining the basic organization of segments in syllables.
Typo logical studies of syllable structure have revealed a universal preference for
simple CV syllables, consisting of a consonant in the onset followed by a vowel in the
PA163
163
nucleus. In OT this preference is captured by various markedness constraints, including

ONSET (Every syllable needs an onset), NoCooA (Codas are prohibited), and *COMP0 Ns
(Complex onsets are prohibited). Other constraints limiting the association of specific
segments with specific syllable positions concern the position of a segment within the
sonority scale. The order of segments on this scale roughly correlates with the openness
of the vocal tract during the articulation, ranging from low vowels, the most open sounds,
to plosives, the maximally closed sounds. In OT, the relevant generalizations are expressed in terms of so-called anti-association constraints, which prohibit the association
of segments with specific syllable positions. The orders for onsets and nuclei are
sketched in (7a, b). The top ranking of "*O Ns/a" ("/a / is prohibited in the onset") in
(7a) means that /a/, due to its maximal sonority, is the worst occupant of the syllable
onset while the low ranking of "*ONs/ PLS" means that plosives, due to their minimal
sonority, are the most preferred segments in onset position. The opposite preferences
obtain for the nucleus position (cf. (7b)) (cf. Prince and Smolensky 1993; for the relevant
descriptive generalizations see also Vennemann 1988).
(7)
a.
b.
*ONs/a. >> *ONsli >> *ONSI R >> *ONs/l >> * O NSINAS >> *ONSI FRC >>
*ONS/ PLS
*Nuc/ PLS >> *Nuc/FRC >> *Nuc/NAS >> *Nuc/l >> *Nuc lR >> *Nuc/i >>
*Nucla
The ranking of faithfulness constraints relative to these anti-association constraints accounts then for the thresholds observed in individual languages: in German only the
least sonorous vowel Iii is allowed in the onset (cf. /ia/ <ja> 'yes' ) whereas in English,
both Iii and /u/ are allowed (cf. lies/ < yes>, /ui/ < we >). These examples also illustrate
the effect that syllabic organization has on the phonetic realization of a phoneme. In
both /ui/ <we > and /iu/ <you> the respective first vowel is pronounced as a short glide
followed by a longer and more steady sound (i.e. [wi:] <we>, [ju:] < you>). This is
because in both words the first vowel is organized as an onset while the second vowel
is organized as a nucleus.
Turning now to complex words in English, there are cases where the phonetics indicates the association of a stem-final consonant with the syllable coda, even when a
stressed vowel follows (cf. Kahn 1980). In (8a) /t/ is flapped before a stressed vowel,
which clearly signals that the vowel belongs to a separate stem. Within si mplexes /t/ is
always phoneticaily aspirated in that context (cf. (8b)). Aspiration also applies when the
plosive occurs in stem-final position before a suffix as in the word hittee shown in (8c).
(8)
a.
b.
c.
Phonetic surface forms:

[kh<ercu]
[1cuh1J
[hit"i :]
Morphological structure:
[kret]sTM[ai]sTM < cat eye >
< latte >
[latci]sTM
< hittee>
[h1t] sTM[i]sFX
The It/ allophony is readily explained with reference to prosodic structure. The trees in
(9) exhibit a hierarchical structure, ranging from composite groups (CG), over phonological words (m), feet o:::), syllables (cr), down to the subsyllabic constituents onset (0),
nucleus (N), and coda (C). In branching nodes, one of the daughters is strong, marked
by the subscript " s" in (9), the others are weak, marked by the subscript "w". Signifi-
PA164
164
cantly, in (9b, c) the /t/ occurs exclusively in the onset, in accordance with general
constraints on the syllabification of consonants before stressed vowels. Aspiration can
be considered a consequence of this prosodic structure. The apparent deviation from this
pattern in (9a) indicates the presence of an internal domain boundary, which prevents
the regular syllabification of It/ as the onset of the stressed syllable. As a consequence,
It/ associates with the coda, resulting in lenition (flapping) rather than fortition (aspiration). (It is assumed here that syllabic tense, high vowels in English associate with the
nucleus and coda.)
(9)
a.
b.
CG
A
ffis
0
I
U>w
I
cr
I
cr
/\
IT\
I I
NC
c.
(J)
CJ
L"
I:w :L.
CJ
cr
/\ /l\
(I
I I
( k re t )oi (Cl i)w
I
0
00
ONC
I I I
a t
i)w
CJ
/\ 11\
ONO
I
(h
[~]
i)(!)
[till
t
[r]
The representations in (9) raise the question of how to capture the associations between
boundaries of prosodic domains and those of specific morphological categories. In OT
such cases of systematic coincidences are described in terms of so-called "alignment"
constraints, a special type of markedness constraints (for a similar idea in pre-OT work
see Selkirk 1986). The general schema of such constraints, referred to as "generalized
alignment", is given in (10) (Cat= category, PCat =prosodic category (e.g., foot, syllable), GCat =grammatical category (e.g., stem, affix), cf. McCarthy and Prince 1993).
(10) Align {Cat 1, Edge!> Cat2 , Edge2) = def
\::/ Cat 1 3 Cat2 such that Edge 1 of Cat 1 and Edge 2 of Cat2 coincide.
Where
Cat 1, Cat2 E PCat U GCat
Edge t> Edge2 E {R (right), L (left)}
The alignment constraints relevant for mapping the morphological structures in (8) to
the prosodic domains (k~t)w (ai)w, ( latci)w, and (hrti)w concern stem and phonological
word boundaries. The "fused" prosodic domain of the complex word (h1ti)w, compared
to the two domains of (k~t)w(ai)w, could then be captured by ranking the constraint
ONSET between the two alignment constraints in (11). ("ALIGN (STEM, L, ro, L)" = "align
the left boundary of every stem w ith the left boundary of a phonological word"; "ALIGN
(STEM, R, ro, R)" = "align the right boundary of every stem with the right boundary of a
phonological word".)
( 11) ALIGN (STEM, L, ro, L) >> ONSET >> ALIGN (STEM, R, ro, R)
PA165
165
The constraint order in (11) accounts for the generalization that the vowel-initial suffix
-ee is integrated into the phonological word of the stem, to satisfy O NSET, while compound members such as eye form separate phonological words, regardless of the ir segmental structure. In functional terms, the grammar in (1 1) expresses the generalization
that in English, it is less important for syllables to have an onset than for boundaries of
individual compound members to be prosodified separately. At the same time, it is more
important for syllables to have an onset than to mark the boundary between a stem and
a suffix. As a result, stem-final It/ occurs in the onset position in (9c) but in the coda
position in (9a). This effect is reached without resyllabification rules employed in rule
based frameworks, where the It/ in the stem hit would disassociate from the coda to
reassociate with the onset position, once the suffix has been attached.
The ranking in ( 11) captures not only the difference in the prosodic organization
between compounds and suffixed words but also predicts potential differences between
vowel-initial and consonant-initial suffixes, which satisfy ONSET independently. The
ranking thus expresses a connection between the phonological shape of a suffix and its
prosodic organization, thereby accounting for the fundamental difference between socalled "cohering" (integrated) and "non-cohering" (non-integrated) affixes (cf. Dixon
1977). At the phonetic level, this difference is manifest in the absence of perfect rhymes
between words with consonant-initial suffixes and simplexes in English. This is because
the pronunciation of phonemes is necessarily affected by their position within higher
prosodic constituents. In English, vowels are potentially lengthened more extensively
when occurring in foot-final position as in (l 2c), compared to foot-internal position, as
in (12a, b).
(12)
a.
b.
U)
crs
cr
crw
11\ If\
C ONC
IV
(v
I 11
s).,
11\ If\
IV
(v
t
[ i:]
<Venus>
crw
ONC ONC
[i:]
[vin~s]sTM
c.
(l)
I 1 1
n <> s)(,,
[vin]sTM[ ~s] sFx

<venous>
CG
f\
cr
cr
If\
If\
/\ \/
I I I
ONC ONC
(fJ i )w n ~ s
t
[i:.]
[fJi]sTM[ n~s]sFx
<freeness>
As a result, words with vowel-initial suffixes such as venous can be homophonous with
simplexes such as Venus, whereas a word with a consonant-initial suffix such as freeness
will exhibit variable but measurable - often noticeable - phonetic differences. The wellaligned prosodic structures in (12c) further account for the systematic "stress-neutrality"
of consonant-initial suffixes in English. As a result of being outside of the phonological
word of the stem, they are outside of the domain of foot formation as well, unable to
influence the stress patterns of the stem. The same generalizations obtain in German,
where again vowel-initial suffixes cohere, whereas consonant-initial suffixes or compound members form separate prosodic domains. (The glottal stop in stressed suffixes
PA169
(17) a.
b.
c.
/ b-J/ <be->, /g-J/ < ge->, !-JR! <er->, / faR / <ver-> , /t 5 -JR/ <zer->, /:mt/ < ent->
hms/ < miss->, /um/ < um->
/yb-JR/ < iiber->, /unt-JR/ <unter- >, /duR<;J < <lurch->, (t) /vid-JR/ <wider->,
(t)/hmt-JRI <hinter->
Additional severe restrictions concern the distribution of consonants. German prefixes

which contain a s ingle voiced obstruent are structured such that this obstruent appears
in initial position followed by schwa (e.g., / b'J/, /g'J/ ). This generalization is captured by
formulating a grammar which for any of the inputs containing no consonant other than
for instance / b/ as in (18a) selects l b-JI as the optimal output. A fragment of such a
grammar is sketched in (18b).
(18) a.
b.
{lb -JI, l'Jbl , l b/, hbl, /bi /, /ubu/, /-Jb-J/} - l b-JI

*Nuc/PLS, *Oas[+VOICE]coo' FATTH(c ) >> FAITH(v ), DEPENDENCE,
M AXIMALITY
The need to faithfully preserve the consonant (FAITH(c)) and organize it prosodically,
along with high-ranking markedness constraints against associating / b / with the nucleus
(due to its low sonority, cf. (7b)), or with the coda (due to its voicedness, cf (3 b)) force
this segment into the syllable onset. A following vowel is needed to satisfy a constraint
requiring every syllable to have a nucleus. Schwa is the optimal choice here as it exhibits
the minimal structure needed to provide a syllable nucleus, without gratuitously violating
either faithfulness or markedness constraints prohibiting various sorts of structure. Full
vowels appear only when necessary to satisfy higher-ranking constraints, including a
constraint prohibiting schwa in the strong syllable of a foot or in a syllable closed by a
consonant other than {/n /, Ill, IR/} .
While /b / is forced into the onset, sonorants seem to avoid this position. This avoidance indicates that high-ranking anti -association constraints militating against sonorants
in the onset (cf. 7a) dominate the faithfu lness constraint LINEARITYp, which requires all
precedence relations among phonemes in the input to be preserved in the output. The
prefixes /urn/ or l 'JRI show that the relevant anti-association constraints als o dominate
the constraint O NSET, which prohibits onsetless syllables.
(19)
a.
b.
c.
{Im/, /mu/, !m-:J!, !-:Jm!, /urn/} - /um/

{/R'Jf/, /fa'J/, /fa/, /foR/, !Rf/, /'JR'Jf/, IR1f/, /Raf} - lfoRI
*ONS/R >> *0Ns/J >> *ONS/NAS >> LINEARITYp >> *ONS/FRC >>
*ONs/PLS
Apparent exceptions indicate the impact of higher-ranking markedness constraints.

Hence the linear order of the segments in the prefix /mis/ <miss> is due to the undominated constraint against Isl in word-initial position in German, along with the constraint
FAITH(c), which ensures that input /s/ surfaces as Isl, rather than /z/ or / t 5/. The necessary
organization of Isl in the coda forces Im/ into the onset. Arranging It/ after /n / in the
prefix /'Jnt/ allows for both consonants to share a place node ([+alveolar]), thereby reducing markedness. The preference for /'Jnt/ over / t-Jn/ or /n-Jt/ is further evidence for the
low ranking of ONSET in the grammar of German verbal prefixes.
169
PA171
171

ing from stem-fi nal Ii/ (cf. (21 b )), which shows the affinity between low sonority nuc lei
and stresslessness mentioned above.
(21) a.
( d1ref(ti)~:)ro
(tJci(ni).r)ro
(tJ~st(tih)ro
(pci(ih)ro
( flJ(ti)~:) ro
(1rn(tih:)ro
< draftee >

< trainee>
< trustee>
< payee>
< fl irtee >
< rentee>
b.
((k5fi)~:) ro
< coffee >

((tJ:}ukih )ro
< trochee >
< tepee>
((tipi):E)ro
(t~u(pi):E)ro ~ ((t:}upi):E)ro
< topee >
(bren(fih)ro ~ ((bcenfih)ro < banshee>
< rupee >
(m(pi)~:)ro ~ ((1upih)ro
Additional English suffixes associated with stable main stress include -eer (cf. suffixed
mountaineer versus simplex cavalier ~ cavalier), -ese (Japanese versus diocese) -ette
(kitchenette versus silhouette ~ silhouette) and -esque (picturesque versus obelisk). Similar cases abound in other languages. In German, final Ii/ is associated with stable main
stress when representing the feminine suffix -ie as in (22a), but appear s in the weak
syllable of the foot otherwise (22b).
(22) a.
b.
[ma.g]sT[iJsF-FEM
[teRa.p] ST [i]sF-FEM
[ melanlkol]sr [i]sF-FEM
[tsJmbiJN-MSC
(Jpa.gcti)N-PL
[ha.Ra.kiRiJN-NEU
(ma.(g i) ~J.,
(teRa.(pih)ro
( melanko(l ih:)ro
< Magie>
< Therapie >
< Melancholie>
'magic'
'therapy'
'melancholy'
( (t5:imb ih)ro
(Jpa.(gctih)ro
(ha.Ra.(kiRi):E)ro
< Zombie >

< Spaghetti>
< Harakiri >
'zombie'
' spaghetti'
'harakiri'
Similar effects can be observed for syllable structure. In French, final Ii/ is regularly
organized in the syllable margin when a syllabic vowel precedes (cf. / bui/ [buj ] < bouille > ' face ', /rei/ [rej] <oeil> 'eye' )), but consistently forms a nucleus when representing
a suffix. A s a result, the simplex /abci/ <abeille > 'bee' ends in a glide ([a.bcj]) while
the suffixed word [abe]sTM[i]sF-FEM < abbaye > ' abbey' ends in a sy llabic vowel ([a.be.i])
(cf. the feminine suffix /i/ in boulangerie ' bakery', charcuterie 'butcher shop').
The presence and stability of the marked structures in question could be captured by
positing high-ranking faithfulness constraints which target the prosodic structure of affixes, specifically the association of certain phonemes with the syllable peak (e.g ., French
[i]sF-FEM) or the stress peak (e.g., German [i]sF-FEM)
(23)
FAITH(PROS)AFF/ro >> MARKEDNESS >> fAITHro >> MARKEDNESS >> fAITHAFF
The ranking in (23) captures the restriction of "deviant" prosodic structure as in (2 1a)
and (22a) to cohering affixes.
The prosodic organization of other affixes is determined entirely by the respective
phonemes, which themselves are severely restricted due to the low ranking of FAITHAFF
in (23). To the extent that the relevant gene raliza tions require reference to fully specified
prosodic output structures they could hardly be captured in traditional genera tive frameworks based on rules for combin ing and modifying individual morphemes.
PA172
172
6. Faithfulness in base-derivative relations

Up to now, reference to faithfulness constraints has been aimed at describing potential
contrastiveness, with a focus on relevant differences between stems and affixes. There
is evidence for a second type of faithfulness in morphology, which concerns the relation
between a base, that is, a word forming the input to word-formation, and the derived
form.
To illustrate the need to distinguish the relevant constraints consider first the potential
contrastiveness between front and back low vowels in English before /m / in (24a), compared to the restriction to the back low vowel before syllable-final h i illustrated in
(24b). Reference to the syllable-final position is important as /re/ does appear before
heterosyll abic h i (cf. /kre1i/ < carry > (Kahn 1980)).
(24) a.
/ kam/ < calm > - /krem / <cam>
b.
/kru/ < car> - */kre1/
The distribution illustrated in (24) is a straightforward case of neutralization, captured in

OT by ranking a specific markedness constraint " *re1)0 " above a faithfulness constraint
FAITH(v), which requires feature values in output vowels to match those given in the
input.
(25)
/ kre1/
*re1)cr
/ kre1/
~
FAITH(Y)
/ kru/
The data motivating the second type of faithfulness are given in (26), where clipped
forms such as /sre1/ retain the vowel quality of their base despite deviating from the
regular patterns. Note that the base Sarah conforms to regular phonotactics as /re/ is not
followed by tautosyllabic /J/ .
(26) /sre1':J/ < Sarah> /lre1i/ <Larry> -
/ sre1/
/lre1/
The generalization that /sre1/ is not a possible word in English, except when related to a
base in which /re/ is fol lowed by (heterosyllabic) h i , is captured by a constraint FAITH 80 ,
which requires structural identity in the relation between a derived form and its base.
Ranking this constraint above the relevant markedness constraint and faithfulness to the
input (henceforth refe1Ted to as FAITtt 10) introduced earlier captures the conditions under
which /sre1/ is grammatical in English (cf. McCarthy and Prince 1995; Benua 1995).
(27)
/sre1 ~/
~
FA ITH 80
/sre1/
/sru/
*re1)cr
*
FAITH10
PA173
173
The concept of faithfulness in the relation between a given base word and its output
accounts for all so-called cyclic effects, including the observation, first noted by Brame
(1974), that cyclicity crucially involves reference to (surface) words rather than smaller
units such as roots. The observation that preservation of structure does not necessarily
concern the entire base word, but can target specific aspects thereof, can be captured
through positional faithfulness . For instance, the retention of the first vowel of lsre1:JI in
(26), as opposed to the last, is achieved by linking the relevant fai thfulness constraint to
a prominent position such as the stressed syllable or the first syllable in the phonological
word.
Recognizing both FAITH 10 and FAITH 80 allows for intricate constraint interactions.
Empirical motivation for such interactions comes from putative "rule ordering paradoxes" in reduplication (cf. McCarthy and Prince 1995, 1999; Russell 1997).
7. Unmarked ness in hypoco ristics

The cases of word-formation illustrated in (28) differ from typical affixations in that the
strings preceding the vocalic endings are not necessarily stems. (This is assuming a
notion of stem consisting of a word with recognizable affixes removed.) In addition, the
formations are atypical in that the endings exhibit few syntactic restrictions, at times
with no apparent effect on the category of the derived word (cf. Schneider 2003: 87 ).
(28) a.
weird]A
nuts]A
sick]A
derelict] N
aggressive]A
ammunitionJN
~
~
~
~
~
weirdo]N
nutso]A
sicko]A,N
dero]N
aggro]A
ammo]N
b.
cook]v
good]A
short]A
comfortable]A
vegetableJN
Australian]A N
~
~
~
~
cookie]N
goodyJintj.
shortyJN
comfyh
veggie]N
Aussie]A N
The final vowels in the formations in (28) appear to have primarily a pragmatic function,
acting as "familiarity markers" characterized by an informal tone and social closeness
(Quirk et al. 1985: 1584). Indeed, it is not just the vowels that serve this function but
the entire shape of the outputs, including the trochaic foot structure. Formations characterized by a fixed phonological shape are known as "templatic" (McCarthy and Prince
1986), seemingly involving the extraction of some specific part of the structure of the
base, which may then combine with additional markers. This procedure is known as
"positive prosodic circumscription" and is also illustrated by the formations in (26).
From an OT perspective, the formations in (28) can be described without assuming
these separate steps but instead indicate dominance of various markedness over faithfulness constraints (cf. McCarthy and Prince 1993; Benua 1995; Lappe 2007). Ranking the
constraint FooTBlN, which requires binary feet, over FAITH 80 captures the fact that each
output in (28) consists of a single trochaic foot. Reference to the foot also accounts for
the choice of the "familiarity markers" as high front vowels and schwa are the only
phonemes which can occupy the nucleus of the weak syllable in a foot in English. This
restriction is revealed by the potential of flapping of intervocalic It/ before just these
vowels in American English (cf. /si[r] i/ < city>, /di[r];)/ < data>, /6[r];)u/ <auto>), as
PA174
174
opposed to necessary aspiration before all other vowels (cf. /la[th]i/ <latte>, / keu[thNu /
< kowtow>, /tu[th]u/ <tutu>, /iu[th)Q.I <Utah>). This difference is indicated in the structures in (29), where /t/ occurs foot-internally, versus the structure depicted in (9b) above,
where It/ occurs foot-initially.
(29)
a.
b.
{J)
(!)
:L
crs
crw
crs
/l\ /l\
/\ 11\
ON ONC
ONC ONC
I I I
(n a i
crw
I\ I
IV
t i )(1)
(b I a. t
I I
u)w
[r]
< blotto>
[r]
<nightie>
Non-occurring "familiarity markers" such as /u/ or la/ can accordingly be ruled out by
independently motivated markedness co nstraints prohibiting their association with the
nucleus of the weak syllable within a foot. In that position, tense Iii , the least sonorous
vowel, will emerge as the least marked vowel, as is reflected in the high productivity
of /ii-formations in English. The choice of !-:Jul as in (28a) presupposes some input
specification, which also accounts for the specific (derogatory) meaning associated with
those formations. Only when /u / is present in the input independently, as part of the
stem, will the relevant negative connotations be absent (cf. amm/u/nition > amm!-:Ju!
< ammo>).
Focusing now on the formation of Engl ish nicknames it appears that markedness
constraints determine not only their overall shape but also the "extraction sites". The
formations in (30a, b) indicate satisfaction of FATTH(w(cr) 80, which requires correspondence between the respective first syllables, regardless of stress. If, however, the wordinitial syllable is both unstressed and begins with a vowel, there is a tendency for the
respective stressed syllable to be "extracted" instead (cf. (30c)):
(30) a.
c.
(Abra):Eham
(Alfredh
(Esther):E
(Agnes)E
(Edward):E
(Elspeth):E
Au(gustus):E
An(t6nius)L
E(mmanuel):E
E(lisa):Ebeth
E(zeki)Lel
Antoi (n ette)~
~
~
~
~
~
~
~
~
~
~
(Abeh
(Alfieh
(Essie):E
(Aggieh
(Eddy):E
(Elspie):E
(Gussieh:
(T6ny):E
(Manny)L
(Lizzy)r
( Zeke)L
(Nettyh:
b.
So(phia):E
M a(ria):E
Ber(nard)L
Vic(t6ria):E
To(bias):E
Chris(tina):E
~
~
~
~
~
~
(S6phie):E
(Maryh:
(Barney):E
(Vicky):E
(T6by):E
(Chrissie):E
PA175
175

From an OT perspective the data in (30a, b) show that FATTH(w(cr) 80 is more important
than FAITH(L(cr)8 D , another positional faithfulness constraint which requires that footinitial syllables in the derived form and the base correspond. The data in (30c) indicate
that the simultaneous violations of both ONSET and FAITH(L) 80 would be worse than
violating FATTH(c.o(cr)80 . This condition can be captured by allowing constraints to conjoin, such that the conjunction is violated whenever each of its components is violated
locally, that is, by a single piece of structure (Smolensky 1993, 2006). Constraint conjunction captures the circumstance that a given structure may be tolerated despi te certain flaws, unless it is flawed in too many regards. The conjunction of ONSET and
FAITH(L(cr) 80 has the desired effect of eliminating the candidate /<igi/ for /a gi\st;:)s/, allowing its rival /gi\si/ to win (cf. (31a)). In (31b, c) the optimal nickname starts with the
same syllable as its base since the high-ranking constraint conjunction is satisfied . This
is because conjoined constraints are violated only when each individual constraint is
violated.
(31) a.
:l( gA.st~sh:
(5g ih:
~ (gA.sih:
b.
FAITH(w(cr)80
*ONSET
FAITH(:!:(cr) BD
*
*
lars(tin~h:
(tinih:
~
c.
{ONSET,
FAITH(:!:)so}
*
*
(laisi):i:
(h ):i:(n:J1):i:
~
(1Th
(n~uih:
The analysis in (3 1) does not account for the fact that several of the non-winning candidates are attested as well. Such variation, to the extent that it is associated with distinct
groups of speakers, indicates differences in the individual grammars, expressed in terms
of different constraint rankings in OT. The important prediction here concerns the limits
imposed by the theory on possible grammars. For instance, it is predicted that "extraction" of non-initial syllables in consonant-initial words such as /tini / from /b1stin;:)/
presupposes extraction in vowel-initial words, but not vice versa. It is further predicted
that extraction of vowel-initial strings such as /5gi / entails the extraction of consonantinitial strings, but not vice versa. These predictions follow from the existence of the
markedness constraint O NSET, which can have the effect illustrated in (3 ] ), while no
ranking of known constraints would produce the opposite effect. It is this sort of explanation and predictive power which distinguishes OT from approaches based on prosodic
circumscription rules ("Pick the word-initial syllable, unless . .. ") and can hardly be
mimicked by those.
A grammar where markedness constraints freely interact with FAITH 8 0 constraints
can capture intricate conditions on word-formation rules which appear outside of the
realm of rule-based approaches. Consider the modifications in the nicknames based on
female names in (32), which indicate satisfaction of the anti -association constraint *ONs/
PA176
176
In some cases h i appears to be replaced by 111, in others by various obstruents, which

have a correspondent in the portion following h i in the respective base words.
1.
(32) a.
Sarah
Mary
Caroline
Laura
---
b.
Sally
Mo!!y
Callie
Lo!!Y
Teresa
- Harriet
Florence
Chari!Y
---
Tessy
Ha!!)'
Flossie
Chattie
The consistent preservation of word-initial Ill as in Rachie from Rachel shows that
*ONSh is dominated by FAITH(ro(cr) 80 (cf. 33a). The choice of the specific underlined
obstruents in the nicknames in (32b) indicates the ranking of FAITH(c) 80, which requires
complete identity of a consonant in the base and the derived form, above CONTIGUITY
(see (4f)).
1itJ:)I < Rachel>
(33) a.
FAITH(c)BD
CoNTIGsn
(1i::i.tfih:
(lEi.tJih:
b.
*ONs/J
FAITH(00( cr) BD
sre1:) <Sarah>
(sre.1ih:
~
(sre.lih:
hre1i:)t < Harriet>
c.
(hre.1ih:
(hre.lih:
~
(hre.tih:
The choice of Sally as the optimal nickname based on Sarah, rather than *Sanny or
*Satty, which have less sonorous and hence better onsets, indicates faithfulness in basederivative relations. Ranking the constraint FAITH(APP) 80, wh ich requires identical
specifications for the feature [approximant], among the relevant anti -association constraints accounts for the selection of the candidate containing I ll.
(34)
sreJ:) <Sarah>
*ONs/J
(sre.1ih
(sre.nih:
~
(sre.lih:
f AITH(APP) BD
*ONs/1
*ONS/NAS
--
Again, the grammars in (33) and (34) express limits on what is possible . The top ranking
of ONsh among the relevant anti-association constraints in (7a) accounts for the fact that
the "replacements" affect precisely h i , rather than one of the less sonorous consonants.
While faithfulness constraints such as FAITH(APP) 80 can have the effect of limiting
distinctness among correspondents (e.g., the replacement of h i by 111, which is also
PA177
177

[+approximant]) there are no constraints which would have the effect of maximizing
distinctness (e.g., the replacement of h i by l pl ). While the high ranking of positional
faithfulness as in (33) can ensure the preservation of hi in word-initial position there is
no corresponding constraint which could ensure its preservation only in non-prominent
positions.
8. Allomorphy and gaps

The term allomorphy refers to the occurrence of phonemically distinct morphemes,
whose distribution is detennined by the linguistic context alone and which consequently
do not differ in meaning. Consider the alternation between the English adjectival suffixes lgl/ and lg1I in (35), the latter occurring only in combination with stems ending in Ill.
(35) a.
l tmib-:;,l/
< tribal>
I remkdgut-dl/ < anecdotal>
l k:}ust-gl/
< coastal>
b.
l pdul-:ul
<polar>
l p1dutdkal-d1 I <protocolar>
lkans:}l-:}JI
<consular>
Phonologically conditioned allomorphy as in (35) is analysed by ranking a markedness

constraint (i.e. *C;VCi ''The flanking of nuclei by identical phonemes is prohibited")
above faithfulness constraints. The constraint FAITH(APP) 80 again has the effect of
choosing the minimally different form ld1I in (36b). Locating the "repair" to the affix,
rather than the stem, can be expressed by ranking faithfulness to the segmental structure
of the stem above that of the affix. As a result, violations of *C;VCi are tolerated within
stems (cf. parallel, lull), but not in stem-affix combinations.
(36) a.
[tmib] STM+[;)l] AFF

IJ$"
FAITH(STM)so
*C;YCi
FAITH(APP)so FAITH(AFF)s o
I uaib-;) II
/ tiaib-;)1/
b.
(p;)Ul] STM+[;)l] AFF

/ p;)U(-;)(/
I pgur-glf
/ pgu(-gn/
Q'
c.
/ pgu[-g1/
[pre1gil]sTM
Q'
I preJ;)11 I
/ preJ;)ltJ/
*
*
The assumption of FAITH 8 0 constraints brings to light the close relatedness between
a llomorphy and gaps. Hence the avoidance of the cand idate l pdul-:}l/ in adjectival -a/suffixation due to the violation of *C;VCi is mirrored in nominal -al-suffixation in (37).
PA178
178
The cases differ in their response to the problem: in one case the suffix is "repaired"
(cf. (35b)), in the other suffixation is avoided altogether (cf. (3 7b)).
(3 7) a.
/ m 0d1~-;)l/
h~phiv-;)l/
/ b1tu:i-;)l/
<withdrawal>
<upheaval>
< betrayal>
b.
/ falf11/
/ b1gl?il/
/mJ;)Ul/
< fulfill >

< beguile>
<enroll>
+ /;)l/ ---+
+ /;)l/ ---+
+ /;)l/ ---+
0
0
0
An OT grammar concisely captures the similarities and differences between the two
suffixes. Both are associated with the undominated markedness constraint *CiVCi and
undominated f AITH(STM) 80 . They differ in that faithfulness to affixes is violable for
adjectival, but not for nominal, -al-suffixation (for more discussion of gaps see Raffelsiefen 2004; Rice 2007).
Returning to the account of allomorphy in (36), the analysis entails neither the assumption of several allomorphs in the input nor a process whereby underlying /;)l/ in the
suffix is changed to /;)J/ in the output. Rather it expresses static constraints concerning
the structural identity in the relation between input and output forms. Deviations from
identity in this relation are conditioned by higher-ranking phonologican markedness constraints and are always minimal. From this perspective many cases of allomorphy which
have been labelled "suppletive" in rule-based frameworks could be reanalysed in terms
of constraint interaction.
Consider the complementary distribution between the third singular masculine pronominal clitics /u / and /h / in Moroccan Arabic, where /u / appears after consonants and
/ h / appears after vowels (Harrell 1962: 136):
(38) a.
/menn-u /
/ktab-u /
/faf-u/
'from - him'
'book - his '
'he saw - him'
b.
/m\a-h /
/xt~a-h/
/fafu-h /
'with - him'
'error - his'
'they saw - him'
The data are consistent with the assumption of /u / in the input, which does not surface
after vowels because the hiatus would violate ONS ET. Assuming that epenthesis or deletion are not allowed, O NSET can be satisfied only when /u/ corresponds to a non-vocalic
segment. / hi is a promising candidate as it agrees with vowels in being [+approximant]
(Clements 1992) and does not introduce place or manner features at odds with those of
/u /. (Historically, the /u /-/h /-a llomorphy goes back to the single morpheme /hu/.)
(39) a.
[menn] sTM+[u JcLT

~
FAITH(STM)so
ONSET
fA ITH( APP)sn
/menn-u/
/menn-h/
b.
FAITH(AFF)s 0
[m\a] sTM+[u]cLT
/m\a-u/
/m\a-k/
~
/m\a-h/
*
*
*
PA179
179
Assuming the possible representation of input morphemes as sets of allomorphs the

correct distribution could also be captured with reference to additional markedness constraints. That is, given the listed allomorphs {/u/, /h/ } in the input, the candidate /mennh/ would no longer violate faithfulness, but could be eliminated as a result of violating
*COMP-CODA (Complex codas are prohibited). An analysis of this sort accounts for the
allomorphy in cases like the passive marking in Turkish, where /n/ after vowel-final
stems corresponds to /ill after consonant-final stems (Underhill 1976: 332). (Additional
alternations are due to vowel harmony constraints and will be ignored here.)
(40) a.
/oku-n/
/de-n/
'to be read'
'to be said'
b.
/kaibed -il/
/ior-ul/
'to be lost'
'to be tired '
In this case it is presumably not possible to account for the allomorphy by positing
a single underlying morpheme. However, the phonological conditions determining the
distribution of the listed allomorphs {[n], [ii]} can be captured by ranking the markedness constraints O NSET and *COMP-CODA above faithfulness constraints (cf. tableau
(41)).
(41)
[oku] sTM+{[n], [ii]}

/oku-il/
ONSET
*COMP- CODA
/o ku-1/
~
FAITH(AFF)so
/oku-n/
[kaibed] sTM+ {[n], [ii]}

~
/kaibed-il/
/ kaibed-n/
Not all cases of allomorphy lend themselves to an analysis based on listed allomorphs.
For instance, positing both allomorphs fol/ and l:J1! in the input for English adjectival
-al-suffixation would cause a problem concerning the elimination of the faithful and
phonologically wellformed, yet ungrammatical, candidate / tmib-:J1/. To express the default status of the allomorph /:Jl/ it would be necessary to specify priority relations among
the listed allomorphs (i.e. {/~l/ < !:J1! } ). This necessity may support the alternative analysis in (36), which is based on a single underlying morpheme and requires no ordering.
In general, decisions concerning underlying forms are based on a principle called
" lexicon optimization" in OT (cf. McCarthy 2002: 76-78). This principle essentially
favors maximal similarity between surface and underlying forms, but di sfavors listed
allomorphs. The questions of whether or not allomorphs must be listed and possibly
ordered are decided based on possible constraint permutations: variation which cannot
be captured in terms of constraint interaction by assuming a single underlying form
requires listed allomorphs. The evidence for listed and ordered allomorphs presented in
the OT literature appears to be limited to inflectional morphology (cf. Mascaro 2007).
Similarly the evidence for assuming candidates consisting of entire paradigms in OT,
rather than individual word forms, concerns only inflectional morphology (Raffelsiefen
1995; McCarthy 2005). Such theory-internally motivated formal differences,. be it multi-
PA182
182
be accommodated by ranking the relevant faithfulness constraints above DISTTNCT( STM).

Objections to this option concern cases where different "anti-faithfulness effects" are
attested in a single language (Horwood 1999). However, the possibility that distinct
word-formation rules may associate with distinct constraint rankings. must be granted
independently, as is shown for instance by the patterns associated with nominal versus
adjectival -al-suffixation in English (cf. (35), (37)). The possible existence of different
"anti-faithfulness effects" is expected then, as long as different cases of word-formation
are concerned.
10. Criticism
OT grammars consist of ordered markedness and faithfulness constraints. There is consequently a prediction that all violations of faithfulness, be it allomorphy or infixation, are
caused by higher-ranking markedness constraints and must accordingly be reflected in
some sort of improvement of phonological output structure. This prediction has been
claimed to be refuted by cases where phonologically conditioned allomorphy or infixation are said to be clearly non-optimizing. Since one of the main characteristics distinguishing OT from other theories of word-formation is its predictive power, the validity
of such counter-examples has to be examined in detail for each individual case. The
investigation of nominalization in the Austronesian language Leti, which has been heralded as a prime example of allegedly non-optimizing allomorphy- and infixation patterns (Blevins 1999; Yu 2007), shows that at least some of the criticism is unfounded.
The regular phonologically conditioned variation in Leti nominalization is illustrated
in (4 7), where the surface affixal material appears in brackets. Before vowel-initial stems,
a prefix [nj] occurs as in (47a) while [nja] precedes stems starting with a consonant
cluster as in (47b). The data in (47c, d, e and f) illustrate the generalization that affixal
material does not precede verb stems starting with a single consonant. Instead the allomorph [nj] is infixed after initial obstruents (47c) and [j] after initial sonorants (47e).
The glide does not occur before high vowels, leaving [n] as the only marker in cases
like (47d). Nominalization is accord ingly not marked at all in verbs with an initial
sonorant before a high vowel (47).
(47) a.
b.
c.
d.
e.
f.
[(n)j]atu
[nja]mnesa
k[nj]asi
t[n]utu
r[j]esi
ruru
'knowledge'
'equality'
'digging'
'support'
'victory', d[j] avra
' trembling'
'cut'
The glide in the affixes represents phonemic I ii and indicates syllabification in onset
position when another vowel follows. The allomorphy can then be shown to be consistent
with the assumption of a single underlying morpheme /ni /, such that all variants can be
captured in terms of markedness constraints dominating F A ITH(AFF) 8 0 . The insertion of
[a] in (47b) is optimizing in that this vowel appears when necessary to avoid the syllabification of Ii i in nucleus position. This condition is captured by ranking the anti-associa-
PA183
183
tion constraint *Nuc/i(oo) >> above DEPENDENCE(AFF) 80 , which prohibits the insertion
of segments in affixes (cf. (4d)). The segment [a] emerges as it is the most sonorous
and hence optimal vowel to fill the nucleus position in the relevant prosodic context (cf.
(16a)). Similarly, the "deletion" of affixal material seen in (47d, e, f) indicates the ranking of MAXIMALITY(AFF) 80 below markedness constraints. The absence of Iii as in (47d,
f) satisfies a constraint against adjacent high vowels while the absence of In/ after sonorants as in (47e, f) satisfies a constraint against onset clusters with even or falling sonority. Reference to sonority may also capture the absence of In/ in cases like d[j]avra 'cut'.
This is because the segment represented with the symbol [d] does not precede other
consonants and hence patterns with liquids rather than plosives in Leti (cf. van Engelenhoven 1995: 53). (Words like lrpa:ril ' they pay' are not counter-examples as lrl ' they'
is a clitic.) In sum, the allomorphy shown in (47) is entirely optimizing, mostly indicative
of markedness constraints concerning sonority and syllable structure.
Turning now to the position of the affixes, it has been claimed that the sound patterns
resulting from infixation in Leti are "exactly opposite of those predicted by OT" (Blevins
1999: 384). Whereas simple prefixation in cases like /ni + kasi l would resunt in a maximally unmarked CV pattern, infixation as in lkniasil has been argued to produce marked
consonant and vowel clusters "for no obvious prosodic or phonotactic gains" (Yu 2007:
5). However, infixation does serve a purpose as can be demonstrated by comparing the
two wellformed cases of prefixation in (48a, b) with the ill-formed case in (48c). The
generalization is that prefixation is grammatical only for cohering affixation, where the
respective initial syllable spans both affix and stem phonemes (cf. (48a, b)), not for non cohering affixation (48c). (The fact that the affix-initial nasal optionally deletes in cases
like [(n)j]atu, but not in [nj a]mnesa, may be due to its occurrence in foot-initial position.)
(48) a.
b.
0)
I
L
/\
(J
(J
/\ /\
c.
(J)
/A
cr
(J
/l\ /\ /\
ON ON
ONC ON O N
/\ I I
(n i a
lni+atu/
I I
CG
(J
/\ I
I I
t u)Q)
I I
(n i a m n e s a)"'
lni+ mnesal
/\
ON
I I
n I
cr
cr
(J
/\ /\
O N ON
I I
I I
(k a S i)m
lni+kasil
Whether or not an affix is cohering is determined by the ranking between markedness

and alignment constraints in (49) ("*ONslcc" = "No two consonants in an onset").
(49) ONSET, *ONslCC >> ALIGN (STEM, L, m, L)
Assuming the distinction in the prosodic organizations in (48) the ungrammaticality of
(48c) can be captured by ranking a constraint *CG, which prohibits non-cohering affixation, above the faithfulness constraint LINEARITY M The observation that *CG is satisfied
not by simple integration of the prefixes, but by infixation, indicates the ranking of
LrN EARITYM below ALIGN (STEM, L, m, L). The precise site of the infixation could fo llow
PA184
184
from a constraint on foot structure, as infixation in well -formed k[nj]asi, as opposed to

*kanisi, satisfies foot-binarity.
While more challenging cases of phonologically conditioned yet allegedly non-optimizing allornorphy or infixation patterns may exist, it appears that much of the criticism
is due to the less than full exploration of the analytical possibilities provided by the
theory. This raises the question of what exactly those possibilities are, in particular, what
the full set of markedness constraints is and on the basis of which criteria they are
established. Another question concerns possible limits on constraint orderings, including
limits on conjunctions (cf. Smolensky 1993, 2006). Finally the theory must tackle phenomena such as gradability and exceptions. For instance, Leti includes a few nominalizations which deviate from the regular patterns in (47), including cases. where the prefix
nia precedes stems not starting with a consonant cluster. Also the grammar in (36), as it
stands, accounts for the preference of polar to polal in English, but says nothing about
the unacceptability of both goalal and goalar as adjectives based on the noun goal.
While differences concerning productivity may be amenable to modeling in OT to some
extent (cf. "stochastic OT") it appears that the phenomenon that semantic coherence and
stability in base-derivative relations may go hand in hand with entirely unproductive
types of exponence (e.g., English hot - heat, strong - strength, hate - hatred) is beyond
all generative theories focused on accounting for grammatical output forms. For an approach in which the modeling of morphological analyses to account for relatedness
among existing words is separated from a model for morphological synthesis, which
then accounts only for productive rules, see Raffelsiefen (2010).
11. Summary
OT is a highly restrictive linguistic theory in that it imposes strict limits on how grammars can be modeled. Restrictions on input forms are ruled out and so is reference to
intermittent steps in the relation between input and output forms. Instead, grammars
assign the optimal output form to any (hypothetical) input form by evaluating candidates
in parallel with respect to ordered constraints. There are only two types of constraints,
markedness constraints concerning well-formedness in output forms and faithfulness
constraints, which require identity (possibly also non-identity) in the relation between
inputs and outputs.
Regarding word -formation, such a grammar can be shown to account for seemingly
drastic modifications of structure, manifest in allomorphy or in varying position of affixes, by assuming a simple IA model. The idea is that all such modifications reflect the
domination of faithfulness by markedness constraints. Whether or not such an approach
is empirically sound or whether or not the theory can be amended to accommodate all
counter-evidence remains to be seen. For now it holds that OT-based analysis have
brought to light numerous connections between observable aspects of morpheme shape
and universal markedness constraints which appear to yield insight into the fundamental
causes of surface deviations from simple IA-structure. Assuming that the relevant connections are not all coincidental it can be said that OT surpasses competing theories with
respect to its explanatory potential.
PA185
12. References
Ackema, Peter and Ad Neeleman
2005
Word-formation in optimality theory. In: Pavo I Stekauer and Rochelle Lieber (eds.),
A lderete, John D.
200 I
Dominance effects as transderivational anti-faithfulness. Phonology 18(2): 201 - 253.
Aronoff, Mark
1976
Wordjormation in Generative Grammar. Cambridge, MA: MIT Press.
Beckman, Jill N.
1999
Positional Faithfulness. An Optionality Theoretic Treatment of Phonological Asymmetries. New York: Garland.
Benua, Laura
1995
Identity effects in morphological truncation. In: Jill N . Beckman, Suzanne Urbanczyk
and D ickey Walsh (eds.), University of Massachusetts Occasional Papers in Linguistics
18: Papers in Optimality Th eory, 77- 136. University of Massachusetts, Amherst.
Blevins, Juliette
1999
Untangl ing Leti infixation. Oceanic Linguistics 38(2): 383-403.
Bloomfield, Leonard
1933
Language. New York: Holt Rinehart and Winston.
Brame, Michael
1974
The cyc le in phonology: Stress in Palestinian, Maltese, and Spanish. Linguistic Inquiry
5: 39-60
Clements, Geoge N .
1992
Phonological primes: Features or gestures? Phonetica 49: 181 - 193.
Dixon, Robert M. W.
1977
Some phono logical rules in Yidiny. Linguistic lnqui1y 8: 1- 34.
Duden
2005
Duden. Das Ausspracheworterbuch. Ed. by Max Mangold. Mannheim: Dudenverlag.
Gregersen, Edgar A .
1974
Consonant polarity in Nilotic. In: Edgar Voeltz (ed.), Third Annual Conference on African Linguistics, 105- 109. Bloomington, TN: Indiana University Press.
Harrell, Richard S.
1962 A Short R eference Grammar of Moroccan Arabic. Washington: Georgetown University
Press.
Harris, John and Edmund Gussmann
2002
Word-final onsets. UCL Working Papers in Linguistics 14: 1- 42.
Horwood, Graham
1999
Anti-faithfulness and subtractive morphology. ROA-466. N ew Brunswick, NJ: Rutgers
University.
Horwood, Graham
2002
Precedence faithfulness governs morpheme position. In: Line Mikkelsen (ed.), Proceedings of the 21st West Coast Conference on Formal Linguistics, 166- 179. Somerville,
MA: Cascadilla Press.
Jones, Daniel
2006
Cambridge English Pronouncing Dictiona1y. Ed. by Peter Roach, James Hartman and
Jane Setter. Cambridge: Cambridge University Press.
Kager, Rene
1999
Optimality Th e01y. Cambridge: Cambridge University Press .
Kahn, Daniel
1980
Syllable-based Generalizations in English Phonology. N ew York: Garland.
1
185
PA188
188
12. Word-formation in construction grammar

l.
2.
3.
4.
5.
6.
7.
Introduction
The hierarchical lexicon
Holistic properties of word structure
Phrasal and phrase-based lexical items
Form-meaning asymmetries
Conclusions
References
Abstract
The notion "construction" that plays a central role in construction grammar, is an
indispensable notion for the analysis of word-formation patterns. In the study of wordformation, we investigate the systematic correspondences between form and meaning at
the word level. Constructional schemas provide an adequate format for expressing these
systematic correspondences. Moreove1~ they are part of a hierarchical lexicon in which
both complex words and morphological patterns of various levels of abstraction can be
specified. An important advantage of the construction morphology approach is that it
can express the relevant similarities between morphological and phrasal lexical expressions, and the paradigmatic relations between morphological and phrasal schemas.
Thus, lexical knowledge is characterized as a complicated network between words and
phrasal expressions on a range of levels of abstractions, varying between individual
words and completely abstract patterns.
1. Introduction
Word-formation is the domain of linguistics in which systematic correspondences between the form and meaning of complex words are studied. Consider the following sets
of English adjectives:
(1)
steady
social
suitable
stressed
sympa thetic
unsteady
unsocial
unsuitable
unstressed
unsympathetic
The meani ng of the adjectives in the column on the right can be circumscribed as 'not
A', where A is the meaning of the corresponding adjective in the column on the left.
That is, there is a systematic correspondence between the presence of un- and the meaning component 'not'. Therefore, we consider the adj ectives in the rightmost column as
complex words, with the word structure [un[x] A]A, where x is a variable for a phonological string. Thus, we assign morphological structure to words on the basis of systematic
paradigmatic relations with other words.
PA189
The structure [un[x]A]A is part of the following constructional schema in which the
correspondence between form and meaning of a morphologically defined class of words
is specified:
The angled brackets demarcate a constructional schema. The correlation between form
and meaning is expressed by the double-arrowed symbol ~. The meaning contribution
of the base word on the right of the arrow is co-indexed with the relevant part of the
formal structure on the left of the arrow. The schema thus represents the meaning of
these un-adjectives as a compositional function of that of their base words. The format
of these schemas derives from the parallel architecture framework, as developed in work
by Ray Jackendoff (Jackendoff 2002; 2009; 20 1O; 2013). The meaning (SEM) of the
base words is specified independently in the lexicon, whereas the meaning contribution
of affixes is specified in constructional schemas, since their meaning is not accessible
outside of the morphological structure in which they occur. Note, moreover, that the
meaning of un- depends on the kind of morphological structure it occurs in. In the
structure [un [x] NJ v for instance, the meaning of unis ' reversative action', as in the
denominal verbs un-cork and un-root. Hence, the meaning of the prefix un- cannot be
specified in isolation of the morphological structure of which it forms a part.
Word-formation patterns can thus be considered constructions at the word level, and
the individual complex words that instantiate these patterns are (morphological) constructs. The constructional schema in (2) differs from the format of the word-formation
rule as used in traditional generative morphology (Aronoff 1976) in that it is neutral as
to production or perception. This schema is a declarative statement that characterizes a
set of existing English complex adjectives, and at the same time indicates how new
adjectives of this type can be formed. This type of morphological knowledge can be
used both in language perception and in language production, and therefore, it is quite
appropriate for morphological regularities to be expressed in declarative form.
The notion "construction" has been shown to be essential for a proper characterization
of the syntax of natural languages, and this theoretical stance is referred to as construction grammar (Goldberg 2006). The use of the notion "construction" in the domain of
morphology has been argued for in my monograph Construction Morphology (Booij
2010). In this article I will present a number of arguments in favour of a construction
morphology approach to word-fonnation. In section 2 the concept of the hierarchical
lexicon is introduced. In section 3 holistic properties of complex words are shown to be
an argument for constructional schemas at the word level. Section 4 argues that there is
no sharp boundary between lexicon and grammar, a basic idea of construction grammar,
and that constructional schemas can be used in order to express the parallelisms between
morphological and syntactic constructs. In section 5 we will see that we need the concept
of paradigmatic relationsh ips between constructional schemas in order to account for
form-meaning asymmetries in complex words. Section 6 presents a summary of the
argumentation for a construction morphology approach to word-formation.
189
PA190
190
2. The hierarchical lexicon

The acquisition of word-formation patterns is based on knowledge of memorized complex words. Once a pattern has been discovered, this can be expressed in a constructional
schema that dominates the individual instantiations of this schema. The schema and its
instantiations co-exist, as the discovery of a schema wi ll not lead to removal from one's
memory of the set of complex words that formed the basis of the constructional schema.
This insight can be expressed in a hierarchical lexicon, in which abstract schemas dominate their instantiations. For instance, the following substructure can be assumed for the
English lexicon:
(3)
< [un[steady]A]A <-4 [not steady]> <[un[social] A] A <-4 [not social]> < [un[suitable]A]A <-4 [not suitable]>
The information about these three adjectives is almost completely redundant, as their
properties are inherited from the general schema by which these adject ives are dominated, and the meaning of their independently specified base words. The only non-redundant
information about these complex adjectives is that they exist, that is, that they are conventionalized complex words of present-day English.
The partial structure of the English lexicon specified in (3) has only one level of
abstraction above the level of the individual complex words. However, it might be necessary to have more levels of abstraction. A clear case for this necessity can be found in
the domain of compounding in Germanic languages. The most general statement about
Germanic compounds is that they are right-headed. This can be expressed at the topmost
level of the substructure of the lexicon for compounds. However, the various classes of
compounds may also have specific properties of their own. Examples are that verbal
compounding is unproductive in languages like Dutch, English, and German, and that
in Dutch AN compounds, unlike NN compounds, the non-head tends to be simplex. For
instance , whereas the AN compound hoog-bouw 'high-building, high rise building' is a
correct compound of Dutch, a compound like huizenhoog-bouw 'houses-high building,
very high rise building', with the complex adjective huizen-hoog lit. 'houses high, very
high' as modifier, is ungrammatical. Hence, the following partial hierarchy is relevant
for Dutch AN-compounds:
(4)
<[Xi Yj]Yk +-+ [SEMj with relation R to SEMi]i?'
I
<[Ai Nj]Nk +-+ [SEMj having property SEM;]k>
Condition: A; = simplex
The lower schema is an instantiation of the upper schema, with the category variables
specified as A and N , and the nature of the semantic relation R specified as 'having
property'; moreover, a restriction on the complexity of the non-head constituent applies.
Another argument for the use of intermediate levels of abstraction for the description
of regularities in compounds is that constituents with in compounds may have lexicalized,
PA191
yet productive meanings. That is, they may have bound meanings dependent on their
occurrence in compounds. An example from Dutch is the use of the noun dood 'death'
as a modifier in NA compounds, with the meaning 'very, to a high degree' as in:
(5)
dood-gewoon
dood-moe
dood-simpel
dood-stil
'very
'very
'very
'very
normal '
tired '
simple'
quiet'
This productive use of the bound meaning 'very' of the noun dood can be expressed in
a subschema for NA compounds that is dominated by the NA compound schema that in
its turn is dominated by the general schema for adjectival compounds:
In this schema one of the slots is lexically specified, and hence this is a constructional
idiom (Jackendoff 2002). The semantic specification for dood in (6) overrules the literal
meaning 'death' of dood when used as an independent word.
Words with bound meanings are often referred to as affixoids, as they are similar to
affixes in having bound meanings. By making use of the concept of ' constructional
idiom' we can avoid introducing a new morphological category for word constituents
besides words and affixes. A constructional idiom is a constructional schema in which
at least one slot is lexically fixed, and at least one slot is open.
We need this type of constructional schema for another form of boundedness as well.
Consider the German word Macher 'maker', discussed in detail in Joeres (1995). Joeres'
observation is that -macher with the regular meaning 'maker' is a very productive constituent of compounds, whereas it has a lexicalized meaning when used as an autonomous lexeme, namely ' strong personality who achieves a lot' . Hence, Joeres (1995: 151)
concluded that -macher can be qualified as a "Halbsuffix", that is, an affixoid. Note,
however, that macher is not one morpheme, as was the case for the affixoids discussed
above, but consists of two, the verbal stem mach- 'to make' and the agentive suffix -er.
Examples of this type of compounding in German are:
(7)
with A as first constituent:

Fit-macher
'fit-maker'
Krank-macher
'ill-maker'
Wach-macher
' awake-maker'
with N as first constituent:
Baby-macher
'baby-maker '
Eis-macher
' ice-maker '
Programm-macher 'program-maker '
Different from what is at stake with the noun dood, the meaning of macher in these
words is completely regular: it has the meaning ' entity that causes or creates something' .
This meaning, however, is only productive within compounds, and not available for the
word Macher in isolation, which only has the lexicalized meaning mentioned above. In
191
PA192
192
order to account for its bounded productivity, we may assume the following constructional idioms for this class of compounds:
(8)
< [[Ai [mach]vj]v er]Nk +-+ [who [CAUSE TO BE]j SEMih

<[[Ni [mach]vj]v er]Nk +-+[who [CREATE]j SEMJ k
These structures presuppose that the -er-nouns are derived from potential but non-existing verbal compounds of the type AV and NV respectively. For instance, Fit-macher can
be seen as a derivation from.fit machen 'to make fit'. Thus, it is predicted that in these
complex words, macher has the expected, fully regular meaning of causer or creator.
The schemas in (8) will be dominated by the general schema for deverbal nouns in -er
in German.
The schema format proposed here makes it possible to account for the co-occurrence
of word-formation patterns. In (8) we saw a case of the productive co-occurrence of
verbal compounding and deverbal -er-derivation in German. Normally, verbal compounding in German is unproductive. That is, this is a case of "embedlded productivity"
(Booij 2010: ch. 3). Language users may use more than one word-formation process at
the same time, in order to construct a multiply complex word. There are, for example,
quite a number of English un-adjectives listed in Van Dale's English- Dutch dictionary
for which the base adjective in -able is not listed as well, for instance:
(9)
listed
un-beat-able
un-cornment-able
un-crush-able
un-say-able
not listed
beat-able
comment-able
crush-able
say-able
We can account for the formation of un-V-able adjectives from verbal bases without the
intermediate word, a deverbal adjective in -able, being in existence, by assuming that
word-formation schemas can be unified and have "a life of their own". The following
unified schema accounts for the adjectives in the left column of (9):
(10) [un[V-able]A]A
To conclude, the schema format is an appropriate format for characterizing word-f01mation processes and sets of complex words as parts of a hierarchical lexicon.
3. Holistic properties of word structure

One of the motivations for construction grammar is that constructions may have holistic
properties that cannot be reduced to properties of their individual constituents. This
applies to morphological constructions as well. A clear example is that reduplication
structures often indicate some form of increase of a semantic property denoted by the
base word. For instance, in Malay the plural form of nouns is created by adding a fu ll
copy of the base noun (e.g., buku 'book' - buku-buku 'books'). The meaning of plurality
PA193
193

does not derive from one of the constituents buku, but is evoked by the copying configuration as such.
An example from Dutch is the semantic interpretation of numerals discussed in detail
in Booij (2010: ch. 8). The numeral drie '3' and honderd ' 100' can be concatenated in
both orders, but with different interpretations: drie-honderd '300', honderd-drie '103 '.
Hence, the nature of the semantic relation between the two numeral constituents (multiplication or addition) is determined by the order in which the lower and the higher
numeral appear. That is, the semantic interpretation is partially a holistic property of the
complex numeral as a whole.
A prototypical example of holistic properties in word-formation is the interpretation
of nominal VN compounds in Romance languages, which denote agents, instruments, or
both, and sometimes other semantic categories such as events. Here are some examples
from Italian:
(11)
apri-concerto
apri-scatole
lancia-fiamme
lava-piatti
spazza-neve
porta-bagagl i
'open-concert, opening act'

'open-cans, can opener'
' throw-flames, flame thrower'
'wash-dishes, dish washer'
'plough-snow, snow plough'
'carry-luggage, luggage carrier I roof rack'
(event)
(instrument)
(instrument)
(agent or instrument)
(instrument)
(agent or instrument)
The semantic component of these VN compounds that is not expressed by either of their
constituents is the meaning 'agent I instrument I event'. It is this meaning that is evoked
by the morphological configuration as a whole. Note that the lexical category N of these
compounds is also a holistic property as these compounds are exocentric, and the N
constituent is not the head of the compound, neither formally nor semantically. These
holistic properties can be captured by a constructional schema of the following type:
(12) <[ViNj]Nk
[Agent/Instrument/ Event of ACTIONi on SEMjh>
This schema specifies a meaning component for which no explicit constituent is available.
In conclusion, holistic properties of word-formation constructions support the construction morphology approach to word-formation.
4. Phrasal and phrase-based lexica l items

An important argument for construction morphology is that it is able to express the
parallelisms between complex words and phrasal expressions with similar functions. A
class of phrasal lexical items that is very relevant for a proper analysis of the syntaxmorphology interaction is formed by particle verbs in Germanic languages (Los et al.
20 12). These particle verbs function as names for verbal concepts (events, actions). They
are phrasal in nature since the particle can be separated from the verb. In German and
Dutch, for instance, root clauses have a word order in which the finite verb appears in
second position, whereas the particle appears at the end of the clause. Particle verbs are
PA194
194
therefore phrasal lexical units, but similar in function to prefixed verbs. An example is
the class of Dutch particle verbs with the particle door. These verbs express continuative
aspect:
(13) verb
eten
fietsen
zeuren
'to eat'
'to cycle'
'to nag'
phrasal verb
door eten
door fietsen
door zeuren
'to continue eating'

'to continue cycling'
' to continue nagging'
The crucial observation is that the word door has a productive, yet bound meaning when
combined with a verb into a phrasal verb, namely that of continuative aspect. The following constructional schema expresses this semantic property:
This schema specifies redundant properties of existing pa1ticle verbs with door, and how
new particle verbs of this type can be coined. The schema in (14) is a good example of
a constructional idiom (Jackendoff 2002), a schema with both variable and lexically
fi lled positions.
In probably all European languages we find phrases with coordination that are semantically transparent, but have a fixed order, such as English salt and pepper, father and
son, ladies and gentlemen. These conventionalized expressions have to be listed because
of the fixed order in which the two words appear. On the other hand, they instantiate
regular and productive phrasal patterns, and we should express this in our description.
Therefore, the syntactic coordination schema [N and N]Nr, which is a sub-case of English coordination, dominates its instantiations, phrases such as salt and pepper. The only
non-redundant information concerning these N and N phrases is that they exist (that is,
are conventionalized), and the order in which the two words appear. Thus, there is no
principled difference between complex words and phrases with respect to the division
of labour between storage and computation. There is massive empirical evidence for
prefabs, word combinations stored in memory (Erman and Warren 2000; Kuiper et al.
2007; Sprenger 2003; Sprenger, Levelt and Kempen 2006; Tremblay and Baayen 2010;
Wray 2002; 2008).
Romance languages use various types of phrases as names for concepts. For example,
in French the patterns NaN and N de N (as in moulin a vent 'windmill' and salle de
bains 'bathroom' respectively) are used for constructing names for entities. This means
that French NPs with the form [N [a/de N)pp] NP have a lexical status, and function as
constructional idioms (Booij 2009).
A correct prediction of the position that both words and phrases are stored as lexical
units is that we will find competition and blocking effects between synonymous phrases
and words. For instance, the existence of the Dutch AN compound! zuur-kool 'sourcabbage, sauerkraut' impedes using the AN phrase zure kool 'sour cabbage' for this type
of cabbage, and inversely, the existence of the AN phrase rode kool 'red cabbage' impedes the coinage of the AN compound rood-kool. Related languages may differ in the
choice between synonymous but structurally different options. For instance, whereas
German is characterized by a very frequent use of AN compounds, Dutch speakers tend
to prefer AN phrases (as in German Rotwein 'red wine' versus Dutch rode wijn 'red
PA195
wine' (Booij 2002; Hi.ining 2010)). Interestingly, it has been shown that the choice
between AN compound and AN phrase in German for a new concept depends on the
family of related expressions. For instance, since there are many AN compounds with
Milch 'milk' as their head, a new AN expression for a new kind of milk will most
probably be a compound (Schli.icker and Plag 2011 ). This shows that new AN expressions are made on analogy to constituent famil ies, the sets of existing AN compounds
or phrases that share a constituent.
In Polish, nouns followed by a denominal relational adjective are used extensively,
where English would have NN compounds (Szymanek 2010):
(15) a.
b.
kosc [[sloni] wow] A-a

bone elephant-ADJ-FEM.SG
'ivory, lit. elephant bone'
sok [[jablk] Now]A-Y
juice apple-ADJ-MASC.SG
'apple juice'
Lexically stored collocational restrictions between A and N in AN phrases may also

affect inflection. In Dutch, the prenominal adj ective that is normally inflected and ends
in the suffix -e /g/ may be used without an ending when the AN phrase is a lexicalized
name, as in het stoffelijk overschot 'the mortal remains', where the expected schwa
ending is omitted (Speelman, Tummers and Geeraerts 2009).
The consequence of these observations is that there is no sharp boundary between
gramm ar and lexicon. Both at the level of the word and the level of the phrase, the
grammar contains schemas that dominate their lexicalized instantiations. Thus, the lexicon becomes a "construction".
The absence of a sharp boundary between lexicon and syntax is also manifested by
the fact that phrases may feed word-formation. It is in particular compounding that
appears to be permissive in this respect. As far as Dutch is concerned, it is far less
common to use phrasal units as bases of derivation, although this is possible for a few
suffixes, such as denominal -er ' -er' and denominal -achtig ' like'. The denominal suffix
-er in Dutch combines easily with classifying phrases:
(16) [[derde klass]NP er] N
[[17de_eeuw] NP er]N
[[doe-het-zelv]s er] N
' third form pupil'

'person living in the l 71h century'
'do it yourself-er'
In the linguistic literature one finds an extensive debate on the consequences of phrasal
compounds for our view of the architecture of the grammar (Meibauer 2007). Wiese
claimed that "a mechanism such as quotation must be held responsible for the existence
of phrasal compounds" (Wiese 1996). Meibauer (2007) objected to this claim, and noted
that often the use of phrases within compounds has an exp ressive function, and does not
always lead to conventionalization into lexical units. Expressiveness is not, however, a
formal precondition for using phrases as building blocks of words, as we wiH see below.
The expressive function of this type of word-formation is dominant when sentences
are embedded in compounds, as in the following English examples:
195
PA196
196
(17) Don't ask, don 't tell policy

(US army policy of dealing with gay soldiers)
One-size-fits-al1-educati on
(Boston Globe 6 March 20 10)
I understand the whole 'live it up, you're only in college once' thing
(Tufts Daily April 7, 2010)
The eat-your-spinach-approach to education
(Boston Globe 13 March 2010)
The sentential constituents in these compounds can be characterized as "fictive interaction" (Pascual and Janssen 2004), and they have indeed a strong expressive character.
Some of these compounds with a sentential constituent are conventionalized, as is the
case for the following Dutch compounds:
(18) God-is-dood-theologie
' God is dead theology'
blijf-van-mijn-lijf-huis
lit. 'stay away from my body-house, house for threatened women'
Not all phrasal compounds possess the property of expressiveness, however. What one
may expect is that when a particular phrasal pattern is used systematicaEly for the purpose
of classification (that is, for naming concepts), the resulting type of phrasal compound
will be quite normal, and will not carry a particular expressive value. This is indeed the
case for right-headed phrasal compounds with AN phrases as their left constituent, as is
illustrated here for Dutch:
'blind intestine inflammation,
appendicitis'
[[centraleA verwarmingsN] NP[monteur] N] N 'central heating technician '
'third-world country'
[[derdeA wereldN] NP [land] N]N
'round table conference'
[[rondeA tafel N] NP [conferentie]N] N
'free time clothing, casual dress'
[[vrije A tijdsN] NP kleding] N] N
'warm gulf stream'
[[warmeA golfNJ NP [stroom]N] N
In these examples, the AN phrase is a conventional lexical unit. However, it is not the
case that Dutch AN phrases can only feed compounding when they are conventional
names for concepts. The cruc ial property of AN phrases as parts o f compounds is that
they can function as classifiers. Hence, there are also phrasal compounds in which the
left AN constituent is not a conventional lexical unit by itself. This is the case for the
following Dutch phrasal compounds:
(20) [[drieNuM sterrenN]NP [hotel]N]N
[[langeA afstandsN] NP [loper] N]N
[[tweeNuM componenten N]NP [lijm]N]N
'three stars hotel'

'long distance runner'
'two components glue'
PA197
[[vierNuM kleurenN] NP [druk] N]N
[[witteA boordenN]NP [criminaliteit)N]N
197
'four colours print'
'white collars criminality'
For instance, there is no lexical unit drie sterren ' three stars' in Dutch. Crucially, these
AN phrases embedded in compounds do not function as referring expressions, but specify the type of N, where N is the head of a compound. This is also why the NP modifier
in these compounds cannot contain a determiner, as this would imply a referential interpretation of the modifying NP. It is reveali ng to compare this absence of a determiner
to the use of AN combinations such as blinde darm 'appendix; lit. blind intestine' when
not embedded in a compound. Since darm 'intestine' is a count noun, it must be preceded
by a determiner, and the determinerless phrase blinde darm is therefore ill-formed, unless
it is embedded in a word. Yet, these AN sequences are phrases, since the adjective is
inflected. The ending -e in blinde, for instance, is required since darm is a singular
common noun, and Dutch adjectives agree in gender and number with their head noun.
Agreement is a clear indication of phrasal status, since there cannot be agreement between parts of words, as this would go against lexical integrity, the defining property of
wordhood.
We thus observe that a particular type of word structure, compounding, licenses a
particular type of phrasal structure: AN phrases headed by count nouns but without
determiners. This once more underscores the point that schemas for morphological structure and those for phrasal structure are interdependent in the formation of lexical units.
5. Form-meaning asymmetries
In section 4 we saw that phrases can feed word -formation. This strengthens the position
that morphology and syntax cannot be completely separated. But what happens if a
word-formation process does not take phrases as bases for affixation? Consider the following examples from Italian:
(21) chitarra elettrica
flauto barocco
tennis da tavolo
'electric guitar'
'baroque flute '
'tennis table'
chitarrista elettrico 'electrical guitarist'

flautista barocco
'baroque flute player'
tennista da tavolo 'table tennis player'
The word sequences on the left and the right are NPs of the form A + N or N Prep N.
The phrase chitarrista elettrico denotes someone who plays the electric guitar. However,
the phrase chitarra elettrica lit. 'guitar electric' is not the formal base for the attachment
of -ista, a suffix that is used in Italian to create personal nouns. This is clear from the
word order (the suffix is not attached at the right edge, *chitarra elettricista), and from
the fact that the adjective elettrico agrees in number and gender with the masculine noun
chitarrista, and not with the feminine noun chitarra. This asymmetry between form and
meaning ("bracketing paradox") is the same as that in the famous English example
transformational grammarian discussed in Spencer (1988). In the case of English, one
might consider the suffix -ian to be formally attached to the phrase transformational
grammar, an option that is available due to English word order and the poor inflection
of English. Such an analysis is obviously impossible in the case of Italian, which only
allows for a phrasal interpretation:
PA198
198
(22) [[chitaIT-ista] N [elettrico] A] NP

This suggests that the adjective elettrico has semantic scope over a part of the complex
phrasal head guitarrista, and that the internal morphological structure of the modified
head word is accessible. In example (22), the adjective elettrico modifies the nominal
base of the head noun. In the following examples AN phrases used as modifiers are
turned into AA sequences, in order to fit the canonical phrase structure of Italian (examples provided to me by Daniele Vergilitto):
(23) energia solare
impiantoN energeticoA solare A
chirurgiaN esteticaA
interventoN chirurgicoA esteticoA
'solar energy'
'solar power plant'
'aesthetic surgery'
'aesthetic surgery intervention'
The same type of bracketing paradox has been observed for Polish. In Polish relational
adjectives follow the head noun. Hence, the structural paradox is the same as for Italian
(Szymanek 2010):
(24) chirurgia plastyczna
chirurg plastyczny
'plastic surgery'
'plastic surgeon'
praca fizyczna
pracownik fizyczny
'manual work'
'manual worker'
In these examples the post-nominal relational adjectives agree in gender with the preceding head noun. Hence the variation in form of the relational adjectives plastyczn- and
fizyczn-.
Let us look at these facts from the perspective of how to encode meaning. When an
Italian speaker knows the lexical phrase guitarra elettrica, and (s)he wants to construct
the lexical unit for denoting a person playing an electric guitar, the suffix -ista that is
used for creating personal names cannot be attached to the phrase guitarra elettrica
because this suffix is attached to words only: the noun *gitarra-elettric-ista is ill-formed.
The alternative is a phrase with a form-meaning mismatch in which the (stem forms of
the) relevant lexemes (guitarra and elettrico) that form a lexical collocation, are inserted
in the availa ble slots in a schema that complies with the restriction that -ista does not
accept phrasal bases. This is the schema [[N-ista] N A JNP The semantic property to be
specified is that the adjective has semantic scope over the base-N only.
Productive patterns with this kind of asymmetry can be accounted for by specifying
a paradigmatic relationship between two phrasal schemas (Booij 2010), where : : : denotes
a paradigmatic relationship:
(25) < [Nj Ak] NPi
+-+
SEM;> :::::;
<[[Nrista]N Ak]NPt +-+ [PERSON with relation R to [SEM;JJi>
Another example of form-meaning asymmetries is formed by words derived from particle verbs in Scandinavian languages. In the Mainland Scandinavian languages Danish,
Norwegian, and Swedish, the particle fo llows the verb, as is the case for English. How-
PA199
199
ever, the particle precedes the verb in past and present participles which have both verbal
and adjectival properties, and in deverbal nouns (Allan, Homles and Lundskaer-Nielsen
1995; Faarlund, Lie and Vannebo 1992; Holmes and Hinchcliffe 2003). Here are some
examples of participles of particle verbs in these three languages:
(26) particle verb
Danish
falde ned
past participle
'to fall down ' ned-faldne
(~bler)
'windfall (apples)'
Norwegian
'to deliver'
bring ud
Posten bliver ud-bragt
'The mail is being delivered'
Swedish
kora om
nng up
den om-korda lastbilen

upp-ringd
'the overtaken lorry'

' phoned'
'to overtake'
'to phone'
In the following deverbal nouns in Norwegian (Faarlund, Lie and Vannebo 1992), we
see again the reversal effect:
(27) particle verb
binde inn
rydde opp
skrive av
trekke opp
'to
'to
'to
'to
bind (a book)'
tidy up'
copy'
pull up'
nominalization
inn-bind-ing
opp-rydd-ing
av-skrif-t
opp-trekk-er
'book cover '

'tyding up'
'copy'
'cork screw'
The deverbal suffixes cannot be attached to the particle since it is not a verb, or part of
a verb. If the suffix were attached to the verb, we would get forms like bind-ing-inn. As
there are no particle nouns, particles cannot occur after a noun. And since Norwegian
compounds are right-headed, bind-ing-inn cannot be interpreted as a compound. Therefore, the nominalizations have the form of a compound, consisting of a particle followed
by the deverbal head noun. Hence, the formal base of these nominalizations is not the
particle verb as such, even though the meaning of these nominalizations is a compositional function of the (often idiomatic) meaning of the corresponding particle verb. In
the nominalizations, the lexical collocation of the particle and the verb is preserved, but
inserted in the reverse order, in the slots of the available compound type [X N]N, which
is the general template for nominal compounds in Germanic languages. There is a formmeaning asymmetry, since the particle verb is not a (non-discontinuous) subconstituent
of the nominalization. Yet, the meaning of these nominalizations is regular and transparent as they relate in a systematic, yet paradigmatic way to the corresponding particle
verbs.
The same expfanation can be invoked for the reversal effect in participles of particle
verbs. Since participles are adjectives, the only possible shape of adjectival forms of
particle forms is that of a right-headed adjectival compound.
The following form-meaning asymmetries have been observed in relation to particle
verbs in Eng lish, Dutch, German, and Italian (Booij 2010: 189):
PA200
200
(28) a.
b.
c.
d.
English particle verbs may carry the suffixes -er and -ing on their verbal head:
look-er-on, runn-er-up, digg-ing up, switch-ing off the lights;
the past participle of Dutch particle verbs is formed by prefixing ge- and
suffixing -ti-di-en to the stem form of the verbal head: aan-val 'to attack ' aan-ge-vall-en, op-bellen 'to phone up' - op-ge-bel-d; ge-nominalization also
applies to the head: rand-spring 'jump around' - rond-ge-spring 'jumping
around';
when German particle verbs undergo nominalization with the affix combination ge-e, this affix combination is attached to the verbal head of the particle
verb: herum-hops-en 'to jump around' - Herum-ge-hops-e 'jumping around'
(Muller 2003; 2006);
in Italian, nominalizing suffixes are attached to the verbaE head (Francesca
Masini, pers. comm.): venire giit 'to come down' - la venuta giit 'the coming
down ', mangiare faori 'to eat out' - la mangiata fuori 'the meal at a restaurant'.
In all cases, canonical morphological f01ms are used for creating denominal forms of
particle verbs. Note that, since Italian has left-headed compounds, the particle can appear
after the deverbal noun, unlike what is the case for Germanic languages such as Norwegian, where a compound like [[ bind-ing]inn] is ill-formed. The relevant patterns can be
accounted for in terms of paradigmatic relations between schemas. For example, the
relation between German herumhopsen and Herumgehopse in (28c) can be expressed by
a paradigmatic relationship symbolized by the sign :::::: between two schemas:
NOM stands for the semantic operator of event nominalization. The structure [Parti [geVre]Nl N in (29) is a right-headed compound structure of the type [XN] N, in which the
head is a deverbal noun with the discontinuous affix ge ... e. What is special about the
structure of these compounds is that its semantic correlate requires. reference to the
semantics of a paradigmatically related structure, that of particle verbs. In (29) we specify that this event nominalization of German particle verbs is formally a head operation:
it is the verbal head that is nominalized, but the nominaliz ing element has semantic
scope over the particle verb as a whole.
In conclusion, the flexibility in encoding lexical units for denoting concepts is greatly
enhanced by allowing for certain types of asymmetries between form and meaning,
mismatches that can be resolved by the language user on the basis of paradigmatic
relationships between (morphologically or syntactically) complex expressions. These observations also show that the internal structure of a lexical unit is not by definition
opaque as an effect of storage: conventional lexical units may preserve their formal and
semantic transparency.
6. Conclusions
The notion "construction" that plays a central role in construction grammar, is indispensable for the analysis of word-formation patterns as well. In the study of word-formation,
PA201
we investigate the systematic correspondences between form and meaning at the word
level. Constructional schemas provide an adequate format for expressing these systematic correspondences. Moreover, they are part of a hierarchical lexicon in which both
complex words and morphological patterns of various levels of abstraction are specified.
An important advantage of the construction morphology approach is that it can express similarities between morphological and phrasal lexical expressions, and the paradigmatic relations between morphological and phrasal schemas. Thus, lexical knowledge
is characterized as a complicated network between words and phrasal expressions at a
range of levels of abstractions, varying between individual words and completely abstract
patterns.
7. References
Allan, Robin, Philip Holmes and Tom Lundskaer-N ielsen
Danish. A Comprehensive Grammar. London/New York: Routledge.
1995
Aronoff, Mark
1976
Booij, Geert
Constructional idioms, morphology, and the Dutch lexicon. Journal of Germanic Lin2002
guistics 14: 301-327.
Booij, Geert
2009
Phrasal names: a constructionist analysis. Word Structure 3: 219-240 .
Booij, Geert
Construction Morph ology. Oxford: Oxford University Press.
2010
Erman, Britt and Beatrice Warren
2000
The idiom principle and the open choice principle. Text 20: 29- 62.
Faarlund, Jan Terje, Svein Lie and Kjell Ivar Vannebo
Norsk Referansegrammatikk. Oslo: Universitetsforlaget.
1992
Goldberg, Adele
2006
Constructions at Work. The Nature of Generalization in Language. Oxford: Oxford
University Press.
Holmes, Philip and Ian Hinchcliffe
Swedish. A Comprehensive Grammar. London/New York: Routledge.
2003
Hiining, Matthias
20 l 0
Adjective + Noun constructions between syntax and word formation in Dutch and German. In: A lexander Onysko and Sascha Michel (eds.), Cognitive Approaches to Word
Formation , 195-218. Berlin: Mouton de Gruyter.
Jackendoff, Ray
Foundations of Language. Oxford: Oxford University Press.
2002
Jackendoff, Ray
2009
Compounding in the parallel architecture and conceptual semantics. In: Rochelle Lieber
and Pavol Stekauer (eds.), The Oxford Handbook of Compounding, 105-129. Oxford:
Oxford University Press.
Jackendoff, Ray
2010
Meaning and the Lexicon. The Parallel Architecture 1975- 2010. Oxford: Oxford University Press.
Jackendoff, Ray
2013
Formal CxG: Constructions in the parallel architecture. In: Thomas Hoffmann and
Graeme Trousdale (eds.), Th e Oxford Handbook of Construction Grammar, 70- 92. Oxford: Oxford University Press .
201
PA203
13. Word-formation in psycholinguistics and neurocognitive research
13. Word-formation in psycholinguistics and

neu rocogn itive research
I.
2.
3.
4.
5.
6.
7.
Introduction
An in-depth look at a seminal study
Word-formation and the fu nctional architecture of the mental lexicon
Morphological parsing and constituent activation
Lingu istic and psycholinguistic isomorphism: assumptions and new a lternatives
Conclusion
References
Abstract
Most words of a language will contain more than one morpheme. Thus, word-formation
is central to the understanding of how words are represented in the mind and brain, how
they are created, and how acts of word-formation, diachronically in the history of the
language and synchronically in acts of language processing, are linked to cognitive
processes and patterns of brain activity. Early studies on morphological processing in
the visual domain revealed that morphological constituents are activated. It is also the
case, however, that effects of the characteristics of whole-word forms have been found.
These have given rise to models that capture the processing of multim01phemic words
as both fully integrated units and as decomposed representations. Recent proposals have
called into question the extent to which traditional theoretical constructs such as morpheme, morphological structure, and mental lexicon actually play a role in lexical representation and processing from a psycholinguistic and neurocognitive perspective. It has
been suggested that we need to recast those constructs in a manner that accords with a
more dynamic view of words in the mind and brain and the role that learning and
experience play in shaping how multimorphemic words are represented.
1. Introduction
In linguistic terms, word-formation can be seen as the means by which the vocabulary
of a language is expanded through the use of existing language elements and patterns to
create new words. Thus, the compound word blueberry is based on the individual lexical
components blue and berry and accords with the general patterns of compounding in
English. It also accords with patterns found in related English compounds such as blackberry and strawberry. Similarly, affixed word pairs such as untwist and twistable and
pairs such as elevator and elevation are formed through the combination of existing
elements in a manner that is consistent with both general language patterns and patterns
associated w ith individual elements.
We might begin the discussion of word-formation in psycholinguistic and neurocognitive research with the term psycholexical to capture the psychological issues associated
with those constructs that we commonly call words. Assuming that the scope of word-
203
PA204
204
formation encompasses derivation and compounding, but does not include inflection, the
psycholinguistic and neurocognitive approach is one that connects those lexical categories to the two fundamental psycholexical questions:
a) How are words represented in the mind/ brain?
b) How are they processed during acts of language production and comprehension?
The psycholexical research issue is thus the extent to which the word-formation status
of a word is required to adequately understand how it is represented in the minds/brains
of language speakers and how it is processed when those speakers engage in acts of
language comprehension and production.
Consider, as an example, a word with a word-fonnation history such as elevator.
There are good reasons to describe this word as multimorphemic. The focus of research
from a psycholexical perspective is thus the extent to which the representation and processing of this word might differ from the mental representation and processing of similarly sounding comparison words without clear morphological components in English
(e.g., elephant or alligator).
The term psycholexical itself is also a potentially useful springboard to framing the
issues. Some readers may be familiar with the term in the context of personality theory
of the late nineteenth and mid twentieth centuries. For many readers, however, it is likely
a new word. In the context of this article, I assume that this word will be much more
useful to the reader than a completely novel form such as, say, pomapel. The reason for
this is that upon first reading, the language comprehender is able to identify and make
use of the major components psycho- and -lexical in order to arrive at an interpretation
of the novel string. How is this done? What computational processes are involved? What
knowledge structures are required? And, if the word remains in the vocabulary of readers, is it encoded in terms of the constituents that were first used to scaffold its comprehension? Are those units (i.e. psycho- and -lexical) stored in the mind? Does the morphological analysis that identified those constituents also result in the decomposition of the
string lexical into the morphological units lex, ic and al?
Finally, is it the case that in the mind/brain, it is reasonable to claim that a form such
as psycholexical has morphological structure at all? Might it be the case that that which
we call morphological structure simply falls out from lexical association, in which psy cholexical might be interpreted by analogy to existing elements, including both the forms
lexical and psychological? These are among the key issues that can be addressed through
psycholinguistic and neurocognitive research.
2. An in-depth look at a seminal study

Speculation on how morphologically complex words are represented in the mind can be
traced back to antiquity (Libben 2010). However, there is no doubt that the paper of Taft
and Forster ( 1975) represented a pivotal moment in the development of the psycholinguistic study of the relation of word-formation to online lexical processing. As I will
discuss below, many of the issues that guided the research agenda in subsequent decades
had their beginnings in the details of the three experiments that Taft and Forster reported
in their initial study. Accordingly, I review those experiments in some depth below.
PA205
The experiments reported by Taft and Forster (1975) employed a visual lexical decision paradigm in which participants are presented with a target string on a computer
screen and must decide as quickly as possible whether that target string is in fact a word
of their language, by pressing either a button labeled "yes" or a button labeled "no".
Response time and response accuracy served as their dependent variables.
There is no doubt about the fact that the lexical decision task (introduced into the
literature by Meyer and Schwanefeldt 1971) has been a cornerstone of research on lexical
processing and our ability to probe the manner in which multimorphemic words are
processed (see Libben and Jarema 2000; Libben, Westbury and Jarema 2012). Indeed
Henderson ( 1982) has claimed that the ascendency of the field of psycholinguistics is
very much linked to the success of the lexical decision task.
Because the lexical decision task requires participants to press either a "yes" or "no"
response button, it necessitates the presence of nonwords in the experiment (typically
the balance of words to non-words is 50 %, to ensure that the experiment does not
produce a response bias). It is most often the case that it is the real words in a lexical
decision experiment that are the primary focus, and lexical decision latencies are interpreted as representing the speed with which words are recognized. In Taft and Forster's
(1975) paper, however, it was the nonwords that were the primary object of study. Specifically, they examined the differences in lexical decision speed and accuracy for words
such as juvenate vs. p ertoire. Critically, both these nonwords are substrings of the words
rejuvenate and repertoire respectively. Both of these words begin with the string (and
possible prefix) re-. Taft and Forster (1975) found that response times to produce a "no"
response for strings such as juvenate were longer than those for strings such as pertoire.
The explanation that they provided for this result set the stage for a great deal of the
reasoning and debate in the field of psycholinguistics on the role of word-formation in
mental representations and lexical processing. The explanation that they presented has
two components. The first is that online lexical processing contains an in-built morphological analysis procedure that is obligatory and non-discriminating. The second is that
the mental lexicon contains not just words, but also morphological substrings such as
juvenate (the poss ible presence of a suffix in the string did not receive special attention
in this early work of Taft and Forster). Thus, in the case of rejuvenate, the potential
prefix re- would be stripped from the string during processing. The stringjuvenate would
then be looked up in the mental lexicon and found to exist. The same procedure would
occur for repertoire, except that in this case, after the potential prefix re- is stripped
from the stimulus and the string pertoire is looked up in the mental lexicon, no representation for p ertoire is found. This, according to Taft and Forster (1975), is the reason why
participants can reject pertoire more quickly than juvenate in a lexical decision task.
There was an unintended way in which the examples of rejuvenate and repertoire
used in the Taft and Forster (1975) study highlights a key challenge in bringing together
the linguistics of word-formation and the psychology of lexical processing. Taft and
Forster were likdy quite right in noting that rejuvenate is understood as containing the
prefix re-. As it turns out, however, etymologically, repertoire also contains the prefix
re-. In this case, it is likely the Latin intensifying prefix re-, attached to the archaic Latin
form parire 'to produce'. Thus, from the very beginning, we were forced to grapple with
the problem of the potential mismatch between the diachrony of a word in the history
of a language and its synchrony in the minds of native speakers in the psycholexical
approach to word-fonnation. From a psycholexical perspective, rejuvenate has morpho-
205
PA206
206
logical structure, because the language user has experience with the formally and semantically related string juvenile.
The case that Taft and Forster ( 1975) presented for obligatory and automatic prefix
stripping in the reading of English words and for the representation in the mental lexicon
of stems that are not free-standing words was strengthened by their use of stimuli such
as dejuvenate vs. depertoire in a subsequent lexical decision experiment reported in the
same study. These stimuli are fonned by taking what Taft and Forster claimed to be
existing elements of the mental lexicon (e.g., juvenate) as compared to representations
that do not exist in the mental lexicon (e.g., pertoire) and attaching the prefix de- to
each, resulting in the creation of nonwords such as dejuvenate and depertoire. They
found that participants took longer to reject the former as compared to the latter. Again,
they took thi s to support the argument that the prefix de- was automatically and obligatorily stripped from the input string. Following that operation, juvenate would be found in
the mental lexicon, creating the response uncertainty that would be manifested in more
errors (i .e. false ''yes" responses) and elevated responses times.
Finally, a third experiment in the seminal Taft and Forster (1975) study set the stage
for another key issue in the study of morphological processing and the mental lexicon.
That is the issue of competition within the mental lexicon. In Experiment 2 of their
report, they focused on strings such as vent. This is a monomorphemic free-standing
word in its own right, but is also part of prefixed words such as invent, prevent, and
circumvent. Taft and Forster compared lexical decision latencies for words such as vent
to latencies for words such as coin, which is only a free-standing word. They also
contrasted vent, for which the bound form is more frequent than the free form, to words
such as card, which also has a bound variant (discard). In the case of card, however,
the frequency of the free form is higher than that of the bound form. The results showed
that "yes" response times for words such as vent were longer than those for matched
free-form-only words such as coin . However, there was no difference between response
times to words such as card and their matched free-only counterparts (e.g., fist). They
concluded that the reason for this is that card has its own representation in the mental
lexicon as a bound stem and that bound stems and free forms can be in competition. In
the lexical decision task, the presentation favours the free form. Therefore, an interference effect is found only for words in which the bound variant is more frequent than
the free one and therefore "gets in the way" .
To summarize, the seminal study of Taft and Forster (1975) created the background
to many of the fundamental psycholexical questions that were to dominate research
programs of the next decades. The study also provides an interesting contrast to the
kinds of issues that have become dominant in just the last few years and which promise
to play a role in the next steps of psycholexical research.
The first issue concerns a shift from viewing the mental lexicon as a static knowledge
store that is composed of discrete items to one that is a dynamic network that changes
through the lifespan. As will be discussed in section 3, there has been a clear development in models of the mental lexicon. It now seems clear that the mental lexicon is
likely not a monolith ic knowledge structure. And, it is very unlikely to have the characteristics of a list. Moreover, it is unlikely that the mental lexicon of a person can be
equated with the lexicon of a language. Generally speaking, vocabulary knowledge
grows in bot h size and richness of association through the lifespan. And, most people in
the world speak more than one language. Therefore, in psycholexica l tenns, the more
PA207
common organization of the mental lexicon across the population is one that can support
multilingual activity.
The second issue concerns the processes of morphological analysis itself. The original
Taft and Forster proposal posited a process of prefix stripping that would enable a multimorphemic word to be accessed through its stem in reading. It is less clear how morphological analysis takes place in the case of suffixed words. And, compound words offer
a particularly chailenging problem because there is no "short list" of affixes that morphological parsing can make use of to identify constituent elements. Moreover, as will be
discussed in section 4, it seems that the morphological processing system is not simply
a parser. It is more likely that morphological parsing falls out from the interaction of
many variables that operate within a system that has as its goal the maximization of
opportunity for mean ing creation. In most cases, by maximizing opportunity for meaning
creation, the psycholexical processing system will produce a parse of an input stimulus
that accords with a morphological description of that stimulus. For some cases, it does
not. And, it is precisely from these cases of morphological mis-analysis in processing
that we can learn a great deal about the characteristics of the psycholexical system.
The third issue, which is discussed in section 5, is the issue of linguistic and psycholinguistic isomorphism. In the original Taft and Forster study, and indeed in very many
studies that have followed, researchers have, as a matter of metaphorical convenience,
described psycholexical representations and processes in a manner that implicitly assumes that the elements (e.g., morphemes) that are used as constructs in linguistic analyses are also constructs that are relevant in the psychological domain. Thus we speak, for
example, of "the activation of morphemes in the mind". However, whether human minds
represent morphological structures, or represent morphemes at all, is an empirical question that has substantial consequences for how we view the relation among the "vocabularies" of linguistic, psycholinguistic, and ultimately neurolinguistic approaches to language.
3. Word-formation and the functional arch itecture

of the mental lexicon
The notion of a mental lexicon is at the heart of our understanding of how language is
represented in the mind. It has also served as a construct that has, perhaps more than
any other, brought together research on formal linguistics, psycholinguistics, neurolinguistics, and aphasiology.
Of course, we must bear in mind that the notion of a mental lexicon is itself a
theoretical construct. There has never been any direct observation of the mental lexicon.
In addition, there is no a priori reason to expect that the neurological instantiations of
an individual's vocabulary store will have a single localized representation in the brain
(Freud 1953 [1891]; Pulvermiiller 201 3; Pulvermiiller, Cappelle and Shtyrov 2013). So,
for the most part, descriptions of the functional architecture of the mental lexicon should
be considered to be metaphors regarding the nature of a theoretical construct.
The metaphor that Taft and Forster ( 1975) used was that of a frequency-ordered list
of entries (Forster 1976). In this view, the list is accessed from beginning to end, and is
ordered in tenns of frequency, with more frequent lexical entries being closer to the top
207
PA208
208
of the list than less frequent ones. It is clear that the particular metaphor of a serial list
was influenced by the ways in which it was common to frame hypotheses concerning
cognitive representations and processing in terms of the data structures that were then
current in the development of computer applications. Nevertheless, a review by Forster
(2007) speaks to the way in which a frequency-ordered mental lexicon provides an
elegant answer to the question of why lower frequency items are slower to recognize
(see, however, Baayen 2010 for an alternative view of the fundamental nature of the
frequency effect).
A rather different approach to constructing a metaphor concerning how words are
represented in the mind has its origins in work by Selfridge (1959) and Selfridge and
Neisser (1960) (see also Grainger, Rai and Dufau 2008). In Selfridge's approach, mental
representations for features can be seen as units which he termed demons, highlighting
their active and independent status. This approach can be seen as a precursor to the
logogen model of Morton (1969) in which logogens are characterized as evidence-collecting units of representation that have specific and independent achvation thresholds
that result from individual experience.
Irrespective of whether representations in the mental lexicon are seen as elements in
a list or as quasi-independent processing units, the question that is most important for a
consideration of the role of word-formation is the question of whether the nature of
lexical units is constrained or affected by the word-formation history of an item. Thus,
we ask whether words such as legal, illegal, and illegality have separate representations
as full forms or whether they are, at some fundamental level, represented in terms of
their constituent morphemes. The hypothesis of a morphologically decomposed representation has, of course, storage efficiencies associated with it, because there are many
fewer morphological elements than there are possible combinations of those elements
into morphologically complex words.
It seems at the present time uncontroversial to claim that the consensus in the field
is that there are mental representations for both multimorphemic words and what we
term morphological constituents. This consensus has been driven largely by data from a
variety of experimental paradigms that have shown that response patterns to the presentation of multimorphemic words is driven both by characteristics of the whole word (e.g.,
its lexical frequency) as well as the lexical characteristics of morphological stems (e.g.,
their lexical frequency). Such findings required the postulation of models of lexical
access and of representation that could accommodate an assertion that multimorphemic
words do have representations in the mental lexicon and that those representations are
structured so that they embody morphological constituency. The supralexical model of
Giraudo and Grainger (2000) provided a framework that claimed that constituent morphemes are accessed through their multimorphemic full word forms, rather than the other
way around, as suggested by Taft and Forster (1975).
Another issue that arose following the early work of Taft and Forster (1975) is whether it is appropriate to assume that representations in the mental lexicon are likely to be
structurally homogenous. Might it be the case, for example, that some prefixed words
are processed through their constituent morphemes, while others are processed and represented as full forms? Such questions were accommodated in a family of dual-route
models that posited competing decomposition routes to the recognition of multimorphemic words and whole-word processing routes. These processing alternatives were
visualized as pathways in a flowchart. Such an approach was represented by the aug-
PA209
mented addressed morphology (AAM) model (e.g., Burani and Caramazza 1987; Burani
and Laudanna 1992). A related approach claimed that the processing alternatives are
carried out by routes that operate in parallel (e.g., Bertram, Schreuder and Baayen 2000;
Schreuder and Baayen 1995). In this type of model, the online processing of multimorphemic words was seen as a type of horserace. In this horserace, one route conducts a
morphological analysis of a multimorphemic word. The other route processes it as a
whole. The process that reaches the finish line first is the one that is employed.
The early models of the mental lexicon that have been outlined above have their
origins in the initial conceptualization of the mental lexicon as a memory store with a
relatively well-defined architecture and with relatively well-defined units of representation. The issue of whether or not this is a good starting point for modeling the mental
lexicon has been central to more recent discussions. Researchers such as Elman (2011)
have argued that the notion of a mental lexicon is not required to explain lexical processing phenomena. He proposes that we see words not as elements of a mental lexicon, but
rather as stimuli that alter mental states in lawful ways. This proposal has as its core the
dynamic nature of processing and the fact that lexical processing in a natural setting is
contextualized in sentences, discourses and situations.
Another recent perspective on the nature of morphemes and lexical representations
has been advanced in the context of naive di scrimination learning (NDL) (Baayen et al.
20 11 ; Baayen, Hendrix and Ramscar 2013). This model is based on the Recorla-Wagner
learning equations (Wagner and Recorla 1972) and has the notions of discrimination and
learning at its core. As has been demonstrated by Baayen, Hendrix and Ramscar (2013),
the NDL model can account for lexical frequency effects without recourse to explicit
morphological representations for prefixed, suffixed or compound words.
As the above new approaches to lexical processing demonstrate, there is considerable
challenge to the notion of a mental lexicon as an inert knowledge store. We can expect
that the mental lexicon will increasingly be characterized as a dynamic cognitive system
that allows for lexical activity. And, a key component of this activity is the lexical
creativity associated with understanding new affixed and compound forms and producing
them in a manner that will be understood by others.
If the ability to produce and understand novel multimorphemic words resides in the
domain that we call the mental lexicon, it is necessary that we consider that domain to
be more active and more flexible than we have in the past. To exemplify this activeness
and flexibility, let us return, for example, to the word repertoire that Taft and Forster
( 1975) used as a monomorphemic contrast to the multimorphemic form rejuvenate. The
word repertoire, is very likely monomorphemic in the minds of most native speakers of
English. But now, what happens psycholexically in the following scenario: Imagine that
a concert pianist is asked to play a familiar piece of music. However, she declines the
request, saying: "I don't play that anymore. It used to be part of my repertoire, but now
I would say that it is part of my ex-pertoire". It is quite likely that the coinage will be
easily understood in that context as a piece of music no longer played. But now, imagine
a slightly different scenario: In this version, a concert pianist is also asked to play a
familiar piece of music. But now she says: " I have never played that piece well enough.
I have tried, but I guess it will never be part of my expert-oire." Here, it seems that the
ability to interpret the novel word as referring to "part of an individual's repertoire that
is played particularly well" is scaffolded by the existence of the lexical items repertoire,
expert, and expertise.
209
PA210
210
Both the example cases above are intended to demonstrate that the word repertoire
is not morphologically inert. Perhaps it is the case that no representation in the mental
lexicon is morphologically inert. Rather, psycholexically, as soon as there are multiple
representations in an individual's vocabulary, and those representations are linked, new
word-formation possibilities are built as expressions of human creative ability. Thus, it
is appropriate to see the mental lexicon not simply as the repository of lexical elements
(i.e. morphemes) and not simply as a repository of the results of word-formation (i.e.
whole multimorphemic words), although it most likely contains representational information for both. The key feature of the mental lexicon is its dynamic nature and the
manner in which it can connect lexical knowledge in order to make word-fonnation
possible and in order to make the comprehension of the results of word -formation possible. Moreover, because the mental lexicon of an individual will very likely encompass
lexical knowledge of more than one language, adequate models need to be able to account for the ways in which linguistic variation in morphological patterns and constraints
can be accommodated.
4. Morph ologica l pa rsing and co nstituent activation

The discussion above of the dynamic nature of the mental lexicon and the need to have
an architecture that can encompass lexical creativity and the results of word-formation
brings to the foreground the issue of how morphological decomposition is actually
achieved. That is, how do listeners or readers actually find the constituent morphemes
in existing and novel multimorphemic words? This is an issue that is. weighted toward
the comprehension side of morphological processing (although there are significant computational issues to be considered in production as well).
As was the case for conceptions of the mental lexicon, we can trace research on
morphological parsing and activation back to the very early work by Taft and Forster in
the mid 1970s. As has been discussed above, Taft and Forster (1975) provided an explicit
proposal for how prefixed words of English are parsed in order for them to be looked
up in the mental lexicon in terms of their stems. In a publication in the following year,
Taft and Forster (1976) addressed the matter of how compound words are processed.
Their proposal was that, during single-word reading, compound words are looked up in
the mental lexicon in terms of their first constituents. They assumed a left-to-right parse
of compound words and further assumed that the prefix stripping approach that was
employed for prefixed words could not work for compound words, because there is no
"short list" of compound initial constituents that would be analogous to the relatively
"short list" of prefixes that exist in English.
Taft and Forster's (1976) evidence for the claim that compound words were accessed
in terms of their initial constituents came from lexical decision "no" latencies to nonwords (as was the case in their 1975 study). They contrasted stimuli such as footmilge
as compared to trowbreak. The crucial difference here is that in the case of footmilge,
the initial constituent is a real word and the final constituent is a nonword. In the case
of trowbreak, the nonword is in initial position. Taft and Forster found that it took
participants longer to reject footmilge and suggested that compounds are processed from
left to right in English in a manner that allowed lookup in terms of the initial constituent.
PA211
Subsequent work by Libben (1994) proposed a compound parsing algorithm in which
compounds are scanned in a recursive beginning-to-end manner. Libben, Derwing and
de Almeida (1999) extended this work to consider "challenging" cases such as the ambiguous English novel compounds cartrifle (which can be parsed as either car-trifle or
cart-rifle). Evidence from words such as these suggests that morphological parsing does
not yield a single univocal parse. Rather, all constituents that can be activated will be
activated. Thus, Libben, Derwing and de Almeida (1999) found that a stimulus such as
cartrifle generated activation of car, cart, rifle, and trifle. Additionally, participants
showed evidence of mis-parses such as cart-trifle.
The theme of morphological mis-parsing and mis-activation has been a dominant in
the morphological processing literature, generating a string of papers w ith the general
title "Is there an X in X?". Sample papers in this genre include "Is there a hat in that"
(Bowers, Davis and Hanley 2005), "The broth in my brother's brothel" (Rastle, Davis
and New 2004), and a set of studies that investigate whether there is corn in corner
(e.g., Longtin, Segui and Halle 2003; Morris, Grainger and Holcomb 2008; Lehtonen,
Monahan and Poeppel 2011; Lavric, Elchlepp and Rastle 2012).
There is still considerable controversy in the field regarding the means by which
recognition of a word corner w ill also activate the representation for corn, and how such
effects can be accommodated in models of lexical representation and processing. If we
accept, however, that such mis-activations do occur, this presents important consequences for the linking of theories of word-formation to psycholexical theories of morphological processing. T he first is that the term morphological processing itself is perhaps
inappropriate. The reason for this is that the processing of corner into corn +er is not a
true morphological analysis. Extracting the lexical representation for broth out of brothel
is certainly not, and extracting hat from that seems to be rather maladaptive from a
morphological point of view. Yet, from a psycholexical perspective, all of these may be
fully adaptive in terms of the function of the system as a whole. It is impotiant to bear
in mind that if there is a goal in the processing of words, that goal is to create meaning,
not just to perform a linguistic or morphological analysis of an input string. Under this
view, the overarching characteristic of the human lexical processing system is that it
seeks to maximize the opportunity for meaning creation (Libben 2010, 2014). In this
way, we might imagine that if multimorphemic words are indeed analyzed morphologically, it is likely not in order to facilitate their lookup in the mental lexicon under this
or that access key. Rather, it is so that activation can occur under any and all access
keys. It may be the case that this, the overlapping access to as many morphological subelements as possible, is the core characteristic of morphological processing. The means
by which this can be achieved will differ from language to language, depending on
morphological patterns within the language and, in the case of reading, the conventions
that govern its expression in writing and print. But despite the potential homogenizing
effects of those conventions, we can expect that morphological processing will be relatively unconstrained, resulting in many "false positive" activations, such as those that
enable corn to be activated in corner.
Finally, it should be noted that this type of over-activation moves us yet further away
from a view of the mental lexicon as an inert list. It seems that, in the case of multimorphemic words, it is likely to be a dynamic place which is characterized by considerable
conflict. The activation of corn interferes with, rather than accords with, the interpretation of corner. The activation of all the potential constituent morphemes in the ambigu-
211
PA212
212
ous form cartrifle will, at some point in the interpretation of the ambiguous word, require
that some potential constituents be deactivated. And finally, for words that we can characterize as semantically opaque (e.g., honeymoon, humbug), the maximization of opportunity for meaning creation will result in over-activation and the need to resolve the
cognitive conflict that may result.
5. Linguistic and psycholinguistic isomorphism: assumptions and

new alternatives
As is often the case in the history of psycholinguistics, the foundational research in
the field of morphological processing demonstrated the extent to which categories and
constructs that play a role in theoretical linguistics also play a role in online processing.
Such investigations may be considered as coming under the general heading of "psychological reality studies". Derwing (1973) argued that in each such case, the adequacy of
a linguistic construct to the psychological domain must be tested against psycholinguistic
data. Authors such as Myers (2007, forthc.) have outlined both the opportunities and
challenges associated with employing the terminology of linguistics in the psycholinguistic domain.
As the discussions in the sections above have underlined, psycholinguistic and neurocognitive research on the representation and processing of multimorphemic words has
been moving to a much more dynamic conceptualization of the mental lexicon and a
much less categorical conceptualization of morphological processing. Thus, if there has
been an assumed isomorphism between linguistic and psycholinguistic constructs in the
past, this is likely to come under a great deal more scrutiny in the future. Indeed indications at present are that the fundamental notions of a morpheme, a word, and multimorphemic structure are likely to be re-thought.
An anticipation of this re-thinking might prompt us to consider the fundamental question of why that which we call morphological constituency would be activated or built
up during online lexical processing at all. One approach to this question is to suggest
that morphemes are activated because words have morphological structure. However, if
we adopt a psycholexical approach, that simply becomes a tautological statement. From
a psycholexical perspective, words do not have any properties that are not fundamentally
psychological in nature.
Another approach, one that requires fewer assumptions and that is situated in the
psycholexical domain, is that, because our minds are organized to maximize the opportunity for meaning creation, we cannot but recognize subunits of multimorphemic strings
when they correspond to elements for which we already have representations (i.e. those
with which we already have experience). So in the case of elevator, there is activation
of elevate. In the case of illegal there is activation of legal. When there is less transparency, for example in the case of illegality, in which there is associated stress change, we
might assume that activation of legal is made somewhat more difficult. It is important
to note here that using the term "experience" may have advantages. Consider a compound example such as blueberry. It seems that there are two kinds of experiences that
are relevant here. One is experience with the free form beny. The other is experience
with the compound head -berry. That latter experience is the one that is more relevant
PA213
to the processing of blueberry. But at the same time, it is the principle of maximization
of opportunity that also gives rise to the activation of berry as a free-form . So, depending
on the weighting of those competing experiences, i.e. their relative frequency, we will
see interference effects rather than facilitation effects between the experience that a
language user has with the free form berry and that user's experience with the compound
head -berry, which we can assume to have psycholexical status as a positionally defined
bound item.
The maximization of opportunity places us in a situation where strict rules of morphological constituency cannot be automatically related to psycholinguistic phenomena. At
the outset of this article, multimorphemic forms such as elevator were contrasted with
monomorphemic fonns such as alligator that are similar sounding and of comparable
length. But as we have noted with the form pertoire above, morphological structure,
from a psychocentric perspective, is a matter of experience. And, that experience grows
and changes throughout the lifespan. Consider how the clipping gator is used, both as a
shortening for alligator and as a compound head in the famous drink Gatorade. This
experience could easily give rise to a novel compound alloygator, drawing by analogy
on alligator, but using -gator as a compound head. Indeed, a Google search will reveal
that such a novel compound has been used commercially as the name for an automobile
alloy wheel protector. The novel compound builds on the whole-word form alligator
(the picture of the product shows an alligator biting the rim of a wheel). It may also be
the case that the product name is trading on the activation of gate (with the connotation
of protection). And, if so, this would be entirely consistent with the notion of maximization of opportunity and a psycholexical conceptualization of morphological activation.
It would be less consistent with the view of morphological processing that is focused on
the formal morphological analysis of the string.
The kind of lexical processing and activation that we have claimed may occur in the
comprehension of the novel compound alloygator accords with the observation by Singh
(2006) that it is much more common for compounds to be formed by switching out
individual elements (creating compound families) than by bringing together two free
morphemes in an entirely novel construction.
The reason for this from a psycholexical perspective is that it is easier to draw on
positional experience. And, it is likely that when such positional experience is activated,
it is not only the form, but many other experiences associated with that form that are
also activated. These would include the semantic relations that have been examined by
Gagne and Spalding (2007, 2009).
An additional important component in the re-thinking of complex and compound
words in the mind/brain involves addressing the question of whether lexican representations themselves can be said to be fixed. Libben (2014) presents a framework in which
they are not. Rather, it is suggested within that framework that those representations that
we typically refer to as morphemes and which take part in complex and compound
representations, develop as a result of the lexical experience of the language user, into
what Libben (2014) calls morphological superstates. The construct of a morphological
superstate draws metaphorically on the notion of superposition in quantum physics. As
applied to morphological processing, the notion suggests that the experience that a language user has with a lexical form (e.g., the word board) creates new morphological
possibilities. So, if those experiences include the use of the form board as a compound
modifier (e.g., boardroom) and a compound head (e.g., keyboard), these develop as new
213
PA214
214
positionally bound morphological elements in the mental lexicon. It is claimed that the
new morphological representation for board can be seen to generate a proliferation of
morphological elements in the mind/brain. This proliferation may be less efficient from
the perspective of storage efficiency in the mental lexicon, but it is much more effective
from the perspective of maximizing the opportunity for meaning creation. It is claimed
that these alternative uses of a form such as board are potentially in competition in the
mental lexicon. One of the ways in which this potential conflict is reduced is for morphological forms with different morphological and semantic functions to be represented in
a superstate form which has specific morphological manifestations (e.g., free lexeme,
compound modifier, compound head) depending on specific usage needs. These manifestations may come into conflict in specific situations. One of those situations would be a
constituent priming paradigm in which a word such as board is followed immediately
by a compound such as boardroom or keyboard that contains the same string, but with
a different morphological function.
6. Conclusion
In this article we have traced the role of word-formation in psycholinguistic and neurocognitive research from its beginnings in the work of Taft and Forster ( 1975, 1976) to
investigations of the electrophysiological correlates of affixation and p seudoaffixation in
Lavric, Elchlepp and Rastle (2012) and, finally, to models that call into question whether
morphemes are useful constructs at all (e.g., Baayen, Hendrix and Ramscar 2013) and
models that challenge the notion of morpheme as a univocal mental structure (Libben
20 14).
It has been almost 40 years since the seminal work of Taft and Forster. Over that
period, there have been very substantial advances in methodology in the field, and a
great deal more is known about how multimorphemic words are processed across lan guages, across specific word forms within languages, and among individuals. Yet there
remains a great deal to be accomplished. In the next years, a key domain of exchange
and advancement will be to find the transdisciplinary language that brings together insights from the study of morphology, the psycholexical study of morphological processing, and the neurocognitive study of brain correlates in the production and comprehension of multimorphemic words. Bringing these domains together will enable the integration of the notion of word-formation as something that takes place in a language and
also something that takes place in a mind and between minds. Indeed, the question could
be framed in the following manner: " If a new word is formed in the forest and there is
nobody there to hear it, is it really a new multimorphemic word?"
This question may contain within it the challenges for the next decade of psycholinguistic and neurocognitive research on word-formation. In some sense, the answer must
be ''yes, a new word is formed". But, on the other hand, the simple enactment of the
word-formation ability of a single language user seems to fail to satisfy the notion of
what it means for something to be "a new word". A new multimorphemic word is
associated with patterns of interpretation in a linguistic community. It is linked to other
representations in the mental lexicons of language users, and it instantiates the links that
must exist across language users in order for communication to take place. These are
PA215
very likely to be aimong the next psycholexical research issues concerning word-formation
and the central role that it plays in language production, comprehension and creativity.
7. References
Baayen, R. Harald
20 IO
Demythologizing the word frequency effect: A discriminative learning perspective. The
Mental Lexicon 5: 436-461.
Baayen, R. Harald, Peter Hendrix and Michael Ramscar
2013
Sidestepping the combinatorial explosion: Towards a processing model based on discriminative learning. Language and Speech 56: 329- 347.
Baayen, R. Harald, Petar Milin, Dusica Filipovic Durdevic, Peter Hendrix and Marco Marelli
2011
An amorphous model for morphological processing in visual comprehension based on
naive discriminative learning. Psychological Review 118: 438- 482.
Bertram, Raymond, Robert Schreuder and R. Harald Baayen
2000
The balance of storage and computation in morphological processing: The role of word
formation type, affixal homophony, and productivity. Journal of Experimental Psychology: Learning, Memory & Cognition 26: 489- 511.
Bowers, Jeffrey, Colin Davis and Derek A. Hanley
2005
Automatic semantic activation of embedded words: Is there a 'hat' in 'that'? Journal of
Memory and Language 52: 131- 143.
Burani, C ristina and Alfonso Caramazza
1987
Representation and processing of derived words. Language and Cognitive Processes 2:
217-227.
Burani, Cristina and Alessandro Laudanna
Units of representation for derived words in the lexicon. In: Ram Frost and Leonard
1992
Katz (eds.), Orthography, Phonology, Morphology and Meaning, 361- 376. Amsterdam:
North-Holland.
Derwing, Brnce
1973
Transformational Grammar as a The01y ofLanguage Acquisition. A study in the empirical, conceptual and methodological foundations of contempora1y linguistics. Cambridge: Cambridge University Press.
Elman, Jeffrey L.
2011
Lexical knowledge without a mental lex icon? Th e Mental Lexicon 60: 1- 33.
[doi: 10. l 075/ml.6.1.01 elm].
Forster, Kenneth
1976
Accessing the mental lexicon. In: Roger J. Wales and Edward Walker (eds.), New Approaches to Language Mechanisms, 257-287. Amsterdam: North-Holland.
Forster, Kenneth
2007
Visual word recognition: problems and issues. In: Gonia Jarema and Gary Libben (eds.),
Core Perspectives on the Mental Lexicon, 56- 78 Oxford: Elsevier.
Freud, Sigmund
1953 [1891]
On Aphasia. A critical study. New York: International University Press.
Gagne, Christine and Thomas Spalding
2007
The availability of noun properties during the interpretation of novel noun phrases. The
Mental Lexicon 2: 239- 258.
Gagne, Christine and Thomas Spalding
2009
Constituent integration during the processing of compound words: Does it involve the
use of relational structures? Journal of Memory and Language 60: 20-35.
215
PA217
Pulvermiiller, Friedemann
2013
How neurons make meaning: Brain mechanisms for embodied and abstract-symbolic
semantics. Trends in Cognitive Sciences 17(9) : 458- 470. [DOI: 10.10 16/j .tics.2013.
06.004)
Pulvermiiller, Friedemann, Bert Cappelle and Yury Shtyrov
2013
Brain basis of words, constructions, and grammar. In: Thomas Hoffmann and Graeme
Trousdale (eds.), The Oxford Handbook of Construction Grammar, 397-416. Oxford:
Rastle, Kathleen, Mathew H. Davis and Boris New
2004
The broth in my brother's brothel: Morpho-orthographic segmentation in v isual word
recognition. Psychonomic Bulletin and Review 11: 1090- 1098.
Schreuder, Robert and R. Harald Baayen
1995
Modeling morphological processing. In: Laurie B. Feldman (ed.), Morphological Aspects of Language Processing, 131-154. Hillsdale, NJ: Erlbaum.
Selfridge, Oliver
1959
Pandemonium: A paradigm for learning. In: D. V. Blake and A. M. Uttley (eds), Proceedings of the Symposium on Mechanisation of Thought Processes, 511-529. Her Majesty's Stationary Office.
Selfridge, Oliver and Ulric Neisser
Pattern recognition by machine. Scientific American 203: 60- 68.
1960
Singh, Rajendra
2006
Whole Word Morphology. In: Keith Brown (ed.), Encyclopedia of Language and Linguistics. Vol. 13, 578- 579. 2nd ed. Oxford: Elsevier.
Taft, Markus and Kenneth Forster
1975
Lexical storage and the retrieval of prefixed words. Journal of Verbal L earning and
Verbal Behavior 14: 630-647.
Taft, Markus and Kenneth Forster
1976
Lexical storage and retrieval of polymorphemic and polysyllabic words. Journal of Verbal Learning and Verbal Behavior 15: 607- 620.
Wagner, Allen and Robert A. Rescorla
1972
A theory of Pavlovian conditioning: Variations in the effectiveness of reinforcement
and nonreinforcement. In: Abraham H. Black and William F. Prokasy (eds.), Classical
Conditioning 11, 64-99. New York: Appleton-Century-Crofts.
Gary Libben, St. Catharines, ON (Canada)
217
PA218
11. Units and processes in word-formation I:

General aspects
14. The delimitation of derivation and inflection
l.
2.
3.
4.
5.
Introduction
Reasons for fuzziness
Criteria for the distinction between derivation and inflection
Conclusions
References
Abstract
This article establishes that the relation between inflectional morphology and derivational morphology is that of cline. The main reasons for this situation are briefly outlined.
The fuzzy nature of the boundary between the two areas of morphology is illustrated by
means of selected criteria for their delimitation that are each confronted with counterexamples taken from various languages of the world.
1. Introduction
Recent decades have witnessed a profound change in evaluating the relationship between
linguistic categories and disciplines. It was cognitive linguistics, in particular, that highlighted their fuzzy nature. The attention of linguists was thus shifted to the problems of
vague boundaries, prototypes and peripheries, good exemplars and poor exemplars. The
situation within the traditional discipline of morphology is no exception. A number of
morphologists, making use of the invaluable contribution of cross-linguistic typological
research, have come to the conclusion that "the inflection/derivation distinction is not
absolute but allows for gradience and fuzzy boundaries [ .. . ] we are deaEing with a continuum from clear inflection to clear derivation with ambiguous cases in between" (Haspelmath 1996: 4 7).
What are the reasons for this situation w ith in morphology? A brief summary of these
reasons is outlined in the first part of the article, fo llowed by an overview of criteria
proposed in the literature for the distinction between derivation and inflection. It will be
demonstrated that none of these criteria are of universal validity due to numerous counter-examples from the languages of the world.
2.
Reasons for fuzziness
2.1. The concept of new word

While it is generally accepted that derivational morphology deals with the formation of
new words, i.e. with giving names to yet unnamed objects, it is not quite as clear what
PA219
a new word actually is. Is a clipping or an acronym a new word? They do not denote a
new object; instead they denote an " old" one by means of a reduced form. Even more
evident is the fuzziness in numerous quantity-related categories. Thus, for example, do
diminutives and augmentatives refer to new "objects", like Slovak beikat' in (1 )?
(1)
beikat'
' to run+oIM'
+---
behat'
'to run '
If so, then the comparative and the superlative degrees of adjectives should also be
treated as derivational forms, although they are mostly discussed within inflectional
morphology. And what about various aktionsart-categories, like frequentatives, duratives, intensifications, as in the following example from Wichi (Veronica Nercesian, pers.
comm.),
(2)
n '-yahin-pej
lsuB-look-ffER
'I look many times'
that are usually considered to be of inflectional nature? While Beard (1982) gives arguments in favour of the derivational status of the category of Plural in general, and Robins
( 1971) argues that number in Japanese is derivational, the plural of verbs has been
perceived as a prototypical inflectional category. But there are also pluractional verbs
like those in the Chadic language Hausa (3) and the Austronesian language Karao (4):
(3)
tuntimii
REo-tunaa
PL.ITER-remind
'to remind many or often'
(Newman, pers. comm.)
(4)
manbabakal
man-cv- bakal
REClP. lMPFT-PL-fight
'to fight each other' (indicates that more than 2 participants are involved in the action)
(Brainard, pers. comm.)
When discussing the situation in Kwakw'ala Anderson (1985: 30) argues in favour of
the derivational status of affixes expressing tense, aspect, voice, modality and plurality,
because they are (a) optional, and present only where necessary for emphasis. or disambiguation; and (b) equally applicable to words of any syntactic function or word class.
Anderson shows that, for example, x wakwana 'canoe' has a future form x wakwanaX 'canoe
that will be, that will come into existence'. Its recent past form is x wakwanaxdi 'canoe
that has been destroyed '. In addition, Stekauer, Valera and Kortvelyessy (2012) point
out that, compared to the "simple, inflectional" plural meaning 'more than one', pluractional verbs have, in principle, an additional meaning, like 'repeated action', 'extent/
degree of action', 'simultaneity of action', etc.
219
PA221
(7)
tsa:lapa:La:ni6:tsa
flee-RPT-PF=now
'he had fled again'
2.5. Systematic ambiguity

While in most languages the inflection/derivation ambiguities are restricted to individual
cases/categories, the situation in Ket and all the other Jenissej languages is much more
complex. Werner ( 1998) demonstrates that the whole system of verbs in these languages
is fuzzy in this respect. Belimov ( 1991: 25) even assumes that word-formation and
inflection are in fact a single process of forming new lexical units. Werner refers to
verbs in Ket, derived by infixation, which have complete inflectional paradigms but lack
an infinitive as a basic citation form:
(8)
ti<j >vij 'southerly wind blows'

ti<j-l->vij "southerly w ind blew'
ur<a>vij ' northerly wind blows'
ur<;;-n -> vij 'northerly w ind blew'
(Werner 1998: 135)
Interestingly, the potential infinitives are identical to nominal compounds given in (9).
(9)
tivej 'southerly wind'

udbej 'northerly wind '
ss ,bej 'upward-rising smoke'
(Werner 1998: 13 5)
2.6. Serving two masters

Some affixes serve two masters. The following example from Karao (Brainard, pers.
comm.) shows that polysemy/homonymy can cross the inflection/derivation boundary
(also because, on a more general level, the boundary between polysemy and homonymy
is rather fuzzy). The Karao affix pan-/impan- can be a nominalizing affix, it can express,
e .g., iterative action on a verb (10) but it can also signal imperative mood (!) as exemplified in (1 1):
(10) p an-chinel-an
mo-ak
ITER-depend.on-IPFV.PAT ERG.2SG-ABS. l SG
'You can always depend on me '
(11)
pan-dotho ka!
IMPER-cook ABS.2SG
'Cook! '
221
PA223
3.2. Range of categories

Inflectional morphology typically involves smaller meaning changes than derivational
morphology.
This meaning-related criterion is closely related to the distribution of derivational and
inflectional morphemes and their respective functions. Since the basic function of derivation is the formation of names for new "objects" it is "natural" that a derivational morpheme is positioned next to the particular lexical morpheme. Since the basic function of
inflection is to express grammatical relations between words it is natural that inflectional
morphemes are placed at the periphery immediately before/after the neighbouring word.
Thus, prototypically, there is an opposition between lexical (conceptual) meaning and
grammatical meaning.
But there are at least two points that need to be taken into account. First, Beard (1981,
1995) postulates 44 fundamental Indo-European functions which underlie both derivation
and inflection and which are realized in various languages differently (or, are not realized
at all ). If this hypothesis is true, the semantic distinctions between inflection and derivation, at least, historically, become obliterated. This hypothesis may be supported by
instances where a category which is commonly used by derivation is, in some languages,
integrated in the inflectional system, and vice versa.
The second aspect of the problem concerns Bybee 's assumption that the derivational/
inflectional status of morphological items correlates with their semantic relevance for
the root and generality of their meaning: "[a] meaning element is relevant to another
meaning element if the semantic content of the first directly affects or modifies the
semantic content of the second" (1985: 13). In addition, Bybee assumes that "the greater
the difference between the meaning of the derived word and the meaning of the base,
the greater the likelihood that the affix is derivational" (1985: 5). Bybee's approach
necessarily leads to a semantic cline, with various degrees of semantic change and semantic relevance.
3.3. Transpa ren cy

Inflectional morphology is usually semantically and formally more transparent than derivational morphology.
Laca's claim (2001: 12 18) that irregularity in derivational morphology manifests itself
in the many-to-one and one-to-many relationships between form and meaning appears
to be too strong. These types of relationships seem to be equally typical of inflectional
morphology, at least in fusional languages which are characterized by numerous cases
of cumulative exponence and syncretism. As an example, each inflectional morpheme in
each of the Slovak paradigms is a portmanteau morpheme expressing three fundamental
meanings: case, number, and gender. Extensive syncretism in Slavic paradigms may be
illustrated with the inflectional morpheme -om in chlap-om, which can mean either singular, instrumental, masculine noun of the pattern ch lap 'man', or plural, dative, masculine noun of the same pattern. This syncretism runs across paradigms, and so -om expresses the same grammatical meani ngs in the (different) paradigm dub ' oak'.
223
PA224
224
II. Units and processes in word-formation I: General aspects

Ricca (2005: 198- 211) refers to portmanteaux morphemes which combine derivational and inflectional categories, this being an important argument against strict separation
of inflectional and derivational morphology. For illustration, French denominal adjectives in -[al] as in national ' national' form masculine plural by -[o] (nationaux). The
suffix not only derives denominal adjectives but also combines the inflectional categories
of masculine and plural (Ricca 2005: 207). Zwarts (2007, pers. comm.) shows that in
Endo, a Southern Nilotic language, the derivation process based on changing a vowe l
from [- ATR] (i.e. non-advanced (retracted) tongue root) to [+ATR] (advanced (retracted)
tongue root) results in agent nouns in plural:
(13) kwaang 'to cook'
---+
kwaang 'cooks'
Interestingly, its singular is derived through the singulative suffix -fin, to give kwaangii'n.
Derzhanski (2005) provides add itional cross-linguistic evidence of the derivation+inflection type of cumulative exponence, in particular, by giving examples of the
plural+diminutive cumulation in languages of various families (e.g., Bulgarian (Slavic),
Fu la (Atlantic-Congo), Swahili (Bantu), Asmat (Trans-New Guinea)).
On the other hand, Plank (1994: 1675) aptly relates the lack of transparency in some
derived words to the process of lexicalization to which inflected forms are usually not
exposed. But examples of a semantic shift in inflected forms do occur, like air-s ' unnatural manner or action intended to impress'; damage-s 'money claimed from a person for
causing damage '; work-s ' moving parts of a machine '. And finally, also derivation may
be highly transparent as is the case with, for example, derivation of adverbs from adjectives in English.
3.4. Productivity and regularity

While rules of inflectional morphology are typically highly productive, the productivity
of derivational morphology is restricted.
While inflection is regular, derivation is fraught with idiosyncrasies.
These two criteria reflect the deep-rooted belief of a number of morphologists and syntacticians, at least since Chomsky's Remarks on Nominalization (1970), that the productivity and regularity of word-formation processes are much lower than that of syntactic
and inflectional processes. However, this has been called into question in recent decades
(Anderson 1982; Strauss 1982; Di Sciullo and Williams 1987; Stekauer 1998; Plag
1999). Plag (1999: 2) emphasizes that " derivational processes are much more regular
than previously conceived", and Anderson (1992: 78) notes that "even a completely
productive process can still, arguably, be derivational, such as the formation of English
nominals in -ing from verbs". On the other hand, some inflectional paradigms are incomplete, defective (cf. singularia tantum, pluralia tantum, mass nouns, proper names; nongradable adj ectives, etc.).
The bias against productivity and regularity in word-formation seems, however, to
rest on unequal criteria.
While the productivity of inflection is usually assessed on the basis of categorial
meanings expressed by the categories of gender, plural number, case, tense, etc., the lack
PA225
of productivity in derivation is usually argued for on the basis of purely formal criteria
(cf. Bauer 2000: 37). Stekauer (1998: 83- 85) shows that it is not important in what
way (by suffixation, compounding, conversion, etc.), for example, AGENTIVE meaning is
expressed. What matters is that, if required, such an AGENTIVE name can be produced
for any AGENTIVE concept. Furthermore, while it is true that the suffix -ion, for example,
does not combine with all verbs, it is also true that not all verbs can be used in the
sentence structure noun - verb - object. The constraint permits only transitive verbs to
be inserted in the relevant slot. Both restrictions (syntactic and derivational) are based
on the same principle of combinability of structural units. And, to give one more example
from the field of inflection, the productivity of the Slovak plural suffix -ovia (as in
hrdinovia 'heroes') - as one of many plural affixes in Slovak - is restricted to one of
12 different substantival paradigms, notably plural of masculine animate nouns ending
in a vowel. From this point of view, however, the assumption of "full productivity" of
inflection vs. significantly restricted productivity of derivational processes does not seem
to be so persuasive.
3.5. Word class

While prototypical inflectional morphology does not change word class, derivational
morphology often does.
Obviously, this criterion is not applicable to cases in which affixation does not change
word class. As counterevidence, our examples of transflexion in section 3.1 above show
the class-changing function of inflection in some languages. In addition, Stump (2005:
53-54) shows that an inflectional category can affect the word class. He refers to cases
like present participles (i.e. inflected forms of verbs), for example discouraging, which
can also function as attributive adjectives, as in the phrase the most discouraging news.
Whi le one might hypothesize a separate word-formation process, the non-derivational
nature of the attributive use of the participle can, in Stump's view, be supported by
evidence from many other languages, in which present participles never appear without
declensional morphology which is unmistakably that of an adjective.
Furthermore, Booij (1994) and van Marie (1995) argue that the so-called inherent
inflection (numerous examples of plural nouns in various languages; infinitives, participles and comparatives in Dutch) can feed word-formation, and that inherent inflection
"expresses like derivation, a certain amount of independent information, whereas the
information expressed by contextual inflection is redundant, and only reflects certain
aspects of the syntactic structure of the sentence" (Booij 1994: 30).
Van Marie's examples from Dutch show that "derivational forms may develop inflection-like properties [ ... ] and that inflectional forms may display derivation-like properties" (Van Marie 1995: 78) as exemplified in (14) and (15):
(14) ([X]quanl
_]NP>
where the slot can be filled by adjectives compulsorily taking -s:
225
PA226
226

(15) iets groen-s
niets waar-s
een heleboel slecht-s
'something green'
'nothing true'
'a lot of bad things'
The -s-suffixation is automatic, and the resulting forms do not occur as independent
words: they can only occur as a part of a construction with the above-given structure.
These are typical inflectional characteristics. The point is, however, that the suffixed
forms are noun-like in contrast to their adjectival bases: (a) at least in some cases, nouns
(without -s) may fill the empty slot in (14), the -s forms are sometimes used as nouns,
and some of them have become lexicalized as nouns (lekkers 'sweets', nieuws 'news').
This suggests that "inflectional forms may develop properties which are usually primarily associated with derivation: the ability to change category. The -s forms indicate, then,
that category changing operations can be found in the realm of inflection as well [ ... ]"
(Van Marle 1995: 74).
Last but not least, Haspelmath ( 1996) gives ample evidence that word-class changing
inflection (for example, German participles, Lezgian verbal nouns) exists in numerous
languages, which authorizes him to conclude that "the myth that word-class-changing
inflection does not exist is no more than a myth [ ... ]" (Haspelmath 1996: 50).
3.6. Recursiveness
While rules ofderivational morphology can be reapplied, rules of inflectional morphology are not recursive.
While the order of derivational affixes reflects the order of semantic operations, the
order of inflectional affixes is fixed and/or semantically irrelevant: only one order is
available.
As pointed out by ten Hacken (1994: 156 ff.) the criterion of recursiveness may be
interpreted in various ways: it may refer to multiple application of different derivational/
inflectional rules, or to the reapplication of one and the same rule. The former criterion
is not of much relevance. In derivation it is very common. An extreme case may be
illustrated by West Greenlandic where there "are around 400 productive affixes in use
[ ... ]. They may combine with each other iteratively, producing a prodigious potential
for the derivational expansion of simple stems; up to ten or more affixes in succession
before the inflectional ending is not particularly unusual [ ... ]" (Fortescue 1984: 313).
On the other hand, at least agglutinative languages admit more than one inflectional affix
in a word (resulting from the application of more than one inflectional rule) as in the
following example from Hungarian:
(16) haz-a-ink-ban
haz-alk:-ink-ban
house-PL-POSS.PL-DATIVE
'in our houses'
That the prediction of the non-existence of recursive application of the same rule in
inflection, as compared to the cyclic application of derivational rules, is not absolute
follows from Dressler's (1989: 8) example taken from Turkish:
PA227
227
(17) ev-ler-de-ki-ler-de
house-PL-LOC-that/which/who-PL-LOC
'in those which are in the houses'
As to the related criterion specified in this section, Anderson (1985: 33) gives an example
from Kwakw'ala by referring to the suffixes -amas 'to cause' and -exsd 'to want'. From
the verb ne'nakw 'to go home', we can form ne'nak"" exsd 'to want to go home'. By
attaching the suffix -amas we obtain ne'nakw'exsdamas 'to cause to want to go home'.
On the other hand, from q'aq'oA.a 'to learn' we can make q'aq'oJ.aamas 'to cause to
learn, teach' and, from this, q'aq'o'Aamadzexsd 'to want to teach' can be formed in turn.
In these examples, the same suffixes appear in opposite orders corresponding to different
meanings 'to cause to want' vs. 'to want to cause'.
3.7. Paradigms
While inflectional morphology is typically organized into paradigms and inflectional
classes, the paradigmatic organization of derivational morphology is much weaker.
While this criterion reflects another deep-rooted false belief (see also, for example, Katamba 1993) van Marie rejects the view that "paradigmatics is of little or no importance
to derivational morphology. The only justifiable conclusion is that, in derivation, paradigmatic structure may manifest itself in a fundamentally different way from the way it
does in inflection" (1994: 2929). This stance may be further supported by Valles' view
that "word-formation patterns emerge from paradigmatic relations" (2003: 14). Let us
also direct our attention to the highly elaborated conceptions of derivationa l paradigms
in Dokulil (1962: 7- 18) and Furdik (2004). They show that the derivational system is
in fact based on paradigmatic relations reflecting the internal structure of the wordformation system of a language as a whole and that word-formation relations cannot be
reduced to those between the motivating and the motivated words. Rather, each such
pair of words is the basis for much more complex relations, including clusters relating
a motivating word w ith a set of motivated words. Thus, to take Dokulil's (] 962) example, the Czech word list ' leaf' as a motivating word is related to a large number of
motivated words like listek 'small leaf', listina 'document', listed 'dossier', olisteny
' provided with leaves', listnary 'deciduous', listovni 'concerning a letter', listovat 'to
browse', listopad 'November ', etc.
A word can also be related to a series of words whose neighbouring members are
related by motivation:
(18) list
'leaf'
listek
' leaflet'N
listkory
'leaflet ' A
listkoviry
'having leaflets' A
listkovitost
'property of having leaflets' N
This brings us to what might be called a derivational paradigm in a narrow sense, correspondi ng to the concept of proportional series:
PA228
228
II. Units and processes in word-formation I : General aspects

(19) list 'leaf'
kvet 'flower'
zub ' tooth'
roh 'corner'
smrk 'spruce'
---
list-ek
kvit-ek
zoub-ek
rnz-ek
smrc-ek
--
--
listk-ovy
kvitk-ovy
zoubk-ovy
rnik-ovy
smrck-ovy
---
listkov-itj
kvitkov-itj
zoubkov-itj
ruikov-ity
smrckov-itj etc.
3.8. Mental lexicon

While accepted derived words are likely to be stored as wholes in memory, inflected
word forms are unlikely to be so.
Booij (2006: 659) refers to the contradictory results of recent psycholinguistic experiments some of which suggest that derived words, even regular ones, are always stored,
while only irregular inflections are stored in the lexical memory (Clahsen, Sonnenstuhl
and Blevins 2003 ). Plag ( 1999: 11) refers to psycholinguistic and structural linguistic
arguments according to which regular complex words can also be stored in the lexicon.
Furthermore, experiments by Sternberger and MacWhinney ( 1988), and Baayen, Lieber
and Schreuder ( 1997) indicate that regular inflectional fonns of high frequency of occurrence are also stored.
3.9. Distribution
Inflectional affixes typically have a more peripheral position in the word form than
derivational affixes.
This criterion is, in fact, a paraphrase of Greenberg's Universal 28: "If both the derivation and inflection follow the root, or they precede the root, the derivation is always
between the root and the inflection" (Greenberg 1966: 93) and has been justified within
various theoretical frameworks (cf. Dressler 1989 for the arguments in favour of this
criterion from the position of natural morphology, Kiparsky l 982a, b and Mohanan 1986
for arguments within lexical phonology, and Anderson 1992, Perlmutter 1998, and Booij
2006 for the arguments based on the theory of split morphology). There are, however,
important exceptions to this criterion. They may concern individual cases or follow from
the existence of languages whose system as such is an exception to the general expectations. It is this latter type that suggests that not all exceptions belong to the periphery of
a language system as believed by Scalise (1988: 566-567). As for the first type of
exception, let us mention the well-known English examples like worsen, betterment,
unhappier, or Hebrew 7 imahut 'motherhood' formed from the plural of 7 em 'mother ',
i.e. pl. 7 imahot by means of the suffix -ut (Schwarzwald 2001: 29). The other type of
exception may be illustrated with Ket. Vajda (pers. comm.) demonstrates that it is diffi cult to separate derivational and inflectional affixes in the finite verb: they are interspersed in between one another in a rigid template of eight prefix position classes:
PA229
(20) subject 8
incorporated root 7
subject or object 6
derivational consonant 5
tense/mood affix 4
neuter class subject or object 3
tense/mood consonant 2
subject or object affix 1
verb root 0
Another example is provided by the agglutinating verbal morphology of Kujamaat J6ola .
Fudeman (in Aronoff and Fudeman 2005: 96) notes that the derivational suffix -u ' from'
can follow inflectional markers, as exemplified in (21 ):
(21)
narilye
-u
-rilJ
3AGR- arrive- HAB- from- redup
' He habitually arrives from'
Rubino (pers. comm.) gives an example of interspersed derivational and inflectional

affixes in Ilocano, an Austronesian language:
(22) nagkinnaawatanda
n-ag-ka(inn)-awat-an-da
PERF-AV-COMIT(RECIP)-understand-NML-3p
'they understood each other'
where ag- is derivational (Av- agentive voice), n- is inflectional based on ag- (perfective),
ka- is derivational (comitative), and -inn- is inflectional (reciprocal). A similar situation
can be found in Udihe (Nikolaeva and Tolskaya 2001 ), an Altaic language. Let us also
mention productively formed noun + noun compounds in the Dravidian language Telugu
(Pingali, pers. comm.) with a plural suffix inside them, for example:
(23) paLLapoDi
tooth-PL.NON .NOM-powder
' toothpowder'
(where pannu 'tooth' takes the plural marker which, due to a process of retroflexion
becomes paLLu. Its non-nominative form is paLLa. All other case markers and postpositions attach to this form).
3.10. Headedness
Only derivational suffixes can function as heads.
This issue has been much discussed in the literature, ranging from Marchand 's ( 1960,
1967) idea of a determinant- determinatum structure of the word-formation syntagma,
229
PA230
230

through Williams' (1981) righthand head rule, Lieber's (1981) feature percolation conventions, Selkirk's ( 1982) revised right-hand rule or D i Sciullo and Williams' ( 1987)
concept of relativized head, with the latter two allowing inflectional morphemes to function as heads, up to Bauer's (1990) head-related skepticism and Stekauer 's counterargument, to name only few of a number of various approaches to the concept of headedness which differ substantially in their treatment of the concept of head. A picture coming out of these different treatments suggests that (a) the concept of head depends on
the particular theoretical approach, and (b) that some of these approaches assign the
status of head also to inflectional morphemes.
3.11. Exponents
While the positioning and the segmental, suprasegmental, syllabic, and morphemic structure ofthe exponents of inflectional categories are relatively similar, those ofderivational
categories are dissimilar.
While analyzing the categories of verbal 3rd person singular agreement and nominal
number, Plank (1994: 1676) maintains that (a) exponents of inflectional categories, if
expressed by affixes, are exclusively suffixes; (b) they are phonetically similar; (c) they
are mostly expressed by monosyllabic or nonsyllabic exponents; (d) they neither carry
nor change the position of stress. While this is true of English, it does not hold universally, and there are languages which express inflectional categories by means of prefixes
as well. An example from Upper Necaxa Totonac (Beck, pers. comm.) is given in (24):
(24) nakinka:tala'hx'a:te:lhaniya:ntunka
na-kin-ka:-ta-la'h-x' a: -te :lha-ni-ya: -n-tunka
FUT-10BJ-PL.OBJ-3PL.SUBJ-DIST-husk.corn-AMB-BEN-IMPF-20BJ-very
' They will go around husking lots of corn for us.'
A similar example comes from Zulu (Van der Spuy, pers. comm.):
(25) be-si-nga-sa-zu-m-biz-a
PAST-we-NEG-still-FUT-him/her-call-VERB
'We were no longer going to call him/her. '
Neither does the issue of phonetic similarity hold universally. Inflectional languages
provide ample counter-examples, for example, various plural markers for various substantival paradigms, to name just one of them. On the other hand, derivational affixes
may show homonymy too: consider the polysemous/homonymous suffixes standing for
AGENTS, INSTRUMENTS, PATIENTS, etc. in English, but not only there.
3.12. Bases
While inflectional morphemes are typically added to existing bases, derivation can also
make use of non-existing bases.
PA231
For illustration, as pointed out by Newman (pers. comm.), Hausa agentives are (usually)
formed from verbs. Many of these verbs are formed from simple nouns by means of the
suffix -taa, e.g., aikii 'work', aikataa 'to work', ma'aikit.cii 'worker'. However, some
agent nouns are based on non-existing verbs. The verb *zinaata does not actually occur
as such, but one has to postulate it as a step in the derivation to make any sense of what
is happening:
(26) mazinaac'f 'adulterer'
zina 'adultery'
The Tzotzil suffix -tik 'it's a place of' combines with intrans1t1ve stems which never
occur independently (Cowan 1969: 102- 103) as in (27):
(27)
te 7-tik
tree/stick-it.is. a.place.of
'a p lace of trees'
Bauer (2009: 407) refers to some compounds in Danish one constituent of which has no
independent existence:
(28) a.
bom-uld
?-wool
'cotton'
b. jom-fru
?-woman
'virgin'
Contrary to postulate 12, there are, however, plural forms in English without any corresponding singular (belonging-s, outskirt-s, particular-s, etc.). Similarly, the comparative
of good, i.e. bett-er, is also based on a non-existing base.
4. Conclusions
Not al l of the criteria proposed in literature could be mentioned here for reasons of
space. However, this brief overview provides adequate support for the v iew that it is not
possible to draw a clear-cut borderline between inflection and derivation and that the
relation between these two areas of morphology is best treated "as a cline rather than a
dichotomy" (Katamba 1993: 21 7), with prototypical cases at both ends of the cline.
5. References
Anderson, Stephen R.
1982
Where's morpho logy. Linguistic Inquiry 13(4): 571-612.
231
PA232
232

1985
Typological distinctions in word formation. In: Timothy E. Shopen (ed.), Language
Typology and Syntactic Description. Vol. 3: Grammatical Categories and the Lexicon,
4- 56. Cambridge: Cambridge University Press.
1992
A-Morphous Morphology. Cambridge: Cambridge University Press.
Aronoff, Mark and Kirsten Fudeman
2005
What is Morphology? Oxford: Blackwell.
Baayen, Harald, Rochelle Lieber and Robert Schreuder
1997
The morphological complexity of simplex nouns. Linguistics 35: 861 - 877.
Bauer, Laurie
Be-heading the word. Journal of Linguistics 26: 1- 31.
1990
Bauer, Laurie
2000
What you can do with derivational morphology. In: Sabrina Bendjaballah, Wolfgang U.
Dressler, Oskar E. Pfeiffer and Maria D. Voeikova (eds.), Morphology 2000. Selected
Papers from the 9 111 Morphology Meeting. Vienna, 24- 28 February 2000, 37- 48. Amsterdam/Philadelphia: Benjamins.
Bauer, Laurie
2009
Compounding in Danish. In: Rochelle Lieber and Pavo! Stekauer (eds.), Oxford Handbook of Compounding, 400-416. Oxford: Oxford University Press.
Beard, Robert E.
1981
The lndo-European Lexicon. A Full Synchronic Theory. Amsterdam: North-Holland.
Beard, Robert E.
1982
The plural as a lexical derivation. Glossa. 16(2) : 133- 148.
Beard, Robert E.
1995
Lexeme-Morpheme Base Morphology. A General Theory of Inflection and Word Formation. Albany, NY: State University of New York Press.
Belimov, Eduard I.
1991
Ketskij sintaksis. Situaczja, propoziczja, predloi enie. Novosibirsk: Izdat. Novosibirskogo
un iversiteta.
Booij, Geert E.
1994
Against split morphology. In: Geert Booij and Jaap van Marie (eds.), Yearbook of Morphology 1993, 27- 49. Dordrecht: Kluwer.
Booij, Geert E.
2006
Inflection and derivation. In: Keith Brown (ed.), Ency clopedia of Language and Linguistics. 2nd ed., 654-661. Oxford: E lsevier.
Bybee, Joan L.
1985
Chomsky, Noam
1970
Remarks on nominalization. In: Roderick Jacobs and Peter Rosenbaum (eds.), Readings
in English Transformational Grammar, 184- 221. Waltham, MA: Ginn.
C lahsen, Harald, Ingrid Sonnenstuhl and James P. Blevins
2003
Derivational morphology in the German mental lexicon: A dual mechanism account.
In: Harald Baayen and Robert Schreuder (eds.), Morphological Structure in Language
Processing, 125 - 157. Berlin/New York: Mouton de Gruyter.
Cowan, Marion M.
Tzotzil Grammar. N orman: Summer School of Linguistics of the University of Okla1969
homa.
Derzhanski, I van A.
2005
On diminutive plurals and plural diminutives. In: Geert E. Booij, Emiliano Guevara,
Angela Ralli, Salvatore Sgroi and Sergio Scalise (eds.), M orphology and Linguistic
PA233
Typology. On-line Proceedings of the Fourth Mediterranean Morphology Meeting
(MMM4_) Catan ia 21-23 September 2003, University of Bologna, 2005 . [http://morbo.
lingue.unibo.it/mmm/].
Di Sciullo, Anna-Maria and Edwin Williams
On the D efinition of Word. Cambridge, MA: MIT Press.
1987
Doku lil, Milos
1962
Tvofeni slov v cdtine. I. Teorie odvozovani slov. Prague: Nakladatelstvi Ceskoslovenske
akademie ved.
Doku lil, Milos
1982
K otazce slovnedruhovych pfovodu a prechodu, zvl. transpozice. Slovo a slovesnost 43:
257- 271.
1989
Prototypical differences between inflection and derivation. Zeitschriftfar Sprachwissenschaft und Kommunikationsforschung 42: 3- 10.
Filipec, Josef and Frantisek Cermak
1985
Ceska lexikologie. Praha: Academia.
Fortescue, Michael
West Greenlandic. London: Croom Helm.
1984
Furdik, Juraj
2004
Slovenska slovotvorba. Prefov: Nauka.
Greenberg, Joseph H .
1966
Some universals of grammar with particular reference to the orde r of meaningful elements . In: Joseph H. Greenberg (ed.), Universals of Grammar, 73- 113. Cambr idge,
MA: MIT Press.
Haspelmath, Martin
1996
Word-class changing inflection and morphological theory. In: Geert E. Booij and Jaap
van Marie (eds.), Yearbook of Morphology 1995, 43-66. Dordrecht: Kluwer.
Katamba, Francis
1993
Morphology. London: MacMillan.
Kiparsky, Paul
1982a Lexical morphology and phono logy. In: In-Seok Yang (ed.), Linguistics in the Morning
Calm. Selected Papers from SICOL-1981 , 3- 9 1. Seoul: Hanshin.
Kiparsky, Paul
1982b From cycl ic phonology to lexical phonology. In: Harry van der Hulst and Norval Smith
(eds.), The Structure of Phonological Representations. Part I, 13 1- 175 . Dordrecht: Faris.
Laca, Brenda
200 1
Derivation. In: Martin Haspelmath, Ekke hard Konig, Wulf Oesterreicher and Wolfgang
Raible (eds.), Language Typology and Language Universals, 1214- 1227 . Berlin/New
York: de Gruyter.
Lieber, Roche lle
On the organization of the lexicon. Ph.D. dissertation, MIT. Bloomington, IN: Indiana
198 1
University L inguistics Club.
Marchand, Hans
1960
The Categories and Types of Present-Day English Word-Formation . Wiesbaden: Harrassowitz.
Marchand, Hans
1967
Expansion, transposition, and derivation. La Linguistique 1: 13-26.
Matthews, Peter H.
199 1 Morphology. An Introduction to the Theory of Word-Structure. 2 11d ed. Cambridge: Cambridge University Press.
Mohanan, Karuvannur P.
1986
Th e Th eory of Lexical Phonology. Dordrecht: Reidel.
233
PA234
234

Niko laeva, Trina and Maria Tolskaya
200 1 A Grammar of Udihe. Berlin/New York: Mouton de Gruyter.
Perlmutter, David M.
1988
The split morpho logy hypothes is: evidence from Yiddish. In: Michael Hammond and
Michael Noonan (eds.), Theoretical Morphology. Approaches in Modern. Linguistics,
79- 100. San Diego, CA: Academic Press.
Plag, logo
1999
Morphological Productivity. Structural Constraints in English Derivation. Berlin/New
Plank, Frans
1994
Inflection and de rivation. In: Rona ld E. Asher (ed.), Encyclopaedia of Language and
Linguistics. Vol. 3, 1671- 1678. Oxford: Pergamon.
Ricca, Davide
2005
Cumulative exponence involving derivation . Jn: Wolfgang U. Dressler, Diete r Kastovsky, Oskar E. Pfeiffer and Franz Rainer (eds.), Morphology and its Demarcations. Selected Papers from the 11 1" Morphology Meeting, Vienna, Februa01 2004, 197- 214.
Amsterdam/Philadelphia: Benjamins.
Robins, Robert H.
197 1 General Linguistics. An Introductory Survey. London: Longman.
Sapir, David J.
1965 A Grammar of Diola-Fogny. A Language Spoken in the Basse-Casamance Region of
Senegal. Cambridge: Cambridge University Press.
Scalise, Sergio
1988
Inflection and derivation. Linguistics 26: 56 1-581.
Schwarzwald, Ora
2001
Modern Hebrew. Munich: LINCOM Europa.
Selkirk, Elisabeth 0 .
1982
The Syntax of Words. Cambridge, MA: MIT Press .
Stekauer, Pavol
1998
An Onomasiological Theory of English Word-Formation. Amsterdam/Philadelphia: BenJamtos.
Stekauer, Pavol
2001 Beheading the word? Please, stop the execution. Folia Linguistica 34(3 -4): 333-355.
Stekauer, Pavo!, Salvador Valera and Livia Kortvelyessy
2012
Word-Formation in the World's Languages. A Typological Survey. Cambridge: Cambridge University Press.
Sternberger, Joseph P. and Bria n MacWhinney
1988
Are inflected forms stored in the lexicon? In: Michael Hammond and M ichael Noonan
(eds.), Theoretical Morphology. Approaches in Modern Linguistics, 101-116. Orlando:
Academic Press.
Strauss, Steven L.
On ' relatedness paradoxes' and related paradoxes. Linguistic Inquiry 13: 695- 700.
1982
Stump, Gregory T.
2005
Word-formation and inflectional morphology. In: Pavo! Stekauer and Rochelle Lieber
(eds.), Handbook of Word-Formation , 49- 71. Dordrecht: Springer.
ten Hacken, P ius
1994 Defining Morphology. A Principled Approach to Determining the Boundaries of Compounding, Derivation, and Inflection. Hildesheim: Olms.
Thompson, C had
1996
The Na-Dene middle voice: An impersonal source of the D-element. International Journal of American Linguistics 62(4): 35 1-378.
PA235
15. Units of word-formation
235
Valles, Teresa
2003
Lexical c reativity and the organ ism of the lexicon. Annual Review of Cognitive Linguistics l : 13 7-160.
van Marie, Jaap
1994
Paradigms. In: Ronald E. Asher (ed.), Encyclopaedia of Language and Linguistics. Vol.
6, 2927- 2930. Oxford: Pergamon.
van Marie, Jaap
1995
The unity of morphology: On the inteiwovenness of the derivational a nd inflectional
dimension of the word. In: Geert E. Booij and Jaap van Marie (eds.), Yearbook of
Morphology 1995, 67- 82. Dordrecht: Kluwer.
Werner, Heinrich
1998
Probleme der Wortbildung in den Jenissej-Sprachen . Miinchen/Newcastle: LINCOM
Europa.
Williams, Edwin
1981
On the notions ' lexically related' and 'head of a word'. Linguistic lnquilJ' 12(2): 245 274 .
Pavo/ Stekauer, Kosice (Slovakia)

I.
2.
3.
4.
5.
6.
7.
8.
Introduction
Sign, morpheme, morph
Word and lexeme
Root, stem, affix
Types of affixes
Morphological processes
Inputs and outpu ts of word-formation rules
References
Abstract
The article discusses the notion of word and the smaller units that words are composed
of (morph and morpheme, root and affix, stem). It provides surveys of different types
of affixes and morphological processes (in particular, reduplication, substitution and
subtraction), with a focus on terms and definitions. It also addresses the question of
which units can be inputs to word-formation rules, with special attention to inflected
forms and phrases as well as the role of word classes.
1. Introduction
The term word-formation suggests that the fundamental unit of thi s linguistic subdiscipline is the word. Formation implies that we are dealing with rules that "form" (create,
PA237
thography or, for foreign scripts, a transliteration although an element of the spoken
language is meant, possibly even together with its content. An indication of the content
is typically placed in single quotation marks ' ... '. For abstract grammatical meanings,
special terms or their abbreviations are available; they are frequently written in capitals
or small capitals, e.g., 'PLURAL' or ' PL'. Otherwise, synonyms, paraphrases or translations
may be employed. Where this is relevant to the discussion, complex signs are analyzed
into their components (e.g., according to the Leipzig Glossing Rules, http:llwww.
eva.mpg.dellingualresourceslglossing-rules.php; cf. also Lehmann 2004).
Some authors regard information on how the sign can combine with others (word
class, subcategorization features, etc.) as a third side of the sign (e.g., Mel'cuk 1982: 40,
2006: 385), but it is equally possible to include it in the content - there is no cogent
reason to restrict the term content to "meaning" in a narrow sense. (Expression can be
understood more broadly as well, cf section 6.5.) It must be emphasized that, like all
linguistic units, both expressions and contents are not observable realities in the extralinguistic world but abstractions which belong to the language system as a theoretical
construct.
An American approach pioneered by Leonard Bloomfield regards a word as one type
of linguistic form, where a form "is a recurrent vocal feature", i.e. a string of phonemes,
and "has meaning" (Bloomfield 1926: 155, emphasis J. M. ). The meaning thus remains
external to the form. Whether a string of phonemes is a form or not depends on whether
it has meaning, but it should not be relevant which meaning it has: -er in great-er and
-er in writ-er should be the same form despite the difference in meaning ('COMPARATIVE'
vs. 'AGENT NOMINALIZER '), whereas -er in winter does not qualify as a form since one
cannot ascribe any meaning to it. In actual fact, Bloomfield assumed that forms can be
different although they are "alike[ ... ] as to vocal features", i.e. phonologically identical,
giving -er ' COMPARATIVE' and -er 'AGE NT NOMINALIZER' as an example (1926: 157).
Such a statement presupposes that the meaning is an integral part of these elements. The
failure to distinguish clearly between signs and their expressions, between two-sided and
one-sided elements, is characteristic of much that is said about morphological units, even
in textbooks and handbooks. In order to achieve greater c larity and consistency, I will
work with the concept of the bilateral linguistic sign.
2.2. Minimal sign

Linguistic signs come in various sizes, from entire texts down to parts of words. But if
we segment a text into progressively smaller signs, we eventually reach a point where
no division on the expression side matches a division on the content side and vice versa.
For instance, English filly has an expression that consists of the phonemes lfl, hi, Ill
and Iii and a content that can be decomposed into 'young', 'female' and 'horse'. A
comparison with colt 'young male horse', mare 'adult female horse', etc. shows that it
is impossible to associate any of the phonemes of filly with any of the meaning components - filly 'young female horse' is a minimal sign, a sign that cannot be segmented
into smaller signs. Less common alternative terms are simple sign and elementary sign;
in the 1970s, several German linguists appropriated Hj elmslev's term plereme for this
purpose, but without much success. Quite often, morph or morpheme are used in this
sense but these terms can also refer to sets of minimal signs (see section 2.3).
237
PA238
238

In order to show that a sign is complex, i.e. not minimal, consisting of the signs A
and C, we should ideally be able to find two further signs B and D that can be substituted
for A and C, respectively, such that all the combinations are signs and satisfy the proportion AC : BC :: AD : BD on both the expression side and the content side, forming a
" square". If A is eat- and C -ing, we can use walk- as B and -s as D to complete the
proportion eating: walking:: eats : walks (cf. Greenberg 1957: 20). A "perfect square"
meets two conditions:
a) The expressions of both parts of the sign to be segmented recur elsewhere, i.e. alone
or in combination with other signs, and
b) their content in these other occurrences is the same as in the sign under analysis.
If only one of these conditions is met, it may still be appropriate to decompose the
sign. The famous word cranberry violates condition a) since there is no X such that
cranberry : blueberry :: cranX : blueX. Yet, cranberry seems to have the same structure
as blueberry. We would therefore like to isolate cran- as a minimal sign, perhaps with
the content "whatever it is which differentiates cranberries from these other kinds of
berries" (Hockett 1958: 126). Such non-recurrent signs are called unique; other terms
are remnant, cranberry morph[eme], blocked morph[eme] or pseudomorph[eme] and the
more specific uniroot and unifix (cf. Mel'cuk 1982: 66, 87- 88) for unique roots and
unique affixes (see section 4.1).
Condition (b) is violated by the following square:
( 1)
-ceive
-sist
-tain
-vert
per-
p erceive
p ersist
p ertain
p ervert
re-
receive
resist
retain
revert
con-
conceive
consist
contain
convert
The parts per-, re-, con- and -ceive, -sist, -fain, -vert recur in many combinations, but
their content is not constant and, in fact, highly elusive. From a semantic point of view,
nothing is gained by decomposing p erceive, etc., but the systematic relationship between
-ceive and -cep- (in perception, reception, conception) or -vert and -vers- (in perversion,
reversion, conversion) is an argument for recognizing -ceive, -vert, etc. as morphological
elements (cf. Aronoff 1976: 11-14; see also article 16 on derivation, section 4.3). Such
a recurrent part without clear content but with some morphological function has been
called quasi-sign or quasi-morph[eme]. Submorph[eme] and submorphemic [element,
unit] may be restricted to quasi-signs but can also be cover term s for a wider variety of
morphological units that lack certain properties of canonical minimal signs (cf. Kubrjakova 2000: 417).
The terms pseudomorph[eme], quasi-morph[eme], submorph[eme], etc. should be taken
as labels that point to certain difficulties but do not refer to well-defined types of units.
The same applies to the expressions defectively segmentable and conditionally segmentable which have been suggested to describe signs that pose formal or semantic problems,
as illustrated by cranberry and p erceive, respectively (cf. Kubrjakova 2000: 419). Faced
with the question "to segment or not to segment?", we must u ltimately make a decision.
PA239
239
E levating doubtful cases to the status of entities of a special kind avoids the issue and
does not lead to greater clarity (see also section 4.2).
Apart from the question of whether to segment and if so, where (cf. also Spencer
2012), a more fundamental difficulty is that expression and content are often not in a
simple 1 : 1 relationship (cf. Coates 2000). A well-known example is that of English
nouns:
(2)
SINGULAR
root /ru:t/
loop / lu :p/
proof /pru:f/
goose /gu :s/
tooth /tu:8/
PLURAL
roots /ru:ts/
loops /Iu:ps/
proofs /pru :fs/
geese /gi :s/
teeth / ti :8/
Since the first three plural forms clearly consist of two parts, the second of which is Isl
'PLURAL', we would like to analyze lgi:sl and lti:01 similarly, as combinations of lgu:sl
and ltu:01 with a minimal sign that has the content 'PLURAL' . However, the segmentation
procedure outlined so far does not allow this in a straightforward way. Some linguists
have responded to the challenge by abandoning the initial assumption that the expression
of a sign must be a continuous sequence of segments, while others have rejected the
building-block model entirely (see section 6.5).
2.3. Morpheme and morph

The segmentation problem on the syntagmatic axis - is this item a single element or a
sequence of two? - has a parallel in the classification problem on the paradigmatic axis:
Are these items two different elements or instances of the same element? At one level
of analysis, the minimal signs Isl 'PLURAL', lzl 'PLURAL' and hzl 'PLURAL' that we can
identify in proofs lpru:f-sl, grooves lgru:v-zl and nooses lnu:s-1z/ are different because
they differ in expression (cf. section 2.1 ). On the other hand, they have the same content
and can therefore be considered variants of a single element at a different level of analysis. Such an element, a set of minimal signs w ith the same content, is a morpheme (for
other meanings of this term, see section 2.5), and the members of the set are its allomorphs (see sectio n 2.4). Conversely, Isl 'PLURAL' can be grouped together with Isl '3RD
PERSON SINGULAR' in writes lra1tsl and Isl 'POSSESSIVE' in Jack's lci;3reksl; such a set of
minimal signs with the same expression is a morph. As defined here, allomorphs are
minimal signs and not morphs, although often only their expression is given. According
to this terminology, both morphs and morphemes are sets and are defined in mirrorimage fash ion - morphs by identity of expression, morphemes by identity of content - ,
but there is no simple term for minimal signs. Alternatively, morph can be used for the
minimal sign (cf. Mel'cuk 1982: 63, 2006: 388), in which case the unit "set of m inimal
signs with the same expression" has no name - a tolerable gap since this unit is of far
less interest than the morpheme. The correspondence between concepts and terms can
be summarized as in (3).
Unfortunately, it is often not clear which of the meanings of morph is intended, i.e.
whether -s in proofs and -s in writes are regarded as the same morph or as two different
ones. Worse still,. many textbook authors introduce the term morpheme in connection
with the demonstration that complex words can be decomposed into several min imal
PA241
(4)
241
S INGULAR
PLURAL
NOMINATIVE
miast-o
miast-a
DATIVE
miast-u
miast-om
Here, number and case are always expressed simultaneously, a phenomenon known as
cumulative exponence or cumulation. (It normally concerns inflectional categories, cf.
Coates 2000: 619- 620; for examples involving derivation, not all of them fully convincing, see Ricca 2005. The terms fusion and/or portmanteau are sometimes used for cumulation, but they are better reserved for cases where the simultaneous expression of two
or more content elements is a deviation from the normal pattern.) In the nominative
plural miasta, for instance, -a expresses both 'PLURAL' and 'NOMINATIVE'. While Polish
has these content features, it does not have distinct morphemes {PL} and {NoM} but a
single morpheme {NOM;PL} .
The Polish example (4) also illustrates another important point. Since only neuter
nouns have a nominative plural in -a, one might consider including 'NEUTER' ('N' for
short) in the content of -a. I prefer to say that 'N' is a feature of miast- which prompts
the choice of -a (see also section 2.4 on morphological conditioning). Thus, the content
of a linguistic sign does not necessarily include all the information that can be gathered
from its occurrence. It need not be a fu ll specification of its meanings and uses either.
For example, there is no necessity to distinguish a perfect participle as in She has written
the letter and a passive participle as in The letter was written for every Eng:Jish verb. In
my view, written is the same participle everywhere (whether we label it 'PERFECT PASSIVE PARTIC IPLE' , 'PAST PARTICIPLE' or whatever), and the fact that the constructions in
which this participle occurs with the auxiliaries have and be are grammatical idioms,
i.e. have meanings that cannot be deduced from their components, is not a matter of
morphology.
While it is true that a morpheme is more "abstract" than its allomorphs, in the sense
that it abstracts away from the differences in their expressions, this does not mean that
allomorphs or morphs are "concrete" or somehow more "real". Attempts at equating the
distinction between morpheme and (allo)morph with dichotomies such as langue vs.
parole, form vs. substance or system vs. text are untenable (cf. Mugdan 1986: 33- 34).
Similarly, it is potentially misleading to describe the relationship between them as one
of "realization" (Bauer 2003: 17, 334- 335) or to discuss allomorphs under the heading
"the pronunciation of morphemes" (Fromkin, Rodman and Hyams 2013: 225- 228). Just
as a toy building block retains its identity whether it is in its box or serves as part of an
edifice, so too a linguistic element remains the same whether we look at it as a component of the language system or of a text. The building-block analogy is not entirely
accurate, however, inasmuch as a given linguistic element can occur an unlimited number
of times in a text. If we are asked how many words there are in They might or they
might not, we can answer 6, counting the two occurrences or tokens of they and might
separately, or 4, counting they and might only once each, as a type. This distinction is
applicable to linguistic units at all levels; the difference between type and token is unrelated to that between X-eme and allo-X. As regards the latter pair, the essential point
is that certain differences which are taken into account at the allo-X level are ignored at
the X-eme level (cf. Mel'cuk 1982: 120-122, 2006: 400), and this is what the definition
of an X-eme as a set of allo-Xs is intended to capture.
PA242
242
2.4. Al lomorphy
When describing morphemes and their allomorphs, three aspects must be taken into
consideratio,n: conditioning, relatedness and domain. My focus will be primarily on terminology; for a fuller treatment and more examples, see article 27 on allomorphy.
a) Conditioning:
Which factor determines the occurrence of one allomorph rather than the other(s)? A
basic distinction is that between phonological conditioning, where only the expression
side is relevant, and morphological conditioning, where features of content are decisive.
Typical phonological conditioning factors are segments or features in the context, word
stress or tone, syllable structure, etc. (for surveys of various subtypes and issues, see
Neef 2000a; Nevins 2011; Bonet and Harbour 2012: 220- 234); sometimes, the distribution of allomorphs is best described by an "output constraint", i.e. with reference to the
resulting structure (cf. Booij 1998).
Some linguists distinguish between grammatical conditioning (or morphological conditioning in the narrower sense) and lexical conditioning, but rarely with sufficient clarity; many explanations mix criteria and the common denominator behind the examples
given is often not obvious. In particular, the distinction seems to be drawn according to
whether the conditioning factor is
- an affix or a stem (cf. "grammatical element" vs. "word" in Katamba and Stonham
2006: 30-31; on the notions affix and stem, see sections 4.1 and 4.3),
- an independently motivated morphosyntactic feature (e.g., [masculine] in a language
with gender agreement) vs. a special "diacritic feature" or "index" that must be recorded in the lexicon just to account for the allomorphy in question, such as [conjugation
class 2] or [undergoes rule X] (cf. Haspelmath and Sims 2010: 26) or
- a feature that defines a larger class of items vs. one that marks individual exceptions
(cf. Bauer 2003: 327, where inflection class is cited as an instance of grammatical
rather than lexical conditioning).
A more comprehensive typology of morphological conditioning must take into account
all of these factors as well as some others, such as conditioning by a morpheme vs. by
a specific allomorph.
Whether syntactic conditioning, i.e. conditioning by the syntactic structure in which
a morpheme appears or by the syntactic function it has, should be admitted as a further
type is controversial. Even examples that allegedly show that the assumption of syntactic
conditioning is inevitable (cf. Neef 2000b: 482-483) are open to alternative analyses
that do not involve allomorphy at all; in any case, they do not pertain to word-formation.
Finally, two or more allomorphs of a given morpheme may occur in the same context:
The plural of /ru:f/ 'roof' can be /ru:f-s/ or /ru:v-zl. This phenomenon is known as free
variation, but the variants are rarely freely interchangeable. They often differ in stylistic
value or correlate with differences in the speaker's age, regional background or social
status. Even where such factors are not detectable, one of the variants is likely to be
more frequent - and the statistics may differ from item to item (cf. Meder and Mugdan
1990: 92- 99).
PA245
also Bonet and Harbour 2012: 199- 201, where terminological choices are mistaken for
theoretical approaches; for a discussion of whether only affixes or also roots exhibit
"allomorphy" in this sense, see Bonet and Harbour 2012: 214-218). Again, this is occasionally confused with lack of phonological predictability, i.e. conditioning (see also De
Belder 2011: 137, where allomorphs may be neither phonologically conditioned nor
strongly suppletive). As a new term for phonologically related alternants, (phonological)
variant may be used; in this parlance, the English plural morpheme has, among others,
the allomorphs lzl 'PL' (with the variants /s/, /z/ and /Jz/) and fan/ 'PL' . Terminological
consistency is, however, a rare virtue, and it is not uncommon for authors to vacillate
between different uses of morpheme and allomorph or to give vague explanations that
leave open what is really meant.
In a French tradition that goes back over a century (cf. Mugdan 1986: 32), morpheme
is used only for minimal signs with grammatical meaning; those with lexical meaning
may then be called semanteme or lexeme (not to be confused with lexeme as defined in
section 3 .4), with moneme as a cover term (cf. Martinet 1960: 20 [ 15-16]). As an offshoot of this usage, Louis Hjelmslev employed morpheme for an individual grammatical
meaning, i.e. a morphosyntactic property (1961: 25-26). Presumably independently,
some English-speaking scholars did so as well (e.g., Lyons 1968: 189- 191; cf. also
Mugdan 1986: 36-37). The difference between this terminology and that adopted here
becomes apparent in cases of cumulative exponence, as in the Polish example (4), where
these authors would speak of 'PL' and 'NoM' as "morphemes". Another term for grammatical (and specifically inflectional) content elements, used chiefly by Russian linguists, is grammeme (cf. Mel'cuk 1982: 30, 2006: 23); although it also has various other
uses, it is a convenient alternative that makes it unnecessary to redefine morpheme in
this way.
3. Word and lexeme

3.1. The universa lity of t he word
The word, "that m ost elusive of all units encountered in the segmentation of utterances"
(Malkiel 1966: 310), is nonetheless "the unit par excellence of traditional grammatical
theory" (Lyons 1968: 194) and the basis of the conventional distinction between morphology and syntax according to which morphology deals with the internal structure of
words and syntax with their external combinations. While it is conceivable that some
languages do not have words and hence no distinction between morphology and syntax,
there is no convincing evidence that this is indeed the case. (My remark that "some
languages evidently lack a word level altogether", Mugdan 1994: 2552, should have
been formulated more cautiously.) The word is therefore generally assumed to be universal although this cannot be proven. Interestingly, not all languages have a word for
'word' and non-linguists are not necessarily conscious of words, especially if they do
not use a writing system that marks word boundaries (cf. Dixon and Aikhenvald 2002:
2- 4, 32- 33).
For a language to "have words" means that a unit word is to be distinguished from
both morpheme and phrase (cf. Bauer 2000a: 255). In languages that are said to be of
the isolating type, the average number of morphemes per word approaches 1 but still
245
PA246
246

remains a little higher so that words and morphemes cannot be equated. In Vietnamese,
for instance, the ratio in a text sample of 100 words was found to be l. 06 (Greenberg
1960: 193) - and subject to how one defines compounds, it may wdl be considerably
higher. As for polysynthetic languages, the impressive textbook examples of extremely
long words that correspond to fairly complex English sentences shouM not be taken to
imply that all sentences (or all phrases) in such languages consist of a single word.
It is notoriously difficult to provide a universal definition of the unit word. Since
there are different criteria that do not necessarily converge, the notion of the word is
neither a predictive concept such that being a word entails a number of logically independent properties (cf. Jacobs 2011 ), nor is it a comparative concept, i.e. we cannot be
sure that what we call word in language A is "the same thing" as what we call word in
language B (cf. Haspelmath 2011). But, as even those who raise these objections admit,
this does not invalidate the usefulness of the concept in practice, especially in the description of individual languages.
3.2. The word in orthography, phonology and the lexicon

The answer to the question "is this one word or a sequence of two?" will depend on
whether we consider orthography (section 3.2. 1), phonology (section 3.2.2), semantics
(section 3.2 .3) or grammar (section 3.3). More detailed surveys and discussions of the
various criteria can be found in textbooks (e.g., Lyons 1968: 199- 206; Bergenholtz and
Mugdan l 979a: 12- 29; Booij 2012: 283- 296), handbook articles (e.g., Bauer 2000a;
Julien 2006) and other publications devoted to the topic of the word (e.g., Peskovskij
1925; Dixon and Aikhenvald 2002: 10- 25; Haspelmath 2011).
3.2.1. Orthography
In many but by no means all writing systems, words are bounded by spaces or, more
rarely, some other symbol (cf. Daniels 2013: 66- 67). An exact definition of the orthographic word is not trivial, however. It must take into account various special symbols
(punctuation marks, digits, etc.) and will have to be different not only for different
writing systems but even for different languages that employ the same writing system
(e.g., English, French and German, which differ slightly in the use of spaces in conjunction with punctuation marks). Moreover, orthographic words do not necessarily correspond to units of phonology or grammar. Even orthographies devised with the participation of linguists (e.g., the Gennan spelling reform adopted in 1998) are not fully convincing in this regard. Apart from that, detailed official rules do not eliminate uncertainty
and fluctuation in practice.
3.2.2. Phonology
The phonological word (or prosodic word,p-word, c.v) is a constituent between the syllable (or the foot) and the phonological phrase in a hierarchy of prosodlic units, i.e. units
PA248
248

3.2.3. Lexicon
If the lexicon is understood as the repository of irregularities, we expect it to contain

simple words because their meanings are arbitrary but not phrases because their meanings can be deduced from the meanings of their parts and the mean ing of the construction. Such compositionality of meaning is lacking in many compounds so that they must
be listed in the lexicon, too. Although there is some semantic relation between the meanings of the parts, it can be almost any relation: Book house (a compound by the criterion
of stress) could mean 'a house that contains many books', 'a house where books are
produced', 'a house where books are sold', 'a house in the shape of a book', etc.; as a
rule, only one of these possibilities is conventionalized. It is often said that "many complex words are semantically compositional in exactly the same way as phrases and
clauses" (Haspelmath 2011: 36). But while it is true that we can easily understand numerous word-formations we have never encountered before, we often needl contextual, pragmatic or extralinguistic knowledge to select the correct interpretation among those that
are conceivable. Still, some complex words do have fully predictable meanings (e.g.,
most English derived adjectives with un-), which would mean that they are not listed in
the lexicon. Conversely, numerous phrases are not semantically compositional, e.g., idioms such as beat about the bush 'avoid the point at issue' or so-called particle verbs
such as put up 'provide food and lodging' (see also article 24 on multi-word expressions). Verbs with "separable prefixes" resemble derivatives to a greater degree because
there are constructions in which the element in question behaves both syntactically and
phonologically like part of a word, e.g., Dutch op in ik moet [hem] opbellen ' I must ring
[him] up', but since there are other constructions in which op clearly is a separate particle, e.g., ik bel [hem] op ' I ring [him] up', such verbs must be regarded as idiomatic
word combinations, too (see also aiiicle 23 on particle-verb formation and articles 3537 on particle verbs in Germanic, Romance and Hungarian). In cases such as Portuguese
moinho de vento 'mill of wind (windmill)', which are sometimes classified as compounds, the fact that plural is marked on the first constituent (moinho~ de vento) suggests
that they are likewise idiomatic syntactic constructions rather than complex words. Some
linguists extend the term lexeme (see section 3.4) to multi-word expressions, but it is
preferable to use a different term for an item that must be listed because its properties
cannot be inferred from those of its parts; a highly successful suggestion is listeme (Di
Sciullo and Williams 1987: 3). Lexical item is more neutral and also applicable in models
which allow the lexicon to contain redundant information, in particular a semantically
compositional whole as well as its pa1is.
3.3. The word in grammar

In syntax, the word is located on the lowest rung of a hierarchy of units that traditionally
reaches up to the sentence, with at least phrase and clause in between. It is generally
assumed that this lowest syntactic unit is simultaneously the highest morphological unit
so that a suitable term for it is morphosyntactic word, whereas grammatical word is also
used in another sense (see section 3.4). Morphosyntactic words are said to be characterized by external autonomy and internal cohesion; more specifically, this may mean:
PA249
- Words can be uttered in isolation, parts of words cannot (section 3.3.1 ).
- A word cannot be interrupted by pauses or "other forms" (Bloomfield 1933: 180), but
it is possible to, separate the words in a sentence from one another in this way (section
3.3.2).
- The parts of a word cannot be reordered whereas the order of words in a sentence
can be changed (section 3.3.3).
- Syntax does not have access to the internal structure of words (section 3.3.4).
3.3.1. Minimum sentence

Since a higher-level element can, in the limiting case, consist of a single lower-level
element, the word is often defined as a minimum sentence (or minimum utterance); thi s
is also what is meant by "minimum free form" (Bloomfield 193 3: 178). However, many
items that we generally call words are not used by themselves as sentences, e.g., in reply
to a question. For instance, the answer to What do you see there? can be Paw-marks (in
the plural) but not Paw-mark (in the singular, without an article), and while Hunting is
a possible response to What are you doing?, one would be hard put to find a question
which can be answered with the present-tense form Hunts alone. (There are languages
in which sentences can consist of a bare count noun in the si ngular or a bare finite verb,
but the set of minimum free forms is still smaller than that of words in the traditional
sense.) If we admit photo captions such as Paw-mark or book titles such as Hunted and
the like, the range of minimum free forms will be larger, but even this will not help in
many cases. Evidently, the minimum sentence criterion is at best a sufficient but not a
necessary one. Moreover, it might lead us to regard paw-mark as a sequence of two
words (cf. Haspdmath 2011: 39), and it is doubtful whether the problem can be solved
satisfactorily by pointing to the fact that mark in isolation has stronger stress than in
paw-mark (cf. Bloomfield 1933: 180).
3.3.2. U ninterruptability
In spontaneous speech, pauses and hesitation phenomena do not necessarily coincide
with word boundaries. It has therefore been suggested that the determination of points
at which pausing is possible should be based on slow and careful speech (Hockett 1958:
166-167), but this version of the criterion is still problematic. When dictating for a
transcriber, literate speakers will be influenced by conventional orthography, whereas
the points at which speakers of unwritten languages pause seem to depend on the language type; apart from that, pauses tend to be more closely related to phonological than
to morphosyntactic word boundaries (cf. Dixon and Aikhenvald 2002: 11-1 2, 23-24).
As regards interruptability by "other forms", we must specify which forms we mean.
Surely, we do not want to say that interrupt consists of two words because we can insert
-cept and dis- to give intercept and disrupt. Also, morphemes may certainly interrupt a
word: In Polish, the insertion of the comparative marker -sz- into mlod-y 'youngNOM.SG.M' results in mlodszy 'younger'. It seems, then, that the inserted material must
consist of whole words (cf. Vardul' 1991: 22) so that the criterion presupposes the unit
249
PA250
250

it is meant to identify (see already Peskovskij 1925: 124- 125). This circularity can be
eliminated if we determine words in stages, first those with the highest degree of autonomy (minimum sentences) and then those w ith lesser degrees of autonomy (cf. Plungjan
2003: 18- 28; see also Bloomfield 1933: 179). Then the criterion can be reformulated as
"a word cannot be interrupted by other words that have already been identified as such";
conversely, the fact that the paw-mark can be interrupted by small demonstrates that this
is a phrase even though the cannot be used alone.
Allegedly, some cases of noun incorporation (see article 22) violate this principle,
e.g., Pawnee (Oklahoma, USA) ta-tu-kut ' l have cut it for you', which can be interrupted
by rfks 'arrow', resulting in ta-tu-rfks-kut 'I have cut your arrow' (Haspelmath 2011: 45,
slightly modified from Julien 2002: 34- 35; this version differs in several details from
the original source, Boas 1911: 31, and the freestanding noun actually needs an absolutive suffix -u, cf. Parks 1976: 97-99 - but examples of this kind are, in principle,
possible). If so, should we perhaps argue that English doors consists of two words
because it can be interrupted by knob (in doorknobs)? As with the definition of the
word as a minimum sentence (see section 3.3.1), the difficulties posed by compounding,
including the subtype known as incorporation, can be partially overcome by showing
that the stem which occurs in the compound is not fully identical with the corresponding
freestanding word. One can also argue that the incorporation of nouns into verbs is much
more restricted than the insertion of words into a phrase (even sequences of words such
as unusually large can come between the and paw-mark) and that verbs with incorporated nouns behave essentially like simple verbs (cf. Julien 2002: 35 with reference to Boas
1911 : 31 - 32, where the line of reasoning is, however, somewhat different).
If words are uninterruptible, the notion of "di scontinuous word" is untenable so that
has come and will stop must be phrases rather than bipartite word forms as is sometimes
maintained (see also section 3.4 on the status of so-called periphrastic forms). One of
the more difficult cases is Russian nikto 'nobody' with the genitive nikogo 'of nobody'
and the instrumental nikem 'by nobody', which are intenupted by prepositions, as in ni
do kogo 'to nobody' and ni s kem 'with nobody' (cf. Vardul ' 1991: 25), but an obvious
solution is to regard ni as a clitic, i.e. a distinct morphosyntactic word (see section 3.2.2).
Other apparent counter-examples can be dealt with in the same way.
3.3.3. Fixed vs. variable order
One proof that I said consists of two morphosyntactic words is that they can occur in
either order, as in "I'll try, " I said and "Then I will go on, " said I. On the other hand,
the permutation test demonstrates that paw-mark is a single word because the opposite
order mark-paw is impossible. Since the order of words in a phrase is often quite rigid,
this is, however, another subsidiary criterion of limited applicability: We cannot transpose I'll try, and the five constituents of Then I will go on occur in only a few of the
theoretically possible 120 orders.
Conversely, there are cases where the parts of a word can be reordered. Many compounds are reversible, normally with a semantic difference as in boat-house vs. houseboat. Frequently, differences in the order of affixes correlate with differences in meaning
and hierarchical structure, i.e. the order in which the affixes are added in the derivational
PA251
hi story of the word (cf. Rice 20 11: 174- 176; see also article 16 on derivat ion, section
3. l ). In Dutch, for instance, some verb prefixes can appear in either order, although
minimal pairs are quite rare. One of them is veronzekerd 'made unsure (disconcerted)'
vs. onverzekerd 'uninsured', both derived from zeker 'sure'. In veronzekerd, the negative
prefix on- has scope over zeker. The addi tion of ver- to on-zeker 'unsure' leads to veron-zeker[en] 'to make unsure (disconcert)' (the bracketed en is the infinitive suffix of
the citation form) with the past participle ver-on-zeker-d. In onverzekerd, the order of
derivation is zeker 'sure' ~ ver-zeker[en] 'to make sure (insure)' ~ ver-zeker-d 'insured' (past participle) ~ on-ver-zeker-d 'uninsured'. It has been suggested that the
function of an affix detennines its position if it is derivational but not if it is. inflectional
(Stump 2001 b: 713 ); depending on how one defines the distinction between derivation
and inflection (see article 14), the existence of contrasting scope relationships among
tense/mood/aspect affixes either contradicts this assumption or shows that these affixes
are not inflectional in the language concerned (cf. Korotkova and Lander 2010: 306,
3 10, 314-315 on Adyghe).
Because of such examples, the principle that the order of morphemes in a word cannot
be varied is often restricted to free variation without a semantic difference (e.g., Dixon
and Aikhenvald 2002: 20). This generalization is sometimes called "Perlmutter 's universal" or " Postal and Perlmutter's universal" (cf. Perlmutter 1970: 234). Again, some counter-examples can be dismissed, more or less plausibly, as involving clitics. rather than
affixes - but not all (cf. Bickel et al. 2007; see also Good and Yu 2005). In fact, the
existence of variable order should not come as a surprise. If there is no scopal relationship between affix A and affix B or if affix A can have semantic scope over affix B but
not vice versa, the order of A and B does not matter; it may therefore be fixed arbitrarily,
vary freely (cf. also Rice 2011: 185- 187) or be governed by some external condition
(for factors that can be relevant, see Manova and Aronoff 2010: 111 - 112; Ryan 20 10;
Rice 2011). Order may also be immaterial if it makes hardly any semantic difference
whether A has scope over B or B over A (cf. Korotkova and Lander 2010: 305-306 on
Adyghe). But even if A can have scope over B and B over A with a clear contrast in
meaning (e.g., 'I'Il want to make her eat' vs. ' I'll make her want to eat'), we occasionally
find that either order is possible and compatible with both interpretations (cf. Caballero
20 10; see also Rice 2011: 187- 188). As for the constituents of compounds, free variation
is highly unusual but does occur: Vietnamese, where the native order (modifier second )
coexists with the Chinese order (modifier first), sometimes allows both options without
any apparent difference in meaning, e.g., in combinations with tru6ng 'leader' such as
tru6ng-l(rp or l6p-tru6ng 'leader of a group' (Bisang 2001: 199).
3.3.4. Invisibility of internal structure

It is commonly assumed that parts of words are not accessible to syntax. (This does not
apply to models in which the basic building blocks of syntax are morphemes rather than
words; cf. Julien 2002: 8- 11, 27- 28 for some discussion.) In particular, the member of
a compound that semantically modifies the other cannot be referred to with a pronoun;
this has been expressed in the slogan that words are "anaphoric islands" (Postal 1969:
205). In the "interpretable but mildly unsettling" utterance I went baby-sitting last night.
251
PA252
252
Boy, was she ugly!, which might look like a counter-example, the referent of she is not
established grammatically, via the compound-member baby, but rather pragmatically, by
our know ledge about the situation (Mithun 1984: 871; apart from its rudeness, the problem with the utterance seems to be the ambiguity of she, which could also refer to the
baby's mother).
Secondly, it should not be possible to modify a part of a complex word by an external
attribute. Here, exceptions are not difficult to find (see also article 16 on derivation,
section 3.4). For instance, a German cookery book from the late 1950s published by a
manufacturer of refrigerators has the title Kalte Kiichenkniffe 'cold cuisine_ tricks', where
kalte semantically modifies the compound-internal constituent Kiiche but agrees in number with the plural Kniffe (the singular would be ein kalte!_ Kiichenkniff 'a cold cuisine_
trick'). While native speakers find this a bit amusing, a much-discussed English example
involving derivation is perfectly normal: nuclear physicist does not mean 'physicist who
is nuclear' but 'specialist in nuclear physics'. Such discrepancies between grammatical
and semantic structure are known as bracketing paradoxes or m01phosemantic mismatches. A suggestion that would eliminate the mismatch is to analyze cases of this kind
as compounding with or derivation from a phrase (kalte Kiiche, nuclear physics), but
many reject this as theoretically unacceptable or empirically inadequate (see section 7 .1).
A third corollary is that words can be coordinated but parts of words cannot. It
appears to be falsified by Dutch hoofd- of nevenaccent 'main [stress] or secondary
stress', zicht- en tastbaar 'vis[ible] and tangible', ijs- en bruine beeren 'ice [bears] (polar
bears) and brown bears' (cf. Booij 2002: 171-172) and analogous constructions in other
languages (cf. Haspelmath 2011: 47- 49). These expressions could, however, be derived
from hoofdaccent of nevenaccent, etc. by ellipsis. Even if the idea that this is a purely
phonological phenomenon (cf. Booij 2002: 172- 173) may be an over-simplification, we
are not dealing with coordination below the word level (cf. also Chaves 2008).
In summary, the generalizations about words as grammatical units are partly theorydependent and partly open to objections that require further fine-tuning of the criteria.
If the tests for word status give intuitively undesirable results, we have a choice between
developing alternative interpretations of the data and giving up our preconceived notions.
Where there is room for legitimate disagreement as to whether a given element is a
morphosyntactic word or part of one, the final decision should be guided by overall
stmctural considerations but may often well be a matter of taste.
3.4. Word-form, grammatical word, lexeme

As regards the classification problem - are we dealing with two different morphosyntactic words or instances of the same word?-, we can draw a distinction that is similar to
that between minimal sign, morpheme and morph (see section 2.3) but involves an
additional pair of concepts, morphosyntactic category and morphosyntactic property (cf.
Matthews 1974: 66, 136). A morphosyntactic category is a set of two or more mutually
exclusive morphosyntactic properties, i.e. content elements, one of which must be chosen
for each word of a certain class. English has a morphosyntactic category number with
the two properties 'SINGULAR' and ' PLURAL' , and all nouns are specified for one of these
properties when used in a text. The term morphosyntactic indicates that the properties
PA253
in question are relevant to syntax and are expressed by morphological means, more
specifically by inflection. Therefore, inflectional category is frequently used instead;
morphological category and grammatical category should be broader notions but often
serve as synonyms of inflectional category. Some define morphosyntactic categories
more narrowly as those involved in government or agreement (notably case, number,
gender and person); categories that are not relevant to syntax in this particular sense but
encode semantic distinctions (e.g., tense, aspect, mood, voice/diathesis) are then called
morphosemantic (Kibort 2010: 64, 80- 83). One of the alternatives to the term (morphosyntactic) property that has already been mentioned is grammeme (see section 2.5).
Confusingly, (inflectional) category is sometimes used as the hyponym, with a somewhat
idiosyncratic hyperonym such as complex of categories (Wurzel 1989: 52) or inflectional
dimension (Haspel math 2002: 60- 61 ). The most important rival to the pair category/
property is f eature/value (e.g., Haspelmath and Sims 2010: 81-83; Kibort 2010: 66; cf.
also Stump 2001a: 38- 39). However,feature can also refer to an individual morphosyntactic property (e.g., Matthews 1991: 40). Besides, features do not always. correspond
one-to-one to categories if we postulate that features must have exactly two values.
Thus, the category person, which typically comprises the properties 'lsT PERSON', '2ND
PERSON' and '3RD PERSON', can be captured by two binary features, [me] and [you],
such that the three properties are reanalyzed as the combinations [+me,
-you], [-me, +you] and [-me, -you], respectively. As this example shows, we usually
enclose feature names in square brackets. Values are indicated by + and - (if there are
two) or by numbers (if there are more). Some linguists prefer a format like [NUM: sg],
where NUM stands for the feature number and sg for its value 'SINGULAR' (e.g., Zwicky
1985: 374; Stump 200la: 39-41). An affix or process (see sections 5-6) which expresses
a morphosyntactic property (or a bundle of properties) is called exponent (cf. Matthews
1974: 144, 1991: 175) or marker (cf. Wurzel 1989: 51), and the relationship between
morphosyntactic properties and exponents is known as exponence (for different types of
exponence, see Coates 2000).
Words that differ only in their morphosyntactic properties but otherwise have the
same meaning, e.g., book 'book.sG' and books 'book.PL', can be viewed as different
words or as variants of the same word, depending on whether the differences in morphosyntactic properties and the corresponding differences in expression are taken into account or not. Widely-used terms for these two senses of word are grammatical word and
lexeme. (For other meanings of these terms, see sections 2.5, 3.2.3, 3.3. Lexeme is attested in the Russian form leksema since the 1920s, cf. Peskovskij 1925: 138.) A lexeme is
a set of grammatical words which have the same content, save for their morphosyntactic
properties. While grammatical words are linguistic signs, a characterization of lexemes
as signs (cf. Aronoff 1994: 9; Montermini 2010: 86) deprives this tenn of its meaning.
Occasionally, a grammatical word is called lex or, in relation to the lexeme it belongs
to, allolex (in analogy to allomorph, see section 2.3). Lexemes are usually indicated by
small capitals, e.g., BOOK for the set {book 'book.sG', books 'book.PL'}. If necessary,
one can add a subscript in order to distinguish the noun BOOKN from the verb BOOKy.
The latter has, among others, the grammatical words booked 'book.PST' and booked
'book.PsT.PTCP', which we distinguish because the past tense and the past participle of
other verbs differ in expression (e.g., I booked/chose my flight two weeks ago and I have
already booked/chosen my flight). If, however, we focus on the expression side, booked
'book.PST' and booked 'book.PST.PTCP' can be regarded instances of the same word. The
253
PA254
254

term word form may be used for this third sense of word, i.e. a set of grammatical words
with the same expression. (The three-way distinction between lexeme, grammatical word
and word form goes back to Matthews 1972: 160- 161; conceptually, it is already formulated in Lyons 1968: 196- 197. According to one definition, a word form comprises not
only grammatical words of the same lexeme but also grammatical words of different
lexemes, such as books 'book.PL' as in Linguistics books are expensive and books
'book.3sG.PREs' as in She always books early; cf. Carstairs-McCarthy 2000: 595.) Some
use word form instead of grammatical word or do not state clearly whether booked
'book.PST' and booked 'book.PsT.PTCP' are the same word form or different ones (e.g.,
Booij 2012: 3; Lieber 2010: 7; cf. section 2.3 on the analogous ambiguity of morph).
Sometimes, word without further specification is restricted to one of the three senses,
e.g., lexeme (Donalies 2005: 19) or grammatical word (Matthews 1974: 26, 1991: 31; I
will follow this practice for the sake of brevity). It is often said that lexemes are abstract
units which are "realized" in texts by concrete word forms - but both ~exe mes and word
forms (or grammatical words) are elements of the system as well as the text, and all
linguistic units are abstractions (see section 2.3 on the relationship between morpheme
and morph). As defined here, the concept of lexeme is not confined to the "major lexical
categories'', the word classes noun, verb and adjective/adverb (as in Aronoff 1994: 910). It should be noted that words of minor categories, e.g., determiners or prepositions,
may be inflected, requiring a distinction between lexeme and grammatical word. And
just as the notion of morpheme is motivated by the existence of variation but can be
extended to cases of non-variation (see section 2.5 on allomorph), so too can we allow
a lexeme to consist of a single grammatical word so that the notion is applicable to all
words.
The grammatical words which belong to a given lexeme form a pattern that is defined
by the morphosyntactic categories and properties involved and can be represented as a
multi-dimensional array. The Lower Serbian nouns DUB 'oak' and ZONA 'wife', for
instance, are inflected for two categories, number and case, and their grammatical words
form a 3 x 6 grid:
(5)
SINGULAR
DUAL
PLURAL
NOMINATIVE
dub
ion-a
dub-a
ion-je
dub-y
ion-y
GENITIVE
dub-a
ion-y
dub-owu
ion-owu
dub-ow
ion-ow
DATl VE
dub-oju
ion-je
dub-om a
i on-oma
dub-am
ion-am
ACC USATIVE
dub
ion-u
dub-a
ion-je
dub-y
ion-y
INSTRUMENTAL
dub-om
ion-u
dub-oma
ion-oma
dub-ami
ion-ami
LOCATIVE
dub-je
ion-je
dub-om a
ion-oma
dub-ach
ion-ach
The term paradigm can refer not only to such an array filled by the grammatical words
of a specific lexeme (e.g., the paradigm of ous) but also to an array filled by the markers
that express these morphosyntactic properties in a particular inflection class (e.g., the
paradigm of one subset of noun lexemes that includes ZONA, characterized by -a in the
nominative singular, -y in the genitive singular, etc.) or an array that is. empty except for
PA255
the morphosyntactic properties that define the cells (e.g., the Lower Sorbian noun paradigm; cf. Carstairs-McCarthy 2000: 596- 597, where it is recommended that paradigm
should be used only in this last sense).
There is a further morphosyntactic category that is relevant to a Lower Sorbian noun
lexeme, that of gender. But in contrast to number and case, for which its grammatical
words vary, each noun has an invariable gender property that is specified in the lexicon;
DUB is masculine, ZONA feminine. Adjectives, however, are inflected for gender in agreement with the noun they mod ify, e.g., strawy dub 'healthy oak' vs. strowa iona 'healthy
wife'. Using the feature/value terminology, we can thus distinguish between fixed and
variable features of a lexeme. As regards the latter, the values of some features are
selected for semantic reasons, e.g., number, which generally corresponds to the quantity
of entities the speaker wants to refer to. Others are dictated by syntax, especially government and agreement; in this case, the feature is called imposed (Zwicky 1986: 85) or
contextual (Booij 1994: 28, 1996: 2-3). Inherent often serves as a cover term for fixed
and selected features (Zwicky 1986: 85; Kibort 20 l 0: 73-77) or refers to selected features only (Booij 1994: 28, misquoting Anderson 1988: 25; Booij 2012: 103- 104). This
is a rather unfortunate terminological choice because inherent suggests invariance (and
is, in fact, used for fixed features by some authors, e.g., Anderson 1988: 25). The hypothesis that selected and imposed features behave differently w ith regard to word-formation
needs to be investigated further (see section 7.2).
Among the grammatical words of a lexeme, one is the citation form that is conventionally used to refer to the lexeme as a whole. For nouns inflected for case and number,
it is normally the nominative singular, which serves to name or label entities and is
semantically unmarked (least specific), e.g., dub and iona in our Lower Serbian examples. The criterion of semantic markedness is less helpful for verbs, where typical citation
forms are the infinitive (e.g., in Russian), the first person singular present (e.g., in Latin)
or the third person singular present (e.g., in Macedonian). Frequently, the citation form
is also a base form in a morphological sense, i.e. one that is morphologically unmarked
and can be regarded as underlying the entire paradigm; this is true of dub (which does
not have a suffix) but not of iona (which has one). A slightly different notion is that of
diagnostic forms, traditionally known as principal parts. They are a subset of the grammatical words of a lexeme from which the rest of the lexeme's paradigm can be deduced
(cf. Wurzel 1989: 112- 121). It should be remembered that a lexeme is not identical with
its citation form; this is particularly important when describing derivational relationships.
For example, the Italian lexeme SICUREZZA 'security' is not derived from the citation
form sicuro of the lexeme SICURO 'secure'; hence, the suffix -o, which is inflectional,
need not be deleted before -ezza can be added (see also section 7.2 on the input to wordforrnation rules). It is also incorrect to identify the lexeme with the "root or underlying
form" (Dixon and Aikhenwald 2002: 6-7) or rather the stern, which is, roughly speaking,
that part which all the grammatical words of a lexeme share, e.g., dub and ion- (see
sections 4.1-4.3).
A fundamental question that has major repercussions for word-formation is what
exactly should be included in an inflectional paradigm (see also article 14 on the delimitation of derivation and inflection). One problem can be illustrated by the following
partial paradigm of the Latin verb ARO 'to plow' (to simplify matters, only 3rd person
present indicative forms are shown):
255
PA256
256

(6)
ACTIVE
IMPERFECTIVE
a rat
'plows'
PERFECTIVE
aravit
'plowed'
PASSIVE
aratur 'is [being] plowed'
Latin verb lexemes do not have a grammatical word with the properties 'PERFECTIVE'
and 'PASSIVE'; instead, 'has been plowed' must be expressed by the construction aratus
est. For the sake of symmetry, this so-called periphrastic form is often included in the
paradigm, but the definition of lexeme given here does not allow such a solution (for
some discussion of periphrasis and paradigms, cf. Haspelmath 2000). The participle
aratus 'plowed ' raises a second question. Like a Latin adjective, it is inflected for gender,
number and case and can be used attributively, e.g., ager aratus 'plowed field' (nominative singular masculine), tenarum aratarum 'of plowed lands' (genitive plural feminine),
etc. Do all these grammatical words belong to the paradigm of the verb ARO (cf. Haspelmath 1996 for a defence of "word-class changing inflection")? Or should we maintain
that a verb paradigm cannot contain an adjective paradigm because inflection can never
change the word class (cf. Scalise 1984: 103)? If so, the formation of participles (and,
for that matter, noun-like infinitives) must be derivation despite its regularity - a conclusion that is somewhat unusual but by no means unreasonable (cf. Plungj an 2003: 122;
see also Mo,tsch 1999: 21, 312- 313). There is a problem of circularity here, however: If
lexeme classes, i.e. morphological as opposed to syntactic word classes (cf. Bergenholtz
and Mugdan 1979a: 137- 141), are defined by paradigms, we cannot distinguish inflection from derivation in terms of preservation vs. change of word class..
4.
Root, stem, affix
4.1. Root vs. affix

Minimal signs fall into two classes which are relevant to the definition of compounding
or composition (see article 20) as opposed to derivation (see article 16), namely root
(German Wurzel, French radical, Russian koren', etc.; the concept was borrowed from
Hebrew linguistics, cf. Lindner 2012: 144- 146) and affix. For example, book-s 'bookPL' is generally considered to consist of a root book and an affix -s. It is, however, not
so easy to name the criteria on which the distinction is based.
According to one explanation, roots are free and affixes are bound. Here, free does
not mean "can be uttered by itself' (as in the definition of the word as a "minimum free
form", see section 3 .3 .1) but "can occur as a word by itself'; by contrast, a bound
minimal sign cannot constitute a word except in conjunction with another sign. It is
therefore more accurate to speak of potentially free and obligatorily bound. This criterion
may work reasonably well for English, but the Lower Sorbian example (5) shows that
the matter is not so simple. While dub is potentially free (since it forms a word by itself
in the nominative and accusative singular), there is no cell in the paradigm of ZONA in
which ion- appears alone. Even dub is not free if we assume zero affixes in the nominative and accusative singular (cf. Mel'cuk 1982: 75; see also section 5.9). How then can
PA257
we uphold the tradi tional view that both dub and ion- are roots? One option is to accept
that while potentially free minimal signs are always roots, obligatory bound ones can be
either roots or affixes, but this requires additional criteria. A lternatively, we can classify
a minimal sign as free even if it must always be combined with some inflectional affix
(cf. Donalies 2005: 20).
According to a related distributional criterion, roots are obligatory components of a
word but affixes are not (cf. Vardul' 1991: 29). There are, however, a number of potential
counter-examples to the assumption that every word contains at least one root (cf. also
D ixon and Aikhenvald 2002: 22). They include Hebrew /b:::>xa/ 'in you', where a prefix
seems to be combined with a suffix, cf. /b:::>-fir/ ' in-song', /fir-xa/ 'song-Poss.2sG (your
song)', and Hungarian naiad 'with/near you', where the suffix -nal 'ADESSIVE' seems to
be combined with the suffix -ad 'poss.2sG', cf. haz-nal ' house-ADESSNE (at [a] house)',
haz-ad 'house-Poss.2sG (your house)'. We might want to say that nal is a root in nil.lad
but an affix in hazna/ (for examples from other languages, some of which may be open
to other analyses, cf. Volodin and Chrakovskij 1991: 110; see also section 3.2.3 on
"separable prefixes") - but there are linguists who exclude such a possibility on principle
(e.g., Vardul ' 199'1: 3 1).
Another distinction that is frequently appealed to is that between lexical meaning and
grammatical meaning. It is based on a variety of factors (cf. Croft 2000: 258- 260);
putting it simply, lexical meanings have something to do with entities, properties and
states of affairs in the extra-linguistic world, and grammatical meanings are relational or
structural. This can also be expressed by the terms autosemantic (having meaning by
itself) and synsemantic (having meaning only in conjunction w ith another item), whereas
the characterization of lexical meanings as concrete and of grammatical meanings as
abstract is too vague to be helpful. A definition of root and affix via lexical vs. grammatical meaning is, however, somewhat problematic. One difficulty is that there are many
meanings that can be expressed - in different languages or even within the same language - either by a root or an affix as defined on distributional grounds (cf. Volodin
and Chrakovskij 1991: 109; Croft 2000: 258). Secondly, it is unclear how to classify
certain kinds of meanings in a given language. Derivational elements are by definition
affixes (cf. Booij 2005: 109) but their meanings are sometimes regarded as grammatical
(e.g., Fleischer and Barz 201 2: 63), sometimes as lexical (e.g., Fleischer and Barz 1992:
26). Conversely, some potentially free minimal signs have a meaning that is evidently
grammatical, e.g., uninflected particles such as Engl ish with, because, etc. and inflected
determiners such as Slovene moj 'my(NOM.SG.M)' I moj-ega ' my-GEN.SG.M' , etc. (as in
moj moi 'my husband', mojega moia 'of my husband ' ). If we nonetheless want to
uphold the definition of roots as minimal signs w ith lexical meaning, we have to give
up the simple bipartition root/affix and introduce additional classes of minimal signs (cf.
Bergenholtz and Mugdan l 979a: 119- 121 ).
Some linguists feel that the term root should be reserved for a diachronic concept
(cf. Chelliah and de Reuse 20 11: 3 14) or avoid it for some other reason. Among the
alternatives are base morpheme (German Grundmorphem or Basismorphem) and core
morpheme (German Kernmorphem). Certain definitions of root are, in fact, definitions
of what is norma lly called stem (see section 4.3; cf. Stump 2001 a: 33; Aronoff 201 2:
29-30; see also Matthews 1974: 40).
257
PA258
258
4.2. Neither root nor affix?

One response to delimitation problems is to abandon an exhaustive subdivision into
mutually exclusive classes in favour of prototypes which are characterized by a certain
bundle of properties. Items can now be assigned to points on a continuum between the
prototypes, based on their similarity to the latter. These points need not include the
prototypes themselves, which then remain ideals that are never attained in practice. In
our case, we can describe minimal signs as more root-like or more affix-like, and we
may or may not find elements that have all the properties of prototypical roots or prototypical affixes. Such a gradual distinction is, of course, more accurate but may not be
very useful for practical purposes. Another solution is to introduce an intermediate category, e.g., semi-affix, between root and affix - but this means that instead of one
delimitation problem, root vs. affix, we now have two, root vs. semi-affix and semi-affix
vs. affix. For this reason, many linguists are reluctant to increase the number of categories although they may use a special label for a group of problematic cases (see also
section 3.2.2 on clitics as an additional category between affixes and particles).
The terms affixoid or semi-affix (and the more specific prefixoid, suffixoid, etc., see
section 5.2) are commonly applied to an element that poses a semantic problem: Although it seems to be identical with a free minimal sign, it does not have the same
meaning (cf. Fleischer and Barz 2012: 58- 61; Booij 2005: 114- 117). A standard example are German nouns with the component -tag that denote assemblies (as events and/or
institutions), e.g., Juristentag 'lawyer assembly' (an organization of lawyers and its annual meeting) and Bundestag 'federation assembly' (the German parliament). One can
discern some semantic connection with the word Tag 'day', but unlike ordinary compounds such as Herbsttag 'autumn day', these complex words do not fit the pattern 'AB
is a (kind of) B'. Another characteristic of "affixoids" that is regularly mentioned is their
high degree of productivity, but that does not necessarily distinguish them from ordinary
compound members. We can therefore classify them as roots if we allow for the possibility that certain meanings of a root occur only in word-formations (see also section 7.3
on compounding stems). It is well known, however, that roots can develop into affixes,
and an affixoid is sometimes described as an element that may diachronically be in the
process of becoming an affix (e.g., Donalies 2005: 24- 25). The term affixoid has also
been used in a completely different sense, for sound sequences that resemble affixes
"without qualifying for that rank", e.g., -er in hammer (Malkiel 1966: 323).
A second kind of minimal sign, often called confix (for a different use of this term,
see section 5.1 ), poses a formal problem: It has lexical meaning but is not free (in the
broader sense that it does not constitute a word by itself or in combination with an
inflectional affix), occurring only in word-formations. The stock example are elements
of Greek or Latin origin such as phob(e) in phobia and xenophobe, but the term confix
has also been applied to native elements like step- in stepmother. Typically, the neoclassical minimal signs were free in the donor language but were borrowed only in
combinations with other signs and later recombined into new words that did not exist in
the classical languages. In many (mostly lndo-European) languages of Europe, neoclassical word-formation constitutes a special subsystem that is largely independent of
these languages (cf. Montermini 2008: 31- 32; see also article 91 on word-formation in
Neo-Latin and articles 93- 96 on foreign word-formation in German, English, Italian and
PA259
Polish). This exceptional situation hardly justifies the introduction of novel units; despite
certain difficulties, a division into roots and affixes seems sufficient.
4.3. Stem
The term stem (German Stamm, French theme) is used in slightly different ways in
different traditions (cf. the comparative table in Chelliah and de Reuse 201 1: 314; for
its history, see Lindner 2012: 135- 144). It is frequently defined as what remains when
inflectional affixes are removed from a grammatical word (e.g., Anderson 1992: 71;
Lieber 20 10: 35). Accordingly, the stem of the plural noun books is book because -s 'PL'
is an inflectional affix. But what about the singular noun book where there is nothing to
be removed? It might be clearer to say that a stem is something to which an inflectional
affix can be added (e.g., Stump 200 1a: 33; Bauer 2003: 341). That would mean that
book has a stem because -s can be added, while uninflected words such as because do
not have one. We may therefore want to amend the definition a little (cf. Schpak-Dolt
20 10: 37) so that all words consist of a stem and possibly one or more inflectional
affixes. It should be noted, however, that something to which inflectional affixes can be
added does not necessarily lack inflectional affixes altogether so that the two definitions
are not equivalent (as is sometimes mistakenly assumed, e.g., Donalies 2005: 19- 20). If
we remove all inflectional affixes from the Bashkir word /awrl-dar-l'bl'o-oan/ 'villagePL-1 PL.POSS- ABL (from our villages)', we a1Tive at the stem /awl'l/ 'village' . But if we
consider the signs to which inflectional affixes can be added , then /awl'ldar/ 'villages'
and /awl'ldarrbl'o/ 'our villages' are stems as well, and stem is a recursive category: A
stem can contain a stem. Even if we decide that only something without any inflectional
affixes should be called stem, we may want to say that the derivational affix 1-daf/ 'co-'
turns the stem /awl'l/ 'village' into the stem /awl'ldaf/ 'co-villager (inhabitant of the same
village)', which occurs in /awrldaftarrbl'ooan/ 'from our co-villagers', etc. (see also
section 7.3. on word-formation as stem formation) - and if we insist that a word has
only one stem and assign /awl'l/ to a different category when it occurs in the derivative I
awl'l-daf/, that category will have to be recursive (e.g., Stammteil 'stem part' in Bergenholtz and Mugdan 1979a: 123-125).
In the Lower Sorbian example (5), the stems of D UB 'oak' and ZONA 'wife' can easily
be identified as dub and ion-. Often, it is more difficult to draw the boundary between
stem and affix (cf. Spencer 2012 for some discussion), and the operation "omit all inflectional affixes" need not give the same result for all the grammatical words that belong
to a lexeme. Lower Sorbian KOTEL 'kettle' has the nominative singular kotel, the genitive
singular kotla and the locative singular kotle, and removal of the inflectional affixes, if
any, leads to three different remainders, kotel, kotl- and kotl-. In such a situation, some
say that the lexeme has a single stem with several stem alternants (cf Halle 1953: 46)
or stem forms (cf. Fuhrhop 1998: 22; Aronoff and Fuhrhop 2002: 454-456). Others
speak of different stems (e.g., Stump 2001a: 33); one can then call the set of all stems
and their distribution stem space (cf. Bonami, Boye and Kerleroux 2009: 106- 108). It
may be desirable to distinguish multiple stems from mere variants of a stem, for which
the term allostem suggests itself (which has, however, also been used for stem alternants
of all kinds) so that a lexeme can have several stems which can in turn have several
259
PA260
260

allostems (cf. Spencer 20 12: 98). What exactly should count as an allostem needs to be
clarified, but the extent to which phonological factors are involved (see section 2.4 on
conditioning, relatedness and domain) certainly plays a decisive role. In the following,
I will use stem alternant as a neutral term where I do not want to commit myself as to
whether we are dealing with multiple stems or allostems of a single stem.
A typology of stem alternations will also need to consider the information content
and the distribution of the alternants. The following partial paradigms of the cognate
verbs VARFA 'to throw' in the Bosco Gurin dialect of Walser German (Ticino, Switzerland) and WERFEN 'to throw' in Standard German serve to illustrate these points:
(7)
Bosco Gurin
! SG
2SG
3SG
I PL
2PL
3PL
Standard German
PRESENT
PRES ENT
PAST
varf-a
verf-ft
verf-t
vuf-~
v1rf-st
v1rf-t
varf
varf-st
varf
vuf-~n
varf-~n
vuf-t
varf-t
vuf-~n
varf-~n
varf-a
varf-at
varf-an
In Bosco Gurin, certain verbs have a special stem altemant with the vowe l le/ instead
of /a/ in the second and third person singular, but the affixes I-ft/ and I-ti are unambiguous exponents of these morphosyntactic properties so that the information provided by
the stem alternation is redundant. Standard German has a similar alternation between
11 and hi, but here the third person singular and the second person plural, /virf-t/ and
/vcrf-t/, have the same affix /-t/ and are distinguished only by the stem alternants. Nonetheless, we may still regard this as allo-variation, i.e. as a side effect without meaning,
because the vast majority of Standard German verbs have the same word form in these
two cells. By contrast, the main difference between the present and past tense forms
cited lies in the stem so that the content of /varf/ includes the feature [+past]. Such a
stem which encodes one or more morphosyntactic properties in addition to the lexical
meaning may be named portmanteau stem (cf. Stump 200la: 208-211; the hypothesis
that stems never serve as exponents of morphosyntactic properties, cf. Spencer 2012:
89, 98, presupposes a different analysis of /varf/).
With regard to the distribution of stem alternants, the question is whether the cells in
which they occur form a natural class in the sense that they can be characterized by a
morphosyntactic feature or combination of features that is associated with all of these
cells and only these cells. Standard German /virf/ and /varf/ occur in cells that form
natural classes since they can be identified by the features [-me, -pl, -past] and [+past],
respectively. This is also expressed by traditional labels such as past (tense) stem, which
are often extended to morphologically complex stems such as German /Jcrf-t/ 'sharpenPAST' in /fcrf-t-~/ 'sharpen-PAST- l sG ([I] sharpened)', etc. with the regular affix /-t/
'PAST'. For a stem alternant which appears in cells that do not form a natural class, the
term morphomic has become fashionable (cf. Aronoff 1994: 25). A we ll-known case are
French verbs of the type PUNIR 'punish'. In the present tense, they have an extended
stem in -iss (e.g., puniss-) in both the indicative and the subjunctive mood except for
PA261
the singular of the indicative - a distribution that cannot be captured by any feature
specification. Another stock example is the Latin "third stem" or "participial stem" (see
article 44 on paradigmatically determined allomorphy). But there is often no need to
resort to morphomic stems if morphosyntactic properties are not taken to be full specifications of meanings and functions (see section 2.3) or if derivational processes are allowed to delete or change morphosyntactic properties (cf. Mel'cuk 2006: 293- 294).
If a lexeme has several stem alternants, word-formation rules may operate on more
than one of them (see also sections 7.2-7.3). For example, French punition ' punishment' is derived from the simple stem puni-, punissable 'punishable' from the extended
stem puniss-. It is an open question which kinds of stem alternants may serve as inputs
to derivation or compounding. Furthermore, alternations in derivatives are not always
due to different inputs. Some have a phonological explanation, others result from different affixes having partially the same effect, as in the case of German umlaut (roughly,
vowel fronting): A number of verbs with the basic stem vowel la:/ have le:/ (orthographically a) in the second and third person singular present, e.g., trag-en 'carry-INF' - trag-t
'carry-3sG'. The noun Trag-er 'carry-AGNR (carrier)' might seem to be derived from the
stem alternant trag-, but a comparison withfahr-en 'drive-INF' - fiihr-t - Fahr-er, jag-en
' hunt-INF' - jag-t - Jag-er and /rag-en 'ask-INF' - frag-t - Frag-er shows that the
occurrence of le:! in the derived noun is independent of its occurrence in the verb forms
(cf. also Eschenlohr 1999: 89-96 on umlaut in German derived verbs).
4.4. Templatic and hierarchical word structu re

Words can have numerous affixes, and there are essentially two ways of accounting for
the possible combinations: templates and hierarchical structures. (Add itional constraints
that have been suggested, e.g., that affixes may belong to mutually exclusive levels, see
article 7 on word-formation in generative grammar, section 4.2, or that certain affixes
prevent further affixation, see article 54 on closing suffixes, will not be considered here.)
A template consists of a series of slots (also known as position classes) for the root or
stem and the affixes that can precede or follow it (cf. Stump 2006; Rice 2011: 188193 ). In each of the affix slots, there is a choice between a limited number of mutually
exclusive affixes. Some of these slots must be filled while others may remain empty.
There may be certain co-occurrence restrictions, e.g., "if slot i is filled, slot j must be
filled as well" or "if slot i is filled with item x, slot j may not be filled with item y", but
by and large the slots are filled quasi-simultaneously and each affix makes its own
contribution to the content of the whole word independently of the others. Sometimes,
the order of the affixes in a template may seem arbitrary (e.g., when derivational and
inflectional affixes are not neatly separated), but it often follows from some more general
principles (e.g., "inflection comes outside of derivation", see section 7.2 and article 14
on the delimitation of derivation and inflection).
Hierarchical structures can be generated by rewrite rules and are represented by tree
diagrams or bracketings as in syntax. Here, affixes are added in a certain order and
typically have semantic scope over the structure they attach to: If we start from roll and
first add -able and then un-, we arrive at [un[rollable]] 'not able to be rolled', but if we
attach the affixes in the opposite order, the result is [[unroll]able ] 'able to be unrolled'
261
PA262
262
(see also section 3.3.3 on Dutch onverzekerd vs. veronzekerd). The principle that all trees
are binary branching is usually taken for granted; cases where three or more constituents
seem to be of equal rank (e.g., blue, white and red in the blue-white-red flag) are fairly
rare and their analysis is controversial. A question which has played a major role in
debates about certain types of word-formations is whether all the intermediate steps in
a hierarchical structure have to be actual words. A well-known case are English and
German verbs of the type endanger 'to expose to danger ' and bestuhl[en] ' to furnish
with chairs' (+----- Stuhl 'chair '; the bracketed -en is the infinitive suffix of the citation
form), where some linguists would rather assume the nonexisting verbs *danger and
*stuhl[en] as intermediate steps than allow prefixes to change word class (cf. Eschenlohr
1999: 106- 108; Donalies 2005: 120- 121 ).
Templates may be more suitable for some morphological systems or subsystems than
for others. In general, templatic structures are typical of inflection and some highly
constrained patterns of derivation, whereas most word-formation creates hierarchical
structures. Therefore, the two models can often be combined so that complex stems are
built up hierarch ically but the combination of stems with inflectional affixes conforms
to a template.
5. Types of affixes
5.1. Criteria
Types of affixes are frequently defined in terms of their position relative to a root, but
affixes can also attach to more complex structures. Depending on the exact definition of
stem (see section 4.3), it may be more correct to describe their positions with respect to
a stem (cf. Stump 2001 b: 708), but it is best to use the neutral term base for whatever
it is that the affix attaches to (cf. Matthews 1991: 13 1; Bauer 2004: 21 ). We can then
classify affixes according to whether or not they disrupt the base, making it discontinuous, and whether or not they themselves are di srupted (cf. Mel'cuk 1963: 34- 37, 1982:
82- 87, 1997: 147- 177, 2006: 299- 300, with further examples). This gives the four types
in (8); affixes that are neither disrupting nor disrupted are subclass ified by position,
before the base (preflX), after the base (suffix) or between two bases (interfix).
(8)
continuous
discontinuous
continuous
prefix, suffix , interfix
infix
discontinuous
circumfix
transfix
There is no generally accepted superordinate term for prefixes and suffixes (and interfixes, which are often not taken into account). Confix, which was proposed in connection
with this classification, is also used in another meaning (see section 4.2). Adjix has the
advantage of being parallel to adposition, the generally accepted cover term for prepositions and postpositions, but it occurs primarily in contradistinction to infix, and it is not
always clear whether it includes only prefixes and suffixes or also in terfixes and/or
PA263
circumfixes. Exfix (Marusic 2003 : 1) and extrafix (cf. Hoeksema and Janda 1988: 204)
are rare and have both been used for other purposes as well.
Another classification recognizes prefixes, suffixes and infixes as basic types and
describes discontinuous affixes as combinations of these: A circumfix is a combination
of a prefix and a suffix, whereas transfix, if defined as in (8), covers any combination
of an infix with one or more other affixes of any type, prefix, suffix or infix. It is only
if transfix is understood more natTowly (cf. section 5.5) that circumfix and transfix leave
a "descriptive gap" (Bauer 1988: 17) that could be filled by additional terms, e.g. ,parafix
for infix plus prefix or infix plus suffix (Blevins 1999: 383-384). A number of cover
tenns for affix combinations in general have been suggested, including polyajfix (Plungjan 2003: 95- 97) and simulfix (Bickel and Nichols 2007: 200), both of which are also
attested in other senses. Synaffix has been proposed for combinations which involve
affixes and/or internal modification (Bauer 1988: 23), and multifixation for morphological processes of any kind (cf. section 6) that operate simultaneously (Hoeksema and
Janda 1988: 216). If such a cover term is needed at all (see section 6.6 for approaches
to multiple marking), multifix would seem the best choice; polyaffix and synaffix are
rather odd word-formations.
5.2. Prefix and suffix

The terms prefix and suffix were coined around 1600 (for details, cf. Lindner 2012: 14614 7) and are firmly established, notwithstanding the fact that the opposite of pre- 'before'
is post- 'after' and not sub- 'under' (cf. preposition vs. postposition). Some linguists do
use postfix instead of suffix; others, especially Slavicists, employ postfix for a derivational suffix that follows inflectional suffixes (cf. Haspelmath 1990: 63, where antefix is
suggested for a pre-inflectional prefix and extrafix as a cover term for affixes outside of
inflectional affixes). In some languages, native equivalents of prefix and suffix (e.g.,
Dutch voorvoegsel/achtervoegsel, Icelandic forskeyti/vioskeyti, Polish przedrostek/przyrostek, Bulgarian predstavka/nastavka) are quite common. German Vorsilbe and Nachsilbe are based on Silbe 'syllable', which is highly misleading because affixes do not need
to coincide with syllables. English ending, German Endung, French desinence, Russian
okoncanie, etc. are often used instead of suffix, but these terms can also refer to a
phonological string at the end of a word regardless of its morphological status (cf. also
Lindner 20 12: 138). Some distinguish between suffix as derivational and ending as inflectional, and others use ending, etc. for a sequence of one or more inflectional suffixes
(Schpak-Dolt 2010: 36-37; cf. Mel'cuk 1997: 147; see also Aronoff 201 2: 48 for the
unsubstantiated claim that ending refers specifically to the last inflectional element in a
word).
Cross-linguistically, prefixes are far less common than suffixes (cf. Stump 200 1b:
709-7 11 ; Montermini 2008: 49-59). This holds true of Europe, too, where a number of
languages are exclusively or almost exclusively suffixing. In this region, prefixes do,
however, often play a significant role in derivational morphology, as a rule without a
change in word class (see also section 7.4). The verb systems of Indo-European languages as well as Hungarian and many Caucasian languages make extensive use of
prefixes (known as preverbs in some grammatical traditions). They often signal local
263
PA264
264
or directional meanings or metaphoric extensions of these, but numerous prefix-stem

combinations have a meaning that is not predictable from the parts. In a number of
languages, prefixes can derive verbs with a specific aktionsart (e.g., Germanic) or aspect
(e.g., Slavic). Prefixes also occur on nouns and adjectives, where they express a wide
variety of meanings (e.g., ex-president, co-worker, postmodern, decentral, etc.); one frequent function is the formation of antonyms (e.g., Albanian pa-barazi 'in-equality',
Farnese 6-vanligur 'un-usual', Welsh an-aml ' in-frequent'). There are other meanings
for which suffixes are clearly preferred, among them diminution (e.g., West Frisian tim-tsje
'garden-DIM', Romanian mic-uf 'small-DIM'). Such correlations between different derivational meanings and prefixing vs. suffixing need to be explored further.
5.3. Circumfix
A circumfix consists of two parts, one that precedes the base and one that follows it, as
in Georgian sidide 'size' from did-i 'large-NoM'. (According to the Leipzig Glossing
Rules, one should write si-did-e 'NR-large-NR', where the repetition of the gloss NR,
short for NOMINALIZER, indicates the circumfix. A more convincing notation, which is
more in line with that for infixes, is si)di(de '(NR)large' with the gloss NR arbitrarily
given for the prefixal part only, cf. Lehmann 2004: 1853.) Less common terms are
simulfix, confix, ambifix and parasynthesis (cf. Hagege 1986: 26; Vardul ' 1991: 54;
Greenberg 1966: 92; Malkiel 1966: 314; see also article 29 on parasynthesis in Romance.)
How can we distinguish this type of affix from a situation where a prefix and a suffix
are attached in one step (or even in two)? According to one definition, neither of the
members of a circumfix "ever appears alone" (Rubba 2001: 679), which should presumably be taken to mean that neither of them occurs elsewhere with a meaning that one
could also attribute to it in the combination. Another definition requires that the parts
"together have a single unitary meaning" and that they "must occur together to provide
that meaning" (Bauer 2004: 29). Such conditions exclude a number of standard examples, e.g., that of the German past participle: In gefragt 'ask.PST.PTCP' (cf. frag-en 'asklNF'), ge- ... -t is often said to be a circumfix. However, ge- can also combine w ith -en
in some irregular verbs, e.g., getragen 'carry.PST.PTCP' (cf. trag-en 'carry-INF'), and verb
stems with an unstressed first syllable do not have ge-, e.g., erfrag-t 'ascertain-PST.PTCP'
and ertrag-en 'bear-PST.PTCP' (cf. erfrag-en 'ascertain-INF', ertrag-en 'bear-INF'). This
shows that the prefixal and suffixal parts are independent of each other. The main reason
for positing a circumfix seems to be that we cannot easily assign any meaning to ge-,
but this does not preclude an interpretation of ge- ... -t as a prefix-suffix combination (cf.
Wiese l 996a: 89-98). Sometimes, a more sophisticated analysis allows us to determine
the distinct functions of the prefixal and suffixal parts even though they do not have a
"single meaning" in the conventional sense (cf. Anderson 1992: 13 7-145 on agreement
markers in Georgian).
In derivational morphology, the assumption of circumfixes may rest on the premise
that all intermediate steps must be actual words (see section 4.4; cf. Lieber 2010: 78).
Thus, one could argue that the Russian verb doidat'sja 'to wait until [the desired result]'
is derived from idat' 'to wait' with a circumfix do- ... -sja because neither *do-idat' nor
PA265
*zdat'-sja exists (Brown 2011: 494- 495). But then again, dokatit'sja 'to roll (intr.) until
[the goal]' must be derived from katit' 'to roll (tr.)' with a prefix do- 'until' and a suffix
-sja 'REFLEXIVE' because dokatit' 'to roll (tr.) until [the goal]' and katit'sja 'to roll (intr.)'
are both in the Russian lexicon. If derivations via potential but unattested words are
permitted, this discrepancy disappears and there is no need for a circumfix. Another
fundamental decision affects the interpretation of a type of word-formation exemplified
by the Slovak adjective pod-koi-n-y 'under-skin-ADJECTIVIZER-NOM.SG.M (subcutaneous)' and the Serbo-Croat noun do-vrat-ak 'next_to-door-NOMTNALlZER.NOM.SG (doorpost)'. If phrases are allowed to be inputs to word-formation rules (see section 7.1), a
derivation from the prepositional phrases pod koi-ou 'under [the] skin-TNS.SG' and do
vrat-a 'next_to [the] door-GEN.PL' is plausible; otherwise, the circumfixes pod- ... -n and
do- ... -ak are an option (see article 165 on Serbian, section 4.1.3).
When derivational patterns are stated in terms of citation forms, it must be remembered that the inflectional affixes of these forms are not part of the pattern (see also
section 3.4). For example, Belarusian has a verb uprygoiye' 'to beautify' from prygoi-y 'beautiful-NOM.SG.M' . Here, the infinitive suffix -ye' must not be mistaken for the
suffixal part of a derivational circumfix u- ... -ye'. It is equally unsatisfactory to set up a
circumfix u- ... -y if the infinitive suffix is taken to consist of -e' only. Rather, the only
overt affix involved is a prefix, irrespective of how one accounts for the fact that the
end result is a verb which belongs to the inflection class with the infinitive suffix -ye'
or the theme vowel y (on this issue, see section 4.4 on English endanger, etc.; cf. also
Schpak-Dolt 2010: 129-134 for a discussion of various analyses of French examples
that leads to a different conclusion).
Some linguists believe that all alleged circumfixes can be reanalyzed as affix combinations (e.g., Vardul' 1991: 54- 55), whi le others explicitly prefer one discontinuous
affix to two continuous ones (e.g., Mel'cuk 2006: 3 16). Even according to the second
opinion, convincing examples of derivational circumfixes are hard to find in the languages of Europe, and it is therefore impossible to make any general statements about
their functions.
5.4. Infix
Infixes are far less common than prefixes and suffixes and it is a generally accepted
universal that a language does not have them unless it also has prefixes and/or suffixes
(cf. Greenberg 1966: 92). Europe is one of the regions where infixes are very rare.
English "expletive infixation" as in fan-blooming-tastie is often cited, but it is more of
a language game than a word-formation pattern - insertion of an entire word is not a
normal form of infixation. A language with several typical derivational infixes is Khmer
(Cambodia). One of them, -n-, is inserted after the first consonant of the base and serves,
among other things, as an instrument nominalizer, e.g., /koah/ 'to raise' --7 /k(n)oah/
' (INsTNR)raise (lever)', /re~rJ/ 'to obstruct' --7 /r(n)e~rJ/ '(INsTNR)obstruct (screen, curtain)'. (In accordance with the Leipzig Glossing Rules, infixes are enclosed in angle
brackets ( ... ); the gloss 'INSTNR' appears at the edge near which the infix is located,
hence before 'raise' and 'obstruct'.)
Contrary to the definition in (8), the term infix is sometimes used for an affix which
occurs closer to the root than others. In some descriptions of Swahili (East Africa), for
265
PA267
5.5. Transfix
The tenn transfix is not very common and typically refers to vowel patterns in Semitic
languages that combine with a root consisting solely of consonants to form a stem with
a particular inflectional or derivational meaning. (Terms that one can occasionally find
instead include intercalated affix and - contrary to normal usage - interfix; cf. Stump
2001 b: 708; Ussishkin 2007: 463- 465.) The Modern Hebrew root /g-d-1/, for instance,
can be combined with -a-a-, -a-e-, -a-o- and -o-e- to give the verb forms /ga'dal/ '[he]
grew' and /ga'del/ '[he] is growing', the adjective /ga'dol/ 'big' and the noun /'godel/
'size'. In the examples that are normally given, all the vowels occur between the root
consonants, and some definitions of transfix explicitly require this (cf. Helmbrecht and
Lehmann 2008: 282). A vowel preceding the first or following the last root consonant
is, however, sometimes clearly part of the transfix and not a prefix or suffix with an
identifiable meaning, e.g., in Maltese ifqar 'poorer.M.SG' (cf. faqar ' poverty'), where
i--a- marks the comparative. In addition, Semitic transfixes may consist of a single
vowel, i.e. strictly speaking an infix, e.g., in fqir 'poor.M.SG'. Since a transfix -@-i- with
a zero vowel (cf. Mel 'cuk 1997: 175) is not to everyone's liking, it may be advisable to
use a special term for the Semitic vowel patterns, e.g., diffix, leaving transfix for other
affixes that are both disrupted and disrupting (cf. Plungjan 2003: 95). The verb forms
ftaqar ' [he] became poor' and faqqar ' [he] made poor ' pose a further problem: Should
the infixed t in ftaqar and the gemination of the second root consonant in faqqar be
regarded as part of the transfix? While this is a debatable point, it is clearly not appropri ate to decompose ftaqartu ' [you] became poor' into fq-r and a transfix -ta-a-tu (cf.
Bossong 2001: 667); rather, the form contains a transfix -ta-a- (or a transfix -a-a- as
well as an infix -t-) and the suffix -tu '2.PL'. Transfixes must also be distinguished from
an intercalation of affix and base that results from phonological metathesis. For instance,
the It/ of the Hebrew prefix /hit/ changes places with a sibilant, as in /hiJtadel/ ' [he] tried
hard ' from the root IJ-d-11 vs. /hitgadel/ ' [he] made himself big (boasted)' from /g-d-1/
(for similar examples with suffixes, cf. Sternberger and Bernhardt 1999: 616-618; Paster
2009: 32).
Transfixes are not a generally recognized type of affix because several other analyses
are possible (cf. Broselow 2000; Rubba 2001 ). Instead of starting from a consonantal
root, one could take a specific stem with vowels as basic, e.g., Hebrew /ga'dal/ '[he]
grew', and derive the others, e.g., /ga'del/ ' [he] is growing' and /gi'del/ ' [he] raised',
from it by changing the vowels (see section 6.3; cf. also Ussishkin 2007: 464- 468 for a
slightly different approach). This may be the reason why transfixation has also been
used for "an alternation at multiple sites throughout the domain" that affects all eligible
segments, as in the Basque affective diminutive /po,\ica/ from /polita/ ' pretty', which
involves palatalization of all coronals (Bye and Svenonius 2012: 431; see article 28 on
affective palatalization in Basque) - but this is not an instance of a transfix as defined
in (8). Another approach, that of autosegmental morphology, holds that in a given type
of inflected or derived Semitic word both the consonantal root and the vowel pattern
must be linked to a template, a skeleton of slots for vowels and consonants, possibly with
some pre-specified segments. In this model, Maltese ftaqar and faqqar must conform to
the templates CtVCVC and CVCCVC, respectively; the -t- in ftaqar is simply part of
the template and gemination of q infaqqar is the answer to the question of how to match
up three root consonants with four C slots.
267
PA268
268

Transfixes play a major role in Maltese, the only Semitic language of Europe (if we
exclude Hebrew and Arabic as the languages of liturgy and study in Judaism and Islam,
respectively). I am not aware of other European languages with discontinuous affixes
that disrupt the base.
5.6. Am bifix
The term ambi.fix, which is sometimes used fo r a circumfix (cf. Greenberg 1966: 92),
normally refers to an affix which can be either prefixed or suffi xed. Such an element
has also been called variable-direction affix (Ussishkin 2007: 460), mobile affix (Paster
2009: 34) or, in German, Wechselaffix (Bossong 2001: 667). The position of an ambifix
may vary freely or it may be determined by phonological, morphological or syntactic
conditions (for examples, see Mel 'cuk 1997: 162- 163; Marusic 2003: 4- 9; Paster 2009:
34-36; Rice 2011: 177- 178); the existence of phonologically conditioned affix placement is, however, a matter of dispute (see also section 3.3.3). One of the few ambifixes
in a European language is the Lithuanian reflexive marker si, which appears as a final
suffix if the stem is not prefixed but as a prefix immediately before the root if the
stem is prefixed, e.g., imperfective perk-a-si 'buy-3sG-REFL ([he] buys for himself)' vs.
perfective nu-si-perk-a 'PERF-REFL-buy-3sG'. Affixes may also occur as prefix and infix
or as infix and suffix or even in all three positions (for some examples, cf. Blevins 1999;
Marusic 2003: 9- 12; Paster 2009: 20; Rice 2011: 177-178; Bye and Svenonius 2012:
470, 4 75-476); one European instance of three-way variation, that of gender agreement
markers in some Northeast Caucasian languages (see articles 202 and 207 on Dargwa
and Khwarshi), is inflectional.
5.7. lnterfix
According to one definition, an inter.fix is an affix that is placed between two roots which
together form a compound (Mel'cuk 1982: 86, 1997: 150, 157, 2006: 298-300), but in
some of the examples given it actually occurs after a derivational suffix and/or before
a derivational prefix. In Russian social'n-o-ekonomiceskij 'social-LINK-economic', for
instance, the interfix -o- comes after a derivational suffix, the adjectivizer -n. The defini tion must therefore be amended to "between two stems". It is often stressed that an
interfix has no meaning in the conventional sense. Although this is sometimes viewed
as a defining property (e.g., Haspelmath and Sims 20 10: 332), it is not a logical necessity
(cf. Mel'cuk 1997: 150). Whether or not "meaningless" elements are signs and belong
to a morpheme is an entirely different issue (pa ce Helmbrecht and Lehmann 2008: 279280). If they do not qualify as signs, not all words can be exhaustively decomposed into
signs (or morphemes) or into roots and affixes. This consequence can be avoided by
recognizing an interfix as an empty sign (one whose meaning is an empty set, cf. Mel' cuk
1982: 51 - 52, 1997: 15- 21) or by regarding the function of indicating the connection
between the two stems as its content, as reflected in the gloss 'LINK' (cf. Bergenholtz
and Mugdan 1979a: 71). Originally, interjix was suggested for an empty sign between a
stem and a derivation al affix (Lausberg 1953: 229), as in the Spanish diminutive avian-
PA269
c-ito 'airplane-LINK-DIM', which has an additional -c- not present in reloj-ito 'watchDIM'. This usage is still widespread in Romance linguistics (see article 31 on interfixes
in Romance), but according to the terminology adopted here, an element such as -c- is
a suffix (if it is recognized as a separate affix at all).
Even after the introduction of interfix, empty signs in both compounding and derivation are still occasionally called infix, but this usage has been rightly criticized. The
tenns linking element or linker (German Fugenelement, also Fugenzeichen 'hnking sign',
Fugenmorphem 'linking morpheme' or simply Fuge) frequently serve as synonyms of
inter.fix as defined here, but some authors who employ linking element for empty signs
in compounding use inter.fix for similar elements in derivation (e.g., Fuhrhop 1998: 24;
cf. also Donalies 2005: 44) or as a cover term for both (e.g., Fleischer and Barz 1992:
32- 33). Occasionally, two or more of the terms interfix, linking element, linking morpheme, etc. are applied to different subtypes of interfixes.
Interfixes frequently go back to inflectional affixes, especially markers of case and/
or number, and are often formally identical with them; the historical developments may,
however, be a good deal more complex (cf. Niibling and Szczepaniak 2013: 68-74 on
German). Some linguists maintain that compounds can indeed contain certain inflectional
affixes, which are then excluded from the category of interfixes (cf. Wiese l 996a: 144145; Donalies 2005: 45- 48 on German), whereas others (including myself) are of the
opinion that the erstwhile inflectional markers no longer have that status synchronically
so that all elements at the boundary between compound members should be treated in
the same way (see section 7.2). An interfix is called paradigmatic if it coincides with
an inflectional affix that occurs in the paradigm of the lexeme in question. In Icelandic,
the interfix -S- in avarp-s-fal/ 'addreSS-LINK-Case(NOM.SG.) (vocative)' is paradigmatic
because AVARP has a genitive singular in -s. In kUpling-s-disk-ur 'clutch-LINK-diskNOM.SG', -s- is non-paradigmatic because KUPLING does not have any word form with
the suffix -s. Some interfixes have no synchronic connection to inflectional affixes, either
because it never existed, as in the case of Greek and Slavic "linking vowels" (cf. Bauer
2000b: 702) or the Iranian "ezafe" (see article 174 on Tat, section 3 .1.1 ), or because the
inflectional affixes they developed from were lost (e.g., the genitive -s in Dutch, cf.
Booij 2002: 34-35, 178-179).
In languages with a variety of interfixes (e.g., many Germanic languages), there are
often no hard and fast rules for their distribution, but one can detect statistical tendencies
that are primarily based on phonological, morphological or semantic properties of the
first compound member. To the extent that interfixes have functions that go beyond
holding the compound together (see article 32 on linking elements in Germanic), they
likewise tend to have more to do with the left than with the right constituent. Phonologically, too, interfixes are more closely connected with the first member of the compound.
In view of these facts, an alternative analysis suggests itself: Linking elements in compounds are not distinct linguistic signs (and hence not affixes) but part of the first stem,
which thus has two or more alternants (cf. also Fuhrhop 1998: 18 7). In the Icelandic
examples, this means that we should segment avarps-fall and kuplings-disk-ur, with
avarps- and kuplings- as alternants of the stems avarp and kUpling - so-called combining
forms or compounding stems (see section 7.2). We could then still describe -s as a linking
element, with the understanding that it is a phonological stem extension, but inter.fix or
linking morpheme would no longer be appropriate. Analogous considerations apply to
an empty suffix before a genuine derivational suffix, e .g., -ul in Romanian frun z-ul-if-a
269
PA270
270

' leaf-LTNK-DJM-SG' fromfrunz-a 'leaf-sG' (vs. gradin-if-a 'garden-DJM-SG' from gradin-
a 'garden-so', without -ul), where we have the additional option of assuming allomorphic variation between -if and -ulif 'oIM' (for arguments in favour of and against the
different solutions, see article 31 on interfixes in Romance; cf. also Malkiel 1966: 3183 19 on Spanish; Wiese 1996a: 100- 103 and Motsch 1999: 14 on German). Instead of
trying to eliminate interfixes by drawing the morphological boundaries differently, some
linguists go in the opposite direction and posit a zero linking element in every compound
that does not have an overt one - but if zero means 'nothing', it means nothing (see
section 5.9).
5.8. Suprafix
Inflectional and derivational meanings can be expressed not only by strings of segmental
phonemes but also by suprasegmental features, especially stress and tone (cf. Hyman
and Leben 2000). A well-known example are the English verb/noun pairs that differ in
their stress pattern: imprint vs. imprint, recoil VS. recoil, etc. Tonal morphology plays
hardly any role in Europe but is very common in Africa. An oft-quoted example is that
of verbs in Ngbaka (DR Congo), which are not specified for tone, e.g., /kpolo/ 'to
return'. Different tone patterns are added to them, as in /kpolO/, /kpolo/, /kpol6/, /kp616/
(where ' marks high tone, - mid tone and ' low tone), to express tense/aspect distinctions
(Nida 1949: 63, 66; this description is not quite correct inasmuch as the tense/aspect
values are co-determined by the tone of the subject pronoun, by a vowel alternation,
here /kpolo/-/kpula/, and by postverbal particles, while it is difficult to assign a unitary
meaning to the tone patterns on the verb; cf. Maes 1959: 13 , 24-25). The term suprajix
was proposed for a suprasegmental affix which is added to a neutral base (Nida 1949:
69; cf. also Matthews 1974: 133). Supeifix is often given as a synonym but originally
referred to the stress pattern of a word considered as a "special kind of affix" by which
a base and its prefixes and suffixes are "combined and unified" (Trager 1948: 157). Of
course, suprasegmental features are not arranged "above" or "below" segmental phonemes but are simultaneous with them so that simuljix might seem a better term (Hockett
1954: 212) - but instead of replacing suprajix, simuljix was put to a variety of other
uses, none of which were ever broadly accepted (see sections 5.1 , 5.3, 6.3).
Although one could conceivably decompose imprint and imprint into a stressless and
word-class neutral stem imprint and a suprafix - ..!.. 'VERB' or ..!.. - 'NOUN', an analysis
that is more in keeping with the overall structure of English takes the verb as basic and
derives the noun by means of a stress shift, i.e. a process and not an affix in the usual
sense (see section 6.3). Some linguists define suprafix or superfix as referring to such a
suprasegmental change (e.g., Bauer 2003: 342, 2004: 98; cf. also Rubba 2001: 680,
where suprafzx is extended to multiple segmental changes). It is not always easy to
decide whether modification or addition of suprasegmental features is involved.
Suprasegmental alternations are not necessarily due to a direct change of features.
Sometimes, it is more plausible to assume an affix with an expression that consists solely
of a "floating" tone which then affects the tones of the base (cf. Hyman and Leben 2000:
589 for examples; see also Akinlabi 20 11: 1945, 1957- 1961). In some languages, stress
placement can be described in terms of stress-relevant properties of the constituents of
PA271
words, and derivational processes can change these properties. One such language is
Abkhaz, where each element of a word (CV, V or C) is either dominant or recessive and
stress falls on the last of an uninterrupted sequence of dominant elements. In the masdar
(verbal noun) /a~a'ra/ 'dawning', the general definite prefix ("article" ) /a-/, the root /~a/
and the masdar suffix /-ra/ are all dominant and therefore /-ra/ is stressed. When an
abstract noun is formed from the masdar, /-ra/ becomes recessive, which has the effect
of moving the stress back: /a'~ara/ 'dawn' (cf. Hewitt 2005: 101, 117; see also artic le
192), but the application of "make /-ra/ recessive" does not result in an overt difference
between the masdar /a'd::ma/ 'knowing' and the abstract noun /a'd::ma/ 'knowledge'.
Here, root-initial /d~/ is dominant and root-final Ir! recessive so that stress must fall on
Id~/ and cannot be affected by the properties of /-ra/ (George Hewitt, p.c.).
5.9. Zero
Returning once again to the inflectional paradigms of Lower Sorbian nouns in (5), we
can observe that all grammatical words of the lexemes DUB and ZONA have a case/
number affix - except for the nominative and accusative singular of DUB. This anomaly
can be rectified if we allow the expression of a linguistic sign to be zero (i .. e. an empty
set; cf. Mel'cuk 1982: 48). We can then postulate the zero affixes -0 'NOM.SG' and
-0 'Acc.sG', which are zero allomorphs of the morphemes {NOM.SG} and {AcC.sG}.
Importantly, these morphemes have non-zero allomorphs as well, e.g., -a 'NOM.SG' and
-u 'ACC.SG' in ion-a and ion-u. English nouns, though, never have an overt affix in the
singular, and if we analyze book as book-0 'book-sG' as opposed to book-s 'book-PL',
the morpheme {so} is a zero morpheme, one that has no non-zero allomorphs. Since the
introduction of zero elements opens a Pandora's box, various principles have been formulated with the aim of preventing an uncontrolled proliferation of zeros, e.g.,
- zero morphemes are not allowed;
- one may not "treat overt distinctions as meaningless and covert distinctions [between
zero and nothing, JM] as meaningful" (Nida 1948: 415);
- zero elements must be recognizable.
Helpful though such guidelines are, each of them excludes some reasonable analyses,
and in the end we have to rely on our good judgment (cf. Bergenholtz and Mugdan
2000: 440- 443; see also Mel'cuk 2006: 469- 516). We should bear in mind, however,
that zero is not jlllst a more fancy word for nothing: A zero affix is a linguistic element
and not the same thing as an unfilled affix slot.
In word-formation, zero elements are much more difficult to justify than in inflection,
and some linguists do not allow derivational zero affixes at all (e.g., Volodin and Chrakovskij 1991: 129; Mel 'cuk 2006: 504-505). Zero derivation, i.e. derivation by means
of a zero affix, is frequently invoked in cases where lexemes of different word classes
(or subclasses, e.g., genders) have homonymous stems, e.g., cook as in please cook the
vegetables and she is a good cook or Latvian direktor-, the stem of a masculine and a
feminine noun which take different case/number suffixes, direktor-s 'headmasterNOM .SG' vs. direktor-e 'headmistress-NOM.SG', etc. If the verb cook is basic, the noun
could be derived from it by means of the affix -0 'AGNR' (short for 'AGENT NOMINALJZ-
271
PA272
272

ER'). A major argument in favour of this zero affix is the existence of overt nominalizers
with the same function, e.g., -er in baker. This approach leads to a multitude of zeros
with different contents (cf. Bergenholtz and Mugdan 2000: 445- 446) and it is not always
clear which overt affix should be taken as the analogue. For example, dloes Dutch zuiver'to make pure' (here shown without the infinitive suffix -en of the citation form) from
zuiver 'pure' have a zero suffix because of blond 'blond' - blondeer- ' to make blond
(bleach)' or a zero prefix because of warm 'warm' - verwarm- 'to make warm' or
perhaps both? There are two major theoretical alternatives to zero derivation of this type:
conversion (see article 17) and multifunctionality. Although often equated with zero
derivation, conversion is by no means the same thing: It is a process that changes syntactic (and possibly also semantic) features and leaves no "visible" traces. Hence, a verb
that was converted from a noun should have the same properties as a primary verb. By
contrast, a zero affix is a linguistic element and should behave like an overt affix, be it
that where overt affixes cannot attach, zero affixes cannot attach either, be it that wordformation processes can distinguish between zero-affixed and primary stems (which does
not seem to be the case; cf. Meder and Mugdan 1990: 100-102; Eschenlohr 1999: 6063, 221 ). Both zero derivation and conversion effect a transposition from one word class
to another (or motion from one gender to another), on the premise that one member of
the pair is basic and the other derived. Multifunctionality, on the other hand, means that
English is assumed to have a single stem cook which can serve as both a noun stem and
a verb stem without any intervening word-fonnation process (see also section 7.4). These
three analyses are not only conceptually quite different and make different empirical
predictions but may describe different phenomena that can co-exist in the same language
(cf. van Lier 2012).
While those who speak, indiscriminately, of "conversion or zero derivation" attach
little importance to the mechanism involved, more theoretically minded linguists have
deliberately postulated zero affixes in order to uphold their hypotheses. For instance, if
one believes in the principle that, at least in English, prefixes may not change word class
(cf. sections 4.4, 5.3), one can dismiss counter-examples such as endanger by positing
a zero suffix that supplies the feature 'VERB'. It goes without saying that zeros of this
kind are highly suspect.
6.
Morphological processes
6.1. Typology of processes

As the comparison of goose/geese with proojlproofs, etc. in (2) has already shown,
affixation (see section 5) is not the only means of expressing a grammatical meaning.
Intuitively, we would say that geese is formed from goose by changing the vowel /u:/
to Ii:/, i.e. by a morphological process (Mel'cuk 1997: 325-350, 2000, 2006: 288-320) also known as morphological operation (Booij 20 12: 35-42) or morphophonological
operation (Spencer 1998) - which changes the expression side of the base (for the
theoretical consequences of this analysis, see section 6.5). If we confine ourselves to
segmental expressions, there are four ways in which a stri ng of segments can be a ltered:
PA273
- addi tion: one or more segments are added;
- substitution: one or more segments are replaced by others;
- permutation: the segments are rearranged in a different order (apparently, the only
type that occurs in natural languages is metathesis, where two segments change places);
- subtraction: one or more segments are deleted.
Each of these changes can occur as the exponent of a grammatical meaning but also as
a purely phonological process or a diachronic sound change. Sometimes, the same term
serves for both morphological and phonological processes (e.g., metathesis), sometimes
the terminologies diverge (e.g., injixation vs. epenthesis and subtraction vs. elision,
apocope, etc.). Among the morphological processes, substitution (see section 6.3) is next
in importance to addition (affixation). Subtraction is rare but has generated a great deal
of interest (see section 6.4). While metathesis as a phonological process is not uncommon (see section 5.5 for a Hebrew example), it is marginal as a morphological one (cf.
Becker 2000; see also Mel'cuk 1997: 296- 297; Spencer 1998: 138- 140) and will not
be discussed further.
Many enumerations of morphological processes include compounding. This is defensible if morphological process is understood in a broader sense, but compounding differs
from the others in that it does not express a grammatical meaning (cf. Mel 'cuk 2006:
288-290, 297-299); it is therefore not included here. Suppletion, which appears in various lists (e.g., Matthews 1991: 139), is not a process but a relationship (cf. Mel 'cuk
2000: 531 , 2006: 309), and it is not correct to say that, for instance, the English comparatives better and worse are "formed by suppletion". One process that does belong here is
reduplication but it does not easily fit into the above classification. (This issue and some
terminological matters are addressed in section 6.2; for general overviews from different
perspectives, see article 25 on reduplication as well as Wiltshire and Marantz 2000;
Raimy 2011; for attempts at more rigorous definitions, cf. Mel'cuk 1996: 41-52, 1997:
273-288; Stolz, Stroh and Urdze 201 1: 1-70.)
6.2. Reduplication
The term reduplication suggests that something is doub led or repeated, and this is indeed
how reduplication is typically described. Such an explanation will do in simple cases,
such as Chuvash total reduplication as in $er ' hundred' -----+ $er$er ' hundreds ', pin ' thousand' -----+ pinpin 'thousands', and Aghul partial reduplication as in fafu 'five' -----+ fa~ faju
'DISTR~five (five each)', jerxi 'six' -----+ je~jerxi 'oISTR~six (six each)'. (According to
Rule 10 of the Leipzig glossing rules, "reduplication" - which apparently means partial
reduplication - "is treated similarly to affixation" but marked by a tilde ~ instead of a
hyphen.) Here, it seems that a certain string of segments is copied and appended to the
base. The copy is often called reduplicant. The highly similar term reduplicand has been
suggested for the "original" that is copied; if it is adopted, it is advisable not to use
reduplicant but some other term, e.g., image or replicate, for the copy (cf. Mel'cuk 1996:
41 - 42, 2006: 301; Stolz, Stroh and Urdze 2011: 40). The reduplicative base is that string
from which the reduplicant may copy. (On the relationship between the reduplicative
base and the entire base to which the reduplicant attaches, see Hogoboom 2003; Haugen
2009; Inkelas 2012: 357- 359.) Another term for the string within which reduplication
273
PA274
274

appl ies is domain (Mel'cuk 1996: 41 - 42; cf. Stolz, Stroh and Urdze 2011: 43- 44). In
partial reduplication, the reduplicant is frequently viewed as a special type of affix that
is defined only in terms of its prosodic constituents (or, for the sake of simplicity, as a
sequence of vowel and consonant slots). Duplifix has sometimes been understood as a
tenn for such reduplicative affixes although this was not its original meaning (see below). Para.fix (see also section 5.1) was proposed for a specific mechanism to describe
reduplication (by means of an affix that is attached in parallel to the stem and only later
linearized, Clements 1985: 4 7) but has also come to be employed for reduplicative
affixes in general. The picturesque term chameleon morph[eme] (Hockett 1950: 79) underscores that the expression of a reduplicative affix is determined by its environment.
A number of observations suggest that reduplication is not necessarily plain repetition.
a) The reduplicant need not be adjacent to the reduplicative base: In Temiar (Malaysia),
the continuative aspect of a triconsonantal verb stem C 1.C2VC 3 is formed by inserting
a copy of the last consonant just before the final syllable, resulting in C 1C 3 .C2VC 3 ,
e.g., s.bg 'to lie_down' ---+ sg.bg (periods indicate syllable boundaries).
b) The segments to be copied need not be contiguous (before or after reduplication has
applied; cf. Hogoboom 2003: 177- 179 on copying from both sides): If a Temiar verb
stem has the shape C 1VC2 , the consonants are copied without the vowel and are
prefixed to give C 1C 2 .C 1VC2 in the continuative aspect, e.g., ggl 'to sit_down' ---+
gl.ggl.
c) The copy may diverge from the original string: Karaim uses a reduplicant CVp- to
express a high degree of a property, as in kap~kara 'very~black', top~tolu
'very- full'. In Akan (Ghana), pluractional verbs have a prefixal redluplicant that consists of the first consonant of the base and a vowel which agrees with the first vowe l
of the base in the features [back] and [tense] but is always [+high], e.g., /sc?/
'resemble' ---+ /s1- sc?/, /sa?/ 'cure' ---+ /s1- sa?/, /so?/ 'seize' ---+ /su- so?/. From the
perspective of the reduplicative base, this is inexact reduplication; viewed from the
reduplicant, this type of reduplication is characterized by prespecified or fixed segments (e.g., Karaim p) or features (e.g., Akan [+high]). Reduplication with so-called
fixed segmentism (Alderete et al. 1997; why -ism?) has occasionally been regarded
as a combination of reduplication and affixation (e.g., Bergenholtz and Mugdan
1979a: 73) and a special term, duplifix , has been suggested for it (Haspelmath 2002:
24), but given that reduplication with a fixed segment is merely the limiting case of
prespecification, it is not reasonable to single it out.
These forms of reduplication can still be described by means of an underspecified affix
which acquires the missing features by copying them from the base, and if we assume
that reduplication first creates a full copy that is subsequently truncated and/or attribute
certain surface oddities to the workings of phonological rules, we can perhaps even
maintain that the underlying operation is repetition. But if reduplication is to be a subtype
of affixation, we must be able to specify the affix involved. This is possible if the
reduplicant has a constant shape, as in Aghul, where it is always a CV sequence that is
reduplicated, whether it is a syllable of the base (<;a~f'afu) or not (je-jer.xi). There are,
however, cases where the base rather than the reduplicant determines how much is copied, e.g., in Yaqui (Mexico). Here, habitual verbs are formed by copying a complete
syllable, be it CV as in VU.Sa 'awaken' ---+ vu. VU.Sa or eve as in vam.se 'to hurry' ---+
PA275
vam. vam.se (cf. Haugen 2009: 507- 508). Sometimes, it is the entire output which must
have a certain shape - this is true of Temiar, where the continuative conforms to the
template CC.CVC for both bi- and triconsonantal stems (cf. Gafos 1998: 234-249; see
also Inkelas 2012: 366- 371). Such facts lend support to the view that reduplication is
not a special kind of addition but a type of modification on a par with substitution, etc.
(e.g., Mel'cuk 2006: 294- 298) or a process sui generis.
Reduplication as a morphological process, and specifically as a productive process of
word-formation, must be distinguished from
- syntactic doubling with a non-repetitive meaning (syntactic total reduplication, cf.
Stolz, Stroh and Urdze 2011: 3, 26- 29) as in Biblical Hebrew 'TS 'TS 'man man (each
man)', where doubled phrases, e.g., yam laS-Sanah yam laS-Sanah 'day for.the-year,
day for.the-year (a day for each year)', provide additional evidence that this is not a
matter of word-formation;
- iterative affixation as in Polish dam 'house' ---+ dom-ek 'house-DIM (little house)' ---+
dom-ecz-ek ' house-dim-DIM (tiny house)', where -ecz and -ek are allomorphs of the
same morpheme (cf. Stolz, Stroh and Urdze 2011: 34- 36 on Hungarian and Turkish);
- reduplicative roots and stems which do not follow a pattern that is regular on both
the expression and the content side, as in English zigzag, where the parts zig and zag
do not occur by themselves and have no meaning (cf. Stolz, Stroh and Urdze 2011:
33-34 on Maltese), or Yiddish pile-pl6im 'great wonders', which has lost the phrasal
structure of its Hebrew source pil' ej pala'Tm 'wonders.of wonders' and is synchronically neither morphologically segmentable nor regularly related to pele 'wonder'.
Unfortunately, these distinctions are not always observed, and what studies of wordformation in individual languages describe under the heading "reduplication" varies
greatly.
6.3. Substitution
The substitution of segments as in goose ---+ geese or of suprasegmental features as in
recoil---+ recoil (see section 5.8) is also called replacement or internal modification (e.g.,
Bauer 2004: 57- 58; for an overview of types of substitution and a discussion of theoretical issues, see Lieber 2000). The term apophony is used either for any substitution or
specifically for a vowel change (Bauer 2004: 17- 18). In a wider sense, apophony includes subtraction and metathesis as well (cf. Mel 'cuk 1982: 98- 102, 2006: 302- 304).
Other terms that can similarly be applied more broadly or to specific phenomena in
individual languages inc lude mutation, gradation and, for vowels, ablaut. Vowel changes
due to assimilation, in particular fronting before a high front vowel in the following
syllable, are known as umlaut; historically, English goose ---+ geese belongs to this type.
In German linguistics, there is a tradition of calling derivational vowel changes (as in
sing ---+ song) implicit derivation as opposed to explicit derivation with an affix (e.g.,
Fleischer and Barz 1992: 51; Donalies 2005: 132- 133).
Attempts at construing substitution as an additive process operate with a "mutating"
infix that manifests itself only through its effect (Bossong 2001: 667) or with a featural
affix that coalesces with one or more segments of the base (Akinlabi 20 11 ; Bye and
275
PA276
276

Svenonius 20 12: 443- 451; cf. also Volod in and Chrakovskij 1991: 116- 117). Such an
affix has been called simulfix (cf. Wallis 1956), but the term is sometimes extended to
substitutions in general irrespective of whether they are analyzed in this way (cf. Lieber
2010: 79; see also section 5 .8 on other meanings of simulfix).
In European languages, substitution quite often accompanies affixation (see section
6.6); more rarely, it is the only or primary exponent of a grammatical meaning. Productive substitution is largely confined to inflection, e.g., plural formation of nouns in several Germanic and Romance language varieties which originally had a suffix that conditioned palatalization of the preceding consonant or fronting/raising of the preceding
vowel. Loss of the suffix resulted in morphologization of the alternation, which could
then be analogically extended to lexemes that did not have it originally. In derivational
morphology, a number of non-productive substitution patterns can be observed; in some
descriptions, they appear under "conversion" on the assumption that the substitution is
merely a side-effect. Verb/noun pairs similar to English sing/song occur in Germanic,
Romance and Balto-Slavic languages; for example, the verb 'to fly' and the noun 'flight'
(without inflectional suffixes) are /vul-/ and /v'Jwl/ in the Bolognese dialect of Italian, I
letG-/ and /lot/ in Polish, /skrid-/ (past tense stem) and /skri:d-/ in Lithuanian. Causative
verbs that are derived from intransitive verbs by a vowel change, e.g., English fall -+
fell 'to cause to fall', are found in other Germanic languages as well. In the Circassian
branch of Northwest Caucasian, intransitive and transitive verb roots often differ only
in the vowel, /a/ vs. /:;,/, e.g., Kabardian /txa-/ 'to be involved in writing' vs. /tx;J-/ ' to
write [something]', and this may be a case of substitution /a/ -+ /:;,/ (see article 193,
section 4.3.3 on Adyghe and article 194, section 4.2 on Kabardian).
The limited role of substitution as a morphological process can be attributed to its
lack of constructional iconicity or diagrammaticity: Additional content is not matched
by additional expression. Languages of the agglutinating type (represented in Europe by
Turkic, Mongolian and, to a lesser extent, Uralic) value iconicity very highly and therefore avoid substitution, whereas the Indo-European languages exploit its main advantage,
brevity, to a limited degree. Where affixation and substitution can express the same
grammatical meaning, e.g., past tense in Germanic weak and strong verbs such as sin/
sinn-ed vs. sing/sang, one might think that a preference for the economy of substitution
over the iconicity of affixation should correlate with a high text frequency of the lexemes. In actual fact, the relationship between frequency and irregularity is much more
complex (cf. Meder and Mugdan 1990: 87-89 on German).
6.4. Subtraction
Subtraction expresses a grammatical meaning by deleting part of the base, prototypically
the final segment (for a survey, cf. Dressler 2000). An example that seems to be more
convincing than most comes from Tohono O'odham (Arizona): Perfective verbs are
formed from imperfective ones by deleting the last C or, in certain cases, VC, e.g.,
/hihim/ -+ /hihi/ 'to laugh', /hi:nk/ -+ /hi:n/ 'to bark', /gigo~id/ -+ /gigo~/ 'to feed' (Kosa
2008: 18). The decisive reason for assuming this direction of derivation is that otherwise
there would be a number of imperfective suffixes, I-ml , I-kl , /-id/, etc., with an unpredictable di stribution. The same argument has been applied to adjectives nn spoken French,
PA277
where the shorter masculine form is predictable from the longer feminine form but not
vice versa, e.g., I otf ~ l o/ 'high', /lurd/ ~ /lur/ 'heavy' . Here, a simple phonological
solution preserves the semantically more plausible derivation of the feminine from the
masculine: If the underlying representations are lotl 'high(M)' and lot-;}I 'high-F' (cf. the
spellings haut and haute), the surface forms can be obtained by applying deletion of the
final consonant and deletion of final /;}/ in that order. In fact, most alleged instances of
subtractive morphology have been reanalyzed in one way or another (cf. Kosa 2008: 515); some are non-productive and concern a very small number of items, some are better
interpreted as derivation of the longer form from the shorter one and some may involve
an opposition between an explicit affix and zero (e.g., Bolognese Italian /skra:na/ 'chair '
vs. /skra:n/ 'chairs', /do:na/ 'woman' vs. /do:n/ 'women', where /-a/ may be a singular
suffix, cf. /raga:sa/ 'girl' vs. /raga:si/ ' girls').
In derivational morphology, subtraction must be distinguished from phenomena that
are sometimes conflated with it, namely clipping and other shortening processes (see
article 19) and backformation (see article 18). The latter is not a morphological process
at all but a diachronic development: Sometimes a word that contains a derivational affix
(or looks as if it did) exists in the lexicon without an underived counterpart and is
therefore not fully segmentable (see section 2.2). This was true of the English noun
sculptor when it was borrowed in the 17111 century without a base verb. Backformation
consisted in the creation of the verb sculpt, and when it entered the lexicon (according
to the Oxford English Dictionary, it is first attested in 1864), sculptor became a word
derived from sculpt notwithstanding the fact that it existed before its base. The argument
that backformation "must be allowed for in a synchronic grammar if it is still a current
method of forming lexemes" (Bauer 1983: 65) is fallacious; it is based on a misinterpretation of synchrony vs. diachrony as "current" vs. "former" and does not distinguish the
technique of badcformation that is available to native speakers from the derivational
relationships that hold between items in the lexicon at a given point in time.
Unlike backformations, clippings such as bike from bicycle and name truncations
such as Liz from Elizabeth (which should perhaps not be lumped together) are indeed
synchronically derived from the longer forms by tmncation. These processes are often
said to differ from subtraction in that they do not change the meaning. However, they
frequently signal endearment, contempt and the like and thus resemble derivations with
hypocoristic or pejorative affixes (see article 88 on the semantics and pragmatics of
Romance evaluative suffixes). The relevant difference seems to be a formal one: Subtraction deletes a specified part of the base (e.g., the last consonant), whereas templatic
truncation (clipping, etc.) deletes whatever is needed to arrive at a specified shape (e.g.,
a single closed syllable); when this is taken into account, another alleged difference
between subtraction and clipping, regularity vs. unpredictability, cannot serve as a criterion either (cf. Alber and Arndt-Lappe 2012 ).
In addition to backformation and templatic truncation, we must eliminate two further
patterns that sometimes figure as derivational "subtraction" in the literature. Both can
be illustrated by Russian lirika ' lyric_poetry' ~ lirik 'lyric_poet'. First, we must bear
in mind that lirika has an inflectional suffix -a ' NOM.SG', whereas lirik has none (or -0):
The stems are identical, and only the citation forms create the illusion of subtractive
word-formation. Second, it could be argued that lirik contains the same derivational
suffix -ik as many other personal nouns; if so, stem-final -ik is truncated before derivational -ik just as it is truncated before -izm in lir-izm ' lyric ism' . Truncation of the base
277
PA278
278

in conjunction with affixation is quite common and can affect both phonological strings
(see section 6.6 on German) and morphological units, e.g., the verbal suffix -ate in
nominate - nominee (cf. Aronoff 1976: 88-97, where truncation refers specifically to
this type). In the latter situation, one could also speak of the replacement of one affix
by another, so-called "paradigmatic word-formation" (cf. Booij 2002: 6- 7, 102- 103).
At any rate, we are not dealing with subtraction as the primary exponent of a derivational
mean mg.
For the remaining examples of derivational subtraction that have been cited, alternative analyses can be offered. In pairs such as Serbo-Croat Hrvatsk-a 'Croatia-NOM.SG' I
Hrvat ' Croat(NOM.SG)' (inhabitant name), the direction of derivation is debatable: Is a
Croat someone who lives in Croatia or is Croatia the country of the Croats? Apart from
that, the great diversity of formal relations between the names of countries, regions and
cities on the one hand and inhabitant names on the other (which cam be observed in
many European languages, with remarkable cross-linguistic differences in detail) makes
one wonder whether derivational rules are at play at all. Moreover, in Hrvatska ---+ Hrvat
as well as in the patterns exemplified by French /fotografi/ ' photography; photograph'
---+ /fotograf/ 'photographer' and Bulgarian misk-a 'mouse-sG' ---+ mis-i ' mouse's-M.SG'
(adjective as in misi zab ' mouse tooth', etc.), the deleted material is not phonologically
defined as it should be in true subtraction. Rather, it is a derivational affix (although in
Bulgarian, in contrast to Russian mys' 'mouse' ---+ myska 'mouse.DIM' , only the derived
form miska exists), and one might consider the poss ibility that both members of each
pair are derived from a common base and not one from the other. In sum, in derivational
as in inflectional morphology, there are no uncontested instances of subtraction. This is
not surprising: As a counter-iconic process in which more content is coupled with less
expression, subtraction is clearly the worst option.
6.5. Elements and operations

Substitution and subtraction present a challenge to building-block models of morphology.
American structuralists therefore came up with the idea that goose and geese have a
discontinuous stem /g .. . s/ and infixes /u:/ 'sG' and Ii:/ 'PL' or that geese consists of an
allomorph of {goose} and a zero suffix 0 'PL' . These counter-intuitive analyses, the
second of which would work for subtraction as well, were soon abandoned in favour of
the tenet that a "morpheme", i.e. the expression of a linguistic sign, can be either a str ing
of segments or a process that manipulates such a string (see also Mel 'cuk 2006: 295297, where both entities and operations are s ign-expressions but are systematically kept
apart). In this way, all morphology involves the concatenation of signs, and terms such
as replacive morph[eme] and subtractive morph[eme] (cf. Nida 1948: 440-441), simulfix
(see section 6.3) and disftx (Hardy and Montier 1988: 389) are symptomatic of the
reinterpretation of processes as entities. (Subtractive morph[eme] is sometimes erroneously used for what is subtracted - it is, however, not the deleted material but the process
of deletion that expresses a grammatical meaning.) Such approaches stem from the belief
that affixation is more "natural" (although there is some experimental evidence that
speakers can learn stem changes as easi ly as affixes, cf. Bybee and Newman 1995) or
follow from specific assumptions about the modular structure of grammar (cf. Bye and
PA279
Svenonius 2012: 428, where non-concatenative effects are attributed to the phonological
component).
At the other extreme, we find models that reinterpret entities as processes and do
away with inflectional affixes as linguistic elements: Morphosyntactic features trigger
certain string manipulations, among which addition has no pride of place. On this view,
(inflectional) morphology is "a-morphous" in the sense that it operates without (affixal)
morphs, i.e. minimal signs (cf. Anderson 1992: 1- 2). The two models are not mutually
exclusive, however, inasmuch as the second describes how one can get from a lexeme
with a set of morphosyntactic feature values to the corresponding word form (e.g., the
1st person plural past indicative of the Catalan verb CANTAR 'to sing' is expressed by
cantavem), which is the perspective of the speaker, while the first shows how one can
infer the meaning of a complex word from the meanings of its constituents (e.g., the
combination of cant- 'sing', -av 'PAST. IND' and -em '1 PL' gives 'we sang'), which is part
of the task of the hearer.
6.6. Mult iple processes

Very often, a given grammatical meaning seems to be expressed by several morphological processes simultaneously - this has been called extended exponence (Matthews
1974: 149- 150; cf. Coates 2000: 622- 623, where extended and multiple exponence are
distinguished). For instance, German diminutives are characterized by a suffix -chen
(e.g., Leiter 'ladder ' -----> Leiterchen) and possibly one or more of the following:
- umlaut, a vowel substitution, orthographically a---->a, o---->o, u---->ii (e.g., Nagel 'nail' ----->
Nagelchen );
- truncation of stem-final -e !'di or -en (e.g., Riemen 'strap ' -----> Riemchen); this can be
treated as a purely phonological phenomenon (according to another opinion, stemfinal -e and -en are always derivational suffixes and do not occur when another nounforming suffix does, cf. Wiese 1996a: 26);
- insertion of -el after the velars g, ng lul, ch /xi but not k (e.g., Strich 'line' -----> Strichelchen; Kuchen 'cake' -----> Kiichelchen illustrates the combination of all processes).
In such a situation, many linguists - tacitly or, more rarely, expressly - follow the
"principle of a single morphological process" (cf. Mel 'cuk 1982: 79- 80, 2000: 533534, 2006: 314- 318) and select one of the morphological processes as the exponent of
the content in question, relegating the others to the status of "meaningless" side effects.
There are two major reasons why in our example the suffix -chen should be regarded as
the exponent of ' DIMINUTIVE'. One is that, other things being equal, an affix is the
preferred exponent because addition ranks higher on a hierarchy of naturalness than
substitution and subtraction. More importantly, chen-suffixation is the process that always occurs; umlaut, e(n)-truncation and el-insertion are obligatory when the phonological preconditions are met but are not applicable to all bases. An even stronger case for
the "meaninglessness" of umlaut can be made with agent nominalizations in -er, where
some of the eligible bases exhibit umlaut and others don 't (see section 4.3 on Trager
'carrier' vs. Fahrer 'driver').
Other rnorphologists - again tacitly or expressly - reject the idea that 'DIMINUTIVE'
has the suffix -chen as its only exponent or marker, whereas it triggers umlaut, e(n)-
279
PA280
280

truncation and el-insertion without being expressed by them (e.g., Wurzel 1989: 51;
Stump 2001 a: 3-7, 10-11 ); this tends to go hand in hand with the rejection of buildingblock models. Instead, it is assumed that all the morphological processes taken together
express the content in question - with the corollary that morphologically conditioned
stem allomorphy (see section 2.4) is logically impossible. Some who take this position
weaken it by admitting that the processes are not equal in rank and identify one of them
as the main exponent (e.g., Matthews 1991: 181).
Suprasegmental processes that accompany segmental ones in inflection are often best
described separately; the complex stress alternations in some Balto-Slavic languages, for
instance, constitute stress paradigms that are fairly independent of the affix paradigms
(see section 3.4). Derivational affixes, on the other hand, are typically characterized by
a particular behaviour with regard to stress or tone: Some are neutral and preserve the
stress or tone of the base, whereas others impose their own pattern or modify the pattern
of the base in a certain way.
7. Inputs and outputs of word-formation rules

Before we can turn to the general questions as to which linguistic units serve as inputs
and outputs of word-formation rules (section 7.3) and whether they are specified for
word class (section 7.4), we have to address two specific issues: Can phrases (section
7.1) and inflectional forms (section 7.2) be constituents of complex words? Examples of
both phenomena have been reported for well over a century (cf. Baudouin de Courtenay
1869 on Polish) and have generated substantial controversy. Non-morphological inputs,
notably letters and syllables (see article 19 on clipping, sections 3.3-3.4), and wordformation that is not rule based (see article 26 on word-creation) will not be considered
here.
7.1. Phrases as input

It is frequently claimed that phrases can serve as inputs to word-formation rules. What
counts as a phrasal compound or a phrasal derivative is, however, often not clearly
defined but simply taken for granted, and the phenomena that have been subsumed under
these rubrics are quite varied. A phrasal input is most likely to be posited if typical traits
of a phrase are still present in the output, as in Dutch heteluchtballon 'hot air balloon',
which seems to preserve the agreement suffix -e of h et~ lucht 'hot air' as well as the
phrasal stress pattern. However, a phrasal base has also been assumed for compounds
like Dutch r6odbuikpapegaai 'red belly palTot', where the constituent r6odbuik lacks the
-e of rod(}_ buik 'red belly' and has compound stress, and for derivatives like red-bellied
in the corresponding English name Red-bellied Parrot. Function words are another characteristic of phrases that appears to occur word-internally, e.g., the conjunction en 'and'
in Dutch peper-en-zout-stel ' pepper and salt set' and the preposition/iir 'for' in German
Wort-fiir-Wort-Ubersetzung 'word for word translation'. In other compounds, comparable
relations remain unexpressed, as in Swedish ost-vdst-konjlikt 'East-West conflict', which
can be paraphrased as konjlikt mellan ost och vast 'conflict between East and West'. For
PA281
some morphologists, ost-vast is nonetheless a compound member of the same kind as
peper-en-zout, be it a phrase or a compound (cf Booij 2002: 146-151 ).
Often, a phrase is postulated as the input to a word-formation rule if the constituents
that belong together semantically do not form an existing word (see section 4.4 on
unattested intermediate steps). Dutch heteluchtballon, for instance, cannot consist of hete
'hot' and luchtballon 'air balloon' (both of which are attested words) because it does
not mean 'air balloon that is hot' but 'balloon that operates with hot air'. But although
hetelucht must be regarded as one of the immediate constituents for semantic reasons, it
does not occur as a word by itself.
Many linguists reject phrasal inputs to word-formation rules altogether because the
model of grammar they advocate does not allow the kinds of interactions between syntax
and morphology that would be required. In order to maintain this "No Phrase Constraint", alleged examples of phrasal compounds and derivatives must be analyzed differently. The alternatives that can be pursued include the following:
a) The complex word has a different bracketing. For example, instead of regarding a
so-called "synthetic compound" such as Dutch groenogig 'green-eyed' as a derivative
from the phrase groen-e og-en 'green-PL eye-PL' (with plural suffixes that do not
occur in the derivative), we could view it as a compound with a secondl component
that does not occur as a word, viz. *ogig 'eyed' (for some discussion, cf. Booij 2002:
158- 160; see also article 33 on synthetic compounds in German and Donalies 2005:
91 - 93).
b) The supposed phrasal constituent is a compound. For the type roodbuikpapegaai, for
instance, one can argue that roodbuik is a compound rather than a phrase. Since
compounds like roodbuik are not productive except as first members of compounds,
this analysis presupposes that possible words are admissible as constituents of complex words (see section 4.4; cf Booij 2002: 148-151).
c) The supposed complex word is a syntactic construction. This option is avai lable
in languages such as English where so-called compounding often consists in mere
juxtaposition. While some regard local body elections as a compound, others classify
it as an entity between a noun (N) and a full noun phrase (NP), with local body being
a nominal modifier of elections (cf. Carstairs-McCarthy 1998: 213).
d) The complex word is created in a single step. In pa1iicular, simultaneous application
of compounding and derivation is a solution for groenogig, etc. that avo ids the need
to posit constituents *groen-oog or *og-ig that do not exist as independent words. A
variant of this approach is to assume a template [A [N-ig]A]A ' having N with property
A', a conflation of the independently motivated templates [A AJA and [N-ig]A (cf.
lichtgroen ' light green' and kleurig 'colourful' ). While it has the bracketing suggested in (a), it takes into account that A has semantic scope over N only and not over
N-ig and that N-ig is not necessarily an attested word (cf. Booij 2002: 159- 160,
2005: 128- 129).
e) The phrase is quoted. It might seem, for example, that no phrase constraint itself
disproves the constraint because no phrase is a phrase, but one could counter that in
this case no phrase is being mentioned rather than used (cf. Carstairs-McCarthy 1998:
214). This analysis is supported by the occasional use of quotation marks, as in the
German newspaper example die ,, Sicherer-als-Bargeld "-Idee 'the "safer than cash"
idea' (Lawrenz 2006: 72) and by compounds with foreign-language items or non-
281
PA282
282

alphabetic signs, e.g., German ihre C'est-la-vie-Haltung 'her c'est-la-vie attitude'
(with a French clause) and das @-Zeichen 'the @-sign' (Wiese 1996b: 186-187),
which demonstrate that linguistic elements of any kind can be quoted as part of a
complex word.
7.2. Inflectional forms as input

In many de scriptions of individual languages, certain derivatives and compounds are
said to contain inflectional forms, for instance Icelandic oro-a-bok 'word-GEN.PL-book
(dictionary)' and Yiddish mishpokhe-s-vayz 'family-PL-wise (by families)'. This assumption contradicts Joseph Greenberg's Universal 28: "If both the derivation and inflection
follow the root, or they both precede the root, the derivation is always between the root
and the inflection" (1966: 93). The generalization can be reformulated in hierarchical
rather than purely linear terms so that it applies even to markers on opposite sides of
the root: Inflection has scope over derivation (cf. Rainer 1996: 83) and likewise over
compounding. If so, word-formation and inflection can be handled by distinct components of grammar that operate strictly in this order. A different architecture of the
grammar is required if compound-internal inflection and inflection within derivation
are allowed. While some linguists explicitly defend this option, others offer alternative
interpretatio,ns so as to avoid a loop from inflection back into word-formation or a conflation of the two components, and some try to reduce the problem by permitting only
certain types of inflection to feed derivation, e.g.,
- only irregular inflectional forms, on the grounds that they are listed in the lexicon (cf.
Anderson 1988: 41; see also Perlmutter 1988: 89), a stock example being English
mice-infested vs. *rats-infested;
- only "inherent" inflection, i.e. selected features such as number as opposed to imposed
features such as case (see section 3.4; cf. Booij 1996; see also Gallmann 1999: 178179, 183 ~ Motsch 1999: 8).
These generalizations are, however, either aprioristic or based on too few examples and
should not be accepted without further empirical evidence.
Compounds and derivatives that seem to be built on inflectional forms can be reanalyzed in a number of ways:
a) The supposed compound is a phrase. This solution is particularly convincing if other
phrasal properties are present as well. In West Frisian, for example, the "genitive
compound" koken-s-doar 'kitchen -GEN-door (door of the kitchen)' has certain phonological and semantic characteristics which it shares with phrases but not with regular
compounds such as koken-doar 'kitchen-door (door of a kitchen)'. But even if a
construction with a case marker is otherw ise quite similar to a complex word, it may
qualify as a phrase. For instance, Kalmyk /ykr-c:e maxn/ 'cow-GEN meat (beef)' seems
to consist of two words - so that it is not of the same type as the compound /ho!
maxn/ 'centre meat (fillet)' - although it differs from normal phrases in that the
components appear in a fixed order and cannot be modified (cf. Pavlov 1983 : 88).
Even semantic non-compositionality as in the case of Khinalug /pt:-e xu/ 'eye-GEN2
water (tear)' is not sufficient evidence against phrase status. Other phenomena that
PA283
may very well be phrases include close-knit combinations with case-marked objects,
e.g., Icelandic fan-um pnydd-ur 'flag-DAT.PL decorate.PsT.PTCP-NOM.SG.M (decorated
with flags)' and Hungarian iskola-ba jar 'school-ILLAT go (attend school)' with the
nominalization iskola-bajar-as 'school-ILLAT go-NR (school attendance)', and instances of "internal inflection" as in English brother-in-law (plural brother~-in-law).
b) The alleged inflectional marker is derivational. This is a possible response to the
claim that comparatives and superlatives are typical examples of "inherent inflection"
(cf. Booij 1996: 5- 6) which can feed word -formation, as in the Ukrainian derived
verb iovt-is-a-ty 'yellow-CMPR-VR-INF (become more yellow)' +----- iovt-is -yj 'yellowCMPR-NOM.SG.M' +----- iovt-yj 'yellow-NoM.SG.M'. Infinitives and participles occur
in derivatives and as both first and second members of compounds, e.g., Dutch
slap-en-s-tijd 'sleep-INF-LINK-time (bed time)', begrens-d-heid ' limit-PAST.PTCP-NR
(limitedness)', boek-bind-en 'book-bind-INF (bookbinding)' and computer-ge-stuur-d
'computer-0-control-PST.PTCP (computer controlled)'. But in these complex words
they behave like nouns and adjectives, as is evident from the linking -s in slapenstijd
and the non-existence of finite forms of *boekbind- and *computerstuur- (cf. Booij
1996: 7). One can therefore either say that infinitives and participles are always
derivatives rather than part of the inflectional paradigm of the verb or that they
undergo conversion to nouns and adjectives before further word-formation rules apply
(see also sections 3.4 and 7.4).
c) What looks like an inflectional fonn is a homonymous special stem or combining
form. For two reasons, special stems that are used only in word-formations must be
assumed in any case. First, not all linking elements resemble inflectional markers
and those that do sometimes appear even in cases where the lexeme in question does
not have a form with this marker. One example is that of Icelandic -s, which is
identical with one of the genitive singular suffixes but is also found in compounds
of the type kuplings-diskur 'clutch-disk' although the genitive singular of KUPLING
'clutch' is kitplingar (see section 5.7). Secondly, constituents of compounds or derivatives often coincide with an inflectional form for which there is no semantic motivation (cf. also Fuhrhop 1998: 192-194; Eschenlohr 1999: 195-206 on German). In
Latvian, for instance, the nominalizer -ums is described as attaching to the past tense
stem of verbs, e.g., meginajums 'attempt' (cf. meginaj-u 'try.PST-1.SG', megin-u
'try.PRS-1.sG' ), but the feature [+past] is not reflected in the meaning of the resulting
nouns (for a similar case involving [+feminine], see article 106 on the Romance
adverbs in -mente). One could allow word-fonnation rules to delete morphosyntactic
properties or devise a mechanism that prevents certain features from contributing to
the meaning of the complex word or simply posit that they are not carried over (cf.
Mel'cuk 2006: 293- 294; Wiese 1996a: 144- 145; Augst 1975: 83, 133), but this
leaves unexplained why inflectional forms should occur in complex words to begin
with. A further problem is that, for no apparent reason, similar word-formations often
exhibit different inflectional forms of the same lexeme. For example, Finnish sadekappa 'rain-coat' and sateen-varjo 'rain-shade (umbrella)' seem to contain the nominative singular sade and the genitive singular sateen of SADE 'rain', respectively,
although the semantic relationship between the compound members is the same.
Thus, there is good reason to regard Finnish sade- and sateen-, Latvian meginaj- and
the like as special word-formation stems that are merely homonymous with inflectional forms or stems but cannot be equated w ith them. (Alternatively, one could
283
PA284
284

assume "morphomic" stems that occur in an arbitrary set of inflectional forms, compounds and/or derivatives; crucially, such stems are not specified for morphosyntactic
properties either, see section 4.3.) In other cases, the evidence in favour of regarding
certain word constituents as inflectional forms is stronger but still not compelling:
- There are numerous complex words with a constituent which not only looks like
an inflectional form but apparently has the grammatical meaning associated with
that form. In German Hiiuserreihe 'row of houses', for instance, the first member
Hauser is identical with the plural of Haus 'house' and could be argued to be
semantically plural because a row involves more than one item (cf. Booij 1994:
36- 37, 2002: 147 on Dutch) - but compounds with Reihe 'row' do not always
contain a plural form, e.g., Lochreihe 'row of holes' (+---- Loch 'hole', plural Locher). Even if Loch- is not viewed as a singular form but as a stem which is not
specified for number (cf. Booij 2002: 145), this demonstrates that the plural reading required by Reihe is independent of number marking on the other constituent.
Conversely, some compound members that must be semantically singular are plural
in form, e.g., Eier-schale 'egg-shell' ( +---- Ei 'egg', plural Eier) as opposed to Eigelb 'egg-yellow (yolk)'. Although the percentage of such mismatches may be
rather small (see Augst 1975: 145-152 for some statistics), their existence raises
the question of whether it is really so obvious that Hauser- in Hiiuserreihe is an
inflectional form.
- Psycholinguistic experiments concerning the Dutch linking element -en, which is
homonymous with a productive plural suffix (cf. Booij 2002: 24, 180), show that
a first compound member with -en is more likely to be interpreted as plural than
as singular (Hanssen et al. 2013) and that informants prefer -en to no linking
e lement in contexts that suggest a plural meaning (Banga et al. 2013) - but a look
at the figures reveals that both tendencies are not particularly pronounced.
- There are minimal pairs such as German Volkskunde 'folklore (as a discipline)' vs.
Volkerkunde 'ethnography' (+---- Volk 'people', plural Volker) and Dutch stedenraad
'council of cities' vs. stadsraad 'city council' (+---- stad 'city', plural steden; see
Booij 2002: 179). Their number is, however, quite small, their meanings are not
fully predictable in terms of singular vs. plural and other pairs with such a formal
difference are synonymous, e.g., the German regional variants Rindsbraten and
Rinderbraten 'beef roast' (+---- Rind 'bovine', plural Rinder).
It is therefore generally possible to assume word-formation stems without morphosyntactic properties instead of treating some word constituents as inflectional forms and others
as stems. While the choice of one word-formation stem rather than another can have a
bearing on the semantic interpretation of the complex word, it is only one of many
factors involved and can be accounted for without recourse to inflectional meanings.
7.3. Words, lexemes, roots or stems?

Compounds are sometimes defined as containing two or more roots (e.g., Jackson 2007:
29), and a classification of affixes according to their position relative to a root (see
section 5.1) implies that derivation likewise operates on roots. While this is true in the
PA285
simplest of cases, it must be remembered that the output of a word-formation rule can
serve as the input to another (or even the same) word-formation rule. Danish sprogvidenskabelig ' linguistic', for example, is a derivative based on the compound sprogvidenskab
'linguistics', which in turn consists of sprog ' language' and the derivative videnskab
'science' (+--- viden 'knowledge' +--- vid- ' to know'). A characterization of word-formation rules as operations on roots is therefore inadequate.
According to another definition, word-formation forms words from words (see section
I); in particular, a compound is said to be "a word wh ich consists of two or more words"
(Fabb 1998: 66). Since word has a number of different meanings (see section 3), we
must clarify which of them is intended here. The output is definitely a lexeme as a whole
and not one of its grammatical words with certain morphosyntactic properties, and it is
sometimes explicitly stated that the "word" which a word-formation rule takes as its
input is a lexeme as well (e.g., Booij 2002: 141; Bonami, Boye and Kerleroux 2009:
124; Montermini 2010: 86). One handbook article that offers such an explanation adds:
"The implication of this is that the forms in which the individual subwords appear may
be differently defined in different languages: a citation form in one, a stem in another, a
specific compounding form in yet a third, a word form in a fourth" (Bauer 2006: 719).
While this statement does not logically follow, it leads to an important point: As a set
of grammatical words, a lexeme cannot appear as such but must be represented either
by a specific word form or by a stem (cf. also Montermini 2010: 87). There is no need
to single out the citation form since it is simply one of the word forms; if it coincides
with the bare stem, it is better analyzed as such, unspecified for morphosyntactic properties (cf. Booij 2002: 145). If a lexeme is allowed to have stems that occur only in wordformations (which requires a slight revision of the definition of stem in section 4.3), the
"specific compounding form" is not a separate case either. On the premise that compounds and derivatives do not contain inflectional forms (see section 7.3), only one of
the four options mentioned in the above quotation remains: The input to a word-formation rule is a lexeme, represented by a word-formation stem.
While special word-formation stems are not a novel idea (cf., e.g., Bloomfield 1933:
229), the implications for the paradigms of lexemes require further study. The suggestion
that lexemes (or at least German nouns, verbs and adjectives) have a stem paradigm
which always contains at least one derivational stem and one compounding stem in
addition to the inflectional stem(s) (Fuhrhop 1998: 22- 26; Aronoff and Fuhrhop 2002:
454- 456) is not convincing because the distinction between derivation and compounding
may or may not play a role in the distribution of word-formation stems. The principle
that all lexemes of a given word class have a fixed number of stems (many of which
have the same expression in certain inflection classes) with the same distribution for
each lexeme is feasible for inflectional paradigms (cf. Bonami, Boye and Kerleroux
2009: 107-108) - but when word-formation sterns are included (cf. Bonami, Boye and
Kerleroux 2009: 120-123), it becomes unwieldy. It seems more realistic to allow lexemes to have a variable number of word-formation stems with distributions that do not
necessarily have a parallel in other lexemes. For example, numerous Finnish nouns have
two word-formation stems, one homonymous with the nominative singular and one
homonymous with the genitive singular (e.g., sade- and sateen- 'rain-', see section 7.2),
but there are no firm rules as to when which of them is used. Such systematic homonymy
between word-formation stems and inflectional forms requires a further refinement of
285
PA286
286

the model: It must permit the expression side of a stem to be deduced not only from
another stem but also from a specific cell in the inflectional paradigm.
A word-formation stem may be special not only in expression but also in content. It
does not necessarily have the full range of meanings of the lexeme it belongs to, and its
meaning in a complex word may well depend on the constituent it combines with; socalled affixoids (cf. section 4.2) can be viewed simply as word-formation stems with
meanings that the simplex word does not have. On the practical side, this implies that
dictionaries should have (sub-)entries for word -formation stems with a description of the
most common patterns in which they occur and the meanings they have there (for some
suggestions, cf. Mugdan 1984: 298- 303). Furthermore, one can allow for the possibility
that the only stems of a lexeme are word-formation stems, e.g., phon- and -ology in
phonology, etc. (cf. section 4. 2 on "confixes"). Then, a distinction between word-formations at different levels, e.g., root, stem and word (cf. Wiese 1996a: 13 1 on German), is
not needed (cf. Montermini 2010: 82, 87).
7.4. The role of word classes and subclasses

Word-formation rules are commonly stated in terms of word classes or lexical categories
(noun, verb, adjective, etc.). In particular, it is often assumed
- that roots are specified for word class,
- that derivational rules operate on bases of a specific word class ("unitary base hypothesis", cf. Aronoff 1976: 47-48) and
- that they generate lexemes of a specific word class ("unitary output hypothesis", cf.
Scalise 1984: 137).
The unitary base hypothesis and the unitary output hypothesis lie behind the popular
characterization of derivatives as denominal verbs, deadj ectival nouns, etc. Proponents
of the unitary output hypothesi s additionally require that the output should be uniform
with regard to subclass (e.g., gender of nouns, transitivity of verbs), inflectional class
and perhaps semantics (cf. Plag 1999: 49- 50); the unitary base hypothesis is usually not
interpreted quite so strictly.
The unitary output hypothesis is w idely accepted but exceptions are easy to find.
Italian -ista and anti-, for example, create words that function as both nouns and adjectives (cf. Montermini 2008: 182- 183), e.g., ottimista ' optimist/optimistic ' (+-- ottim-o
'optimal-sG'), antifurto ' burglar alarm/anti-theft' (+--fart-a 'theft-sG'). Here, it is difficult to justify two homonymous affixes or a conversion from noun to adjective or vice
versa (see also De Beitler 20 11: 152- 170 on Dutch). Frequently, derivational affixes
which violate the unitary output hypothesis violate the unitary base hypothesis as well
and are "transparent" with regard to word class or subclass. Evaluative suffixes in Romance and Slavic languages which preserve the gender of the base and can sometimes
be used with words of different classes are a typical example, e.g., the Kashubian
diminutive -k in the masculine noun grzibk +-- grzib ' mushroom', the feminine noun
gqbka +-- gaba 'face' (with the nominative singular suffix -a) and the verb r6bkac +robic 'to work' (with the infinitive suffix -c preceded by a theme vowel). Yet, evaluative
suffixes do not always behave in this way. For instance, the diminutive affixes of Dutch,
PA287
Frisian and German attach to nouns of any gender but the output is always neuter, in
conformity with the unitary output hypothesis; Albanian has diminutive suffixes which
are gender-specific with regard to both input and output, -th for masculine nouns as in
djaleth +--- djale ' boy' and -z for feminine nouns as in arkez +--- arke 'box'. On the other
hand, suffixes that are transparent to gender are not necessarily evaluative; one example
is Breton -ad as in tasad 'cupful(M)' +--- tas 'cup(M)', bolad 'bowlful(F)' +--- bol
' bowl(F)'. Prefixes do not constitute a systematic exception either. Many are transparent
to word class, but some do conform to the unitary output hypothesis and determine the
word class of the output, e.g., English en- in endanger (unless one resorts to one of the
debatable analyses mentioned in sections 4.4 and 5.9). Thus, it is not predictable which
derivational affixes obey the unitary output hypothesis and which do not, and perhaps
the possible outputs should not be characterized in terms of output w ord c lasses at all
(cf. De Belder 2011: 170- 195).
The unitary base hypothesis is almost impossible to falsify because of two immunization strategies that can be employed if a given affix attaches to bases of different word
classes (cf. Plag 1999: 47-48, 2004: 198):
a) If the word classes in question form a natural class, the latter can be regarded as a
unitary base. For example, if nouns are defined as [+N,- V], adjectives as [+N,+V]
and verbs as [-N,+V], an affix that attaches to both nouns and adjectives, e.g., -en
in strengthen and deepen, has the unitary underspecified base [+N] and one that
attaches to both verbs and adjectives, e.g., -th in growth and warmth, has the base
[+V].
b) If the word classes in question do not form a natural class, e.g., nouns and verbs in
the feature system given in a), one can assume homonymous affixes, e.g., -ful 1 in
thoughtful and -ful2 in forgetful.
These strategies are highly problematic. First, the featural decomposition of word classes
is somewhat arbitrary; for instance, one can easily devise a system in which nouns and
verbs do go together. Second, it is frequently not clear whether one should assume
underspecification or homonymy. Farnese -fig, for example, can be combined with
nouns, adjectives or verbs, as in andlig- 'spiritual'(+--- and- 'spiritN' ), blidlig- 'kind'(+blio- 'kind A') and damlig- 'pleasant ' (+---dam- ' likev'; the obligatory inflectional affixes
are not shown). Should one set up three distinct affixes or only two - and if so, which?
Third, the assumption of homonymy is often not convincing because all the meanings
of the affix in question are closely related and are not fu lly determined by the word class
of the base (cf. Plag 2004: 213-2 14, 217-218 on English un-). Generally, less subjective
criteria for the distinction between polysemy and homonymy would be desirable (see
De Belder 2011: 132- 152 for a proposal).
A number of case studies (e.g., Plag 2004: 202-216 on English) lead to the conclusion that restrictions on the input to word-formation rules are primarily semantically
based and not word-class based. This is supported by the observation that a derivational
affix which prototypically takes bases of a particular word class may be app:tied to other
bases, too, as long as the output can be interpreted in a similar way. For example,
German adjectives in -bar, which are often used w ith the negative prefix un-, are normally based on transitive verbs, e.g., (un)trinkbar ' (un)drinkable' ( +--- trink-en 'drink-INF').
In 1990, an advertising campaign for reusable PET bottles introduced unkaputtbar ' unbreakable', from the informal adjective kaputt 'broken'. This was fo llowed by unplattbar
287
PA288
288

( +-- platt 'flat') for a bicycle tyre that cannot be punctured, and although the pattern was
originally deviant and could therefore serve as an attention-getting device, it may well
become productive and unremarkable in the course of time. Such developments are
amply attested (cf. also Eschenlohr 1999: 120- 122 on German er- and De Belder 2011:
191 - 195 on Dutch nonce formations). An important prerequisite for the success of unA-bar 'cannot be made A' was its close resemblance to un-V-bar 'cannot be V-ed'; at a
more abstract level, both instantiate the same pattern. Thus, word-formation patterns or
schemas form a hierarchy, and newly created compounds and derivatives need not follow
the most specific schema (see also article 12 on word-formation in construction grammar,
section 2).
The specification of roots for word class runs into difficulties when lexemes of different classes have identical roots, e.g., SAWN as in he plays music on a saw and SAWv
as in she sawed the fish with a knife . Many linguists take one of the lexemes as basic
and derive the other(s) by means of conversion or zero derivation (see section 5.9).
Leaving aside the diachronic question as to which lexeme is attested earlier, there are a
number of criteria for determining the direction of derivation (cf. Marchand 1964; Umbreit 2010: 308-314). The most important among them is semantic dependence as shown
by a paraphrase test: Since, so the argument goes, the meaning of the verb saw must be
explained with the help of the noun ('to cut with a saw') but not vice versa (one need
not describe a saw as 'instrument for sawing'), the noun is basic andl the verb derived
(cf. Marchand 1964: 12). In reality, it is often debatable which paraphrase is superior to
the other (cf. Umbreit 2010: 310) and typically both of them have their weaknesses. It
has also been noted that lexeme A is not necessarily semantically dependent on lexeme
B in all of its meanings (cf. Marchand 1964: 12; Plank 2010). For these and other
reasons, native speakers often disagree about which lexeme is derived from the other
(cf. Bergenholtz and Mugdan 1979b: 350- 352).
In many cases of alleged conversion, it is not necessary to posit a change of word
class at all. The Russian term for the grammatical subject, podleiascee (literally 'that
which underlies' ), is a typical example. It is the nominative singular neuter form of the
present active participle of podleia-t' 'underlie-INF', with adjectival inflection and the
syntactic functions of nouns (head of a noun phrase, etc.). On some accounts, the participle as a form of the verb is converted into an adjective and the adjective into a noun.
The first step is unnecessary because one can regard the participle as an adjective derived
from the verb (see section 3.4). The second step is unnecessary if we distinguish between
two kinds of word classes, morphological word classes (lexeme classes.) defined in terms
of inflectional paradigms and syntactic word classes defined in terms of syntactic functions (cf. Bergenholtz and Mugdan 1979a: 133- 141 ). Russian adjectives (lexemes inflected for gender, number and case) can generally serve the same syntactic functions as
nouns (lexemes with inherent gender, inflected for number and case) but not vice versa.
If conversion is admitted as a process of word-formation, it should involve a change in
lexeme class, e.g., from the adjective dobr-yj 'good-NOM.SG.M' (neuter dobr-oe with the
genitive dobr-ogo) to the abstract noun dobr-o ' good(N)-NOM.SG' (with the genitive
dobr-a), whereas the syntactic behaviour and the semantic specialization of podleiascee
do not involve the formation of a different lexeme.
An alternative to conversion or zero derivation is to give the lexemes in question
equal status by means of underspecification or multiple specification of the root (cf.
Eschenlohr 1999: 70- 71). For English saw, underspecification means that it is neither
PA290
290

Alderete, John, Jill Beckman, Laura Benua, Amalia Gnanadesikan, John McCarthy and Suzanne
Urbanczyk
1999
Reduplication with fixed segmentism. Linguistic inquiry 30: 327-364 [originally 1997
in Rutgers Optimality Archive, avai lable at http://roa.rutgers.edu/files/226-1097/roa-226alderete-l.pdf].
1988
Inflection. In: Michael Hammond and Michael Noonan (eds.), Theoretical Morphology.
Approaches in Modern Linguistics, 23- 43. San Diego, CA: Academic Press.
1992
A -Morphous Morphology. Cambridge: Cambridge Univers ity Press.
Aronoff, Mark
1976
Word Formation in Generative Grammar. Cambridge, MA/London: MIT Press.
Aronoff, Mark
1994
Morphology by Itself Stems and inflectional Classes. Cambridge, MA/London: MIT
Press.
Aronoff, Mark
2012
Morphological stems: What William of Ockham really said. Word Structure 5: 28- 51.
Aronoff, Mark and Nanna Fuhrhop
2002
Restricting suffix combinations in German and English: Closing suffixes and the monosuffix constraint. Natural Language and Linguistic Th eory 20: 451 - 490.
Augst, Gerhard
1975
Uber das Fugenmorphem bei Zusammensetzungen. In: Gerhard Augst, Untersuchungen
zum Morpheminventar der deutschen Gegenwartssprache, 71 - 155. Tubingen: Narr.
Banga, Arina, Esther Hanssen, Anneke Neijt and Robe1i Schreuder
2013
Preference for linking element -en- in Dutch noun-noun compounds: Native speakers
and second language learners of Dutch. Morphology 23: 33- 56.
Baudouin de Courtenay, Jan
1869
Wortformen und selbst satze, welche in der polnischen sprache zu stammen herabgesunken sind. Beitriige zur vergleichenden Sprachforschung 6(2): 204- 210 [volume dated
1870].
Bauer, Laurie
1983
English Word-formation. Cambridge: Cambridge University Press.
Bauer, Laurie
1988
A descriptive gap in morphology. In: Geert Booij and Jaap van Marie (eds.), Yearbook
of Morphology, 17- 27. Dordrecht/Providence, RI: Foris.
Bauer, Laurie
2000a Word. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.) , Morphology. An
international Handbook on Inflection and Word-Formation. Vol. l , 247- 257. Berlin/
New York: de Gruyter.
Bauer, Laurie
2000b Compounding. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-Formation. Vol. 1, 695707. Berlin/New York: de Gruyter.
Bauer, Laurie
2003
Introducing Linguistic Morphology. 2"d ed. Washington, DC: Georgetown Un iversity
Press [ P 1 ed. 1988].
Bauer, Laurie
2004
A Glossary of Morphology. Washington, DC: Georgetown University Press.
Bauer, Laurie
2006
Compound. In: Keith Brown (ed.), Ency clopedia of Language and Linguistics. Vol. 2,
7 l 9-726. 2"d ed. Amsterdam: E lsevier.
PA291
Becker, Thomas
2000
Metathes is. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphology. An international Handbook on Inflection and Word-Formation. Vol. 1, 576-581.
Bergenholtz, Henning and Joachim Mugdan
1979a Einfiihrung in die Morphologie. Stuttgart: Kohlhammer.
1979b Ist Liebe primar? Ober Ableitung und Wortarten. In: Peter Braun (ed.), Deutsche Gegenwartssprache. Entwicklungen, Entwiirfe, Diskussionen, 339- 354. Mi.inchern: Fink.
2000
Nullelemente in der Morphologie. In: Geert Booij, Christian Lehmann and Joachim
Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-Formation. Vol. l , 435-450. Berlin/New York: de Gruyter.
Bickel, Balthasar, Goma Banjade, Martin Gaenszle, Elena Lieven, Netra Prasad Paudyal, lchchha
Puma Rai, Manoj Rai, Novel Kishore Rai and Sabine Stoll
Free prefix ordering in Chintang. Language 83: 43- 73.
2007
Bickel, Balthasar and Johanna Nichols
2007
Inflectional morphology. In: Timothy Shopen (ed.), Language Typology and Syntactic
Description. Vol. 3: Grammatical Categories and the Lexicon, 169- 240 . 2nd ed. Cambridge: Cambridge University Press.
Bisang, Walter
200 1 Syntax/morphology asymmetry in Vietnamese - a consequence of contact by writing
between Vietnamese, Chinese and Standard Average European Languages. In: Birgit
Igla and Thomas Stolz (eds.), "Was ich noch sagen wollte ... ". A Multilingual Festschrift
for Norbert Boretzky on Occasion of His 65 1" Birthday, 189-201. Berlin: Akademie
Verlag.
Blevins, Juliette
Untangling Leti infixation. Oceanic Linguistics 38: 383- 403.
1999
Bloomfield, Leonard
1926
A set of postulates for the science of language. Language 2: 153-164.
Bloomfield, Leonard
1933
Language . New York: Holt [British edition London: Allen & Unwin 1935].
Boas, Franz
1911
Introduction. In: Franz Boas (ed.), Handbook of American Indian Languages. Part 1, 183. Washington, DC: Government Printing Office.
Bonami, Olivier, G illes Boye and Fran9oise Kerleroux
2009
L'allomorphie radicale et la relation flexion-construction. In: Bernard Fradin, Fran9oise
Kerleroux and Marc Plenat (eds.), Apen;us de morphologie dufrani;ais, 103- 125. SaintDenis: Presses Universitaires de Vincennes.
Bonet, Eulalia and Daniel Harbour
2012
Contextual allomorphy. In: Jochen Trommer (ed.), The M orphology and Phonology of
Exponence, 195- 235. Oxford: Oxford University Press.
Booij, Geert
Against split morphology. In: Geert Booij and Jaap van Marie (eds.), Yearbook of Mor1994
phology I993 , 27- 49. Dordrecht: Kluwer [dated 1993].
Booij, Geert
1996
Inherent versus contextual inflection and the split morphology hypothesis. In: Geert
Booij and Jaap van Marie (eds.), Yearbook of Morphology I995 , 1- 16. Dordrecht:
Kluwer.
Booij, Geert
1998
Phonological output constraints in morphology. In: Wolfgang Kehrein and Richard
Wiese (eds.), Phonology and Morphology of the Germanic Languages, 143- 163.
Ti.ibingen: N iemeyer.
291
PA292
292

Booij, Geert
2002
The Mo1phology of Dutch . Oxford: Oxford University Press .
Booij, Geert
2005
Compounding and derivation. Evidence for Construction Morphology. In: Wolfgang U.
Dressler, Dieter Kastovsky, Oskar E. Pfeiffer and Franz Rainer (eds.), Morphology and
its Demarcations. Selected Papers from the 11'" Morphology Meeting, Vienna, February
2004, 109-132. Amsterdam/Philadelphia: Benjamins.
Booij, Geert
2012
Th e Grammar of Words. An Introduction to Linguistic Morphology. 3rd ed. Oxford:
Oxford University Press [l st ed . 2005].
Bossong, Georg
2001
Ausdrucksmoglichkeiten fiir grammatische Relationen. In: Martin Haspelmath, Ekkehar d Konig, Wulf Oesterreicher and Wolfgang Raible (eds.), Language Typology and
Language Universals. Vol. I , 657- 668. Berlin/New York: de Gruyter.
Broselow, E llen
2000
Transfixation. In: Geert Booij, C hristia n Lehmann and Joachim Mugdan (eds .), Morphology. An International Handbook on Inflection and Word-Formation. Vol. 1, 552557. Berlin/New York: de Gruyter.
Brown, Dunstan
2011
Morphological typology. In: Jae Jung Song (ed.), Th e Oxford Handbook of Linguistic
Tj;pology, 487- 503. Oxford: Oxford University Press.
Bybee, Joan L. and Jea n E. Newman
1995
Are stem changes as natural as affixes? Linguistics 33: 633 - 654.
Bye, Patrik and Pe ter Svenonius
2012
Non-concatenative morphology as epiphenomenon. In: Jochen Trommer (ed.), The Morphology and Phonology of Exponence, 427- 495. Oxford: Oxford University Press.
Caballero, Gabriela
20 IO
Scope, phonology and morphology in an agglutinating language: Choguita Rara 'muri
(Tarahumara) variable suffix ordering. Morphology 20: 165- 204.
Carstairs[-McCarthy], Andrew
1988
Some implications of phonologically conditioned suppletion. In: Geert Booij and Jaap
van Marie (eds), Yearbook of Morphology, 67-94. Dordrecht/Providence, Rl: Fa ris.
Carstairs-McCarthy, [Andrew]
1998
[Contribution to] Discuss ion of the papers by Sadock and Ba ker. In: Steven G. Lapointe,
Diane K. Brentari and Patrick M. Farrell (eds.), Morphology and Its Relation to Phonology and Syntax, 213-2 15. Stanford, CA: CLSI.
Carstairs-McCarthy, Andrew
2000
Lexeme, word-form, paradigm. In: Geert Boo ij , Christian Lehmann and Joachim Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-Formation.
Vol. I , 595- 607. Berlin/New York: de Gruyter.
Chelliah, Shobhana L. and Willem J. de Reuse
2011
Handbook of Descriptive Linguistic Fieldwork. Dordrecht: Springer.
Chaves, Rui P.
2008
Linearization-based word-part ellipsis. Linguistics and Philosophy 31: 261 - 307.
C lements, G[eorge] N.
1985
The problem of transfer in nonlinear morphology. Cornell Working Papers in Linguistics
7: 38-73.
Coates, R icha rd
2000
Exponence. In: Geert Booij, C hristian Lehmann and Joachim Mugd!an (eds.), Morphology. An International Handbook on Inflection and Word-Formation. Vol. 1, 6 16- 630.
PA293
Croft, W illiam
2000
Lexical and grammatical meaning. In: Geert Booij , Christian Lehmann and Joachim
Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-Formation. Vol. 1, 257- 263. Berlin/New York: de Gruyter.
Daniels, Peter T.
2013
The history of writing as a history of linguistics. In: Keith A llan (ed.), The Oxford
Handbook of the History of Linguistics, 53- 69. Oxford: Oxford University Press.
De Belder, Mar ijke
2011
Roots and Affvces. Eliminating Lexical Categories from Syntax. Utrecht: LOT. Available
at http://www.lotpublications.nl/publish/articles/004300/bookpart.pdf.
Di Sciullo, Anna Maria and Edwin Williams
1987
On the Definition of Word. Cambridge, MA/London: MIT Press.
Dixon, R[obert] M. W. and A lexandra Y. Aikhenvald
2002
Word: A typological framework. In: R. M. W. Dixon and Alexandra Y. Aikhenvald
(eds.), Word. A Cross-Linguistic Typology, 1- 41. Cambridge: Cambridge University
Press.
Donalies, Elke
2005
Die Wortbildung des Deutschen. Ein Oberblick. 2nd ed. Tilbingen: Narr [1 st ed. 2002].
2000
Subtraction. In: Geert Booij, Christian Lehmann and Joachim Mugdan (ed!s.), Morphology. An international Handbook on Inflection and Word-Formation. Vol. 1, 581 - 587.
Eschenlohr, Stefanie
1999
Vom Nomen zum Verb. Konversion, Priifigierung und Riickbildung im Deutschen . Hildesheim: Olms.
Fabb, Nigel
1998
Compounding. In: Andrew Spencer and Arnold M. Zwicky (eds.), Th e Handbook of
Morphology, 66-83. Oxford: Blackwell.
Fleischer, Wolfgang and lrmhild Barz
1992
Wortbildung der deutschen Gegenwartssprache. Tilbingen: Niemeyer.
2012
Wortbildung der deutschen Gegenwartssprache. 4th ed. Berlin/Boston: de Gruyter.
Fromkin, Victoria, Robert Rodman and Nina Hyams
2013
An Introduction to Language . I 01h ed. Independence, KY: Cengage Learning [dated
2014, pt ed. 1974].
Fuhrhop, Nanna
1998
Grenzfiille morphologischer Einheiten. Tilb ingen: Stauffenburg.
Gafos, Diamandis
1998
Elim inating long-distance consonantal spread ing. Natural Language and Linguistic
Theory 16: 223 - 278.
Gal lmann, Peter
1999
Fugenmorpheme als Nicht-Kasus-Suffixe. In: Matthias Butt and Nanna Fuhrhop (eds.),
Variation und Stabilitat in der Wortstruktu1: Untersuchungen zu Entwicklung, Erwerb
und Varietiiten des Deutschen und anderer Sprachen, 177- 190. Hildesheim: O lms.
Garvin, Paul L.
1945
Pure-relational suffixes and postpositions in Hungarian. Language 2 1: 250- 255.
Giegerich, He inz J.
The English compound stress myth. Word Structure 2: 1-17 .
2009
Good, Jeff and Alan C. L. Yu
2005
Morphosyntax of two Turkish subject pronominal paradigms. In: Lorie Heggie and Francisco Ord6fiez (eds.), Clitic and Affix Combinations. Th eoretical Perspectives, 3 15- 341.
Amsterdam/Philadelphia: Benjamins .
293
PA295
Hockett, Charles F.
1950
Peiping morphophonemics. Language 26: 63-85.
Hockett, Charles F.
1954
Two models of grammatical description. Word JO: 210- 234.
Hockett, Charles F.
1958 A Course in Modern Linguistics. New York: Macmillan.
Hockett, Charles F.
1961
Linguistic elements and their relations. Language 37: 29-53.
Hoeksema, Jack and Richard D. Janda
1988
Implications of process-morpho logy for categorial grammar. In: Richard T. Oehrle, Emmon Bach and Deirdre Wheeler (eds.), Categorial Grammars and Natural Language
Structures, 199-247. Dordrecht: Reidel.
Hogoboom, S. L. Anya
2003
Definition of the base. Proceedings of the[ ... } Annual Meeting of the Berkeley Linguistics Society 29(1): 173-183.
Hyman, Larry M. and William R. Leben
2000
Suprasegmental processes. In: Geert Booij, Christian Lehmann and Joachim Mugdan
(eds.), Mo1phology. An International Handbook on Inflection and Word-Formation. Vol.
I, 587- 594. Berlin/New York: de Gruyter.
Inkelas, Sharon
2012
Reduplication. In: Jochen Trammer (ed.), Th e Morphology and Phonology of Exponence, 355- 378. Oxford: Oxford University Press.
Jackson, Howard
2007
Key Terms in Linguistics. London/New York: Continuum.
Jacobs, Joachim
2011
Grammatik ohne Worter? In: Stefan Engelberg, Anke Holler and Kristel Proost (eds.),
Sprachliches Wissen zwischen Lexikon und Grammatik, 345-372. Berlin/Boston: de
Gruyter.
Julien, Marit
2002
Syntactic Heads and Word Formation. Oxford: Oxford University Press.
Julien, M[arit]
2006
Word. In: Keith Brown (ed.), Encyclopedia of Language and Linguistics. Vo l. 13, 617624. 2nd ed. Amsterdam: Elsevier.
Katamba, Francis and John Stonham
2006
Morphology. 2nd ed. Basingstoke/New York: Palgrave Macmillan [ts' ed. by Francis
Katamba 1993].
Kibort, Anna
2010
Towards a typology of grammatical features. In: Anna Kibort and Greville G. Corbett
(eds.), Features. Perspectives on a Key Notion in Linguistics, 65- 106. Oxford: Oxford
University Press.
Korotkova, Natalia and Yury Lander
2010
Deriving affix ordering in polysynthesis: Evidence from Adyghe. Morphology 20: 299319.
Kosa, Loredana Andreea
2008
An argument for process-based morphology. Subtractive morphology in Tohono O 'odham (Uta-Aztecan). M.A. Thesis, Department of L inguistics, Simon Fraser University.
Available at http://summit.sfu.ca/item/8989 .
Kubrjakova, Elena S.
2000
Submorphemische Einheiten. Obersetzung und Bearbeitung Joachim Mugdan. In: Geert
Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphology. An International
295
PA296
296

Handbook on Inflection and Word-Formation . Vol. L, 417- 426 . Berlin/New York: de
Gruyter.
Lausberg, Heinrich
1953
[Brief notice on Malkiel 1949 .] Romanische Forschungen 65: 229- 230 .
Lawrenz, Birgit
2006
Moderne deutsche Wortbildung. Phrasale Wortbildung im Deutschen. Linguistische Untersuchung und sprachdidaktische Behandlung. Hamburg: Kovac.
Lehmann, Christian
Interlinear morphemic glossing. In: Geert Booij, Christian Lehmann, Joachim Mugdan
2004
and Stavros Skopeteas (eds.), Morphology. An International Handbook on Inflection and
Word-Formation. Vol. 2, 1834-1857. Berlin/New York: de Gruyter.
Lehmann, C hristian
2008
Roots, stems and word classes. Studies in Language 32: 546- 567.
L ieber, Rochelle
2000
Substitution of segments and features. In: Geert Booij, Christian Lehmann and Joachim
Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-Formation. Vol. 1, 567-576. Berlin/New York: de Gruyter.
L ieber, Roche lle
Introducing Morphology. Cambridge: Cambridge University Press.
2010
L indner, Thomas
2012
Jndogermanische Grammatik, Band JV/I. Komposition. Lieferung 2 [pages 71-1 48].
Heidelberg: Winter.
Lyons, John
Introduction to Theoretical Linguistics. London: Cambridge University Press.
1968
Maes, V[edaste]
1959
Dictionnaire ngbakafranr;ais-neerlandais precede d'un aperr;u grammatical. Tervuren:
Musee Royal du Congo Beige.
Malkiel, Yakov
1949
Studies in the Hispanic infix -eg-. Language 25: 139-181.
Malkiel, Yakov
1966
Genetic analysis of word formation. In: Thomas A. Sebeok (ed.), Current Trends in
Linguistics. Vol. 3: Theoretical Foundations, 305- 364. The Hague/Paris: Mouton.
Manova, Stela and Mark Aronoff
2010
Modeling affix order. Morphology 20: 109- 13 1.
Marchand, Hans
1964
A set of criteria for the establishing of derivational relationship between words unmarked by derivational morphemes. Indogermanische Forschungen 69: 10- 19.
Martinet, Andre
1960
Elements de linguistique generate. Paris: Colin [later editions with different pagination].
Marusic, Franc
2003
*A.ff-STEM-ix : On discontinuous morphology. Available at http://www.ung.si/- fmarusic/
pu b/marusic_ 2003 _ cirkumfixi.pdf.
Matthews, P [eter] H.
1972
Inflectional Morphology. A Theoretical Study Based on Aspects of Latin Verb Conjugation. Cambridge: Cambr idge University Press.
Matthews, P [eter] H .
1974
Morphology. An Introduction to the Theory of Word-Structure. Cambridge: Cambridge
University Press.
Matthews, P[eter] H.
1991
Morphology. 2"d ed. Cambridge: Cambridge Un iversity Press.
PA297
McArthur, Tom
1998
Concise Oxford Companion to the English Language. Online edition . Oxford: Oxford
University Press. Available at http://www.oxfordreference.com/view/ 10.1093/acref/
9780 192800619.001.0001/acref-9780192800619 [last access 10 Sept 20 14].
Meder, Gregor and Joachim Mugdan
1990
Alie reden von Hiiufigkeit .... Anmerkungen zum Thema ,,Frequenz" in der Morphologie. Tn: A[rmin] Bassarak, D[agmar] Bittner, A[ndreas] Bittner a nd P[etra] T hie le (eds.),
Wurzel(n) der Natiirlichkeit. Studien zur Morphologie und Phonologie IV, 87- 108. Berlin: Zentralinstitut fur Sprachwissenschaft.
Mel'cuk, I[gor'] A.
0 "vnutrennej fleksii" v indoevropejskich i semitskich jazykach. Voprosy jazykoznamja
1963
12(4): 27- 40.
Mel'cuk, l [gor] A.
1982
Towards a Language of Linguistics. Ed. by Ph[ilip] Luelsdorff. Miinchen: Fink.
Mel'cuk, Igor
Cours de morphologie generate (theorique et descriptive). Vol. 3/3: Moyens morphologi1996
ques I Quatrieme partie. Syntactiques morphologiques. Montreal: Presses de l 'U niversite
de Montreal; [Paris:] CNRS Editions.
Mel'cuk, Igor
Cours de morphologie generate (theorique et descriptive). Vol. 4/5: Cinquieme partie.
1997
Signes morphologiques. Montreal: Presses de l' Universite de Montreal; [Paris:] CNRS
Editions.
Mel'cuk, Igor
2000
Morphological processes. In: Geert Booij, Christian Lehmann and Joachim Mugdan
l , 523- 535. Berlin/New York: de Gruyter.
Mel'cuk, Igor
Aspects of the Theory of Morphology. Ed. by David Beck. Berl in/New York: Mouton de
2006
Gruyter.
M ithun, Marianne
1984
The evolution of noun incorporation. Language 60: 847- 894.
Montermini, Fabio
II lato sinistro della morfologia. La prefissazione in italiano e nelle lingue de/ mondo.
2008
Milano : FrancoAngeli.
Montermini, Fabio
20 IO
Units in compounding . In: Sergio Scalise and Irene Vogel (eds.), Cross-Disciplinary
Issues in Compounding, 77- 92. Amsterdam/ Philade lphia: Benjamins.
Moravcsik, Edith A.
2000
Infixation. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphology.
An International Handbook on Inflection and Word-Formation . Vol. I, 545 - 552. Berlin/
Motsch, Wolfgang
Deutsche Wortbildung in Grundziigen . Berlin/New York: de Gruyter.
1999
Mugdan, Joachim
1984
Grammatik im Worterbuch. Wortbildung. In: Herbert Ernst Wiegand (ed.) , Studien zur
neuhochdeutschen Lexikographie V, 237- 308. Hildesheim: Olms.
Mugdan, Joachim
1986
Was ist eigentlich ein Morphem? Zeitschriftfar Phonetik, Sprachwissenschaft und Kommunikationsforschung 39: 29- 43.
Mugdan, Joachim
Morphological units. In: R [onald] E. Asher (ed.), The Encyclopedia of Language and
1994
Linguistics. Vol. 5, 2543- 2553. Oxford: Pergamon Press.
297
PA299
Plank, Frans
20 IO
Variable direction in zero-derivation and the unity of polysemous lexical items. Word
Structure 3: 82- 97.
Plungian [= Plungja n], Vladimir A.
200 1
Agglutination and flection. In: Martin Haspelrnath, Ekkehard Konig, Wulf Oesterreicher
and Wolfgang Raible (eds.), Language Typology and Language Universals. Vol. 1, 669678. Berlin/New York: de Gruyter.
Plungjan, Vladimir A.
2003
ObScaja mmfologija. Vvedenie v problematiku. 2nd ed. Moskva: URSS [l st ed. 2000] .
Postal, Paul M.
1969
Anaphoric islands. Papers from the Regional Meeting of the Chicago Linguistic Society
5: 205-239.
Raimy, Eric
2011
Reduplication. In: Marc van Oostendorp, Colin J. Ewen, Elizabeth Hume and Keren
Rice (eds.), Th e Blackwell Companion to Phonology. Vol. 4: Phonological Interfaces,
2383- 24 13. Chichester: W iley-Blackwe ll.
Rainer, Franz
Inflection inside derivation: Evidence from Spanish and Portuguese. In: Geert Booij and
1996
Jaap van Marie (eds.), Yearbook of Morphology 1995, 83- 9 l. Dordrecht: Kluwer.
Ricca, Davide
2005
Cumulative exponence involving derivation: Some patterns for an uncommon phenomenon. In: Wolfgang U. Dressler, Dieter Kastovsky, Oskar E. Pfe iffer a nd Franz Rainer
(eds.), Morphology and Its Demarcations. Selected Papers from the 11'h Morphology
Meeting, Vienna, February 2004, 197-2 13. Amsterdam/Philadelphia: Benjamins.
Rice, Keren
20 11
Principles of affix ordering: An overview. Word Structure 4: 169-200 .
Rubba, Johanna
200 l
Tntrotlection. In: Martin Haspel math, Ekkehard Konig, WulfOesterreicher and Wolfgang
Raible (eds.), Language Typology and Language Universals. Vol. 1, 678- 694. Berlin/
Ryan, Kevin M.
20 IO
Variable affix order: Grammar and learning. Language 86: 7 58-79 l.
1916
Cours de linguistique generate. Publie par Charles Bally et Albert Sechehaye avec la
collaboration de A lbert Riedlinger. Lausanne/Paris: Payot [later editions with different
pagination].
Scalise, Sergio
1984
Generative Morphology. Dordrecht/Cinnaminson, NJ: Foris.
Schiering, Rene, Balthasar Bickel and Kristine A. Hildebrandt
The prosodic word is not un iversal, but emergent. Journal of Linguistics 46: 657- 709.
2010
Schpak-Dolt, Nikolaus
20 IO
Einfiihrung in die franzosische Morphologie. 3rd ed. Berlin/New York: de Gruyter [l st
ed. 1992].
Spencer, Andrew
Morphophonological operations. In: Andrew Spencer and Arnold M . Zwicky (eds.), The
1998
Handbook of Morphology, 123- 143. Oxford: Blackwell.
Spencer, Andrew
2012
Identifying stems. Word Structure 5: 88- 108.
Sternberger, Joseph P. and Barbara Handford Bernhardt
1999
Contiguity, metathesis, and infixation. In: Kimary Shahin, Susan Blake and Eun-Sook
299
PA300
300

Kim (eds.), The Proceedings of the Seventeenth West Coast Conference on Formal Linguistics, 610-624. Stanford, CA: Cente r for the Study of Language and Information.
Stolz, Thom as, Cornelia Stroh and Ania Urdze
Total Reduplication. The Areal Linguistics of a Potential Universal. [Berlin]: Akademie
2011
Verlag.
Stump, Gregory T.
200 La Inflectional Morphology. A Theory of Paradigm Structure. Cambridge: Cambridge University Press.
Stump, Gregory T.
2001b Affix position. In: Martin Haspelmath, Ekkehard Konig, Wulf Oesterreicher and Wolfgang Raible (eds.), Language Typology and Language Universals . Vol. 1, 708- 7 14.
Stump, G[regory] T.
2006
Template morphology. In: Keith Brown (ed.), Encyclopedia of Language and Linguistics. Vol. 11, 559- 562. 2nd ed. Amsterdam: Elsevier.
Trager, George L.
1948
Taos I: A language revisited. International Journal of American Linguistics 14: 155160.
Umbreit, Birgit
2010
Does love come from to love or to love from love? Why lexical motivation has to be
regarded as bidirectional. In: Alexander Onysko and Sascha Michel (eds.), Cognitive
Perspectives on Word Formation , 301 - 333. Berlin/New York: De Gruyter Mouton.
U ssishkin, Adam
2007
Morpheme position. In: Paul de Lacy (ed.), The Cambridge Handbook of Phonology,
457-472. Cambridge: Cambridge University Press.
van Lier, Eva
2012
Reconstructing multifunctionality. Theoretical Linguistics 38: 119- B 5.
Vardul ', I[gor'] F.
1991
Morfema v tipolog iceskom issledovanii. In: I. F. Vardu l' (ed.), Morfema i problemy
tipologii, 6-85 . Moskva: Nauka.
Volodin A [lelksandr] P. and V[iktor] S. Chrakovskij
1991
Tipologija klassifikacij morfem. In: I[gor'] F. Vardul' (ed.), Morfema i problemy tipologii, I 08- 134. Moskva: Nauka.
Wallis, Ethel
1956
S imulfixation in aspect markers of Mezquita I Otomi. Language 32: 453-459 .
Wiese, R ichard
1996a The Phonology of German . Oxford: Oxford University Press.
Wiese, Richard
1996b Phrasal compounds and the theory of word syntax. Linguistic Inquiry 27: 183- 193.
Wiltshire, Caro line and Alec Marantz
2000
Reduplication. In: Gee1t Booij, C hristian Lehmann and Joachim Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-Formation . Vol. 1, 557567. Berlin/New York: de Gruyter.
Wurzel, Wolfgang Ullrich
1989
Inflectional Morphology and Naturalness . Translated by Manfred Schentke. Berlin:
Akademie Verlag; Dordrecht: Kluwer [originally German 1984].
Yu , A lan C. L.
2007
A Natural Histmy of In.fixation . Oxford: Oxford University Press.
Zwicky, Arnold M.
How to describe inflection. In: Mary Niepokuj, M . V. Clay, V. Nikiforidou and D. Feder
1985
(eds.), Proceedings of the JI'" Annual Meeting of the Berkeley Linguistics Society, 372386. Berkely, CA: Berkeley Linguistics Society.
PA303
16. Derivation
practice. At the end of the artic le I briefly di scuss evaluative morphology, which appears
to create pragmatically modified versions of a lexeme rather than a new lexeme as such.
Although derivation typically creates a new lexeme, it is difficult to u se this as a
hard-and-fast criterion because there is no way of deciding when two words constitute
distinct lexemes as opposed to two usages of one and the same lexeme. This can be
thought of as the central problem of lexicology. In particular, it is often difficult to know
whether a new lexeme has been created when new meanings are coerced. For instance,
in English it is often possible to coerce a count meaning from a mass term (and hence
allow that term to be pluralized: three coffees meaning ' three cups of coffee', the "Universal Packager"), or to coerce a mass meaning from a count noun (there's egg on your
tie, the "Universal Grinder"). However, this coercion relation is limited to specific lexemes, in general, and is therefore not a fully productive part of the language. It is
therefore difficult to know whether we are dealing with different allowable w ays of using
a single lexeme (mass/count 'coffee', 'egg' ) or whether we have a lexically restricted
derivational category expressed by conversion giving rise to systematic polysemy (mass
'coffee' - count 'coffee'). Similar problems arise, for instance, with intransitive/transitive verb alternants such as BREAK: The vase broke vs. Tom broke the vase and a whole
host of other lexical relations (see also article 17 on conversion).
In this connection we must bear in mind that in some respects polysemous variants
of a lexeme very frequently behave like distinct lexical entries with respect to wordformation. To give just one simple example, from the adjective poor we cam create the
property noun poverty but only where poor means ' not rich (in something)', either concrete (the poverty of the unemployed) or abstract (the poverty of the stimulus). Where
poor means just 'bad' we have to use the nominal form poorness: the poorness!*poverty
of the team's performance. This suggests we have two distinct lexemes (see also Plank
2010 on Germanfett 'fat'). Abundant examples can be found in deverbal actional nominals. Thus, commit (a crime) gives commission, commit (to trial) gives committal whi le
commit (oneself to something) gives commitment (Carstairs-McCarthy 1992a). It may
turn out that we can use such derivational patterns as a test of lexemehood.
2.
Formal aspects: how derivation 1s rea lized
2.1. Stems
It is customary to define the root of the base lexeme as the starting point for derivational
processes. However, in practice, we usually find that the interesting generalizations about
derivation (and inflection) are defined over sets of stems (the root is often the default
stem form). Traditionally, a stem is what remains when we remove inflectnons, though
this definition hinges on us being able to distinguish inflectional formatives from stem
formatives (see Spencer 2012 on segmenting stems and inflections).
In the descriptive tradition a stem is often associated with a meaning or inflectional
property ('the oblique case stem', ' the imperfect stem ' and so on). However, Aronoff
(1994; see also article 44 on paradigmatically determined allomorphy) introduces the
notion of ' morphomic stem'. This is a form that does not realize a meaning or morphosyntactic property in itself, and which is used as the basis for inflectional, derivational
303
PA304
304

and sometimes compounding morphology. Aronoff illustrates this notion with the third
stem in Latin. A regular 1st conjugation verb (AMO ' I love') has a first ('imperfect')
stem ending in the conjugation theme vowel (ama:) and a second ('perfect') stem derived by adding -v to the first stem: ama:-v-. For some verb classes there are irregularities
and subregularities. For example, the 3rd conjugation REGO (reg-) ' I rule, govern',
lengthens the root vowel and adds -s: re:ks-i: (= re:gsi:) ' I ruled'.
Latin verbs also distinguish a third stem in -t: ama:-t '(having been) loved'. This
stem is not necessarily passive in meaning, however, because it is the base for another
stem, that of the future active participle in -u:r-: ama:-t-u:r-(us) 'being about to love
(-M.SG.NOM)'. This derivation is found with irregular verbs, too: e.g., the suppletive:
FERO 'I carry', infinitive ferre, 3rd stem la:t- ' carried', future active participle la:tu:rus.
It would be incorrect to think of the third stem as purely inflectional, however, because
it is the basis of several derivational processes, including the agent nominal in -or:
ama:re ----+ ama:t-or 'lover', an event nominal in -io(n-): cogita:re ' to think' ----+ cogita:tio(n-) 'thought' and an iterative verb in -it: scri:bere > script- > scriptit- 'write often'.
It is sometimes thought (mistakenly) that stems are typically irregular in form (partly
because discussion tends to focus on irregular stems) but the three Latin stems are entirely regular for regular verbs: [root + theme vowel], [root + theme vowel+ w -] and [root
+ theme vowel + t]. On the other hand, in many languages stems are irregular in the
important sense that they consist of two or more discontinuous parts, whose combination
is unpredictable. Typically, this is found with verbs and is the result of grammaticalizing
a verb root with an adverbial, an incorporated root or whatever. Athapaskan languages
illustrate this pattern (see Spencer 1991: 208-211 for examples from Navajo) as does
the very distantly related Siberian (Yenisseian) language, Ket. Verbs in Ket can become
very complex, being extended by a host of prefixes. Here I follow the position class
description in Vajda (2004: 45), shown in (1 ). (Vajda calls agreement markers "valence"
markers. For a more complex analysis of Ket position classes see Werner 1997.)
(1)
Ket position class template

P8
agr
P7
mcorporate
P6
agr
PS
adpos
P4
dur/
or agr
P3
agr
P2
past/
imp
p]
agr
PO
base
P-1
agr
The crucial positions are PO, the position of the basic root, P5 and P7. All but about 100
simplex verbs, and all productive patterns, require an overt element in position P7. This
may be a productively incorporated element, for instance a direct object, an instrument,
or a meaningless element, sometimes accompanied by a (meaningless) element (noted
as "L") in P5. In some words none of the components of the stem can be assigned any
meaning, as in (2), where "MS" is a "morphotactic separator", a meaningless connector,
and the P7, PO elements are both cranberry morphs:
(2)
deqsaq
di 8
1SG.SUBJ
' I hear'
eq7
(s)
L7
(MS)
PA305
16. Derivation
2.2. Morphologica l devices realizing derivation

In addition to standard patterns of affixation, a number of European languages (especially
English) employ conversion (see article 17). Infixation is uncommon (though it seems
to have been an exponent of derivation in earlier stages of lndo-European). Reduplication
(see article 25) is found, though not in the derivational morphology of Inda-European
languages. (However, a type of reduplication process has entered conversational French
evaluative morphology: bete 'stupid' - bebete 'a bit stupid'.) Turkic languages and
languages of other groups influenced by Turkic often have full or partially reduplicated
forms akin to English bye-bye or higgeldy-piggeldy, what is sometimes referred to as the
mz.Jma.J-type, after Turkish (see, for instance, article 198 on Aghul and article 200 on
Khinalug, which explicitly points out the influence of Azeri). Such formations are almost
always onomatopoeic or evaluative/expressive in character. Unlike many language
groups outside Europe we don't find, say, reduplication of the first syllable of a root to
express prototypical derivational relations such as subject nominalization or causative
verb in typical European languages. However, in some Daghestanian languages reduplication does have a genuinely derivational function, for instance, verb root reduplication
to express aktionsarten or aspectual categories in Tsezic languages.
In inflection, languages may deploy non-affixational or non-concatenative formal devices. These include ablaut/apophony (change in the root vowel), consonant mutation
(change usually in the initial or final consonant(s) of the root), and prosodic alternations,
including alternations in stress, accent, or tone, segment length (gemination/ lengthening
or shortening of consonants or vowels), word-domain prosodies such as nasanity, palatality, or advanced tongue root. Non-affixational devices seem to be less common and less
systematic in (canonical) derivation, but it is not uncommon to find such devices being
used as part of the systematic morphology. For instance, in English there are noun-verb
pairs related only by stress shift: a contrast ~ to contrast, transport ~ to transport. This
only affects disyllabic verbs which are etymologically of the form prefix-stem and it
reflects a tendency for disyllabic nouns to be forestressed (syntax) and disyllabic verbs
to be endstressed (relax).
In a number of language groups we see non-concatenative morphology, in which
morphological properties are realized by changing the disposition or the length of vowels
and consonants in the root rather than by affixation. The most well-known systems are
often referred to as 'root-and-pattern' morphology (also called "transfixation" by Bauer
1988). It is in the Afro-Asiatic languages and especially the Semitic languages such as
Hebrew and especially Arabic that we see well-developed systems of this kind. The
system is described in any handbook of Arabic (and many textbooks of morphology)
but a convenient summary of Modern Standard Arabic (with transcribed examples) is
presented in Kaye (2007). Of the languages of Europe, Maltese (see article 183) preserves a good many of these patterns.
Derivational relations are often expressed by multi-word expression (see article 24)
and in the course of grammaticalization the grammatical components of such expressions
often become more affix-like. Thus, Toivonen (2003) argues that in English the particle
of particle-verb constructions (see also article 23 and article 35 on particle verbs in
Germanic) such as turn on or throw out has become a suffix when it appears after the
verb and before the direct object (turn# on the light, where # represents an affix boundary). In post-verb position, however, (turn the light on) the particle is a non-projecting
305
PA306
306

word, that is, a word which cannot form the head of a multi -word phrase. If this analysis
is correct it shows that we have a morph which is sometimes an affix and sometimes a
syntactically independent form. This complicates the morphological definition of "derivation" somewhat.
A particularly interesting set of cases is represented by preverbs in languages such as
German and especially Hungarian (Ackerman and LeSourd 1997, article 35 on particle
verbs in Gennanic and article 3 7 on particle verbs in Hungarian) in which the separable
preverb can attach as a prefix to the verb stem and the whole construction can then
undergo further derivation. For example, from Hungarian old 'to dissolve' we obtain
meg-old- 'to solve (a problem)' (with the (meaningless) preverb meg-), from which we
can derive the adjective megold-hatatlan 'unsolvable' (Ackennan and LeSourd 1997:
89).
Even in languages in which the preverb is always fixed to the verb stem morphologically we find that the preverb may not cohere as tightly to the base as we would expect
from a bona fide prefix. Languages such as Greek and Sanskrit exhibit interesting instances of inflection between the preverb and the verb. The so-called Greek augment is a
case in point (see also a1iicle 172). In (3) we see examples of present and past tense
verb forms (Holton, Mackridge and Philippaki-Warburton 1997: 161 f..):
(3)
Prefixation in Greek verbs

Present
grafo
'write'
epivalo
'impose'
sim-veni
'happen'
ep-ana-lamvano 'repeat'
Past Imperfective
e-grafa
ep-e-vala
.
.
sm-e-v1
ep-an-e-lava
The tense prefix is e- but the derivational prefixes epi-, sim-, ana- appear external to
that tense marker. This is an instance of what Stump (2001: chapter 4) refers to as headmarking morphology, because the inflections effectively ignore the non-head, prefixal,
part of the representation.
Another type of word-formation that is not mediated through affixes as traditionally
defined is iilustrated by verbs with reflexive pronominal clitics. In a number of Slavic
languages weak pronominal forms, including reflexives, are realized as second position
(Wackernagel) clitics. As in many of the world's languages, reflexive verbs have a generally detransitivizing function, but the meanings of the reflexive can be very varied,
ranging from a straightforward pronominal object (The children washed themselves) to
no discernible meaning change whatever. One familiar function is that of a passive voice
as seen in (4) from Czech, where the verb kosit 'to mow ' is used in an imperfective
passive sense (the examples are adapted from Smilauer 1972):
(4)
Obili se
kosi
corn REFL mow.3sG

'The corn is being mown. '
This represents word-formation to the extent that you believe that a regular passive
alternation is a species of derivation. A very common use of reflexive forms is to detransitivize causatives:
PA307
307
16. Derivation
(5)
dvefe
a. Zavirat
close.INF.IPFV doors.Ace
'to close the doors'
Czech
b. Dvefe
se
zaviraji
Doors.NOM REFL close
' The doors are closing.' [warning on Prague underground railway]
This process can give rise to reflexiva tantum interpretations, as in (6c) compared with
(6a, b):
(6)
a. Zlomil
vetev
broke.M.SG branch
'He broke the branch.'
Czech
b. Vetev
se
zlomila
branch(F) REFL broke.F.SG
' The branch broke.'
se
mu
hlas
c. Zlomil
broke.M.SG REFL him.DAT voice(M)
' His voice broke. '
Another use of reflexive forms common in Slavic is to create a kind of intensive aktionsart (often accompanied by prefixation). Such constructions can sometimes be difficult to
render in languages such as English, cf. the Czech examples in (7):
(7)
spat ' to sleep' ~ vyspat se ' to sleep off, get a good night's sleep'
de/at 'to do, make ' ~ nadelat se nepfijemnosti 'to create a lot of unpneasantness'
There are a good many reflexiva tantum in Czech, including some very common verbs
such as bat se (+genitive) 'to fear, be afraid of', smat se ' to laugh ', ptat se ' to ask', zdat
se 'to dream (of); seem '. In some cases, or in some meanings, the reflexive is optional,
as with blyskat (se) 'to flash (e.g., lightning)', or zaCit (se) ' to begin' and skoncit (se)
'to finish'.
The examples so far have all illushated the accusative form of the reflexive clitic
pronoun, se. Czech also has a dative reflexive form, si, and this, too, participates in
word-formation. An example of a dative reflexivum tan tum verb is vsimat si 'to notice',
which takes a genitive case complement.
3. Syntagmatic and paradigmatic aspects of derivational processes

By "syntagmatic aspects of derivation" I mean principally the order of stems and affixes.
That aspect has been subject to a good deal of discussion. Rice (2011) provides a very
good summary of recent research into the determinants of affix ordering cross-linguistically in inflection and also in derivation. Paradigmatic relations are normally discussed
in relation to inflection, but some linguists have argued that they can be discerned in
PA308
308

derivation, too. Finally, there are phenomena that have both syntagmatic and paradigmatic aspects, which I discuss briefly in the final subsection.
It is often said that derivation is "recursive". What is meant by this is that a wordformation rule gives rise to a lexeme and it is lexemes which are the base of wordformation rules, so the output of such a rule can serve as the input to another rule. Thus,
we can find chains of the kind [N cipher] ~ [v de [N cipher]] ~ [A [ v de [N cipher]]
able] ~ [A in [A [v de [N cipher]] able]] ~ [N [A in [A [v de [N cipher]] abil] ]ity].
However, there are generally phonological, grammatical and semantic restrictions on the
applicability of a given word-formation process (see Part V of this handbook) so the
recursion is not as general as in the case of syntax. Some languages have affixes which
are obligatorily word final (see article 54 on closing suffixes and article 55 on closing
suffix patterns in Russian).
3.1. Affix order and scope

Cross-linguistically, we often find claims that semantic scope determines affix order.
However, in European languages there are fewer clear instances of scope-driven ordering. Here I will look briefly at two recent studies which argue for scopal differences in
ordering, one for suffixes, the other for prefixes.
Kiefer and Koml6sy (2011) discuss the behaviour of the deverbal causative (-(t)At)
and factitive (-tAt) suffixes, the diminutive/iterative suffix (-gAt) (actually, a better term
would be "attenuative", since this is not an instance of evaluative morphology), and the
possibility suffix (-hAt) in Hungarian. Kenesei, Yago and Fenyvesi (1998: 359) treat
the causative and factitive suffixes in Hungarian as a single causative suffix. However,
Kiefer and Koml6sy argue that they should be distinguished because they have different
syntax. Their causative only attaches to non-agentive verbs (hence, onEy to intransitives)
behaving rather like a transitivizer, while their factitive behaves more like a standard
causative suffix and can causativize transitives. The two can be combined in one word in
the order stem < causative < factitive: mozg-at-tat 'to make somebody move something'.
Likewise, the factitive can combine with a deadjectival causative verb formed with -it:
nagy-it 'to make bigger, enlarge', nagy-it-tat 'to let make bigger, let enlarge'.
The diminutive/iterative suffix can combine with verbs derived with the causative
and/or factit ive suffix. Either order of diminutive/iterative and factitive is possible, depending on the sense: from olvas 'to read' we have olvas-tat-gat 'from time to time
(gat) it is true that x lets (tat) y read' vs. olvas-gat-tat 'to let read from time to time'.
As is generally the case cross-linguistically, the affix furthest from the stem takes the
nearer affix in its scope: olvas-tat-gat = [[[READ]-CAUSE]-ITERATIVE], olvas-gattat = [[[READ]-ITERATIVE]-CAUSE].
The possibility suffix -hAt behaves rather differently. This suffix always seems to
come finally in the word and fails to take most of the verbal inflections expected of a
verb stem. Many grammarians treat it as (effectively) inflectional (see, for instance,
Kenesei, V ago and Fenyvesi 1998: 3 59). Kiefer and Koml6sy (2011 ), however, suggest
that this means that the suffix is not (prototypically) inflectional and not (prototypically)
derivational.
Bulgarian, like other Slavic languages, has several layers of verb prefixation. Markova (2011) draws the traditional distinction between lexical prefixes, which create entirely
PA309
16. Derivation
new verb lexemes and other types. For instance, from the root znaja 'to know' we have
the verbs po-znaja 'to be acquainted with; guess, tell', su-znaja 'to realize', u-znaja 'to
learn, find out', pri-znaja 'to acknowledge, recognize; confess, admit' (Markova 2011:
247; examples given in 3sg present tense form). The semantic contribution of these
prefixes is generally non-compositional. However, there are other kinds of prefix which
impart aspectual or aktionsart meanings ("supra-lexical prefixes"). Markova argues that
there are two sorts, inner and outer (aspectual) prefixes. The inner prefixes modify the
verb's argument structure by specifying a spatial/locative, or a causative or a quantificational meaning. Examples of the spatial/locative prefixes are: bjaga 'to run' vs. do-bjaga
'to run to (place)'; pluva (v rekata) 'to swim (in the river) ', intransitive vs. pre-pluva
(rekata) 'to swim (the river)', transitive. A causative prefix is raz- as in place bebe-to
'cries baby-the, the baby cries' (intransitive) vs. raz-placa hebe-to 'I CAUS-cry babythe, I make the baby cry' (transitive). The cumulative and distributive suffixes induce
quantification over the verb's complement, as in the case of the cumulative na-: gotvja
'to cook' vs. na-gotvja supi ' to prepare a lot of soups'. Markova's outer prefixes have
no particular effect on verb arguments but instead introduce aspectual or phasal meanings: za-place 'to start crying (place)', pri-boli 'to suddently start hurting (boli)', poplace 'to cry for a little while', ot-boli 'to stop hurting', pre-proda 'to sell again' (from
pro-da 'to sell', with the lexical prefix pro-). A second set of outer prefixes modulate
the intensity of an action, giving intensive or attenuative readings: pre-jade 'to eat excessively (jade)', po-jade 'to eat a little'. Finally, there are prefixes which denote manners:
s-migne 'to blink rapidly (migne)', raz-V-Urie 'to untie (vurie)'.
Up to four prefixes may be found in a single verb form: iz-po-na-[raz-kaie} 'to
narrate ([raz-kaze]) completely (iz) many (na) stories one by one (po)'. Markova follows
Cinque (1999) in assuming that all the different meanings of the inner and outer affixes
are represented in a universally fixed hierarchy which gives rise to a strict linear ordering
(though some of the prefixes appear in more than one place on the hierarchy). This is
said to govern the combinability of the inner and outer prefixes. For example, we can
combine the outer prefix do- 'completion' and inner prefix na- 'many' in that order in
do-na-[pri-gotvja} 'finishing preparing many (materials)' but not in the reversed order,
*na-do-[pri-gotvja}. (See also article 160 on Ukrainian.)
3.2. Morphophonologica l rest rictions

Earlier generative approaches to affix ordering generally, and derivational affix ordering
in particular, proposed that affixes in English (and by extension in other languages) can
be split into two complementary groups, "Class I" and "Class II" (Siegel 1974; Kiparsky
1982; see the summary in Spencer 1991: 79-81 ). The Class I affixes, such as -ity, trigger
and undergo idiosyncratic allomorphy and affect stress while the Class II affixes, such
as -ness, are morphophonologically inert. According to this "strata!" or "cyclic" model
the C lass I affixes are always closer to the word root than Class II affixes. Many morphologists were unconvinced by these proposals even for English (Aronoff 1976) and
especially for other languages such as Polish (Booij and Rubach 1984, 198 7), not least
because of contradictions and blatant counterexamples. Fabb ( 1988) demonstrated that
an empirically better solution could be found relying on affix-specific selection restric-
309
PA310
310

tions. Hay (2002) argued for a psycholinguistically-based account in terms of "parsability": some affixes are easily integrated into the morphophonology of a word and are
therefore hard to identify as affixes. For instance, word forms ending in the past tense
suffix allomorph -t, such as sent, bent, kept and so on have the same shape as monomorphemic roots such as dent, apt, but the -ed suffix of jogged, rubbed gives rise to the
word final sequences -gd, -bd which are never found in monomorphemic forms. Thus,
the -d suffix is easily parsed out of these word forms. Hay argued that the -ment in
government is less parsable in these phonotactic terms than, say, the -ment of bajjlement.
For this reason it is easier to treat government as a possible base for -al suffixation,
witness governmental vs. *bajjlemental. Hay and Plag (2004) demonstrated that a very
good empirical fit for English could be obtained by a judicial combination of selection
restrictions jn the manner of Fabb with phonotactically determined parsability, and this
conclusion has been extended in psycholinguistic work by Plag and Baayen (2009).
3.3. Paradigmatic aspects of derivational processes

The most obvious way in which we can speak of paradigmatic relations in derivation is
seen with the phenomenon of competing affixation. It is typically the case that a given
derivational relationship, say, subject-nominal formation, can be expressed by more than
one affix depending on the nature of the base lexeme. Thus, in English, deverbal subject
nominals (or agent nominals, see article 74) are formed with the suffixes -ant/ent (claimant, applicant, student, resident, etc.), by conversion (cook), by total suppletion (pilot
from fly (an aircraft)), but mostly by -er suffixation (driver). We can think of these
various processes as being in paradigmatic opposition to each other, with -er suffixation
being the default. In this way, subject-nominal formation is morphologically similar to
plural formation.
However, there are more subtle types of paradigmatic relationship, which involve
whole words. For instance, Spencer (1988) discusses personal nouns of the type baroque
flautist, electrical engineer, East German . These are each motivated by two distinct
expressions: {baroque flute, flautist}, {electrical engineering, engineer}, {East Germany, German}. The phrasal personal nouns are derived from the nominal expressions
baroque flute, etc., so that a baroque flautist is 'one who plays the baroque flute (habitually, professionally)'. The morphology of the head noun.flautist, however, is determined
by the head noun. This can lead to a situation in which it appears that the meaning of
'person who' is expressed by a process of truncating the suffix of the phrasal noun's
head. However, not all expressions permit such personal nouns. Although most baroque
flutes are made out of wood we cannot say of a baroque flautist that she is a wooden
flautist. The conect way to think of such constructions is to say that personal noun
formation of this sort is completely productive provided that the base expression is a
lexicalized expression. Where the base expression is a normal syntactic phrase such as
wooden flute the paradigmatic lexical relationship no longer holds and the word-formation is not licensed. The linear model of derivation, under which a derivational process
creates a new lexeme, which can then further serve as the base of lexeme formation, has
difficulty in accounting for those kinds of derivational relationship which appear to be
organized paradigmatically (Booij 1997, 2008; Spencer 1988; artic le 46 on word-forma-
PA312
312

adjectives in Slavic languages generally (see section 4.2 below). Thus from bratr 'brother' we obtain the adjective bratrow. However, the original base noun bratr can itself be
modified by a possessive pronoun (Corbett 1987: 300):
(8)
a. mojeho
bratr-ow-e
dieCi
my.M.GEN.SG. brother(M)-POSS.A-NOM.PL child.NOM.PL
'my brother's children'
b. naseho
wucer1-ow-a
zahrodka
our.M.GEN.SG teacher(M)-POSS.A-F.NOM.SG garden(F).NOM.SG
'our teacher 's garden'
These cases are particularly odd because the possessive pronominal adjectives mojeho,
naseho not only modify the base nouns bratr 'brother' and wucer 'teacher' but agree
with them in number, gender, and in case, as though the noun were in the genitive rather
than in the possessive adjective form. Other Slavic languages demonstrate (or used to
demonstrate) this type of "category mixing" to a much lesser extent, as reported in detail
by Corbett (1987).
Further examples of denominal adjectives whose nominal base is itself modified are
found in a number of languages, and especially in the Tungusic group of languages
spoken mainly in Russia (Nikolaeva 2008). These languages have a productive category
of proprietive adjective, that is, a denominal adjective with the meaning 'having N' (such
languages often have privative or caritive adjectives w ith the meaning 'without, lacking
N'; cf. English headless 'without a head' ). In English proprietive meanings are expressed
by compounds such as short-sleeved ('having short sleeves'), blue-eyed, or by derivations such as hairy ' possessing hair ', colourful, headed (as in linguistics terminology: a
headed (i.e. endocentric) compound). Proprietive adjectives in the Northern Tungusic
language Evenki (spoken in Russia) are productively derived from nouns with the suffix
-Ci-. This adjective can combine with number-inflected nouns and it then agrees in case
with the noun it modifies:
(9)
aja-l
oro-l-Ci-du
asi:-du
good-PL reindeer-PL-PROPR-DAT woman-DAT

'to the woman with good reindeer '
Example (9) shows that the proprietive marker is a lexical suffix, not a clitic or phrasal
affix, because it can be shown independently that the case markers are suffixes and not
clitics. Moreover, the proprietive marker may be fo llowed by diminutive and augmentative affixes that semantically modify the base noun:
(10)
a. sa1;ari-Ci-ka:r
hole-PROP.A-DIM
'with little holes'
b. xegdi dere-Ci-ke:ku:n
big face-PROP.A-AUG
'with a big face'
PA313
16. Derivation
Similar constructions can be found in Samoyedic languages of Russia, such as Selkup
(Nikolaeva 2008; Spencer 2013) and Nenets (Nikolaeva 2003).
4. Semantico-syntactic aspects
4.1. Prima ry sema nt ic f unctions of derivat ion - exa mples fro m Chukchi
The prime role of derivation is to create new lexemes by adding a semantic predicate to
that of the base lexeme. Article 74 on agent and instrument nouns, article 76 on place
nouns and article 80 on spatial and temporal relations in German word-formation provide
details of some of the more common derivational semantic relations. A1ticle 183 on
Maltese provides information about the meanings expressed by Semitic "verbal measures". For comparison, I describe the derivational patterns found in Chukchi, spoken in
the northeastern part of the Russian Federation (Skorik 1961; Dunn 1999). In this language we see a wide variety of derivational meanings expressed mainly by suffixes.
In the Tables below I list the principal derivational affixes of Chukchi, inc luding
aktionsart affixes. One noticeable feature is that several of these affixes appear with
more than one meaning/function, an instance of "separationism" (Beard 1995). These
include: ciklcek, -l?et/l?at, tam1Je, e- ... -ke, mec/mac-, -tkultko, em-lam-, -r?ulr?o, -twa,
et/at, ew/aw. In addition some aktionsart suffixes are also used to form verbs from nouns:
!?et/!?at, -r?u, -tku.
Tab. 16.1: Chukchi denominal affixes
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
-et/at 'general verbalizer'

-ewl-aw 'general verbalizer'
-twa 'general verbal izer (for statives)'
-tku/tko 'activity with given object'
-l?etl-l?at 'to travel by N'
-r?ulr?o ' developing event (of meteorological phenomena)'
-jpl-ep 'donning clothing'
-tw 'removing clothing'
-ulo 'to hunt', 'to eat', 'to remove '
-l)ithyet 'to spend period of time'
te/ta-N-1) 'creation of N '
r- ... -ewlaw 'causative'
r- ... -et/at 'causative'
r- ... -twetltwat 'causative '
-twiltwe 'inchoative'
Tab. 16.2: Chukchi denumeral affixes

16.
17.
-qewlqaw (i.e. ordinal suffix) intransitive verbalizer 'act n times'

r- ... -qewlqaw transitive verbalizer 'act on s.o./sth. n times'
313
PA314
314

Tab. 16.3: Chukchi affixes for expression of aktionsarten (see article 85 on word-formation and
aspect in Samoyedic for similar morphology in an unrelated language group)
L.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
-l?etl-l?at ' pro longed continuous action'

-cir 'prolonged interrupted action'
-c?etl-c?at 'occasional action'
-cit 'a lternating action'
-r?ulr?o 'distributive'
-sqacetl-sqacat 'accelerated action'
tale-ltala- 'gradual action'
-jiwal-iwa ' intensity'
-tku 'weakened sporadic action'
-l)IJO ' inceptive'
-platku ' terminative, completive'
le1J-lla1J- 'extreme completion'
mec-lmac- ' incomplete action'
cik-lcek- 'partial action'
em-lam- 'delimitative'
Tab. 16.4: Related verb forms

1.
2.
3.
4.
5.
6.
lagi-llage- 'genuine action'

tam1Je- ' in vain, not seriously'
-ca 'scattered distributive '
-qeetl-qaat 'diminutive'
-cgat 'augmentative'
re-Ira- ... -I] 'desiderative'
The affixes ine-/ena- and -tku/tko are particularly polyfunctional. They are used as antipassive markers and also as parts of the transitive agreement paradigms, and in their use
as aktionsart markers and in denominal verb formation they are often found as more or
less meaningless derivational formatives. The suffix -tku/tko additionally has a derivational function meaning 'activity with given object'. In these functions it is distinct
morphophonologically from the homophonous iterative aktionsart marker (weakened
sporadic action), which triggers "palatalization" of Ill, unlike the other occurrences of
-tku/tko.
In (11) I cite representative examples from Moll and Inenliqej 's (1957) dictionary
(including some Russian glosses which are especially difficult to translate into English
out of context), with the derivational formatives in boldface.
(11) Derived lexemes in Chukchi
ra-talewak 'to move sth. from one place to another; continue'

[peredvigat'; prodolfat']
ine-n-talewak 'to lead, carry' [vodit', vozit']
(Note that the transitivizing prefix ra- has the allomorph n in non-initial position)
ine-n-talewa-tku-k 'to lead, guide; govern, manage'
[vesti, provodit'; pravit', upravljat']
PA315
16. Derivation
ra-citkuk 'to govern, manage' [pravit', upravljat']
ine-n-citkuk 'to govern, manage; rectify, touch up; touch'
[pravit', upravljat'; podpravit'; trogat']
ra-jegtelewak 'to save; leave alive' [spasat'; ostavljat' v zivyx]
ine-n-jegtelewak 'to save; preserve life' [spasti; soxranit' zizn']
4.2. Transpositions and va lency alternations

There are several types of lexical relatedness which border on standard derivation but
which some morphologists would regard as a distinct type of relation from derivation
proper. It is often difficult to distinguish these other types of relatedness from standard
derivation. The kinds of relations I am thinking of are (see also the articles 67- 70):
- Lexical relatedness which is mediated through addition of a semantic predicate which
arguably forms part of the inflectional paradigm ("inherent inflection"), see articles
84- 86.
- Lexical relatedness which is not mediated through addition of a semantic predicate
but which invo,lves a change in morpholexical or syntactic word class (transposition).
- Lexical relatedness which is mediated through an alteration in the argument structure
of a predicate (usually a verb), either with the addition of a semantic predicate (causative verbs, stative/middle verbs) or without the addition of a semantic predicate (passive and antipassive verb forms, applicative verbs, reflexive/reciprocal verbs). See
articles 60 on word-formation and argument structure and 83 on valency-changing
word-formation.
An example of borderline lexical relatedness is the possessive adjective construction
found in many languages, for instance, throughout Slavic (for Ukrainian see article 160;
for Polish see 156). These adjectives are similar to relational adjectives. Relational adjectives, however, simply create an adjectival form of a noun so that it can be used as an
attributive modifier, as prepositional (from preposition) in the phrase prepositional
phrase. Thus, in Russian the noun kniga 'book' gives the adjective kniinyj 'pertaining
to a book' as in kniznyj magazin 'book shop'. In Polish relational adjectives differ syntactically from qualitative adjectives because, provided there is only one of them, it will
appear in postnominal position (Szymanek 2010: 82, 85- 91). Possessive adjectives, on
the other hand, seem to add a meaning component of 'possession' (either alienable or
inalienable), though it is very hard to know how to characterize that meaning. Thus, in
Russian the nouns Mama ' Mummy' or the name Ivan form adjectives mam-in 'Mummy's', Ivan-av ' Ivan's', as in mamina rukalsestralsapkalmasina/ .. .lproblema 'Mummy's
hand/sister/hat/car/ .. ./problem'. The adjective can express more or less any pragmatically mandated relation between the "possessor" and the "possessed". In Russian, possessive adjectives inflect in a different way from ordinary qualitative adjectives (or relational adj ectives such as kniznyj), but they are clearly denominal adjectives. Russian, like a
number of Slavic languages, also has a class of adjectives which are formed from names
of animals: ryba ' fish' ~ ryb-ij 'pertaining to (a) fish ' which has yet another declension
pattern (see also Szymanek 20 10: 93 for corresponding fonns in Polish).
315
PA316
316

One of the unclear aspects of derivational morphology is the position of argument
structure alternations and derivation (see articles 49 on argument-strnctural restrictions
on word-formation patterns and 60 on word-formation and argument structure). There
are two ways in which argument structure is important in a classification of derivation.
First, it is very common for morphology to target a verbal argument to derive a subjector object-denoting noun such as English employer or employee respectively (see article
74 for examples of subject nominals and article 52 on English -ee derivatives).
A closely related type of derivation is found in languages which permit conversion
of adjectives to nouns. In many cases the language preferentially or obligatorily converts
an adjective into a noun denoting a person (see Spencer 2002 for discussion). Thus, in
Russian the adjective bol'noj 'sick, ill' can be used as a noun to mean 'patient' . When
such conversion applies to deverbal transpositions we obtain nouns with meanings of
the kind 'person that VERBs' or 'person that is VERB-ed ', and these have essentially
the same function as -erlee nominals in English. Thus, the normal word for 'school
pupil/student' is a (reflexive) present participle, inflected as an adjective but functioning
syntactically like a noun: ucascijsja (from the verb uCit'sja 'to learn'). Similarly, the
word obvinjaemyj means '(male) defendant (in a court trial)' and is the present passive
participle from the verb obvinjat' 'to accuse'. In some cases the present passive morphology is found in the absence of the base verb lexeme, for instance, the word podsudimyj
'defendant' (sud 'trial', pod 'under'). Although the verb sudit' 'to judge' exists as well
as various prefixed forms, there is no verb *podsudit'.
Such converted personal nouns are fairly widespread cross-linguistically. For instance, Skorik (1961: 345) describes a "noun-participle" form in Chukchi. This is formed
from nouns, verbs, and adjectives with the suffix -l?on. Skorik stresses that such derived
forms are indeterminate between attributive modifiers and personal nouns (where they
can denote humans), serving to respond to either 'one who is doing what?' or 'which
person?'. Skorik provides numerous examples, including ro)ul?on 'one who is herding
(reindeer); herder', from rojuk 'to herd (reindeer)', ?ott?ol?on 'one who has dogs; dogowner', from ?ott?on 'dog ' , gontil)ol?on 'one who is handsome; handsome person'.
4.3. A-sema nt ic lexical relatedness: meaningless derivatio n

Aronoff (1976) pointed out that English has a number of verbs derived from Latin or
French in the main which appear to consist of a meaningless prefix (such as con-, de- ,
e(x)-, per-, re-, trans- and others) and a meaningless root (such as -fer, -late, -mit, -port).
He argued that such words really are morphologically complex on the grounds that the
roots show consistent a llomorphy, for instance, in their nominalized forms: com-mit ~
com-mission, e-mit ~ e-mission, p er-mit ~ per-mission , trans-mit ~ trans-mission and so
on. Now, it's unclear whether such words really are best thought of as complex (see
Carstairs-McCarthy l 992b for discussion). However, there exists a small number of
verbs in English with a similar prefix-stem structure for which it's clear that the stem
represents a motivating verb which has no semantic relationship to the derived verb:
under-stand, with-stand, up-hold, with-hold and others. The reason we know such verbs
are motivated (formally) by STAND, HOLD is that they share the same irregularity in
inflection: withstood, withheld, etc. Thus, English has verbs which are clearly complex
morphologically but in which both the prefix and the base are meaningless ("cranberry")
morphs.
PA317
16. Derivation
This phenomenon is fairly marginal in English, but in other languages the existence
of prefixed verbs consisting entirely of cranberry forms is widespread. As a rough approximation I would estimate that at least one third of the verb vocabulary of Slavic
languages such as Russian is of this character (Spencer 2000). Gennan is another language whose verb stock is mainly populated with prefixed verbs. A glance at any German dictionary will demonstrate that the majority of German verbs are formed with one
of a small number of prefixes such as be-, ein-, ver- (see article 134 on German). Some
of the prefixes, the so-called "separable prefixes'', have special behaviour (see below
and also article 139 on Faroese). In many prefixed verbs it is impossible to assign a
meaning reliably either to the prefix or to the base or to both.
In many cases it is possible to discern some meaning relationship with the base
lexeme. Thus, (sich) ver-sprechen in its meaning of 'to make a speech error' is motivated
by sprechen 'to speak' and is likewise related to a small number of other verbs (also in
construction with the accusative reflexive pronoun) with a similar meaning: (sich) verlesen 'to misread', lesen 'to read'; (sich) ver-schreiben 'to make a slip of the pen',
schreiben 'to write', (sich) ver-sehen 'to overlook', sehen 'to see'. In other cases verseems to function as a deadjectival causative prefix: sicher 'sure, certain', ver-sichern
'to assure'. In yet other cases, the prefixed verb is essentially synonymous with the
unprefixed base lexeme which motivates it but it differs in somewhat subtle ways semantically. Thus, veriindern ' to change' consists of the morphs ver-ander-n (with umlaut of
the /a/ of the base form, that is vowel fronting apophony /a/ ---+ /el), clearly begins with
a prefix ver- and is clearly derived from the verb andern 'to change'.
In many other cases, however, it is impossible to motivate the ver- prefixed verb by
its morphological base. Thus, the 'mis-' class of verbs with ver- illustrated above all
have other (usually more frequent) meanings: ver-sprechen 'to promise' (and with a
dative, rather than accusative, reflexive pronoun sich etwas vonjemandem/etwas versprechen 'to have expectations of something'), verlesen 'to sort, clean (fruit, salad, etc.)',
verschreiben 'to prescribe', versehen 'to provide with' and so on. In other cases there is
no semantic relation between the ver- prefixed verb and its base: geben 'to give', vergeben 'to award; forgive', verfugen 'to have at one's disposal, possess',fugen 'to place'.
In some cases the verb merely has the form of a prefixed verb but there is no (longer
any) motivating lexeme: ver-gnugen 'to amuse' in the absence of *gnuglgnug (though
this is presumably related etymologically to the adjective genug 'enough ').
Languages such as German, Russian and a good many others demonstrate that we
need to adopt the distributional approach to word-formation as well as the semanticallybased approach if we are to understand how derivation works. Verbs in these languages
obey a constructional schema (see article 12 on word-formation in construction grammar) that licenses words of the form prefix-base without necessarily providing a strict
set of rules or meanings.
4.4. Eva luative morphology

Evaluative morphology is an expression of the speaker's emotional attitude towards the
referent of the word, either positive or negative. Evaluative morphology applied to nominals (nouns or adjectives) tends to derive from morphology realizing size: diminutives
317
PA319
16. Derivation
Ackerman, Farrell and Philip LeSourd
1997
Toward a lexical representation of phrasal predicates. In: Alex Als ina, Joa n Bresan and
Peter Sells (eds.), Complex Predicates, 67- 106. Stanford: CSLI.
Ackerman, Farrell, Gregory T. Stump and Gert Webelhuth
2011
Lexical ism, periphrasis, and implicative morphology. In: Robert D. Borsley and Kersti
Borja rs (eds.), Non-transformational Syntax. Formal and Explicit Models of Grammar,
325- 358. Oxford: Blackwell.
1992 A -morphous Morphology. Cambridge: Cambridge University Press.
Aronoff, Mark
1976
Aronoff, Mark
Morphology By i tself. Cambridge, MA: MIT Press.
1994
Bauer, Laurie
1988
A descriptive gap in morphology. In: Geert Booij and Jaap van Marie (ed s.) , Yearbook
of Morphology 1989, 17- 27. Dordrecht: Kluwer.
Beard, Robert
1995
Lexeme Morph eme Base Morphology. Stony Brook, NY: SUNY Press.
Bochner, Harry
Simplicity in Generative Morphology. Berlin: Mouton de Gruyter.
1993
Booij, Geert
1997
Autonomous morphology and paradigmatic relations . In: Geert Booij and Jaap van Marie (eds.), Yearbook of Morphology 1996, 35-54. Dordrecht: K luwer.
Booij, Geert
2008
Paradigmatic morphology. In: Bernard Fradin (ed.), La raison morphologique. Hom mage a la memoire de Danielle Corbin, 29-37. Amsterdam/Philadelphia: Benjamins.
Booij, Geert
2009
Lexical integrity as a forma l universal: A constructionist v iew. In: Serg io Scalise, Elisabetta Magni and Antonietta Bisetto (eds.), Univerals of Language Today, 83-100. Dordrecht: Springer.
Booij, Geert and Jerzy Rubach
1984
Morphological and prosodic domains in lexical phonology. Phonology Yearbook I: 1- 27.
Booij, Geert and Jerzy Rubach
1987
Postcyclic versus postlex ical rules in lexical phonology. Linguistic lnquity 18: 1- 44.
1992a Current Morphology. London: Routledge .
1992b Morphology without word-internal constituents: A review of Stephen R. Anderson's AMorphous Morphology. In: Geert Booij and Jaap van Marie (eds.), Yearbook of Morphology 1992, 209- 234. Dordrecht: Kluwer.
Cinque, Guglielmo
1999 Adverbs and Functional Heads. A Cross-Linguistic P erspective. Oxford: Oxford University Press.
Corbett, Greville G.
1987
The morphology-syntax interface. Language 63: 299- 345.
Dunn, Michael
1999
A gramm ar of Chukchi. Ph.D. dissertation, Australian National University, Canberra.
Fabb, Nigel
1988
English suffixation is constrained only by selectional restrictions. Natural Language and
Linguistic Theory 6: 527- 539.
Halle, Morris
1973
Prolegomena to a theory of word formation. Linguistic Inquiry 4: 3- 16.
319
PA320
320

Hay, Jennifer
2002
From speech perception to morphology: Affix ordering revis ited. Language 78: 527555.
Hay, Jennifer and Ingo Plag
2004
What constrains possible suffix combinations? On the interaction of grammatical and
processing restrictions in derivational morphology. Natural Language and Linguistic
The01y 22: 565- 596.
Holton, David, Peter Mackridge and Irene Philippaki-Warburton
1997
Greek. A Comprehensive Grammar of the Modern Language. London: Routledge.
Jackendoff, Ray S.
197 5
Morphological and semantic regularities in the lexicon. Language 5 l: 639- 671.
Kaye, A lan S.
2007
Arabic morphology. In: Alan S. Kaye (ed.), Morphologies of Asia and Africa. Vol. 1,
211-247. Winona Lake, IA: Eisenbrauns.
Kenesei, Istvan, Robert M. Yago and Anna Fenyvesi
1998 Hungarian. London: Routledge.
Kiefer, Ferenc and Andras Koml6sy
2011
On the order of word-class preserving derivational suffixes in the Hungarian verb. Word
Structure 4: 201-2 14.
Kiparsky, Paul
1982
From cyclic phonology to lexical phonology. In: Harry van der Hulst and Norval H.
S. Smith (eds.), The Structure of Phonological Representations. Part 1, 131-175. Dordrecht: Foris.
Koenig, Jean- Pierre
1999 Lexical Relations . Stanford, CA: CSLI.
L ieber, Rochelle
1992
Deconstructing Morphology. Chicago: University of Chicago Press.
L ieber, Rochelle and Sergio Scalise
2007
The lexical integrity hypothesis in a new theoretical universe. In: Geert Booij , Luca
Ducceschi, Bernard Fradin, Em iliano Guevara, Angeliki Ralli and Sergio Scalise (eds.),
On-line Proceedings of the Fifth Mediterranean Morphology meeting (MMM5), Frejus
15 - 18 September 2005. Bologna: University of Bologna. http://mmm.lingue.unibo.it/
Markova, Angelina
2011
On the nature of Bulgarian prefixes: Ordering and modification in multiple prefixation.
Word Structure 4: 244-271.
Moll, T. A. and P. I. Inenliqej
1957
Cukotsko-russkij slovar'. Leningrad: Nauka.
Niko laeva, Irina A.
2003
The structure of the Tundra Nenets noun phrase. In: Marianne Bakr6-Nagy and Karoly
Redei (eds.), Onnepi kotet Hon ti Laszlo tiszteletere, 3 15- 327. Buda pest: MTA.
Nikolaeva, Irina A.
2008
Between nouns and adjectives: A constructional view. Lingua 118: 969- 996.
Plag, Ingo and Harald Baayen
2009
Suffix ordering and morphological processing. Language 85: 109- 152.
Plank, Frans
2010
Structure 3: 82-97.
Rice, Keren
2011
Principles of affix ordering: An overview. Word Structure 4: 169- 200.
Riehemann, Susanne
1998
Type-based derivational morphology. Journal of Comparative Germanic Linguistics 2:
49- 77.
PA321
16. Derivation
321
Selkirk, Elizabeth 0.
1982
Th e Syntax of Words. Cambridge, MA: MIT Press.
Siegel, Dorothy
1974
Topics in English morphology. Ph.D. dissertation, Department of Linguistics and Philosophy, MIT, Cambridge, MA.
Skorik, Pjotr Ja.
Grammatika cukotskogo jazyka. Vol. I. Leningrad: Nauka.
1961
Smilauer, Vladimir
1972
Nauka o ceskem jazyku. Prague: Statni pedagogicke nakladatelstvi.
Spencer, Andrew
1988
Bracketing paradoxes and the English lexicon. Language 64: 663- 682.
Spencer, Andrew
1991
Morphological Theory. Oxford: Blackwell.
Spencer, Andrew
2000
Morphology. In: Mark Aronoff and Janie Rees-Miller (eds.), Handbook of Linguistics,
213-237. Oxford: Blackwell.
Spencer, Andrew
2002
Gender as an inflectional category. Journal of Linguistics 38: 279- 3 12.
Spencer, Andrew
2007
Extending deponency: Implications for morphological mismatches. In: Matthew Baerman, Greville C. Corbett, Dunstan Brown and Andrew Hippisley (eds.), Deponency and
Morphological Mismatches, 45- 70. Oxford: Oxford University Press.
Spencer, Andrew
2010
Lexical relatedness and the lexical entry - a formal unification. In: Stefan Muller (ed.),
Proceedings of the HPSGJO conference, 322-340. Stanford: CSLI.
Spencer, Andrew
2012
Identifying stems. Word Structure (Guest editors Olivier Bonami and Gilles Boye) 5:
88-108.
Spencer, Andrew
Lexical Relatedness. Oxford: Oxford University Press.
2013
Stump, Gregory T.
How peculiar is evaluative morphology? Journal of Lingustics 29: 1-36.
1993.
Stump, Gregory T.
Inflectional Morphology. A Theory of Paradigm Structure. Cambridge: Cambridge Uni2001
versity Press.
Szymanek, Bogdan
1989
Introduction to Morphological Analy sis. Warsaw: Panstwowe Wydawnictwo Naukowe.
Szymanek, Bogdan
A Panorama of Polish Word-Formation. Lublin: Wydawnictwo KUL.
2010
Toivonen, Ida
Non-Projecting Words. A Case Study of Swedish Particles. Dordrecht: Kluwer.
2003
Vajda, Edward J.
Ket. Miinchen: LJNCOM Europa.
2004
Werner, Heinrich
1997
Die ketische Sprache. Wiesbaden: Harrassowitz.
Andrew Spencer, Essex (UK)
PA322
322
17. Conversion
1.
2.
3.
4.
5.
6.
7.
8.
Introduction
Background
Conversion as word-class change plus formal identity
Conversion as word-class change without formal identity
Conversion without word-class change, but with formal identity
Conversion in European languages
Conclusion
References
Abstract
Conversion is recorded throughout the European languages. Its manifestations vary according to the theoretical standpoint on word-formation and of related issues, and also
according to the morphological structure of each language. This article reviews the
essentials of the concept (section 2), and the cases that are described as conversion in
several European languages according to the two conditions that are set in the concept
of conversion (sections 3 to 5). An overview of conversion in European languages shows
that the well-known difficulties that it poses in certain languages occur in other languages too and are likely to be inherent in the concept of conversion (section 6).
1. Int roduction
Derivation is described in article 16 based on the formal, syntactic and semantic features
that characterize the relation between base and derivative. According to that framework,
and unlike derivation, conversion appears as a word-fonnation process where the form of
the converted item does not change, while its inflectional potential, its syntactic function
and its meaning do, such that the item displays inflectional, syntactic and semantic properties of a new word class. The definition of conversion is therefore built upon the conditions
of formal identity and word -class change of the base form (Bauer 1983: 32, 2003: 38), as
in the follow ing examples borrowed from the British National Corpus (at http://corpus.
byu.edu/bnc/) and the Corpus do portugues (at http://www.corpusdoportugues.org/),
respectively:
(1)
He was put up for sale before the start of this season.

Each time it was a great way to start the season.
(2)
A planta
tern
folhas
longas
e
The p lant-FEM.SG have-3.sG.PR leaves.FEM.PL long-FEM.PL and
estreitas [ ... ]
narrow-FEM. PL
'The plant has long and narrow leaves [ ... ]'
English
Portuguese
PA324
324

For a review of the theories of conversion, a number of specific ana~yses are available
in the literature which have become classic references themselves for the study of conversion (e.g., Bergenholtz and Mugdan 1979, 2000; Lieber 1981; Pennanen 1984; Sanders 1988; Olsen 1990; Don 1993; Stekauer 1996; Plag 1999).
The concept of conversion is a function of the concept of word-class change, which
in turn relies on the concept of word class (Bauer 2005). In cross-linguistic research, the
categories inherited from the Western tradition are used as the framework and new concepts are added as necessary, e.g., converbs or masdars as intermediate categories for
specific languages. The notion of intermediate categories is closely related to the flexibility with which the categories should be understood. Most current approaches define the
separation between categories as a gradient, so the differences between at least some
categories, specifically nouns, verbs and adjectives, are of degree, rather than of kind,
with respect to the prototype according to which each category has been defined (cf.
Danes 1966.; Lipka 1971, 1986; Ross 1972; Croft 1984). In the prototype approach, each
word c lass is defined by canonical members that have the potential to display all the
morphological, syntactic and semantic features that are associated with the members of
the category (see article 16 on derivation and also article 112 on the evolution from
Latin to Romanian for deverbal nominal conversion in Romanian). Category members
may deviate from the prototype in various degrees by not displaying all the possible
features of the category and taking on instead features of other categories (for a review
of word classes as prototype-based categories, cf. van der Auwera and Gast 2011 ).
There is also the view that the above relativism is unnecessary as regards at least the
categories noun, verb and adjective (cf., among others, Colombat 1988; Baker 2003).
The combination of properties prototypical of each category is then replaced by a defining feature, e.g., syntactic function in Baker (2003: 301 - 302), even if it projects into the
morphology and the semantics of each category. For ease of reference, but also because
the categories in question coincide largely with the ones that have been described as
universal (noun, verb and adjective, plus the word class adverb, which has been considered systematically in this handbook), this article uses the conventional system of word
classes, even if its categories may not always reflect accurately what they are supposed
to represent in different languages (for a review of word classes in cross-linguistic research, cf. Bisang 2011 ).
Neither of these two views on word classes is without difficulty in the identification
of conversion, partly due to the framework, partly due to the operation of related processes. This article does not make the distinction between conversion and! processes whose
output is synchronically similar to the output of conversion but that may not qualify as
conversion diachronically. This is a relevant distinction according to the differences that
may be found between actual cases of conversion and processes that may give similar
results. The distinction is relevant as regards the process too. The word classes and the
type of diachronic process may differ in each language, e.g., nouns and! verbs in English,
nouns and participles in Hungarian (see article 70 on nominalization in Hungarian),
and nouns and verbs in Romanian (see article 112 on the development from Latin to
Romanian).
In general, some degree of overlap between conversion and historical processes lies
at the core of several Indo-European languages, and is present in languages of several
genera, e.g., Danish, French, Turkish and Awar (see articles 140, 147, 184 and 206,
respectively). This overlap is well-known in the literature on conversion in English (cf.
PA325
17. Conversion
325
Pennanen 1984; cf. also Pennanen 1975 on the connection between conversion and backformation). It is also present in the description of a number of issues in word-formation,
e .g., article 49 on argument-structural restrictions on word-formation patterns which considers apparent instances of prefixation in German described as conversion and also
article 8 1 on adverbial categories.
A more relevant separation is between syntactic processes that have as a result a
similar profi le to conversion and actual lexical conversion (cf. Marchand 1967, 1969;
Plag 2003: 113-116; M anova and Dressler 2005; cf. also Eschenlohr 1999 on German
and the identification of transposition vs. derivation based on the former 's higher productivity, semantic predictability, low lexicalization, resistance to change their inflectional
and governmental properties and low availability). These syntactic processes are not
usually considered to be lexical conversion, e.g., when nouns premodify other nouns as
heads of noun phrases, or when adjectives appear as heads of noun phrases. Of these,
the former case (nouns interpreted as adjectives to whatever extent as a result of their
prenominal position) is more questionably lexical conversion, even if the literature may
have presented it as a case of partial conversion (see also articles 159 on Slovak, 178
on Permic or 179 on Mari on similar cases as conversion). When this position entails
not only apparent syntactic change but also semantic change, as described in article 38
on noun-noun compounds in French for cases like marron (where the nomina l meaning
'chestnut' changes into a colour as an adjective), it can be argued that conversion has
occurred (see also a11icles 148 on Ladin, 150 on Italian, and 151 on Romanian).
As to adjectives that appear as heads in noun phrases w ithout nominal inflection, a
case for ellipsis is often cited. This structure is well-known in a number of languages of
a number of genera, from English or Ukrainian to Khwarshi (see aiticles 1.35, 160 and
207, respectively, to name only some). The elliptical noun may typically denote a person
or thing, and one of these, or a hyponym, is reported to be usually recoverable in the
structure:
(3)
Les bienheureux sont [ ... ]
French
' The blessed are [ ... ]'

(4)
Die Armen sind [ ... ]
German
'The poor are [ .. . ]'

(5)
L'impossibile
e [...]
Italian
'Whatever imposs ible is [ ... ]'

(6)
Lo extranjero es [ ... ]
Spanish
'Whatever foreign is [ ... ] '

The structure may result in varying degrees of lexicalization (see article 118 on the
development from Old Hungarian to Modern Hungarian for nouns and verbs or adjectives and nouns). The structure may also reach the point that the adjective acquires the
categorial meaning of the noun and the elliptical noun is no longer recoverable because
it causes ungrammatical redundancy (e.g., Spanish present participles in -ante or -ente,
as in amante or presidente; cf. also Ramat and Hopper 1998: 2). A s imilar situation is
reported to be widespread cross-linguistically in article 16 on derivation for Russian and,
similarly, there are cases where the present participle remains in this lexicalized form,
PA326
326

even if the original verb is no longer present in the language (e.g., Spanish agente from
agere). These cases have also been described for Latin (see article 91 on word-formation
in Neo-Latin). They appear in the literature as plain conversion (see articles 66 on collectives and 118 on the evolution from Old Hungarian to Modern Hungarian), and as partial
conversion. It is usually thought that they do not qualify as complete conversion unless
they take on nominal inflection and the categorial meaning of nouns (cf. Marchand 1969:
362; Colombat 1988: 62; Gutierrez Ord6fiez 1994 for the role of syntactic processes,
specifically of ellipsis; cf. also Dokulil l 968b: 58 on the descriptive problem that these
constructions pose in English and German).
These cases are allegedly less productive than conversion between noun and verb of
the type found in Romance languages, but they are at the heart of the interpretation that
nouns and at least a subcategory of adjectives (classifying adjectives) are essentially the
same word class. The productivity of each type depends on the view of conversion being
assumed. In languages like English or Spanish, the restrictions on the type of adjectives
that can gain access to the syntactic structure described above are not entirely clear, e.g.,
with asymmetries between positive and negative predicates, as in Spanish un tonto vs.
*un inteligente that do not always hold (cf. un tonto vs. un listo). This structure, where
nouns and adjectives converge, is recorded in a number of languages of a range of
genera, e.g., Gennanic, Romance, Slavic, lndo-Iranian, Uralic, Basque, Turkic and
Northeast Caucasian. Actually, the high frequency with which this structure is cited in
Indo-European endorses Kruisinga's (1927: 103) claim that conversion of adjectives to
nouns is related at least to primitive Indo-European grammar.
A number of other cases can be mentioned here which have not always attracted
much attention in relation to conversion. One is conversion from noun to adverb by
lexicalization of oblique forms of nouns. This can be found rarely in English, as in
forward_, and in German, as in abend_ ' in the evenings', and is reportedly present in a
number of other languages too, like Lithuanian, Latvian, Lak and Bezhta (see articles
169, 170, 201 and 203, respectively). Their interpretation depends largely on whether
the original inflectional paradigm of the noun is available and the nominal base can take
other inflections proper of the original word class. Another case concerns participles, for
their borderline status between the categories verb, noun and adjective (see article 156
on Polish; cf. also Haspelmath 1996). Although participles are reported as input for
conversion in a number of European languages, it should first be decided whether these
really entail word-class contrast and in which respects, i.e. to what extent this change
really means new inflectional, syntactic and/or semantic contrast, or whether participles
should be considered an intermediate, multifunctional category.
3. Conversion as word-c lass change plus formal identity

Canonical conversion requires word-class change and formal identity. This matches the
standard profile of conversion in English (see article 135 on English) and the one on
which the description of conversion in other languages has been modeled. A number of
other European languages can be added to this profile depending on the word classes
involved and the direction of conversion. Other aspects of conversion in addition to this
profile have achieved a lower degree of unanimity as regards the essential elements
involved in the concept, namely the process, the input, the output, the productivity and
the relationship.
PA327
17. Conversion
Whether viewed as a dynamic process with a base and a derivative or not, as is
possible in some of the theoretical approaches listed in the introduction, conversion
responds to a lexical need. As in the definition used as a starting point for this article,
the output retains the form of the input and assumes the inflectional potential, the syntactic function and the categorial meaning of the new word class. The input is usually the
citation form of the base, but it can also be an inflected form, as is the case with
participles when they are taken to be the result of conversion (see also section 3). The
input and especially the output are mainly members of open word classes, although
closed word classes and a range of multiword items have also been recorded as input
for conversion (see article 135 on English; see also article 129 on word-formation and
orthography for a number of examples in German).
Transfer between nouns and verbs, between adjectives and nouns, and between adjectives and verbs are the usual cases. The first type of transfer is the most common pattern,
in either direction, but it is not recorded in every language: this handbook shows that
Tat, Khinalug, Bezhta and Khwarshi (see articles 174, 200, 203 and 207), the latter three
within the genus North-Caucasian, are examples of word-class change and formal identity, but none of them involves the word classes noun and verb. Hungarian must also be
mentioned here, even if situations become particularly complex with the relationship of
one and the same form with nouns, adjectives and participles as in English -ing and its
interpretation as conversion or as a multifunctional suffix (also with intervening diachronic considerations which may, however, not be entirely a reflex of contemporary
word-formation; see article 70 on nominali zation in Hungarian).
The output of conversion tends to come in line with a set of semantic patterns that
are specific for the word class of the output (cf. Whorf 1945: 9; Nida 1949: 50; Marchand
1969: 3 75; Cetnarowska 1996: 17- 18; Plag 1999: 219- 221 ). Converted words are versatile and adapt themselves to the context so they may carry different meanings as required
by the context in which they are used (cf. Clark and Clark 1979; Cetnarowska 1996;
Buck 1997; Plag 1999: 225- 226). This is a property of conversion that helps understand
its productivity. Conversion may also be related to figures of speech, specifically with
metonymy and metaphor. The role of these figures in conversion has received less attention than it deserves, even if they have been cited in the literature occasionally (cf. Stein
1977: 228- 230; Kastovsky 1989: 182).
The output of conversion is usually considered a new lexeme (cf. Dokulil 1968b),
but not in the interpretation of conversion as word-class underspecification, where no
new item is formed and the lexeme limits itself to taking one or the other specification
as regards word class. The interpretation of the productivity of conversion varies according to the theoretical position, and specific constraints have been defined in several
languages (cf. Don 1993 for Dutch; Plag 1999: 22 1- 222 for English and Neef 1999 and
2005 for German).
The relationship between the base and the derivative is interpreted in terms of formal
identity, whether it is homonymy (Lipka 1986, 1990: 140; Saeed 1997: 63- 64), polysemy or heterosemy (Goddard 1998: 20), homomorphy (Quirk, Greenbaum, Leech and
Svartvik 1985: 70-71; Hockett 1994: 180-181 ), or zero-derivation (as a relation instead
of a process, Sanders 1988: 157), according to the significance granted to the contrast
implicit in the word class between the base and the derivative. The interpretation of the
relationship between the input and the output as polysemy or as homonymy is by necessity implicit in how conversion-related pairs are recorded in dictionaries: they can be listed
327
PA328
328

under the same entry or under separate entries (see article 131 on dictionaries). It has
been claimed that finer distinctions in the analysis of the lexical change involved in
conversion are necessary for the identification of conversion-related pairs as one or as
several lexemes, e .g., as in the comparison between the noun and the verb can and the
noun and the verb corner, or the noun and the verb book, and the semantic contrast that
the syntactic difference carries with it in the latter two cases and that may argue against
the classification of the latter pairs as one and the same lexeme (cf. Cruse 1986: 7980).
As regards the relationship between the pairs involved in conversion, the emphasis is
otherwise on directionality (for a review of the synchronic and di achronic criteria used
in directionality and of their degree of agreement in English, cf. Bram 20 11 ). It has been
suggested that the usual approach to directionality needs to identify the base and the
derivative by senses rather than by lexemes, because several bases and several derivatives may exist within one and the same lexeme according to the senses of the input and
the output (Plank 2010). This multiplies the difficulty inherent in the identification of
directionality in conversion, but is consistent with the relative independence of senses
within lexemes as regards their use in word-formation, or at least with the fact that
derivation can pick up on a specific sense for creation of a new word and not on others
within the same lexeme (see article 16 on derivation).
4.
Conversion as word-class change without formal identity
4.1. Formal contrast
by (supra-)segmental change
Two main types of formal variation are usually accepted in conversion: minor formal
changes involving segmental or suprasegmental material (e.g., vowel or consonant alternation, deletion, voicing, stress shift), and formal changes involving morphology that
may be considered inflectional or derivational (see section 4.2).
The former types of formal contrast have been regarded as of less importance than
other morphologically, syntactically and semantically more relevant variables. They have
been included as conversion often, with and without the qualification that they are of a
marginal or a special type. Well-known cases are stress shift and the less discussed issue
of voicing with and without orthographical reflex in English noun/verb conversion (cf.
Kastovsky 1996: 234; Payne 1997: 30; Beard 1998: 63; Carstairs-McCarthy 2006: 752
on d ifferent types of internal modification). Formal changes may be in other languages
too, e.g., associated stress shift in noun/verb conversion in Lithuanian (see article 169)
and, under different circumstances, in Slovene and Greek deadjectival adverbialization
(see article 81 on adverbial categories; cf. also adjective to noun conversion in Ukrainian,
see article 160):
(7)
gar-as
'steam, vapour'
vs. gar-uo-ti
'to evaporate'
(8)
Lepo
vs. lep6
'beautifully'
'beautiful'
Lithuanian
Slovene
PA329
17. Conversion
(9)
f3if3awr:;
veveos
'certain'
329
vs. f3sf3aiwr:;
veveos
'certainly'
Greek
Vowel contrasts have also been cited as conversion in German (see article 134, cf. also
Neef 1999: 199 on the interpretation of vowel mutation as conversion in German):
(10) Kopf
'head'
vs. kopfen
'to behead'
German
Ablaut in Frisian may be interpreted to belong synchronically in conversion (see article

137). The opposite view is possible too, and vowel contrast is understood as outside
conversion for Icelandic and presented instead as sound change (see article 143) and as
ablaut in Greek (see article 172).
Other formal changes can also be considered here, e.g., involving consonant and
vowel in noun to verb conversion in Danish (see article 140), the linking morpheme -k
or -tin noun to verb conversion in Frisian (see article 137), vowel variation or deletion
in adjective to adverb conversion in Albanian (see article 171 ), and fir alternation in
noun to verb conversion, and internal Sandhi adjective to verb conversion in Basque
(see article 182):
(11) gaffe !
'fork'
vs. gafle
'to take with a fork'
Danish
(12) tun
'garden'
vs. tunkje
'to garden'
Frisian
(13) sinne
' sun'
vs. sintsj e
' to sunbathe'
Frisian
(14) (i) shpejte

'fastADJ '
vs. se shpejti
'very soonAnv'
Albanian
(15) aferm
' closeAn/
vs. se afermi
'close UPAov '
Albanian
(16) aske
'free'
vs. askatu
'to free'
Basque
(17) ugari
'plenti(ful)'
vs. ugaldu
'to increase'
Basque
These are less well-known than the cases cited for English but fall within the same group
as regards formal variation of the citation form of the base and of the derivative. The
occurrence of each of these changes needs consideration of each specific case separately,
so the description reflects their status as specific processes, or as associated features of
conversion.
PA330
330
4.2. Stem-based conversion

The second issue in fonnal changes and conversion revolves around the question as to
whether systematic formal variation is allowed between the base and the derivative. This
is partly related to the often cited connection between conversion and English, or to the
alleged higher productivity of conversion in English than in other languages, which has
been attributed to the loss of inflections since Old English. Viewed from the opposite
end, a relation has sometimes been established between inflection and conversion, such
that inflectionally rich languages tend to constrain conversion because they mark word
classes explicitly (for a review, cf. Pennanen 1984; see also artic les 140 on Danish, 174
on Tat, 183 on Maltese and 202 on Dargwa).
This constraint has been rejected in the literature based on the argument that conversion in inflecting languages only manifests itself differently than in non-inflecting languages, i.e. conversion appears according to what the morphology of each language
requires and, within that, to what the morphology of each word class requires (see article
145 on Spanish). Inflecting languages may use stems as their input and then apply
deletion of the ending in the base form and/or addition of a new ending, e.g., in Bulgarian
and Russian (see articles 54 on closing suffixes and 55 on closing suffi x patterns in
Russian). This appears as a more complex process than word-class change without any
subtraction or addition (examples (18) and (20) are from Manova and Dressler 2005:
73- 74):
(18) vik
'a cry'
vs. vikam
' I cry'
(19) pa6oma
rabota
'work'
VS.
pa6omamb
rabotat'
' to work'
(20) hod
'walk, gate, pace'
VS.
h6dati
'to walk, pace'
Bulgarian
Russian
Serbo-Croatian
Alternatively, this has also been described as an inflectional method of word -formation
(see article 179 on Mari). Other terms are paradigmatic derivation, as in article 156 on
Polish, or transflexion, as in article 15 9 on Slovak, even if the latter may not always
entail word -class change (cf. also Pennanen 1984: 79- 80 for the description of morphologically related pairs where one of the pair members by default must take a word-class
mark, namely the infinitive mark in terms other than conversion, specifically as derivation immediate, transfert grammatical or unmittelbare Ableitung; cf. also Kastovsky
1994: 104 on the nature of derivation al and inflectional endings).
In the case of English, this means that conversion may not be any more frequent than
in highly inflectional languages like Spanish or Latin, and that the loss of endings has
not resulted in a higher influence on the productivity of conversion, as has been claimed
in the literature (Marchand 1969: 362- 364; cf. also Quirk and Wrenn 1957: 104- 106
for the link between these processes in stem-based morphological frameworks like Old
English and word-based morphological frameworks like Modern English, even if both
are not explicitly presented as the same process in nouns and adjectives with respect to
verbs). The availability of this process, at least for the categories noun and verb, has
PA331
17. Conversion
331
been reported for highly inflected Indo-European languages like Latin, Sanskrit or Old
English (cf. Biese 1941: 18-30; Marchand 1969: 363-364; Pennanen 1984: 88; Sanders
1988: 170; Kastovsky 1992: 392). It is also a profitable process in such highly inflected
languages as Italian and Romanian, for example as regards denominal verbs (see articles
84 on word-formation and lexical aspect: deverbal verbs in Italian and 112 from Latin
to Romanian):
(21) hoop
'hope'
vs.
hopen
'to hope'
(22) tanssi
'dance '
vs.
tanssia
'to dance '
(23) strega
'witch'
vs.
stregare
' to bewitch'
(24) vorba
'talk'
vs.
vorbi
'to talk'
Dutch
Finnish
Italian
Romanian
The view that this pattern of word-class change is conversion even if it may necessarily
mark the derivative as belonging to a specific word class is also consistent with the
claim that affixless denominal verbal derivation has been described as deeply engrained
in lndo-European languages as well as more common than suffixal derivation (Kastovsky
1994: 94). It should however be noted that what today is described as conversion may
historically have been derivation of different word classes from a common root in lndoEuropean, so there is no actual relationship between those word classes, but a relationship with a common source (Kastovsky 1994: 110). Although this does not necessarily
have an effect on synchronic description, it may help to explain the existence of similar
difficulties in the description of conversion in lndo-European languages.
The argumentation that word-based and stem-based conversion are not substantially
different types of processes except that they manifest themselves differently according
to the different morphology of each language finds support from a different perspective
in the study of language disorders. Omission of the verbal inflection appears as replacement of the correct inflection with the infinitive ending, i.e. the occurrence of the verbal
stern requires the infinitival ending for verbal denorninal conversion in Italian but not in
English (see article 122 on word-formation in aphasia). When morphologically related
noun/verb pairs express themselves with and without the same infinitival ending in Italian and English, they do, as in aphasia, as a result of the morphological pattern of each
language rather than as a result of different processes.
5. Conversion wit ho ut word-class change, but with form al identity

The difference between this type and the first case discussed lies in whether change of
secondary word class should qualify as conversion and, therefore, as formation of a new
word (insofar as conversion is a word-formation process). In this interpretation, the
standpoint is that conversion covers unmarked change of secondary word-class categories such as countability or gender in nouns, valency in verbs, or gradability in adjectives,
PA332
332
to name some of the most common. If this change involves fonnal variation too, this is
the least frequent of the cases considered in the outline of the introduction, i.e. conversion without word-class change and without formal identity (see article 148 on Ladin).
The above changes of subclass may entail a new inflectional potential, new structural
capacities and a new nuance of meaning (see articles 179 on Mari (in-)transitive and
reflexive verbs and 160 on Ukranian, which deals with relational to qualitative adjectives). To the best of our knowledge, the view that this is conversion, even if it is of a
secondary type, dates back to Leech (1974: 2 14- 216), but this is not a widespread
approach. This view of conversion has been extended to categories which have traditionally been considered to be grammatical categories, like gender in nouns (see articles 64
on gender marking, 91 on word-formation in Neo-Latin, 145 on Spanish, 150 on Italian,
156 on Polish or 160 on Ukrainian), or grammatical categories, as can be argued of
Polish aspect (see article 156 on Polish):
(25) lider
' leader.FEM'
vs.
lider
' leader.MASC'
(26) ragazza
'girl'
vs.
ragazzo
' boy'
Spanish
Italian
The semantiic contrasts sometimes go as far as involving such specific oppositions as the
one between tree vs. its fruit (see articles 145 and 150 on Spanish and Italian respectively), or proper to common noun (see article 14 7 on French; for a review of these cases,
see also article 156 on Polish):
(27) naranja
'orange'
vs.
naranjo
' orange tree'
(28) mela
'apple'
vs.
melo
'apple tree'
Spanish
Italian
The implication of this interpretation is that conversion then ceases to be word-class

changing derivation by definition (see article 14 on the delimitation of derivation and
inflection for several languages where this approach is used). Another implication is
that conversion does not necessarily belong to word-formation. Considering the lexical
proximity between the two terms of the opposition, secondary word-class conversion
may be different uses of one and the same word, i.e. a new meaning structure (see article
77 on intensification as regards the word class adjective). When these contrasts are
described in terms other than conversion, the literature uses the terms morphological
alteration (see article 6 on word-formation in onomasiology) or valency-changing wordformation (see article 83, see also, among others, Payne 1997: 25, 41-43 on some of
these changes and their value as derivation).
6. Conversion in European languages

Unlike derivation or compounding, conversion is non-iconic, i.e. the changes in the
semantic structure of the input are not represented formally in the output, except for the
PA333
17. Conversion
new inflectional characteristics that the word-c lass change entails. This property leads
to the expectation that conversion should be rarer than other, morphosemantically more
transparent word-formation processes (cf. Dressler 1987: 105, 116; Beard 1998: 63).
This property also establishes a contrast with derivation or compounding that has been
felt since the earliest approaches to word -formation outside grammar in the 19th century.
As a result, conversion is often listed along with minor word-formation processes, even
if its productivity is usually described as higher than that of those minor word -fonnation
processes, and close to that of derivation and compounding (see article 2 on wordformation research from its beginnings to the 19th century; cf. also Sanders 1988: 165).
To the best of our knowledge, no procedure for the measurement of the productivity of
conversion has been put forward as there is, for example, for suffixation. Quantitative
assessments of the productivity of conversion are therefore not available in the literature.
However, conversion is often reported to be productive in Germanic languages, like
English and to a lesser extent German, less productive in Romance and Slavic languages
(see articles 76 on place nouns, 110 on historical word-formation in English and 120 on
word-formation in first language acquisition; see Neef 1999, 2005) and marginal in
Uralic languages (see article 117 on the history of word-formation in Uralic).
It has been questioned whether conversion is justified outside lndo-European languages, or whether it is as justified as in the languages which have modeled linguistic
description. However, although conversion is recorded in every language genus contemplated in this handbook, it does not exist in every language, e.g., not in Welsh (see article
153 ), in Mordvinic (see article 180), in Gagauz (see article 188), in Karaim (see article
189) or in Akhwakh (see article 205), and perhaps neither in Kabardian, where the
productivity is low too and multifunctionality and blurred word-class limits obscure a
clear argument for conversion (see article 194 on Kabardian).
Although it does not entail formal contrast to signal the new meaning, conversion
stands out among word-formation processes for its formal simplicity (cf. Kerleroux 1999:
100). This formal simplicity, along with its semantic versatility (see section 2) explains
its use before other more iconic but also more complex processes (e.g., certain types of
affixation), and its wide occurrence over a number of languages (see article 120 on
word-formation in first language acquisition).
The range of structures that are presented here as conversion have the disadvantage
that some may not qualify as conversion for a number of readers and are thus a lax
interpretation. It has the advantage that it exhausts the possibilities of interpretation of
conversion in a range of languages. The resulting picture also offers the possibility of
viewing conversion as applying to varying degrees as regards the conditions set on the
notion. The view of degrees of conversion is frequent where a separation is made between true conversion and the rest (see article 117 on the history of word-formation in
Uralic; the notion of degrees of conversion is also implicit in article 55 on dosing suffix
patterns in Russian with its consideration of a "(next-to-)zero nominalization" ).
The manifestations of conversion may vary because word classes and the properties
on which they are defined may vary across languages, and also because the value of
formal identity and, more importantly, the degree of freedom in the formal differentiation
of words may also vary across languages. In this respect, it should be stressed not only
that conversion is difficult to identify cross-linguistically (Kerleroux 1999: 93), but also
that English, which has set the pattern for the description of conversion in the less wellknown languages, is not the best possible frame of reference for conversion because it
333
PA334
334

is not representative of the morphological framework in which conversion has to operate
in other languages.
The essential requirements for the ideal of this word -formation process apply differently not only across languages according to their own morphology, but different types
of conversion may also co-exist within one and the same language. This idea lies at the
core of specific studies on conversion like Don (2005), where directionality is taken to
the fore as a criterion for the identification of different types of conversion between
nouns and verbs in Dutch. It is also the conclusion of Bauer's (2005) review of a number
of cases of infinitival constructions in a set of languages and of his discussion of what
it means to change a lexical class: not all languages may exhibit the canonical type of
conversion that certain languages do for a number of cases within certain word classes.
As conversion does not use a different explicit form with which each of its meanings
can become associated, it is usually considered to be one and the same process regardless
of the base it operates on and regardless of the derivative it forms. However, taking to
the extreme the position that finer distinctions are necessary in the study of conversion,
it may be argued that these are different types of conversions when different inflectional
paradigms, different categorial meanings and/or different semantic patterns within the
same new categorial meaning are involved in conversion according to the word classes
involved. If different semantic patterns of conversion also display different properties
(such as different degrees of productivity, different constraints, etc.), then they must be
considered different types of conversion, even if the patterns do not ca1Ty different formal
markings (see article 145 on Spanish). The same applies as regards directionality as
proposed in Don (2005). Deverbal nominal conversion is reportedly less frequent than
denominal verbal conversion in several respects. Thus, e.g., in English child language
deverbal nominal formation is not mainly by conversion, but by compounding, whereas
the opposite, i.e. denominal verbs, are mainly by conversion; in French child language,
it is reportedly not productive (see article 120 on word-formation in first language acquisition). The constraints involved in the conversion patterns that exist between nouns and
verbs in English have been referred back to the origins of Indo-European (Buck 1997).
These patterns belong to different levels of lexis formation, so their differences are more
complex than just directionality and involve, for example, different potential for further
derivation (see article 54 on closing suffixes).
The hypothesis that there are as many different models of conversion as there are
word classes and directions of conversion from one word class into another is conversion's counterpart to the diversity of patterns of affixation, again according to the direction of application and the word class. The same goes for compounding. The difference
is that in conversion the output does not mark the different meanings (including the
different word classes that may be created by the same form). This a lso goes for the
different internal relations that are possible between the base lexemes of compounds but
are not marked formally. This does not presuppose formal, syntactic or semantic unity
within one process and within one word class, where considerable variation may exist
between the inflectional paradigms and the categorial meanings associated with the target
word class (e.g., regu lar vs. irregular patterns as regards inflection, and different semantic
patterns within, e.g., noun to verb conversion, even if they all form denominal verbs).
The specialized literature does not use this position (cf. Plag 1999: 223- 225). In an
output-oriented approach, there is no need to consider formal specifications because
there are not any, and fairly small sets of semantic patterns can account successfully for
PA335
17. Conversion
the semantic diversity of the output. In a process-oriented approach, the fact that all the
examples comply with the general conditions of conversion regardless of the word classes involved, of the formal specifications that are abandoned or taken on, or of the semantic patterns which are used does not mean that they are one and the same process in
each case.
7. Conclusion
Conversion has a range of interpretations in European languages. The difficulties of
identification reside in the widely varying morphological structures of European languages and in the concept of conversion itself.
Some conceptual agreement exists on the nature of the output of conversion, but
otherwise there is considerable conceptual diversity. This is an effect of the dependence
of conversion upon the theoretical framework used for the description of word classes
or the interpretation of word-formation in general and the concept of formal identity in
particular.
Acknowledgements
I would like to thank the editors of this volume for extremely helpful comments on
earlier versions of this paper. Special thanks are also due to Prof. Laurie Bauer, Dr. Ana
Diaz-Negrillo and Dr. Juan Santana Lario for insightful discussion. I would also like to
thank Prof. Nicofas Ballier and Dr. Jesus M. Garcia Gonzalez for their comments on
French and Modern Greek, respectively. Responsibility for all errors remains mine. Preparation of this paper was supported by the Spanish Ministry of Economy and Competitiveness' project with reference FFI2012-39688.
8. References
Auwera, Johan van der and Volker Gast
2011
Categories and protoypes. In: Jae Jung Song (ed.), The Oxford Handbook of Language
Typology, 167- 189. Oxford: Oxford University Press.
Baker, Mark C.
2003
Lexical Categories. Verbs, Nouns, and Adjectives. Cambridge: Cambridge University
Press.
Bauer, Laurie
1983
Bauer, Laurie
2003
Introducing Linguistic Morphology. 2nd ed. Edinburgh: Edinburgh University Press.
Bauer, Laurie
Conversion and the notion of lexical category. In: Laurie Bauer and Salvador Valera
2005
(eds.), Approaches to Conversion/ Zero-Derivation, 19-30. Munster: Waxmann.
Beard, Robert
1998
Derivation. In: Andrew Spencer and Arnold M. Zwicky (eds.), Handbook of Morphology, 44-65. Oxford/Malden, MA: Blackwell.
335
PA336
336

Be rgenholtz, Henning and Joachim Mugdan
1979
Ist L iebe primar? O ber Ableitung und Wortarten. In: Peter Braun (ed.), Deutsche Gegenwartssprache, 339-354. Miinchen: Fink.
Be rgenholtz, Henning and Joachim Mugdan
2000
Nullelemente in der Morphologie. In: Geert Booij, Christian Lehmann and Joachim
Mugdan (eds.), Mo1phology. An International Handbook of Inflection and Word-Formation. Vol. 1, 435- 450. Berlin/New York: de Gruyter.
Biese, Yrjo M.
1941
Origin and development of conversions in English. Anna/es Academiae Scientiarum
Fennicae B 45(2): 1-495.
B isang, Walter
2011
Wo rd classes. In: Jae Jung Song (ed.), Th e Oxford Handbook of Language Typology,
280-302. Oxford: Oxford University Press.
Bram, Barli
2011
Major total conversion in English: The question of directionality. Ph.D. dissertation,
School of L inguistics and Applied Language Studies, Victoria University of Wellington.
Buck, Rosem ary A .
1997
Words that are their opposites: N oun to verb conversion in English. Word 48( 1): 1-14.
2006
Internal modification. In: Keith Brown (ed.), Encyclopedia ofLanguage and Linguistics.
Vol. 5, 752- 755. Oxford: E lsevier.
Cetnarowska, Bozena
1996
Constraints on affix less derivation in Po lish and E nglish: The case of action nouns. In:
He nryk Kardela and Bogdan Szymanek (eds.), A Festschrift for Edmund Gussmann
from his Friends and Colleagues, 15-28. Lublin: T he University Press of the Catholic
University of Lublin.
Colombat, Be rnard
1988
Les "parties du d iscours" (partes orationis) et la reconstruction d ' une syntaxe latine au
XVIe siecle. Langages 92: 5 1- 64.
C la rk, Eve V. and Herbert H. Clark
When nouns surface as verbs. Language 55(4): 767- 8 11.
1979
Croft, William
1984
Semantic and pragmatic correlates to syntactic categories. In: David Testen, Veena M ishra a nd Joseph Drogo (eds.), Papers from the Parasession on Lexical Semantics, 53- 70.
Chicago, IL: Chicago Linguistic Society.
Cruse, Alan
1986
Lexical Semantics . Cambridge: Cambridge University Press.
Danes, Frantisek
1966
The relation of centre and periphery as a language universal. Travaux Linguistiques de
Prague 2 : 9- 2 1.
Dokulil, Milos
I 968a Zur Frage der Konversion und verwandter Wortb ildungsvorgange und -bezie hungen.
Travaux Linguistiques de Prague 3: 2 15- 239.
Dokulil, M ilos
1968b Zu r Frage der sog. Nu llableitung. In: Herbert E. Brekle and Leo nhard L ipka (eds.),
Wortbildung, Syntax und Morphologie, 55- 64. The Hague: Mouton.
Don, Jan
Mo rpholog ical convers ion. Ph.D. d issertation, Utrecht University.
1993
Don, Jan
A note on conversion in Dutch and German. In: Leonie Cornips and Paula Fikkert (eds.),
2003
Linguistics in the Netherlands 2003, 33-43. Amsterdam/Philadelp hia: Benja mins.
PA338
338

Kruisinga, Etsko
1927
On the history of conversion in English. English Studies 9: 103-108.
Leech, Geoffrey
1974
Semantics. Harmondsworth: Pengu in.
Lieber, Rochelle
1981
Morphological conversion within a restricted theory of the lexicon. In: Michael Moortgat, Harry van der Hulst and Teun Hoekstra (eds.), The Scope of l exical Rules, 161 200. Dordrecht: Faris.
L ipka, Leonhard
1971
Grammatical categories, lexical items and word-formation. Foundations of Language 7:
21 1- 238.
L ipka, Leonhard
1986
Homonymie, Polysemie und Ableitung im heutigen Englisch. Zeitschrift fiir Anglistik
und Amerikanistik 34(2): 128- 138.
L ipka, Leonhard
1990
An Outline of English Lexicology. Lexical Structure, Word Semantics, and Word-Formation. Tiibingen: Niemeyer.
Manova, Ste la
2011
Understanding Morphological Rules. With Special Emphasis on Conversion and Subtraction in Bulgarian, Russian and Serbo-Croatian. Dordrecht/New York: Springer.
Manova, Ste la and Wolfgang U. Dressler
2005
The morphological technique of conversion in the inflecting-fusional type. In: Laurie
Ba uer and Salvador Valera (eds .), Approaches to Conversion/Zero-Derivation, 67-102.
Munster: Waxmann.
Marchand, Hans
1967
Expansion, transposition and derivation. La Linguistique 1: 13-26.
Marchand, Hans
1969
The Categories and Types of Present-Day English Word-Formation .. A Synchronic-Diachronic Approach. 2nd ed. Miinchen: Beck.
Neef, Martin
1999
A declarative approach to conversion into verbs in German. In: Geert Booij and Jaap
van Marie (eds.), Yearbook of Morphology 1998, 199-224. Boston/Dordrecht: Kluwer.
Neef, Martin
2005
On some a lleged constraints on conversion. In: Laurie Bauer and Salvador Va lera (eds.),
Approaches to Conversion/Zero-Derivation, 103-130. Munster: Waxmann.
Nida, Eugene A.
1949
Morphology. The Descriptive Analysis of Words. 2nd ed. Ann Arbor, MI: T he Un iversity
of Michigan Press.
0 lsen, Susan
1990
Konversion als kombinatorischer Wortbi ldungsprozel3. Linguistische Berichte 127: 185216.
Payne, Thomas E .
1997
Describing Morphosyntax. A Guide for Field Linguists. Cambridge: Cambridge University Press.
Pennanen, Esko V.
1975
What happens in back-formation? In: Even Hovdhaugen (ed.), Papers from the Second
Scandinavian Conference of Linguistics, Lysebu, April 19- 20, 1975, 216- 229. Oslo:
University of Oslo.
Pennanen, Esko V.
1984
What happens in conversion? In: Haken R ingbom and Matti Rissanen (eds.), Proceed-
ings from the Second Nordic Conference for English Studies, Hanasaari!Hanaholmen,
19'"-21s1 May 1983, 79-93. Abo: Abo Akadem i.
PA339
17. Conversion
339
Plag, Tngo
1999
Morphological Productivity. Structural Constraints in English Derivation. Berl in/New
Plag, lngo
2003
Plank, Frans
2010
Structure 3(1): 82-97.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik
Comprehensive Grammar of the English Language. London: Longman.
1985
Quirk, Randolph and C. L. Wrenn
1957
An Old English Grammar. 2"d ed. London: Methuen.
Ramat, Anna Giacalone and Paul J. Hopper
1998
Introduction. In: Anna Giacalone Ramat and Paul J. Hopper (eds.), Th e Limits of Grammaticalization, 1-10. Amsterdam/Philadelphia: Benjamins.
Ross, John Robert
1972
The category squish: Endstation Hauptwort. In: Paul M. Peranteau, Judith N. Levi and
Gloria C. Phares (eds.), Papers from the Eighth Regional Meeting, 316-328. Chicago,
TL: Chicago L inguistic Society.
Saeed, John I.
1997
Semantics. Oxford: Blackwell.
Sanders, Gerald
1988
Zero derivation and the Overt Analogue Criterion. In: Michael Hammond! and Michael
Noonan (eds.), Theoretical Morphology. Approaches in Modern Linguistics, 155- 175.
San Diego, CA: Academic Press.
Siegel, Muffy E. A.
1980
Capturing the Adjective. New York/London: Garland.
Stein, Gabriele
1977
The place of word-formation in linguistic description. In: Herbert E. Brekle and Dieter
Kastovsky (eds.), Perspektiven der Wortbildungsforschung. Beitrii.ge zum Wuppertaler
Wortbildungskolloquium vom 9.- 10. Juli 1976, 2 19- 235. Bonn: Bouvier Verlag Herbert
Grundmann.
Stekauer, Pavol
1996
A Theory of Conversion in English. Frankfurt/M.: Lang.
Vogel, Petra Maria
Wortarten und Wortartenwechsel. Zu Konversion und verwandten Erscheinungen im
1996
Deutschen und in anderen Sprachen . Berlin/New York: de Gruyter.
Whorf, Benjamin Lee
1945
Grammat ical categories. Language 21: 1- 11.
Salvador Valera, Granada (Spain)
PA340
340
18. Backformation
I. Introductio n
2. Approaches to backformation
3. Conclusions
4. References
Abstract
This article discusses the nature of one of the most problematic processes of wordformation that has been traditionally labeled as backformation or back-derivation. It
provides an overview of a range of theoretical approaches which differ considerably in
their respective accounts of the nature of this process. Finally, a new approach to this
phenomenon is proposed.
1. Int roduction
According to the theory of natural morphology the most natural process in inflectional
and derivational morphology is captured by the principle of diagrammaticity, i.e. constructional iconicity (cf. Mayerthaler 1981 ; Dressler 2005; Dressler, Mayerthaler, Panagl
and Wurzel 1987). This principle requires that a new meaning be represented by a corresponding form. Therefore, suffixation and prefixation processes are most natural, while
conversion is not natural, and subtraction processes, including backformation, are counter-natural.
This counter-naturalness of backformation is reflected in the limited number of languages that make use of backformation as a (productive) word -formation process. For
example, Stekauer, Valera and Kortvelyessy (201 2) show that backformation only occurs
in 10 out of 55 analyzed languages. What is even more striking, none of these ten
languages make use of backformation productively. Examples of backformed words are
given in (1):
(1)
kvalitetssakra
' to assure quality'
kvalitetssakring
'quality assurance'
(Olofsson, pers. comm.)
stofzuig
' to vacuum-clean'
stofzuiger
'vacuum-cleaner'
(Booij 2010: 41)
Swedish
Dutch
Cross-linguistic research by Stekauer, Valera and Kortvelyessy (20 12) also indicates that
backformation is more or less a European phenomenon which occurs in Germanic, Romance and Slavic languages. Backformation in Finno-U gric languages is non-existing,
or very rare. According to Kilgi (pers. comm.), there are some examples of backformation in Estonian as in (2), but it is not a productive word-formation process nowadays.
In Finnish it depends on the account of relatively rare compound verbs traditionally
explained either as calques or backfonnations (3) (Laakso, pers. comm.):
PA341
18. Backformation
341
(2)
eelis
'advantage'
+-
eelistama
'to prefer'
(Kilgi, pers. comm.)
(3)
hiekkapuhaltaa
'to sandblast'
+-
hiekkapuhallus
'sand-blasting'
(Laakso, pers. comm.)
Estonian
Finnish
There are only three exceptions to the "Eurocentric nature" of backformation in the
sample collected for the above-mentioned research (Tamil, Hebrew, and Marathi), here
exemplified with a backformed word from Marathi:
(4)
kara
'hand'
+-
karaNe
'to do'
(Dixit, pers. comm.)
Marathi
Thus, the question may be raised as to whether the non-occurrence of backformation in

other parts of the world is a methodological issue, i.e. what counts as backformation in
European linguistic tradition may be possibly evaluated in a different way (see section
2 for various theoretical accounts), or whether it is not a purely diachronic process as
proposed by Marchand. The latter option might suggest that the absence of backformation in the word-formation descriptions of the languages of the world might be related
to the lack of historical data, or better, to the absence of this kind of investigation.
While there is a consensus among morphologists that backformation is a word-formation process in which a morpheme or a pseudomorpheme (or even - more generally "something" (Booij 2010: 40)) is deleted, they differ in accounting for the nature of
this process. Therefore, in the following sections various approaches to the concept of
backformation will be presented and various answers to the questions raised above will
be discussed. Moreover, it will be shown that there are basically two groups of views,
those which only admit the diachronic relevance of backformation, and those who stress
that, synchronically, it is a productive word-formation process (in English), no matter
what its nature stems from.
2. Approaches to backformation
2.1. Backformation as zero-derivation with subsequent clipping
In Marchand 's view ( 1960, 1969), paralleled by the views of, for example, Kastovsky
(1968, 1982), Quirk et al. (1972), Aronoff (1976), Kiparsky (1982), the process called
back-derivation (backformation) is of diachronic relevance only. That peddlev is derived
from peddlerN is of historical interest. For syn chronic analysis Marchand 's equation is
(5)
peddle : peddler
write : writer
which means that backfonnation is analyzed analogically with suffixation. In order to

adhere to his principle of the syntagmatic nature of word-formation, Marchand suggests
PA342
342

that all backformed verbs are zero-derived from their suffixal bases (1960: 310). In other
words, the "basic" verb is arrived at through zero-derivation and the subsequent dropping
of the (pseudo)morpheme.
Hansen et al. (1982: 13 5) d istinguish between the type peddle - pedlar/peddler that
can be analyzed as a 'person who peddles', which, synchroni cally, should be classified
as suffix derivation, and the type burglar - burgle with the paraphrase 'to act as a burglar',
which belongs in the field of backformation owing to the more complex semantic structure of the "backformed" coinage. This brings us to Marchand 's assumption that "content
must be the final criterion of derivational relationship for any pair of words" (1969: 392).
This position is reflected in a series of semantic criteria for determining the direction of
derivation (Marchand 1964). For illustration, the verb saw is backfonned from the noun
saw because the semantics of the verb saw depends on the semantics of the corresponding noun: ' to use a saw, cut with a saw' (ibid.: 12). This idea was later taken over by
Nagano (2007).
Marchand's approach suffers from serious problems. Zandvoort calls into question
the assumption that backformation is of diachronic relevance only, and, furthermore,
claims that "[d]erivation without a derivational morpheme somehow reminds me of Bottom's Dream" (1961 : 123). Pennanen (1966, 1975), like Adams (1973: 38- 42), shows
that, first, a set of semantic criter ia designed to determine the direction of zero-derivation
is not reliable, w ith there being many counter-examples (cf., for example, Stekauer
2000). Therefore, no semantic criterion can be the only and primary criterion for backformation. Certainly, semantics can not be ignored as illustrated w ith the verb unwish which
should not be treated as backformed from unwished because their meanings do not match
(' to revoke, retract, cancel a wish' vs. 'not desired, not wished').
Second, the zero-derivation approach is called into question by many cases in which
the backformed verb occurs much earlier than the converted verb from the same base,
which means that "it is not conceivable how conversion shou ld have a share in the
formation of the BF (backformation, P. S.]" (Pennanen 1975: 221 - 222). As an example,
to beg backformed from beggar dates back to 1225 whereas the converted verb beggar
is first instanced three centuries later in 1528. Moreover, these two verbs have different
semantics: beg 'to ask for alms', beggar 'to make a beggar of, to impoverish, to make
look like a beggar'. Consequently, Pennanen, like Blad in ( 1911: 51 ff.) and Biese ( 1941:
245) before him, suggests that the reduction of converted verbs be classified as clipping.
In addition, while the majority of conversion-backformation pairs differ semantically,
the few pairs that are not semantically distinct differ in frequency, register, or grammatical features. For illustration, while the verb j ell is intransitive, jelly is both transitive and
intransitive.
The view of the diachronic validity of backformation is also rejected by Szymanek
(1998) who draws attention to "one quite productive pattern where backderivation reveals itself as a synchronically powerful device" ( 1998: 94), in particular, to "compound
verbs" like caretake +--- caretaker; air-condition +--- air-conditioning; and henpeck +henpecked. In add ition, Szymanek points out that Polish has a productive way of deriving
augmentative/pejorative nouns on the basis of nouns ending in the suffix -ka, which is
associated with a diminutive meaning (ibid.: 95):
(6)
beka
bu/a
'big barrel'
'big roll'
beczka 'barrel'
bulka 'roll '
PA343
18. Backformation
343
Finally, the "diachronic-only" position is criticized by Bauer (1983) who maintains that
"this can be seen not to be true. While it is true that now the paradigms:
meddle
meddler
peddle
peddler
sound identical, and it is not possible to tell that the second was a case of backformation,
at the time the form peddle was first used as a lexeme, there must have been some
synchronic process which allowed the analogy [... ]. Backformation must be allowed for
in a synchronic grammar, it is still a current method of forming lexemes" (1983: 6465). All in all, Bauer emphasizes that backformation is a synchronically productive process different from suffixation.
2.2. Backformation as compounding

Kiparsky (1982) also assumes that the term backformation has historical relevance only.
This position is supported by the fact that the assumption that backformed verbs are
historically derived from nouns and adjectives comes as a surprise to linguistically unsophisticated speakers, it is not accessible to a learner, and it has no structural significance.
Diachronically, Kiparsky speaks of the process of reanalysis, where morphologically
simple words come to be perceived as derived from verbs:
(7)
[beggar]N reanalyzed as [[beg]v ar] N

[injury]N reanalyzed as [[injur]v Y]N
Synchronically, " backformations" of the air-conditionv type are, according to Kiparsky,

explained by scheme (8):
(8)
[Y Z]x
i.e. as compounds generated by insertion of stems Y and Z into a categorial frame X

where X is a verb.
Verbs such as beg, edit, peddle, eavesdrop, partake, sy llabify are treated by Kiparsky
as basic, and the fact that they are historically backformed from nouns and adjectives
has no structural significance.
2.3. Backformation as reinterpretation
In relation to word-pairs like peddler - peddle, Pennanen (1966, 1975) assumes that the
starting point for the process in question is the re-interpretation of a given form as a
complex word. Pennanen, like Koziol (1937: 194), maintains that reinterpretation is
based on the backward application of a particular derivative pattern. To use Nagano's
example (2007: 34), beggarN is originally a monomorphemic word, but is reanalyzed as
having the structure [[begg]-ar], based on which backformation takes place and brings
PA345
345
18. Backformation
Becker does not agree with merging these two rules into one bidirectional relation because "this would dissolve the difference between backformation and other formations"
(ibid.). This is an important difference from Plag's schema-based model discussed m
section 2.5.
2.5. Sche ma- based model of backf orm at ion

Plag calls into question an analysis of backformation based on suffix deletion by pointing
out that there is no productive process of suffix deletion in English. In his schema-based
model which systematically relates two sets of words, new words can be formed in both
directions (hence the bidirectional arrow in the schema below). Plag assumes that "the
existence of backformation is to be expected in a schema-based model, because there is
no inherent directionality in the relationship between the two sets of words that are
related by the schema" (2003: 187). Even though he does not provide any specific
schema for backformation one of them might look as follows:
(11)
[ ~> ]
'X'
<Xer>
[
~~gent performing X'
While this model copes with Nagano's class (a) it still cannot account for the other two
problematic classes.
2.6. Backformatio n as conversion w ith su bseq uent cli pping

Nagano (2007) accounts for backformation as a kind of conversion rather than zeroderivation. The synchronic relevance of backformation is, in his view, evidenced by the
productivity of backformation in English. Nagano resumes Marchand's idea of semantic
relationship and, in addition, introduces a quantitative-distributional criterion in order to
determine the derivational relationship among the three patterns illustrated in ( 12) and
(13 ):
(12) [xvolve]v,..., [xvolute]v- [xvolution JN
circumvolve
convolve
devolve
evolve
intervolve
involve
revolve
convolute
devolute
evolute
involute
revolute
circumvolution
convolution
devolution
evolution
intervolution
involution
revolution
PA346
346

(13) [xsolve ]v"' [xsolute]v"' [xsolution] N
absolve
dissolve
exsolve
resolve
solve
dissolute
resolute
solute
absolution
dissolution
exsolution
resolution
solution
Nagano assumes that the quantitative-distributional criterion makes it possible to identify

the derivational relations in the direction xvolvev - xvolutionN - xvolutev, and xsolvev
- xsolutionN - xsolutev, respectively, which gives support to the backformed nature
of the fonns [xvolute]v and [xsolute]v.
Nagano points out that the three problematic classes of backformation (a)- (c) (see
section 2.3) are coped with in Marchand 's zero-derivation approach because the attachment of zero is fo llowed by clipping a particular phonological sequence, a pseudomorpheme, which has no derivational significance. By implication, as Nagano says, this
approach makes it possible to delete a non-productive affix in (a), to ignore the categorial
selectional property of an affix in (b), and to delete a non-affixal element in (c).
However, given the extensive criticism of the theory of zero derivation, Nagano
comes up with a modified version of Marchand 's theory and claims that backfonnation
is a type of conversion with subsequent clipping: "BF results from these two processes.
Take the BF pair of the simple-word type televisionN, televisevi and the BF pair of the
compound type baby-sitterN, baby-sitvi, for instance. Just as conversion turns the nouns
catalog and chairman into verbs (catalogv, chairmanv), it turns the nouns television and
baby-sitter into verbs (televisionv, baby-sitterv), and just as clipping shortens the forms
of the nouns cocaine and doctor to the forms coke and doc, it shortens the forms of the
verbs televisionv and baby-sitterv to the forms televise and baby-sit. Tlhe only difference
between the converted verbs catalogv and chairmanv and the 'back-formed' verbs televise and baby-sit is that the latter have their fonns 'adjusted' by clipping" (2007: 56).
While the main reason for clipping can be seen in the universal tendency towards the
economy of expression, Nagano assumes that, w ith backformation, one more reason is
at play, in particular, the adjustment of a converted word 's form to its category. When,
for example, an input has a nominal or adjectival (pseudo) suffix (e.g., televisionN),
conversion to a verb yields a categorially verbal but formally nominal/adjectival output
(e.g., televisionv). In such cases, conversion uses clipping to remove the categorially
obstructive element, i.e. the (supposed) suffix, to adjust the output form to the output
category. Nagano (2007: 65) maintains that since clipping is not a systematic grammatical process a backformed verb may have a converted rival as in usherv and ushv for the
same reason that we have the doublet nouns doctor and doc.
Like all the other accounts this one also is not watertight. First, the lack of systematicity in the clipping of converted words can also be used as a counter-argument to the
theory of backformation as clipped conversion. The question may be raised as to why
the majority of converted words do not undergo the clipping process, and why a few of
them co-exist with the clipped counterparts. And why, in fac t, one and the same word
should exist both clipped and not clipped if they emerge due to the same underlying
process, especially in the cases when the backformed and the converted verbs have
identical meanings.
PA347
18. Backformation
Furthermore, the existence of converted verbs whose form has not been categorially
adjusted to the category of verbs calls into question the claim that the purpose of clipping
in backformation is to adjust the form of the converted word to that without an affix of
a different syntactic category.
2.7. Backformation as a synchronica lly non-rule-based process

Allen (1978) assumes that backformation is not a productive word-formation process
because "it is limited to applying only to lexicalized compounds with idiosyncratic meanings (i.e. compounds in the permanent lexicon), and does not apply to productive compounds with a variable range of possible meanings. In other words, there is no rule
of English morphology which directly derives non-occurring compound verbs such as
*[[furniture][move]]v, *[[watch][wind]]v, *[[bracelet]Liangle]]v or occurring compounds
such as baby-sit, type-write, tape-record" (1978: 214).
Similarly, Anderson ( 1992: 191) does not consider backformation of the type editor edit to be a rule-based process; rather, it is a sporadic process. For example, submersible
falls within the class of -able words. A speaker may then presume that inspite of there
not being a corresponding base verb, there ought to be one. This "enables" him to
produce submerse. The coming into existence of submerse is, in Anderson's view, a
matter of inference, which might be distinguished from the direct application of wordformation rules.
Finally, let us mention Beard (1995) who, in addition to automatic L(exical)-extension
rules, introduces another type of L(exical)-rules. They are intentional, conscious, extragrammatical rules. He demonstrates his ideas on the basis of backformations, acronyms,
and clippings. Thus, for example, the backderivation of lase from laser is not an automatic process, it is the product of an individual, non-ideal, speaker. It presupposes the
existence of a prior L-extension rule (in this case, the -er suffixation rule generating
instrumental nouns) to analyze and subsequently backform lase in analogy with such
words as shak-er, blow-er, etc. Consequently, unlike automatic L-rules, instances of
backformation are irregular, non-automatic, and unpredictable.
2.8. O no masiologica l approach

The so-called backformations are treated in accordance with the general onomasiological
model (Stekauer 1998, 2005) on the basis of the principle of morpheme-to-seme assignment that relates the onomasiological and the onomatological levels of the word-formation model. This can be exemplified by stage-manager and stage-manage in (14) and
(15). The examples show that these naming units are produced independently:
(14) Conceptual level: 'a person who manages a stage'
Onomasiological level:
Object
+--Action
- Agent
stage
manage
er
347
PA348
348

(15) Conceptual level: 'to manage a stage'
Onomasiological level:
Object
+----Action
stage
manage
In the case of naming units of the peddler-type only the "longer" word falls within the
scope of word-formation. To derive it, one postulates a potential verb peddle that, later,
became "actualized".
2.9. Noun inco rporation?

It may be assumed that what has been traditionally analyzed as backformation (mostly
based on diachronic data indicating that the "longer" form appeared historically before
the "shorter" one) represents, formally, two different processes. It may be proposed that
one of these processes that brings into existence naming units of the N+V structure
(baby-sit, proof-read, brainwash, sheepsteal, housebreak, caretake, fortune-hunt, bookkeep, spring-clean, sight-read, strap-hang, sleep-walk, chain-smoke, vacuum-clean,
wire-pull, window-shop, air-condition, brainstorm, tape-record, spotweld, factory-automate, etc.) has developed into a new productive word-formation process in English, i.e.
the process of noun incorporation. If noun incorporation is understood as a verb-fonning
process whereby a nominal stem is fused with a verbal stem to yield! a larger, derived
verbal stem, and the incorporated noun functions as an argument (usually object) of the
predicative verb, all new English "backformed" verbs comply with this definition and
most of them have the prototypical direct object-verb stmcture. Furthermore, if noun
incorporation is viewed as a special type of compounding as it combines two stems, this
approach is, in principle, in accordance with the above-mentioned conception of Kiparsky. There are at least five reasons for th is proposal: (i) compounding is globally, and
also in English, a highly productive word-formation process; (ii) it is a concatenative
process which better corresponds to the psychological expectations of language users;
furthermore, subtractive morphology appears to be an alien element in the overwhelmingly concatenative system of English word-formation; (iii) a high number of new coinages of this kind indicate a high productivity of this process. It may be assumed that
noun-incorporation emerged on the basis of analogy (as is usally the case with new wordformation patterns) and later on - with a growing number of instances - it developed
into a new word-formation pattern, a situation not unique cross-linguistically in the field
of word-formation; (iv) if it is the suffix which is deleted in backformation - as postulated in the "classical" account of backformation - the suffix must have first been - logically - attached to the stem (at least potentially). Thus, what is proposed by the classical
account is a cumbersome two stage model in which at first the affix is attached and
subsequently it is deleted. The noun-incorporation-based account of these formations is
much simpler and more straightforward, corresponding to the principle of constructional
iconicity as specified in natural morphology; (v) as demonstrated by Bromser (1985:
101) there are numerous recent N/A+V coinages without a motivating nominal or adjectival basis.
Admittedly, there are views rejecting this option, but they do not seem to be wellfounded. Thus, Booij opts for backformation in the case of words like babysit because
"there is no general process of N + V compounding in English" (2010: 41), a view
PA349
18. Backformation
which, in fact, rei terates Marchand 's rejection of this type of compounds (1969: 100).
Plag (2003: 154), too, claims that verbal compounds with nouns as non-heads are impossible in English, and that verbs cannot incorporate adj ectival/adverbial non-heads. First,
it is not quite clear why verbal compounds with nouns as non-heads should be impossible. Each language undergoes changes, and the development of English from Old English
to Modern English provides ample evidence that what once may have seemed to be
impossible has become reality. Moreover, Bromser (1985) - who also argues in favour
of an incorporation-based theory of "backformation" (within the framework of functional
grammar) - points out that verbal compounds existed as early as Old English. Second,
American Indian languages and some other languages of the world give evidence that it
is not only nouns that are incorporated (Hall 1956; Stekauer, Valera and Kortvelyessy
20 12).
Therefore, nothing precludes verbs like merry-make, rough-ride, white-wash from
being considered adjective incorporations. Semantically, various functions can be fulfi lled by the incorporated noun in languages of the world, including the category of
Time as in spring-clean, the category of manner as in strap-hang, etc. Thus, contrary to
what is claimed by Plag, nothing prevents the potential incorporations like book-read,
car-steal, fast-drive, slow-move from becoming actual verbs of English.
All in all, it may be argued that the original process of backformation in English gave
rise to numerous analogical formations no longer relying on "longer" counterparts. The
growing number of such formations may have triggered a synchronically productive
process of noun/adj ective incorporation. This view is not new. In 1956, Hall, in a paper
with a highly indicative title, How we noun-incorporate in English, conclusively argues
that noun incorporation is a productive method of English word-formation which has
been used for centuries (the fac t that contradicts the above-mentioned views of Booij
and Plag). A more careful position is taken by Kastovsky (1986: 419) who predicts
gradual development of a productive process of noun incorporation, a view recently
strengthened in Kastovsky (2009: 338) by saying that "this type has become rather
popular in Modern English as a result of the re-interpretation of denominal back-formations". A high productivity of this pattern is also confirmed by Szymanek (2005: 434)
who speaks of "considerable growth over recent years". Let us conclude this section by
Bromser's words who assumes that "there is no essential difference between English
verbal compounds and incorporation constructions in those languages which make use
of such a process for the construction of sentences" ( 1985 : 111 ).
3. Conclusions
This article has provided an overview of various approaches to the phenomenon known
in (predominantly) European languages as backformation or back-derivation . It has been
shown that there is no agreement in assessing the nature of this process. Differences
concern its diachronic vs. synchronic validity, different theoretical approaches to this
phenomenon as well as the systematicity vs. idiosyncrasy of the process. It may be
surmised that one can hardly describe various aspects of "backformation" by a unified
theory. It seems that rather than a single word-formation process we face processes of
different sort, including, on the one hand, incorporation, and on the other hand, a mixture
349
PA350
350

of reinterpretation, analogical formation, and clipping, all thi s clearly connected with the
general tendency (especially strong in recent decades) towards the economy of expression. Whether or not, in what way and to what degree this general tendency is implemented appear to depend, primarily, on productivity and the related morphosemantic
transparency of the corresponding word-formation rules. No wonder then that about 87 %
of English backformed words are verbs (Pennanen 1975). Namely, as Rainer (1993: 86)
correctly points out, deverbal derivatives such as action or agent nouns are produced by
highly productive rules and generally allow unequivocal reconstruction of the semantics
of the base word - the c ircumstances that establish a favourable framework for the
processes in question.
4. References
Adams, Valerie
1973
An Introduction to Modern English Word Formation. London: Long man.
Allen, Margaret R .
1978
Morphological investigations. Ph.D. dissertation, University of Connecticut, Storrs, CT.
A-Morphous Morphology. Cambridge: Cambridge University Press.
1992
Aronoff, Mark
1976
Bauer, Laurie
1983
Beard, Robert E.
Lexeme-Morpheme Base Morphology. A General Th eory of Inflection and Word Forma1995
tion. Albany : State University of New York Press.
Becker, Thomas
1993
Back-formation, cross-formation, and 'bracketing paradoxes' in paradigmatic morphology. In: Geert Booij and Jap van Marie (eds.), Yearbook of Morphology 1993, 1-25.
Dordrecht: Kluwer.
Biese, Yrjo M.
1941
Origin and Development of Conversions in English. Helsinki: Suomalainen T iedeakatemia.
Bladin, Wilhelm
Studies on Denominative Verbs in English. Uppsala: Alquist and Wiksell.
1911
Booij, Geert
The Grammar of Words. An i ntroduction to Morphology. Oxford: Oxford University
2010
Press.
Bromser, Bernd
1985
On the derivation of English verbal compounds. In: Wilfried Kurschner and Rudiger
Vogt (eds.), Grammatik, Semantik, Textlinguistik. Vol. 1, 99- 11 3. Tiibingen: N iemeyer.
Dressler, Wolfgang
2005
Word-formation in Natural Morphology. In: Pavol Stekauer and Rochelle Lieber (eds.),
Handbook of Word-Formation , 267- 244. Dordrecht: Springer.
Dressler, Woffgang U ., Wi lli Mayerthaler, Oswald Panagl and Wo lfgang U. Wurzel
1987
Leitmotifs in Natural Morphology. Amsterdam/Philadelphia: Benjamins.
Hall, Robert
How we noun-incorporate in English. American Speech 31: 83-88.
1956
Hansen, Barbara, Klaus Hansen, Albrecht Neubert and Manfred Schentke
1982
Englische Lexikologie. Leipzig: VEB Verlag Enzyklopadie.
PA351
18. Backformation
Kastovsky, Dieter
1968
Old English Deverbal Substantives Derived by Means of a Zero Morpheme. Esslingen:
Langer.
Kastovsky, Dieter
1982
Wortbildung und Semantik. Tiibingen/Diisseldorf: Francke.
Kastovsky, Dieter
1986
Diachronic word-formation in a functional perspective. In: Dieter Kastovsky and Aleksander Szwedek (eds.), Linguistics across Historical and Geographical Boundaries. In
Honour of Jacek Fisiak on the Occasion of his Fiftieth Birthday, 409- 421. Berlin/New
Kastovsky, Dieter
2009
Diachronic pe rspectives. In: Rochelle Lieber and Pavo! Stekauer, (eds.), The Oxford
Handbook of Compounding, 323- 342. Oxfo rd: Oxford University Press.
Kiparsky, Paul
1982
Lexical M orphology and Phonology. In: In-Seok Yang (ed.), Linguistics in the Morning
Calm. Selected Papers from SICOL-1981, 3- 9 1. Seoul: Hanshin Publishing Company.
Koziol, Herbert
1937
Handbuch der englischen Wortbildungslehre. Heidelberg: Winter.
Marchand, Hans
1960
The Categories and Types of Present-Day English Word-Formation. W iesbaden: Harrassowitz.
Marchand, Hans
1964
A set of crite ria for the establishing of derivational relationsh ip between words unmarked by derivational morphemes. l ndogermanische Forschungen 69: 10-19.
Marchand, Hans
1969
The Categories and Types of Present-Day English Word-Formation. A Synchronic-Diachronic Approach. 2nd ed. Miinchen: Beck.
Mayerthale r, Willi
198 1 Morphologische Naturlichkeit. Wiesbaden: Athenaion.
Nagano, Akiko
2007
Marchand's analys is of back-fo rmation revis ited: Back-formation as a type of conversion. Acta Linguistica Hungarica 54(1): 33-72.
Pennanen, Eska
1966
Contributions to the Study of Back-formation in English. Tampere: Ju lkaisija Yhteiskunnallinen Korkeakoulu.
Pennanen, Eska
1975
What happens in back-formation? In: Even Hovdhaugen (ed.), Papers from the Second
Scandinavian Conference of Linguistic, 216- 229. Oslo: Oslo University Department of
Linguistics.
Plag, Inga
2003
Word-formation in English. Cambridge: Cambridge University Press.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik
1972
A Grammar of Contemporary English. London: Longman.
Rainer, Franz
1993
Spanische Wortbildungslehre. Tiibingen: Niemeyer.
Stekauer, Pavo!
1998 An Onomasiological Theory of English Word-Formation. Amsterdam/Philadelph ia: Benjamms.
Stekauer, Pavo!
2000
English Word-Formation. A History of Research (J 960- 1995). Tiibingen: Narr.
351
PA352
352

Stekauer, Pavol
2005
Onomasiological approach to word-formation. In: Pavo! Stekauer and Rochelle Lieber
(eds.), Handbook of Word-Formation, 207- 232. Dordrecht: Springer.
Stekauer, Pavol, Salvador Valera and Livia Kortvelyessy
2012
Word-Formation in the World's Languages. A Typological Survey. Cambridge: Cambridge University Press.
Szymanek, Bogdan
1998 introduction to Morphological Analysis. Warszawa: Wydawnictwo Naukowe PWN.
Szymanek, Bogdan
2005
The latest trends in English word-formation. In: Pavol Stekauer a nd Rochelle Lieber
(eds.), Handbook of Word-Formation, 429- 448. Dordrecht: Springer.
Pavo! Stekauer, Kosice (Slovakia)
19. Clipping
I.
2.
3.
4.
5.
Introduction
Historical outline
General survey of the most common forms of clipping
Semantic and pragmatic aspects of clipping
References
Abstract
Clipping - including acronyms and hybrid forms of word shortening - has been accepted as a part of word-formation for many years now. This article presents an overview
of research history up to modern (prosodic, prototypical, semantic, interdisciplinary)
approaches. It also highlights the most common techniques of word shortening in different European languages and focuses on some pragmatic features of word shortening.
1. Int roduction
Clipping, in general, refers to a word-formation process in which the cutting of a lexeme
(one or more words) does not result in a change of meaning, i.e. the new, condensed
word carries the same meaning as the (longer) original. In this article, the term clipping
(also known as "truncation") is therefore used for a variety of word shortening processes
that result in new word-forms, i.e. a new word is produced by abbreviating a longer
form. This process is not limited to individual words; in fact, an existing lexeme can
also contain several words, and even phrases can be clipped (e.g., as soon as possible
is commonly cli pped to ASAP and also to saspo). Furthermore, a clippnng usually occurs
PA353
19. Clipping
simultaneously with its longer original, at least when it is first being formed. This article
describes different results of this clipping process.
There are various other types of word shortening that do not result in new words
and that therefore, strictly speaking, do not belong to the field of word-formation (like
abbreviations). This article provides an overview of the types of word shortening that
generate accepted new words. Unfortunately, the term clipping is not used consistently
in the description of word-formation processes. In fact, researchers tend to generally
bemoan the simultaneous existence of a variety of definitions. Nevertheless, this drawback has never been resolved.
In order to lay a convenient foundation for the discussion of different kinds of word
shortenings, clipping should be understood primarily as a general term for word shortening as a word-formation process, comprising lexemes that have retained - at least for a
while - (nearly) the same meaning as their longer originals. In the literature, the term
clipping is often used only for cases where one part of a word stands for the whole, e.g.,
condo(minium), (tele)phone or (in)flu(enza). Given that a strict classification of all versions of word shortening is hardly possible, it seems helpful to see clipping as the
umbrella term for word shortenings in the context of word-formation (for further details
of definition and classification see section 3 .1 ).
Section 2 provides a brief overview of the history of research on word shortening.
Since the techniques of word shortening differ significantly, the resulting words need to
be differentiated as well. In the last decades, numerous attempts have been made to
establish a stringent typology of clipped word forms. While no generally accepted typol ogy has been found to date, most scholars agree on the basic terminology. Following a
brief presentation of word shortenings that do not belong to clipping, such as blending
or backformation, this section also discusses divergent approaches to word shortening
and various typologies.
Section 3 presents the most common classes of clipping. Since clipping is to be found
in most modem languages today, this article offers examples of word shortenings from
a variety of languages, with a concentration on European languages. Kreidler (2000)
presents further information on clipping in other languages, including Japanese. A German anthology on linguistic shortness in general (Bar, Roelcke and Steinhauer 2007)
includes analyses of "brevity" in various European languages. Thornton ( 1996) discusses
shortenings extensively in Italian, and Lopez Rua (2004) concentrates on English and
Spanish "Acronyms & Co.". For detailed information on clipping in different languages
see also the articles 134- 207 in this handbook. Finally, in section 4 semantic and pragmatic aspects of clipping are discussed.
2. Historica l outline
2.1. Word shorten ing and its resea rch
Clipping as a word-formation process is seen as a phenomenon of the late 19th and the
20th and 21st century. While many shortened forms were found in ancient times (like in
Latin IMP for imperator, COS. for consul, INRI for Jesus Nazarenus Rex Iudaeorum,
SPQR for senatus populusque romanus) or in medieval manuscripts (Greule 2007:
353
PA355
19. Clipping
(2008). Since all languages differ in their approaches to word shortening andl prefer their
own ways of clipping, it is not feasible to discuss all of them here. (For shortening in
German, English, French, Italian, Swedish, Danish, Norwegian, Icelandic and Ukrainian
see also the adequate articles in Bar, Roelcke and Steinhauer 2007, namely Busse and
Schneider 2007, Mayer and Rovere 2007, Nilbling and Duke 2007, Sandhop 2007,
Schmitt 2007 .)
2.2. Typological and term inologica l variation

For a few decades, scholars mainly concentrated on the shortening process itself, describing what parts of the long form were taken to create the shortened form. This became the
basis for hand ling the phenomenon on a scientific fundament. Miscellaneous publications
appeared, each with its own terminology, and often only discemable by minute details.
Since there are numerous ways to shorten a word and consequently various different
types of clippings, it is immensely complicated to establish a typology that applies to all
cases. Accordingly, scholars are usually forced to refer to a great variety of clipping. As
Lopez Rua states, "the multifarious nature of the categories forces them to complete
those descriptions with long enumerations of particular cases which do not follow the
general rule" (Lopez Rua 2004: 112).
In the last few years, research has begun to examine the linguistic features that surpass
the pure process of shortening, following the trend to leave structuralism in favor of a
more pragma- and sociolinguistically oriented approach. Today, research discusses clipping as a phenomenon of "extragrammatical morphology" (cf. Doleschal and Thornton
2000) containing submorphemic elements. Especially the aspect of prosodic morphology
has surfaced as significant. As early as 1996, scholars began to use the theory of prosodic
morphology to explain the strategies of word shortening: Anna M. Thornton underlined
the existence of prosodically governed morphological phenomena in Italian (Thornton
1996) and Elke Ronneberger-Sibold examined the "phonotactic and prosodic properties
of shortenings in German and French" (Ronneberger-Sibold 1996); for the English language, Ingo Plag discussed clippings in the context of prosodic morphology (cf. Plag
2003 and article 135 on English). One example illustrates the underlying idea of examining clippings on the basis of prosodic morphology: As Thornton shows in her study, the
minimal prosodic word in Italian is disyllabic, has trochaic stress, and ends in a vowel.
Various examples in fact support her thesis that this typical minimal word serves as a
template for the process of clipping in Italian.
The last decades have elucidated the impracticality of classifying the great variety of
existing types of shortening into one typology. Still, the prototypical approach (e.g.,
Lopez Rua 2002 and 2004; Michel 2006) might offer a solution to this dilemma, since
it focuses on the main issue, namely that there are central, peripheral and borderline
cases of shortenings that stretch across a kind of continuum - some being more typical
than others, and again others being central but not typical. The idea of a continuum
matches the need for classification best since it allows for locating any clipping in the
appropriate position. Following this method enables observations such as, "the most
'perfect' examples of acronyms (such as laser) look like actual words, both in shape and
in meaning (that is, speakers use them as words and are no longer aware that they once
replaced a whole expression)" (Lopez Rua 2004: 125).
355
PA356
356

Another aspect of research on clipping is the semantic approach (cf. Michel 201 1):
Why are originals actually clipped? What happens to their meaning and the meaning of
the clipped form? When do the meanings differ from each other? Other linguists examine
clippin g from a lexicographical point of view (e.g., Schroder 2000), focus on grammatical aspects (cf. Harley 2004; Ronneberger-Sibold 2007) or employ an interdisciplinary
approach (Fandrych 2008a, 2008b).
3. General survey of the most common forms of clipping
3.1. Word shortenings not belonging to clipping

Clipping is different from most types of classic word-formation as it neither accumulates
morphemes, nor does it attach something to a base (like in composition and affixation).
Rather, it is a process of deleting material but it needs to be distinguished from other
cases of shortening a longer base.
First, there are examples of shortening that do not result in a new word, namely the
purely graphic abbreviations. These should not be treated in the context of word-formation since they are far from being "real" words. Purely graphic abbreviations that are
always dismantled in spoken language cannot count as new words of their own, they are
never pronounced in the shortened form, which means that no new word is being formed
(like Mr. for Mister in English, Hbf for Hauptbahnhof in German, qc. for quelque chose
in French, sing. for singular, Singular and singulier in all the three languages). As stated
earlier, abbreviations constitute a rather old phenomenon that can be traced back to
ancient times when writing texts was an intricate and arduous piece of work and authors
often felt pressed to shorten words for economical reasons. Also, the numerous abbreviations occurring in medieval handwritings did not find their way into spoken language,
and have never become lexemes.
Furthermore, the definition of clipping claims that the clipped form - at least when
being created - represents a kind of synonym with its "original'', the longer form. This
implies that the longer form is a lexeme by itself. This does not apply to blends (see
article 21 on blending) and their longer forms: the basis for a blending never existed as
only one lexeme itself. The words breakfast and lunch exist, but not a combination of
both, whereas the blend brunch, wh ich contains parts of both, is a lexicalized word of
its own. Also, backformation (see article 18 on backformation) is not an example of
clipping.
On the contrary, formations that result from clipping like the ones often called acronyms
(Laser from light amplification by stimulated emission ofradiation, NATO from North Atlantic Treaty Organization, with orthoepic pronunciation, i.e. pronounced as a word) and
the ones often called initialisms (CIA from Central Intelligence Agency, pronounced as a
series of letters) today are clearly seen as part of word-formation, even if they are not considered to be morphological. Therefore they are treated in this article as well as the "classic"
forms of clipping like the ones that are a single subpart of the original.
PA357
19. Clipping
3.2. Back clipping, fo re-clipping, and related forms

Some archetypes of clipping occur in many languages. While various current attempts
to explain the phenomenon of clipping concentrate on other issues than the purely formal
typology of short forms, an introduction to the great variety of clippings is helpful in
order to gain an overview.
In order to avoid further confusion resulting from inconsistent terminology, thi s article
centers on describing only the most common types of clipping, each as a kind of prototype. Terminologically, it follows the majority of research on word shortening by calling
the longer form that is reduced the original.
The first group of clippings comprises words that consist of one continuous part of
the original; they can go back to "back clipping'', "fore-clipping" or "middle clipping".
a) Back clipping: The original is clipped at its back and the beginning is retained. Examples in which the original is a simple word are English ad(vertisement), app(lication),
exam(ination), gas(oline), gym(nasium), lab(oratory), mag(azine), math(ematics), recap(itulation), vet(erinarian); German Abi(tur), Dino(saurier), Frust(ration), Perso(nalausweis), Prof(essor), Reha(bilitationsmajJnahme), Uni(versitat); French ampli(ficateur),
anar(chiste), bac(calaureat), croco(dile), impro(visation), manif(estation), proc(ureur),
promo(tion); Dutch bib(liotheek), bios(coop), aso(ciaal), ju/ (from juffrouw). The clipping process differs from one language to another: German Kina no longer possesses an
original, whereas English cinema does not have a short form. In comparison, most Romance languages employ both a longer and a shorter form: French cine and cinema,
Italian cinema and cinematografo, Portuguese cine and cinema, or Spanish cine and
cinemat6grafo. In Russian, too, one can find the clipped form and the original: kino and
kinoteatr.
In German, the second part of a complex noun is often deleted, whereas the first
morpheme constitutes the clipping. Especially foreign words utilize back clippings:
Auto(mobil), Chiro(praktik), Derma(tologie), Dia(positiv), Echo(lot), Endo(skopie),
Ergo(therapie), Foto(grafie), Kilo(gramm), Mammo(graphie) , Neuro(logie), Ortho(padie), Sono(graphie). Scandinavian languages similarly favor this kind of clipping, e.g.,
Swedish arr(angemang) and el(ektricitet), Danish kval(ifikasjonsrunde) and stud(erende), Norwegian krim(inalroman) und perm(isjon). Icelandic, however, rarely uses
back clippings, and instead clearly prefers fore-clippings (cf. Niibling and Duke 2007).
Sometimes the clippings lose their link to the original (see section 4) or the original
is not recalled at once; this is the reason why some examples of back clipping have a
gender different from their original form. Such a change of gender can be registered for
instance in German der Prust (masculine noun) but die Frustration (female noun), die
Mikro (female noun) but das Mikrowellengerat (neuter), das Foto (neuter) but die Fotografie (female noun).
As evidenced in the examples above, the clippings that result from shorteni ng can be
arbitrary parts or existing morphemes; prosodic morphology tries to answer the question
why words are clipped in a certain way. Additionally, examples exist in which the original is rather complex and consists of several words: English cap(ital letter), pub(lic
house), or German Zoo(logischer Garten). Sometimes the clipped form is slightly adjusted, e.g., in bike (bicycle) or coke (Coca Cola).
357
PA358
358

Following a rather ancient tradition, first names are also often found among clippings:
Al, Ben, Fred, Lu, Nick, Ray. Furthermore, back clippings of proper names with an
additional endearing suffix are popular: Andy, Benny, Charlie, also Lindy (Charles Lindberg), Oppy (Robert Oppenheimer). In Germany, famous soccer players are honored this
way, e.g., Klinsi (Jiirgen Klinsmann) and Schweini (Bastian Schweinsteiger). In both
English and German, a hypocoristic suffix cannot only be added to proper names but to
common nouns as well: English looney (lunatic), Aussie (Australian), bookie (bookmaker), cabby (cabman), telly (television), German Campi (Computer), Knobi (Knoblauch),
Pulli (Pullover), Schlaffi (Schlaffer), Studi (Student/-in).
b) Fore-clipping, in which the original is clipped at the beginning and the end of the
word is retained, is less common in English - (we)blog, (omni)bus, (violon)cello,
(ra)coon, (alli)gator, (tele)phone, (tum)pike, (aero)plane, (cara)van, (uni)versity; and
very rare in German (there are Bus and Cello, like in English, and the rather colloquial
Sch/and for Deutsch/and), French, and other European languages. Only Icelandic has a
greater number of fore-clippings, such as (automo)bill, (s6sialdemo)krati, (siga)retta or
flensa for influenza (cf. Niibling and Duke 2007). First names originating in fore-clipping
are, for example, Bella, Dora, Tilda, Tina, Trix, with an additional hypocoristic suffix in
Betty and Netty.
Fore-clippings must not be mistaken for similar-looking short forms such as German
Rad for Fahrrad or Kanzlerin for Bundeskanzlerin. In such cases (frequently found in
German) determinative compounds are shortened to their head; no new words are being
constructed.
c) Clippings with the middle of the word retained are rather infrequent. Flu for influenza,
fridge for refrigerator in English, or niver for aniversario in Portuguese may serve as
examples. In German, they are virtually unknown except for some proper names such
as Lisa (Elisabeth) or Basti (Sebastian); French is merely acquainted with strass for
l' administration. Typically, the term middle clipping is employed in English to describe
these formations. Not surprisingly, this expression often leads to confusion, since the
terms back clipping and fore-clipping emphasize the part of the original that is clipped,
whereas the term middle clipping stresses the part that is retained.
Generally speaking, back clipping surfaces as the rule in most languages when retaining
whole parts of the original, while the omission of the first part is much less common.
While most of the clipped words are nouns, some adjectives can be found: English
bi(sexual), or with hypocoristic suffix comfy (comfortable) and loony (lunatic), German
bio(logisch), French sympa(thique). In German, for example, the trend for clipping
seems to be more dominant in colloquial language or youth slang than in standard language. Often, back and front-clipping is employed to create novel, humorous, or witty
words such as the adjectives schizo and phren (both in use for schizophren), or depri
deriving from deprimierend.
3.3. Acronyms, initialisms, and related forms

At first glance, this category appears closed and easily identifiable; it includes shortenings that consist of different separate letters of the original. These are the most frequent
PA359
19. Clipping
word shortenings in all modern languages, with the most common type containing three
letters of the original (FBL JFK, UNO), but also with the possibility for two (UK), four
(YMCA) or more letters (UNESCO).
Such words are usually labelled acronyms or initialisms, generally discerned on the
basis of their pronunciation: There is first the alphabetic, letter-naming, letter-recitation
type, when every letter of the word is pronounced on its own (EFSF for English European Financial Stability Facility and Gennan Europaische Finanzstabilisierungsfazilitat,
OON for Organizacija Ob "edinennyx Nacij, the Russian version of UNO, or Polish PKO
for Polska Kasa Oszcz(jdnosci, Poland's largest bank), and there is second the orthoepic,
letter-sounding type, in which the shortened form is pronounced as a "normal" word
(NATO for North Atlantic Treaty Organization in English, OTAN for Organisation du
Traite de I' At/antique Nord in French, Havo for Hoger Algemeen Vormend Onderwijs in
Dutch, TAP for Transportes Aereos Portugueses in Portuguese). Very long acronyms
characteristically belong to the second category since it would be rather exhausting to
spell a word such as UNESCO. The orthoepic acronyms are also more likely to be
integrated into the lexicon of a language since, typically, the original eventually falls out
of the speakers' awareness: Who can remember spontaneously that UNESCO stands
for United Nations Educational, Scientific and Cultural Organization, or laser for light
amplification by stimulated emission of radiation? Even shorter ones like Aids and PISA
are not always known to stem from a longer original, and only few people remember
the original (acquired immune deficiency syndrome and Programme for International
Student Assessment).
The distinction between alphabetic and orthoepic acronyms is not always viable. For
example, there are cases in which both types of pronunciation can be observed: The
acronym VIP is pronounced as [vi: ai 'pi:] and as [vip], in German you can speak of the
FAZ (Frankfurter Allgemeine Zeitung) as [ef a: tset] or as [fats] - the latter pronunciations being much more informal. Similarly, the name of the lung disease SARS (severe
acute respiratory syndrome) was pronounced as [es a: er es] when it first emerged in
Germany in 2007. After a period of intense media coverage, both the pronunciation and
the notation were adjusted to that of other substantives: Sars. In fact, there are even
words that combine alphabetic and orthoepic pronunciation: LSAT (Law School Aptitude
Test) [El s.:et] in English, CPAM [ce pam] and FIDL [fidEl] in French, or EuGH (Europaischer Gerichtshoj) [oi ge: ha:] in German. And WLAN(Wireless Local Area Network)
in German is pronounced as a mixture as well [ve: la:n]; in English this short form is
not known, it is commonly called Wi-Fi, a term resembling the long-established term
Hi-Fi (for high fidelity), but originally not being a short form for wireless fidelity. As to
the pronunciation of French "sigles'', Plenat found out that apart from hi storical, sociological and lexical factors there are some phonological regularities like the distribution
of consonants and vowels ("syllababilite"), the length of the acronym ("taille") and the
hiatus that have an impact on the way an acronym is pronounced (Plenat 1998).
The basis for an acronym is most often a complex lexeme, but there are also phrases
and even sentences that are shortened this way. Especially electronic communication
makes use of acronyms: English bbs!BBS (be back soon) or lol!LOL (laughing out loud),
German HDL!hdl (hab dich lieb) or wkw (Wer kennt wen?) (for different kinds of word
shortening in electronic communication cf. Balnat 201 1).
359
PA360
360
3.4. Hybri d form s of clipping

The problematic lack of a cohesive terminology becomes most evident in the case of
word-formations that do not fit into any single category. There are words like, for example, radar (radio detection and ranging) that take one syllable and three further initials;
Benelux (Belgium, Netherlands, Luxembourg) that takes syllables from different parts of
the original;. Portuguese ABRALIN (Associac;ii.o Brasileira de Lingiiistica) that takes one
letter and two syllables; or German BaFin (Bundesanstalt fur Finanzdienstleistungen)
that takes two initials from parts of the first compound and one syllable of the last word.
Syllable clippings such as these occur rather frequently in German and Dutch. They
are often called acronymic clippings or clipping compounds: German Azubi (Auszubildender), Kita (Kindertagesstiitte), Schiri (Schiedsrichter); and Dutch doka (donkere
kamer), minco (minderwaardigheidscomplex), stufi (studiefinanciering). But syllable
clippings are found in Russian as well: politruk (politiceskij rukoviditel'), Komintern
(Kommunisticeskij internacional), Minzdrav (Ministerstvo zdravooxranenija), or zavkaf
(zavedujuscij kafedroj). Fascinatingly, German Kripo (Kriminalpolizei) must be considered such a hybrid clipped form, while the similar-looking Krimi for Kriminalroman or
Kriminalfilm as one continuous part of the original counts as a back clipping, a fact that
challenges the terminology. Most of these clippings originate in rather informal speech,
with many of them becoming increasingly popular and hence incorporated in quotidian
language.
Other examples of hybrid forms of clipping are combinations of one letter standing
for one part of the original compound with the second word unclipped, such as English
b-day (birthday), e-mail (electronic mail), e-zine (electronic magazine), or German
H-Milch (haltbare Milch), K-Frage (Kanzlerfrage), 0-Saft (Orangensaft), and U-Haft
(Untersuchungshaft).
4. Semantic and pragm atic aspects of clippin g

As evidenced, the use of clippings often expresses familiarity and hints at implied social
bonds amon.g the people employing them: one needs to belong to the same social group
to easily decode the meaning of a clipping. Such groups and their specific languages
can be as diverse as languages for special purposes (such as technonogy, the military,
police, medical sciences, linguistics, and countless other disciplines) or, for example,
youth slang. Especially slangy or colloquial contexts witness a change of stylistic level
between the short form and its original.
Often we find analogies among word shortenings: the term yuppie (young urban
professional), which emerged during the early eighties of the twentieth century, was
followed by dinkie (double income no kids, also Dink), Yettie (young, entrepreneurial,
tech-based twenty-something), and lately Fruppie (frustrated urban professional). Other
clippings employed to characterize people are LOHAS (lifestyle of health and sustainability), or Lobo (Lifestyle of bad organization).
Gradually, clippings can lose the link to their original, and their initial meaning might
be lost as a consequence. This can entail communicative problems, such as attempts of
concealment on the side of the user, or misunderstandings on the side of the recipient
PA361
19. Clipping
because of ambiguities, e.g., English sub: this clipping can refer to such diverse words
as submarine, substitute, subscription, subeditor or substitute teacher. At the same time,
this very phenomenon might help to ease communication when only a short form is
vocalized and the listener knows exactly what it signifies. This applies to examples such
as NATO, UNESCO, or WHO. Since communication mostly depends on context, all
partners know what they are talking about - they are acquainted with the subject, and
hence usually with the meaning of shortened forms. Especially in LSP (languages for
special purposes), clippings are indispensable since economic communication is essential.
A fonnidable communicative advantage of clippings is the possibility of further wordformations. Even German with its characteristic vast compounding possibilities could
never assemble words like OECD-Studie, or ARD-Korrespondent without using clipped
forms. Clipping also results in new verbs such as English to voip and German voipen
('to make a telephone call over the Internet using Voice over Internet Protocol technology'), English to blog and German bloggen ('to maintain or add content to a blog =
weblog'), or German simsen ('to send a message via Short Message Service on a cell
phone').
Often, letters are deliberately chosen in clippings to coin a word that speaks for itself
because it coincides with an already existing word; the homonym CARE (Cooperative
for American Remittances to Europe) might serve as an example. Another communicative aspect of clipping is language creativity necessarily employed for (re-)
interpreting known clippings. The case of the internationally operating German company
SAP might serve as an example for a situation in which the holder of a dipped name
requires (re-)interpretation: SAP originally stood for Systemanalyse und Programmentwicklung and can now equally denote the company after it changed its name to Systems, Applications, and Products in Data Processing. Similarly, the names of the German
companies AEG (initially Allgemeine Elektricitiits-Gesellschaft) and A&P (formerly The
Great Atlantic and Pacific Tea Company) grew to be "Aus Erfahrung gut" and "Attraktiv & Preiswert", respectively, in advertising and commercials. And the name FIAT
(Fabbrica Jtaliana Automobil Torino) came to stand colloquially in Germany for Fehler
in alien Tei/en, Fiir Jtalien ausreichende Technik, or Fahre immer am Tag - all of which
insinuate ironically a lack of quality and reliability in the cars.
Especially the last examples point to the huge variety of possibilities word shortening
procures. In the future, contrastive research should focus significantly on clipping in
different languages. Further investigation following the example of the recent publications concentrating on more than purely structural and formal aspects of clipping (cf.
section 2.2) indicates a very promising strategy that allows for analyzing the various
aspects - and specifically the semantic role - of word shortening.
5. References
Balnat, Vincent
2011
Kurzwortbildung im Gegenwartsdeutschen. Hildesheim: Olms.
Bar, Jochen A., Thorsten Roelcke and Anja Steinhauer (eds.)
2007
Sprachliche Kiirze. Konzeptuelle, strukturelle und pragmatische Aspekte. Berlin/New
York: de Gruyter.
Bellmann, Gunther
1980
Zur Variation im Lexikon: Kurzwort und Original. Wirkendes Wort 30: 369- 383.
361
PA362
362

Biv ille, Frederic
1989
Un processus derivationnel meconnu du latin: la derivation par troncation. L'lnformation
Grammaticale 42: 15- 22.
Busse, Ulrich and Dietmar Schneider
2007
Ktirze im englischen Wortschatz. In: Jochen A. Bar, Thorsten Roelcke and Anja Steinhauer (eds.) Sprachliche Kurze. Konzeptuelle, strukturelle und pragmatische Aspekte,
159- 180. Berlin/New York: de Gruyter.
Cannon, Garland
1989
Abbreviations and acronyms in English word-formation. American Speech 64(2): 99127.
Doleschal, Ursula and Anna Thornton (eds.)
2000
Extragrammatical and Marginal Morphology. Mtinchen: LINCOM Europa.
Fandrych, Ingrid
2008a Submorphemic elements in the formation of acronyms, blends and! clippings. Lexis 2:
105- 123.
Fandrych, Ingrid
2008b Pagad, Chillax and Jozi: A multi-level approach to acronyms, blends, and clippings.
Nawa Journal of Language and Communication 2: 71-88.
Greule, Albrecht
1996
Reduktion als Wortbildungsprozef3 der deutschen Sprache. Muttersprache 106: 193-203.
Greule, Albrecht
2007
Kurzworter in historischer Sicht. In: Jochen A. Bar, Thorsten Roelcke and Anja Steinhauer (eds.), Sprachliche Kurze. Konzeptuelle, strukturelle und pragmatische Aspekte,
118-130. Berlin/New York: de Gruyter.
Harley, Heidi
Why is it the CIA but not *the NASA? Acronyms, initialisms, and d!efinite descriptions.
2004
American Speech 79: 368-399.
Kobler-Tri11, Dorothea
1994
Das Kurzwort im Deutschen. Eine Untersuchung zu Definition, Typologie und Entwicklung. Ttibingen: Niemeyer.
Kreidler, Charles W.
1979
Creating new words by shortening. Journal of English Linguistics 13: 24- 3 6.
2000
Clipping and acronomy. In: Geert Booij, Christian Lehmann and Joachim Mugdan
1, 956- 963. Berlin/New York: de Gruyter.
Leuschner, Torsten
2008
Kurzwortbildung im Deutschen und Niederlandischen. Grundlagen und Ergebnisse eines
prototypischen Vergleichs. Germanistische Mitteilungen 67: 247- 261.
Lopez Rua, Paula
2002
On the structure of acronyms and neighbouring categories: A prototype-based account.
English Language and Linguistics 6: 31- 60.
Lopez Rua, Paula
2004
Acronyms & Co.: A typology of typologies. Estudios lngleses de la Universidad Complutense 12: 109- 129.
Mayer, Maurice and Giovanni Rovere
2007
Ktirze im italienischen Wortschatz. In: Jochen A. Bar, Thorsten Roelcke and Anja Steinhauer (eds.), Sprachliche Kurze. Konzeptuelle, strukturelle und pragmatische Aspekte,
211 - 226. Berlin/New York: de Gruyter.
Michel, Sascha
2006
Kurzwortgebrauch. Pladoyer fur eine pragmatische Definition und Prototypologie von
Kurzwortern. Germanistische Mitteilungen 64: 69- 83.
PA363
363
19. Clipping
Michel, Sascha
2011
Das Kurzwort zwischen ,Langue' und ,Parole' - Analysen zum Postulat der Synonymie
zwischen Kurzwort und Vollform. In: Hilke Elsen and Sascha Michel (eds.) , Wortbildung
im Deutschen zwischen Sprachsystem und Sprachgebrauch. Perspektiven - Analysen Anwendungen, 135- 163. Stuttgart: ibidem.
Niibling, Damaris and Janet Duke
2007
Kiirze im Wortschatz skandinavischer Sprachen. Kurzworter im Schwedischen, Danischen, Norwegischen und Is!andischen. In: Jochen A. Bar, Thorsten Roelcke and Anja
Steinhauer (eds .), Sprachliche Kiirze. Konzeptuel/e, strukturelle und pragmatische Aspekte, 227- 263. Berlin/New York: de Gruyter.
Plag, Inga
2003
Word-formation in English. Cambridge: Cambridge University Press.
Plenat, Marc
1998
De quelques parametres intervenant dans l'oralisation des sigles en franvais. Cahiers
d'Etudes Romanes (CERCLJD) 9: 27-52.
Ronneberger-Sibold, Elke
1996
Preferred sound shapes of new roots: On some phonotactic and prosodic properties of
shortenings in German and French. In: Bernhard Hurch and Richard A. Rhodes (eds.),
Natural Phonology. The State of the Art, 26 1- 292. Berlin/New York: Mouton de Gruyter.
2007
Zur Grammatik van Kurzwortem. In: Jochen A . Bar, Thorsten Roelcke and Anja Steinhauer (eds.), Sprachliche Kiirze. Konzeptuelle, strukturelle und pragmatische Aspekte,
Sandhop, Martin
2007
Kiirze im ukrainischen Wortschatz. In: Jochen A. Bar, Thorsten Roelcke and Anja Steinhauer (eds.), Sprachliche Kiirze. Konzeptuelle, strukturelle und pragmatische Aspekte,
Schmitt, Christian
2007
Kiirze im franzosischen Wortschatz. In: Jochen A. Bar, Thorsten Roelcke and Anja
Steinhauer (eds.), Sprachliche Kiirze. Konzeptuelle, strukturelle und pragmatische Aspekte, 181-210. Berlin/New York: de Gruyter.
Schroder, Marianne
2000
Kurzworter im Worterbuch: Lexikographische Aspekte der Kurzwortbildung. In: Irmhild
Barz, Marianne Schroder and Ulla Fix (eds.), Praxis- und lntegrationsfelder der Wortbildungsforschung, 91- 105. Heidelberg: Winter.
Steinhauer, Anja
2007
Kiirze im deutschen Wortschatz. In: Jochen A. Bar, Thorsten Roelcke and Anja Steinhauer (eds.), Sprachliche Kiirze. Konzeptuelle, strukturelle und pragmatische Aspekte,
Steinhauer, Anja
2000
SprachOkonomie durch Kurzworter. Bi/dung und Verwendung in der Fachkommunikation. Tiibingen: Narr.
Thornton, Anna M.
1996
On some phenomena of prosodic morphology in Italian: Accorciamenti, hypocoristics
and prosodic delimitation. Probus 8: 81 - 112.
Anja Steinhauer, Wiesbaden (Germany)
PA364
364
20. Composition
I.
2.
3.
4.
5.
6.
7.
Definition of composition
Categories of compounds
Exceptions to the head generalization
Bound compound constituents
Compound meaning
Conclusion
References
Abstract
This article first defines the term composition and explains the major structural properties of compounds. It then sets out to examine the different categories of compounds
proposed in traditional works on compounding with the intention of contrasting them
with current terminology and checking their coherence. Exceptions to the regular properties of compounds, in particular the head generalization and the criterion offree lexeme
constituents, are discussed. A basic reasoning strategy is proposed to explain the broad
range ofmeaning types possible for novel compounds. Finally, general pragmatic limitations on the compounding process deriving from the fanction and use of compounds are
discussed that restrict the nature of compound meaning.
1.
Definition of composit ion
1.1. Hist orica l develop ment of t he category

Composition is a major process of vocabulary extension in natural language. It refers to
the combination of two or more lexemes (roots, sterns or freely occurring words) in the
formation of a new, complex word, termed a compound. Over the course of time, the
scope of the term composition has shifted somewhat in tandem with a change in content
of its major counterpart, derivation. In his diachronically oriented Deutsche Grammatik
[German Grammar], Jacob Grimm (1826) defined derivation narrowly as the addition of
a formative to the end of a root for the purpose of deriving a new word, thus limiting the
class of explicit derivational processes to suffixation. Prefixes, lacking the transpositional
function of suffixes, were considered equivalent in function to the initial constituents of
compounds, hence, Grimm called combinations of prefixes with lexemes Priifixkomposita [prefix compounds] and treated them together with compounds. Wilrnanns and Paul
in their diachronic works of 1896 and 1920 respectively, both also entitled Deutsche
Grammatik, followed Grimm's example in treating prefixations together with compounds
and contrasting the class of compounds with explicit derivation, i.e. suffixation. This
organization of the major categories of word-formation was adopted by Henzen (194 7)
in his historical Deutsche Wortbildung [German Word-formation] and was also represented in Romance linguistics where the reasoning resembled that found in Grimm, namely
PA365
20. Composition
the obvious similarity between the set of prefixes and the freely occurring prepositions
and adverbs that formed first constituents of compounds, thereby functioning as modifiers of the lexeme they combine with rather than as formatives of new words, cf. Bauer
(2005) and article 2 on word-formation research from its beginnings to the 19th century.
This view is present in certain relatively contemporary studies of English as well. For
example, Marchand (1967) claims that all syntagmas are made up of a determinant and
determinatum and represent the two basic word-formation categories of expansions and
derivations, depending on whether the determinatum is an independent or a dependent
morpheme. The class of compounds is defined in the following manner: "All combinations whose determinata are independent morphemes [... ] are expansions. Expansions of
which both the determinatum and the determinant are words [ ... ] are called compounds."
Thus, compounds form a subclass of the category of expansions to which prefixations
also belong: in rewrite the independent determinatum write is modified by the determinant re- (Marchand 1967: 323).
The second edition of Henzen's Deutsche Wortbildung, which appeared in 1957, documents a shift in perspective. The class of prefixations was removed from the category
composition and assigned an intermediate status of its own between composition and
derivation. This reduction in the original scope of composition and new three-way division of concatenative word-formation with prefixation constituting its own category
served as the basis of Fleischer's 1969 Wortbildung der deutschen Gegenwartssprache
[Word-formation of the contemporary German language]. Fleischer 's revised work in
collaboration w ith Irmhild Barz in (1992) altered this categorization further to reach the
view, already represented in Erben ( 1983) and generally accepted in theoretical works
today, in which suffixation and prefixation together constitute the category of explicit
derivation which is set apart from composition.
1.2. Properties of compounds

Composition is a productive process of word-formation in which the component lexemes
are chosen from the lexical categories of a language forming a combination in which
one constituent functions as the head that is modified by the other constituent. In cherry
tree, for example, tree forms the head determining the nominal category and the related
morphosyntactic features of the complex word as well as constituting its general concept,
while the first constituent cherry modifies the head, creating a complex or subordinate
concept whose meaning is a hyponym of the meaning of tree. A certain amount of
recursion is permitted resulting in hierarchically arranged, binary groupings of constituents as in rug repairman [[rug] [repair + man]], brass knuckle method [[brass+ knuckle]
[method]] or fruit fly attention span [[fruit + fly] [attention + span]]. But frequently
compounds appear as simple combinations made up exclusively of lexical categories
that combine freely with one another. For instance, in the Germanic languages that display regular and highly productive compounding patterns, the compounding process is
not restricted by the categorial or selectional properties of the head constituent as is
generally the case in their derivational and syntactic configurations (see sections 2.4 and
3 for the more restrictive nature of Slavic and Romance compounds). Because of their
lack of functiona l categories, compounds do not overtly express the grammatical or
365
PA366
366

conceptual relations that exist between their constituents and are, consequently, inherently ambiguous. For instance, a lawyer joke can be a 'joke about lawyers', 'a joke thought
up by a lawyer' or 'a joke typically told by lawyers', etc.
It is generally the final constituent that functions as the head of a compound transferring its categorial and morphosyntactic features to the whole word; this generalization
is captured in Williams' (1981) righthand head rule. The most common heads crosslinguistically are found among the nominal and adjectival categories. Perhaps not quite
as common, but certainly possible, are verbal compounds. All the lexical categories
occur as the modifier constituent (nouns, adjectives, verbs, adverbs and prepositions)
and, occasionally, certain minor categories such as pronouns and focus particles may
appear as the nonhead constituent as well, although these are exceptional (cf. she-wolf,
self-indulgence, G. Noch-Kanzler 'still-chancellor '). Compounds with plural and phrasal
first constituents are not uncommon, cf. suggestions box, sky's-the-limit promise,
G. Mitgliederka.rte 'members card', Versuch'-Dein-GlUck-Spiel 'try-your-luck game', cf.
Johansson ( 1980) and Meibauer (2007). But, again, these non lexical categories are only
found as nonheads. Since words often carry inflectional markings for case, number,
person, etc. and since inflection is expressed on the head constituent (cf. Hoeksma's
1985 head operations), a nonfinal head would be a nonoptimal morphological structure
due to the fact that the head inflection (for plural, case, etc.) would break up the structural
integrity of the word. But exceptions to the head generalization can be found: there are
both compounds without a head and compounds that are left-headed. It is also necessary
to distinguish between the formal and semantic aspects of head hood, i.e. there are cases
where a constituent serves as the formal but not the semantic head of the compound.
These different deviations from the default pattern will be discussed in section 3. In
general, composition creates potentially recursive, binary branching, headed word structures or morphological objects (cf. DiSciullo and Williams 1987) that carry a unified
accent, have a unique inflection expressed on the head and constitute both formally and
semantically a coherent unit in which the individual constituents cannot be separated by
other material, establish a referent on their own or be modified independently of one
another, cf. Giegerich (2009).
2.
Categories of compounds
2.1. Determinative, copu lative and possessive compounds

Although compounds are found in most natural languages (cf. Bauer 2001; Dressler
2006), each language determines the parameters that define its particular class of compounds. The Romance and Slavic languages are not highly compounding languages (especially if the default case of lexical combinations of basic stems without formatives or
functional categories are the focus of attention, see article 38 on noun-noun compounds
in French and article 43 on compounds and multi-word expressions in Slavic), while
compounding is much more productive in the Germanic languages. In the early lndic
language of Sanskrit compounds were extremely prevalent, so much so that the description of the compound types found in the grammars of Sanskrit was adopted as the
descriptive basis for compounds in the historical grammars of Europe (see article 3 on
PA367
20. Composition
367
word-formation in historical-comparative grammar). Whitney's (1993) Sans"/..Tit Grammar distinguished! three classes of compounds, cf. also Mayrhofer ( 1965: 101-103 ). The
most productive class were the determinative compounds, illustrated in (la), in which
the first constituent depends on or describes the second. These were the tatpuru.$a compounds, with the descriptive subclass sometimes referred to separately as karmadharaya
compounds. Copulative compounds, illustrated in ( 1b), are those whose constituents are
coordinated such that the relation between them can be rendered explicitly by the conjunctions 'and' or 'or' . They were usually marked for dual number and were termed
dvandvas ' pair; lit. two-two', a term that exemplifies the subclass of repetitive compounds expressing repetitional meaning (referred to as amredita compounds). Possessive
compounds, cf. (1 c ), had the structure of determinative compounds but were used adjectivally to express possession. This class is also referred to as bahuvrThi compounds, the
term itself an example of the class meaning 'having much rice; lit. much-rice ', cf. Whitney (1993: 481 - 511).
(1)
a.
Determinative:
b.
c.
Copulative:
Possessive:
padodaka
mahar.$i
candradityau
brhadratha
'water for the feet'

'great sage'
'moon and sun'
'having great chariots; lit. great-chariot'
Many synchronic treatments of compounds in the European languages make use of this
classification as well, sometimes having it serve as a basis of the description (as, e.g.,
Fleischer and Barz 2012) and others as established terminology deserving some discussion (e.g., Schmid 2011 ). The categories are exemplified in (2) again with examples
from English and German:
(2)
a.
b.
c.
Determinative:
Copulative:
Possessive:
beach towel I Bierflasche 'beer bottle'

oncologist-surgeon I Dichter-Komponist ' poet-composer'
tenderfoot I Dickbauch '(person with a) fat belly'
From a modern viewpoint, it is not clear that possessive compounds should! comprise a
class of their own distinct from the class of determinative compounds. (Whitney 1993:
481 concedes that possessive compounds have the structure of detenninative compounds
but "with the idea of possessing added".) Today, a number of linguists would argue that
the possessive interpretation of these compounds arises on the basis of their determinative structure via a metonymic shift in meaning (for example, Fleischer and Barz 2012:
179; Schmid 2011: 125; Motsch 2004: 376; Olsen 2012: 2121), an idea which can apparently be traced back at least to Coseriu (1977). Metonymic shift is a semantic process
that applies freely to (both simplex and complex) words, see article 61 on word-formation and metonymy: just as it is possible to say the suits referring to 'persons wearing
suits', one can use hardhat as a metonymic epithet for 'a person wearing a hardhat'. The
shifted word identifies the intended referent as the possessor of the particularly salient
property that it expresses. A different view is held by Marchand (1969), however, who
assumes that an actual morpheme without phonetic content (i.e. a zero morpheme) performs the function of shifting the reference away from the actual head to a metonym of
the head so that birdbrain 'one having a birdbrain' would have the structure [[bird +
brain] 0] parallel to pigtailer [[pig + tail] er] where the overt morpheme -er signals
PA369
20. Composition
369
vessels-containers' as well as the alternative subtype from Mordvinic vest' -kavkst' 'once
or twice'. Co-compounds represent - to different degrees - loose word structures as can
be seen by the fact that they can carry inflection on both constituents indicating that
both stems are visible to the syntax (cf. the Mordvinic nominal co-compound ponks. tpanar. t 'clothes; ht. trouser.PL-shirt.PL' with the !-suffix encoding plural and as well as
the verbal kui.i-valg.i 'fly up and down; lit. ascend.PRES.SG-descend.PRES.SG' with the
i-suffix indicating present singular). Other functional categories can accompany both
stems as well. For example, the generalizing co-compound in Mari jiid.et-kec.et 'day
and night; lit. night.also-day.also' has an additive focus particle et 'also' repeated after
each stem. These co-compounds are described in Walchli (2005: 13 7- 158, see also
articles 179, 180, 187, 190 and 206 on Mari, Mordvinic, Crimean Tatar, Chuvash and
Avar). Furthermore, Arcodia, Grandi and Walchli (2010: 180) mention cases where cocompounds display even more clearly a discontinuous structure like that found in the
East Asian language Hmong Daw, cf. (4):
(4)
kuw
I
lub
CL
teb
land
tub
CL
chaw
land
Here the classifier lub precedes the two nominal heads teb and chaw, both meaning
'land'' that together with kuw, a possessive pronoun in the 1st person, glossed by the
authors as ' I', give the meaning 'my land', an example of a synonymic co-compound.
Such properties are not shared by the coordinative appositional compounds of the
more western European languages. The view is often expressed in the literature that they
have two semantic heads, e.g., Plag (2003: 146), but formally speaking, at any rate,
coordinative appositive compounds have a unique head. The neuter gender of German
Zeuge-Opfer, as an example, demonstrates this; it correlates with the neuter gender of
the second constituent and not with that of the masculine first constituent. The formal
head is also unique in Romance coordinative appositives where the compounds are in
general left-headed: the gender of coche casa 'van; lit. car-house' matches that of the
masculine left constituent, not the feminine second constituent. The picture is complicated somewhat in Romance due to the fact that the appositive compounds quite strictly
inflect for plural on both stems, cf. Sp. poetas-pintores 'painter-poets'. But it needs to
be kept in mind that compounding in Romance is less typically morphological than in
Germanic. Article 144 on Portuguese documents that double inflection for plural can
even occur with nominal determinative compounds whose second constituent functions
much like an adjective as in palavra-chave ---+ palavras-chaves 'key word ' . With their
left-headed structure and productive "improper" or "syntagmatic" NPrepN patterns (cf.
Fr. pomme de terre), there appears to be more syntactic influence present in the compounds of the Romance languages, even though these compounds must be considered
lexical (cf. Rainer and Varela 1992; Kornfeld 2009; Fradin 2009) and not syntactic (DiSciullo and Williams 1987: 81) patterns, cf. section 3.
2.3. Endocentric vs. exocentric compounds

Bloomfield (1933) did not adopt the traditional classes of Sanskrit compounds as the
basis of his discussion. Instead, he proposes a general classification according to whether
PA370
370

the compound has the same grammatical structure found in the phrases of the language
(cf. blackbird and a black bird), in which case it is termed "syntactic", or whether there
is no parallel construction in the syntax, in which case the compound is considered
"asyntactic", cf. door-knob. Intermediate cases in which an equivalent phrasal structure
is present but the word order of the compound is different (to housekeep vs. to keep
house) or in which the compound lacks a functional category necessary in the syntax
like an article (turnkey vs. to turn the key) are termed "semi-syntactic". Copulatives like
bittersweet are "semi-syntactic" because of the missing conjunction, cf bitter and sweet
(Bloomfield 1933: 233-235).
In addition to these classes, Bloomfield proposes another fundamental distinction that
crosscuts the previous classification, namely whether the compound has a head or not.
Compounds like blackbird, doorknob and bittersweet that as a whole have the same
semantic "function" as their head constituents (bird, knob and sweet) are termed "endocentric"; but turnkey 'jailer; lit. one who turns a key' and bittersweet (the name of a
plant) that do not are "exocentric". Red-head and longhorn are also exocentric because
their referent is not understood as belonging to the semantic category of the head. Examples like chimney-sweep and clambake, on the other hand, must be considered endocentric due to the fact that their nominal head arises on the basis of the regular process of
conversion (to sweep/bake ~ a sweep/bake).
2.4. Synthetic vs. primary/root compounds
In the course of his discussion of exocentric structures like loudmouth, Bloomfield touches on endocentric nominal constructions that carry the same possessive meaning as the
exocentric redhead type, but whose possessive meaning is signaled by an overt morpheme such as -ed, -d or -t as exemplified by blue-eyed, cf. Bloomfield ( 1933: 23 1233 ). He terms this structural type synthetic and their deverbal counterparts meat eater
semi-synthetic because of the word order difference to the phrase to eat meat. Marchand
(1969: 15- 18) uses the same terms for his hunchbacked and watchmaker types. The
term synthetic compound apparently originates with von Schroeder (l 851-1920), see
article 33 on synthetic compounds in German, where it was coined to describe the formation process of German examples like Machthaber ('power holder ' ) that seem to involve
a synthesis of compounding and derivation in their genesis: neither the compound
*machthab 'to power hold' nor the derivation *Haber ' holder' exist as words in German - the construction arises only when all three morphemes join together. From the
start, the discussion of these structures has revolved around the question of their correct
analysis. Are they to be considered compounds with the structure in (5a) or derivations
as in (5b)? The original term synthetisches Kompositum is, in fact, no longer in use in
German linguistics, but was replaced by the term Zusammenbildung in order not to
prejudice the issue by referring to such structures as compounds (see article 33 on synthetic compounds in German and Leser 1990). Although both Bloomfield and Marchand
see a parallel between the deverbal type meat eater, watchmaker and an underlying
sentence structure, they nevertheless consider these composites to be compounds, structured similarly to the primary compounds that contain purely nominal constituents like
steamboat, oilwell, cf. Marchand (1969: 18).
PA371
20. Composition
(S)
a.
b.
371
[[watch]N [makev + er]N]N

[[watchN makev]v -er]N
Many linguists including Allen (1978), Selkirk (1982), Lieber (2004), Jackendoff (2009),
ten Hacken (2010) and Olsen (2012) have espoused the analysis in (Sa) as well. The disadvantage of the alternative analysis in (Sb) is that the derivation of synthetic compounds
from a NV base cannot account for their extreme productivity, since NV compounds are
not productive in Gennanic (cf. Booij 1988). If the compound structure in (Sa) is supplemented with the idea of argument inheritance, however, the awkwardness of the deverbal
head alone can be explained along the same lines as the use of a transitive verb without its
obligatory argument: the deverbal noun maker inherits a modified version of the argument
structure o f the transitive verb make (cf. Fanselow 1988, Bierwisch 1989 and article 60 on
word-formation and argument structure) and assigns the inherited object to its first constituent. Without the satisfaction of its argument, maker is as incomplete as the verb phrase *he
makes. It was most likely the insights of this early discussion that opened the way for modern theories of word-formation to extend the content of the term synthetic compound from
its original sense as a synthesis of compounding and derivation to become equated with
the class of compounds based on a verbal interpretation, i.e. those compounds whose first
constituent is understood as an argument of their deverbal head, regardless of whether the
head can occur alone (programm er - computer programmer) or not (??keeper - house
keeper). Due to the verbal nature of their head, synthetic compounds (also termed verbal
compounds) have a narrow range of unambiguous interpretations that can be contrasted
with the more open and often ambiguous meanings of compounds whose heads are not
deverbal, but consist of nominal roots or stems called primary (or root) compounds. This
can be illustrated by comparing the verbal interpretations of cost reduction, snow removal,
law enforcement, oil spillage and window installer with the semantically more variable interpretational possibilities of primary or root compounds like dance music, merit pay, designer drug, sweetheart contract and monsoon wedding. Synthetic compounds are also
found in Slavic, cf. Cz . dfev-o-rub-ec ' woodcutter ' and R. kanat-o-chod-ec 'ropewalker '
where they represent a more productive alternative to the less productive nomin al root compounds (see article 43 on compounds and multi-word expressions in Slavic).
3. Exceptions t o the head generalization

3.1. No overt head, forma l or semant ic
The class of synthetic compounds is lacking in Romance. The Romance languages,
rather, di splay a pattern of agentive and instrument compounds that are semantically
similar to the synthetic compounds of Germanic and Slavic, but without a d erived head
(see article 39 on verb-noun compounds in Romance as well as articles 144, 14S, 147,
lSO and lSl on Portuguese, Spanish, French, Itali an and Romanian):
( 6)
a.
b.
c.
cuentachistes
rabat-joi
cantastorie
'joke teller; lit. tell jokes'

'spoil sport; lit. reduce joy'
'street singer; lit. sing stories'
Spanish
French
Italian
PA372
372

d. limpa-chamines
e. fura-becuri
'chimney cleaner; lit. clean chimneys'

'tall person; lit. steal lightbulbs'
Portuguese
Romanian
Remnants of this pattern are found in English as well where they were most likely
borrowed during Middle English times from French (cf. Kastovsky 2009), for instance:
scofflaw 'law violator', spoilsport, sawbones 'surgeon ', daredevil ' recklessly bold person', etc. This VN pattern cannot be considered synthetic because it does not contain an
overt signal of derivation, but consists of the combination of the (Yd person singular or
imperative) stem of a verb together with a noun marked for plural if it is a count noun
(cf. Fr. coupe-legumes 'vegetable cutter; lit. cut-vegetables') or as an indefinite singular
if it belongs to the category of mass nouns (cf. Sp. quitaesmalte 'polish remover; lit.
remove-polish') that is understood as the object of the verb. Article 150 on Italian shows
that the nature of the stem can be more precisely determined in Italian where a few
unambiguous cases are found that show it to be formally identical to the imperative. An
interesting aspect of the productive VN pattern is that the words are headless, both
formally and semantically: Sp. abrepuertas denoting a 'doorman' (lit. ' open-doors') is
not a feminine plural noun as puertas but carries the default gender of masculine singular; furthermore it denotes neither a subcategory of the constituent lexeme 'open' nor of
'door'. Article 145 on Spanish reports that the pattern creates nouns even when no noun
is present as in matasiete 'braggart; lit. kill-seven' and mandamas 'boss; lit. order-more'.
Such headless composites constitute an exception to the generalization formulated in
section 1.2 that compounds are headed structures. In fact, these structures are both formally and semantically headless. Formally they are assigned the default gender masculine and semantically they are perhaps best interpreted as examples of coercion, i.e. a
shift from the property they denote (e.g., to open doors) to the person or instrument that
is typically .associated with this property (cf. Olsen 20 12 and article 61 on word-formation and metonymy). Using Bloomfield's terminology, they are "exocen.tric" since neither
constituent functions as the head of the compound.
The term "exocentric" in Bloomfield's sense is broader than the Sanskrit term bahuvrfhi introduced in section 2.1 in that it was conceived to encompass both the semantically and formally head less VN compounds (Bloomfield's examples are E. pickpocket and
lickspittel 'a fawning subordinate') as well as semantically headless AN compounds
such as redhead and longhorn (cf. Bloomfield 1933: 235- 237) with a metonymically
shifted meaning. The term bahuvrfhi is defined more narrowly as NN or AN compounds
of the descriptive determinative type "with the idea of possession added, turning them
from nouns into adjectives" (Whitney 1993: 481). Bloomfield's notion "exocentic",
moreover, inc ludes two further Sanskrit classes beyond the bahuvrfhis, namely the numeratives (dvigu) such as sixpense,fortnight and adverbials with a nominal head (avyayfbhava) like bareback, uphill (Bloomfield 1993: 23 7).
The PN pattern of Romance as in Sp. sinvergiienza '(one) without shame', Fr. sanscoeur '(one) without heart' and It. senzatetto '(one) without roof' must be considered
"exocentric" as well: the property expressed by a preposition together with a noun that
is understood as its object is coerced into a reading as a person who exemplifies this
property (Olsen 2012: 2142). Another class of exocentric, or semantically and formally
headless word structures, are the scalar co-compounds characteristic of East and South
East Asia. In this case, two adjectives that denote the opposite poles on a scale hyponym-
PA373
373
20. Composition
ically express the quality that names the scale; thus a noun arises on the basis of two
adjectives, cf. Walchli (2005: 152-154).
(7)
a.
b.
c.
srab-mthug
xaluun xiijten
tsopats mkalto
'density; lit. thin-thick'

'temperature; lit. heat-cold'
'size; lit. big-small'
Tibetan
Khalkha
Tokharian
3.2. Formal, but no semantic head

The pure possessives or core cases of bahuvrzhis, like E. tenderfoot 'newcomer, beginner' or G. Dickbauch lit. 'fat-belly', are exocentric in Bloomfield's sense because their
second constituents foot and Bauch have a different function in the combinations which
mean 'inexperienced person' and 'fat person', respectively, than when they occur as
independent words. They are not the semantic heads of the construction. On the other
hand, they do function as the formal heads of the word: that is the nouns foot and Bauch
determine the gender and plural class of the whole compound (tenderfeet, Dickbiiuche).
Such cases as these demonstrate the necessity of distinguishing between the semantic
and the formal notion of head.
3.3. Left-headed compounds

Another deviation from the head generalization are the left-headed compounds of the
Romance languages. Morphologically speaking, these structures do not represent the
default case because - when considering the NN pattern - pluralization disrupts the
structural integrity of the words, cf. Fr. timbre~ paste 'postage stamps', Sp. hombre~
rana 'frogmen', and It. vagon letto 'sleeping cars' . This is even more obvious in the
productive NPrepN pattern of so-called "improper" or "syntagmatic" compounds, exemplified by the following forms (see articles 144, 145, 147 and 150 on Portuguese, Spanisch, French and Italian), where a semantically bleached remnant of a functional category
from syntax signals the subordination of the modifying constituent to the head.
(8)
a.
b.
c.
d.
bota de lluvia
chemin de fer
mulino a vento
casa de banho
'rain boot'
'railroad'
'windmill'
'bathroom'
Spanish
French
Italian
Portuguese
The literature reports cases of such NPrepN combinations that have been reduced to the
more typical NN structure such as Sp. tren de mercancias ~ tren mercancias 'freight
train' (Rainer and Varela 1992: 120) and Fr. stylo a bille ~ stylo-bille 'ballpoint pen'
(Fradin 2009: 433), demonstrating a tendency to level the marked to the unmarked word
form. There have been attempts to view the syntactically opaque preposition as a linking
morpheme similar to the opaque case markings in German compounds Freundeskreis
'circle of friends' or Sonnenschein 'sun shine', cf. Ralli (2008) and article 32 on linking
elements in Germanic. This view, however, is highly controversial since the prepositions,
PA374
374

although weakened semantically, still define clear semantic patterns among the compounds. In fact, Romance has a true linking morpheme that does parallel the Germanic
linking morphemes in function, namely the -i- as in Sp. puticlub 'hostess club' and It.
fruttivendolo 'fruit seller', see articles 145 and 150 on Spanish and Italian.
4.
Bound compound constituents
4.1. Combin in g forms of neoclassical compounds

Neoclassical compounds differ from regular compounds in a number of ways. First, in
defiance of the definition of compounds in section 1. 1, their constituents are not freely
occurring lexemes. Rather, they consist of combinations of elements taken over from the
Classical languages, Greek and Latin, that do not occur freely in the modern languages.
However, the combinations themselves are formed according to modern compounding
rules, hence the term "neo"classical. Neoclassical compounds are usually comprised of
simple combinations of two elements; the recursion found in regular compounding is
usually absent. They appear in all the language families of Western Europe and are so
prevalent in the international terminologies of science, technology, scholarship, etc., that
they are often referred to as "internationalisms" (cf. Bauer (1998) and the articles 93,
94, 95, and 96 on foreign word-formation in German, English, Italian and Polish). Examples are E. neurology, democracy, stethoscope, suicide (Bauer 1998), Fr. aerodrome,
geographe, anthropomorphe, pathogene (Zwanenburg 1992), Polish neojita 'neophyte',
poligamia, ksenofobia 'xenophobia', neurologia (Szymanek 2009) and Basque telefono,
mikrobiologia,jilologia, elektromagnetismo (see article 182 on Basque).
Because the lexemes involved in neoclassical compounds represent a class of forms
that do not occur freely in the compounding languages, they are often termed combining
forms. In their formal aspects, combining fonns seem to have much in common with
prefixes and suffixes. The majority are specialized to either the initial or final position
of a combination; e.g., astro-, bio-, electro-, geo-, gastro-, tele- occur initially wh ile
-cide, -cracy, -graphy, -phobe, -scope occur finally (cf. Plag 2003: 155 ff.). They are
also transparent in meaning and form the basis of productive patterns, cf. anthropology,
astrology, geology, pathology, psychology, etc., and geology, geography, geometry, geophagy, etc. Nevertheless, combining forms cannot be equated with affixes because prefixes and suffixes do not combine with one another as the combining fonns typically
do. Their lexical category and, in the case of nouns, their gender and inflectional class
are not set at a default for the entire set of patterns, but vary with the particular head.
Although they do not constitute free lexemes, the individual combining forms (at least
those that appear in head position) are lexically marked for these features which they
project onto, the whole word, cf. the Gennan nouns die Biologie, pl.: die Biologien; der
Psychologe, pl.: die Psycho logen; das Pathogen, pl.: die Pathogene. The righthand head
structure of these internationalisms is taken over even into languages whose native compounds typically display lefthand heads (cf. section 3.1 and 3.3 as well as article 95 on
foreign word-formation in Italian and article 144 on Portuguese): e.g., Port. gastropatia,
Fr. biologie, It. termometro.
The combining forms of the class of neoclassical compounds differ in their transparency and productivity from the bound roots of the Latinate vocabulary in English as
PA375
20. Composition
represented by -ceive in conceive, deceive, perceive, receive, -mil in omit, submit, transmit, remit and -duce in deduce, induce, reduce, etc. The most pronounced difference
between the two is that the bound roots of the Latinate vocabulary do not share a common semantic component: -tain, for example, does not have a uniform meaning in contain, detain, retain, etc. And, secondly, whereas the combining forms of neoclassical
compounds are productive and continually spawn new forms, the Latinate patterns do
not permit novel creations. These verbs entered the English language from Romance
early on as complete words in specialized meanings so that a transparent preposition +
root structure was never reconstructed. Even though speakers may be aware subconsciously that the roots of these verbs represent a morphological unit (they share the same
allomorphic form under suffixation, cf. conception, deception; submission, remission;
deduction, reduction and detention, retention, cf. Aronoff 1976), their semantic opacity
prohibits speakers from identifying them as morphemes capable of new combinations.
4.2. Positional ly bound lexemes

Another type of bound lexeme not only occurs regularly in native compounds, but is in
actual fact very typical of compounding. What is at issue here are not the results of the
grammaticalization processes by which certain types of compound constituents evolve
over time into affixes or even affix-like units that are classified in word-fonnation grammars as "semi-affixes; semi-prefixes, semi-suffixes" (G. "Affixoide; Prafixoide, Suffixoide'', see article 104 on grammaticalization in German word-formation for a detailed
discussion of both these notions with illustrative examples). Rather, what is central to
this section is another phenomenon that has come to light as a matter of interest in the
structure of the mental lexicon. Psycholinguistic experimentation has demonstrated that
both transparent and opaque constituents of compounds can be primed andl, hence, are
psychologically available to speakers, cf., e.g., Zwitserlood (1994), Libben, Buchanan
and Colangelo (2003) and Libben et al. (2003). In theories of parsing that assume dual
access to entries in the mental lexicon such that decomposition and whole-word access
are simultaneously employed in the retrieval of compound meaning, the activation of
constituent meaning together with the whole-word meaning w ill lead to conflicting information whenever a compound contains a nontransparent constituent. For example, if both
air and line are activated while processing airline (or Hand and Schuh in G. Handschuh),
incongruous information is called up because the compound constituent -line (or -schuh)
does not correspond in meaning to the independent lexeme line (or Schuh). Conflicting
information requires a resolution. Since meaning specialization is widespread in compounds, access to both the individual lexemes and the whole-word meaning of the compound will result in the activation of superfluous information. Solutions to this problem
have been sought that include missing links as well as inhibition processes. that do not
allow access to or suppress the irrelevant meaning of the free lexeme. Libben (2010)
proposes an alternative suggestion that circumvents much of the problem (cf. also article
13 on word-formation in psycholinguistics and neurocognitive research): the specialization in meaning of a compound constituent (what he terms "morphological transcendence") encourages the creation of a positionally bound variant of that constituent that
corresponds to its specialized meaning in the context of the compound. This means, for
375
PA376
376

example, that in addition to the entry for the independent word Line, the mental lexicon
would contain a positionally bound entry -line which would be targeted during the decomposition of airline. This bound constituent would then cover busline, shipline, cruiseline as well. The same would apply to -lace in necklace, shoelace, bootlace, etc., vis-avis the related, but no longer relevant, free word lace. Positionally bound constituents
will differ in their degree of opacity. For example, industry refers in its core meaning to
the production of goods or resources within an economy. Typical compounds with this
lexeme are metal industry, petroleum industry, iron industry, auto industry, meatpacking
industry, electronic industry, etc. In recent creations like entertainment industry and happiness industry, the head lexeme deviates from its core meaning in as far as entertainment
and happiness denote concepts too abstract to fall under its usual production output. The
lightly bleached meaning of the constituent -industry might therefore at some time establish its own entry in the mental lexicon, allowing further compound creations with the
same meaning extension.
Positionally bound constituents often spawn larger patterns: -free in sugar-free, fatfree, salt-free, gluten-free, cajfein-.free, error-free, toll-free, etc. is related to its free counterpart free, but is not used in its core sense since it simply signals absence of an unwanted object. Likewise, smart-, which implies the presence of information-based technology
in smart phone, smart card, smart key, smart board, etc., is related to, but simultaneously
differs from, the independent adjective smart. More opaque cases like E. pig and G.
Schwein 'pig' could have the positionally bound variants pig- and schweine- w ith the
semantics of an intensifier as in E. pig ignorant, pig arrogant and G. schweineteuer
'extremely expensive', schweinekalt 'extremely cold'. Seen in this light, these lexemes
would not have to be - implausibly - classified as prefixoids. Positionally bound entries
in the mental lexicon would capture speakers ' knowledge of regular patterns and relations in their inventory of words. First constituents in German compounds suggest the
need for an independent entry of compound constituents at any rate, due to the fact that
they occur in unpredictable combinatorial variants (see article 32 on linking elements in
Germanic).
5.
Compound meaning
5.1. Compound template

Earliest studies of compound meaning attempted to capture the general semantic patterns
instantiated by established compounds and use these relatively fixed sets of mea ning as
a basis for predicting the meanings that are possible in novel compounds. The patterns
were described in the form of underlying syntactic relations (subject, object, object of
preposition, attribute, predicate, etc.) or basic semantic-thematic roles (purpose, location,
instrument, material, and so on). Most often a combination of the two were used, cf.
Jespersen (1909), Hatcher (1960), Marchand (1969), Lees (1970), Adams (1973), Levi
( 1978) and Warren (1978). Work by Zimmer (1981) and Downing (1977) shifted the
focus away from classifying the relations that held between the constituents to the idea
that compound meanings could arise on the basis of any relation that w as "appropriately
classificatory". Bauer (1978: 122) captured the basic intuition of this new approach by
PA377
20. Composition
postulating a single covert PRO-verb that was understood as establishing a plausible
relation. Allen (1978: 93) introduced her "variable R condition" in an attempt to characterize the range of possible compound meanings by requiring a matching of semantic
features between the constituents. Dowty (1979: 316) formalized the idea of an open
relation between the constituents by relating the external arguments of the two predicates
involved by means of an unspecified relation variable R indexed with the feature "appropriately classificatory". Around the same time, Fanselow ( 1981) isolated the class of
"Rektionskomposita" [governed compounds], i.e. compounds with relational heads
whose interpretation was based on the inherent argument structure of the nominal or
deverbal head, from two further sets, namely those displaying basic relations ('and',
'location' and ' part of') and those with numerous options available. For the latter he
formalized a series of semantic rules that allowed stereotypical relations to be inferred
on the basis of the semantic content of the constituents. In fairly recent work, Jackendoff
(2009: 122) postulates two general compound schemata - one based on argument structure yielding the class of synthetic interpretations and another based on modification
allowing for the " unlimited set of possibilities" exemplified by primary compounds.
Olsen (2012: 2138-2140), on the other hand, proposes that compound interpretation
proceeds by means of a uniform reasoning strategy that relates the predicates of a compound without distinguishing between primary and synthetic interpretations. These different meaning types are simply the result of the specific choices made during the instantiation of the unspecified relation connecting the component predicates. A compound
template, adopted from the modification templates of Higginbotham (1985), Maienbom
(2003) and Bucking (2009, 2010), is used to implement this strategy. This compound
template can be thought of as a lexical schema in the grammar that combines two arbitrary predicates and relates their referential arguments by means of an unspecified relation.
In the transition from the grammar to an actual utterance, the content of the implicit
relation will be fleshed out. As an example of such a specification, the relation 'made
of' might be inferred on the basis of the predicates chocolate and cake in (9a), yielding
the interpretation 'cake made of chocolate'. For the predicates pepper and grind in (9b)
the cognitively most prominent relation is already found in the semantic structure of
grind producing the probable reading 'something that grinds pepper' :
(9)
a.
b.
c.
chocolate cake
pepper grinder
waitress-actress
Hence, the template does not enforce an artific ial distinction between two classes of
meaning, the so-called primary and synthetic interpretations. Many borderline cases exist
between the two reading types that make clear-cut decisions notoriously difficult. For
example, are relations such as 'synthesize', 'certify', 'steal', 'trap', 'cover' and '(to)
pilot' in speech synthesis, death certificate, car thief, bear trap, bed cover and airplane
pilot argument relations or are their readings based on modification? Furthermore, a
single structure can receive both types of meaning: horse rescuer can be 'one who
rescues horses' or 'one who rescues something on horseback' and a shop corner can be
'corner of a shop' or 'corner with a shop located on it'.
The proposed interpretative strategy accommodates coordinative appositive interpretations as well. Since the predicates in (9c) carry information about career types, the
377
PA379
20. Composition
adding the meanings of the conjuncts together to form an arbitrary group, but rather by
subsuming the conjuncts under a common conceptual frame which Lang (1991: 605) terms
a "common integrator" and which is integrated into the semantic structure of the sentence.
Co-conjuncts that are formally and semantically similar will be easily assimilated to a common integrator. In such cases, a more natural and direct interpretation is possible. This is
the case in the Mordvinic example where the common integrator of 'trousers and shirts' is
clearly 'clothes', supplying the meaning of the co-compound. Interestingly, the additive,
collective, generalizing and scalar interpretation groups of the co-compounds postulated
by Walchli (2005) and discussed in section 2.2 and section 3.1, provide confirmation of
this. Their compound meaning is equivalent to the common integrator of their constituents
and is, therefore, hyperonymic. Furthermore, the directness of the common integrator as
the superordinate category of the constituents is the source oftheir naturalness: the Mordvin
combination t' et' a.t-ava. t 'father.PL-mother.PL' denotes the hyperonymic concept 'parents', Chuvash set-Su 'milk-butter' picks out the superordinate concept 'dairy products',
Vietnamese ban ghe ' table-chair' refers to 'furniture ' and Chinese gao ai ' high-low', conveys the meaning ' height', etc. (cf. Olsen 2014). Hence, Arcodia, Grandi and Walchli's
(2010: 182) assumption that co-compounds may have originated via the grammaticalization of an asyndetic phrasal coordination is a highly plausible explanation for the source of
their meaning.
5.3. Pragmatic co nsiderations

The simple, non-formal, open and ambiguous character of the basic compound template
discussed in section 4.1 yields basic morphological objects that are assumed here to be
the default case of compounding. But in spite of their pragmatic flexibility, even these
structures are subject to clear restrictions. In the previous sections the influence of the
morphological head on the interpretation possibilities of a compound was shown. There
are further restrictions of a pragmatic nature that are placed on a compound''s interpretation by virtue of its function in communication and as a name for a relevant category in
the speaker's experience. The most general restriction requires, in accordance with the
cooperation principle of Grice (1975), the interpretation of a compound to be inferable
on the basis of the overt information present. If it is not possible to find a plausible
interpretation, the compound fails to fulfill its communicative purpose. For example, the
novel compound weed bicycle would require a great deal of context or common background in order to make sense. In addition, several linguists have noted that, although
the range of possible relations between the constituents is very broad, relations based
purely on negation, absence or dissimilarity are nevertheless not possible, cf. Zimmer
(1971: Cll), Downing (1977: 825) and Dowty (1979: 316). Notions such as 'not located
at', 'not similar to' or 'not for ', open up such a large number of possibilities that they
would not provide a clear basis for classification. A reading based on an inferred relation
that contains an element of negation, on the other hand, will be possible (cf. mud guard
'guards against mud'). Next, a plausible referent needs to be chosen from one of the
domains in the human ontological system. Hence, cowboy hat cannot have a coordinative
appositive denotation *'x is a cowboy and a hat' although the determinative meaning
'hat worn by cowboys' is fine. And, as Olsen (2004: 23) further explains, the coordina-
379
PA380
380

tive reading of father-daughter is permissible in the nonhead (modifier) position of a
compound as in father-daughter dance, but it cannot occur in an unembedded position
in a coordinative interpretation without violating the Principle of Ontological Coherence,
cf. *father-daughter, because no single individual can simultaneously be a father and
daughter. Hence, speakers do not entertain a coordinative interpretation for German Vatertochter, but immediately construe either a determinative 'daughter favored by her
father' or an argument reading 'daughter of the father' . On the other hand, E. actressdaughter and G. Schriftsteller-Freund can have both a coordinative ('actress and daughter', 'writer and friend') and a determinative reading ('daughter of an actress', 'friend
of a writer'), depending on the context. Such considerations demonstrate the openness
of interpretation predicted by the underspecified relation of the compound template: It
is not the combination of the lexemes father and daughter that is deviant - such combinations are structurally possible. But restrictions - pertaining both to the human conceptual system and the context of use - apply to limit the set of possible meanings of a
combination, cf. Meyer (1993: 171 ff.).
Furthermore, since the function of a modifier is to restrict the meaning of the head
that it modifies, a headed compound violating this endocentric structure will be deviant.
Compounds like ??water fish or ??building house are extremely odd for this reason,
unless a context can be construed revealing a set of alternatives which would allow the
recovery of a sensible hyponymic reading. In order to pick out a subcategory that is
relevant to the speaker's cognitive organization in a coherent manner, the modifier must
make reference to a property that is not necessarily shared by all members of the head
category. Thus, *whisker cat has no raison d'etre, unless it can be contrasted with a type
of cat that has no whiskers. Changes in society allow combinations to emerge that once
would have had no function: a generation ago cord telephone would most likely have
been eschewed, but is possible today due to the contrast with cordless phones. Again,
such examples attest to the correctness of the open pragmatic interpretation strategy
sketched in section 4.1.
Subsumptive, clarifying, redundant and synonymous compounds, on the other hand,
serve a function, namely to make the unclear or demotivated meaning of one constituent
more perspicuous, cf. E. tuna fish, pebble stone or G. Maultier 'mule; lit. mule animal',
where the meaning of Maul 'mule' is now unclear to speakers. Also the word p ea was
added to E. chick < ME chick 'pea' to clarify its meaning, cf. chick pea, and G. Kichererbsen 'chick pea' was created for the same reason from MHG kicher 'pea' and Erbse
'pea' . Furthermore, synonymic (cf. E. pathway, courtyard, G. Zeitalter 'time age', Streifzug 'tract passage'), imitative (nap schmap 'not a real nap'), rhyming and alliterative
(razzle-dazzle, teeny-weeny, topsy-turvy, G. Techtelmechtel 'flirtation') and reduplicating
compounds (hush-hush , goody-goody, G. Pinkepinke 'money' ) may occur sporadically.
Their use in emphatic or playful language allows them to circumvent the pragmatic
restrictions that are operative in normal discourse.
5.4. Origin of compounds in a protolanguage?

It has been suggested that compounds originated, as far as their evolution is concerned,
outside of grammar proper in a primitive pragmatic system of communication. Fanselow
PA382
382
tion a broad spectrum of compound structures conforms to quite simp~e basic principles
of concatenation and meaning constitution: two predicates conjoin and - guided by the
formal modifier-head structure which they instantiate - result in the designation of a
hyponymic or subordinate concept. One type of exception to this generalization occurs
via the application of the regular semantic process of metonymic shi ft to a headed morphological structure. A shift in reference, but not in category, results yielding an exocentric interpretation. Nonheaded morphological structures are also subject to the semantic
process of metonymic shift or coercion and in this case a new category results that is
not represented in the formal structure of the base. In all these cases, general ontological
and pragmatic principles are at play, placing restrictions on the process of inference and
guaranteeing the comprehensibility of the resulting structures. Loose word structures that
differ from the default case of morphological structures in that they make more than one
head visible to the syntax, on the other hand, are subject to different principles. They
result in hyperonymic meanings, the choice of which is guided by the hierarchical organization of the human conceptual system.
7. References
Adams, Valerie
1973
An Introduction to Modern English Word Formation. London: Longman.
A llen, Margaret
1978
Mo rpholog ical investigations. Ph.D. dissertation, University of Connecticut.
Arcodia, Giorgio F., Nicola Grandi and Bernhard W i.ilchli
Coordination in compounding. In: Sergio Scalise and Irene Vogel (eds.), Cross-Disci20 l 0
plinary Issues in Compounding, 177- 197. Amsterda m/Philadelphia: Benjamins.
Aronoff, Mark
1976
Bauer, Laurie
1978
The Grammar of Nominal Compounding with Special Reference to Danish, English and
French. Odense: Odense University Press.
Bauer, Laurie
Is there a class of neoclassical compounds, and if so is it productive? Linguistics 36:
1998
403-422 .
Bauer, Laurie
2001
Compounding. In: Martin Haspelmath, Ekkehard Konig, Wulf Osterreicher and Wolfgang Raible (eds.), Language Universals and Language Typology. Vol. 1, 695-707.
Bauer, Laurie
2005
The borderline between derivation and compounding . In: Wolfgang U . Dressler, Dieter
Kastovsky, Oskar E. Pfeiffer and Franz Rainer (eds.), Morphology and its Demarcations,
97- !08. Amsterdam/Philadelphia: Benjam ins.
Bauer, Laurie
2008
Dvandva. Word Structure 1: 1-20.
Bickerton, Derek
1990
Language and Species. Chicago, IL: The University of Chicago Press.
Bierwisch, Manfred
1989
Event nominalizations: Proposals and problems. In: Wolfgang Motsch (ed.), Wortstruktur and Satzstruktur, 1-73. Berlin: Akademie Verlag.
PA383
20. Composition
Bloomfield, Leonard
1933
Language. Chicago: U nivers ity of Chicago Press.
Booij, Geert
1988
The relation between inheritance and argument linking: Deverbal nouns in Dutch. In:
Martin Everaert and Arnold Evers (eds.), Morphology and Modularity. In Honour of
Henk Schultink, 57- 73 . Dordrecht: Foris .
Bu cking, Sebastian
2009
How do phrasal and lexical modification differ: Contrasting adjective-noun combinations in German. Word Structure 2: 184- 204.
Bucking, Sebastian
2010
German nominal compounds as underspecified names for kinds. In: Susan Olsen (ed.),
New i mpulses in Word-Formation , 253- 28 1. Hamburg: Buske.
Coseriu, Eugenio
1977
Inhaltliche Wortbildungslehre (am Beispiel des Types ,,coupe-papier"). In: Herbert E.
Brekle and Dieter Kastovsky (eds.), Perspektiven der Wortbildungsforschung, 48- 61.
Bonn: Bouvier.
Downing, Pamela
1977
On the creation and use of English compounds. Language 53: 810-842.
Dowty, David
1979
Word Meaning and Montague Grammar. Dordrecht: Kluwer.
2006
Compound types. In: Gary Libben and Gonia Jarema (eds.), The Representation and
Processing of Compound Words, 23-44. Oxford: Oxford University Press.
DiSciullo, Anna Maria and Edwin Williams
1987
On the D efinition of Word. Cambridge, MA: MIT Press.
Erben, Johannes
1983
Einfuhrung in die deutsche Wortbildungslehre. 2"d ed. Berlin: Schmidt.
Fanselow, Gisbert
1981
Zur Syntax und Semantik der Nominalkomposition. Tubingen: N iemeyer.
Fanselow, Gisbert,
1985
Die Stellung der Wortbildung im System kognitiver Module. Linguistische Berichte 96:
91- 126.
Fanselow, Gisbert
1988
Word syntax and semantic principles. In: Geert Booij and Jaap van Marie (eds.), Yearbook of Morphology, 95-122. Dordrecht: Foris.
Fleischer, Wolfgang
1969
Wortbildung der deutschen Gegenwartssprache. Tubingen: Niemeyer.
2012
Wortbildung der deutschen Gegenwartssprache. 4th ed. Tu bingen: Niemeyer.
Fradin, Bernard
2009
IE, Romance: French. In: Rochelle Lieber and Pavo! Stekauer (eds.), The Oxford Handbook of Compounding, 417- 435. Oxford: Oxford University Press.
Giegerich, He inz
2009
Compounding and Lexicalism. In: Rochelle Lieber and Pavo! Stekauer (eds.), The Oxford Handbook of Compounding, 178- 200. Oxford: Oxford University Press.
Grice, Herbert P.
1975
Logic and conversation. In: Peter Cole and Jerry L. Morgan (eds.), Syntax and Semantics. Vo l. 3: Speech acts, 4 1- 58. New York: Academic Press.
Grimm, Jacob
1826 Deutsche Grammatik, 2. Theil: Ableitung und Zusammensetzung. Gottingen: Dieterich.
Hatcher, Anna
1960
An introduction to the analysis of English noun compounds. Word 16: 356- 373.
383
PA384
384

H enzen, Walter
1947
Deutsche Wortbildung. Ti.ibingen: N iemeyer.
Higginbotham, James
1985
On semantics. Linguistic Inquiry 16: 547- 593.
H oeksema, Jack
1985
Categorial Morphology . New Yo rk: Garland.
Jackendoff, Ray
1999
Possible states in the evolution of the language capacity. Trends in Cognitive Sciences
3: 272- 279.
Jackendoff, Ray
2002
Foundations of Language. Oxford: Oxford University Press .
Jackendoff, R ay
2009
Compounding in the parallel a rchitecture and conceptual semantics. In: Rochelle L ieber
and Pavol Stekauer (eds.), The Oxford Handbook of Compounding, 105-128. Oxford:
Jespersen, Otto
1909 A Modern English Grammar on Historical Principles. Part I: Sounds and Spelling.
London: G eorge Allen & U nwin.
Johansson, S tig
1980
Plural Attributive Nouns in Present-Day English. Lund: LiberLiiromedel.
Kastovsky, Dieter
2009
Diac hronic perspectives. In: Rochelle Lieber and Pavo! Stekauer (eds .), Th e Oxford
Handbook of Compounding, 323-340. Oxford: Oxford University Press.
Kornfeld, Laura Malena
2009
IE, Romance: Spanish. In: Rochelle Lieber and Pavo) Stekauer (eds.), Th e Oxford Handbook of Compounding, 436- 452. Oxford: Oxford U niversity Press.
Lang, Ewald
1991
Koordinierende K onjunktionen. In: Arnim von Stechow and Dieter Wunderlich (eds.),
Semantics. A Handbook, 597- 623 . Berlin/New York: de Gruyter.
Lees, Robert
1970
Problems in the grammatical analysis of Eng lish nominal compounds. In: M anfred Bierwisch and K arl-Erich Heidolph (eds.), Progress in Linguistics, 174- 186. The H ague:
Mouton.
Leser, Martin
1990
Das Problem der ,Zusammenbildungen '. Eine lexika/istische Studie. Trier: Wissenschaftli cher Verlag Trier.
Lev i, Judith
1978
Th e Syntax and Semantics of Complex Nominals. New York: Academic Press.
Libben, Gary
2010
Compound words, semantic transparency, and morphological transcendence. In: Susan
Olsen (ed.), New Impulses in Word-formation, 3 18-330. H amburg: Buske.
L ibben, Gary, Lori Buc hanan a nd Annette Colangelo
2003
Morphology, semantics and aphasia: The fa ilure of inhibitio n hypothesis. Logos and
Language 4(1): 45-53.
L ibben, Gary, Martha Gibson, Yeo Born Yoon and D ominiek Sandra
2003
Compound fracture: T he role of semantic transparency and mo rpho logica l headedness.
Brain and Language 84: 50- 64.

L ieber, Rochelle
2004
Morphology and Lexical Semantics. Cambridge: Cambridge U niversity Press.
M aienborn, C laudia
2003
Event-internal modifiers: Semantic underspecification and conceptual interpretation. In:
Ewald Lang, Claudia Maienbom and Cathrine Fabricius-Hansen (eds.), Modify ing Adjuncts. An Introduction, 475-509. Berlin: Mouton de Gruyter.
PA385
20. Composition
Marchand, Hans
1967
Expansion, transposition, and derivation. In: Dieter Kastovsky (ed.), Studies in Syntax
and Word-Formation . Selected articles by Hans Marchand, 322-337. Mi.inchen: Fink.
Marchand, Hans
1969
Th e Categories and Types of Present-Day English Word-Formation. A Synchronic-Diachronic Approach. 2nd ed. Mi.inchen: Beck.
Mayrhofer, Manfred
1965
Sanskrit- Grammatik mit sprachvergleichenden Erliiuterungen. Berl in: de Gruyter.
Meibauer, Jorg
How marg inal are phrasal compounds? Generalized insertion, expressivity, and I/Q2007
interaction. Morphology 17: 233 - 259.
Meyer, Ralf
1993
Compound Comprehension in Isolation and in Context. Ti.ibingen: Niemeyer.
Motsch, Wolfgang
Deutsche Wortbildung in Grundziigen. 2nd ed. Berlin/New York: de Gruyter.
2004
Olsen, Susan
Copulative compounds. A closer look at the interface between morphology and syntax.
2001
In: Geert Booij and Jaap van Marie (eds.), Yearbook of Morphology 2000, 279-320.
Dordrecht: Kluwer.
0 !sen, Susan
2004
The case of copulative compounds. In: Alice ter Meulen and Werner Abraham (eds.),
The Composition of Meaning. From Lexeme to Discourse, 17- 37. Amsterdam/Phi ladelphia: Benjamins.
0 lsen, Susan
20 12
Semantics of compounds. Tn: Claudia Maien born, Klaus von Heusinger and Pau I Portner
(eds .), Semantics. An International Handbook of Natural Language Meaning. Vol. 3,
2120- 2 150. Berlin/Boston: De Gruyter Mouton.
Olsen, Susan
201 4
Coordinative structures in morphology. In: Antonio Machicao y Priemer, Andreas N olda
and Athina S ioup i (eds .), Zwischen Kern und Peripherie. Untersuchungen zu Randbereichen in Sprache und Grammatik, 269- 286. Berlin/New York: de Gruyter.
Paul, Hermann
1920
Deutsche Grammatik. Vol. 5: Wortbildungslehre. Halle/S.: Niemeyer.
Plag, Tngo
2003
Rainer, Franz and Soledad Varela
1992
Compounding in Spanish. Rivista di linguistica 4( 1): 117- 142 .
Ralli, Angela
Compound markers and parametric variation. Language Typology and Universals 61:
2008
19- 38.
Sandra, Dominiek
1990
On the representation and processing of compound words: Automatic access to constituent morphemes does not occur. Th e Quarterly Journal of Experimental Psychology 42a:
529- 567.
Schmid, Hans-Jorg
English Morphology and Word-Formation. An Introduction. Berlin: Schmidt.
2011
Selkirk, Elisabeth
1982
Szymanek, Bogdan
2009
IE, Slavonic : Polish. In: Rochelle Lieber and Pavo! Stekauer (eds.), The Oxford Handbook of Compounding, 464-477. Oxford: Oxford University Press.
385
PA386
386
ten Hacken, P ius

2010
Synthetic and exocentric compounds in a parallel architecture. In: Susan Olsen (ed.),
New Impulses in Word-Formation, 233- 251. Hamburg: Buske.
W iilchli, Bernhard
2005
Co-Compounds and Natural Coordination. Oxford: Oxford University Press.
Warren, Beatrice
Semantic Patterns of Noun-Noun Compounds. Lund: Acta Universitatis Gothoburgensis.
1978
Whitney, William Dwight
1993
Sanskrit Grammar. Delhi: Motila l Banardsidass.
Williams, Edwin
On the notions ' lexically related' and ' head of a word'. Linguistic lnquily 12(2):
1981
245- 274.
Wilmanns, W ilhelm
1896
Deutsche Grammatik. Gotisch, Alt-, Mittel- und Neuhochdeutsch. Zweite Abteilung:
Wortbildung. 2nd ed. Straf3burg: Tri.ibner.
Z immer, Karl
1972
Appropriateness conditions for nominal compounds. Working Papers on Linguistic Universals 8: 3-10.
Zwanenburg, Wiecher
1992
Compounding in French. Rivista di Linguistica 4( 1. ): 221 - 240 .
Zwitserlood, Pienie
The role of semantic transparency in the processing and representation of Dutch com1994
pounds. Language and Cognitive Processes 9(3): 341-368.
Susan Olsen, Berlin (Germany)
21. Blending
l.
2.
3.
4.
5.
6.
7.
Introductio n
Characteristic properties of blends
Blending and grammar
Phonological conditioning
Semantic conditioning
Delimiting blending
References
Abstract
In this article, the most typical features of blending are first presented and the way they
shape the status of blending relative to grammar is discussed. The issue of the phonology
of blends, which plays a crucial role in the way their parts are combined, is then addressed. Many of the approaches that have been proposed are discussed and the main
PA388
388

Blending is not generally recognized as a legitimate word-formation process in Slavic.
It is dealt with in one article on Russian (Janko-Trinickaja 1975) and mentioned, together
with other units of word-formation, in Uluxanov (1996: 52 ff., 192). There are no monographs dealing directly with blends and blending is neither mentioned in the Russian
Academy Grammar (Svedova 1980) nor in a monograph on recent tendencies in Polish
word-formation (Jadacka 2005) that otherwise provides a good discussion of the development of Polish word-formation. Since blends that arise in Slavic are predominately
stylistically marked occasionalisms, blending is usually considered a type of playful
word-creation (cf., e.g., Il'jasova and Amiri 2009). Recent works document the fact that
blending is active in contemporary Russian (Xrusceva 2011).
2. Characteristic properties of blends

Let's admit the uncontroversial claim that word-building processes belonging to the
grammar of a given language L normally satisfy the properties stated in (1 ). The first
clause expresses compositionality, the second one guarantees that a given morphological
process is embodied in a fixed phonological pattern, and the third one states that the
association of sound with meaning is stable for any given morphological process.
(1)
The meaning of a morphological construct is the regular combination of the

meaning of the source lexemes together with the instruction associated w ith
the rule itself.
(ii) In a given morphological process, the phonological link between the input
unit(s) and the output unit (the construct) is held constant for each construct
which belongs to the process in question.
(iii) There is a link between the way phonological parts are combined, on the one
hand, and the way semantic contents are, on the other, and this linking is
held constant for each construct exemplifying the process.
(i)
If we agree that compounding belongs to plain morphology (Booij 2005: ch. 4), then
compounds should meet the requirements in (1). Indeed, French VN compounds satisfy
properties (i) and (iii) since they are interpreted in 90 % of cases as 'x such that x V'
N", where V' is the semantic counterpart of V, and N' that of N, e.g., garde-barriere
' level-crossing keeper' = 'x such that x keeps level-crossing'. The same is true of Germanic nominal determinative compounds, even though their construction meaning often
remains largely underspecified. Their interpretation follows their structure (Olsen 2000:
898) which, in turn, depends on the position of stress; then, in certain varieties of German, while Lebensmittelpunkt, the structure of which is 'N 1 (N2 N 3 ) ', means 'centre of
life', Lebensmittelpunkt, the structure of which is '(N 1 N 2 ) N 3 ', is interpreted as 'marker
on groceries' (Neef 2009: 394). The compounds just mentioned also satisfy property (ii)
insofar as the phonology of the compound results from the concatenation of the appropriate stem s associated with the lexemes involved in the compounding process, following
the order lexeme, , lexeme2 (lexeme 0 + 1). If we compare these compounds with French
binominal blends hippidemie, elevache, and ordinosaure given in Table 21.1, we see that
the latter do not satisfy (iii) because the linear adjustment of the segments correlated to
PA389
389
21. Blending
Tab. 21. l : Form meaning combination in blends
Lexeme l
hippie
ele<vage>
ordi<nateur>
G loss
Lexeme 2
Gloss
'hippie'
' breeding'
'computer'
<epi>demie
vache
<di>nosaure
'epidemic'
'cow'
'dinosaur'
Pho no I.
1 EB 2
1 EB 2
1 EB 2
Meaning
'epidemic of hippies'
'cow breeding'
'very old computer'
the base-lexemes does not allow us to predict the way these lexemes are semantically
combined. For instance, whereas the semantic head is lexeme2 in hippidemie, it is lexeme 1 in elevache and ordinosaure. In addition, (i) is not satisfied because the semantic
combination of the lexemic contents varies: in hippidemie the content of lexeme, is what
the epidemic is the vector of; in elevache, the content of lexeme 2 is the argument of the
verb underlying lexemei. while in ordinosaure the connotative content of lexeme2 is
predicated of the first unit. Property (ii) is not satisfied either, since the combination of
the phonological stems belonging to each of the source lexemes greatly varies (cf. section
4). The more striking feature is that the integrity of the lexical stems corresponding to
each of the source units is rarely kept intact in blending, and the part which is kept
cannot be known in advance. In most cases, the stems are shortened in a way that
depends on strictly local phonological parameters (section 4). As Table 21. l shows, the
suppressed segments, which are between angle brackets, can be part of either one of the
source lexemes or of both.
This sample of examples, however small, shows that blending does not comply with
the most basic principles of grammatical word-formation, namely (lii) and (liii).
If we take into account paradigmatic relations (in the Saussurean sense), we discover
another feature distinguishing blending from compounding that we might call the hapactic nature of blends: blends do not constitute series (see section 6.5) and this is all the
more true the more the phonological overlap (cf. section 4.1) motivating the original
blend has disappeared. For instance, on the model of elevache it is utterly awkward to
coin *elechien 'dog breeding' or *elanguille 'eel breeding'. But when the overlap is
preserved and complete, series may emerge, e.g., Ger. Fairkehr (cf. Verkehr 'traffic' ),
Fairkauf (cf. Verkauf 'selling' ), fairteilen (cf. verteilen 'to divide'). To sum up, blends
show three salient properties (in a weak sense, since all are negative properties):
a) No preservation of lexical integrity. In contrast to derivational morphology and compounding, the integrity of the stems corresponding to each of the source units is
rarely maintained in blending and, moreover, the manner in which they are altered
does not follow a predetermined pattern. Hence c lause (1 ii). In most cases, they are
shortened in a way that depends on the interaction of one stem upon the other (sections 4.1, 6.3).
b) No fixed pattern of compositionality. Insofar as there is no specific instruction, however coarse, attached to the process of blending (cf. (1i)), the semantic combination
of the source units' content is left floating. The absence of clause (1 iii) strengthens
this effect. It would be inaccurate however to say that blends are not compositional,
insofar as their meaning results from the combination of the meaning of their parts.
But this combination is constrained only by general conceptual categories. Blends
show as much compositionality as the most underspecified compounds do, namely
PA390
390

attributive (determinative) N-N compounds, e.g., Ger. (attributive) Fischfrau may
have half a dozen meanings (Neef 2009: 395) except the 'fish maid' meaning, whereas (coordinate compound) Kinder-Soldat means 'child soldier' .
c) Blends are type hapaxes. Unlike derived or compound units, they cannot form series.
Each one is a (lexeme) type which is the only one to instantiate the morphological
pattern it belongs to.
3.
Blending and grammar
3.1. Clarifying t he issue

The idea that the way (parts of) lexemes are combined in blending does not follow any
explicit grammatical rule indicating how the combination has to be carried out, in contrast to what happens in compounding, has been defended by Dressler (2000) and Bauer
(1988: 39) among others. Other linguists, however, hold the view according to which
blending is rule-governed and should be considered as a completely grammatical phenomenon on a par with other word-building morphological processes (Kubozono 1990;
Bat-El 1996; Plag 2003). These opposing views raise the question of the link between
blending and grammar. Actually, the debate involves two separate issues corresponding
to the conceptual oppositions regular vs. irregular on the one hand and grammatical vs.
extragrammatical on the other. The key point is to determine what the criteria could be
that allow one to say whether a given process pertains to grammar or not.
Advocates of the view according to which "blending is part of derivational morphology" (Bat-El 1996: 316) emphasize that the phonology of blends is regular and obeys the
general prosodic constraints that apply throughout the lexicon. Plag recalls that truncations (cf. section 4.1) that take place in blends are highly systematic and follow the same
patterns as those observed in clipping (cf. definition section 6.3), which shows that
blending is "part of the morphological competence of the speakers" (Plag 2003: 177).
Plag (2003: 123) claims that rule (2) may account for the most frequent types of blends
in English, e.g., guess x estimate ----+ guestimate (B or C may be null).
(2)
A B+ CD
----+
A D
Actually, the fact that the phonology of blends is not completely special is not disputed
by anybody. Dressler even says that "universal preferences" are expected to apply more
consistently to extragrammatical phenomena than to the morphological rules encapsulated
within grammar (Dressler 2000: 6). Saying that blending pertains to extragrammatical morphology does not mean that it lies out of the realm of linguistic mechanisms, especially the
phonological ones (Fradin, Montermini and Plenat 2009), or that it is governed by a "distorted version of them" (Pifieros 2004: 209). Assuming that a given process is extragrammatical only entails that it includes properties which do not match the grammar of the language in question, or that it lacks properties regularly associated with processes of similar
type (in the language in question). (Extra)grammaticality should be measured against the
grammar of a particular language, not in general. From this perspective, besides (i) regular
vs. irregular, the other relevant dimension to take into account is (ii) universal vs. specific.
The grammar of a particular language may include phenomena which are:
PA392
392

In Dutch NA compounds, the N provides a point of comparison with respect to the
property expressed by the adjective, e.g., boter-zacht 'as soft as butter; lit. butter-soft;
lit. butter-soft', ijzer-sterk 'as strong as iron; lit. iron-strong' (Booij 2002: 155-156). But
the N can function as a mere intensifier, in which case it bears the main stress of the
word, e.g., boterzacht 'very soft; lit. butter-soft', etc. Meaning specializations of this sort
presuppose the emergence of a compounding subpattern which is the basis out of which
the variation grows. No such patterns can emerge with blends because each blend is by
nature a hapax.
3.3. Consequences of a major gap

There is one property constitutive of blending which strongly supports the extragrammatical view, namely the non-satisfaction of (1 iii). To fully acknowledge this poi nt, it is
worth recording that, if we leave aside the issue of shortening, the phonology of blends
is rule-governed (by very general rules or constraints) and their semantics is underspecified (as it is for some compounds) but not at all specific (section 5). The crucial point
is that no sound/meaning correlation is stated contrarily to what is required in morphology (Zwicky 1992). The first consequence is the fact that the burden of combining the
(parts of) source units hinges on phonology. The second is the fact that the well-formedness of blends qua signs is referred to semiotic principles, inasmuch as no morphological
rules state the correlation between the phonological and the semantic plane (cf. section
3.4). This has an immediate impact on category and (gender) assignment. This assignment is stable and well-defined for any word-building process. For instance, in Romance
VN compounds are always masculine Ns, while N-N compounds inherit their category
and gender from the first noun, e.g., Fr. camion-citerne ' tanker-truck' is a masculine
noun (as camion) but emission-phare 'flagship program; lit. program lighthouse' is a
feminine noun (as emission). In blends, this assignment is not fixed. When it involves
two nouns, 1exeme2 is the morphological head, e.g., hippidemie, ordinosaure but sometimes there is a conflict between the semantic and the morphological head: suiciderurgie
( +-- suicide 'suicide' x siderurgie 'steel metallurgy') is a femin ine noun as siderurgie,
yet it means ' suicide of steel metallurgy' . Lexeme 1 is the semantic head, as in normal
compounds, e.g., chou-fleur 'cauliflower; lit. cabbage flower', but in hippidemie, it is
lexeme2 .
The position assumed here entails that blending should not be considered a wordformation process, contrary to what is claimed in the majority of studies devoted to
blending (Pharies 1987; Kubozono 1990; Gries 2004a: 639; Renner 2006). What is true,
nevertheless, is that blends quite often enter the lexicon of languages as the result of
lexicalization. But this is not special to blends, since almost any kind of expression, be it
constructed by morphology or syntax, may eventually be lexicalized (Hohenhaus 2005).
Lexicalization is orthogonal to word-formation and this is why erratic formations can
enter the lexicon and become well-behaved lexemes, e.g., Fr. chandail '(thick) jumper'
by aphaeresis of marchand d' ail 'garlic retailer ' . If blending is no more a word-formation
process, it becomes meaningless to speak of the productivity of blending, let alone of
blends, because there is no way to measure it. Following Baayen and his colleagues,
many people agree that metric P, based on hapax legomena, gives a fanrly good approxi -
PA394
394

by some instance, and (ii) that the switching point, i.e. the point where the first constituent of the blend ends and the second begins, has also to be determined by a mechanism
sensitive to fine-grained phonological distinctions at the level of syllable or below. Phonologically oriented approaches assume that both tasks (i) and (ii) are in the realm of
phonology (Bat-El 1996: 288; Kubozono 1990). This assumption is probably founded
for what regards (ii), but it is much more controversial for what regards (i). Actually,
even though the decision about (i) is based on information concerning the phonology of
W 1 and W2 (number of syllables, etc.), the conditions allowing one to choose the best
order are not strictly phonological (cf. section 4.3). In order to have a better grasp of the
mechanisms carrying out tasks (i) and (ii), we must have an idea of the variety of blends
that may exist. To achieve this goal, we shall classify blends according to their use of
the three operations mentioned above, A) truncation, B) linearization and C) overlap. If
we assume that the commonest blends involve two source units only, the combination
of these parameters results in Table 2 1.2, which corresponds to Gries's classification of
blends (Gries 2004a: 646; see also Pharies 1987: 284). Actually this classification is a
two-lexeme combination classification. It is helpful precisely because it allows us to
delimit what place blending occupies among the morphological phenomena involving
two (or more) lexemes. The notation follows Bat-El's (1996): non-occurring material is
enclosed in angle brackets, similar segments are underlined, and the bullet indicates a
lexeme boundary.
Tab. 21.2: The classification of blends and two-lexeme units (cf. Gries 2004: 646)
A. Trunc. = both
B. Trunc. = I
+ov
+UN
daxp6r
daxaf x laxp6r
knauros
knastos x auras
+ov
- UN
dialugisch
dialogisch x Li.ige
carnibbleous
carniverous x nibble
- ov
+LIN
brunch
breakfast x lunch
klafuda
klara x Juda
- ov
- UN
C. Trunc. = 2
wildschOn
wild x bildschon
smothercate
smother x suffocate
-
D. Trunc. = 0
Paradiesel
Paradies x Diesel
hypocritiquement
hypocritement x
critique
salkal
sal x kill
rajolivissant
ravissant x Joli
a) Truncation (trunc) can be distributed over the constituents of the blend or limited
to just one. Truncation means here that a segment of a source lexeme is without a
correlate in the blend. For instance, in A 1 Table 21.2 the segments /af/ and /lax/,
present in daxaf and laxp6r respectively, have disappeared from daxp6r. But dies
/di :z/ in D 1 has not since this segment occurs in the blend. What we have instead
here is a case of overlap (cf. below).
b) Linearization of constituents. The constituents of a blend are linearly ordered (+lin)
when no part of a blend correlated to the first source lexeme needs to be processed
after parts belonging to the second source lexeme have been processed (where a part
includes at least one onset). This usually corresponds to configuration PARTl < ... ... >
PART2. Otherwise, the order is non-linear (-lin). Parameters involved in ordering can
be tied either to the source units, or to systemic constraints on blending, which impose
PA395
21. Blending
restrictions on the output or take advantage of the phonotactics of the language (section 4.3).
c) Overlap of constituents (ov). Two phonologically similar segments overlap when
both appear in the source lexemes whereas only one remains in the blend. The overlap
can be local or global (section 4.4). Although overlap includes contiguous segments
(+con) in most languages, this is not always so (section 4.5.2). Overlap is the same
as haplology.
Al)
A2)
A3)
A4)
Bl)
B2)
B3)
B4)
Cl)
C2)
C3)
C4)
Dl)
D2)
D3)
D4)
Heb. daxp6r 'bulldozer'+--- da~af 'to push' x lagJ6r 'to dig', da<~af la>~6r
Ger. dialiigisch +--- dialogisch ' dialogal' x Liige ' lie', dia<lo>Lii<ge> gisch
E. brunch +---breakfast x lunch, br<eakfast-l>unch. Agitprop +---agitation x propaganda, agit<tation>prop<aganda> would illustrate this case too.
no attested example
Gr. knauros 'ripe but green' +--- knastos 'ripe' x auros 'unripe' , kn<~stos>~uros
E. carnibbleous +--- carniverous x (to) nibble, car<niver>nibbleous (Pound 1914:
45)
Gr. klafuda ' branch with a tuft ' +--- klara 'branch' x Juda ' tuft', kla<ra>fUda
no attested example
Ger. wildschon +---wild 'wild ' x bildschon ' beautiful', wild<bild>schon
No attested example (most intercalated blends show no truncation cf. 4.4)
E. smothercate +--- smother x suffocate, smother<suffo>cate
no attested example
Ger. Paradiesel +--- Paradies 'paradise' x Diesel 'diesel', Para<dies>diesel
Fr. hypocritiquement (Rabelais) +--- hypocritement ' hypocritically' x critique 'critical', hypo<crite>critiquement (*critiquement does not exist by itself).
Heb. salkal 'baby car seat' +--- sat ' basket' x kal 'easy, light', salkal
Fr. rajolivissant +--- ravissant 'charming' x Joli 'pretty', rajolivissant
Not all cells of Table 21.2 correspond to widespread patterns, neither cross-linguistically
nor language internally. The prevalent ones are those that combine overlap and contiguity
(line 1), most of which including a truncation (Al = rule (2), Bl , Cl). The patterns in
line 2 illustrate "intercalative blends", which are also quite common unlike patterns
without overlap (lines 3 and 4 ). These blends strongly prefer overlapping (section 4.5 .1 ),
hence their absence in C2. Actually, cell D3 may include expressions whose pattern is
outside of the blending system, namely compounds, e.g., E. bow-tie; salkal is not a
compound in Hebrew (Bat-El 1996: 287). This is also the case of patterns in cell A4
(section 4.4). Cell D4 includes untypical blends, which "are not very witty" precisely
because their components have not been shortened but "just stuck together" (Gries 2006:
538- 539). These are hardly blends, which is confirmed by the fact that they are sparse
and even non-existent in most languages with blends (more examples in Fradin 2000).
Table 21.2 provides us with a systematic classification of blends but gives no information
about the actual possibilities or impossibilities we observe in languages. These possibilities depend on fine-grained phonological constraints and structural (segmental and prosodi c) preferences (section 4.4). Moreover, some patterns allow variations on overlap or
truncation which seem to be language dependent (section 4.5.2).
The classification of blends in Table 21.2 does not preclude one from di stinguishing
among blend classes based on other properties, such as the interlocutor's capacity to
understand the blend. The most widely accepted di stinction among linguists is the one
395
PA396
396

between telescope (Piiieros, Ronneberger-Sibold)/overlap (Kemmer) blends on the one
hand, e.g., A 1, A3, Bl, B3, C3, and substitute (Kemmer)/portemanteau (Piiieros)/intersective (Ronneberger-Sibold) blends on the other, e.g., B2, Cl , Dl , D 2 (section 3.4.2).
This classification is based on semiotic instead of formal properties. In telescope/overlap
blends (= A blends), the source lexemes are juxtaposed like in a phrase or compound,
in intersective/substitute/portemanteau blends (= B blends), one source lexeme includes
(part of) the other as part of its sound chain.
4.2. General constraints on source units

The issue of the nature of sources has seldom been addressed, although it should be once
one assumes the distinction between word-form (the inflected form) and lexeme (the abstract unit) (Matthews 1991 ). Some linguists claim that the bases of blends are "full surface
forms" (Bat-El 1996: 290) and many speak of "source words" (SW), but it is not certain
that they are refetTing to "word-form'', since the above distinction is not operative in their
works. Even though attested examples support the idea that the sources of blends are wordforms, e.g., ]t. Alitalia (company name)~ ali wing:PL x Italia 'Italy', it is difficult to elaborate an undisputed argument for this view, because the number (or gender) feature can be
an inherent inflectional feature and as such may appear in derivation, e.g., Nld. helden-dom
hero:PL-NZR 'heroism' (Booij 2002: 84). What is sure, is that the sources of blends are fullfledged forms generated by morphology and endowed with prosodic structure. For lack of
evidence to the contrary, I will continue to assume that blends are formed with lexemes
even though, for reasons of convenience, I will make use of word when discussing the
views of authors who use this term.
Kubozono contends that the source lexemes must be of the same category (Kubozono
1990: 3). This is certainly not a mandatory requirement. There are many cases where
the constituents are N and A, e.g., Ger. damondiin ~ Damon 'demon' x mondan 'chic',
or N and V, e.g., Fr. giraffoler ~ girafe 'giraffe' x raffoler 'to be crazy about'. Crosslinguistic differences are expected to exist for thi s parameter.
The fact that blending does not confine itself to strict phonology but, in many cases,
involves the graphemic dimension has been noticed by several authors (Gresillon 1984;
Cannon 2000; Gries 2004a; cf. also article 130 on word-formation and visuality). For
instance, Cannon (2000: 954) notes that in smog the shared 'o' is pronounced differently
in smoke and Jog, which implies that the overlap is graphemically licenced by the spelling conventions of the language. Blends uniquely motivated by graphemic aspects are
those where the pronunciation of the blend is identical to the pronunciation of the longer
source lexeme, e.g., Fr. Jessetival ~Jesse ' buttock' x festival 'festival', E. jewbilee ~
Jew x jubilee. They are sometimes called "inclusive blending", e.g., Ger. alternatief ~
alternativ 'alternative' x fief 'deep' (Ronneberger-Sibold 2006). Although they are quite
numerous, these blends will be left aside in this article since I consider the phonemic
dimension as the fundamental one.
4.3. Ordering issues

Several authors have suggested that the order of occurrence of the source lexemes within
a blend is correlated to their length. Kubozono observes that in 86 % of the error blends
PA397
21. Blending
in English and 80 % of blends in hi s database " the righthand source word and the resultant blend form consist, in most instances, of the same number of syllables" (Kubozono
1990: 12). Even though many counter-examples can be found, e.g., Ger. Denkmaler ~
Denkmal 'monument' x Maler 'painter', E. ambisextrous ~ambidextrous x sex, Kubozono's observation indicates that the shorter source lexeme tends to come first, a view
that is explicitely advocated by Kelly on a basis of a statistical study of 320 two-word
English blends (where the source units co-occur sequentially as if they were coordinated,
e.g., scandalous x ridiculous ~ scandiculous). He demon strates that length effect may
combine with frequency and even prototypicality effects, whereby the source lexeme
which denotes the most prototypical entity in a given domain appears in the first position,
e .g., spoon, fork ~ spork vs. *foon (Kelly 1998). In the same vein, Bat-El argues that
segmental maximization determines the order of the base lexemes in exocentric blends
(i.e. without a semantic head), since it leads speakers to adopt "the number of syllables
from the longer rather than the shorter base word" (Bat-El 2006: 67- 68). When the
number of syllables is identical, segments are substituted for syllables. The maximization
of the number of segments makes a lexeme with a complex onset occur first and a
lexeme w ith a complex coda second. Hence the contrast: glaze ~glare x gaze vs. *gare
~ gaze x glare. Pifi.eros argues contra Kelly that, in Spanish, phonological resemblance
between source words is what determines their order in the blend (Pifi.eros 2004: 226).
The order of elements in blends can also be determined by systemic constraints following from the lexical integrity constraint or from higher level constraints. The Linearity Constraint, which guarantees that the respective order of the subparts of the source
lexemes is preserved in the blend, is of the first type. It aptly rules the order of subparts
in blends. For instance, since /by/ preceeds /sefal/ in Bucephale and /lobil/ in BuffaloBill, the only possible blend will be Bucephalo-Bill. Phonotactic constraints belong to
the second type and any blend must comply with them, e.g., E. bang, smash ~ bash vs.
*smang. Since English does not al low the sequence CVC where the two Care nasal (BatEl 2006: 68), the only licit order is bang x smash. A very general constraint prohibits a
process from having an output phonologically similar to its input (Bauer 1983). It has
been restated by Bat-El as the Uniqueness Constraint (Bat-El 1996: 288), e.g., Heb.
mesuxzar 'reconstructed' x mesupac 'renovated' ~ mesuxpac vs. *mesupac , which excludes the reverse order (other examples in Koutita-Kaimaki and Fliatouras 2001: 128).
A similar constraint is in force in derivation too, e.g., Fr. papier ' paper' (Agent nominal
deri vation) ~ papier-ier paper-AGT ' paper-maker' (haplology) ~ *papier ' paper-maker '
(the actual form is papet-ier).
4.4. Maximization
The semiotic challenge any blend has to face is to allow the listener to recover the source
units while having no access to their integral phonological form. Among the means
available to the speakers to help them reach this goal, and thereby to further maximization, there are length limitations, segment sharing, similarity in prosodic structure and
phonemic/graphemic clues associated with lexical items (the situation is slightly different
for blends which are trademarks). Many authors have claimed that limitations are imposed on the length of blends. They must not have more syllables than the longer of
two source e lements (Cannon 2000), which "facilitates the semantic recoverability of
397
PA398
398

the base word" (Bat-El 2006: 67- 68). Piiieros notes that Spanish blends tend to be
faithful to the source lexeme that functions as the head with regards to prosodic structure
(Pifieros 2004: 219). Exceptions to the above-mentioned proposals are numerous however (cf. B2, D2).
Overlap at the edges is the more common segment sharing phenomenon in blending.
Overlap clearly weakens the effect of shortening since it allows one to shorten the outcome without truncating the source lexemes, which makes them readily identifiable, e.g.,
Ger. Fantasiegel +-- Fantasie 'imagination' x Siegel 'seal'. For many authors (Gresillon,
Plag, Pifieros), overlap is a crucial feature of blending and for some a typical blend must
have an element that has correspondents in both source lexemes (Bat-El, Cannon). The
fact that overlapping acts as a driving force in blending can be deduced from several
facts: (i) there are more patterns with overlap than without, (ii) these patterns exhibit
many examples, (iii) overlap furthers the creation of intercalated blends allowing the
insertion of a segment of which it is a subpart, e.g., Fr. autoimmobiliste +-- automobiliste
'driver' x immobile 'immobile' autoimmobile< mobil >iste, (iv) there are no intercalated blends w ithout overlap (cf. A2, B2; D4 is highly marked), (v) examples without
overlap are less numerous and more contrived (in a given language) (cf. also Piii.eros
2004: 234- 235). However, overlap depends on the language. From a comparative study,
Bertinetto concludes that it is more frequent in German and French than in English or
Italian (Bertinetto 200 l ). The overlap need not be completely faithful (Pharies 's "imperfect overlapping"). Segments may not be identical in all their features, e.g., Sp. candafe
+-- canda ' lousy' x Santafe 'a Bogota soccer team', canda<santa>fe [d i- t] (Pifieros
2004: 219), E. smog. Even metathesis is licensed in some languages (but not in others
as Hebrew), e.g., Fr. caveaubulaire +-- caveau 'sepulcher' x vocabulaire 'vocabulary',
caveau<voca >bulaire [kavo i- voka] (more examples in Fradin 2000) .. These data show
the importance of taking into account the whole phonological/prosodic structure instead
of just the segments. According to Kemmer "speakers are operating with a facility for
global pattern-matching that allows similarities on many different dimensions to count,
as long as there are many of them" (Kemmer 2003: 77), and the more familiar the
patterns, the easier the recovering of the source lexemes is. The necessity of adopting a
global viewpoint to deal with the communicative transparency of blends also motivates
the introduction of "contour blends" which are characterized by their "overall rhymical
contour defined by its number of syllables and the place of its main stress" (RonnebergerSibold 2006: 170-173). Contour blends are reminiscent of an earlier proposal stating
that blends must retain the same stress that occurs in one of the source words (Cannon
2000).
The idea according to which the match between a blend and its source lexemes has
to be globally and not locally evaluated has been taken up again by Gries in a more
cognitive perspective (Gries 2004a). Starting from the claim made by Kaunisto that the
shorter lexeme in a blend contributes a larger percentage of itself to the blend than the
longer lexeme in order to preserve its recognizability (Kaunisto 2000), Gries proposes
to calculate how much each source word (SW) contributes to the blend. The conhibution
is based on the respective number of phonemes or graphemes each SW share with the
blend. He argues that the graphemes (phonemes) taken into account should not be limited
to the switching point area, as most authors do, but extended to elements similar in the
whole SW. According to this method, the contribution of both source lexemes is very
similar in chunnel, whereas li miting the similarity around the switching point (noted by
PA403
403
21. Blending
a) additive reading: the blend denotes "pairs, each cons1stmg of the parts A and B"
(Walchli 2005: 13 7 and article 40). More generally, the meaning of the blend is the
union of the meanings of its source lexemes (Bauer 2008). But for categories other
than Ns, this union may be controversial. Here are the glosses of the newly mentioned
blends: Gr. psir6s +-- psil6s 'slender' x xodr6s 'fat'; Heb. sezifarsek 'nectarine' +ezik 'plum' x ?afarsek 'peach'.
b) synonym reading: the parts A, B and the whole blend have almost the same meaning
(Walchli 2005: 143 and article 40), e.g., Gr.frimos 'more than terror' +-frici 'horror'
x tr6mos 'terror'.
The second pattern (II) corresponds to intersective meaning. Such interpretation occurs
in what Walchli (2005) calls appositional compounds, e.g., Ger. beruhmt-beruchtigt 'famous and ill-famed'. The semantics of many blends follows this pattern.
Tab. 21.5: Pattern (II). Intersective interpretation
Lg
A, N
E.
Fr.
Sp.
Gr.
Blend
P'(x) /\ Q'(x)
Jantabulous
autoimmobiliste
brujeres
saxlaburizo
fantastic'(x) /\ fabulous' (x)

car_driver'(x) /\ immobile'(x)
witch'(x) /\ women'(x)
to talk_nonsense'(x) /\ to makejokes'(x)
Glosses: Sp. brujeres +--- bruja 'witch' x mujeres 'women'; Gr. saxlaburizo 'to make nonsense
jokes' +--- saxlaro 'to talk nonsense' x kalaburizo 'to make jokes' (Koutita-Kaimaki and Fliatouras
2001: 123-124; cf. also Ralli 2009).
Pattern (III) includes blends where one of the source lexemes provides the second lexeme
with a predicate of which this second lexeme is an argument. The predicate can be the
source lexeme itself or a verb inferable from the source lexeme. Variable e is Davidson's
(1967) eventuality variable.
Tab. 21.6: Pattern (Ill). Argumental interpretation
Lg
N, N
N, V
Ger.
E.
Fr.
Blend
P'(x1> .. .,
Witzenschaft
sneakret
elevache
Glosses: Ger. Witzenschaft (E . Jandl)
Xn,
e) /\ Q'(x;)
science'(x) /\ object_of'(x, y) /\ joke'(y)

to sneak'(x, y, e) /\ secret'(y)
to breed '(x, y, e) /\ cow'(y)
+---
Witz 'joke' x Wissenschaft 'science' .
The fourth pattern (see Tab. 21.7) involves a causal relation between an event, of which
one of the source lexemes is an argument, and a causer, which may be an agent or
another event. The relation can appropriately be recast in terms of a result relation, but
for simplicity reasons I refrain from modifying the proposed formu lation.
A fifth minor pattern (see Tab. 2 1.8) involving the equative relation 'isa' could be
added to the preceding relations. This relation makes explicit the class of the referent of
the second source lexeme, which hence must be a proper noun. It has a strong connotative flavour since it introduces a side predication, actually an implicature (Potts 2005),
about the referent denoted by the blend. Using Chirouette the speaker both (i) refers to
Chirac and (ii) says that he is somebody who always changes his mind (metaphorical
PA404
404

Tab. 21.7: Pattern (TV). Causal interpretatio n
Lg
Fr.
Heb.
CAUSE(x, e) /\ P'(yi, ... , Yn, e) /\ Q'(x)
Blend
s' etrangueuler
ricpaz
CAUSE(e1> e 2)
CAUSE(x, e 1)
/\
/\
to have_a_row'(x, y, e 1) /\ to choke'(x, e 2 )
to shine'(y, e 1) /\ floor '(y)
Glosses: Fr. s' etrangueuler +--- s' engueuler 'to have a row' x s' etrangler 'to choke'; Heb. ricpaz
'floor cleaning detergent' +--- ricpa 'floor' x paz 'shining'.
sense attached to gimuette 'weathercock'). Such blends illustrate the conceptual integration exhibited by conceptual blends (according to cognitive linguistics), where the
blend 's denotatum belongs to a mental space sharing properties of the two input spaces
containing the source referents (Fauconnier and Turner 2002; Kemmer 2003).
Tab. 21.8: Pattern (V). The equative interpretation
Lg
N, N
Fr.
Blend
Chirouette
isa'(x, y) /\ P'(x) /\ Q '(y)

isa'(x, y) /\ Ch irac'(x) /\ weathercock'(y)
Two final remarks. First, as evidenced long ago by Lehrer ( 1996), blends are more easily
identified in context, and when the context proves to be indispensable, it means that the
blend is too opaque to stand by itself (cf. magalog (magazine, catalog) quoted in BrdarSzab6 and Brdar 2008: 175). Second, blending permits interpretive combinations not
allowed by grammatical morphology. Fr. elevache (elevage, vache) is a case in point
since N-N compounds such as Ger. Schijjbau 'shipbuilding', where one of the Ns is a
deverbal of which the semantic content of the other N is an argument, are generally
not well-fonned in Romance languages, e.g., Fr. *traitement-donnees 'data processing'
(although the model trasporto latte 'milk transportation' is speading in Italian).
6.
Delimiting blending
6.1. Blending and speech errors

Gries has shown that speech-error blends (also called interference blends, or contaminations), e.g., E. stragedy +---strategy x tragedy differ significantly from intentional blends
in a way predicted by Kaunisto's hypothesis (section 4.4; Gries 2004a, b). This reason
is sufficient to treat them separately (cf. also Pharies 1987).
6.2. Neocla ssica l compounds

Blends differ from neoclassical compounds, e.g., anthropomorphous in several properties. In the majority of cases the sources for the former are lexemes belonging to the
native language, while the parts stem from the classical languages in the case of neoclassical compounds; the former have no interfix while the latter generally have one, e.g.,
PA407
21. Blending
407
IRAN. Hence /rangate = 'scandal connected with Iran involving h igh-ranking people'.
Secreted affixation is therefore markedly distinct from blending, since neither operation
(ii) nor (iii) is involved in blending. It is also distinct from shortening, insofar as both
only share the abbreviation operation (i). Secreted affixation, like all derivational processes, crucially depends on series of forms showing identical correlations, such as (6),
since a bstraction presupposes repetition. Jespersen is the first one who pointed out the
phenomenon of secretion and gave it a name (Jespersen 1922: XIX 13).
6.5. Concealed compounding

Repetition obviously does not always result in a secreted affixation rule. In (7), substrings
-ware, -tique, san- are not secreted affixes but subparts of full-fledged lexemes, respectively
SOFTWARE, JNFORMATIQUE 'computer science', SANITARNYJ 'sanitary' (Mel'cuk 1997: 93).
(7)
a. fontware, freeware, groupware, shareware, vapourware ...

b. bureautique, domotique, educatique, ludotique, monnetique ...
c. sanvrac, sanpunkt, sanobrabotka, sanprosvescenie, santexnika ...
English
French
Russian
Actually, thi s is the situation we had with telecast, where tele- abbreviates television,
e.g., telefilm, telegenic, teleplay, televisual. Contrary to what is often claimed, the constructs in (7) cannot be blends because they constitute series and blends are antithetic to
series. The only property they share w ith b lending is (i) i.e. shortening. They do not
involve secretion proper nor, by the same token, abstraction since the meaning of the
substring is equivalent to that of the whole source lexeme. This is confirmed by the fact
that their interpretation is the combination of the meaning of the two source lexemes,
on the model of what happens in compounding, e.g., E. fontware 'software for fonts';
Fr. bureautique 'computer science devoted to office'; Rus. sanvrac 'sanitary doctor', etc.
For this reason, these constructs have been dubbed concealed compounds (more on this
in Fradin 2000: 47). Some of them may stem from series based on a source lexeme
which is a reduced phrase (Mel'cuk 1996: 58) such as (5b) or even (5a), e.g., Rus.
profsojuz +---- professional'nyj sojuz professional union 'trade-union' abbreviated as prof
in profsobranie 'trade-union meeting', profrabotnik 'trade-union worker', etc . Concealed
compounding is quite common in many languages, especially in technological or administrative speech. ]n Chinese, this phenomenon seems to be widespread in ordinary language too (Arcodia and Montermini 2012), e.g., feiji fly-machine 'airplane ' ---.. jfchang
machine-field 'airport', jfcang machine-cabin 'airplane cabin', where jf stands for 'airplane'.
Abbreviations
ACT.PRSPT
AZR
INS
RFX
active present participle

adj ectivizer
instrumental
interfix
PA408
408
7. References
Algeo, John (ed.)
1991
Fifty Years among the New Words. A Dictionaty of Neologisms. Cambridge: Cambridge
Un iversity Press.
Arcadia, Giorgio Francesco and Fabio Montermini
2012
Are reduced compounds compounds? Morphological and prosodic properties of reduced
compounds. In: P ierre J.-L. Arnaud, Franc,:o is Ma niez and Vincent Renner (eds.), CrossDisciplinary Perspectives on Lexical Blending, 93- 114. Berlin/New York: Mouton de
Gruyter.
2002
Restricting suffix combinations in German and English: C losing suffixes and the monosuffix constra int. Natural Language & Linguistic Theory 20(3): 451-490.
Baayen, Harald R.
1992
Quantitative aspects of morphological productivity. In: Geert Boo ij and Jaap van Marie
(ed s.), Yearbook of Morphology 1991 , l 09-149. Dordrecht: Kluwer.
Baayen, Harald R.
1993
On frequency, transparency and productivity. In: Geert Booij and Jaap van Marie (eds.),
Yearbook of Morphology 1992, 181-208. Dordrecht: Kluwer.
Baayen, Harald R. and Rochelle Lieber
1991
Productivity and English derivation: A corpus-based study. Linguistics 29( 5): 80 1-843.
Bat-El, Outi
1996
Selecting the best of the worst: The grammar of Hebrew blends. Phonology 13: 283328.
Bat-El, Outi
2006
Blends. In: Keith Brown (ed.), Encyclopedia of Language & Linguistics. Vol. 2, 2 11d ed.,
66- 70. Oxford: Elsevier.
Bauer, Laurie
1983
Bauer, Laurie
1988
Introducing Linguistic Morphology. Edinburgh: Edinburgh University Press.
Bauer, Laurie
2008
Dvandvas. Word Structure 1(1): 1-1 8.
Bertinetto, Pier Marco
1999
Psycholinguistic evidence for syllable geometry: Italian and beyond. ln: John R. Rennison and Klaus Kiihnhammer (eds.), Phonologica 1996. Syllables!?, l-28 . The Hague:
Holland Academic Graphics.
Bertinetto, P ier Marco
Blends and syllabic structure: A four-fo ld comparison. In: Merce Lorente, Nuria A ltura,
200 1
Emil Boix, Maria-Rosa Lloret and Lluis Payrat6 (eds.), La gramatica i la semantica en
I' estudi de la variacio, 59- 11 2. Barcelona: Promociones y Publicaciones Universitarias,
S .A .
Bonami, O livier, Gilles Boye and Frarn;oise Kerleroux
2009
L'allomorphie radicale et la relation tlexion-construction . In: Bernard Fradin, Frarn;oise
Kerleroux and Marc Plenat (eds.), Apen;us de morphologie fram;aise, 103-125 . SaintDe nis: Presses Universitaires de Vincennes.
Booij, Geert
2002
Th e Morphology of Dutch. Oxford: Oxford University Press .
Booij, Geert
2005
The Grammar of Words . Oxford: Oxford University Press .
Brdar-Szab6, Rita and Mario Brdar
2008
On the marginality of lexica l blending. Jezikoslovije 9( 1- 2): 171 - 194.
PA410
410

Gries, Stefan Th.
2006
Cognitive determinants of subtractive word formation: A corpus-based perpective. Cognitive Linguistics 17(4): 535-558.
Hohenhaus, Peter
2005
Lexicalization and institutionalization. In: Pavo! Stekauer and Rochelle Lieber (eds.),
Hong, Sung-Hoon
2005
An optimality theoretic analysis to English blends. Korean Journal of Linguistics 30(3):
535-557.
Iljasova, Svetlana V. and Ljudmila P. Amiri
2009
Jazykovaja igra v kommunikativnom prostranstve SM! i reklamy. Moskva: F linta Nauka.
Jadacka, Hanna
2005
System slowotw6rczy polszczyzny (1945-2000) . Warszawa: Wydawnictwo naukowe
PWN.
Janko-Trinickaja, N adija
1975
Me:l.duslovnoe nalo:lenie. In: Elena A. Zemskaja (ed.) , Razvitie sovremennogo russkogo
jazyka. Slovoobrazovanie. Clenimost' slova, 254- 258. Moskva: Nauka.
Jespersen, Otto
1922
Growth and Structure of the English Language. Garden City, NY: Doubleday & Company.
Kaunisto, Mark
2000
Relations and proportions in English blend words. In: Conference Handbook of the
Fourth Conference of the International Quantitative Linguistics Association. Prague August 24- 26. Prague: Qualico.
Kelly, Michael H.
1998
To " brunch" or to "brenc h": Some aspects of blend structure . Linguistics 36(3): 579590.
Kemmer, Suzanne
2003
Schemas and lex ical blends. In: Hubert Cuyckens, Thomas Berg, Renee Dirven and
Klaus-Uwe Panther (eds.), Motivation in Language. From Case Grammar to Cognitive
Lingustics. A Festschrift for Giinter Radden, 69- 97. Amsterdam/Philadelphia: Benjam10s.
1996
Syllable and foot in French clipping. In: Bernhard Hurch a nd Richard A. Rhodes (eds.),
Natural Phonology. The State of the Art, 135- 152. Berlin /New York: Mouton de Gruyter.
Koutita-Kaimaki, Myrta and Asimakis Fliatouras
200 1 Blends in Greek dialects: A morphosemantics analysis. In: Angela Ra lli, Brian D. Joseph
and Mark Janse (eds.), Proceedings of the First International Conference of Modern
Greek Dialects and Linguistic Theory, 11 7- 130. Patras: University of Patras.
2000
Clipping and acronymy. In: Geert Booij, C hristian Lehmann and Joachim Mugdan
1, 956- 963. Berlin/New York: de Gruyter.
Kubozono, H aruo
1990
Phonological constraints on blending in English as a case for phonology-morphology
interface. In: Geert Booij and Jaap van Marie (eds.), Yearbook of Morphology 1990, 120. Dordrecht: Kluwer.
Lehrer, Adrienne
1996
Ide ntifying and interpreting blends: An experimental approach. Cognitive Linguistics
7(4): 359-390.
PA411
21. Ble nding
Leturgie, Arnaud
20 l l
Un cas d'extragrammaticalite particulier: Les amalgames lexicaux . Lingulstica 5 l: 87104.
Leturgie, Arnaud
2012
Predire la structure des amalgames lexicaux du franc,:ais. In: Frank N eveu, Valeria MuniToke, Peter Blumenthal, T homas Kingler, Pierluigi Ligas, Sophie Prevost and Sandra
Teston-Bonnard (eds.), 3e Congres Mondial de Linguistique Fram;aise - Ly on, 4-7
Jui/let 2012, 1351 - 1368. Paris: ILF/EDP Science. Available at: www.linguistiquefrancaise.org [last access 20 Jan 20 15].
Matthews, Peter Hugoe
1991
Morphology. 2nd ed. Cambridge: Cambridge University Press.
McCarthy, John
1986
OC P effects: Gemination and antigemination. Linguistic inquiry 17(2): 207-263.
McCarthy, John J. and Linda Lombardi
1991
Prosodic circumscription in Choctaw morphology. Phonology 8( 1): 37- 7 1.
Mel'cuk, Igor A.
1993
Cours de morphologie generate. Premiere partie: Le mot. Vol. I. Montreal: Presses de
l'Universite de Montreal - CNRS Editions.
Mel'cuk, Igor A.
1996
Cours de morphologie generate. Troisieme partie: moyens morphologiques. Quatrieme
partie: syntactiques morphologiques . Vol. 3. Montreal: Presses de l'Universite de Montreal - CNRS Editions.
Mel'cuk, Igor A.
1997
Cours de morphologie generate. Cinquieme partie: signes morphologiques. Vol. 4. Montreal: Presses de l'Universite de Montreal - CNRS Editions.
Montermini, Fabio
2007
Hypocoristiques et minimalite en russe. In: Elisabeth Delais-Roussarie and Laurence
Labrune (eds.), Des sons et des sens: donnees et mode/es en phonologie et en morphologie, 198-213. Paris: Lavoisier/Hermes.
Muller, Peter 0. and Cornelia Friedrich
2011
Kontamination. In: Hilke E lsen and Sascha Michel (eds.), Wortbildung im Deutschen
zwischen Sprachsystem und Sprachgebrauch. Perspektiven - Analysen - Anwendungen,
73- 107 . Stuttgart: ibidem.
Neef, Martin
2009
TE, Germanic: German . In: Rochelle Lieber and Pavo l Stekauer (eds.), The Oxford
Handbook of Compounding, 386-399. Oxford: Oxford University Press.
0 !sen, Susan
2000
Composition. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.) Morphology An International Handbook on Inflection and Word-Formation . Vol. 1, 897- 9 16.
Berlin/Ne w York: de Gruyter.
Pharies, David A.
1987
Blending in Spanish word-formation. Romanistisches Jahrbuch 38: 271 - 289.
Pierrehumbert, Janet
2012
The dynamic lexicon. In: Abigail Cohn, Mary K. Huffman and Cecile Fougeron (eds.),
Handbook of Laboratory Phonology, 173- 183. Oxford: Oxford University Press.
Pierrehumbert, Janet and Rami Nair
1995
Word games and syllable structure. Language and Speech 38(1): 77-114.
Pifieros, Carlos-Eduardo
2004
The creation of portmanteaus in the extragrammatical morphology of Spa nish. Probus
16(2): 203- 240.
Plag, Ingo
2003
411
PA412
412

Plenat, Marc
1984
Toto, Fanfa, Totor et meme Guiguitte sont des ANARS. In: Franyois Dell, Daniel Hirst
and Jean-Roger Yergnaud (eds.), Forme sonore du langage. Structure des representations en phonologie, 161-181. Paris: Hermann.
Plenat, Marc
Une approche prosodique de la morphologie du verlan. Lingua 95(1): 97- 129.
1995
Potts, Christopher
2005
The Logic of Con ventional Implicatures. Oxford: Oxford University Press.
Pound, Louise
1914
Blends, their Relation to English Word Formation. Heidelberg: Winter.
Prince, Alan and Paul Smolensky
1993
Optimality Th eory. Constraint Interaction in Generative Grammar. Report no. 2, Rutgers University Center for Cognitive Science. Cambridge, MA: MIT Press.
Ralli, Angela
2009
Modern Greek VY dvandva compounds: A linguistic innovation in the history of the
Indo-European languages. Word Structure 2(1): 48-68.
Reischer, Jiirgen
2008
Die Wortkreuzung und verwandte Verfahren der Wortbildung. Eine kotpusbasierte Analyse des Phiinomens ,, Blending" am Beispiel des Deutschen und En:glischen. Hamburg:
Kovac.
Renner, Vincent
2006
Depasser les desaccords: pour une approche prototypiste du concept d 'amalgame lexical.
In: Miryam Pereiro and Henry Daniels (eds.), Le desaccord, 137- 147. Nancy: AMAES.
Riehemann, Suzanne z.
1998
Morphology and the hierarchical lexicon. Journal of Comparative Germanic Linguistics
2: 49- 77.
2006
Lexical blends: Functionally tuning the transparency of complex words. Folia Linguistica 40(1 - 2): 155- 181.
20 l 0
Word creation: Definition - function - typology. In: Franz Rainer, Wolfgang U. Dressler,
Dieter Kastovsky and Hans Christian Luschiltzky (eds.), Variation and Change in Morphology, 20 1-216. Amsterdam/Philadelphia: Benjamins.
Svedova, Natalija Ju. (ed.)
1980
Russkaja grammatika. Moskva: Nauka.
Thornton, Anna M.
1993
Italian blends. In: Livio Tonella and Wolfgang U. Dress ler (eds.), Natural Morphology:
Perspectives for the Nineties, 143- 155. Padova: U nipress.
Thornton, Anna M.
1996
On some phenomena of prosodic morphology in Italian: Accorciamenti, hypocoristics
and prosodic delimitation. Probus 8: 81 - 112.
Thornton, Anna M.
2004
Parole macedonia. In: Maria Grossmann and Franz Rainer (eds.), La formazione de/le
parole in italiano, 567- 571. Tilbingen: Niemeyer.
Tournier, Jean
1985
Introduction descriptive a la lexicogenetique de /' anglais contemporain . Paris/Geneve:
Champion - Slatkine.
Uluxanov, Igo r' S.
1996
Edinicy slovoobrazovatel'noj sistemy russkogo jazyka. Moskva: Astra sem'.
van Marie, Jaap
1985
On the Paradigmatic Dimension of Morphological Creativity. Dordrecht: Foris.
PA413
22. Incorporation
413
Villoing, Florence
2012
Contraintes de taille dans Jes mots composes: quand la phonologie entre en concurrence
avec Jes contraintes morphologiques. In: Frank Neveu , Valeria Muni-Toke, Peter Blumenthal, Thomas Kingler, P ierluig i Ligas, Sophie Prevost and Sandra Teston-Bennard
(eds.), 3e Congres Mandia! de Linguistique Franr;aise - Lyon, 4- 7 juillet 2012, 14251440. Paris: ILF/EDP Science. Available at: www.linguistiquefrancaise.org [last access
20 Jan 20 15].
Walchli, Bernhard
2005
Co-compounds and Natural Coordination . Oxford: Oxford University Press.
Warren, Beatrice
1990
The importance of combining forms. In: Wolfgang U . Dressler, Hans Christian Luschutzky, Oskar E. Pfeiffer and John R. Rennison (eds.), Contemporary Morphology,
111 - 132. Berlin/New York: Mouton de Gruyter.
Xrusceva, Oksana A.
2011
Universal' nye i lingvokul'turnye osobennosti blendinga. Ph.D. dissertation, Celjabinskij
gosudarstvennyj universitet, Ce ljabinsk.
Zwicky, Arnold M.
1992
Some choices in the theory of morphology. In: Robert Levine (ed.), Formal Grammar.
Theory and implementation, 327-371. Oxford: Oxford University Press.
Bernard Fradin, Paris (France)
22. Incorporation
1.
2.
3.
4.
5.
6.
Introduction
Noun incorporation
Incorporation as a syntactic process
Applications and extensions
Conclusion
References
Abstract
This article overviews two distinct senses of the term incorporation. The first refers to a
word-formation process akin to compounding, and is typically applied to the particular
noun-verb compounding process known as noun incorporation. The second refers to a
syntactic approach to word-formation innovated by Mark C. Baker, head-movement,
which was proposed to account for noun incorporation and other verb derivations such
as causatives and applicatives. Competing theoretical approaches to noun incorporation
are discussed and evaluated, and then the application of Baker's incorporation approach
to other derivational contexts, especially deadjectival and denominal verbs, is discussed
and assessed.
PA414
414
1. Introduct ion
Discussion of the word-formation process known as "incorporation" is potentially
fraught with difficulties at the outset because this term is used for different (not necessarily commensurate) meanings by different authors. In morphology, the two most common usages of the term incorporation are: (i) some type of word-formation process akin
to compounding, especially in reference to noun incorporation (a special type of nounverb compounding); and (ii) a word-formation process involving the syntactic derivation
of morphologically complex words via head movement. The first meaning is a grammatical notion based on a long tradition of descriptive and typological research on various
languages from around the world, whereas the second is a relatively recent theoretical
innovation involving a syntacticocentric view of morphology which was pioneered by
Baker (1988) in order to account for the descriptive process referred to in (i).
As it turns out, the first (and perhaps most common) usage of incorporation may be,
at first blush, of less relevance to many European languages, since most of those are
of a morphological type which does not heavily utilize incorporation in the traditional
(descriptive) sense. If Baker's syntactic view of that process is on the right track, however, then other construction types (beyond N-V compounding) may make use of the
same syntactic processes (e.g., head-movement) which form the traditionally-recognized
incorporation constructions.
This article will thus first illustrate the most prototypical incorporation construction,
noun incorporation, and discuss the major implications of this construction for different
theories of morphology (section 2); here we will primarily rely on data from non-European languages. I will then discuss the second notion of incorporation, Baker's syntactic
head-movement analysis of noun incorporation (henceforth, NI) (section 3). Next, I will
illustrate the potential application of Baker's incorporation theory to the morphology of
many languages, including English and other European languages, beyond traditionally
recognized NI languages (section 4). I will then conclude in section 5.
2. Noun incorporation
"Incorporation", in principle, can be understood to involve the combination of two different lexemes of any category to create a complex word, in a manner very similar to, if
not actually identical with, compounding. However, this concatenation is often described
as a verb (somehow) taking (the head of) its complement into its own morphological
structure. Matthews (1997) thus defines one meaning of incorporation as follows:
(1)
A regular process by which lexical units which are syntactically complements of

verbs can also be realized as elements within the verb itself: e.g. schematically,
hunt-rabbit-PROG-3sG 'He is hunting rabbits, is rabbit-hunting ' .
(Matthews 1997: 173)
The prototypical instantiation of this process, which is illustrated by the quotation from
Matthews above, involves a verb incorporating a nominal. This nominal is usually taken
to be the verb's notional object argument, and when that is the case then it instantiates
PA415
22. Incorporation
415
the prototypical instance of "noun incorporation" (NI). But in some cases incorporati on
can also involve a nominal indicating the manner of the action expressed by the verb,
or the instrument used to perform the action of the verb . These are sometimes referred
to as "manner incorporation" (Harley 2005) and " instrumental affixes'', respectively. On
manner incorporation, more will be said in section 4.
With respect to instrumental affixes, it is important to note that while the etymological
sources of such affixes are often nominal, and while many of these constru ctions may
have originated from the actual morphological process of noun incorporation, in p ractice
these elements can have adverbial (rather than strictly nominal) meanings, e.g., indicating the type of motion or manner of use rather than the actual specific type of instrument
used (Mithun 1999: 119). Thornes (2003) gives the examples in (2) to show that different
instrumental prefixes in Northern Paiute (Uta-Aztecan) can be used to describe different
uses for the same actual instrument - in this case it could be a wooden spoon - which
is used in differe nt manners for creating different culinary products (e.g., soup vs. gravy,
the latter requirin g a more horizontal positioning of the instrument to make the stirring
motion):
(2)
Northern Paiute
a. tsi-kwidu?i
IP/sharp-stir
'stir (e.g., soup)'
b.
wi -gwidu?i
IP/ long-stir
'stir (e.g., gravy)'
(Thornes 2003: 357 [7])
The verb root is the same in both of the above examples, the difference being which
" instrumental prefix" gets incorporated onto the verb stem. Given the fact that these
" instrumental affix" constructions involve adverbial usages of wh at once used to be
nominal roots, and that verbs can also serve as the etymological sources for such affixes
(Thornes 2000), this construction type has more in common with manner incorporation
(see below) than N I, strictly speaking.
Returning to NI proper, it would be difficult to find another single construction type in
modern linguistics which has had both such a long history of debate and which has played
such a p ivotal role in motivating new theoretical developments. With NI serving as a paradigmatic case of a morphological construction existing at the interface of the morphology with
the syntax, different theoretical approaches have either tried to pigeonhole it into one grammatical module or the other (i.e. the morphology or "lexicon", on the one hand, vs. the syntax,
on the other), or to allow the two modules to overlap to some degree.
The locus classicus of the controversy surrounding the proper analysis of N I can be
found in a series of papers authored by the eminent Americanist linguists A. L. Kroeber
and Edward Sapir, from 1909 to 1911 , which resulted from their observation that NI
constructions can commonly be found in the indigenous languages of the Americas. The
crux of their argument is still at the heart of many debates in modern theoretical lingu istics, and it involves the fundamental question of how separate the morphological and
syntactic components of the grammar ought to be. Sapir (1911) held the view that wordformation (i.e. morphology) was separate from syntax, whereas Kroeber (1909, 1911)
wanted to allow some bleeding between the two components. As discussed in detail in
PA416
416

Haugen (2008a, 2009), this debate has been maintained into the present era under the
guise of lexicalist (or, what I have termed the "neo-Sapirean") vs. syntacticocentric (or,
what I have termed the "neo-Kroeberian") approaches to morphology.
In the more modem era, which has seen various developments in theorizing since the
"Chomskyan Revolution" of the 1950s and 1960s, the same issues have been at the
forefront of debate, and discussion of alternative analyses of NI has led to the development of a variety of different theoretical perspectives on the issue. Another prominent
disagreement hinging on the appropriate analysis of NI can be found in the mid-l 980s
back-and-forth between Marianne Mithun ( 1984, 1986) and Jerrold Sadock ( 1980, 1986).
Mithun (1984), for instance, offers a typology of cross-linguistic NI constructions
with an eye toward explaining their diachronic development in different languages. Crucially, for Mithun, morphology must be fundamentally distinct from syntax, as per her
dictum that NI is "the most nearly syntactic of all morphological processes" (Mithun
1984: 889, emphasis added). In each of the Comanche NI examples in (3), for instance,
a nominal root is morphologically attached to a verb stem to create a complex verb stem:
(3)
NI in Comanche
a. narinoo' -rikisaddle-put.away
'to saddle up'
b.
pihi-tsah-kwe'yafuzz-by.hand-remove
'to skin an animal'
c.
wana-roh-peticloth-by.force-throw
'to gamble'
d.
waa-hima
cedar.tree-take
'to celebrate Christmas'
(Mithun 1984: 855)
For Mithun, this compounding process is nearly syntactic because the complex word
relates constituents of a sentence that are normally related syntactically (i.e. a verb with
a nominal object argument), and the process can be very productive (which is another
prototypical characteristic of syntax). However, this compounding process must be regarded as morphological, for Mithun, because they are not entirely free (even if they
may be productive), and, as word-formation processes, they are quite prone to lexicalization; e.g., see examples (3c) and (3d) above, which have idiomatic meanings (and are
thus not transparently syntactic). With this clear bifurcation of grammatical modules in
mind, then, Mithun thus adopts a neo-Sapirean perspective on the topic of NI.
Sadock (1980, 1986), on the other hand, argues that NI in West Greenlandic displays
very clear syntactic properties, including the modification of incorporated nominals by
external adjectives (and other nominal modifiers). An example is shown by the contrast
between (4a) and (4b):
PA417
22. Incorporation
(4)
Incorporation in West Greenlandic

a. Sapanngamik kusanartumik
pzszvoq
bead-INST
beautiful-NOM-INST thing-get-INDIC-3sg
'He bought a beautiful bead. '
b. Kusanartumik
sapanngarsivoq
beautiful-NOM-INST bead-get-INDIC-3sg
' He bought a beautiful bead.'
(Sadock 1980: 307)
(4a) illustrates a nominal root (sapanng- ' bead') in an object position NP, whereas (4b)
shows the same nominal root as incorporated into the verb stem (the inflected form of
'to get'). Note that the nominal modifying adjective, kusanartumik 'beautiful-NOM-INST',
can appear with both construction types. Sadock claims that this and other external
modifiers form a constituent with the incorporated nominal in (4b), derived from (the
equivalent of) a deep structure like (4a). If this is the case, then perhaps there can't be
such a clear separation of morphology from syntax. For this reason, Sadock adopts more
of a neo-Kroeberian view of the process of incorporation.
One important difference between Sadock's West Greenlandic data and cases of NI
presented by Mithun, however, lies in the fact that the verbal stems in Greenlandic are
obligatorily bound elements - i.e. they are affixal in nature. Mithun (1986) maintains
that NI applies only to those situations in which the verbal element itself may be a free
morpheme without nominal support. She regards Sadock's Greenlandic case as an instance of denominal verb formation (i.e. the derivation of a verb from a nominal root),
rather than NI per se. This position cleverly leaves open the possibility that denominal
verbs may interleave morphology and syntax while still maintaining the further possibility that true NI would remain morphological. However, plenty of other languages do
allow external modification of incorporated nominals even in the case of NI with optionally free verb roots, so this tactic does not actually resolve the issue for those languages.
(We will return to the issue of denominal verbs possibly involving incorporation in
section 4 below.)
Consider a paradigmatic case of NI from Mohawk:
(5)
Incorporation in Mohawk
a. Ka-rakv ne sawatis hrao-nuhs-a'?
3N-white DET John
3M-house-SuF
'John's house is white'
b. Hrao-nuhs-rakv ne sawatis
3 M-house-white DET John
' John's house is white'
(Baker 1988: 65)
The difference between (5a) and (5b) shows clearly that predicates such as ka-rakv 'to
be white' can incorporate their internal argument (as in 5b, where the root nuhs ' house'
is part of the complex verb stem), but they do not have to (cf. 5a). Baker (1988) analyzes
such constructions as involving a movement operation wherein the underlying theme or
object moves up the tree into the verbal position, given the connection between the
incorporated nominal head's external modifier, here the possessor DP sawatis ' John ', as
417
PA418
418

well as an attractive theoretical position that theta-roles are always assigned to the same
syntactic positions (which explains the semantic "deep structure" connection between
such alternations as the active vs. the passive, where the latter is derived from the former
via transformational rules or their equivalents). (Baker 1988: 46 codifies this proposition,
i.e. that thematic relations are universally related to specific syntactic configurations at
deep structlllre, as the uniformity of theta assignment hypothesis, or UTAH).
Baker calls the syntactic movement process itself incorporation, and a large subsequent literature has followed this usage. We will consider Baker's analysis in more detail
in the next section. Before that, however, let us consider two more prominent theoretical
approaches to the phenomenon of NI: autolexical syntax and lexicalist theories.
Autolexical syntax is a theory that was proposed by Sadock ( 1985, 1991) to account
for the NI facts of West Greenlandic and other languages. For Sadock, morphology and
syntax are two completely autonomous modules of grammar which contain their own
principles and rules, and they may assign different structural descriptions to words,
phrases, and sentences. For example, in an English example with an auxiliary clitic like
John's here> the clitic = 's can be dually analyzed: i.e. in the syntax it can be analyzed
as a V heading a VP with an adverbial complement (here), whereas, simultaneously, in
the morphofogy it is attached at the word level to the N John. Sadock offers a notation
with two different tree structures attaching to the sentence, as shown in (6). Sadock's
convention for constructing these trees is to put the morphological representation on the
top, and the syntactic one on the bottom.
(6)
The autolexical syntax analysis of English aux cliticization
---------
ADV
John
's
here
John
's
here
I
I
ADV
NP
VP
--------------s
(Sadock 1985: 385)
For the case of NI, Sadock extensively discusses and analyzes cases from West Greenlandic (among other languages). A simple example is given in (7):
(7)
Hansi illo-qar-poq
Hans house-have-INFL
'Hans has a house '
(Sadock 1985: 400)
PA420
420

place in the lexical component of the grammar (see, e.g., DiSciullo and Williams 1987).
Weaker lexicalist theories may allow some word-formation processes, such as inflection, to
have access to the syntactic component (relegating other processes, such as derivation and
compounding, to the lexicon) (see, e.g., Anderson 1982).
Lexicalist theories of N I, such as that proposed by Rosen (1989), hold that NI simply
results from the process of compounding a noun and a verb in the lexicon to yield a
complex verb, which is then fit for insertion into a verbal node in the syntax. This
theoretical perspective is in line with what Sapir (1911) had proposed from a descriptive
standpoint and what Mithun ( 1984) holds from a functionalist and typological standpoint.
This type of analysis is sketched in (9):
(9)
Lexicalist analysis of NI: The Lexicon feeds the Syntax

[RootN] + [Rootv]
deer
butcher
[Root-Root]v
deer-butcher
v
[Root-Root]v
deer-butcher
The Lexicon
The Syntax
In most ver sions of this kind of theory, an important assumption is that the internal
structure of words, including compounds, should be invisible to the syntax. This is encapsulated in such notions as the " lexical integrity hypothesis", "the thesis of the atomici ty of words" (Di Sciullo and Williams 1987), and related notions. An important consequence of this stance is that, in derivations like that shown in (9), the syntax only "sees"
the result of morphological operations, in this case the fact that there is a word of
category V, deer-butcher. The further fact that in the derivational history of this word
there was once a nominal root, deer, which was compounded onto a verbal root, butcher,
via the morphological process of NI, should not play a role in how the syntax treats the
overall resulting word. A clear prediction of this approach is that syntactic processes
should not target only one member of the N I construction, i.e. neither the noun or the
verb, independently of the compound itself qua compound.
The notion of " lexical integrity" for NI constructions is counter-exemplified by data
from Hiaki (also known as Yaqui, Ute-Aztecan), which allows inflectional reduplication
to target the verbal head of the N -V compound in NI constructions, rather than the
compound word itself (as would be expected if NI is merely the process of the lexical
creation of a new verb for use by the syntax). The lexicalist analysis in (9) above predicts
that habitual reduplication in Hiaki, which is prefixal, should target the edge of the N V compound. This is not the case (cf. lOa). R ather, reduplication occurs inside the
compound, targeting the verbal head (lOb):
(10) Head-marking reduplication in H iaki
a. *Peo *ma-ma.so.peu.te I *Peo *ma.so-ma.so.peu.te
Peo RED-deer.butcher
(< maasoN+p eutev 'deerwbutcherv' )
'Peo is always butchering deer '
b.
Peo ma.so-peu-peu.te
(< maasoN+peutev ' deerwbutcherv')
Peo deer-RED-butcher
'Peo is always butchering deer '
(Haugen and Harley 2013)
PA421
22. Incorporation
While this is a serious problem for weak lexicalist theories (which by architectural design
require that inflection must occur after derivation), it is less of a problem for stronger
lexicalist theories which have all word-formation processes, including inflection, occur
in the Lexicon. Such theories could in principle allow interleaving of derivation and
inflection - but in doing so they actually overpredict such interactions, as with DiSciullo
and Williams' (1987) examples like ?parts-supplier, ?choirs-boy, etc. , which not all English speakers agree are acceptable. Such theories still face serious challenges, though,
from the fact that NI constructions are typically limited to certain thematic relations
(e.g., patient or theme) while robustly ruling out others (e.g., agents), a fact which is not
necessarily expected if NI results from simple N-V compounding; see Haugen and Harley (2013) for further discussion of challenges to lexicalist theories from head-marking
reduplication in Hiaki NI (and other compounding) constructions.
There is an alternative to lexicalist theories which holds that the NI word-formation
process occurs outside of, and prior to, syntax; namely, to actually allow the syntax itself
to participate in the word-formation process. This approach has been adopted by Baker
(1988) in his incorporation theory, and this perspective has been followed by many
others in subsequent research.
3. Incorporation as a syntactic process

We have seen that some linguists, adopting theoretical orientations which presuppose
that the domains of morphology and syntax must be kept distinct, note that NI comes
close to breaching the divide between them. Such a proposal has been offered both as a
typological generalization (e.g., Sapir) and a theoretical proposition (e.g., Mithun 1984,
from a functionalist perspective; as well as Rosen 1989, from a more formal, i.e. generativist/lexicalist, perspective). The just mentioned theoretical perspectives have adopted
this partition of morphology from syntax in various guises, either by maintaining that
morphology and syntax are indeed autonomous components which interface with one
another (e.g., Sadock's autolexical syntax), or that one component, morphology, can feed
into the other component, syntax (e.g., lexicalism).
There exists, however, a third type of theory which allows the domains of morphology
and syntax to overlap - specifically, by allowing syntactic structures and processes to
build words (i.e. morphological structures). We may refer to this kind of theory as "syntacticocentric". In the case of NI, the most prominent syntacticocentric theory has been
proposed by Baker (1988), who allows the nominal head of VP-internal (object) NPs to
move into the verbal position by syntactic head -movement. This process itself is termed
"incorporation" by Baker, and much work has adopted and followed the essence of his
analysis.
By employing syntactic head movement Baker ( 1988) accounts for a few key features
of the NI construction. First, it derives the placement of the nominal head in the verb
complex from the syntactic representation. This analysis of incorporation explains the
pattern of NI demonstrated in Mohawk, as shown in ( 11) which repeats the data from
(5) above:
421
PA423
22. Incorporation
423
In regard to the exclusion of subjects from incorporation constructions, it is important

to note here that adherents of Baker's approach adopt the unaccusative hypothesis of
Perlmutter (1978), wherein some intransitive "subjects" are actually objects in deep
structure. Such nominals are able to incorporate in some languages, e.g., Nahuatl:
(14) Unaccusative subject incorporation in Tetelcingo Nahuatl

a. toonal-kisa
sun-emerge
'the sun comes out'
b.
b'.
VP
~
s
~
VP
~
NP
kisa
toonali-kisa
emerge
NP
I
N
toonal
ti
sun
(Tuggy 1986: 457)
Such nominals are typically themes or patients, which are theta roles assigned within
VP given assumptions about theta role assignments in Baker's theory. Baker's analysis,
then, is intended to exclude VP-external subjects (as with unergative intransitives), which
are prototypically agents, which is largely in accord to the empirical facts about NI
which have been observed since at least Kroeber (1909). This account predicts that
unergative verbs should not be able to incorporate their subjects, which are presumed to
be generated external to the VP (e.g., in an external functional head, e.g., voice 0 or v 0 ;
cf. Kratzer 1996; Harley 1995, and much other work). As we will see below in our
discussion of Hale and Keyser's (1993, 2002) extension of Baker's incorporation theory
to denominal verbs, unergatives are able to incorporate their internal (i.e. theme and
patient) object arguments.
While more recent work in generative syntax has developed a more complicated
syntactic tree structure (e.g., more functional heads) than what was originally envisioned
by Baker (1988), the crucial insights of Baker's incorporation analysis have been largely
maintained in many current (generativist) theoretical approaches to morphology as well
as to syntax. External subjects (agents) are projected external to a functional projection
above the VP, typically regarded as vP or voiceP, and they therefore cannot move down
the tree into the V position.
Other extensions of Baker 's theory have also been proposed in light of innovations
in generativist syntactic theorizing since principle and parameters theory, such as the
minimalist program (Chomsky 1995) and its variety of interrelated theoretical frameworks, e.g., the bare phrase structure of Chomsky (1994) and the antisymmetry approach
of Kayne ( 1994). This latter approach makes the strong proposal that the underlying
structure of all phrases is universal (and, in the clausal context, of the form SVO), with
deviations involving various kinds of movements. Barrie (2011) adopts this position and
applies Moro's (2000) dynamic antisymmetry theory to account for the patterns of head-
PA427
22. Incorporation
d.
e.
f.
g.
427
jolas
oihu
lo
zurrunga
egm 'to play'

egm 'to shout'
egm 'to sleep'
egm 'to snore'
(Hale and Keyser 2002: 117 [22])
Hale and Keyser argue that the Basque situation is the general case, where a nominal semantically composes with a light verb to yield something along the lines of 'to do N'.While
Basque does not actually involve incorporation per se, some languages do require incorporation, including the English-style cases (which by hypothesis involve null light verb affixes) and other languages which have overt functional heads, such as derivational suffixes,
requiring incorporation. This latter kind of pattern is ubiquitous in the Uta-Aztecan language family (Haugen 2008b); some examples from Tohono O'odham are shown in (20):
(20) Tohono O 'odham unergatives = [N + -t( Vdo)]
a. ki:
'house'
ki:-t
'to build a house'
b. jufi 'cactus candy' jun-t
'to make cactus candy'
c. boa 'basket '
hoa-t 'to make a basket'
ha'a-t 'to make a pot, olla'
d. ha'a 'pot, olla'
e. si :l 'saddle'
si :l-t
'to make a saddle '
(Hale and Keyser 2002: 133 [48])
The incorporation process would be the result of language-specific parameter-setting, and
may be driven by "phonological defectivity" (Hale and Keyser 2002: 63) of the light verb
in English and other languages requiring incorporation for their unergative denominals.
Even more complex denominal verb derivations can also be explained by Hale and Keyser 's incorporation account, including location denominal verbs like shelve, the derivation
of which is shown in (21):
(21) The derivation of a location denominal verb
----------/\
~
v'
VP
/\v
/\
p
I
shelf
NP
(her books)
V'
A
IA
PP
NP
I
N
I
'----t
(adapted from Hale and Keyser 1993: 58 [8])
PA428
428

In this derivation, N incorporates into P (-on N); this complex P then incorporates into
a stative V head (- be on N); and this complex then itself incorporates into a higher
verbal head indicating causation, i.e. v (-cause to be on N). Each movement operation
obeys Travis' head movement constraint, which requires moving up the tree to the next
highest head without skipping any heads en route.
Other verbs which are claimed to derive via similar incorporation processes include
locatum verbs (which denote the object transferred to a location, e.g., bandage, butter,
etc.) and deadjectival verbs (e.g., redden, lighten, etc.).
Several criticisms have been lodged against this approach, including the issue that this
type of incorporation, in English at least, does not seem to allow "stranded" modifiers
of the type seen in other incorporation cases seen above (e.g., in West Greenlandic and
Mohawk). For example, an incorporation analysis of the structure in (21) might lead one
to expect possible sentences like *He saddleid the horse Western ti, which are not licit
in English. (See Kiparsky 1997 for th is and other criticisms of the Hale and Keyser
incorporation approach to English denominal verbs, and an alternative approach to these
constructions.) A possible response to this objection would be that "modifier-stranding"
is itself dependent on the independent existence of "null-head" modifiers in a given
language. If this is the case, though, then this weakens one of the key motivations for
positing a movement analysis for NI in the first place, since the verb could be formed
from noun-verb compounding in the Lexicon with this lexically-derived complex verb
taking a complement NP with no overt N but with modifiers (i.e. a null-head NP). (See
Rosen 1989 for related discussion and a lexicalist theory of NI exactly along these lines.)
Even if such criticisms can be surmounted, a crucial empirical issue facing the syntactic head-movement approach to denominal verb-formation, as well as to NI as traditionally envisioned, is the fact that these constructions can often co-occur w ith an overt
nominal in the complement position. This overt nominal is either cognate with (i.e. rootidentical to) the incorporated noun, or it is interpreted as a more specific instance of the
incorporated noun. When such a hyponymous object appears, it is not obvious how
the supposedly incorporated nominal creating the verb would be related to the object
position:
(22) The hyponymous object problem

b'.
a.
VP
~
v
NP
VP
~
?!
v
I
NP
ti /dnlWB
a )lgiJj
dance
?!
Baker's original analysis of NI in Mohawk did not have to deal with this issue because
Mohawk, as a polysynthetic language, does not have overt NPs in argument positions.
Some other noun-incorporating languages, such as Hopi (Hi ll 2003), though, clearly do
PA429
22. Incorporation
have object NPs that originate in these positions, as evidenced, in the Hopi case, by such
facts as relatively strict SOY word order, overt accusative case-marking, etc.
Haugen (2008a, 2009) offers a solution to this problem by adopting a crucial notion
from distributed morphology (among other approaches): Late insertion, which holds that
morphemes, containing their phonological instantiations, do not actually appear in the syntactic structure until after the syntax has performed operations such as head-movement (or,
in a slightly different technical implementation, the copying of abstract morphosyntactic
features). At spell-out, it appears to be possible (in some languages at least) to insert different roots into the head and tail of an incorporation chain. The hyponymy relation between
the two different roots is one of induced pragmatic inference, along the lines of Grice's
maxim of quantity, because hyponymy readings can be coerced in situations where two
roots otherwise do not necessari ly encode 'more specific than' meanings in relation to one
another. For example, in Hiaki one can both horse-have a dog (i.e. ' have a dog and use it
as one would canonically use a horse') or dog-have a horse (i.e. 'have a horse and use it as
one would canonically use a dog'); note that this account ofhyponymous objects would be
difficult to implement in Sadock's autolexical syntax framework, discussed above. This is
not a fact about speakers' knowledge or opinions about horses or dogs in Hiaki culture, but
rather an artifact of the coercion of meaning that occurs by putting either one root or the
other at the head or the tail of the chain of movement.
A similar analysis for denominal verb constructions and NI constructions is particularly appealing given the similar nature of these constructions. Some languages, such as
Hopi, seem to have identical properties for the two construction types, including modifier-stranding, cognate and hyponymous objects, and the introduction of nominal discourse
referents (Hill 2003; see a lso Haugen 2008a, 2009 for additional discussion). Some
languages do show some differences between the two construction types, however, including Hiaki, which tends to only have intransitivizing NI which does not allow for the
stranded modifiers or hyponymous objects which are quite acceptable in denominal verb
constructions (cf. Haugen 2008b). A further difference is that NI is not fully productive
in this language, unlike the situation for denominal verbs. See Gerdts and Marlett (2008)
for additional discussion of denominal verb constructions and a comparison of these
with NI, pertaining especially to the indigenous languages of the Americas.
Nevertheless, the close connection between denominal verbs and NI constructions
necessitates some consideration of the possibility that they may justify a unified analysis.
A further empirical fact about English denominal verbs, which drives this point home
particularly clearly, is that cognate direct objects tend to be judged as redundant when
they do not co-occur with some kind of external modification. Contrast (23) and (24):
(23) Pragmatic redundancy with cognate objects
a. Hortense is laughing/dancing.
b. ?Hortense is laughing a laugh/dancing a dance.
c. ?Hortense is laughing the laugh/dancing the dance.
(24) Less redundancy with modifiers for cognate objects
a. He laughed a false laugh that held genuine bitterness.
(Michael Chabon, Gentlemen of the Road, p. 187).
b. Hortense is dancing a happy dance.
429
PA432
432
Chomsky,
1981
Chomsky,
1994
Noam
Lectures on Government and Binding. Dordrecht: Faris.
Noam
Bare Phrase Structure. Cambridge, MA: MIT Department of L inguistics and Philosophy.
Chomsky, Noam
1995
The Minimalist Program. Cambridge, MA: MIT Press.
DiSciullo, Anna-Maria and Edwin Williams
1987
On the Definition of Word. Cambridge, MA: MIT Press.
Gerdts, Donna B. and Stephen A. Marlett
2008
Introduction: The form and function of denominal verb constructions. International
Journal of American Linguistics 74(4): 409- 22.
Hale, Kenneth L. and Samuel Jay Keyser
1993
On argument structure and the lexical expression of syntactic relations. In: Kenneth Hale
and Samuel Jay Keyser (eds.), The View from Building 20. Essays in Linguistics in
Honor of Sylvain Bromberger, 53- 109. Cambridge, MA: MIT Press .
Hale, Kenneth L. and Samuel Jay Keyser
2002
Prolegomenon to a Theory of Argument Structure. Cambridge, MA: MIT Press.
Halle, Morris and Alec Marantz
Distributed morphology and the pieces of inflection. In: Kenneth Hale and Samuel Jay
1993
Keyser (eds.), The View from Building 20. Essays in Linguistics in Hon or of Sylvain
Bromberger, 111-17 6. Cambridge, MA: MIT Press.
Harley, Heidi
Subjects, events, and licensing. Ph.D. dissertation, MIT, Cambridge, MA.
1995
Harley, Heidi
How do verbs get their names? Denominal verbs, manner incorporation, and the ontolo2005
gy of verb roots in English. In: Naom i Shir and Tova Rapoport (eds.), The Syntax of
Aspect, 42-64. Oxford: Oxford University Press.
Harley, Heidi and Jason D. Haugen
2007
Are there really two different classes of instrumental denominal verbs in English? Snippets 16: 9- 10.
Haugen, Jason D.
2008a Morphology at the Interfaces. Reduplication and Noun Incorporation in Uta-Aztecan.
Amsterdam/Philadelphia: Benjamins.
Haugen, Jason D.
2008b Denominal verbs in Uta-Aztecan. International Journal of American Linguistics 74(4):
439- 70.
Haugen, Jason D.
2009
Hyponymous objects and late insertion. Lingua 119: 242- 262.
Haugen, Jason D. and Heidi Harley
2013
Head-marking inflection and the architecture of grammatical theory: Evidence from
reduplication and compounding in Hiaki (Yaqui). In: Shannon T. Bischoff, Deborah
Cole, Amy V. Fountain and Mizuki Miyashita (eds.), The Persistence of Language.
Constructing and Confronting the Past and the Present in the Voices of Jane H. Hill,
133- 174. Amsterdam/Philadelphia: Benjamins.
Hill, Kenneth C.
2003
Hopi denominal and noun- incorporating verbs. In: Luis M. Barragan and Jason D. Haugen (eds.), Studies in Uta-Aztecan, 215- 244. Cambridge, MA: MIT Press.
Julien, Marit
2002
Syntactic Heads and Word Formation. Oxford: Oxford University Press.
Kayne, Richard
The Antisy mmetry of Syntax. Cambridge, MA: MIT Press.
1994
PA434
434

Sadock, Jerrold M .
1986
Some notes on noun incorporation. Language 62: 19- 3 1.
Sadock, Jerrold M.
1990
Rev iew of Mark Baker's incorporation: A theory of grammatical function cha ng ing.
Natural Language and Linguistic Theory 8: 129- 141.
Sadock, Jerrold M.
1991
Autolexical Syntax. A Theory of Parallel Grammatical Representations. Chicago: University of Chicago Press.
Sapir, Edward
1911
The problem of noun incorporation in American languages. American Anthropologist
13: 250-282.
Thornes, Timothy J.
2000
Instrumental prefixes in Northern Paiute: A functional and typological perspective. In:
Zarina Estrada Fernandez and Isabel Barreras Aguilar (eds.), Quinto Encuentro Internacional de Linguistica en el Noroeste. Torno I: studios Morfosintacticos. Vol. 1, 243272. Hermosillo: Editorial Unison.
Thornes, Timothy J.
2003
A Northern Paiute grammar with texts. Ph.D. dissertation, University of Oregon.
Travis, Lisa deMena
1984
Parameters and effects of word order variation. Ph.D. dissertation, MIT, Cambridge,
MA.
Tuggy, David H.
1986
Noun incorporation in Nahuatl. In: Scott DeLancey and Russell Tomlin (eds.), Proceedings of the Second Meeting of the Pacific Linguistics Conference, 455- 470. Eugene,
OR: Department of Linguistics, University of Oregon.
Van Valin Jr., Robert
1992
Incorporation in universal grammar: A case study in theoretical reductionism. Journal
of Linguistics 28: 199-220.
Weggelaar, C .
1986
Noun incorporation in Dutch. International Journ al of American Linguistics 52(3): 301314.
Jason D. Haugen, Oberlin (USA)
23. Particle-verb formation

l.
2.
3.
4.
5.
6.
What is a particle (verb)?

Semantic and argument-structural properties of particle verbs
The syntax of partic le verbs
Particle verbs and morphology
Particle verbs as morphological objects (complex heads)?
References
PA437
437
23. Particle-verb formati on

structural case position like arguments of passive or unaccusative verbs (Svenon ius 2003;
Mcintyre 2007: section 2.7).
(6)
a.
wipe the table off/wipe off the table (cf. wipe the dust off the tab le)
b. pump the cellar out/pump out the cellar (cf. pump the water out of the cellar)
c. Du solltest dem Mann nicht hinterher laufen
you should the mandat not after
run
' You shouldn't run after the guy.'
d.
German
Dem Mann hinterher] ist keiner gelaufen

the man after
is nobody run
'Nobody went after the guy.'
[pp
A harder case is (6c). The adposition hinterher occupies the particle position adjacent to
the clause-final verb. The dative dem Mann could be analysed as the ground argument
of hinterher, similarly to the dative in (6d ), where hinterher is a dative-assigning postposition forming a full PP. Such constructions would be problematic for the particles-quacomp/ementless-prepositions constraint in (3c) only on an analysis which assumes that
(6c) contains a PP like that seen in (6d). The non-adjacency between the particle and
dative in (c) makes this approach hard to defend; presumably hinterher would have to
reanalyse with the verb (without incorporating, since V can move away from hinterher),
allowing the dative to scramble despite German's usual ban on P-stranding. There are
alternative analyses which do not counterexemplify (3c). Zeller (200lb: 2 18-225) adumbrated a non-movement account involving percolation of P 's dative feature. Another
possibility for at least some such constructions is that the dative is not grammatically
represented as an argument of the particle but is a free dative. For more on datives with
particle verbs, see Mcintyre (2007) and Oya (2009).
2. Semantic and argument-structural properties of particle verbs

Resultative or directional particles like those in (7) predicate a result or direction over a direct
argument (i.e. direct object or unaccusative subject). As is often observed with resultative
constructions, the direct argument need not correspond to the verb's usual selection restrictions, cf. (7c) and *vote the government (Mcintyre 2007 and references).
(7)
a.
b.
c.
She put a hat on.

She walked in.
We voted the government in.
[on her head]

[into a contextually present room/building]
[into office/power/parliament, etc.]
In (7) the direct argument is a figure/theme while the ground is implicit. There are three
main deviations from this standard type of resultative particle. Firstly, in structures like
(6) the ground is realized instead of or along with the figure. Secondly, there is a rare
and rarely-discussed type of unergative particle construction like look in, phone through
where the particle coexists with an agent without predicating over it (Mcintyre 2004 ).
Thirdly, (8) shows a "pleonastic" construction, widespread in German and Dutch but
PA441
4.2. Particle verbs as input to derivational morp hology

and compounding
In head-final languages like Dutch and German, particle verbs freely feed word-formation processes, as ( 15) illustrates with German data. (15a, b) are compounds with particle
verbs as nonheads (here there is no evidence that the particle verbs have been nomina lized ; German freely forms compounds with verb stems as nonhead, cf. Gast 2008).
P article verbs inside other types of derviational morphology are seen in (15c-g).
(15) a.
b.
c.
d.
e.
f.
g.
ANziehsachen ' on.put.things = clothes'

WEGgehabend 'away.go.evening= evening where someone goes out'
ABfahrbereit 'off.go.ready = ready to leave'
WEGwerffreudig 'away.throw.joyful= who likes throwing things out'
RUMsteherei ' round.stand.Af =standing around '
ANbieter 'to.offerer = provider'
unAuFhorlich ' un .up.hear.ly =unceasing, from AUFhor 'cease; lit. up.hear'
ANnehmbar 'to.take .able = acceptable'
RUMgdabere 'around.Af.chat.Af = incessant chatter'
EINgekaufe ' in.Af.buy.Af = negative ly evaluated shopping'
UMzug ' moveN' - UMzieh- ' move houses/flats '
AUFnahme 'taking up' - AUFnehm- 'take up'
ANkunft 'arriva l' - ANkmmn - 'arrive'
UNTERkunft 'accommodation ' - UNTERkomm- 'find/have accommodation '
The types in ( 15e - g) raise special problems. In (e) the nominalizing circumfix Ge- ...
-e takes the whole particle verb as its semantic input (as is clear from cases where the
particle verb is lexicalised), even though ge- intervenes between the particle and verb
(Liideling 2001; Muller 2003; Stiebels and Wunderlich 1994). Structures like (15f- g)
("root nomina lizations") do not involve synchroni cally productive affixation or vowel/
consonant mutation. They could be analysed either as idiosyncratic spellouts of the verb
roots plus nominalizers, or as a llomorphs of the verb roots which surface in nominal
environments. In (g) the -kunft-nominals show the additional complication that -kunft
only co-occurs with particles. For more on root nominals, see Stiebels and Wunderlich
(1994), Becker (1993: 15f.).
Here it is expos itionally expedient to prefigure the question of section 5 regarding
the status of particle verbs as compound-like morphological obj ects (complex heads). It
is often argued that instances of particle verbs feeding affixation or compounding need
not show that particle verbs are morphological structures, since clearly phrasal constituents can also feed affixation and c ompounding. Instances of this include phrasal compounds like the particle-verbs-are-compounds stance and nominalizations like the destruction of the evidence immediately, where the adverb indicates that the nominalizer
attaches to a verb but to a VP or larger constituent (Fu, Roeper a nd Borer 2001). There
are, however, several complications. As part of a n argument th at partic le verbs, though
initially merged in phrasal configurations, reanalyze as X 0 items inside complex verbs,
Zeller (2001 b: ch . 6) notes that many German affixes disallow phrasal input. One can
add that compounds like (15a) do not evince the expressive, jocular or stylistically
marked feel of phrasal compounds discussed in Meibauer (2007), and that compounds
441
PA443
question for future research would concern the reasons for the differing behaviour of the
affixes in (18) and (16).
(18) a. the turning {off} of the lights {*off}
b. their turning {off} the lights {off}
c. sawnoff shotguns; un-written-up/under-worked-out ideas; falling-down houses
4.3. Verb stems in particle verbs

We now discuss some problems concerning the types of e lements which can function as
verb stems (or: non-particle elements) in particle verbs. Firstly consider the English and
German examples in (19). Here the non-particle elements lack (relevant) counterparts
which can be used as verbs without the particle. Explaining this by treating the particles
as heads of the construction is not an option since verbal inflection is realized on the
non-particle element and since the particles act like prepositions according to certain
tests (e.g., preposition-specific modifier: dumb it back down, soldier right on). The alternative is that the structures exploit the conversion (zero derivation) patterns available
without particles, with the additional complication that certain conversions are (in some
yet-to-be-understood sense) licensed only in the presence of a particle. (The same problem is found in clearly phrasal structures like worm one's way in, where worm is not
independently usable as a verb, and one's way is clearly a syntactic entity, cf. corporations have wormed their insidious way into governments.) On data like (19) see especially Stiebels (1998), as well as Booij (2002), Kolehmainen (2006: 48- 52), Mcintyre
(2002), Muller (2002: 337f.), Olsen (1998), Zeller (2001b: 211 - 215).
(19) a. slot in, ferret/bottom out, muck up/around, soldier on, beaver away, pig out
b. wise up, dumb down
c. AUFbahr- 'to put on a stretcher', EINsack- 'to put in a sack', EINsarg- 'to put
in a coffin'
d. AUFfrisch- 'to freshen up', ANreicher- 'to enrich'
Perhaps unsurprisingly, particles do not usually exhibit the types of morphological and
phonological restrictions found with affixes. One potential exception is Fraser's (1976:
13-16) observation that most English particle verbs have monosyllabic or forestressed
bisyllabic stems, cf. {ringlphone/ *telephone} her up. A related observation is that in
(20) the non-particle element appears in a clipped form not found without the particle.
The theoretical significance of these observations is unclear given the sporadic nature of
data like (20) and the existence of exceptions to the tendency: partition off, separate
out, continue on, gallyvant around.
(20) sum(*marise) up, (*con)fess up, (con)glom(erate) together
Another constraint affecting verb stems in particle verbs is the resistance to morphologically complex stems seen in (2 la). Fraser (1976: 15) sees the blockage as phonological,
while Keyser and Roeper (1992) assume that particles and prefixes originate in the
443
PA445
rable and thus obey constraints like lexical integrity (which bans syntactic manipulation
of parts ofX 0 items) or the oft-posited ban on excorporation (e.g., Baker 1988: 73). This
puts an explanatory burden on the complex word view, but whether this burden breaks
its back is another, partly theory-dependent question. To see this, we have to distinguish
two types of complex word approaches:
(i) The particle and verb only enter a complex word configuration in some of their uses
(for instance when the particle incorporates into, or reanalyses with, the verb). These
uses do not include the uses where the particle and verb are separated syntactically.
(ii) The particle and verb are always initially inserted into syntax as a complex word
but are separated in certain contexts.
Proponents of (ii) have the harder task given that complex heads are not normally separable. It is not problematic for (ii) that affixes do not separate from their hosts, since any
theory would presumably need (the empirical equivalent of) a stipulation in the lexical
entry of an affix which makes it inseparable from its base, and one can simply say that
particles lack this stipulation. The real challenge to (ii) arises when we compare particle
verbs to compounds. Here the options seem to be as follows. One could try to argue that
verbs separate from particles in, say, verb-second contexts like (2a) because verb movement (or its empirical equivalents) targets verb stems, and there is no reason to piedpipe the particle, since it is not an affix. To explain why constituents of compound nouns
or adjectives do not normally separate, one could assume firstly that nonnal compound
formation turns one of the constituents into an affix-like element, an operation which is
sometimes visible, cf. conjunct forms like oft (usable only in compound adjectives: oftderided) or linking morphs (interfixes) like the -s in tradesman. A second option would
be to assume that there are simply no syntactic operations which could legitimately target
parts of normal compounds. This would be possible if one assumes that nouns and
adjectives do not undergo (the empirical equivalent of) head movement operations, and
that compound non-heads cannot undergo phrasal movement (Car he is not a [N P tear
driver]) either because they are not phrases of the appropriate category (e.g., DP) or
because such movements would violate island constraints (e.g., the left-branch condition). These positions might be correct, but a satisfying defense would require a substantial cross-linguistic discussion of the reasons for the observation that compounds are
cross-linguistically normally inseparable.
The issue of particle modification often crops up in discussions of the complex word
view of particles. Particles (or at least complementless prepositions homophonous with
particles) can uncontroversially be modified, see (23a). Proponents of position (ii) above
would presumably have to assume a phrasal compound [push [right in]] here. This seems
hard to defend given that the "close union" phenomena discussed in section 1 are excluded with modified particles: English disallows particle-object order in the presence of a
modifier, cf. (23 b, c ), and it is hard to find convincing examples of particle-verb nominalizations where the particle is modified. If some variant of the complex word analysis of
particle verbs is right, then data like (23) seem to favor option (i) above. The modifier
would prevent the particle from compounding with, reanalyzing with, or incorporating
into, the verb because English simply does not license phrasal compounds of the type
[push [right in]], or because the modifier projects some structure which interrupts the
adjacency between particle and verb, so that reanalysis would be excluded by the head
movement constraint or some empirically equivalent principle.
445
PA446
446

(23) a.
b.
c.
d.
She pushed the pin [right in].

*She pushed [right in] the pin.
*She pushed in the pin right.
pin pusher-(* right )-in-er
(* if right modifies in)
A final point to note concerning the complex word view of particle verbs is that Englishstyle verb-initial particle verbs violate the right-hand head rule of Williams (1981 ), see,
e.g., den Dilkken (1995: 88). To evaluate this point, note firstly that some writers, including Williams himself (1981: 249f.), do not see this rule as exceptionless. Williams'
putative cou nterexamples include denominal prefix verbs lacking unprefixed verbal
counterparts (enthrone but *thronev). This argument seems dubious given that particles
can also "license" zero derivations or conversions but are clearly not heads of particle
verbs, recaU (19). Stronger evidence for left-headed morphological structures in English
comes from data like (24), which are attested and acceptable to many speakers not
influenced by language purism. Here plural -s appears on the right edge of the construction. This suggests that the construction is analysed as a morphological object, since
plural -s does not otherwise behave like a phrasal affix (the discussion(s) yesterday(*s),
the car(s) of his daughter(#s)). Nevertheless, structures like mother-in-law and passerby are clearly left-headed, witness, e.g., their semantics and the fact that some speakers
additionally attach the affix to the leftmost element, which seems to be a straightforward
case of head marking (recall again Stump's 1994 Breton examples).
(24) mother(s)-in-laws, passer(s)-bys, hanger(s)-on(s), fixer(s)-uppers
This section has hopefully clarified some of the issues regarding the debate concerning
the complex word view of particle verbs and shown why the last word on the (non)existence of morphological objects consisting of particles and verbs has not been said.
6. References
Baker, Mark
1988
In corporation. Chicago: University of Chicago Press.
Basilico, David
2008
Pa rticle verbs and benefactive double objects in English: High and low attachments.
Natural Language and Linguistic Theory 26: 731 - 773.
Becker, Thomas
1993
Back-formation, cross-formation, and 'bracketing paradoxes' in paradigmatic morphology. In: Geert Booij and Jaap van Marie (eds.), Yearbook of Morphology 1993, 1- 25.
Dordrecht: K luwer.
Berg, Thomas
1998
The (in)compatibility of morpheme orders a nd lexical categories and its historical implications. English Language and Linguistics 2: 245-262.
Blom, Corrien
2005
Complex predicates in Dutch. Utrecht: LOT.
Booij, Geert
1990
The boundary between morphology and syntax: Separable complex verbs in Dutch. In:
Geert Booij and Jaap van Marie (eds.), Yearbook of Morphology 1990, 45- 63. Dordrecht: Kluwer.
PA448
448

Kaufmann, Ingrid and Dieter Wunderlich
1998
Cross-linguistic patterns of resultatives. In: Theorie des Lexikons. Arbeiten des Sonderforschungsbereichs 282. Nr. 109. Heinrich-Heine-Universitat Dusseldorf.
Kayne, Richard
1985
Principles of particle constructions. In: Jacqueline Gueron, Hans-Georg Obenauer and
Jean-Yves Pollock (eds.), Grammatical Representations, 101- 140. Dordrecht: Faris.
Keyser, Samuel J. and Thomas Roeper
1992
The abstract clitic hypothesis. Linguistic Inquiry 23: 89- 125.
Kolehmainen, Leena
2006
Prafix- und Partikelverben im deutsch-finnischen Kontrast. Frankfurt/M.: Lang.
L indner, Susan
1983 A Lexico-Semantic Analysis of English Verb Particle Constructions. Bloomington, IN:
Indiana University Lingu istics Club.
Ludeling, Anke
2001
On Particle Verbs and Similar Constructions in German. Stanford: CSLI.
Ludeling, Anke and Nivja de Jong
2002
German particle verbs and word formation. In: Nicole Dehe, Ray Jackendoff, Andrew
Mcintyre and Silke Urban (eds.), Verb-Particle Explorations, 315-333. Berlin/New
Marchand, Hans
1969
The Categories and Types of Present-Day English Word Formation .. A Synchronic-Diachronic Approach. 2nd ed. Munchen: Beck.
Meibauer, Jarg
2007
How marginal are phrasal compounds? Morphology 17: 233-259.
Mcintyre, Andrew
2001
German Double Particles as Preverbs. Tubingen: Stauffenburg.
Mcintyre, Andrew
2002
Idiosyncrasy in particle verbs. In: Nicole Dehe, Ray Jackendoff, Andrew Mcintyre and
Sitike Urban (eds.), Verb-Particle Explorations, 97- 118. Berlin/New York: Mouton de
Gruyter.
Mcintyre, Andrew
2003
Preverbs, argument linking and verb semantics. In: Geert Booij a nd Jaap van Marie
(eds.), Yearbook of Morphology 2003, 119-144. Dordrecht: Kluwer.
Mcintyre, Andrew
2004
Event paths, conflation, argument structure and VP shells. Linguistics 42: 523- 571.
Mcintyre, Andrew
2007
Particle verbs and argument structure. Language and Linguistics Compass 1: 350-397.
Miller, Gary
2010
On the syntax of morphology. Ms., University of Florida.
Muller, Stefan
2002
Complex Predicates. Stanford: CSLI.
Muller, Stefan
2003
Solving the bracketing paradox: An analysis of the morphology of German particle
verbs. Journal of Linguistics 39: 275- 325.
Neeleman, Ad and Fred Weerman
1993
The balance between syntax and morphology: Dutch particles and resultatives. Natural
Language and Linguistic Theory 11: 433-75.
Nicol, Fabrice
2002
Extended VP-shells and the verb-particle construction. In: Nicole Dehe, Ray Jackendoff,
Andrew Mcintyre and Silke Urban (eds.), Verb-Particle Explorations, 165- 190. Berlin/
New York: Mouton de Gruyter.
PA449
449

Olsen, Susan
1996
Pleonast ische Direktionale. In: G isela Harras and Manfred Bieiwisch (edls.), Wenn die
Semantik arbeitet. Klaus Baumgartner zum 65. Geburtstag, 303-329. Tilbingen: Niemeyer.
Olsen, Susan
Ober den lexikalischen Status englischer Partikelverben. In: Gisa Rauh and Elisabeth
1997
Lobel (eds.), Lexikalische Kategorien und Merkmale, 45- 71. Ti.ibingen: Niemeyer.
Olsen, Susan (ed.)
Semantische und konzeptuelle Aspekte der Partikelverbbildung mit ein-. T i.ibingen:
1998
Stauffenburg.
Oya, Toshiaki
2009
Ground a rguments in German particle verbs. Journal ofLinguistics 21: 257-296.
Ramchand, Gillian and Peter Svenonius
The lexical syntax and lexical semantics of the verb-particle construction. Jn: Lina Mik2002
kelsen and Chris Potts (eds.), Proceedings of the West Coast Conference on Formal
Linguistics 21, 387- 400. Somerville, MA: Cascadilla Press.
Stiebels, Barbara
Lexikalische Argumente und Adjunkte. Berlin: Akademie Verlag.
1996
Stiebels, Barbara
1998
Complex denominal verbs in German and the morphology-semantics interface. In: Geert
Booij and Jaap van Marie (eds.), Yearbook of Morphology 1997, 265- 302. Dordrecht:
Foris.
Stiebels, Barbara and Dieter Wunderlich
1994
Morphology feeds syntax: The case of particle verbs. Linguistics 32: 913 - 968.
Stump, Gregory
The uniformity of head marking in inflectional morphology. In: Geert Booij and Jaap
1995
van Marie (eds .), Yearbook of Morphology 1994, 245- 296. Dordrecht: Kluwer.
Svenonius, Peter
2003
Limits on P: Filling in holes vs. falling in holes. Nord/yd 31: 431 - 445.
Svenonius, Peter
2010
Spatial Pin English. In: Guglielmo Cinque and Luigi Rizzi (eds.), Mapping Spatial PPs,
127- 160 . Oxford: Oxford University Press.
Toivonen, Ida
Non-Projecting Words. Dordrecht: Kluwer.
2003
Toivonen, Ida
On continuative on. Studia Linguistica 60: 181-219.
2006
Walker, Jim
2009
Double -er suffixation in English. Lexis. -Journal in English Lexicology I: 5- 14.
Williams, Edwin
1981
On the notions 'lexically related' and ' head of a word' . Linguistic lnquiryr 12: 245- 74.
Winkler, Susanne
Focus and Secondary Predication. Berlin/New York: Mouton de Gruyter.
1997
Zeller, Jochen
200 l a How syntax restricts the lexicon: Particles as thematic predicates. Linguistische Berichte
188: 461 - 494.
Zeller, Jochen
200 1b Particle Verbs and Local Domains . Amsterdam/Philadelphia: Benjamins.
Zeller, J ochen
2002
Particle verbs are heads and phrases. In: Nicole Dehe, Ray Jackendoff, Andrew Mcintyre
and Silke Urban (eds.), Verb-Particle Explorations, 233- 267. Berlin/New York: Mouton
de Gruyter.
Andrew Mcintyre, Berlin (Germany)
PA453
24. Multi-word expressions
This means that uttering a MWE involves the reproduction or retrieval of the phrase as
a whole from the mental lexicon rather than the production or computation of the individual parts (for discussion, see, e.g., Wray 2008). MWEs are regarded as lexical units
because they obviously form semantic units, that is, they function as an expression for
a particular concept, just as words do.
Second, in the view of traditional phraseological research, prototypical MWEs by
definition have non-compositional (or idiomatic) meaning, e.g., to keep one's fingers
crossed I German die Daumen driicken lit. 'to press the thumbs'. Although this is true
for many MWEs, it is nowadays - thanks to the seminal paper by Nunberg, Sag and
Wasow (1994) and others - generally accepted that non-compositionality should not
count as a defining criterion for MWEs. This is why many approaches distinguish between several degrees of idiomaticity/non-compositionality of MWEs. However, the notion of non-compositionality is usually associated with several aspects of meaning, such
as opacity, unanalysability or figurative meaning, which are often confusedl (or at least
not made explicit) in the literature, cf. Svensson (2008). Generally, a distinction is made
at least between fully opaque, non-decomposable MWEs (e.g., red herring), decomposable MWEs which contain one or several words with an idiomatic meaning (e.g., black
market) and fully compositional, non-idiomatic MWEs (e.g., fish and chips ). The latter
are often referred to as "collocations" (cf., e.g., Barz 1996; Riehemann 2001; Burger
2010) or as "institutionalised phrases" (cf. Sag et al. 2002; Villavicencio et al. 2005).
Third, MWEs have traditionally often been regarded as syntactically fixed expressions. However, as in the case of semantics, it is widely accepted nowadays that, al though many MWEs are indeed syntactically fixed or otherwise deficient, this is not
necessarily the case. Instead, MWEs exhibit a continuum of syntactic fixedness, which
is often related to the degree of compositionality of meaning (see, for instance, Fellbaum
2011). On the one end, there are, rather infrequently, fully invariant expressions (e.g.,
English by and large I German im GrojJen und Ganzen). Some MWEs contain unique
elements, i.e. items that only appear in one particular MWE and do not have a meaning
on their own ("cranberry collocations", cf. Moon 1998), e.g., klipp und klar 'clear as
daylight; lit. (klipp) and clear'. Other MWEs are restricted w ith regard to syntactic
operations such as anaphoric reference, passivization, relative clause formation, topicalization, modification, and others (cf. Nunberg, Sag and Wasow 1994; Dobrovol'skij 1997;
Moon 1998; Donalies 2009; Burger 2010, among others). For example, Dutch rode kool
' red cabbage' allows (in its MWE meaning) neither the modification of the adjective
nor the insertion of another prenominal adjective, cf. *erg rode kool 'very red cabbage',
*rode dure kool ' red expensive cabbage', cf. Booij (2009). However, although the word
order is fixed in these cases, they are not fully invariant forms, as can be seen from
similar examples in German. In spite of the syntactic restrictions, they exhibit regular
inflection, e.g., der blaue Fleck, dem blauen Fleck 'blue mark, bruise'. On the other
hand, many (often relatively complex) MWEs allow occasional variation in creative
speech, e.g., by means of adjectival modification (cf. Ernst 1981 ; Fellbaum and Stathi
2006; Stathi 2007; Burger 2010), e.g., German etwas unter den politischen Teppich kehren 'to brush something under the political carpet'. Finally, there are underspecified
MWEs, i.e. patterns with open slots. Two subclasses can be distinguished: (i) MWEs
with a more or less fixed group of lexical items that may be inserted in these slots,
resulting in expressions which are relatively similar semantically (cf. Dobrovol'skij
1988), e.g., to hit the roojlthe ceiling; German jemandem eins aufs Dach I auf den
453
PA454
454
Deckel I auf den Hut I auf die Nase I auf die Riibe geben lit. 'to give somebody something on the roof I on the hat I on the nose I on the conk, to conk somebody' . (ii) Patterns
with a variable slot to be filled by a non-restricted group of content words, such as the
NPN construction (N by N, N for N, N after N, etc., e.g., day after day, cf. Jackendoff
2008). The NPN construction partially coincides with constructions known as binomials
in traditional phraseology (e.g., Lambrecht 1984; Moon 1998; Burger 2010), e.g., Danish
to og to 'pairwise; lit. two and two', med hud og har 'neck and crop; lit. with skin and
hair ', German null und nichtig ' null and void', Hab und Gut 'goods and chattels'. Other
examples of patterns with open slots are the 'time' -away construction (Bill slept the
afternoon away, cf. Jackendoff 1997) or collocational sequences such as al an N of or at
the N of (e.g., a kind/lot/number of, at the end/time/beginning of; Renouf and Sinclair
1991 ; Biber 2009).
Finally, there is another aspect of MWEs' stability as units which has often been
referred to in the literature as "habitualness" or "recurrence". MWEs are combinations
of a minimum of two words which language users prefer over alternative combinations
with an equivalent meaning, so they occur more frequently (cf. Erman and Warren 2000).
This criterion is of particular importance for the identification of collocations as they
lack other properties such as deviant semantics or syntax. The question as to what should
count as "more frequently" as well as methodological questions of measuring frequency
have been subject to extensive discussion in the corpus linguistic literature, cf. Ba1isch
(2004), Biber (2009) for an overview. According to one common view, MWEs are a
combination of lexical items whose frequency of co-occurrence is larger than would be
expected on the basis of chance alone, cf. Gries (2008).
3. Multi-word express ion s and word-formation

The relation between MWEs and word-formation can be considered under three aspects:
The demarcation between MWEs and word-formation units. - Just like MWEs,
word-formation units are complex units made up from a minimum of two constituents. Cross-linguistically as well as language-specifically, languages differ greatly
with regard to the degree to which the distinction between morphological and syntactic complex entities is formally marked. For instance, while the distinction between nominal compounds and phrases can easily be made on the basis of stress
and inflection in German, Dutch and Danish, this is much more difficult in languages like English, French or Spanish. This difference is also reflected by the fact
that in the latter languages, compounds are not, or not consistently, written as one
word, in contrast to German, Dutch and Danish, which exhibit a consistent distinction between morphological and phrasal complex nominals in terms of spelling.
Particle verbs such as English to look up, German aufgeben 'to give up' or Dutch
opbellen 'to phone' are another case in point as they appear either as one single
unit or as two words, depending on the kind of sentence they appear in, e.g., Er
will den Plan au[geben ' He wants to give up the plan' vs. Er gab den Plan au['He
gave up the plan I gave the plan up'.
(ii) MWEs as alternative forms to word-formation units. - Often, MWEs and wordformation can be regarded as alternative means for naming a particular concept,
(i)
PA456
456

(cf. section 2), MWEs must be regarded as potential lexical units, too, rather than as
being lexical by definition. Acknowledging phrasal patterns of this kind means that there
must also be non-lexicalized MWEs, i.e. MWEs that have been occasionally coined as
instantiations from a certain phrasal pattern for use in a particular situation but have not
become lexicalized.
Closely related to their lexical status is another property shared by MWEs and wordformation units: their function. Both fonns function as linguistic signs for specific concepts, i.e. they are names. However, as has been repeatedly mentioned in the literature,
word-formation units - and in particular compounds - may also be used as mere descriptions, i.e. as expressions that describe a concept but do not name it - just like "regular'',
non-lexicalized phrases do. A case in point is the well-known example apple juice seat
when used in a particular conversation to indicate a seat in front of which a glass of
apple-juice has been placed (cf. Downing 1977). It is obvious that this use of apple
juice seat does not involve the existence of a corresponding self-contained concept.
Accordingly, one may wonder whether MWEs can have a descriptive function, too.
However, this question has - as far as we know - not yet been discussed in the literature.
Finally, both MWEs and word-formation units may have compositional or non-compositional semantics, and both may (but need not) contain constituents with a metaphorical meaning.
3.2. Differences between multi-word expressions

and word-form ation unit s
The main difference between MWEs and word-formation units is that, while both are
complex expressions, MWEs are phrases, i.e. syntactic entities, whereas word-formation
units are words, i.e. morphological entities.
A first question related to this difference in status between MWEs and word-formation units has to do with their mutual relations: should MWEs and word-formation units
be regarded as alternative means that complement each other or as competitive processes
instead? Obviously, lexical categories differ greatly with regard to their attraction to
MWE formation and word-formation. According to Fleischer (1996b, 1997b) and Barz
(2007), MWEs are abundant in the verbal domain, but they are less frequent with nouns
and adverbs and even more infrequent with adjectives in Gennan (and presumably also
in other (Germanic) languages). This unequal distribution can be related to the corresponding word-formation processes: whereas verbal compounding does not exist in German, or exists there only marginally, and the number of verbal prefixes is relatively
restricted, nominal compounding is highly productive. That is, MWEs and word-formation units seem to complement each other, supporting the view of MWEs and wordformation are alternative means of expanding the lexicon. Of course, other factors are
important in this connection, too. For instance, MWEs (especially verbal ones) are often
said to exhibit a high degree of expressivity (e.g., to sweat blood I G erman Blut und
Wasser schwitzen). Likewise, German MWEs seem to include a high proportion of technical terms and proper names (e.g., Echte Kamille 'German chamomile', Totes Meer
' Dead See'). Thus, stylistic factors also play an important role in the alternation between
MWEs and word -formation.
PA457
24. Multi-word expressions
There also seems to be a major difference in between the way in which MWEs and
word-formation units come into existence. Word-formation units, we can assume, are the
result of regular, more or less productive word-formation rules (or patterns or schemas,
depending on the particular framework used). On the other hand, many MWEs are made
from existing phrases in a secondary process, i.e. through semantic reinterpretation or
specialization, or just by becoming habitual through frequent use (cf. Fleischer l 997b;
Barz 2007). These changes may also affect the syntactic properties of the phrase, for
example by causing it to lose its syntactic flexibility. Such a view of MWEs and wordformation units in fact implies a fundamental difference between them: word-formation
is a primary process, it is regular and - given extralinguistic motivation - it is more or
less predictable, whereas MWE formation is a secondary process, i.e. an unsystematic
and idiosyncratic lexicalization of phrasal units, and it is not predictable at all. However,
in addition to highly idiosyncratic MWEs resulting from the alteration of existing phrases, there are also many MWEs that are obviously based on patterns or schemas. The
NPN construction and the 'time' -away construction mentioned in section 2 are pertinent
examples of this. Also, lexical A+N phrases (e.g., German blauer Fleck, Dutch rode
kool, cf. section 2) can be regarded as instantiations of an abstract lexical schema that
is equipped with additional morphosyntactic restrictions (compared to regular, non -lexicalized A+N phrases, cf. Booij 2009, Schlilcker 2014).
Under the assumption that MWEs may be formed from abstract underlying patterns
or schemas, MWEs become more similar to word-formation units, and the difference
between MWEs and word-formation is less fundamental than described above. This
view, however, makes important assumptions about the architecture of the language system as it assumes the existence of regular and productive phrasal patterns within the
lexicon. Obviously, such a view is incompatible with a modular view of the grammar
system, i.e. a strict separation of grammar and the lexicon as set forth in mainstream
generative grammar. On the contrary, it is in full agreement with the fundamental assumptions behind a number of related frameworks known as "constructionist" theories
(e.g., Goldberg 1995, 2006; Croft 2001; Jackendoff 2002), and, similarly, cognitive
grammar (e.g., Langacker 1987, 1991 ; for an overview of different linguistic frameworks
and MWE research, see Wray 2008; Gries 2008). Leaving aside many details and differences, these theories share the assumption that "constructions'', i.e. symbolic units that
are pairings of form and meaning, constitute the basic elements of the language system,
and that there is no such thing as a strict divide between grammar and lexicon. Not
surprisingly, much recent work on MWEs is construction-based, for example Riehemann
(2001 ), Masini (2005, 2009), Booij (2002, 2009, 2010), Jackendoff ( 1997, 2008), PoJ3
(2010). It can be said without doubt that the increased interest of theoretical lingu istics
in research into MWEs is connected with the development of constructionist frameworks
during the past decades: work on specific constructions such as the let alone construction
in Fillmore, Kay and O'Connor (1988) and the insight that MWEs are central to language
and cannot be disregarded as marginal by linguistic theory (e.g., Jackendoff 1995) were
fundamental for the further development of constructionist theory. Even earlier, similar
ideas of considering MWEs as being variable syntactic patterns with open slots within
the lexicon have been developed in more traditional research on phraseology under the
name of Modellbildung (cf. Hausermann 1977) and Phraseoschablone (cf. Fleischer
l 982/ l 997a).
457
PA458
458
3.3. Com petition between phrasal and morphologica l pattern s

Fellbaum (2011: 454) speculates that "perhaps the most important function of many
idioms, which may account for their universality and ubiquity, is that they provide convenient, pre-fabricated, conventionalized encodings of often complex messages". In contrast to "regular" phrases, the central functions of MWEs are the encoding of complex
messages and the providing of names for (complex) concepts ("naming function", cf.
Fleischer 1982: 129). This can be illustrated by comparing the MWE black market with
its paraphrase (taken from Merriam-Webster): "illicit trade in goods or commodities in
violation of official regulations; also: a place where such trade is carried on."
MWEs like black market share their naming function with morphologically complex
words like compounds (cf. Barz 2007). If, therefore, a clear formal distinction cannot be
drawn between complex morphological and phrasal entities, this functional equivalence
may lead to the assumption of "mixed" rather than pure morphological or syntactic
entities. For instance, it has been proposed that some English A +N sequences might be
constructions that are neither fully lexical nor fully syntactic and that there might therefore be a significant area of overlap between syntax and lexicon (cf. Sadler and Arnold
1994; Giegerich 2005).
The shared function of complex morphological and phrasal entities becomes even
more significant through language comparison. Comparing English A+N units with their
counterparts in Dutch and German reveals that speakers of Dutch very often use MWEs
(zwarte markt 'black market', rode wijn 'red wine') where German has compounds for
the same concepts (Schwarzmarkt, Rotwein) (cf. section 2 for some criteria for distinguishing phrases from compounds in Dutch and German). Dutch, like German, has A+N
compounds (jijnstof, Feinstaub 'fine particles', zuurdeeg, Sauerteig 'sour dough'), and
German, like Dutch, allows for phrases being used as lexical units (saurer Regen, zure
regen 'acid rain'). However, these forms are not equally distributed in both languages:
a great many of lexicalized A+N compounds in German correspond to A+N phrases in
Dutch (Dunkelkammer - donkere kamer 'darkroom', Vollmond - volle maan 'full
moon'; cf. Booij 2002; Hlining 2010). Within one language, lexicalization of one form
often blocks the other, corresponding form (as indicated by#), e.g., grune Welle 'progressive signal system; lit. green wave' vs. #Grunwelle, die Dunkelkammer 'darkroom' vs.
#die dunkle Kammer (although this is a well-formed phrase, it is blocked for the specific
interpretation expressed by the compound). The same is true for Dutch, where the potential compound #zuurregen and the potential phrase #dunne druk are blocked by the
existence of zure regen 'acid rain' and dundruk 'lightface'.
The fact that lexicalization of a phrase can be blocked by the existence of a compound
and that compounding can be blocked by the existence of a MWE shows that the syntactic and the morphological pattern are in competition with each other. This has been taken
as a further argument against theories in which syntax and lexicon are seen as distinct
modules of a language, since blocking only takes place among lexical elements (cf.
Booij 2002, 2010, among others).
The question, then, is which factors determine the choice of one of the two patterns
for encoding a certain concept. One factor is probably analogy with existing forms. For
German A+N sequences, the choice between both patterns seems to be sensitive to type
frequency effects, as shown by Schliicker and Plag (20 11). The realization of a nove l
PA460
460

one word. This can therefore be regarded as a case of "univerbation" . It is an endocentric
compound and the adjective functions as the syntactic and semantic head. As in the
phrasal expression, the comparison is made explicit by giving the "tertium comparationis" (the property denoted by the adjective), which is modified by the first element of
the compound (the noun). The comparison can thus be expressed syntactically/phraseologically or morphologically without a d ifference in meaning: ihre Haut war weijJ wie
Schnee I schneeweijJ - 'her skin was as white as snow I snow-wh ite'.
This parallel, however, only holds for the adjectival type of phrasal similes. Verbal
comparisons cannot be expressed by means of a compound: frieren wie ein Schneider
' to be very cold; lit. to freeze like a tailor ' - *schneiderfrieren; schimpfen wie ein
Rohrspatz 'to rant and rave; lit. to rant like a reed bunting' - *rohrspatzschimpfen. This
type of noun incorporation is ungrammatical, and verbal compounding is in general
highly restricted in German. This illustrates once more (as in the A+N case above)
that the syntactic patterns are less restricted, and accordingly more flexible, than the
morphological ones.
For the adjectival type, the coexistence of both the MWE and the compound is wellestablished in many cases. Pairs like flink wie ein Wiesel 'as quick as a flash; lit. as
nimble as a weasel ' - wieseljl.ink or schlank wie eine Gerte lit. '(as) slender as a whip' gertenschlank are well-known to native speakers of German. In many other cases, however, the compound is not conventionalized. Words like haubitzenvoll (< voll/blau wie
eine Haubitze 'as drunk as a skunk; lit. drunk like a howitzer') or bohnenstrohdumm
(< dumm wie Bohnenstroh 'as thick as a brick; lit. dumb like bean straw') are grammatical and can be found via Google, but they are not conventionalized and their use is
highly marked. The relationship between the two patterns is asymmetrical: all compounds expressing a stereotyped comparison have a corresponding phrasal comparison
but not vice versa. There are many phrasal similes without a corresponding compound.
Another difference concerns the syntactic behaviour of similes and compounds. Both
can be used predicatively or as an adverbial, but only the compound can be used prenominally as an attribute in an NP (der Kleinwagen war wieseljlink I jlink wie ein Wiesel
' the compact car was as nimble as a weasel' - der wieseljl.inke Kleinwagen). MWEs, on
the other hand, can only be used in postnominal position to modify a noun (as an apposition): der Kleinwagen, jlink wie ein Wiesel.
Thus, as in the A+N case, the morphological and the phrasal comparative pattern are
competitive only with regard to a relatively small subset of all possible similes. Compounding is, again, much more restricted than MWE formation. On the other hand,
compounds are words, and this makes them more versatile with regard to their syntactic
range of use.
3.4. Constructionalization
In phrasal similes the noun tends to lose its literal meaning. The comparison often seems
to be "strange" or far-fetched (Donalies 2009: 76). Why is Bohnenstroh ' dumb' (dumm
wie Bohnenstroh sein 'to be as thick as a brick; lit. dumb like bean straw')? What has
drunkenness to do with a skunk (as drunk as a skunk)? Other comparisons seem to be
motivated. For instance, it seems reasonable to consider a weasel as agile (jlink wie ein
PA462
462

adjective usually go back to formations expressing a comparison. The German compound
stocksteiflit. 'stickstiff', for example, got its meaning 'very stiff' through the comparison
steif wie ein Stock 'as stiff as a stick'. By analogy, new compounds with the constituent
stock have been coined (cf. Fleischer and Barz 1995: 231): stockblind 'stone-blind ',
stockkonservativ 'conservative to the core', stockburgerlich 'philistine to the core', stockkatholisch 'catholic to the core', stockreaktionar 'extremely reactionary', etc. This use
of stock- might also be related to the existence of adjectival participle verstockt 'obdurate' and refated words. In any case, the literal meaning of the noun Stock has been lost
in these adjectives. They can be accounted for by assuming a subschema of the one
given above, in which the position of the N is lexically filled: [stock+ A]A f-> 'very A I
extremely A I A to the core', cf. Hiining and Booij (forthc.).
The element stock is no longer (synchronically) identical to the noun Stock, it has
become a "prefixoid" or even a "prefix" (for a discussion of the concept of "affixoid",
see, e.g., Stevens 2005, Booij and Hiining 2014). The comparison is no longer part of
the meaning, so these compounds do not correspond to a MWE expressing a stereotyped
comparison. In a case like this, the phrasal simile can be seen as a starting point for the
development of a new morphological pattern. Through "univerbation", the comparison
can be expressed by a compound. The compound becomes lexicalized, the meaning may
become more abstract (intensification), and the relation to the meaning of the corresponding noun becomes opaque. Through the use in a series of compounds, the first element
(the noun) is reinterpreted and eventually becomes a bound morpheme. Ultimately, this
can change the status of the compounds in question. They are interpreted as instantiations
of a productive derivational word-formation process rather than as examples of compounding.
Summing up, the developments outlined here strongly support the idea of a hierarchical lexicon, containing words, MWEs, constructions and constructional schemas on different levels of abstraction (cf. Jackendoff 2008; Booij 2010). Subschemas allow for
generalizations of subsets of words and MWEs within a morphological category or within a certain phrasal construction. The examples presented here show that there is a
functional overlap between syntax and morphology (and the lexicon). MWEs and compounds often share certain functions and meanings, but as a result of their different
origin and structural status (syntactic phrase vs. morphological compound) they also
differ with respect to their range of use. The two patterns compete such that it sometimes
is not really possible to make a distinction between an MWE and a complex word, as
in the case of English A+N constructions. The overall picture is, however, that MWEs
and compounds are largely a complementary means for creating lexical units.
4. References
Bartsch, Sabine
2004
Structural and Functional Properties of Collocations in English. A Corpus Study of

Lexical and Pragmatic Constraints on Lexical Co-occurrence. Ttlbingen: Narr.
Barz, Irmhild
1988
Wortbildung und Nomination. Zur Theorie der Wortbildung im Deutschen. Dem Wirken
Wolfgang Fleischers gewidmet, 19-24. (Sitzungsberichte der Akaidemie der Wissenschaften der DDR 4/G). Berlin: Akademie Verlag.
PA464
464

Fellbaum, C hristiane and Ekaterini Stathi
2006
ldiome in der Grammatik und im Kontext: Wer briillt hier die Leviten? In: Kristel Proost
and Edeltraut W inkier (eds.), Von /ntentionalitiit zur Bedeutung konventionalisierter Zeichen. Festschriftfur Gisela Harras zum 65. Geburtstag, 125- 146. Tllbingen: Narr.
Fiedler, Sabine
2007
English Phraseology. A Coursebook. Tiibingen: Narr.
F illmore, Charles J., Pau l Kay and Mary Catherine O 'Connor
1988
Regularity and idiomaticity in grammatical constructions. Language 64: 50 1-538.
Fle ischer, Wolfgang
Phraseologie der deutschen Gegenwartssprache. Leipzig: VEB Bibliographisches Insti1982
tut.
Fleischer, Wolfgang
1992
Konvergenz und Divergenz von Wortbildung und Phraseologisierung . In: Jarmo Korhonen (ed.), Phraseologie und Wortbildung - Aspekte der Lexikonerweiterung, 53-65.
Tubingen: Niemeyer.
F le ischer, Wolfgang
1996a Phraseologische, terminologische und onymische Wortgruppen als Nominationseinheiten. In: Clemens Knobloch and Burkhard Schaeder (eds.), Nomination, fachsprachlich
und gemeinsprachlich, 147- 170. Opladen: Westdeutscher Verlag.
Fleischer, Wolfgang
1996b Zurn Verhaltnis von Wortbildung und Phraseologie im Deutschen . In: Jarmo Korhonen
(ed.), Studien zur Phraseologie des Deutschen und des Finnischen II, 333- 344. Bochum:
Brockmeyer.
Fleischer, Wolfgang
1997a Phraseologie der deutschen Gegenwartssprache. 2"d ed. Ti.ibingen: Niemeyer.
F le ischer, Wolfgang
1997b Das Zusammenwirken von Wortbildung und Phraseologisierung in der Entwicklung des
Wortschatzes . In: Rainer Wimmer und Franz-Josef Berens (eds.), Wortbildung und Phraseologie, 9- 24. T i.ibingen: Narr.
Fle ischer, Wolfgang and Irmhild Barz
Wortbildung der deutschen Gegenwartssprache. 2nd ed. Ti.ibingen: Niemeyer.
1995
Fraser, Bruce
1970
Idioms with a transformational grammar. Foundations of Language 6: 22- 42.
Gibbs, Raymond W. and Herbert L. Colston
Psycholinguistic aspects of phraseology: American tradition. In: Harald Burger, Dmitrij
2007
Dobrovol'skij, Peter Kuhn and Neal R. Norrick (eds.), Phraseology. An International
Handbook of Contemporary Research. Vol. 2, 819- 836. Berlin/New York: de Gruyter.
Giegerich, Heinz J.
Associative adjectives in English and the lexicon-syntax interface. Journal ofLinguistics
2005
41: 571 - 591.
Goldberg, Adele
Constructions. A Construction Grammar Approach to Argument Structure. Chicago:
1995
Un iversity of Chicago Press.
Goldberg, Adele
Constructions at Work. The Nature of Generalization in Language. Oxford: Oxford
2006
Un iversity Press.
Granger, Sylviane and Magali Paquot
2008
Disentangling the phraseological web. In: Sylviane Granger and Fanny Meunier (eds.),
Phraseology. An interdisciplinmy perspective, 27- 49. Amsterdam/Philadelphia: Be njamtns.
PA466
466

Moon, Rosamund
1998 Fixed Expressions and Idioms in English. A corpus-based approach. Oxford/New York:
Clarendon Press.
Moon, Rosamund
2007
Corpus linguistic approaches with English corpora. In: Harald Burger, Dmitrij Dobrovol'skij, Peter Ki.ihn and Neal R. Norrick (eds.), Phraseology. An International Handbook o.fContemporary Research . Vol. 2, 1045- 1059. Berlin/New York: de Gruyter.
Newmeyer, Frederick
1974
The regularity of idiom behaviour. Lingua 34: 327- 342.
Nunberg, Geoffrey, Ivan Sag and Thomas Wasow
1994
Idioms. Language 70: 491-538.
PoJ3, Michaela
2010
Under Construction. Cognitive and Computational Aspects of Extended Lexical Units.
Utrecht: LOT.
Renouf, Antoinette and John Sinclair
1991
Collocational frameworks in English. In: Karin Aijmer and Bengt Altenberg (eds.), English Corpus Linguistics, 128-143. London/New York: Longman.
Riehemann, Susanne Z.
2001
A constructional approach to idioms and word formation. Ph.D. dissertation, Stanford
Un iversity.
Sadler, Louis.a and Douglas J. Arnold
1994
Prenominal adjectives and the phrasal/lexical distinction. Journal of Linguistics 30: 187226.
Sag, Ivan, Timothy Baldwin, Francis Bond, Anne Copestake and Dan Flickinger
2002
Multiword expressions: A pain in the neck for NLP. In: Alexander Gelbukh (ed.), Computational Linguistics and Intelligent Text Processing, I -15. Mexico City/Berlin:
Springer.
Sailer, Manfred
2007
Corpus linguistic approaches with German corpora. In: Harald Burger, Dmitrij Dobrovol'skij, Peter Kiihn and Neal R. Norrick (eds.), Phraseology. An international Handbook of Contemporary Research . Vol. 2, l 060- 1071. Berlin/New Yo rk: de Gruyter.
Schli.icker, Barbara
2014
Grammatik im Lexikon. Adjektiv+ Nomen- Verbindungen im Deutschen und Niederlandischen. Berlin/Boston: de Gruyter.
Schli.icker, Barbara and Ingo Plag
2011
Compound or phrase? Analogy in naming . Lingua 121: 1539-1551.
Schuster, Saskia
2013
Variation und Wandel. Zur Konkurrenz morphologischer und syntaktischer Muster im
Deutschen und Niederlandischen. Ph.D. dissertation, Freie Universitat Berlin.
Sprenger, Simone
2003
Fixed expressions and the production of idioms. Ph.D. dissertation, Max Planck Instituut
voor PsycholinguYstiek, Nijmegen.
Stathi, Katerina
2007
A corpus-based analysis of adjectival modification in German idioms. In: Christiane
Fellbaum (ed.), Idioms and Collocations. Corpus-based Linguistic and Lexicographic
Studies, 81 - 108. London: Continuum.
Stevens, Christopher M.
2005
Revisiting the affixoid debate: On the grammaticalization of the word. In: Torsten
Leuschner, Tanja Mortelmans and Sarah De Groodt (eds.), Grammatikalisierung im
Deutschen, 71 - 83. Berlin/New York: de Gruyter.
PA468
468
1. Introduction
" Reduplication" is the systematically and productively employed repetition of words or
parts of words for the expression of a variety of lexical and grammatical functions.
As a morphological device exhibiting a range of theoretically challenging phonological
characteristics, it has been the subject of an ever-growing field of intense linguistic
research since the 1970s. Terminologically, besides occasionally also being referred to
by the less redundant form "duplication" (e.g., Inkelas 2008; see also Naylor 1986: 175),
other labels sometimes attached to the phenomenon in question as well as to related or
similar (but nonetheless distinct) phenomena include "(morphological) doubling" (e.g.,
Inkelas and Zoll 2005), "(re-)iteration" (e.g., Aboh, Smith and Zribi-Hertz 20 12), "repetition" (e.g., Gi l 2005) and "replication" (e.g., Mel'cuk 2006: 301 -302). Notwithstanding
this terminological diversity and its potential for causing confusion, over the years the
tenn "reduplication" has proven to be the most stable and most w idely-used one among
linguists engaged in the topic.
In addition to giving an overview of formal and functional properties (see also article
15 on uni ts of word-formation) displayed by reduplication in spoken languages from all
over the world (including creoles; see Kouwenberg 2003) as well as sketching an outline
of the phenomenon 's interpretation within different theoretical frameworks, thi s article
specifically concentrates on reduplication as a means of word-formation and its status in
the languages of Europe (for reduplication in sign language see Pfau and Steinbach
2005; Wilbur 2005; see also article 126 on word-formation and sign languages).
2. Forms
In modern twentieth-century linguistics, at the latest si nce the seminal dissertation by
Wilbur ( 1973 ), reduplication is commonly recognized as a morphological process characterized by the phonological peculiarity of providing the exponents for various linguistic
categories (i.e. the "reduplicants") by full or partial repetition of the respective base
forms. During the past four decades the at times surprising formal properties exhibited
by di fferent types of the process have been the main theoretical focus of reduplication
research.
2.1. General formal properties

" Full (or complete or total) reduplication" is the repetition of any morphological unit,
preferably from the root up to the whole word (see below). For example, Basque forms
intensive adj ectives like argi~argia 'very clear' (following the Leipzig glossing rules,
from here on a tilde will be used to indicate the boundary between base and reduplicant)
by reduplicating the adjectival base form without inflectional suffixes, in the case at
hand the singular determiner -a (Hualde and Ortiz de Urbina 2003: 360; see also article
182 on Basque). In Afrikaans, a noun including a plural suffix can be reduplicated to
express considerable number, e.g., bottel-s 'bottles' ~ bottels~bottels ' bottles and bottles
PA469
25. Reduplication
(i.e. many bottles)' (Botha 1988: 92). Falling in line with an apparent privilege of morphological roots to be at least partly repeated in all kinds of reduplication (cf. Inkelas
2012: 3 58), full reduplication of affixes alone is rather rare; one language exemplifying
the phenomenon is Fijian, where a collective or distributive prefix vef- can be reduplicated to express greater number of a noun, e.g., vanua 'country' - vef-vanua 'various
countries' - vef~vef-vanua 'larger number of countries' (Schiltz 1985: 367; a somewhat
more complicated example of prefix reduplication in the European language Hungarian
is discussed at length in Kiefer 1995-96; see also articles 37 on particle verbs in Hungarian and 181 on Hungarian).
"Partial reduplication" involves phonological and prosodic categories smaller than
the morphological base undergoing the process, offering a number of subtypes and classificatory parameters. Leaving possible but often controversial cases of single-segment
reduplication like gemination aside (but see El Zarka 2005), partial reduplication can
manifest itself in any way from simply reduplicating a consonant and vowel to the almost
complete repetition of a base, while the respective redu plicants can attach initially, internally or finally to, the latter. Illustrative examples come from N giyambaa attenuation by
an initial disyllabic foot reduplicant (cf. Donaldson 1980: 69-70) as in bara:y 'fast' bara~bara:y 'more or less fast, fastish' (Donaldson 1980: 72), Mangarayi final consonant-vowel-consonant reduplication as in the plural dyadic kin term gab;bam~bam-yi
'spouses' (additionally requiring the proprietive suffix -yi) from gab;bam 'spouse' (Merlan 1982: 215) and Chamorro medial consonant-vowel continuative reduphcation as in
hugando 'play' - hu~ga~gando 'playing' (Topping 1973: 259; the accent marks primary stress). As can be seen from the first and last example, the prosodic make-up of the
base is not necessarily exactly reflected in the reduplicant (the Ngiyambaa foot reduplicant only shows the onset consonant and nuclear vowel of the second base syllable,
while in Chamorro the sole reduplicant syllable lacks the base coda), pointing toward
an independently fixed shape for partial reduplicants merely to be filled by segmental
information from the base (see Moravcsik 1978: 307- 308, 311- 312, 315).
A further special characteristic when comparing a base and its redupnicant is the
structural reduction in terms of markedness occasionally found in partial reduplication.
In the example from Ngiyambaa above, vowel length is reduced in the reduplicant;
another widespread case of reduced marked structures is the simplification of consonant
clusters vis-a-vis the base as in Tagalog so-called proposed verbal forms, exemplified
here by mag-ta~trabahoh 'X will work' (the prefix mag- indicates agent focus) from
trabahoh ' work' (French 1988: 23). The latter example moreover demonstrates the possibility of a discontinuous string of segments in the base to form a partial reduplicant.
Intensive adjective formation in Turkish via partial reduplication as in sarz 'yellow'
- sap~sarz 'yellow like a quince (i.e. bright yellow)' (Muller 2004: 87; see also articles
77 on intensification and 184 on Turkish) reveals important characteristics as well. On
the one hand, such an example demonstrates that the reduplicant and base do not necessarily have to appear directly adjacent to one another. On the other hand, it can be seen
that what specifically interferes in this case is a so-called "fixed segment" in the reduplicant, i.e. a segmental unit not found in the base. Fixed segments are not restricted to
partial reduplication, though; they can also appear as sound substitutions or additions
with full reduplication in formations commonly known as "echo-words". Relevant examples for the latter can once again be adduced from Turkish, e.g., hasta 'sick' - has-
469
PA472
472

'thing' - mina~mina 'everything' (Terrill 2003: 36). Another frequent function of nomi nal reduplication (though quite distinct from any type of plurality) is diminution as in
Malagasy alahelo 'sadness' - alahelo~helo ' little sadness' (Keenan and Razafimamonjy 1998: 167; the grave and acute accents mark secondary and primary stress, respectively).
With verbs, the most common meanings expressed by reduplication belong to the
broad field of verbal plurality, including continuity or progressivity as in Swahili -cheka
' laugh' - - cheka~cheka ' keep laughing' (Novotna 2000: 65), repetition or iterativity as
in Eastern Oromo bab~baas-e 'took out often' (with the past tense suffix -e) from the
verbal base baas 'to take out' (Owens 1985: 84), frequentativity or habituality as in
Hmong Njua qua} 'to cry loudly' - quaj~quaj 'to always cry' (Harriehausen 1990: 4 7)
and distributivity or dispersion as in Plains Cree kah~kiwike-wak 'they visit from time
to time/here and there ' (the h indicates devoicing following the long fixed reduplicant
vowel, while the ending -wak adds specifications for person, number, animacy and transitivity) from the base kiwike- 'to visit' (Ahenakew and Wolfart 1983: 375). A lso, redupli cating a verb may intensify its meaning, e .g., Kwaza kahc- 'to bite' ~ kahc~kahc- ' to
keep on biting (ferociously)' (van der Voort 2003: 75). Furthermore, similarly to diminution in nouns, one frequently finds attenuation of verbal meanings with reduplication too
as in Malagasy manome 'gives' - manome- me 'gives a bit' (Keenan and Razafimamonjy 1998: 166).
Adjectives and adverbs often reduplicate for plural reference or agreement as well as
for distributivity, e.g., Amele ben ' big' - ben~ben ' many big things' (Roberts 1991:
121), Somalifiican 'good' - fiic~ican 'good (plural)' (Berchem 199 1: 159) and Georgian axal-i 'new' (with absolutive suffix -i) ~ axal- axali ' new (obligatorily distributed
over head noun)' (Gil 1988: 1042- 1043). Intensification as inBagirmiyou 'quickly' - youyou 'very quickly' (Stevenson 1969: 161) is also a very common function of adjectival
and adverbial reduplication, and as with verbs, attenuation is found as well, frequently
in reduplicated colour terms like Modern Hebrew tsahov 'yellow' - tsahav- hav 'yellowish' (Levkovych 2007: 152; the vowe l a in the reduplicated syllable is fixed and
appears in the base as well).
Minor word classes and certain subclasses of major word classes are also prone to
reduplication, with similar semantic effects as found in (and exemplified above with)
nouns, verbs, adjectives and adverbs. Pertinent examples are distributive numerals like
Lezgian c'uwad~c'uwad 'fifteen each' (Haspelmath 1993: 235) and indefinite pronouns
as in Gayo sahan 'who' ~ sahan- sahan 'whoever' (Eades 2005: 57).
The semantic interpretation of echo-reduplication (called the "Et Cetera interpretation" by Singh 2005: 266) is relatively uniform across d ifferent word classes as well as
across languages, context-dependently conveying different mixtures of generality, vagueness, plurality, indefiniteness, pejorative connotations and sometimes also intensification
(see Keane 2001: 56-58), e.g., Tamil puli 'tiger '~ puli- gili 'tiger and the like', vandu
' to come' ~ vandu- gindu 'to come, etc.', motti 'fat' - motti- gitti 'fat and the like'
and avan 'he ' - avan- givan 'he, etc.' (Abbi 1992: 2 1).
There are also reduplicative phenomena with unclear semantic effects. For example,
it is hard to tell whether in reduplications with simultaneous affixation as in Lavukaleve
reciprocal formations (Terri ll 2003: 366- 367) showing a suffix -ria and initial consonant-vowel reduplication of the stem (e.g., numa 'to choose' - nu- numa-ria 'to choose
each other') the reduplicative part itself contributes any meaning (in the Lavukaleve
PA473
25. Reduplication
case this is especially doubtful because reduplication in reciprocal forms is only obligatory with stems longer than two syllables) or if it is rather some kind of stem formation
process upon which morphological rules act (cf. Saperstein 1997: 160; see also Niepokuj
1997: 83- 86).
It should be noted that most of the above meanings can be summarized as somehow
expressing the concept of increased quantity, with the subtypes quantity of referents
(loosely plurality) and amount of emphasis or intensification (cf. Moravcsik 1978: 317).
The remaining meanings revolve around the quite opposite notions of diminution and
attenuation. The theoretical interpretation of such rather stark semantic contrasts is a
strong dividing line between two broad approaches to reduplication semantics to be
di scussed in the next subsection. There are, however, reduplicative meanings which do
not easily fit into any of the above categories too, e.g., perfectivity, future, possession,
inchoative and associative qualities (see Moravcsik 1978: 325; lnkelas and Zoll 2005:
14; Rubino 2005b: 20- 21). At the same time, some meanings seldom or never seem to
occur with any type of reduplication, e.g., gender, case and negation (for a longer list
exclusively dedicated to full reduplication see Stolz, Stroh and Urdze 2011: 194).
The above survey has left two important functions of reduplication completely out of
the picture, namely the creation of new words and word-class change, both often correlated with an additional change involving plurality, intensity or diminution/attenuation (cf.
Moravcsik 1978: 324). As the former functions constitute prime examples of wordformation, they will be described more thoroughly in the remaining main sections (for
further cross-linguistic overviews of reduplication semantics see Key 1965; Niepokuj
1997: 65- 87; Regier 1998).
3.2. Theoretical approaches to reduplication function

Although Moravcsik ( 1978) stands out as a notable exception in early modern reduplication research in also containing a detai led discussion of reduplicative meanings, the
paper's overall conclusion on this point nevertheless reads as follows: "Given that reduplication is neither the exclusive expression of any one meaning category in languages,
nor are the meanings that it is an expression of all subsumable under general classes, no
[emphasis mine] explanatory or predictive generalization about the meanings of reduplicative constructions can be proposed" (Moravcsik 1978: 325). The author thereby further
supported from a functional-typological perspective an opinion which has more or less
naturally fallen out from the formally oriented reduplication studies since Wilbur (1973).
The latter were either not concerned with reduplicative semantics at all or they saw no
reason to regard the meanings occurring with reduplications in different languages as
special in any way. This was particularly true in reduplication-as-affixation approaches
(e.g., Marantz 1982), in which the assumption ofreduplicants as merely being segmentally underspecified affixes could be paralleled without difficulty by the view that the same
range of meanings is in principle possible for reduplication and other types of affixation.
Despite her above quotation, Moravcsik (1978: 330) did acknowledge "a tendency
[... ] for languages to use reduplicative patterns - i.e. quantitative form differentiation for the expression of meanings that have something to do with the quantity of referents",
however. This already touched upon a topic of interest which would later on be taken
473
PA474
474

up more seriously especially by functional approaches to morphology (most notably
natural morphology; see article 9) under the heading of iconic form-meaning relationships or "(constructional) iconicity" (for an early example see Mayerthaler 1977). In
general, iconicity can be characterized as the non-arbitrary motivation of lexical items,
morphological processes and aspects of morphological and syntactic structure, either
directly by virtue of a sign's homology with the entity signified or, more abstractly, via
diagrammaticity, the latter being the mirroring of relations among concepts or elements
of discourse by the make-up of linguistic strnctures (cf. Downing and Stiebels 2012:
379). Reduplication is special in that it very often reflects both these aspects of iconicity;
it exhibits diagrammaticity because more complex concepts are expressed by more complex structures (cf. Mayerthaler 1977: 34), and it additionally shows homology because
repetition of form mirrors repetition of meaning (cf. Downing and Stiebels 2012: 394).
Observations like these led Kouwenberg and Lacharite (2005: 534) to propose the iconic
principle of reduplication: "More of the same form stands for more of the same meaning", where "more of the same meaning" must be interpreted in a metaphorical sense to
be able to include meanings like distributivity, continuity and intensification as well (cf.
Downing and Stiebels 2012: 395).
The obvious question now arises if and how widespread reduplicative meanings like
diminution and attenuation can also be motivated by the above principle. In her semantic
study of Malayo-Polynesian noun and verb reduplication, Kiyomi (1995) regarded reduplication as both iconic and non-iconic, with a so-called "consecutive" as well as a
"cumulative iconic process" bringing about all sorts of plurality and intensity meanings,
respectively (similarly, Fischer 2011: 59 speaks of repetition on a horizontal and a vertical axis here), and a non-iconic process akin to affixation being responsible for sundry
notions like diminution. On the other hand, Regier (1998) and Fischer (2011) both took
a similar stance in granting reduplication a cognitively grounded iconic base that is
semantically extendable via metaphoric and metonymic processes. The two authors also
proceeded similarly in deriving the problematic diminutive and attenuative meanings
from meanings ultimately connected to the concept of baby (Regier 1998) or the babytalk register in general (Fischer 20 11 ; see also Niepokuj 1997: 72-73). From a somewhat
different angle (and also in slight contrast to their earlier work on reduplication in Caribbean creoles, e.g., Kouwenberg and Lacharite 200 l ), Kouwenberg and Lacharite (2005:
540) subsumed diminution and attenuation in reduplication under the iconic principle by
proposing an extension of dispersive readings from discontinuous occurrence to approxi mation, e.g., Jamaican Creole yala~yala 'yellow-spotted', in which "[t]he real-world
effect of such scattered distribution of colour is to tone down rather than intensify the
colour, to diminish rather than augment it" (Kouwenberg and LaCharite 2005: 538),
ultimately yielding the meaning 'yellowish' (see Abraham 2005 for a critical reaction to
Kouwenberg and LaCharite 2005). Finally, invoking a statement going all the way back
to Pott (1862: l 02), Stolz (2007: 342-345) developed a revised model of reduplicative
iconicity, claiming that the latter is not primarily based on diagrammaticity or homology
but rather on the conceptual deviation a redupli cated form expresses vis-a-vis the norm
or prototype encoded in its unreduplicated counterpart. Obviously, by this the author was
not only able to accommodate diminution and attenuation but all kinds of other reduplicative meanings problematic for a traditional conceptualization of iconicity as well (see
also Stolz, Stroh and Urdze 2011: 178- 191).
PA476
476

many languages. Relevant examples include, among plenty of others, Afrikaans kort
'short ' ~ kort~kort 'every now and again' (Botha 1988: 118), Swahili bata (denoting a
kind of duck) ~ -bata~bata 'waddle' (Novotna 2000: 65) and Portuguese esconde~esconde (a traditional game of hide-and-seek), cf. esconder 'to hide' (Kroll 1991: 33).
According to the standard interpretation rule of Afrikaans reduplications and a number of additional conceptual devices presented in Botha (1988), kort~kort above should
mean 'for a very short period', its actual meaning thus being a case of lexicalization, a
property also typical of many compounds and derived words (cf. Botha 1988: 118; see
also Niepokuj 1997: 68). Apart from an apparent change in word class (see section
4.1.2), the Swahili case, on the other hand, is an example of a transfer of meaning based
on similarity (see Novotna 2000: 65), a notion conceptually close to attenuation (cf.
Moravcsik 1978: 323), which in turn is very often expressed by reduplication as has
already been pointed out in the previous main section. In a similar vein, the Portuguese
game-name reduplication can be interpreted as being connected to plurality via the conceptual link that the playing of a game typically (in hide-and -seek especially) involves
more than one participant (cf. the discussion of Afrikaans reduplicated game names in
Botha 1988: 122-128).
4.1.2. Word-class derivation
Changing the word class of a base is a very common function of reduplication in various
languages. It can affect all major word classes and displays all possible directions of
change. To give just a handful of examples, Marshallese reduplicates a noun like bahat
'smoke' to form a verb bahat~hat ' to smoke' (Harrison 1973: 439), while in Papiamentu
an adjective like zeta~zeta 'very oily' can be formed by reduplicating a noun, cf. zeta
'oil' (Kouwenberg and Murray 1994: 21 ). Similarly, Chamorro verbs like kanno' 'to eat'
and hatsa 'to lift' can be nominalized or changed into a modifier via reduplication, cf.
ka~kanno' 'eater' (Topping 1973: 182) and ha~hatsa 'lifting (attributive)' (Topping
1973: 103), respectively. Finally, word-class changing reduplication from adjectival bases can be illustrated by Swahili -tamu 'sweet' ~ tamu~tamu 'sweets, confectionary'
(Novotna 2000: 63) and Nama (a click language in which morpheme-final nasals can
also carry tone) !6rh 'difficult' ~ !6rh~ !om 'to make (something) difficult' (Hagman
1977: 18).
Although in general word-class derivation is functionally quite distinct from the broad
meaning categories of plurality, intensification and diminution so often found in reduplication (and discussed in the previous main section), it frequently occurs simultaneously
with such notions in reduplicative constructions (cf. Moravcsik 1978: 324), for example
in the Papiamentu denominal adjective above which also expresses intensification.
4.1.3. Intra-category changes and other functions
Reduplication can also bring about clearly derivational changes within a given word
class, a typical intra-category change of this kind being one that affects the transitivity
of a verb as in Mokilese koso 'to cut (transitive)' ~ kos~kos 'to cut (intransitive)'
PA477
25. Reduplication
(Harrison 1973: 415). Another clear example of reduplicative derivation is the diminution of nouns which has already been exemplified in the previous main section. However,
many cases of reduplication can pose problems when trying to classify them in terms of
their morphological function.
It has long been noted in linguistics that the difference between inflection and derivation is not necessarily a clear-cut one (see article 14 on the delimitation of derivation
and inflection). While the previous main section showed that the meanings and functions
of reduplication are at least to a great extent easily characterized in a general fashion
(theoretically in terms of iconicity, for example, if one subscribes to such a concept),
the question as to whether morphologically these meanings and functions are inflectional
or derivational in nature is often much harder to answer (see also Inkelas 20]4). Typically problematic instances are all sorts of meanings pertaining to plurality (cf. number in
nouns and adjectives as well as certain aspectual differentiations in verbs) and intensification (cf. the degree of adjectives) but also less common reduplicative functions like
forming the perfect (i.e. tense) as in some Latin and Ancient Greek verbs (for the latter
see article 116 on from Ancient Greek to Modern Greek). However, it should be pointed
out that many of these meanings can be related to inherent (i.e. derivation-like) inflectional categories in the sense of Booij (1993, 1996). What is more, even at first sight
clearly inflectional meanings like plural agreement often turn out to lack some decisive
properties like obligatoriness when expressed by reduplication (e.g., Owens 1985: 93 on
Eastern Oromo). The question thus arises if indeed there is unambiguous reduplicative
inflection at all in the languages of the world (see also Saperstein 1997: 161 - 163).
4.2. Lexica l reduplication

Lexical reduplicative formations (often called frozen, fossilized or pseudo-reduplications) lack an unreduplicated counterpart but may nevertheless exhibit striking formal
and functional parallels to productive reduplication. Many of the forms in question probably hail from a once productively used pattern, others may have a more spontaneous
and transparent origin (for general diachronic aspects ofreduplication, including possible
scenarios of grammaticalization, see Bybee, Perkins and Pagliuca 1994: 166- 174; Niepokuj 1997; Hurch and Mattes 2005; Stolz 2008; Stolz, Stroh and Urdze 2011: 147204).
4.2.1. Inherent plural ity and other semantic fields
There is a cross-linguistic tendency for words containing some degree of inherent plurality to be formally reduplicated, especially zoological expressions, certain kinds of movement and botanical terms, e.g., Welsh pilipala ' butterfly' (see article 153 on Welsh),
Koasati wananatlin ' to shiver ' (Kimball 1988: 438) and Portuguese lemba-lemba (a local
term from Sao Tome) meaning 'liana, cord' (Kroll 1991: 28), respectively.
Further semantic fields found fairly frequently in lexical reduplication relate to human
beings or bodyparts, e.g., Bagirmi fitik 'bowels' (Stevenson 1969: 18), nature or natural
phenomena, e.g., Lavukaleve lamulam 'storm' (Terrill 2003: 106), diseases or sickness,
477
PA478
478

e.g., Ngiyambaa giral)gira 'sickly' (Donaldson 1980: 37), and colours, e.g., Mokilese
imwpwilapwil 'pink' (Harrison 1973: 437), to name just a few (for a systematic compilation of Portuguese lexical reduplications according to such categories see Kroll 1990,
1991 ).
4.2.2. Expressive formations
Expressives of all sorts (see also article 62 on the pragmatics of word-formation), often
with negative connotations, are also prone to be lexically reduplicated, e.g., German
plemplem (sein) '(to be) nuts' (see article 134 on German; for German reduplication and
word-formation see also Wiese 1990; Schindler 1991 ), Portuguese xexe ' ridiculous person' (Kroll 1991: 40) and Swahili halahala 'immediately !, at once !' (Novotna 2000:
68).
EspeciaUy widespread are reduplicated expressive forms of a sound-symbolic nature,
i.e. onomatopoeia and ideophones, e.g., Danish klipklap (the sound of wooden shoes or
horseshoes; see article 140 on Danish), Spanish cucu (cry of the cuckoo; see article 145
on Spanish), Polish and Slovak chi-chi (laughter; see articles 156 on Polish and 159 on
Slovak), Albanian bu(m)bullin ' to thunder' (see article 171 on Albanian), Ossetic c'ipc'ip (sound produced by young chickens; see article 173 on Ossetic ), Tat jiv-jiv zere 'to
chirp' (ideophonic coverb; see article 174 on Tat), Udmurt zup-zup (heartbeat; see article
178 on Permic ), Mari vrek-vrek (a cat or hare jumping; see article 179 on Mari) and
Chuvash cuccu 'swing' (noun based on sound imitation; see article 190 on Chuvash).
5. Reduplication, word-formation and the languages of Europe

Putting the focus on Europe, striking observations on the status of redupl ication in general as well as the phenomenon's specific connection to word-formation can be made. As
the examples provided throughout the previous main sections have already suggested,
reduplication is normally found with onomatopoeia in European languages, next to some
expressive forms and words relatable to certain common semanti c fie lds. Compared to
such cases of lexical reduplication, productive reduplication of whatever kind seems to
be rare in this part of the world (see also article 16 on derivation). This state of affairs
is normally related to the fact that the predominant language family spoken in Europe
is (western) Indo-European, which is generally believed to lack reduplicating languages
in a narrow sense. In additional support of this, important exceptions to the above characterization of Europe as an essentially reduplication-free area typically come from languages like Basque (an isolate) and Hungarian (Uralic) as well as from other non-IndoEuropean family members encountered when moving towards or straddling the outer
borders of the continent (e.g., Turkish, Georgian and Modern Hebrew).
The occurrence of productive reduplication thus appears to become markedly higher
with an increasing di stance from uncontroversial mainland Europe and western IndoEuropean languages, a fact which incidentally is nicely reflected in the sketches of individual languages collected in the present handbook (cf. articles 134-207). In principle,
all kinds of reduplicative forms and functions show up, e.g., fu lly reduplicated plurality
PA479
25. Reduplication
in Komi iun 'day' - lun-iun 'every day' (see article 178 on Permic), partially reduplicated intensification (with fixed segmentism) in Gagauz koca 'big' - kos~koca ' huge' (see
article 188 on Gagauz), attenuating echo-word formation in Akhwakh lemada 'liquid'
- lema~jemada ' more or less liquid' (see article 205 on Akhwakh) as well as wordclass derivation, cf. Tat para ' piece' - para~para 'scattered ' (see article 174 on Tat),
Karaimjyrach 'far' - jyrach--jyrach 'far away' (see article 189 on Karaim) and Sardinian curre 'run' - curre curre 'in a hurry' (see article 149 on Sardinian). In coming from
a Romance language, the latter example is hinting at a controversial topic which has
only recently gained fresh interest to be discussed below.
In line with the mainstream opinion presented above, the typological map of The
World Atlas of Language Structures (Haspelmath et al. 2005) pertaining to reduplication
portrays Europe as a more or less blank spot concerning this feature (see Rubino 2005a:
116-117). Among other things as a reaction to the methods reflected in Rubino (2005a),
Stolz, Stroh and Urdze (2011) set out to redraw the European reduplicative landscape in
an extensive areal-typological study concentrating on full reduplication. By quantitatively and qualitatively investigating a large corpus of literary texts instead of solely re lying
on written grammars, the authors take up several languages of Europe into the productively reduplicating class which more or less have been excluded before, but their approach crucially involves the expansion of the definition of reduplication so as to also
include what many other researchers would rather count as syntactic repetition (see Stolz,
Stroh and Urdze 2011: 26). Also, the reduplicative functions they discover often shaddle
the line between emphasis and intensification, i.e. they make it hard to unambiguously
assign certain forms as belonging to either pragmatics or grammar (see Stolz, Shoh and
Urdze 2011: 137-147). What is more, their results indeed only pertain to full reduplication as the authors explicitly state themselves that they find "no compelling evidence of
P[artial] R[eduplication] as a systematically employed grammatical device in Europe"
(Stolz, Stroh and Urdze 2011: 490; see also article 16 on derivation). It thus remains an
open question to what extent reduplication in most European languages is really comparable to reduplication as it is found in the rest of the world, be it from a formal or from
a functional perspective.
Acknowledgemen t s
Many thanks are due to Bernhard Hurch for his suggestions concerning the overall
outline of this article as well as some points of detail. Furthermore, much is owed to my
former position as a research ass istant in Hurch 's Graz reduplication project (duration
2005- 2010) funded by the Austrian Science Fund (project number P18173-G03), a main
outcome of which was the Graz database on reduplication (see Hurch and Mattes 2007,
2009) from where most of the examples given here have been drawn (http://reduplica
tion.uni-graz.at/).
6. References
Abbi, Anvita
1992
Reduplication in South Asian Languages. An Areal, Typological and Historical Study.
New Delhi: Allied Publishers.
479
PA481
25. Reduplication
French, Koleen Matsuda
1988
Insights into Tagalog Reduplication, Infixation and Stress from Nonlinear Phonology.
Dallas: Summer Institute of Linguistics.
G il, David
1988
Georgian reduplication and the domain of distributivity. Linguistics 26( 6): 1039- 1065.
Gil, David
From repetit ion to reduplication in Riau Indonesian. In: Bernhard Hurch (ed.), Studies
2005
on Reduplication, 31- 64 . Berlin/New York: Mouton de Gruyter.
Hagman, Roy S.
1977
Nama Hottentot Grammar. Bloomington, IN: Indiana University Publications.
Harrie hausen, Bettina
1990
Hmong Njua. Syntaktische Analyse einer gesprochenen Sprache mithilfe datenverarbeitungstechnischer Mittel und sprachvergleichende Beschreibung des siidostasiatischen
Sprachraumes. Ti.ibingen: Niemeyer.
Harrison, Sheldon P.
1973
Reduplication in Micronesian languages. Oceanic Linguistics 12(1-2): 407-454.
Haspelmath, Martin
1993
A Grammar of Lezgian. Berlin/New York: Mouton de Gruyter.
Haspelmath, Martin, Matthew S . Dryer, David Gil and Bernard Comrie (eds.)
Th e World Atlas of Language Structures. Oxford : Oxford University Press.
2005
Hendricks, Sean
200 I
Bare-consonant reduplication w ithout prosod ic templates: Expressive reduplication m
Semai. Journal of East Asian Linguistics 10(4): 287- 306.
Hualde, Jose Ignacio and Jon Ortiz de Urbina
2003
A Grammar of Basque. Berlin/New York: Mouton de Gruyter.
Hurch, Bernhard and Veronika Mattes
2005
Ober die Entstehung von partieller Reduplikation. In: Gertraut Fenk-Oczlon and Christian Winkler (eds.), Sprache und Natiirlichkeit. Gedenkbandfiir Willi Mayerthaler, 137156. Ti.ibingen: Narr.
The Graz database on reduplication. In: Alexis Michaud and Aliyah Morgenstern (eds.) ,
2007
La Reduplication , 191-202. Paris: Ophrys.
2009
Typology of reduplication: The Graz database. In: Martin Everaert, Simon Musgrave
and Alexis Dimitriadis (eds.), The Use of Databases in Cross-Linguistic Studies, 301327. Berlin: Mouton de Gruyter.
Inkelas, Sharon
2005
Morphological doubling theory: Evidence for morphological doubling in reduplication.
In: Bernhard Hurch (ed.), Studies on Reduplication, 65-88. Berlin/New York: Mouton
de Gruyter.
Inkelas, Sharon
2008
The dual theory of reduplication. Linguistics 46(2): 35 1- 401.
Inkelas, Sharon
2012
Reduplication. In: Jochen Trommer (ed.), The Morphology and Phonology of Exponence, 355-378. Oxford: Oxford University Press .
Inkelas, Sharon
2014
Non-concatenative derivation: Reduplication. In: Rochelle Lieber and Pa vol Stekauer
(eds.), The Oxford Handbook of Derivational Morphology, 169- 189. Oxford: Oxford
University Press.
Inkelas, Sharon and Cheryl Zoll
2005
Reduplication. Doubling in Morphology. Cambridge: Cambridge U niversity Press.
481
PA483
25. Reduplication
McCarthy, John J. and Alan S. Prince
1999
Faithfulness and identity in prosodic morphology. In: Rene Kager, Harry van der Hulst
and Wim Zonneveld (eds.), The Prosody-Morphology Interface, 2 I 8- 309. Cambridge:
Cambridge University Press.
McCarthy, John J., Wendell Kimper and Kevin Mullin
20 I2
Reduplication in harmonic serialism. M0tphology 22(2): 173- 232.
Mel'cuk, Igor
2006
Aspects of the Theory of Morphology. Berlin/New York: Mouton de Gruyter.
Merlan, Francesca
Mangarayi. Amsterdam: North-Holland Publishing Company.
1982
Moravcsik, Edith A.
1978
Reduplicative constructions. In: Joseph H. Greenberg (ed.), Universals of Human Language. Vol. 3: Word Structure, 297- 334. Stanford: Stanford University Press.
Millier, Hans-Georg
2004
Reduplikationen im Tiirkischen. Morphonologische Untersuchungen. Wiesbaden: Harrassowitz.
Naylor, Paz Buenaventura
On the semantics of reduplication. In: Paul Geraghty, Lois Carrington and Stephen
1986
Adolphe Wurm (eds.), FOCAL I. Papers from the Fourth International Confe rence on
Austronesian Linguistics, 175-185. Canberra: The Australian National University.
Niepokuj , Mary
1997
The Development of Verbal Reduplication in lndo-European . Washington: Institute for
the Study of Man.
Novotna, Jana
2000
Reduplication in Swahili. Afrikanische Arbeitspapiere 64: 57- 73.
Owens, Jonathan
A Grammar of Harar Oromo (Northeastern. Ethiopia). Hamburg: Buske.
1985
Pfau, Roland and Markus Steinbach
Backward and sideward redupl ication in German Sign Language. Jn: Bernhard Hurch
2005
(ed.), Studies on Reduplication, 569- 594. Berlin/New York: Mouton de Gruyter.
Pott, August Friedrich
1862
Doppelung (Reduplikation, Gemination) als eines der wichtigsten Bildungsmittel der
Sprache, beleuchtet aus Sprachen al/er Welttheile. Lemgo/Detmold: Verlag der Meyer'schen Hofbuchhandlung. http://reduplication.uni-graz.at/pott/index.html [last access
18 Apr 2013].
Raimy, Eric Stephen
2000
The Phonology and Morphology of Reduplication . Berlin/New York: Mou ton de Gruyter.
Regier, Terry
1998
Reduplication and the arbitrariness of the sign. In: Morton Ann Gernsbacher and Sharon
J. Derry (eds.), Proceedings ofthe Twentieth Annual Conference ofthe Cognitive Science
Society, University of Wisconsin -Madison, August 1-4, 1998, 887-892. Mahwah: Erlbaum.
Roberts, John R.
199 1
Reduplication in Amele. In: Tom Dutton (ed.), Papers in Papuan Linguistics No. I ,
115- 146. Canberra: The Australian National University.
Rub ino, Carl Ralph Galvez
2005a Reduplication. In: Martin Haspelmath, Matthew S. Dryer, Dav id Gil and Bernard Comrie (eds.), Th e World Atlas of Language Structures, 114- 117. Oxford: Oxford University
Press.
483
PA484
484

Rubino, Carl Ralph Galvez
2005b Reduplication: Form, function and distribution. In: Bernhard Hurch (ed.), Studies on
Reduplication, 11- 29. Berlin/New York: Mouton de Gruyter.
Saperstein, Andrew D.
1997
A word-and-paradigm approach to reduplication. Ph.D. dissertation, Department of Linguistics, The Ohio State University.
Schindler, Wolfgang
1991
Reduplizierende Wortbildung im Deutschen. Zeitschrift fur Phonetik, Sprachwissenschaft und Kommunikationsforschung 44(5): 597- 613.
Schutz, Albert J.
1985
The Fijian Language. Honolulu: University of Hawaii Press.
Singh, Rajendra
2005
Reduplication in Modem Hind i and the theory of reduplication. In: Bernhard Hurch
(ed.), Studies on Reduplication, 263-281. Berlin/New York: Mouton de Gruyter.
Steriade, Danca
1988
Reduplication and syllable transfer in Sanskrit and elsewhere. Phonology 5(1): 73-155.
Stevenson, R. C.
1969
Bagirmi Grammar. Khartoum: University of Khartoum.
Stolz, Thomas
2007
Re: duplication: Iconic vs counter-iconic principles (and their areal correlates). In: Paolo
Ra mat and Elisa Roma (eds.), Europe and the Mediterranean as Linguistic Areas. Convergencies from a Historical and Typological Perspective, 317- 350. Amsterdam/Philadelphia: Benjamins.
Stolz, Thomas
2008
Grammatikalisierung ex nihilo : Totale Reduplikation - E in potentielles Universale und
sein Verhiiltnis zur Grammatikalisierung. In: Thomas Stolz (ed.), Grammatikalisierung
und grammatische Kategorien , 83- 109. Bochum: Brockmeyer.
Stolz, Thomas, Cornelia Stroh and Aina Urdze
2011
Total Reduplication. The Areal Linguistics of a Potential Universal. Berlin: Akademie
Verlag.
Terrill, Angefa
A Grammar of Lavukaleve. Berlin/New York: Mouton de Gruyter.
2003
Topping, Donald M.
1973
Chamorro Reference Grammar. Honolulu: University of Hawaii Press.
Vollmann, Ralf
2009
Reduplication in Tibetan. Grazer Linguistische Studien 71: 115-134.
Voort, Hein van der
2003
Reduplication of person markers in Kwaza. Acta Linguistica Hafniensia 35: 65- 94.
Wiese, Richard
Uber die Interaktion von Morphologie und Phonologie: Reduplikation im Deutschen.
1990
Zeitschrift far Phonetik, Sprachwissenschaft und Kommunikationsforschung 43(5): 603624.
Wilbur, Ronnie Bring
Th e Phonology of Reduplication . B loomington, IN: Indiana U niversity Linguistics C lub.
1973
Wilbur, Ronnie Bring
2005
A reanalysis of reduplication in American Sign Language. In: Bernhard Burch (ed .),
Studies on Reduplication, 595- 623. Berlin/New York: Mouton de Gruyter.
Thomas Schwaiger, Graz (Austria)
PA485
26. Word-creation
26. Word-creation
l.
2.
3.
4.
5.
Definition: word-creation versus word-formation

Creative techniques
Functions of word-creation
Word-creation in a cross-linguistic perspective
References
Abstract
The term word-creation ref ers not only to the coining of new lexemes not based on any
previously existing meaningful linguistic elements. It also covers linguistic operations
such as shortening, blending, alienation, and so-called extragrammatical derivation,
which are deliberately peiformed on the basis of existing words or phrases, but outside
the productive models or rules of word-formation. The present article discusses the diff erences between word-creation and regular word-formation and presents an overview
of the basic creative techniques, their fan ctions and their cross-linguistic implementation.
1. Definition: word-creation versus word-formation

The tenn word-creation is traditionally reserved for the coining of a new lexeme without
using any previously existing meaningful linguistic element (German Urschopfung, cf.
Paul 1920: 174-175, taken up, e.g., in Fleischer and Barz 20 12: 18-20). In a wider
sense, which is adopted in this article, the term word-creation also refers to linguistic
operations such as shortening, blending, alienation, and others, which are deliberately
performed on the basis of existing words or phrases, but outside the productive models
or rules of word-formation. In contradistinction to the latter, such operations are here
called creative techniques. In this article, the focus is on those techniques which are not
treated in separate articles in this volume (see articles 19 on clipping, 2 1 on blending,
25 on reduplication).
The distinction between regular word-formation and word-creation is a matter of
debate among mo,rphologists, mostly argued out with respect to blending and/or clipping
(Ronneberger-Sibold 2010).
The major arguments in favor of a principled difference between regular word-formation on the one hand and word -creation on the other are the following:
a) The output of a creative technique is not predictable from its input, because the way
in which a technique is applied is not fixed in every detail. E.g., blending.flimsy and
miserable could result in (attested) mimsy as well as in (unattested) jliserable (Bauer
2003: 47).
b) Because of this indetenninacy, the output of a creative technique cannot be evaluated
according to its grammaticality. It makes no sense to ask native speakers of English
whether mimsy or jliserable are acceptable or non-acceptable as blends of.flimsy and
miserable. All they can possibly evaluate is the efficiency of these words in fulfilling
485
PA486
486

their purpose as an amusing word play, and in this respect, mimsy might be the better
solution.
c) The transparency of word-creations with respect to their linguistic input is reduced
as compared to regular formations, for most creative techniques shorten or modify
their linguistic input in a way that impedes its recoverability (Marchand 1969: 2- 3).
Even if all input elements are contained in full in the output, they are combined in
an unexpected manner, hampering the semantic interpretation of the output. Regular
word-formational operations, in contrast, are normally not designed in a way to make
their input difficult or impossible to recover. In fact, this would run counter to the
very purpose of normal linguistic utterances which is to make oneself understood.
d) Unlike with regular word-formation, the products of some important creative techniques such as shortening and alienation are not new words with respect to form and
meaning/part of speech, but with respect to their form only (Bauer 2003: 40). E.g.,
ad has the same denotation as advertisement. Only the stylistic value may be
changed, cf. below.
In contrast to this, the following arguments have been put forward against a principled
difference between regular word-formation and word-creation:
a) The phonetic outputs of creative techniques are most often regular (frequently even
preferred or minimal) sound patterns of the respective language, especially as far as
prosody is concerned (cf. section 4). Note, however, that this in no way implies the
regularity of the techn iques producing these patterns (see also article 21 on blending).
b) The products of word-creation are often integrated into the inflectional system of the
respective language (though with certain licenses, cf. Barz 2009: 736). Of course,
this argument also applies to the products, but not to the techniques of word-creation.
c) Like creative techniques, even regular operations such as suffixation can be outputdriven, for the resulting words have to fulfill certain prosodic requmrements (cf., e.g.,
Plag 2003: 116 with respect to English deadjectival abstract nouns in -ity).
d) Although a certain reduction of transparency, e.g., through clipping, cannot be denied,
the respective techniques are devised in such a way as to maximize the recoverability
of the source form by different constraints (for a discussion cf. Lappe 2007: 171205).
e) Even though creative techniques such as shortening do not change the denotation of
their source forms, they nevertheless provide a special stylistic value of sloppiness,
confidentiality or expressiveness (Zwicky and Pullum 1987: 335), which normally
qualify the creations for special pragmatic functions. Occasionally they may even
modify the denotation of their source forms (Lappe 2007: 20- 29).
Taking into account some or all of these arguments, several views on the relation between word-formation and word-creation have been advanced:
a) Both word-formation and word-creation are based on regular rules or models (cf.,
e.g., Soudek 1978, Plag 2003: 116 and the articles cited in section 4.1 ), the only
problem being that the rules or models of word-creation have not yet been completely
discovered and described. Many manuals of word-formation (including this handbook) treat at least some creative techniques on a par with regular word-formational
operations without or with only little discussion of the problem. E.g., in the standard
grammar of German, shortening is explicitly classified as regular word-formation
PA487
26. Word-creation
b)
c)
d)
e)
together with compounding, derivation and conversion (Barz 2009: 660), whereas
blending and other creative techniques follow as "other kinds of word-formation"
(Barz 2009: 669-673).
Neither word-formation nor word-creation is based on regular rules or models, but
both are directly founded on cognitive principles. This is assumed in the schemabased approach within cognitive grammar, which deliberately renounces any "'building-block' style theories of morphology" even for regular word-formation (Kemmer
2003: 69). This approach quite efficiently describes certain types of transparent
blends which crucially rely on a formal similarity between words. However, nowhere
nearly all creative techniques are of this kind.
Only productive word-formation is regular and therefore relevant to the linguistic
system proper. Word-creation, on the contrary, is a matter of performance in general
or at best of stylistics in particular. This is the view of "pure" structuralist morphology (cf., e.g., Marchand 1969: 441 with respect to clipping and 451 to blending) and
of generative morphology (e.g., Aronoff 1976: 20).
Only productive word-formation is completely regular, but word-creation is not completely irregular. Rather, it is a borderline phenomenon between regular morphology
and phonology or between regular word-formation and an (unspecified) kind of irregularity. This view is held by many morphologists, e.g., Gresillon (1984), Cannon
(1986).
Word-creation is a subtype of extragrammatical morphology, characterized by intentionality. This is the view of natural morphology (cf., e.g., Dressler 2000, 2005). In
this theory, "extragrammatical" is a cover term for different morphological operations
"outside of grammar'', governed entirely (at the first, pre- and protomorphological
stages of language acquisition) or in part (at later stages) by cognitive principles such
as iconicity or figure vs. ground. The fact that these principles remain available to
language users even after the acquisition of grammatical morphology explains why
similar creative techniques are applied and their products understood with some consistency in many different languages, although these techniques are not part of the
linguistic systems proper. (This effect is explained by a kind of metalinguistic competence for stylistic effects assumed by Sobkowiak 1991 ; cf. also section 4.)
2. Creative techniques
For a complete typology of creative techniques exemplified by product names see article
124 on word-formation and brand names and Ronneberger-Sibold (2010: 209). In the
following, only the structure of this typology will be sketched and the techniques not
covered by articles 19 on clipping, 21 on blending, and 25 on reduplication will be
described in some detail. Given the principled creative freedom of the techniques, this
description can only cover their basic features, such that every non-basic instance of
word-creation can be interpreted as a variation or combination of these features.
Moreover, it is important to note that the techniques are described from the viewpoint
of the creator only, for, as a consequence of the reduced transparency of word-creations,
the perspective of the receiver is not simply the converse of the creator's perspective.
For instance, the hearer or reader of an opaque word-creation such as Persil, name of a
487
PA488
488

washing powder, cannot decide whether this word was created by shortening, blending,
or some other technique. For the creator, on the contrary, these techniques exhibit important differences. (For the perspective of the receiver cf. Ronneberger-Sibold 2004:
590 ff.)
2.1. Word-creation without a linguistic input: phonetic symbolism

On the highest level of the proposed typology, a distinction is made between creative
techniques which are based on a morphological input and others which are not. The
latter produce free creations in the traditional sense mentioned in section 1 (German
Urschopfungen).
Free creations are, strictly speaking, not new words but new roots not assigned to
any particular part of speech, e.g., German husch, English woosh. The integration of
these roots into grammar as interjections (husch!), verbs (huschen 'to dart', huscheln 'to
do something sloppily'), nouns (das order Husch '(the) jiffy') or adjectives (huschelig
'sloppily') and possible conversions between these parts of speech concern regular derivational and inflectional morphology as well as syntax, but not the creative technique as
such (Paul 1920: 183 ). The same applies mutatis mutandis to semantic shifts frequently
accompanying grammatical integration, e.g., the metonymic use of a sound-iconic root
depicting a bird's call, such as German Uhu [u:.hu:] 'eagle owl' for the bird itself.
Most free creations are motivated by what is currently called "phonetic symbolism",
e.g., by Marchand (1969: 397) and Hinton, Nichols and Ohala (1994: 5), subdivided by
the latter into a) "imitative phonetic symbolism'', i.e. onomatopoeia imitating an auditory
sensation (e.g., bang, splash, miaow), b) "synaesthetic phonetic symbolism" imitating a
visual and/or haptic sensation (e.g., dump, plump), and c) "conventional phonetic symbolism", imitating the sound shape of other words. As with respect to a) and b) the use
of the term symbolism is at variance with Peircian semiotic tenninology (cf. Jakobson
and Waugh 1979: 178), they are here called sound-iconic, the third, conventional one
sound-symbolic.
Synaesthetic sound-iconic creations are based on (universal?) associations between
certain sound shapes or even individual sounds and sensory concepts such as 'bright' or
'dark', 'hard' or 'soft', 'strong ' or 'weak', 'big' or ' small', etc., e.g., between the vowel
Iii and 'small size' (Jespersen 1933). For an excellent overview of the problem and the
older literature cf. Jakobson and Waugh (1979: 177- 204), including milestones such as
Grammont (1901), Sapir (1929) and Jespersen (1933). An in-depth empirical psycholinguistic study is presented by Ertel (1969), a variety of languages are contrasted in
Fischer-10rgensen (1978) and Ultan (1978), cf. also O 'Boyle, Miller and Rahmani (1987)
with further references. For different psychological explanations of synaesthesia based
on auditory inputs cf. Tsur (2006).
Sound-iconic imitations of haptic sensations often depict movements, especially if
they are accompanied by sounds. Thus, German husch, English woosh imitate the sensation and sound of an airflow produced by a quick, furtive movement. In an even more
iconic variant of haptic sound-iconicity the speaker's mouth while pronouncing the
sound-iconic word imitates the depicted movement (Marchand 1969: 397). E.g., in pronouncing the name Maoam the speaker automatically imitates the movements of the
PA489
26. Word-creation
lower jaw when chewing this sweet. Moreover, the bilabial closure of the initial and
final bilabial resonant [m] imitates the movement of the lips during swallowing and
intensive tasting, thus indirectly even imitating a gustatory experience.
An important device for depicting rhythmically recurring sounds or movements or
both is sound-iconic reduplication, possibly accompanied by vowel alternations, especially between high and non-high vowels in so-called ablaut combinations, e.g., English
ding-dong rendering bell chimes and at the same time the swinging movement of the
bell, or English and French zigzag for a movement in opposite directions. As pointed
out by Marchand ( 1969: 43 6), the motivating factor in ablaut combinations is the polarity
of the vowels in an identical consonantal frame. E.g., in zigzag, the identity of the
movements is rendered by [z ... g], their opposite direction by the opposition between the
highest English short vowel [1] and its lowest counterpart [re]. Another variant with a
similar motivation are rhyme combinations such as English helter-skelter.
Of course, sound-iconic reduplication is a free creation only if it is not based on any
existing morpheme. Cases such as German Zickzack 'zigzag', synchronically and probably also etymologically based on Zacke 'sharp point' (Dressler 2000: 5), though not
according to a regular let alone a productive model, are better classified as a type of
extragrammatical derivation (cf. section 2.3). Moreover, reduplication as a creative technique has to be kept apart from regular derivational and inflectional reduplication found
in many languages (but cf. Malkiel 1977 on the difficulty of delimitation).
Sound-symbolic free creations (i.e., instances of "conventional phonetic symbolism"
according to Hinton, Nichols and Ohala 1994: 5), are motivated by a phonetic resemblance to the sound shapes characteristic of certain groups of words in the lexicon sharing certain semantic features. Two types of such lexical groups are to be distinguished:
The first group contains prototypical loanwords from a certain language or language
group sharing a number of phonetic and semantic (mostly connotative) features. E.g., in
German, comparatively recent, unadapted loanwords and proper names from Italian,
Spanish, or Portuguese such as Milano, Espresso, Polenta, Spaghetti, Torero, Veranda,
etc. are phonetically characterized by three or more syllables with full vowels, the last
of which ends in [-o ], [-a], or [-i], and by penultimate stress (Ronneberger-Sibold 2002).
As semantically most of these words belong (for Germans) in the context of Mediterranean holidays, elegance and enjoyment of life, the above-mentioned sound shape is capable
of evoking these associations even in completely meaningless free creations often used
as names for products designed for making life pleasant, elegant and easy, such as Zafira,
Meriva, Antara (names of car models made by Opel, artificially created by a computer
program, cf. Lotscher 2007).
In the second type of sound symbolic words serving as models for free creations, the
common phonetic features are not abstract sound shapes, but so-called phonaesthemes
(Firth 1957: 39: "phonaesthetic function"), also called submorphemes (Dressler 1990),
i.e., concrete sounds or sound clusters recun-ing in a number of words which share a
common semantic feature without, however, allowing for a morphological analysis. A
famous English example is the initial cluster [gl-] in words such as gleam, glimmer,
glisten, glitter, glow, gloss, etc. All of them share a semantic feature 'light, shine', but a
segmentation of a morpheme *[gl-] ' light, shine' is impossible, because the remaining
parts such as *-earn, -immer, -isten, etc. do not exist as separate morphemes (Marchand
1969: 411; Tournier 1985: 139). Free creations containing a phonaestheme [-on], suggesting a scientifically developed chemical substance are, e.g., artificially created Ger-
489
PA490
490

man names for synthetic fibers such as Dacron, Orlon, Antron, probably by an original
analogy with Nylon (Lotscher 1992: 84). (In an alternative analysis, -on would be considered as a suffix, in which case only the bases Daer-, etc. would be free creations, whereas
the suffixation by -on would be an instance of extragrammatical derivation, cf. section
2.3.)
Of course, free creation is not the only way in which sound-iconic or sound-symbolic
sound shapes may come into being. In many cases, they are created by other techniques,
relying on previously existing morphological material. E.g., in the German name Eduscho [e.' du:Jo] for a coffee brand the above-mentioned " Italian" sound shape has been
created by shortening Eduard Schop/, name of the fou nder of the firm. Moreover, they
may also be borrowed or result from diachronic change (Marchand 1969: 398; Meier
1975: 13- 17). In fact, the production of such sound shapes "on demand" is an important
motive for word-creation in cases where no borrowed or diachronically arisen word is
available (Ronneberger-Sibold 2001).
All free creations cited so far are motivated by their sound shape. In fact, contrary to
what is sometimes claimed (e.g., in Bu/3mann 2008: 800), it is extremely difficult to find
examples of totally unmotivated new free creations. (Incidentally, both examples given
in Bu/3mann are in fact based on morphological material.) Even artificially created
product names such as Meriva, etc., mentioned above, were selected because of their
sound-symbolic sound shape. Only creations in nursery rhymes such as eeny-meenyminy-moe are completely unmotivated, because they lack any meaning with respect to
which they could possibly be motivated (Zelinsky-Wibbelt 1983: 38).
Of course, the original motivation of free creations may be lost by diachronic change
(Jespersen 1922: 406). This seems to have occurred in many so-called expressive interjections such as ouch, ugh, etc. or orders such as whoa! Although they are today highly
conventionalized, they possibly originated in "natural" gestures and expressions of feelings (cf. Paul 1920: 179- 180 and, for a typology of interjections, Kleiber 2006).
2.2. Modifi cation of an existi ng source f orm : shortening and alienation

As for the creative techniques based on a linguistic input, the linguistic elements of this
input may either be combined for the first time by extra-grammatical word-formation in
order to create a word which is new regarding its meaning and form, or the input may
be a previously existing word or phrase, here termed source form, which is merely
modified with respect to its phonological and/or graphical shape, but not with respect to
its meaning (except for its stylistic register). Two important techniques of the latter type
are shortening and alienation. Shortening has been studied for various languages and in
different theoretical frameworks (see articles 19 on clipping and 124 on word-formation
and brand names). Contrary to this, alienation (see also article 130 on word-formation
and visuality) is only touched upon in investigations of different linguistic varieties such
as the language of advertising, journalism, the new media, etc., but, to my knowledge,
there is no typology, let alone a comprehensive study of this phenomenon. In the following, a typology will be sketched, based on Ronneberger-Sibold (2004) and RonnebergerSibold, Kazzazi and Potsch-Ringeisen (forthc.).
Alienation (German Verfremdung) is a cover term for different techniques reducing
the recoverability of a source form without, however, shortening it. Contrary to shorten-
PA491
26. Word-creation
ing, alienation aims less at simplifying the task of the speaker or writer than at surprising
and impressing the hearer or reader. The proposed typology distinguishes between three
types: orthographical alienation, orthographical and phonetic alienation, and phonetic
alienation.
In an orthographical alienation, only the orthography of the source form is modified,
leaving its pronunciation intact. (Note that purely graphical variation of graphemes within the same orthography is not covered by this definition.) There are two subtypes:
plain orthographical alienation and alienation by homophones. An example of a plain
orthographical alienation is the German brand name Vileda for an artificial chamois
leather from the German phrase wie Leder ' like leather ', spelt in the IPA system and
thus giving the impression of an Italian or Spanish loan word. This type of alienation is
quite common in bi- or multilingual and -scriptal speech communities where words of
one language may be written in the orthography or even the script of another language
for humorous or euphemistic purposes or for camouflage. In alienation by homonyms,
the source form is replaced by a previously existing or regularly formed new homophone
word. E.g., the German phrase (was) kostet das? '(How much) does this cost?' in colloquial pronunciation is replaced by the Greek name Kostas in a German advertisement
for Greek specialties by McDonalds (Wabner 2003: 113). Although the homophone word
(in this case Kostas) is not new with respect to its form, it is new with respect to its
meaning. Therefore, such cases may be considered as word-creations, provided that the
two meanings are indeed entirely different from each other, i.e., are neither synchronically in any metaphorical, metonymic, generalizing, etc. relation, nor combined in the creation like in a compound. (In the latter case the creation is better considered as a blend,
cf. below.)
Alienation by homophones may be combined with plain orthographical alienation in
rebus-like creations, where numbers or alphabet names of letters written as ciphers or
letters respectively replace homophone (parts of) words, thus yielding a graphical shortening at the same time. This shortening effect makes the technique especially attractive
for texting or advertising, e.g., in English CU for see you, U2 for you too or Car2Go
for car to go (Crystal 2008).
An orthographical and phonetic alienation may proceed from a modification of the
sound shape or of the spelling of the source form. The sound shape can be alienated by
extending it or by replacing part of it by elements without a morphological function, as
in German Schauma, name of a shampoo, from Schaum 'foam' and German Wella from
Welle 'wave', a brand name for hair cosmetics, respectively. Moreover, elements of a
sound shape such as syllables, may be mixed up, e.g., in French verlan from (a) l'envers,
name of a special jargon popular among young people, which consists in modifying the
order of syllables in a word (cf., e.g., Goudailler 2002). The orthographic alienation
precedes the phonetic one in anagrams, where the letters of the source form are "read
backwards" or mixed up, e.g., in Lycra, name of a fabric, from Acryl).
A mere phonetic alienation without a corresponding orthographical change is performed in word plays obscuring the morphological structure of the source form, e.g., by
a stress shift in German puns such as [blu'msntopfeud~] instead of [blu:m~ntopfeud~]
Blumentopferde ' flower-pot soil '. Although the output of the alienation sounds like a
compound with the word Pferde ' horses' as its second member, it is not designed to talk
about an obscure kind of horses, but simply to conceal the source form in a playful way,
by alienating its sound shape.
491
PA492
492
2.3. Extra-grammatical word-formation: blending and extra-grammatical

derivation
Contrary to shortening and alienation, the techniques subsumed here under the term
extra-grammatical word-formation do not modify a given source fonn, but combine
lexical and derivational morphemes into new words, thereby breaking the rules or models
of regular word-formation. If the combined morphemes are lexemes, the respective techniques are different kinds of blending; if lexical bases are combined with affixes or
combining forms, we call this extra-grammatical derivation. As blending is treated in
article 21 (see also article 124 on word-formation and brand names), the present article
focuses on extra-grammatical derivation.
In extra-grammatical derivation, regular derivatives are used outside their regular
functions. For instance, the German loan suffix -al is regularly used to form relational
denominal adjectives, (e.g., phanomenal from Phanomen), possibly nominalized, especially in professional jargons (e.g., Chloral for a chemical substance containing chlorine,
Gennan Chlor). However, in the name Verona! for a barbiturate, the suffixation by -al
neither produces an adj ective meaning ' related to Verona' nor a noun meaning 'a substance containing Verona', but has simply been chosen because it provides the name
with a sound shape characteristic of Latin and Greek loan words, associated by Germans
with scientific reliability. I.e. for the hearer or reader, -al here functions more as a
phonaestheme comparable to -on, mentioned above, than as a genuine suffix. This is
why such "derivations" are here considered as extra-grammatical creations.
A similar argument can be made for creations w ith combining forms (German Konfixe), such as therm 'warm ' or 'warmth' (Marchand 1969: 131 - 132, Dona lies 2000). In
regular use, they have full lexical meanings, but, unlike lexemes, they do not occur as
free forms. Moreover, most of them are fairly productive, like affixes. Their principal
domain is neoclassical word-formation, where they occur in initial or final position (e.g.,
Thermometer lit. 'measure of warmth', endotherm lit. ' inside warm ' ) or even as bases
(Thermik). In extra-grammatical use, this distribution is superficially imitated, but w ithout a clear structural and semantic function of the combining forms, especially if these
forms were themselves created through shortening, such as flex from flexible, tex from
textile or text, med from medicine or medical, mat(ic) from automat(ic), etc. (Cf. Thornton 2000; Ronneberger-Sibold 2009 for a delimitation between regular and extra-grammatical use; Schmitt 1996 on the international use of such elements; Fradin 2000 on the
process of their creation.) Examples are brand names such as Blend-a-med for a tooth paste, combining the ideas of brilliance (German blendend) and medicine, but not according to the regular models of German or neoclassical word-fonnation.
As mentioned in section 2.1, even reduplication, possibly accompanied by "Ablaut'',
might be considered as an instance of extra-grammatical derivation, if it is based on
existing morphemes but not derived by regular rules or models (e.g., German Zickzack).
3. Functions of word-creation
The most basic function of word-creation is an extension of the lexicon by words exhibiting two characteristics which, normally, may result from diachronic change or borrow-
PA494
494

and professional jargons requiring short sound shapes which are at the same time
maximally distinctive and easy for production and perception (Kobler-Trill 1994;
Ronne berger-Si bold 1996; Steinhauer 2000; Balnat 2011 ),
b) a gain in prestige through the imitation of typical sound shapes of prestige languages
(cf. section 2.1).
A semiotic purpose of specific sound shapes is the imitation of sensory impressions by
phonetic iconism, used in literary texts in the widest sense including literature for children
and comics (e.g., Kayser 1932; Nodier 2008: 15- 20 and passim; Grunewald 2000: 14).
4. Word -creation in a cross- linguistic perspective

If word-creation is not entirely governed by the word-formational rules or models of
individual languages, but, at least in part, based on universal cognitive principles (cf.
section 1), a cross-linguistic survey should be of particular interest for research on linguistic universals and typology, because it allows for the study of the interaction between
universal tendencies and language-specific features more directly than the grammars of
individual languages. In the following, some examples of this interaction will be discussed separately for the creative techniques and for their products.
4.1. Products of wo rd-creation

Research on the sound shapes of word-creations, especially of clippings and blends, has
to a great deal focussed either on their congruence with the minimal or at least the
preferred prosodic patterns of the language in question, thus emphasizing the languagespecific character of word -creations (e.g., Monnot 1971 for English and French; Plenat
1993 for French; Wiese 1996 and Fery 1997 for German; Bat-El 2000 for Hebrew), or,
on the contrary, on the emergence of universally unmarked structures in word-creations,
thus stressing their universal properties (e.g., Kubozono 1990 for Enghsh and Japanese;
Kilani-Schoch 1996 for French; Montermini 2007 for Russian). A median position is
taken in Ronneberger-Sibold (1996) on the basis of a detailed comparative study of
Gennan and French shortenings. It seems that word-creation serves, among other things,
to re-establish an optimal balance between different universal tendencies in cases where
this balance has been upset by diachronic change in the normal vocabulary of a language.
E.g., in the history of French, repeated deletion of final consonants due to a continuous
pursuit of the universal preference for open syllables has yielded numerous homophones,
thereby coming into conflict with the universal need for distinctiveness. In the hi storical
development of German phonology, on the contrary, final consonants were far more
stable. Therefore, somewhat perversely at first sight, the preference for open syllables is
stronger in German than in French shortenings (cf. also Nlibling 200 I on German and
Swedish).
A field where language-specific phonology interacts with universally identical patterns is onomatopoeia. Although the im itated sounds, e .g., certain birdcalls (Callebaut
1985), are universally (almost) identical, their linguistic imitation has to take into ac -
PA496
496
5. References
Aronoff, Mark
1976
Word-Formation in Generative Grammar. Cambridge, MA: MIT Press.
Balnat, Vincent
2011
Kurzwortbildung im Gegenwartsdeutschen. H ildesheim: Olms.
Barz, Irmhild
2009
Die Wortbildung. In: Dudenredaktion (ed .), Duden. Die Grammatik, 634- 762. 9111 ed.
Mannheim: Dudenverlag.
Bat-El, Outi
2000
The grammaticality of "extragrammatical" morphology. In: Ursula Doleschal and Anna
M. Thornton (eds.), Extragrammatical and Marginal Morphology, 61-84. Mtinchen:
LINCOM Europa.
Bauer, Laurie
2003
Introducing Linguistic Morphology. 2"d ed. Edinburgh: Edinburgh University Press.
Bellmann, Gunther
1980
Zur Variation im Lexikon: Kurzwort und Original. Wirkendes Wort 30: 369- 383.
Bul3mann, Hadumod (ed.)
2008
Lexikon der Sprachwissenschaft. 4th ed. Stuttgart: Kroner.
Callebaut, Bruno
1985
Onomatopees et noms d'oiseaux en franc;:ais. Le Fram; ais Moderne 53(1- 2): 49- 77.
Cannon, Garland
1986
Blends in English word-formation. Linguistics 24: 725-753.
Crystal, David
2008
Tx tng. The Gr8 Db8. Oxford: Oxford University Press.
Donalies, E lke
2000
Das Konfix. Zur Definition einer zentralen Einheit der deutschen Wortbildung. D eutsche
Sprache 28: 144- 159. [Reprinted in: Muller, Peter 0. (ed.) 2005 Fremdwortbildung.
Th eorie und Praxis in Geschichte und Gegenwart, 179- 198. Frankfurt/M.: Lang.]
1990
Sketching submorphemes within natural morphology. In: Julian Mendez Dosuna and
Carmen Pensado (eds.), Naturalists at Krems, 33-41. Salamanca: Ediciones Universidad
de Salamanca.
2000
Extragrammatical vs. marginal morphology. In: Ursula Doleschal and Anna M. Thornton
(eds.), Extragrammatical and Marginal Morphology, 1-10. Miinchen: LINCOM Europa.
Dressler, Wolfgang U .
2005
Word-formation in natural morphology. In: Pavol Stekauer and Rochelle L ieber (eds.),
Ertel, Suitbert
1969
Psychophonetik. Gottingen: Hogrefe.
Fery, Caroline
1997
Unis und Studis: Die besten Worter des Deutschen. Linguistische B erichte 172: 461 89.
F irth, John R.
1957
The use and d istribution of certain English sounds. In: John R. Firth , Papers in Linguistics 1934- 1951 , 34- 46. Oxford: Oxford University Press.
Fischer-JeJrgensen, Eli
1978
On the universal character of phonetic symbolism with spec ial reference to vowels.
Studia Linguistica 32: 80- 90.
Fleischer, Wolfgang and Irmhild Ba rz
20 12
PA497
26. Word-creation
Forgacs, Erzsebet and Agnes Gondocs
1997
Sprachspiele in der Werbung. Studia Germanica Universitatis Vesprimiensis 1: 49-70.
Fradin, Bernard
2000
Combining forms, blends and related phenomena. In: Ursula D oleschal and Anna M.
Thornton (eds.), Extragrammatical and Marginal Morphology, 11- 59. Miinchen: LTNCOM Europa.
Goudaillier, Jean-Pierre
2002
De !'argot traditionnel au fran~ais contemporain des cites. Linguistique 38(1): 5-23.
Grammont, Maurice
1901
Onomatopees et mots expressifs. Revue des langues romanes 44(5 eme serie, tome 4eme) :
97-158. [Reprint Nendeln: Kraus 1970.]
Gresillon, A lrnuth
La regle et le monstre. Le mot-valise. Interrogations sur la langue, a partir d'un corpus
1984
de Heinrich Heine. Ti.ibingen: Niemeyer.
Grunewald, D ietrich
2000
Comics. Tubingen: Niemeyer.
Hinton, Leanne, Johanna Nichols and John J . Ohala (eds.)
1994
Sound Sy mbolism. Cambridge: Cambridge University Press.
Jakobson, Roman and Linda R. Waugh
1979
The Sound Shape of Language. B loomington/London: Bloomington University Press.
Jespersen, Otto
1922
Language. Its Nature, Development and Origin. London: Allen & Unwin.
Jespersen, Otto
1933
Symbolic value of the vowel I. In: Otto Jespersen, Linguistica. Selected Papers in English, Fren ch and German, 283-303. Copenhagen: Levin & Munksgaard, London: Allen & U nw in.
Kayser, Wolfgang
1932
Die Klangmalerei bei Harsdorjfer. Ein Beitrag zur Geschichte der Literatur, Poetik und
Sprachtheorie der Barockzeit. Leipzig: Mayer & Muller.
Kemmer, Suzanne
2003
Schemas and lexical blends: Studies in honor of Gunter Radden. In: Hubert Cuyckens,
Thomas Berg, Rene D irven and K laus-Uwe Panther (eds.), Motivation in Language,
69-97. Amsterdam/Philadelphia: Benjamins.
1996
Syllable and foot in French clipping. In: Bernhard Hurch and Richard Rhodes (eds.), Natural Phonology. The State ofthe Art, 26 1- 292. Berlin/New York: Mouton de Gruyter.
Kleiber, Georges
2006
Semiotique de !' interjection. Langages 161: 10- 23.
Kobler-Trill, Dorothea
1994
Das Kurzwort im Deutschen. Eine Untersuchung zu Definition, Typologie und EntwickLung. Tubingen: Niemeyer.
Kubozono, Haruo
1990
Phono logical constraints on blending in English as a case for phonology-morphology
interface. In: Geert Booij and Jaap van Marie (eds.), Yearbook of Morphology 1990, 120. Dordrecht: Faris.
Lappe, Sabine
2007
English Prosodic Morphology. Berlin: Springer.
Lotscher, Andreas
Von Ajax bis Xerox. Ein Lexikon der Produktenamen. 2nd ed. Zurich: Artemis & Winkler.
1992
Letscher, Andreas
2007
Von Mercedes zu Xsara - Modellnamen fiir Automobile in geografisch-historischer
Perspektive. In: Ludger Kremer and Elke Ronneberger-S ibold (eds.), Names in Commerce and Industry. Past and Present, 115-129 . Berlin: Logos.
497
PA498
498

Lopez-Rua, Paula
2005
Shortening devices in text messaging: A multilingual approach. Neuphilologische Mitteilungen 106: 139- 155.
Malkiel, Yakov
1977
From phonosymbolism to morphosymbolism. In: Michel Paradis (ed.), The Fourth Lacus Forum, 511- 529. Columbia, SC: Hornbeam Press. [R eprinted in: Yakov M al kie l
1990: Diachronic problems in Phonosymbolism . Edita and lnedita, 1979- 1988. Vo l. l ,
157- 175. Amsterdam/Philadelphia: Benjamins.]
Marchand, Hans
1969
The Categories and Types of Present-Day English Word-Formation .. A Synchronic-Diachronic Approach. 2nd ed. Miinchen: Beck.
Meid, Wolfgang
1977
Beziehungen zwischen iiuI3erer und innerer Sprachform: Verschriinkte Zeichen und fusionierte l nhalte. Veroffentlichungen der K ommission far Linguistik und Kommunikationsforschung. Anzeiger der philosophisch-historischen Klasse 114( 6): 294- 304. Wien:
Verlag der Osteneichischen Akademie der Wissenschaften.
Meier, Harri
1975
Primare und sekundi:ire Onomatopoien und andere Untersuchungen zur romanischen
Etymologie. Heidelberg: Winter.
Monnot, Michel
1971
Examen comparatif des tendances de syllabation dans les mots abreges de l'anglais et
du francyais. Le Fran<;ais Moderne 39: 191- 206.
Montermini, Fabio
2007
Hypocoristiques et minimalite en russe. In: Elisabeth Delais-Roussarie and Laurence
Labrune (eds.), Des sons et des sens, 198-2 13. Paris: Lavoisier/Hermes.
N odier, Charles
2008
Dictionnaire raisonne des onomatopees franr;aises. Edition etablie, presentee et annotee
par Jean-Frarn;ois Jeandillou . Geneve/Paris: Droz [Original edition 1808].
Niibling, Damaris
2000
Auf der Suche nach dem idealen Eigennamen. Beitri:ige zur Namenforschung 35: 275302.
Niibling, Damaris
200 1 Auto - bi!, Reha - rehab, Mikro - mick, Alki - alk is: Kurzworter im Deutschen und
im Schwedischen. Skandinavistik 31: 167- 199.
Niibling, Damaris, Fabian Fahlbusch and Rita Heuser
2012
Namen. Eine Einfahrung in die Onomastik. Tiibingen: Narr.
O ' Boyle, Michael W., David A. Miller and Fahim R ahmani
1987
Sound-meaning relationships in speakers of Urdu and E nglish: Evidence for a crosscultural phonetic symbolism. Journal of Psycholinguistic Research 16(3): 273 - 288.
Paul, Hermann
1920
Pr.inzipien der Sprachgeschichte. 5111 ed. Tiibingen: Niemeyer.
Plag, Ingo
2003
Plenat, Marc
1993
Observations sur le mot minimal francya is: L'oralisation des sigles. In: Be rnard Laks and
Marc Plenat (eds .), De natura sonorum. Essais de phonologie, 143- 172. Saint-Den is:
Presses Universitaires de Vincennes.
R onneberger-Sibold, E lke
1996
Preferred sound shapes of new roots: O n some phonotactic and prosodic properties of
shortenings in German and French. In: Bernhard Hurch and Richard Rhodes (eds.) ,
Natural Phonology. The State of the Art, 261- 292. Berlin/New York: Mouton de Gruyter.
PA499
26. Word-creation
2000
Creative competence at work: The creation of partial motivation in German trade names.
In: Ursula Doleschal and Anna M. Thornton (eds.), Extragrammatical and Marginal
Morphology, 87-105. Munchen: LINCOM Europa.
2001
On useful darkness: Loss and destruction of transparency by linguistic change, borrowing, and word creation. In: Geert Booij and Jaap van Marie (eds.), Yearbook ofMorphology 1999, 97- 120. Dordrecht: Kluwer.
2002
On the phonostylistic function of prototypical nonnative sound shapes in contemporary
German: Evidence from the history of brand names. In: Katarzyna Dziubalska-Kolaczyk
and Jaroslaw Weckwerth (eds.), Future Challenges for Natural Linguistics, 211 - 231.
Milnchen: LINCOM Europa.
Ronneberger-S i bo Id, E Ike
2004
Warennamen. In: Andrea Brendler and Silvio Brendler (eds.), Namenarten und ihre
Erforschung. Ein Lehrbuch fur das Studium der Onomastik, 557-603. Hamburg: Baar.
2009
Thermodur, Blend-a-med, Sivitrex: Konfixe in deutschen Markennamen: Typen - Geschichte - Funktionen. In: Peter 0. Muller (ed.), Studien zur Fremdwortbildung, 141 193. Hildesheim: Olms.
2010
Word creatio n: Definition - function - typology. In: Franz Rainer, Wolfgang U. Dressler,
Dieter Kastovsky and Hans Christian Luschi.itzky (eds.), Variation and Change in Mor-
phology. Selected Papers from the 13 111 International Morphology Meeting (Vienna, February 2008), 20 1-216. Amsterdam/Philadelphia: Benjamins.
2012
Blending between grammar and universal cognitive principles: Evidence from German,
Farsi, and Chinese. In: Vincent Renner, Franyois Maniez and Pierre J. L. Arnaud (eds.),
Cross-Disciplinary Perspectives on Lexical Blending, 115- 143. Berlin/Boston: Mouton
de Gruyter.
2014
Tuning morphosemantic transparency by shortening: A cross-linguistic perspective. In:
Franz Rainer, Francesco Gardani, Hans Christian Luschi.itzky and Wolfgang U . Dressler
(eds.), M otphology and Meaning. Selected Papers from the 15111 International Morphology Meeting, Vienna, February 2012, 275- 287. Amsterdam/Philadelphia: Benjamins.
Ronneberger-Sibold, Elke, Mir Kamaladin Kazzazi and Stefanie Potsch-Ringeisen
forthc. Intendierte morphologische Irregularitiit im Sprachvergleich: Schriftbasierte Wortschopfung im Deutschen, Farsi und Chinesischen. In: Andreas Bittner and Klaus-Michael
Kopcke (eds.), Regularitat und Irregularitat. Berlin/Boston: Akademie Verlag.
Sapir, Edward
1929
A study in phonetic symbolism. Journal of Experimental Psychology 12: 225-259. [Reprinted 1949 in: David G. Mandelbaum (ed.), Writings of Edward Sapir in Language,
Culture and Personality, 61 - 72. Berkeley, CA: University of California Press.]
Schmitt, Christian
1996
Euromorphologie: Perspektiven einer neuen romanistischen Teildisziplin. In: Wolfgang
Dahmen, Gunter Holtus, Johannes Kramer, Michael Metzeltin, Wolfgang Schweickard
and Otto Winkelmann (eds.) , Die Bedeutung der romanischen Sprachen im Europa der
Zukunft, 119- 146. Ti.ibingen: Narr.
Sei ler, Hansjakob
1975
Die Prinz ipien der deskriptiven und der etikettierenden Benennung. In: Hansjakob Seiler
(ed.) , Linguistic Workshop III, 2-57. Munchen: Fink.
499
PA500
500

Sobkowiak, Wfodzimierz
Metaphonology of English Paronomasic Puns . Frankfurt/M.: Lang.
1991
Soudek, Lev t
1978
The relation of blending to English word-formatio n: Theory, structure, and typological attempts. In: Wolfgang U. Dressler and Wolfgang Meid (eds.), Proceedings of the Twelfth
International Congress ofLinguists, 462-466. Innsbruck: Institut fur Sprachwissenschaft.
Steinhauer, Anja
2000
Sprachokonomie durch Kurzworter. Bi/dung und Verwendung in der Fachkommunikation. Tiibingen: Narr.
Thornton, Anna M.
2000
On -ex and -tex. In: Ursula Doleschal and Anna M. Thornton (eds.), Extragrammatical
and Marginal Morphology, 107-126. Miinchen: LINCOM Europa.
Tournier, Jean
1985
Introduction descriptive a la lexicogenetique de I' anglais contemporain. Paris/Geneve:
Champion & Slatkine.
Trubetzkoy, N[ikolaj] S[ergeevic]
1977
Grundziige der Phonologie. 61h ed. Gottingen: Vandenhoeck & Ruprecht.
Tsur, Reuven
2006
S ize-sound symbolism revisited. Journal of Pragmatics 38(6): 905- 924.
Ultan, Russell
1978
S ize-sound symbolism. In: Joseph H. Greenberg (ed.), Universals of Human Language.
Vol. 2 : Phonology, 525-568. Stanford, CA: Stanford University Press.
Wahner, Matthias
2003
Kreativer Umgang mil Sprache in der Werbung. Eine Analyse der Anzeigen- und Plakatwerbung von McDonald's. Networx, Nr. 32. <http://www.mediensprache.net/networx/
networx-32.pdt>.
Wiese, Richard
1996
Th e Phonology of German. Oxford: Oxford University Press.
Zelinsky-Wibbelt, Cornelia
1983
Die semantische Belastung von submorphematischen Einheiten im Englischen. Eine empirisch-strukturelle Untersuchung. Frankfurt/M .: Lang.
Zwicky, Arnold M. and Geoffrey K. Pullum
1987
Plain morphology and expressive morphology. In: Jon Aske (ed.), Proceedings of the
Thirteenth Annual Meeting (14- 16 February 1987). General Session and Parasession
on Grammar and Cognition, 330- 340. Berkeley, CA: Berkeley Linguistics Society.
Elke Ronneberger-Sibold, Eichstatt (Germany)
27. Allomorphy
l.
2.
3.
4.
5.
6.
7.
Introduction
General overview
Conditions of allomorphy
Exclusion of zero allomorphs
Word-creation
Conclusion
References
PA501
27. Allomorphy
Abstract
Allomorphy, which violates biuniqueness and canonical word-formation, has to be restricted. This is true more for word-formation than for inflection but is otherwise rather
similar in stem vs. affix allomorphy and in types of conditioning. Allomorphy is rare in
most types of extragrammatic word-creation.
1. Int roduction
Allomorph (of a morpheme) and allom01phy are terms created by Nida (1948: 420) in
analogy to the terms allophone and allophony with the primary purpose of conceptualising stem variation and affix variation in inflectional morphology (cf. Carstairs-McCarthy
2005). Allomorphs thus form a set of formal variants of the same morpheme but must
have an identical content in terms of syntactic, semantic and (I must add) pragmatic
features, which is more precise than Bloomfield's (1933) principle of identical function
and complementary distribution. Thus the German adj ective-forming suffixes -ig, -Lich,
-isch (cognate to E. -y, -ly, -ish) are not only formally similar but also have the same
syntactic and semantic content as heads of derived adjectives, except for the fact that
-isch adjectives may be pejorative, as in weib-lich 'woman-ly' vs. pejorative weib-isch,
kind-lich 'child-like' vs. kind-isch 'child-ish', and inherently pejorative adjectives have
also adopted the pejorative suffix -isch in their diachronic development, as in former
neid-ig > neid-isch 'envious'. Because of these differences in meaning, the three similar
German suffixes are not to be considered allomorphs of the same morpheme but rather
competing and functionally overlapping morphemes. For the same reasons, the Spanish
adjective-forming suffixes -ado, -udo, -ido cannot be classified as allomorphs (pace Malkiel 1966: 335 f.). But when semantic differences occur only rarely, then we are in a
transition zone between allomorphy and morphematic contrast, as in Stekauer, Valera
and Kortvelyessy's (2012: 137-138) Slovak instances of contrast between the prefix
forms pre- and prie- in predictable derivations, e.g., pre-beh-nu-t 'to run (perfective)'
---+ verbal noun prie-beh 'course (of events)' vs. the unpredictable contrast between deverbal pre-chod 'transition' and prie-chod 'gangway'.
Content identity holds only for word-formation meanings (Marchand 1969; cf. also
Corbin's 1987 concept of "sens construit" ), usually not for word meanings. Thus the
word-formation meaning of E. (a) build-ing is 'something which has been/is being built',
where the meaning of build- is identical with the lexical meaning of the verb build,
whereas the word meaning of (a) building has idiosyncratic meaning restrictions in respect to the verbal base. As a consequence not everything that has been or is being built
can be called a building. And such deviation of word meaning from its word-formation
meaning holds for all result nouns, and, at least to a certain degree, for all derivations
in general. In other words, the word-formation meaning of a complex/derived word is
fu lly compositional, whereas the word meaning is never fully compositional (cf. Frege
1892).
Another criterion of allomorphy advanced by many (e.g., Thornton 2005: 40; cf. also
Mugdan 1994: 2547) is complementary distribution. But since allophony is not necessarily in complementary distribution, allowing also for free allophones (e.g., several rhotic
50 1
PA502
502

variants of the same /r/ phoneme in the same context), it is not self-evident why complementary distribution should be obligatory for allomorphs. English examples of free variation between allomorphs are dependence and dependency, expectance and expectancy
(Plag 2003: 87; further examples in Stekauer, Valera and Kortvelyessy 2012: 193). On
the other ha nd, the presence of complementary distribution is an argument for allomorphy, such as Plag's (2003: 29) example of English adj ectivizing -al vs. -ar (after basefinal -l) in caus-al vs. pol-ar.
What remains, in any case, as a distinctive criterion for allomorphy is non-contrastive
distribution in meaning, a consequence of the above-mentioned identity-of-content constraint, as in dependence - dependency.
So far we have treated allomorphy in terms of alternation in representation. But
process models, such as some generative word -fo rmation models treat it, at least partially, in terms of adjustment rules, including allomorphy rules. For example, Aronoff (1976
and later; cf. also Stekauer 2000: 146) derives complex forms such as electri-jic-ation
from electri-fy with an allomorphy rule which replaces the suffix -fy with its a llomorph
-fie- (cf. Scalise 1984: 57-69).
For quite some time, allomorphy in word-formation has not been the only or the main
topic of any single publication, but it has been dealt with together with other basic
concepts of morphology in all introductory books and handbooks of morphology (albeit
with a focus on inflectional morphology). This holds also for the most formal classification of allomorphy in Mel'cuk 1993- 2000 Vol. 4: 214- 228).
2.
Genera l overview
2.1. Generalities
This article focuses on word-formation as part of grammatical morphology; operations
of extragrammatical morphology (cf Dressler 2001) will be mentioned only marginally,
such as abbreviations, blends or sophisticated terminography, as in the truncated base of
laps+o+logy 'error analysis' from lapsus.
Allomorphy violates the general preference for biunique relationships between meaning and form (Dressler 2005: 274, 1987: 111), i.e. between the meaning and the form of
a morpheme. Since biuniqueness is difficult to attain in morphology and since it plays a
greater role in (especially scientific and juridical) terminology, allomorphy should be
less frequent in sophisticated, conscious terminology than in ordinary fanguage . But this
is hardly the case, because the different types of conditioning of allomorphy (studied in
section 3) can be hardly avoided even in very sophisticated terminology.
In Corbett's approach to canonical typology, allomorphy inherently deviates from
canonical derivational morphology (implicit in Corbett 20 10). Allomorphy also provides
a problem for learnability, insofar as it presents a problem for young children, which so
far has only been investigated systematically for inflectional morphology, cf. Bernhardt
and Sternberger (1998: 658 ff.), Peters (1997: 169 ff.). But in the acquisition of diminutives and hypocoristics (Savickiene and Dressler 2007), at least Russian , Italian, Spanish
and Hungarian children do not seem to have particular problems acquiring competing
suffixations with stem allomorphy, nor do Flemish children with Dutch suffix allomor-
PA505
27. Allomorphy
mentioned E. -al, -ar, to interfixes, as in G. Garage-n-besitzer 'proprietor of a garage' vs.
Schwan-en-hals 'swan's neck', or to inflectional (but apparently not to truly derivational)
infixes.
Stem allomorphy refers to variants of the stem, which is prototypically the lexical
base, affix allomorphy to affix variants, as in the above examples. This d ifference is,
however, blurred in opaque cases of multiple affixation as in E. person-ific-ation, where
the verb person-ify is the base and thus the complex stem of person-ific-ation. Derivational morphology has both types of allomorphy, whereas compounding has only stem
allomorphy, if at all: for the first member of compounds note Fuhrhop's (1998) analysis
of the first part of G. Garagen-platz 'garage place (of a car)' as stem allomorph of the
noun Garage. But an alternative, traditional analysis is the assumption of an interfix or
linking element In/ inserted between the two compound members (see ar ticle 31 on
interfixes in Romance). An undisputed case of stem allomorphy in compounding is constituted by the following change from -a to -i in earlier Chicago Lithuanian, which is
even applied to English loan words, e.g., masc.nom.sg. kara-s (< E. car) ~ street-kari-s
'street-car, tram'. The same stem change, inherited from Proto-Inda-European, occurs in
Latin derivatives of the type in-ermi-s 'unarmed', from arma 'arms', genitive armorum.
In Latin and many other Inda-European (incl. modem Romance and Slavic) languages, stem allomorphy has been assumed (predominately since Aronoff 1994), not
only for inflectional but also derivational deverbal morphology, starting with the participial stem as an allomorph of the present and perfect stems, such as in Latin action and
agent nouns: ora-re 'to speak out, pray', perf. oravi, past participle oratus ~ orat-io
'speech', orat-or 'speaker', auge-re 'to increase', perf. aux-i ' I increased ', past participle
auct-us ~ auct-io 'auction', leg-e-re ' to read ', perf. leg-i, past participle Leet-us ~ lectio 'reading', feet-or 'reader'. Aronoff ( 1994) calls this participial stem allomorph of
derivational morphology "morphomic", i.e. a stem variant implying no semantic change,
whereas an older, alternative name is "parasitic", i.e. morphosernantically opaque derivation.
We can speak of stem allomorphy in the strict sense, if a single stem change is not
the only morphological change involved. Thus we can establish English stem allomorphy
in bread+th vs. broad, drunk+ard vs. drink, divin +ity vs. divine (with vowel change in
the second syllable). Here the addition of a suffix is the primary signal of der ivation,
stem vowel change the secondary signal, which can therefore be classified as stem allomorphy in the strict sense. In contrast, the German partially synonymic deverbal nouns
Trank, Trunk 'drink, potion' from trink+en 'to drink' do not fulfil this criterion and thus
can be classified only as allomorphs in a loose sense, whereas we can identify stem
allomorphy in the strict sense in the partial synonym Ge+triink ' beverage' .
All cases of a llomorphic changes discussed so far, have been cases of segmental
allornorphy. But prosodic changes may accompany or even determine segmental changes
in allornorphy (cf. Drachrnan, Kager and Malikouti-Drachman 1995, and article 11 on
word-formation in optimality theory, in particular the sections on output constraints).
For example, the a djective manager-ial derived from the noun manager shows two allornorphic stem changes: segmental change in the vowel of the last stem syllable plus
stress shift from the first to the last stern syllable (cf. Kaisse 2005: 32). A prosodically
automatic constraint also affects derivational suffixes in Slovak, due to its often discussed rhythmic law, which prohibits a sequence of long vowels in adj acent syllables at
the word level, as in the allornorphic contrast between the long basic variant -ik and the
505
PA506
506

shortened variant -ik, e.g., recnik ' speak-er' vs. basn-ik ' poet' (Stekauer, Valera and
Kortvelyessy 2012: 189-190).
Prosodic changes occur in ancient Indo-European languages together with the stem
allomorphy of ablaut (apophony) which can change vowel duration and vowel quality.
For example, in Ancient Greek (classical Attic dialect) the word for mother shows the
following five allomorphs /me:te:r/ (nom.sg. with rising acute pitch-accent on the first
syllable) vs. /me:ter/ in the acc.sg. /me:ter+a/ (with accent shift to the second syllable)
vs. /meter/ (voc.sg. with rising-falling circumflex) vs. /me:tr/ in several oblique cases,
e.g., gen.sg. /me:tr+6s/ (with accent shift and deletion of the second stem vowel) vs.
adjective derivations /a+me:to:r/ 'without a mother ' and /dys+me:to:r/ 'of a bad mother '
(with final vowel change).
This Ancient Greek example illustrates the fact that, at least in ancient Indo-European
languages, stem allomorphy is much richer in inflection than in word-formation. The
opposite is true in Semitic morphology. Note the Standard Arabic stem allornorphy of
the root /ktb/ 'to write' in derivational morphology: kiitab-a 'corresponded ', a-ktab-a
'dictated', kith-a 'book's transcription', kitiib 'book', kutub-i 'book-seller', katib-a 'written document, diploma', kiitib 'writer', kuttiib 'writing school', mu-kattib 'master of
writing', ma-ktab ' letter' vs. the inflectional allomorphs of the verb, as they appear in
katab-a 'he wrote', imperfect ya-ktub(u), and either inflectional or dlerivational in the
participle katib (also agent noun, see above) or in the causative kattab-a. Thus the relative amount of stem allomorphy is a typological criterion.
Due to the much greater morphological richness in English of Latinate derivational
morphology than of inflectional morphology, English also has, in absolute numbers,
more stem a llomorphy in derivation than in inflection, i.e. in strong verbs, as in keep kept+t, rise - ris-en (and rose as a stem allomorph in the loose sense, because vowe l
change is here the only signal of the category of preterit). In derivational morphology,
we have many more instances of stem allomorphy, such as in deep - dep+th, divine divin +ity, manager - manager+ial, porous - poros+ity, etc. Prosodic change in terms
of stress shift, as in export ~ to export are instances of stem allomorphy in the loose
sense, because it is the only signal of word-formation.
2.3. Primary vs. secondary allomorphs

Among both stem and affix allornorphs we can distinguish between primary vs. secondary allomorphs. Primary (also called basic or principal) allomorphs are the basic or
major variants in allomorphy (Mel'cuk 1993- 2000 Vol. 4: 217- 218), most clearly when
they represent the default in derivation, i.e. the least restricted variant (cf. Scalise 1984:
60). Instances in affix allomorphy are E. -able as primary vs. -ible as secondary or minor
allomorph, a nalogically adjectival -al vs. -ar (cf. sections I and 3.2), cf. G. +heit in
Frei+heit 'free+dom' vs. its secondary allomorphs in Bitter+keit 'bitter+ness' and
Neu+igkeit 'new+ness' . In stem allomorphy, however, the essential criterion for the dichotomy between primary and secondary allomorphs is their occurrence in underived
bases vs. derived stems. Thus E. electric with word-final /kl is considered as primary
allomorph vs. secondary stem allomorphs in electric+ity [elek'trisiti], electric+ian
[e lek'triJ~n]. because these secondary allomorphs occur only in affixed forms. The occur-
PA507
27. Allomorphy
rence of the primary allomorph in other derivations, such as electric+al, is of minor
importance for justifying its status as primary. Obviously often both occurrence in the
base of derivations and frequency of occurrence in derivations may converge, as in the
case of the primary stem allomorph in the base elect and in the derivations elect+or,
elect+ress, elect+ive vs. the secondary allomorph in the fusional derivation elect+ion
[e' lekJ~n].
3.
Conditions of allomorphy
3.1. Phonological conditioni ng

Phonological conditioning of either affix or stem allomorphy entails, first of all, the
application of phonological rules (Booij 2000: 336 ff.; Dressler 1985a: 4 1 ff.; Kaisse
2005; cf. also article 50 on phonological restrictions on English word-formation). First,
their optional application in casual speech must be mentioned, but need not be discussed
in detail. For example, in casual speech, final /n/ of the first member of a compound can
be assimilated to the initial obstruent of the second member, as in E. sun-burn, raincoat. Or the final nasal of the English (and German, etc.) negative prefix un- can be
assimilated in place of articulation to a following obstruent, phonemically (i ..e. neutralizing) before labials, as in E. un-pleasant, un-believable, un-married, allophonmcally before
labiodentals, as in un-.faithful, and velars, as in un-committed. Also the Latinate negative
prefix in- (cf. section 3.5) shows the same optional allophonic assimilation to word-initial
velars (e.g., in-credible) and labiodentals (e.g., in-frequent). Before labials assimilation
is obligatory (e.g., im-possible), and it depends on the respective phonological approach
whether this conditioning is classified as phonological or morpho(pho )nological (see
section 3.2). In any event such cases should not be simply collapsed (as in Stekauer,
Valera and Kortvelyessy 2012: 192).
Obligatory phonological conditioning can be illustrated with Bauer's (1983: 126; cf.
also Plag 2003: 28) examples of stress-change-induced stem allomorphy between diphthongs and schwa and between tense and lax vowels in photograph - photograph-y,
cremate - cremat-orium, regal - regal-ia. Of course, such alternations occur only in
word-formation. For quantitative conditioning of allomorphs in Slovak cf. section 2.2.
Phonological conditioning of allomorphy is often used, when there are also phonological conditions for the application of morphonological rules or even for suppletion, as in
the allomorphy of the English indefinite article a/an (Plag 2003: 27; cf. Iacobini 2000:
874, also article 11 on optimality theory in the sections that focus on phonological,
especially prosod ic output constraints), and sections 3 .2, 3.4, and 3.5.
3.2. Morpho(pho)nological cond iti oning

Morphophonological conditioning of allomorphy can be defined as a phonological
change which occurs only in certain morphological alternations (cf. Booij 2000: 336 f.),
such as the English alternations which go back historically to the Great Vowel Shift, as
in the stem allomorphy of divine - divin-ity, wild - wild-er-ness (many examples in
Bauer 1983: 126 ff.).
507
PA510
510

' three' (without a final consonant in pronunciation) vs. trois en/ants 'three chi ldren',
trois-ieme 'third', where <s> is realized as /z/. This relation between liaison and wordformation was already observed by the ancient Indian grammarian Par;iini ( 1989) in his
conception of external vs. internal sandhi.
Free phonological allomorphy, such as in the word-initial vowel variation between
[1] and [s] in explain (Plag 2003: 28), is usually always lexically restricted.
Another type of lexical condition has been attributed to analogical lexical influence
in output patterns within optimality theory, also subsumed under the concept of external
allomorphy (cf. Mascaro 2007; Lloret 2011). For example prosodic stem allomorphy in
E. bureaucrat-ism vs. bureaucrat can be regarded as influenced by parallel accent shift
in bureaucrac-y, i.e. by analogy within an actual word family.
A further type of lexical condition lies in the appurtenance to different lexical strata.
Thus in English, there is the default restriction that Latinate derivation is limited to
Latinate bases. This is clearly the case with the suffixation of adjectivising -ive (phonologically restricted to stems ending in the dentals /t, s/ and only partially transforming
It/ into Isl, as in permit ---+ permiss-ive) or with velar softening (cf. section 3.2). The
German counterpart is much more restricted, namely to Latinate nouns ending in -ik,
when transformed into a verb with the derivational suffix /i:r/, as in Fabrik ' industrial
plant', Rubrik ' rubrique' ---+ fabriz-ier-en, rubriz-ier-en. But this constraint is looser than
in English, because also the Hungarian loan-word Paprika 'peppers' was derived to
papriz-ier-en.
A similar lexical division is responsible for the allomorphy in French quality noun
formation in -te vs. -ite, as in etrange-te 'strange-ness' vs. banal-ite 'banal-ity' (Koehl
20 12).
3.5. Suppletive allo morphy

Lexical conditioning is at its extreme in suppletion, i.e. in the combination of a high
degree of morphosemantic transparency and formal irregularity usually accompanied by
a high degree of morphotactic opacity (cf. Mel'cuk 1993-2000 Vol. 2: 349, 392, 397;
Vol. 4: 361, 369-392; Dressler 1985b), as in Russian suppletive stem allomorphy between xolod-n-yj 'cold' and its prefixed verbal derivatives o-xlad-et' ' to get cold',
o-xlad-it' 'to cool ' (Mel'cuk 1993-2000 Vol. 2: 349). In weak suppletion sometimes
high morphotactic transparency remains as in the stem allomorphy of G. Risiko 'risk'
vs. risk-ant 'risky' or in G. lnsel 'island' vs. adj. insul-ar and in the outdated (translational) name of Indonesia lnsul-inde. The relation between the English and French correspondents, E . island ---+ insular, Fr. ile ---+ insul-aire, is already more opaque. More
important than word-final or word-internal identities in weakly suppletive allomorphy
are word-initial identities and especially prosodic identity of identical segments as in the
above examples. Thus stem allomorphy in Fr. eponge 'spunge' vs. adj. spong-ieux, E.
moon - lun-ar is more opaque, although it is difficult to rank opacity in regard to E.
law vs. leg-al or even also morphosemantically deviant lay-al. And the suppletion between the French toponym Fontainebleau and its inhabitant name Bellifontain must already by considered as an instance of strong suppletion, inspite of having seven or eight
identical segments. Of course, suppletion is strongest in cases such as Fr. aveugle ' blind'
PA512
512
5.1. Backformation and popular etymology

Backfonnation creates non-existing bases in order to motivate polysyllabic words, i.e.
words which have more syllables than usual simple words and whose word length thus
resembles the length of morphologically complex words. It tends towards maximisation
of morphosemantic transparency but allows low degrees of morphotactic transparency.
This may lead to the creation of irregular allomorphs, as in E. beggar > beg, where the
noun has been reanalysed as containing an itTegular allomorph of the suffix -er. In
contrast, the backformation editor > edit has been formed in analogy to the normal
allomorph -or after stem-final /ti, as in creat-or. This allomorph is not simply an orthographic variant of -er (Plag 2003: 189), in view of the derivations manag-er-ial vs. editor-ial.
In contrast, popular etymology, which also creates motivation of longer words, tends
towards partial or full morphotactic transparency, thereby caring little for morphosemantic transparency. Thus it avoids allomorphy. A case in point is Hock's (1991: 203) example of Old French a(u)ndier, borrowed into Middle English as aundyre 'fire dog', and
then first motivated morphotactically in its second part as and-iron and finally fully
motivated as hand-iron (and- could have been regarded as a Cockney variant of hand,
thus not allowing stem allomorphy). In contrast to the production of popular etymologies,
the secondary interpretation of words, which represents an evaluative subpart of popular
etymology, may allow for irregular allomorphy (Olschansky 1996: 193 ff.), as in G.
Attentat 'assault', when related by popular etymology to Attacke 'attack' and Tat 'deed'.
5.2. Blending
In blends (cf. article 21) allomorphic variants are rare but possible, e.g., Fr. technoval
and technival from techno and festival, tigron and tiglon (from tigre 'tiger ' and lion
' lion' ). A llomorphy due to alternative order of constituents does not seem to exist in
adult blends, because one of the conceivable two variants is more salient. Thus E. smog
from smoke and fog is more salient than the conceivable variant *Joke. But in early
(~ Papa
acquisitional stages of child language such variants exist, e.g, Fr. Pama/Mapa
.
.
and Maman ' Daddy-Mama'; Russian Barna, Maha (~ babuska and mama 'GrannyMama' ). These blends have the meaning of coordinate compounds.
5.3. Abbreviatory devices

In extragrammatical word-formation subtractive stem allomorphy occurs by definition
by means of abbreviation (cf. article 19 on clipping), as in Fr. professeur ~ prof 'professor', but due to official or at least social norming there is little variation in the forms of
the abbreviatory targets. This contrasts with the large amount of allomorphy in hypocoristic abbreviations, e.g., E. Elizabeth ~ Liz, Bet, Eliza (plus the possibility of adding
hypocoristic suffixes: Lizz-y, Bett-y), G. Elisabeth ~ Lisa, Liese, Ella, plus a great variety
via hypocorjstic suffixes: Lies-I, Lis-i, Lies-chen, Jss-i, Siss-i (cf. below), Ell-i, Bett-i.
PA513
27. Allomorphy
Variants exist much less in other abbreviations, e.g., Fr. proculproc(ureur) 'prosecutor' , super-prodlsuper-produc(tion) 'superproduction'.
Hypocoristic abbreviation plus subsequent reduplication, derived from proper names
has generally several variants, e.g., Fr. Sophie~ Soso, Fifi, Dominique~ Dodo, Mimi,
Valerie ~ Veve, Lili.
6. Conclusion
The notion of allomorphy has fuzzy boundaries, particularly in the transition to strong
suppletion (section 3.5), to overlapping competing morphemes (section 2.1 ), to zero
allomorphy (which we excluded for derivation but not for inflection: section 4). And
in Amerindian languages there may be even fuzzy boundaries between stem and affix
allomorphy. Although syntax may condition affix competition, it apparently does not
condition allomorphy in word-formation. This fact can be derived from assumptions
on modularity which allow for morphological, morpho(pho)nological, phonological and
lexical conditioning. Apparent semantic and pragmatic conditioning can be reduced to
lexical and sociolinguistic conditions.
7. References
Aronoff, Mark
1976
Word Formation in Generative Grammat: Cambridge, MA: MIT Press.
Aronoff, Mark
1994
Morphology by Itself. Stems and inflectional classes. Cambridge, MA: MIT Press.
Bauer, Lauri
1983
Bernhardt, Barbara H. and Joseph P. Sternberger
1998
Handbook of Phononological Development. San Diego: Academic Press.
Bloomfield, Leonard
1933
Language. Lo ndo n: Allen & Unwin.
Bonami, Boye and Franyoise Kerleroux
2009
L'allomorphie radicale et la relation flexi on-constmction. In: Bernard Fradin, Franyoise
Kerleroux and Marc Plenat (eds.), Apen;us de morphologie dufran<;ais, 103-126. Paris:
Presses Universitaires de Vincennes.
Bonet, Eulalia and Daniel Harbour
2012
Contextual allomorphy. In: Jochen Trommer (ed.), Th e Morphology and Phonology of
Exponents, 195-235 . Oxford: Oxford University Press.
Booij, Geert
1995
The Phonology of Dutch. Oxford: C larendon Press.
Booij, Geert
2000
Morphology and phonology. In: Geert Booij, Christian Lehmann and Joachim Mugdan
(eds.), Morphology. An International Handbook on Inflection and Word-formation . Vol.
l , 335- 343. Berlin/New York: de Gruyter.
Booij, Geert and Ariane van Santen
1995
Morfologie. De Woordstructuur van het Nedelands. Amsterdam: Amsterdam University
Press.
513
PA514
514

2005
Basic terminology. In: Pavo) Stekauer and Rochelle Lieber (eds .), Handbook of WordFormation , 5- 23. Dordrecht: Springer.
Canonical derivational morphology. Word Structure 3: 141- 155.
2010
Corbin, Danielle
Morphologie derivationnelle et structuration du lexique. Tiibingen: Niemeyer.
1987
Drachman, Gaberell, Rene Kager and Angelika Malikouti-Drachman
Greek allomorphy: An optimality account. University of Trondheim Working Papers in
1995
Linguistics 28: 345- 361.
1985a Morphonology. Ann Arbor: Karoma.
1985b Sur le statut de la suppleance dans la morphologie naturelle. Langages 78: 41-56.
1987
Naturalness in word formation. In: Wolfgang U. Dressler, Willi Mayerthaler, Oswald
Panagl and Wolfgang U. Wurzel (eds.), Leitmotifs in Natural Morphology, 99- 126. Amsterdam/Philadelphia: Benjamins.
2000
Subtraction. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-formation. Vol. 1, 581 - 587.
200 1 Extragrammatical vs. marginal morphology. In: Ursula Doleschal a nd A nna Thornton
(eds.), Extragrammatical and Matginal Morphology, 1-10. Miinchen: LINCOM Europa.
Word-Formation in Natural Morphology. In: Pavo) Stekauer a nd Rochelle Lieber (eds.),
2005
Dressler, Wolfgang U. and Lavinia Merlini Barbares i
1994
Morphopragmatics . Berlin: Mouton de Gruyter.
Frege, Gottlob
1892
Uber Sinn und Bedeutung. Zeitschrift far Philosophie und philosophische Kritik N.F.
I 00: 25 - 50.
Fuhrhop, Nanna
1998
Grenzfiille morphologischer Einheiten. Tiibingen: Stauffenburg.
Giegerich, Heinz J.
Lexical Strata in English. Cambridge: Cambridge University Press.
1999
Hock, Hans Henrich
1991
Principles of Historical Linguistics. Berlin: Mouton de Gruyter.
Tacobini, Claudio
2000
Baise and direction of derivation. In: Geert Booij, Christian Lehmann and Joachim Mugdan (eds.), Morphology. An International Handbook on Inflection and Word-formation.
Vol. I, 865- 576. Berlin/New York: de Gruyter.
Janda, Richard D.
On limiting the form of morphological rules: German umlaut, diacritic features, and the
1982
'cluster-constraint' . Proceedings of the Annual Meeting of the North Eastern Linguistic
Society 12: 140-152.
Kaisse, Ellen M.
2005
Word-formation and phonology. In: Pavo! Stekauer and Rochelle Lieber (eds.), Handbook of Word-Formation, 25- 47. Dordrecht: Springer.
Kenesei, Istvan, Robert M. Yago and Anna Fenyvesi
1998
Hungarian. London: Routledge.
PA515
27. Allomorphy
Kiparsky, Paul
1982
From cyclic phonology to lexical phonology. In: Harry van der Hulst and Norval Smith
(eds.), The Structure of Phonological Representations I, 131-175. Dordrecht: Faris.
Koehl, Aurore
2012
La construction morphologique des noms desadjectivaux suffixes en franc;ais. Ph.D.
dissertation, ATILF; Universite de Lorraine.
Lloret, Maria-Rosa
2011
La alomorfia en la teoria de la optim idad. In: Jose Pazo, Irene Gil and Maria Angeles
Cano (eds.), Teoria morfol6gicay morfologia def espanol, 133-161. Madrid: Ediciones
Universidad Aut6noma de Madrid.
Malkiel, Yakov
1966
Linguistics 3, 305-369. The Hague/Paris: Mouton.
Manova, Stela
2011
Understanding Morphological Rules. With special emphasis on conversion and subtraction in Bulgarian, Russian and Serbo-Croatian. Dordrecht: Springer.
Manova, Stela and Wolfgang U. Dressler
2005
The morphological technique of conversion in the inflecting-fusional type. In: Laurie
Bauer and Salvador Valera (eds.), Approaches to Conversion/Zero-Derivation, 67- 10 I.
Munster: Waxmann.
Marchand, Hans
1969
The Categories and Types of Present-Day English Word Formation. A Synchronic-Diachronic Approach. 2d ed. Mi.inchen: Beck.
Mascaro, Joan
2007
External altomorphy and lexical representation. Linguistic Inquiry 38: 715-735.
Mel'cuk, Igor
1993- 2000
Cours de morphologie generale. 4 Vol. Montreal: Les Presses de l' Universite de
Montreat
Menn, Lise and Brian McWhinney
1984
The repeated morph constraint. Language 60: 519- 541.
Mohanan, Karuvannur P.
1986
Th e Th eory of Lexical Phonology. Dordrecht: Reidel.
Mugdan, Joachim
1994
Morphological units. In: Robert E. Asher (ed .), The Encyclopedia of Language and
Linguistics, 2543- 2553. Oxford: Pergamon Press.
Nida, Eugene
1948
The identification of morphemes. Language 24: 414- 44 l.
0 lschansky, Heike
1996
Volksetymologie. Ti.ibingen: Niemeyer.
PaQini
1989 A$fiidhyiiyT. Roman Transliteration and English Translation by Sumitra M. Katre. Delhi:
Byanarsidass.
Peters, Ann M.
1997
Language typology, prosody, and the acquisition of grammatical morphemes. In: Dan
Slobin (ed.), The Crosslinguistic Study of Language Acquisition. Vol. 5, 135- 197. Hillsdale: Erlbaum.
Plag, Inga
2003
Plank, Frans
1994
Inflection and derivation. In: Robert E. Asher (ed.), The Encyclopedia of Language and
Languages. Vol. 5, 1671-1678. Oxford: Pergamon Press.
515
PA517
111. Units and processes in word-formation 11 :
Special cases
28. Affective palatalization in Basque
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
Introduction
Sound substitutions
Affective palatalizatio n and sound change
Semantic evolution
Productivity
Child-d irected speech
Hypocoristic formation
Affective pa latalization in the a llocutive conjugation
Other types of sound symbo lism in Basque
References
Abstract
Affective or expressive palatalization is a phenomenon in Basque whereby certain consonants are replaced by (pre)palatals to create diminutives. Affective palatalization is also
found in child-directed speech, where it may have its origin.
1. Introduction
Affective palatalization is a special morphological process in Basque by which, in its
canonical form, dental or alveolar consonants are replaced with "palatals" (including
palatals and postalveolars) to give rise to diminutives, as in the examples in Table 28.1,
where forms in the first column are unmarked or plain and those in the second column
are diminutives derived by affective palatalization. The phenomenon is also known as
Tab. 28.1 : Examples of affective palatalization

P lain form
tan ta
lagun
bero
nabar
zezen
sagu
zoko
adio
Diminutive
ttantta
llagun
bello
nabar
xexen
xagu
txoko
addio
G loss
'drop'
'friend'
'hot'
'many-colored'
'bull'
'mouse'
'corner'
'goodbye'
Sound substitution
It/
-+
/cl
Ill
Ir!
/n/
-+
I I.I
I I.I
-+
Ip /
/~/
-+
/3/
/'fi./
/fl
/fl
-+
Id /
-+
-+
-+
ltfl
!j !
PA518
518
III. Units and processes in word-formation II: Special cases

expressive palatalization and has been related to sound symbolism (see Onederra 1990,
2002). As is common in other languages, diminutive forms may indicate small size or
may be used to convey affection, either positive or negative (contempt, dislike).
Note on Basque orthography: As can be seen in Table 28.1, <tt> is a voiceless palatal
stop, IPA [c]; < dd > is a voiced palatal stop or approximant, [J] ~ [j]; < II > is a palatal
lateral, [A:]; <fi> is a palatal nasal, [J1]; < x> is a voiceless postalveolar (or prepalatal)
fricative, [J]; < tx> is a voiceless postalveolar affricate, [!f]. The fricahves <s> and <z>
are both voiceless and differ in place of articulation, which is apico-alveolar for <s>,
[]; and !amino-alveolar or lamino-dental for <z>, [~]. The affricates <ts> and <tz>
differ in the same way. I will refer to the class of palatal and prepalatal/postalveolar
consonants involved in these alternations as "palatals" in quotation marks.
Affective or expressive palatalization is thus an instance of sound symbolism, where
certain sounds, "palatals", are imbued with meaning.
2. Sound substitutions
In Table 28.1, some examples of sound replacements in affective palatalizations were
given. Not all of them are equally frequent or have the same dialectal distribution. In
general, affective palatalization has much more vitality in eastern than in western dialects. The most common replacements involved in the creation of affectively palatalized
words are the following:
The sibilant fricatives <s> and <z> are both replaced with prepalatal <x>, as in sagu
'mouse' ---+ xagu ' little mouse ', zahar 'old' ---+ xahar 'oldy', gizon 'man' ---+ gixon ' little
guy', kaskar 'mediocre, insignificant' ---+ kaxkar 'contemptible'. Notice that affective
palatalization causes phonemic neutralization in the case of the sibilants; that is, from
the affective form of a word we cannot recover whether the plain form has <s> or <z>.
When word-initial, the affricate <tx > is frequently used instead of < x> in the central
or Gipuzkoan area, zuri 'white' ---+ xuri - txuri, except that, even in Gipuzkoa, <x> is
kept when the word contains another sibilant fricative, as in zezen ' bull' ---+ xexen ' little
bull' .
The affricates < ts> and <tz> are likewise replaced by < tx> in forms derived by
affective palatalization, as in bi/dots ' lamb' ---+ bildotx 'cute little lamb', beltz 'black' ---+
beltx, bihotz 'heart' ---+ bihotx 'my heart, term of endearment'.
The voiceless dental stop /t/ is replaced by a voiceless palatal stop written <tt>, as
in kutun ' dear ' ---+ kuttun. The use of orthographic gemination to represent palatalization
of /ti in Basque has a long tradition (unlike many other orthographic conventions in
Basque, which are very recent). The influential writer Pedro de Axular, in the prologue
to his 1643 work Gero 'Later', explains that he writes <tt> by analogy with the use of
< 11> in Spanish for a palatal lateral.
The affective palatalization of Id/ is less frequent and is nowadays essentially confined to the eastern dialectal area, except for a few hypocoristics, although some examples of lexicalized variants show that it was once more widespread. The phoneme /d/ is
replaced by either a voiced palatal stop or, between vowels, an approximant. The grapheme <dd >, which is used to represent the affective correlate of /di, is ambiguous between
these two realizations; e.g., andereder 'weasel' (lit. andere ' lady' + eder 'beautiful') ---+
PA519
andderedder, Baztan hildotsa 'the lamb ' --+ hilddotxa. The word dundu 'blue', used only
in the extinct Roncalese dialect, had the affective form ddunddu 'bluish'. In the Zuberoan
dialect, spoken in French territory, the "palatal" correlate of /d/ is a prepalatal voiced
fricative /3/, and is usually represented as <j >, so that the diminutive/affective form of
eder ' beautiful' is ejer or eijer in Zuberoan. Also, sometimes, the grapheme < i> or,
word-initially, <j > are employed for the approximant, as in aio 'bye' (from adio) and
iantza, jantza 'dance' (from dantza).
The affective palatalization of liquids is a much less common phenomenon than that
of obstruents, nowadays also restricted to the eastern area. The alveolar lateral < l> is
replaced by the pa latal lateral < II >, as in lahe ' oven' --+ llahe, lagun 'friend ' --+ llagun.
An example found also further to the west is lo 'sleep' --+ llo llo (with reduplication) in
child-directed speech. The rhotic tap is also replaced by < II > or <dd> in eastern dialects
that allow the palatalization of this consonant, Roncalese maro 'slowly' --+ mallo, Roncalese hero 'hot' --+ hello, Baztan lorea 'flower ' --+ lloddia (Salaburu 2005: 56). In hypocoristics, both <11> and < i> are also found replacing <r>, as in Pe/lo ~ Peio ' Peter' (<
Old Sp. Pero, in some Basque areas Peru). The rhotic trill, on the other hand, does not
enter in these alternations.
Occasionally, both coronal and non-coronal word-initial stops may be replaced by
<tx> or this consonant may be added to vowel-initial words, as in pispildu --+ txispildu
'to get burned, to become happy after drinking', inurri --+ txinurri 'ant'. In some local
dialects the fo1111 w ith initial < tx> is the only one that is used in these examples, so that
the original expressive value has been lost (on this, see section 4). This replacement is
also found in some hypocoristics, Domingo --+ Txomin, Peru --+ Txeru.
The word onddo 'mushroom' most likely derives historically from Spanish hongo
with expressive palatalization.
Initial <tz>, which is very rare in Basque, may replace < z> to create expressive
forms with an augmentative or pejorative flavor (see Ofiederra 1990: 64). From a semantic point of view, this may be considered the opposite of affective palatalization. In the
Navarrese dialect of Baztan, zahar 'old' has the diminutive xahar and the pejorative tzar
'evil' (Salaburu 2005).
3. Affective pa lata lization and sound change

It was noted above that in Gipuzkoan Basque, the affricate < tx> is usually employed
word-initially as a replacement for <s> and <z>, except when the word contains more
than one such sound, as in xexen ' little bull', not *txexen. The effects of the semantics
of palatalization on sound change are even more striking. In all of Gipuzkoa and some
neighboring areas, earlier If I was retracted to a velar point of articulation, exactly as in
Spanish and most likely under its influence. For instance, the verb Jan 'to eat' can be
/jan/, /3an/, If an/ or /xan /, depending on the dialect, and an evolutionary path *lean/ >
/jan/ > /3an/ > !Jani > /xan/ can be inferred. Although the last step in this chain of
changes, velarization, is a regular sound change in Gipuzkoan Basque, it has failed to
apply to If / resulting from affective palatalization in words like /fefen / 'little bull' or I
gof o/ 'sweet'. It thus appears that the sound-symbolic value assigned to "pa latals" prevented the velarization of the prepalatal fricative, which would have destroyed the correlation (Michelena 1985: 194- 195, 201- 202; Trask 1997: 155- 157).
519
PA520
520
4. Semantic evolution
In some cases, an etymologically affective form has become unmarked, and the originally unmarked form may have acquired an augmentative or pejorative meaning. Examples
of markedness reversals for some speakers or dialects are txa"frur 'dog' vs. za"frur 'big,
menacing dog', txerri 'pig' vs. zerri 'big pig, used as an insult', goxo 'sweet' vs. gozo
'excessively sweet' (Michelena 1985: 180).
The plain form has also been lost in some cases, so that the etymologically diminutive
form remains as the only variant. For instance, for many speakers in western areas, the
only word for 'dog' is txakur, which has a diminutive created by suffixation, txakurtxo
'little dog'. In some areas of Navarra, for instance, the word for 'mouse' is always xagu,
the plain form sagu having been lost. As mentioned above, in the Zuberoa dialect, to
give another example, eder 'beautiful' is usually edder, also written eijer, and pronounced with a voiced prepalatal fricative. In Baztan the word anddere 'queen bee' lacks
a nonpalatal form, since the word andere 'lady' has been lost in the dialect (Salaburu
2005: 56).
The semantic connection between the plain and the affective word has occasionally
been lost, even if both forms survive. Thus, among scholars of Basque historical linguistics, it is widely assumed that zori 'luck' is related to txori 'bird' by affective palatalization (Trask 1997: 148-149; Rijk 2008: 15). Although for a scholar it is easy to see the
semantic link in old superstitions regarding birds and omens (in essentially the same
way that the Latin words surviving as English avian and auspicious are related), this
connection is no longer evident to native Basque speakers.
Even though it derives from Spanish celebre 'famous', the Basque word x elebre
(and its less common "plain" form zelebre) means 'funny, peculiar'. From this word,
xelebrekeria 'peculiar or funny behavior' may be derived.
The word deus 'nothing' may be a case of false restoration of a dental from an
affective fonn jeus (also attested), if its origin is, as seems plausible, Latin genus, cf.
Catalan gens 'nothing' (Michelena 1965: 112- 117). Another example of false restoration
is found in jostatu ~ dos ta tu 'to have fun', a borrowing from Old Occitan jostar.
5. Productivity
Most likely the phenomenon of affective palatalization is very old in Basque. However,
the "palatal" consonants are not postulated in the consonantal phonemic inventory of
Ancient Basque in the standard reconstruction (Michelena 1985: 3 74; Trask 1997: 12).
The reason is that these segments are considered mere variants of their non-palatal (dental or alveolar) counterparts, even if their occurrence, instead of being conditioned by
the phonological context, was semantically conditioned. That is, the sound-symbolic
value of palatals may be reconstructed for Ancient Basque.
Nowadays, the productivity of affective palatalization as a morphological procedure
differs significantly among Basque dialects. Whereas in eastern varieties (parts of Navarra, in Spain, and the French Basque region) the phenomenon is still fairly productive, in
the western province of Bizkaia it has lost all productivity, so that diminutives are now
exclusively formed by suffixation and the originally diminutive forms. that are found in
PA521
the lexicon have lost their original diminutive meaning. The most common dim inutive
suffix is -txo (variants -txu, -tto), which has a "palatal" consonant, e.g., txakur-txo ' little
dog', gizon-txo 'little man'.
A factor in the loss of productivity of affective palatalization has been an increased
presence of palatals from other sources, either in borrowings or, more importantly, because of a phonological rule of palatalization that applies to some dental and alveolar
consonants in contact with a preceding high vowel or glide, thus blurring the soundsymbolic value of palatals.
In general, in a given dialect a "palatal" consonant is more likely to retain its soundsymbolic value if that same consonant is not also produced by the contextually conditioned palatalization of coronals. Thus Iverson and Ofiederra (1985) report that in some
Gipuzkoan areas, affective palatalization affects only obstruents (coronal fricatives, affricates and stops), since the sonorants /n/ and /1/ undergo a phonologically-conditioned
rule of palatalization in intervocalic position after Ii/ in this dialect, in words like langife
'worker', mi!!:_a 'the pain'.
An often repeated (but most likely apocryphal) anecdote is that of a preacher in
whose native dialect sibilant fricatives are automatically palatalized after Ii /, so that
gizon 'man' is always pronounced gixon (as it happens in some areas of Bizkaia) and
who goes to preach in another town without this phonological process, where he manages
to offend his male listeners by repeatedly referring to them as ' little men'.
The importance of sound symbolism in an area of Navarra where affective palatal ization remains productive was demonstrated in a study by Yarnoz Yaben (2002: 180-188).
In one experiment, participants were asked to match nonce words presented auditorily
with one of two pictures, differing in size. The auditory stimuli contained either a "palatal" or a non-palatal sibilant fricative or affricate. The result was that subjects unanimously associated made-up words like trampaza and janatsa with larger objects than
the corresponding nonce forms trampaxa, janatxa. This result is particularly interesting
because in another perception test where no semantic clue was used, the pairs of sound s
represented orthographically by <s> and <x> (apico-alveolar/postalveolar) and, especially, the affricates < ts> and < tx> were often confused. That is, the presence of a semantic
cue somehow enhanced the perceptual distinctiveness of the sounds for these Navarrese
subjects.
6. Child-directed speech
Affective palatalization, besides serving to create diminutive forms, is widely used in
child-di rected speech in those eastern areas where it remains a fully productive process.
Salaburu (1984: 118) provides the fo llowing examples for the Navarrese dialect spoken
in the Baztan Valley. Plain zazi ta erraiozu tortzeko 'go and tell him/her to come',
becomes in child-directed speech xaxi ta erraixu ttortxeko, where all coronal obstruents
are replaced by their "palatal" counterparts. Similarly otz iten du ' it is cold ' becomes
otx itten ddu, where /n/ and /d/ are also palatalized. Epelde (2003: 78) also describes the
phenomenon for Larresoro, in French territory, providing examples like zer pasatzen da?
'what happens?' produced as xer paxatxen da ? in the speech of a woman addressing her
grandchildren.
521
PA522
522
7. Hypocoristic formation
Affective palatalization is widely used in the creation of hypocoristics or nicknames.
Thus Pella and Txeru are hypocoristics of Peru ' Peter', Mattin or Matxin is the hypocoristic form of Martin, Txomin is from Domingo and Maddi is from Maria. Also Ixabel,
Maddalen, Te(r)exa, Joxe, Jexux, Xalbador, Ifi.axio and Bittor are forms of the Spanish
names Isabel, Magdalena, Teresa, Jose, Jesus, Salvador, Ignacio and Victor, respectively. The same process may be applied to surnames, thus Xubiri can be the nickname of
someone whose family name is Zubiri or who lives in a house known as Zubiri or
Zubiria.
8. Affective palatalization in the allocutive conjugation

In general, Basque makes a distinction in second person singular pronouns between a
familiar treatment of hi 'thou' and an unmarked treatment of zu 'you' (although many
speakers use only zu). In the Eastern Low Navarrese dialect (spoken mostly in France
but extending also to a few localities south of the political border), there is a third second
person pronoun xu, created by affective palatalization of zu, which is intermediate in
formality (see A lberdi 1995, 1996). This form of treatment can perhaps be defined as
friendly but polite. The choice of second person address permeates the di scourse in
Basque, because of the nature of verb agreement. First of all, in Basque the verb shows
agreement with subject, direct object and indirect object: ikusi duzu 'you have seen it',
ikusi ~aitut ' I have seen you', eman dizut ' I have given it to you' (all examples are given
in normalized standard Basque). In these forms, <z> is replaced by < x> if the addressee
is given xu treatment (ikusi duxu, etc.). In addition, the treatment in hi requires adding
a mark for the addressee in all verbs in main clauses. For instance plain egin du 'slhe
has done it', when the addressee is given hi treatment, becomes egin dik (same meaning,
's/he has done it', but addressing a close male friend). This phenomenon is known as
allocutivity in Basque linguistics. In the easternmost dialects, Eastern Low Navarrese
and Zuberoan, allocutivity has been extended to the treatment in zu (and xu). In this
manner, the choice between zu and xu for second person in dialects with this distinction
will be reflected in every sentence. Our example egin du 's/he has done it' will become
either egin dizu or egin dixu depending on whether the addressee is given zu or xu
treatment. Thus, it should be clear that far from being limited to the use of a pronoun,
the choice of xu as form of address implies an extremely high frequency of the "palatal"
sound If/ in the discourse, thus impregnating the interaction with affective meaning. (For
details on the set of verbal forms in local varieties where the treatment in xu is employed,
see Yrizar 1999.)
9. Other types of sound symbolism in Basque

Besides the correlations involved in affective palatalization, there are other instances of
sound symbolism in Basque (see Ibarretxe-Antunano 2006a, 2006b, 2006c, 2009). In
PA525
29. Parasynthesis in Romance
napa(jvvBswv. In Latin grammar, this latter term was rendered as decompositum. While
this terminology continues to be used - parsimoniously - in the same way in most
philologies (cf., for example, Henzen 1957: 222 on German "Dekomposita oder Parasyntheta" ), in Romance philology it has undergone a semantic change, denoting nowadays
a word-formation process whereby a prefix and a suffix are simultaneously attached to
a lexical base (cf. Darmesteter 1875; Elliot 1884; Malkiel 1941 ; Allen 1981 ; Crocco
Galeas and Iacobini 1993 ; Brachet 1999; Iacobini 2010). The standard example is constituted by Fr. embarquer 'to load, board ', a verb consisting of a base barque 'ship', a
prefix em- and a suffix -er, which is the infinitive ending. Neither *embarqu-, nor *barquer are actual, or even possible French words.
This terminological tradition in Romance philology goes back to Darmesteter (1875:
79-80):
Cette sorte de composition est tres-riche: les mots qu ' elle forme, et que l'on designe du nom
de parasynthetiques, offrent ce remarquable caractere d 'etre le resultat d 'une composition
et d 'une derivation agissant ensemb le sur un meme radical, de telle sorte que l'une ou l'autre
ne peut etre suprimee sans amener la perte du mot. C'est ainsi que de barque l'on fait embarqu-er, de-barqu-er, deux composes absolume nt uns et dans lesquels on ne retrouve ni Jes
composes debarque, embarque, ni le derive barquer, mais le radical barque. La langue tire
les deux composes immediatement du radical, sans !'aide d' aucun intermedianre.
[T his type of compounding is very rich: the words that it forms, whic h are called parasynthetic, show the special characteristics of being the result of compounding and derivation
on the basis of the same root, in a way that neither of them could be eliminated without the
word disappearing as such. In this way, from barque one creates em-barqu-er, de-barqu-er,
two compounds com pletely indivisible, a nd in which it is not possible to find either the
compounds debarque, embarque, or the derived word barquer, only the root barque. The
language fo rms those two compounds directly from the root, without the help of any gobetween.]
As one can see, Darmesteter 's use of parasynthesis still follows the Classical tradition,
since he considered the em- of embarquer and the de- of debarquer to be prepositions
(i.e. Fr. en 'in' and de ' from') and not prefi xes, and hence the combination of em- and
de- plus barque a s cases of compounding. However, when during the 20th century elements such as em- and de- came to be viewed as prefixes, the term parasynthesis was
retained and reinterpreted as the simultaneous use of prefixation and suffixation, instead
of compounding and suffixation.
This terminological tradition has been consolidated among early 201h-century Romance linguists from different origins, among them Menendez Pidal (1904: 130), Nyrop
(1908: 206) and Thorn (1909: 8). Most Romance linguists have agreed on the need to
treat simultaneous double affixation as a necessary condition for parasynthesis, and tried
to differentiate these formations from others where the presence of prefixes and suffixes
is the result of consecutive processes (cf. Alemany Bolufer 1920: 152; Brendal 1943:
125; Badia Margarit 1962: 394-395; Lloyd 1964: 736; Malkiel 1966: 3 14; Tekavcic
1968: 145; Reinheimer-Rlpeanu 1973: 487; Brea 1977: 127-128, etc.).
There were few exceptions to this restrictive view of parasynthesis. Tollemache
( 1945: 110) starts from a wider perspective, but eventually keeps the name parasynthetic
for those cases where the affixing processes take place at the same time ("parasinteti
simultanei", simultaneous parasynthetics, in his terminology). A real exception is Asan 's
525
PA529
sis involving an inflectional suffix (Port. a + baix(o) + ar 'to lower', from baixo ' low',
with inflectional -ar; cf. Valente et al. 2009).
Some of the proposals reviewed above can be included under a wider view of parasynthesis (cf. Serrano-Dolader 1995). The derivational character of the infinitive ending
(or part of it, i.e. the thematic vowel) seems plausible. If inflection cannot change the
category of the base, it is obvious that in Sp. lider (noun) 'leader' ------> liderar 'to lead',
atras (adv.) 'back' ------> atrasar 'to put back', curioso (adj.) 'curious, nosy' - curiosear
'to pry', the category change leads us to think that the element added to the stem should
be considered as a derivational transcategorisation device. It seems illogical to accept
the derivational character of this ending in these verbs while assigning to it a different
character in parasynthetic forms such as brib6n (noun) 'rascal'------> abribonar 'to become
a rascal', prisi6n (noun) 'prison' ------> aprisionar 'to put into prison, to trap'. A possible
interpretation for guaranteeing the parasynthetic character of the whole verbal paradigm - and not just that of the infin itive, which is a simple conventionalised citation
form of the paradigm - is to assign the derivational value just to the thematic vowel (cf.
Scalise 1984: 205).
As argued in Serrano-Dolader (1995), such a proposal can be also applied to the
analysis of parasynthetic forms with an explicit suffix. Thus, in examples such as Sp.
entristecer 'to sadden' ( +----- triste 'sad'), anochecer 'to get dark' ( +----- noche 'night'),
encanecer 'to go grey' (+----- cana 'grey hair') or Port. enrouquecer 'to become hoarse '
(+----- rouco 'hoarse'), esbravejar 'to bellow'(+----- bravo 'rude'), the corresponding thematic
vowel would be still considered a verbalising element, and consequently, it would not
be necessary to assign a transcategorial or verbalising character to the affixes -ec- and
-ej -.
On the other hand, there are conjugated fonns where it is difficult to find the thematic
vowel. In fact, the verbal derivational form can be instantiated by a thematic vowel (as
in the infin itive) or by a zero morpheme, wh ich precedes certain verbal inflectional
forms: Sp. engordAr 'to fatten', engordAmos 'we fatten' vs. engord@o ' I fatten'.
In sum, the thematic vowel and the alternating zero morpheme show a double nature:
that of a derivational morpheme, if the verb is considered as derived from the previous
base; that of a morpheme inherent to the verbal category, if the verb - whether derived
or not - is considered as part of its corresponding inflectional paradigm. This solution
safeguards the derivational value of the thematic vowel in parasynthetic verbs and protects their status (cf. Serrano-Dolader 1995).
3.2. The role of the prefix

Several linguists have claimed that parasynthesis is a counterexample to the principle
that states that prefixation does not allow a transcategorisation of the base. For instance,
Hall (1948: 136- 165) already considered cases that had been traditionally characterised
as parasynthetic as examples of exocentric prefixation (cf. also Wagner 1952: 54 and
Tekavcic 1972: 146). Corbin, as we have already seen in section 2.3, considers the prefix
itself to be responsible for the change in category involved in what are traditionally
considered to be parasynthetic verbs.
The opposite view, as we have also seen, is held by Scalise (1984) and Alcoba (1987).
Both authors reject the view that the prefix is a transcategorisation device in parasynthet-
529
PA533
compound that does not exist independently - with a derivational suffix: Sp .. [*[[macho]
[hembra]] -ar] 'a male + a female + verbal suffix' --7 machihembrar ' to dovetail ' (-ibeing a linking vowel typical of compounds), [* [[corcho] [tap6n ]] -era] 'cork + plug +
suffix' --7 corchotaponero 'relative to the industry of cork plugs' .
This word-formation process seems to be marginal and not very productive in Romance languages, with only a few examples that are difficult to systematise. On the
other hand, the underlying structure in formations created by compounding and suffixation does not always correspond to the parasynthetic compounding model, since it is
possible for suffixation to precede compounding ([[casco] [[mul(a)] [-efio]]] 'hoof +
mule + suffix' --7 casquimuleno 'with small hoofs like those of mules') or vice versa
([[[sal] [pimienta]] -ar] 'salt + pepper + suffix' --7 salpimentar 'to season, spice'); in
these cases, suffixation and compounding do not act jointly and therefore, they do not
give rise to parasynthetic structures.
Parasynthetic compounds in Romance languages have been scarcely studied due to
the heterogeneity of the formations that seem to correspond to the double process of
compounding and suffi xation. Although the concept of "parasynthesis in compounding"
is not new in Romance linguistic studies (cf. Serrano-Dolader 1995: 199-262), this
term - as well as the corresponding concept - has only recently reappeared in studies
on Romance, Germanic, and Slavic languages (cf. Bisetto and Melloni 2008; Melloni
and Bisetto 2010).
These formations are closely related to synthetic compounds ("Zusamrnenbildungen",
in German), such as possessive adjectives of the type E. blue-eyed and G. rothaarig
'red-haired' or nouns of the type E. truck driver and G. Lastwagenfahrer (on synthetic
compounds in German see article 33). Synthetic compounds, however, are not discussed
under the heading of parasynthesis, neither in Germanic nor in Romance linguistics (cf.
Gaeta 20 I 0: 219).
Apart from the lack of an accepted terminology, there is also a deep disagreement
among scholars when they try to delimit the different subtypes of formations that are
likely to be classified as parasynthetic compounds in Romance languages (cf. Bisetto
and Melloni 2008; Melloni and Bisetto 20 IO; Serrano-Dolader 20 l 2b).
5. References
Alcoba Rueda, Santiago
1987
Los parasinteticos: Constituyentes y estructura lex ica . Revista Espanola de Lingiiistica
17(2): 24 5- 267.
A lemany Bolufer, Jose
1920
Tratado de la formaci6n de palabras en la lengua castellana. Madrid: Libreria General
de Victoriano Suarez.
A llen, Andrew S.
198 1 The development of prefixal and parasynthetic verbs in Latin and Romance. Romance
Philology 35: 79- 87.
Aronoff, Mark
1976
Word Formation in Generative Grammar. Cambridge, MA/London: MIT Press.
Asan, F inuta
1965
Formatii paras intetice in limba romana. Limba Romana 14: 87- 95 .
Badia Margarit, Antoni M.
1962
Gramatica catalana. Vol. 2 . Madrid: Gredos.
533
PA534
534

Bisetto, Antonietta and Chiara Melloni
2008
Parasynthetic Compounding. Lingue & Linguaggio 2: 233-259.
Bosque, Ignacio
La Morfologia. In: Francisco Abad and Antonio Garcia Berrio (eds .), Introducci6n a la
1983
lingii,istica, 115-153. Madrid: Alhambra.
Brachet, Jean-Paul
Les preverbes ab-, de-, ex- du latin. Etude linguistique. Villeneuve d' Ascq: Presses du
1999
Septentrion.
Brea, Mercedes
1977
La parasintesis en las Cantigas d' escarnho e de mal dizer. Verba 4: 127- 136.
Brnndal, Viggo
1943
Essais de linguistique generate. Copenhague: Munksgaard.
Corbin, Danielle
1980
Contradictions et inadequations de )'analyse parasynthetique en morphologie derivationnelle. In: Anne-Marie Dessaux-Berthouneau (ed.), Theories linguistiques et traditions
gramaticales, 181- 224. Lille: Presses Universitaires de Lille.
Corbin, Danielle
1987
Morphologie derivationnelle et structuration du lexique. Ti.ibingen: Niemeyer.
Crocco Galeas, Grazia and Claudio Iacobini
1993
Parasintesi e doppio stadia derivativo nella formaz ione verbale de! latino. Archivio Glottologico Italiano 78: 167-199.
Darmesteter, Arsene
1875
Traite de la formation des mots composes dans la languefranr;aise comparee aux autres
langues romanes et au latin. Paris: Franck.
Darmesteter, Arsene
1877
De la creation actuelle de mots nouveaux dans la langue franr;aise et des Lois qui la
regissent. Paris: Vieweg.
Darmesteter, Arsene
1891- 1897
Formation des mots et vie des mots. Cours de grammaire historique de la langue
franr;aise (troisieme partie). Paris: Delagrave.
Dubois, Jean
1962
Etude sur la derivation suffixale en franr;ais moderne et contemporain. Paris: Larousse.
E lliott, A. Marshall
1884
Verbal parasynthetics in a- in the Romance languages. American Journal of Philology
5: 186- 199.
Gaeta, Livia
2010
Synthetic compounds: With special reference to German. In: Sergio Scalise and Irene
Vogel (eds.), Cross-Disciplinary Issues in Compounding, 2 19- 235 . Amsterdam/Philadelphia: Benjamins.
Hall, Robert A. Jr.
Descriptive Italian Grammar. Ithaca, NY: Cornell University Press and Linguistic Socie1948
ty of America.
Harris, Zellig S.
1945
Discontinuous morphemes. Language 2 1: 12 1- 127 .
Henzen, Walter
1957
Deutsche Wortbildung. 2"d ed . Ti.ibingen: N iemeyer.
Hockett, Charles F.
1947
Problems of morphemic analysis. Language 23: 321- 343.
Hockett, Charles F.
1950
Peiping morphophonemics. Language 26: 63-85.
PA535
Tacobini, Claudio
2010
Les verbes parasynthetiques: de )'expression de l'espace a )'expression de l'action. De
lingua Latina 3, < http://www.paris-sorbonne.fr/IMG/pdf/lacob ini_parasynthetiques.pdf>
[last access 9 Sept 20 14].
Lindner, Thomas
2011
Indogerrnanische Grammatik. Vol. 4/1: Komposition. Heidelberg: Winter.
Lloyd, Paul M.
1964
An analytical survey of studies in Romance word formation. Romance Philology 17:
736-770.
Ludtke, Jens
2005
Probleme einer fu nktionellen romanischen Wortbildungslehre: Gibt es Parasynthese? In:
Carmen Kelling, Judith Meinschaefer and Katrin Mutz (eds.), Morphologie und romanische Sprachwissenschaft. Akten der gleichnamigen Sektion beim XXIX Deutschen
Romanistentag, Saarbriicken 2005. Arbeitspapier Nr. 120, 125- 139. Konstanz: Fachbereich Sprachwissenschaft der Universitat Konstanz. < http://kops.ub.unn-konstanz.de/
bitstream/handle/urn:nbn:de:bsz:352-opus- l 8 l 22/ AP_ 120.pdf?sequence= l > [last access
9 Sept 201 4].
Ludtke, Jens
20 11 La "parasynthese": Une fausse piste? Romanische Forschungen 123(3): 312- 330.
Malkiel, Yakov
1941
Atristar - entristecer: Adjectival verbs in Spanish, Portuguese and Catalan. Studies in
Philology 38: 429- 46 1.
Malkiel, Yakov
1966
Linguistics. Vol. 3, 305-365 . T he Hague/Paris: Mouton.
Marchand, Hans
1955
Synchronic analysis and word-formation. Cahiers Ferdinand de Saussure 13: 7- 18.
Mascaro, Joan
1985
Morfologia. Barcelona: Enciclopedia Catalana.
Melloni, Chiara a nd Antonietta Bisetto
Parasynthetic compounds: Data and theory. In: Sergio Scalise and Irene Vogel (eds.),
20 10
Cross-Disciplinary Issues in Compounding, 199-217. Amsterdam/Philadelphia: Benjamins.
Menendez Pidal, Ramon
1904
Manual elemental de gramatica hist6rica espaiiola. Madrid: Libreria general de Yictoriano Suarez.
Nyrop, Kristoffer
1908
Grammaire historique de La Langue frani:;aise. Vol. 3. Copenhague: Gyldendalske Boghandel Nordisk.
Padrosa Trias, Susanna
2007
Argument structure and morphology: The case of en- prefixation revisited. Anuario de/
Seminario de Filologia Vasca Julio de Urquijo. International Journal of Basque Linguistics and Philology 41 (2): 225- 266.
Pottier, Bernard
1962
Systematique des elements de relation . Paris: Klincksieck.
Reinheimer-R'ipeamu, Sanda
1972
Suffixe zero? Revue Roumaine de Linguistique 17: 261-269.
Reinheimer-R'ipeam.1, Sanda
1973
Differents types de parasynthetiques. Revue Roumaine de Linguistique 18: 487- 49 1.
Reinheimer-R'ipeanu, Sanda
1974
Les derives parasynthetiques dans /es langues romanes. The Hague/Pa ris: Mouton.
535
PA536
536

Scalise, Sergmo
1984
M01fologia Lessicale. Padua: CLESP.
Scalise, Sergio
1986
Serrano-Dolader, David
1995
Lasformaciones parasinteticas en espaiiol. Madrid: Arco/Libros.
1999
La derivaci6n verbal y la parasintesis. In: Ignacio Bosque and Violeta Demonte (eds.),
Gramatica Descriptiva de la Lengua Espanola . Vol. 3, 4683- 4755 . Madrid: Espasa
Cailpe/Real Academia Espanola .
2012a Sobre los adjetivos l,parasinteticos? locativos (submarino, intramuscular, interdigita/).
In: E lisenda Bernal, Carsten Sinner and Martina Emsel (eds.), Tiempo y espacio en la
formaci6n de palabras de/ espaiio/, 65-78. Miinchen: Peniope.
2012b Sobre los compuestos (para)sinteticos len espaiiol? In: Antonio Fabregas, E lena Feliu,
Josefa Martin and Jose Paz6 (eds.), Los limites de la morfologia. studios ofrecidos a
Soledad Varela Ortega, 427- 442. Madrid: Servicio de Publicaciones de la Universidad
Aut6noma de Madrid .
Tekavcic, Pavao
1968
Fonnazione de lie parole nell' istroromanzo dignanese. Lingua e Stile 3: 125- 180.
Tekavcic, Pavao
1972
Grammatica storica dell'italiano. Vol. 3: Lessico. Bologna: ll Mulino.
Thiele, Johannes
1984
Reflexiones comparativas sobre verbalizaciones con bases adjetivas en espafiol y aleman. Linguistische Arbeitsberichte 45: 19-24.
Thorn, Anders C hristopher
1909
Les verbes parasynthetiques enfranr;ais. Lund: Lunds Universitets Arsskrift.
Tollemache, Federico
1945
Le parole composte net/a lingua italiana. Roma: Edizioni Rores di Nicola Ruffolo.
Valente, Ana Carolina Mrad de Moura, Caio Cesar Castro da Silva, Carlos Alexandre Gorn;:alves and
Maria Lucia Leitao Almeida
2009
Enfoques sobre parassintese em portugues: Da tradic;:ao gramatical a lingiiistica cognitiva. Revista Virtual de Estudos da Linguagem (ReVEL) 7: 75-89.
Wagner, Robert-Leon
1952
Remarques sur la valeur des preverbes a- et en- (<IN) en ancien franc;:ais. In: Festgabe
Ernst Gamillscheg. Zu seinemfonfundsechzigsten Geburtstag am 28.. Oktober 1952, 51 65 . Tiibingen: N iemeyer.
David Serrano-Dolader, Zaragoza (Spain)
PA538
538
2. The definitional scope of pleonasm

The ultimate etymon of pleonasm is the Ancient Greek nA.covacr6~ meaning 'superabundance, excess'. Thus, the etymology of pleonasm does not reveal in which ways the
excess that it describes occurs.
The most systematic treatment of the topic has been provided by Lehmann (2005),
who studies pleonasm at different leve ls of linguistic analysis and identifies hypercharacterization as "pleonasm at the level of grammar" (2005: 119). Prior to this, by reference
to Paul (1920: 162), Haspelmath (1993: 297) defines affix pleonasm as "the semantically
vacuous addition of a transparent affix to a word that is already characterized for the
morphosyntactic property expressed by th is affix". Focusing on cases such as children,
evolved in Middle English as a compromise between the forms child(e)r and chi/den,
Hock (1986: 189-190) treats affix pleonasm as a non-systematic process under the heading of blending - a position that Haspelmath ( 1993: 300) rejects by arguing that pleonasm does not necessarily result from a morphological compromise of two existing
occurring forms: for example, Vulgar Latin esse-re 'to be' cannot be seen as the result
of blending, because Latin does not attest any form *es-re to combine with the infinitive
esse. Admittedly, there are cases of affix pleonasm which match true affix blending, for
example, in some varieties of German, rundlicht ' round ish' (synonymous to standard
rundlich) shows the suffix -Licht, which is a blend of the suffixes -lich and -icht (see
Paul 1920: 162; Plank 1981: 77- 79; Haspelmath 1993: 307 n. 12). Therefore, affix
blending can be considered a hyponym of pleonasm.
Malkiel (1957: 79) provides the following both descriptive and explanatory definition
of the phenomenon (that he terms hypercharacterization), in which the diachronic dimension is explicitly highlighted:
If a given linguistic formation develops in such a way as to allow, at a certain point, one of
its distinctive features to stand out more sharply than at the immediately preceding stage, one
may speak of hypercharacterization (or hyperdetermination) of that feature, in the diachronic
perspective.
Another term which recurs in the literature is overcharacterization, described as "adding

a suffix that is strictly speaking superfluous, and hence a pleonastic addition" (Booij
2007: 273).
All these definitions have in common that they do not reveal anything about how to
constrain the conceptual scope, that is, the intension of the definition of pleonasm. In
fact, phenomena running under the heading of pleonasm can be motivated pragmatically
or phonologically: for example, the Spanish plural forms pie-s-es 'feet', cafe-s-es 'coffee-s', and Old Latin 3PL.PRS da-n-unt 'they give' (besides Classical da-nt) are prosodically motivated (Dressler, Dziubalska-Kolaczyk and Spina 2001: 123). But, then, is this
not motivation enough to discard the characterization of these forms as pleonastic? A
way to solve this doubt would be to exclude pragmatic and prosodic motivation from
the definitional scope of pleonasm and to only account for both semantic and functional
(i.e. morphosyntactic) motivation, but this would not do justice to the synchronic variation appearing in language.
A further aspect relating to the definitional intension of pleonasm concerns the way in
which pleonasm is realized. In general terms, the realization of semantic and functional
PA539
30. Affix pleonasm
properties can be explicit (or overt) and implicit (i.e. lexical). Based on the distinction
between explicit vs. implicit characterization, the following types of pleonastic, that is,
over-characterized, marking, can be derived:
l. Implicit pleonastic marking is the combination of a lexeme implicitly carrying a semantic or functional value with one element which explicitly (morphologically or
syntactically or both) codes the same value. For example, an explicit marker having
a value Vis combined with a lexeme which inherently contains V: a case in point is
the historical addition of the overt feminine marker -aa to inherently feminine nouns
in Hausa (Newman 1979).
2. Explicit pleonastic marking is the addition of double or multiple explicit (morphological or syntactic or both) marking of a value V to a lexeme which does not code V
inherently, for example, the Italian NP tre cani 'three dogs' with realization of the
plural via both the numeral and the suffix -i, as well as the Spanish example cafe-ses mentioned above. A further subtype of explicit pleonastic marking wou ld be the
realization of V via double or multiple explicit (morphological or syntactic or both)
marking on an item which implicitly holds this value. While this option is virtually
possible, no case of this type is known to me.
For the sake of completeness, the remaining part of this section briefly dis.cusses what
kinds of phenomena can be confused with, but do not count as pleonasm. Crucially,
pleonasm must be kept apart from both affix replacement and general loanw ord integration devices. To the first type belong, e .g., Old Aragonese a/fayante 'tailor', which
resulted from the renewal of a/fayate through suffi xation with -ante (Malkiel 1957: 108).
To the second type belong, for example, "loanverb markers'', such as the suffix -oa in
Yaqui (Uta -Aztecan) which is added to verbs borrowed from Nahuatl and Spanish in
order to facilitate accommodation, e.g., mediar-oa [mediate-LYM] 'to mediate' from
Spanish mediar (Wohlgemuth 2009: 226).
Sometimes scholars apply the label " interference suffix" (see Kolb 1980: 283; Muller
2005: 38) to cases of integration of borrowed lexemes into inflectional classes of the
recipient language, e .g., Latin operari > Old High German opferon 'to sacrifice' . This,
too, must be seen as a separate procedure, as it is just the result of the application of
productive inflectional rules (the infinitive marker -on) to the base of borrowed items in
order to make the m fit for syntax (cf. Gardani 201 3).
3. Pleonasm in language
In this section, I detail at which levels of language analysis and processing pleonasm
can occur, before deepening the phenomenon of affix pleonasm in derivational morphology in section 4.
At the level of discourse, some languages have pleonastic forms acting as discourse
markers, for example, in Yiddish, the expletive es instructs the hearer that the subject
does not represent an entity already evoked in the discourse: es geyt epes in vald a yid
'Some Jew seems to be walking in the woods; lit. it goes something in wood a Jew'
(Prince 1993: 176); also Dominican Spanish ello, European Portuguese ele, and Balearic
Catalan ell all fun ction as markers that encode sentence pragmatics, e.g., Balearic Catalan Ell no n'hi ha! 'It does not exist; lit. It not (not) there exists!' (Hinzelin 2009: 17).
539
PA540
540

In the lexicon, pleonasm often occurs in the form of pure addition of modifiers to
lexemes which inherently encode the meaning expressed by the modifier, e.g., dead
corpse, briefly sketch; Spanish aniversario anual 'annual anniversary'; French conversation orate; German mundliches Gesprach 'oral conversation'; or the additive use of a
synonymous grammeme, e.g., Italian ma pero lit. 'but but'.
An originally lexical item can also be the source of syntactic pleonasm due to grammaticalization. This is famously the case of the French sentential negation ne ... pas
(Rowlett 1998): initially, the noun pas 'step', from Latin pass(um), served as a reinforcement in the clauseje ne vais 'I do not go'-----+ je ne vais pas 'I do not go any step'; later,
pas was desemanticized and acquired the grammatical function of negation; from this
time, the negation pas is a pleonastic addition to the negative adverb ne. Moreover, in
the syntax, there is pleonasm in expletives such as the non-standard Enghsh 'excrescent 's',
for example in Does anyone see what's the tactic is? This element reinforces or emphasizes the WH interrogative word, just as in (non-standard) how's about, how's come,
what's about (Zwicky 20 12).
As has been observed (Lehmann 2005: 137-138), pleonasm is very frequent in the
expression of spatial relations. For example, in German, particle verbs license (pleonastic) "directional PPs" (Olsen 1996; Okamoto 2002; Rehbein and Genabith 2006), such
as in Peter lief[pp durch den Wald] durch 'Peter ran through the forest'. Similarly, Latin
attests the pleonastic realization of spatial relations via the combined use of a preposition
and a preverb, e.g., ex urbe ef-fugere 'to flee out of town ' (Lehmann 2005: 138). Moreover, Latin has what has been called "pleonastic reflexive", e.g., suo sibi lautum sanguine
tepido [his:ABL.SG RFL bath:PTCP.PST.ACC.SG blood:ABL.SG warm:ABL.SG] 'bathed in his
own warm blood': here, the reflexive sibi is a fossilized omissible expression of possession (Cennamo 1999: 117).
Frequently, pleonasm occurs in the realization of the values of comparative and superlative, producing syntactic (i.e. analytic) constructions which combine an adverb and an
inflected fonn to express a single predicate, e.g., Middle English more strenger (Wlodarczyk 2007: 196), substandard French le plus meilleur, Spanish el mas mejor (Lehmann 2005: 139). In other languages, implicit pleonastic marking of the superlative
value may occur by adding dedicated inflectional formatives to inherently marked lexemes, e.g., German (das) bestmoglich-ste (Lehmann 2005: 140), Late Latin minimissimus for minimus (Malkiel 1957: 86). Instead, in Modern Greek, we have explicit pleonastic realization of the superlative via double comparative suffixation, e.g., o kali-tero-ter(os) [good-c oMP-COMP] 'bestest' (Petros Karatsareas, p.c.). In south-eastern dialects
of Lithuanian, the superlative may be realized, at once, by the simultaneous application
of the inherited superlative suffix and a superlative prefix nai- borrowed from Slavic,
e.g., nai-gardz-iaus-i obuoliai 'the most delicious apples' (Grinaveckiene 1969: 222;
Wiemer 2009: 353).
Possessive constructions are not immune to pleonasm either, as shown by the syntactic augment in Spanish SU casa + de el [POSS house + of him] 'his house ' (Malkiel
1957: 100), as well as the non-standard southern German 'dative + possessive adjective
construction' of the type dem Peter sein Buch 'Peter's book'.
In syntax-dependent (i.e. contextual) inflection, we find the perhaps most common
manifestation of pleonasm in grammar - viz. agreement. Agreement is the redundant
realization of feature values by means of discontinuous affixes in order to facilitate
understanding for the hearer (Corbett 2006: 274- 275). Nowadays, the most common
PA541
30. Affix pleonasm
terms used to describe this phenomenon of morphological asymmetry are "extended
exponence" (Matthews 1972), "multiple exponence" (Halle and Marantz 1993 ), and "exuberant exponence" (Harris 2008). For example, in Batsbi (Northern Caucasian), class
marking (contextually: gender-number agreement) can occur in numerous positions within a single verb form, such as in tisin c'a dan d-ex-d-o-d-ano [old house(d/d).ABS PY
CM-destroy-cM-PRS-CM-EVIDI] 'they are evidently tearing down the old house', via triple
affixation with -d- (Harris 2009: 267- 268).
A particular type of pleonasm is word-internal agreement (Stolz 2007), which has
been claimed to result from a process of externalization of inflection (Haspelmath 1993).
A well-described case is the realization of definiteness in Lithuanian adjectives. Here,
the adjective stem is inflected for case, number, and gender, and this is followed by a
suffixal definiteness formative which is again marked for the same features (Stolz 2010:
236), e.g., balt-os-i-os [white-NOM.PL.F-DEF-NOM.PL.F] 'white'.
Apart from agreement, the following morphosyntactic features (Corbett 2012) may
be realized pleonastically, in the sense of reinforcement of single formatives:
1. case, as exemplified by the genitive (singular) of the Latin pronouns eius
(< *e(syo)(-s)), cuius, huius (Malkiel 1957: 98), or, in German, of both common
nouns, e.g., Hasens 'of the rabbit', and proper names, e.g., Mariens ' Mary's' (Paul
1920: 162);
2. number, in particular the value of plural, as frequently attested in European languages,
e.g., Portuguese alvara-z-es 'charters ', fllh6-s-es 'fried doughs' (Malkiel 1957: 98);
Dutch kind-er-en 'children'; northern German Junge-n-s 'boys'; moreover, it is particularly recurrent in loanwords, e.g., English spaghett-i-s; Dutch lied-er-en 'songs';
Afrikaans vrou-en-s 'women' (Thomason 1988: 304). In Maltese, there are cases in
which templatic morphology is combined with concatenative morphology, e.g., trufijiet [ends-PL] 'ends', where truf is the broken plural of tarf and the formative -ijiet
realizes plural,. too (Maris Camilleri, p.c.). In southwestern dialects of Hungarian, we
find cases in which the plural is doubly realized via the regular plural formative, -Ek,
and the plural possessive prefix, i-, e.g., in tehen-i-m-ek 'my cows' (Imre 1972: 320);
3. gender: some varieties of Italian spoken in Garfagnana give examples of Giller6n's
"linguistic therapeutics" (Gillieron 1921: 11 ), that is, inflectional class shift of lexemes from a gender-opaque class to a class that marks a certain gender value more
neatly, e.g.,fiumo 'river' for standard Italian.flume (masculine): this is an instance of
implicit pleonasm, because the inflectional class libro libri realizes the gender value
masculine on a noun which is inherently masculine (Malkiel 1957: 81).
On the side of verbs, pleonastic realization of person is found, for example, in the Greek
dialect spoken in the village of Ochthonia in Euboea, in forms such as erx-es' tane-s
'you were coming', which reshape the opaque 2sG medio-passive imperfect form *erxes'tane by adding the 2SG (final) -S formative, yielding more paradigmatic transparency
(Pantelidis 2010: 3 23 ).
A quite common case of pleonasm concerns the realization of the morpho-semantic
feature of tense. MacKay (1979: 487) refers to pleonastic past tense suffixation in child
language: for example, smashted is due to reanalysis of smasht as a present tense form
and addition of the past tense formative -ed (further examples in Bowerman 1982: 327).
As concerns word-formation, pleonasm is attested in compounding, abbreviation, and
derivation. Pleonastic compounds are not uncommon in the languages of Europe: Frisian
541
PA543
30. Affix pleonasm
il-ebil-(in)-ir [go-PASS-ABTL-PASS-AOR] ' it is possible to go' (Ash Gokse l, p.c.). Moving
our way down from primarily inflectional towards primarily derivational categories, we
find instances of affix pleonasm in the realization of causative. An intriguing case is
found in the Romani variety of Selice (Slovakia), where in the perfective forms, two
causative allomorphs can be used pleonastically, that is, without yielding a double-causative interpretation: for example, the aorist form an-av-a-d'-a [do-cAus-cAUS-PFV3 sG.PFv] regularly means 's/he had (sth) ordered; lit. s/he makes (so) make (so) bring
(sth)' but can also mean 's/ he ordered (sth)' (Viktor E lsik, p.c.).
In polysynthetic Adyghe (Northwest Caucasian), pleonasm can affect the expression of participant-internal possibility: for example, in se nwi"'e-r qa-s-fe-?eta-sva-K-ep
[ lsG stone-ABS DIR- lsG.IO-BEN-raise-ABIL-PST-NEG] ' I could not raise the stone', possibility is realized by means both of the abilitive suffix -sva and (a modal use of) the
benefactive applicative prefix/e- (Yury Lander, field notes).
A clearly derivational operation which is often accompanied by affix pleonasm is
derivational gender marking, e.g., German Hindin 'female deer' for Hinde (Covington
1981: 35), Diakonissin 'deaconess' for Diakonisse (Malkiel 1957: 86).
Some few languages display pleonastic formation of agent nouns (see article 74),
especially when they denote the agent's profession or rank. Pleonastic agentive suffixation occurs in German, e.g., Vorfahr-er 'ancestor' (Paul 1920: 62); Old French lamaneur 'pilot' (Malkiel 1957: 107); Spanish VN compounds, by means of the agentive suffix
-era, e.g., picapedr-ero 'stonemason' for picapiedras (Rainer 1993: 268, 487). In Dutch,
pleonastic affixation is also found in acronyms which are enriched with the suffix -er,
resulting in denominal names, such as UD-er 'university teacher' (Universitair Docent)
(Booij 2007: 273).
Processes of intensification such as diminution, augmentation, and iteration are particularly prone to affix pleonasm (Lehmann 2005: 145). Just as, in inflection, pleonastic
marking of the plural seems to be the most frequent case of affix pleonasm, in derivation,
the champion role is played by diminutive marking. This is richly attested in the languages of Europe: see in Polish child-directed speech Monisienko [Monika:DIM:DIM.voc]
' little Monika!' (Wierzbicka 2003: 53); eastern Yiddish majlxl ' little mouth', from Standard Yiddish moil 'mouth' + xa +!(Jacobs 2005: 69; Herzog and Baviskar 2000: 120);
in Lithuanian, puod-as 'pot' may be diminuted as puod-uk-as 'small pot, cup', but also
as puod-uk-el-is 'small nice pot', via the diminutive suffixes -uk and -el (Gaters 1977:
60-6 1); in German (cf. article 134), fo1111s such as See-lein-chen 'small lake', Austrian
German Schatz-i-lein ' darling; lit. little treasure' are not uncommon; in Icelandic, there
are combinations of prefixoids (pinu-, sma-) and suffixes (e.g., -lingu,r), e.g., pinu-disklingur or sma-disk-lingur 'tiny diskette ', though these cases are far from common there
(porsteinn lndrioason, p.c.); in Greek, manulitsa 'mommy' is formed via -ul(a) + -its(a)
suffixation of the base man(a) 'mother ' (Petros Karatsareas, p.c.); in Turkish, a sequence
of the two diminutive suffixes, -Aczkl-kik and -Cik, is ungrammatical and would produce
an ill-formed form such as *kU<;iiciikciik (from {kii<;iik+icik+cik} ), but it works when a
possessive suffix is added, e.g., kii<;iiciikciigiim 'my little tiny one' (Ash Goksel, p.c.).
Interestingly, in Italian, while double diminutive suffixes can effect further denotative
diminution, diminutivizing interfixes are denotationally meaningless, e.g., libr-ino/-etto
(smaller than libro 'book') ~ libre-ett-ino (still smaller), but libr-ic(c)-ino (not smaller
than librino) (Dressler and Merlini Barbaresi 1994: 540). Sti 11, they seem to have con no-
543
PA545
30. Affix pleonasm
deverbal derivational affixes can be themselves borrowed items, as is the case of the
suffix -avy, bormwed from Slavic languages in the Lithuanian dialect of Zietela, e.g.,
den-avy-ti 'to lay down' (vs. standard deti(s)) or grii-avy-ti-s 'to return' (vs. standard
griiti) (Wiemer 2009: 360).
5. Why affix pleonasm?

After having provided, without any claim to completeness, a survey of the types of
derivational affix pleonasm occurring in the languages of Europe, this final section tries
to overview how linguists have approached pleonasm and, in particular, its motivations
from a theoretical viewpoint.
For this purpose, I turn to Haspelmath's definition of affix pleonasm, as. reproduced
in section 2. In fact, this defin ition suggests that the morphosyntactic features of those
word forms that we define as pleonastic, are already realized by primary affixes, to
which the secondary (i.e. pleonastic) affixes are claimed to be a "semantically vacuous
addition". While the existence of primary affixes is the theoretical foundation for acknowledging the secondary affixes as being de facto pleonastic, the status of both primary and secondary affixes needs to be discussed in more detail.
The issue at hand here is whether pleonasm qualifies as a complete or rather a transitory phenomenon in terms both of diachronic evolution and language acquisition. Clearly, in a diachronic perspective, pleonasm can refer to a transitory stage, as can be exemplified with the evolution of sentential negation in French from stage (1), Old French
jo(u) ne vais, via stage (2), je n(e) vais pas, to stage (3), contemporary colloquial j'vais
pas ' I don't go' (see Malkiel 1957: 90). However, in synchronic tenns, one form can be
definitive or more variants can be in competition with each other: for example, in contemporary French, both (2) and (3) are possible variants, though they reflect a difference
in register.
In scenarios of early phases of language acquisition (Dressler 1997), Turkish children
produce forms such as manav-cz 'greengrocer-cz' (Ash Goksel, p.c.), English children
have plurals such as feets or past-tense forms such as earned (Covington 1981: 35).
These cases, in fact, represent an intermediary stage. Also in language contact, as we
have seen in some of the cases reported, pleonasm can be considered, at least partly, the
reflex of an intermediary stage: for example, German verbal, nominal have not been
extended via -isch (*verbalisch, *nominalisch).
These facts lead us to raise the question of whether affix pleonasm exists at all. In
other words, do pleonastic affixes have any psycholinguistic salience in the grammars
of the speakers or do they only exist in the heads of linguists? Dressler, DziubalskaKolaczyk and Spina (200 1: 124) argue that "hypercharacterization" (i.e. pleonasm) "is an
imprecise concept that is only justified in a very superficial morphotactic or panchronic
perspective", because synchronically, the secondary affixes are, in fact, the only markers
which are relevant in terms both of productivity and generality.
Thus, if secondary (pleonastic) markers alone realize meanings and values, what are
the conditions that make speakers use these markers? What are the conditions under
which their occurrence is necessary? While Dressler, Dziubalska-Ko!aczyk and Spina
(2001: 124) claim that pleonasm is only an apparent phenomenon, thus does not represent a violation of biuniqueness, other linguists consider it to be the source of allomorphy, of uneconomical and non-uniform coding and, thus, a violation of the elsewhere
545
PA547
30. Affix pleonasm
CM
COMP
DEF
DIM
DIR
EVlD
class m arker
comparative
definite
diminutive
directional
evidential
547
PTCT
PY
RFL
SG
TRR
participle
preverb
reflexive
singular
transitivizer
6. References
Booij, Geert
2007
The Grammar of Words . Oxford: Oxford University Press .
Bowerma n, Melissa
1982
Reorganizational processes in lexical and syntactic development. In: Eric Wanner and
Lila R. G le itman (eds.), Language Acquisition. The State of the Art, 3 19-346. Cambridge: Cambridge University Press.
Cennamo, Michela
Late Latin pleonastic reflexives and the Unaccusative Hypothesis. Transactions of the
1999
Philological Society 97( I): I03-150 .
2006
Agreement. Cambridge: Cambridge U niversity Press.
2012
Features . Cambridge: Cambridge University Press.
Covington, Michael A.
Evidence for lexicalism. A critical review. Bloomington, IN: Un iversity Linglll istics C lub.
198 l
1997
"Scenario" as a concept for the functiona l explanation of language change. In: Jadranka
Gvozdanovic (ed .), Language Change and Functional Explanations, 109- 142. Berlin:
Mouton de Gruyter.
Hypercharacterisation and productiv ity in inflectional morphology. In: Thomas Krisch,
2004
Thomas L indner and Ulrich Muller (eds.), Analecta homini universali dicata. Arbeiten
zur lndogermanistik, Linguistik, Philologie, Politik, Musik und Dichtung. F estschrift far
Oswald Panagl zum 65. Geburtstag, 5 15-524. Stuttgart: Heinz.
Dressler, Wolfgang U ., Katarzyna Dziubalska-Kolaczyk and Rossella Spina
200 1
Sources of markedness in language structures. Folia Linguistica Historica 22(1-2):
103- 136.
Dressler, Wolfgang U. and Lavinia Merlini Barbaresi
1994
Morphopragmatics. Diminutives and Intensifiers in Italian, German, and Other Languages. Berlin/New York: Mouton de Gruyter.
Fleischer, Wolfgang and Jrmhild Barz
20 12
Gardani, Francesco
Dynamics of Morphological Productivity. The Evolution of Noun Classes from Latin to
2013
Italian. Le iden/Boston: Brill.
Gaters, A lfreds
1977
Die lettische Sprache und ihre Dialekte. The Hague: Mouton.
Gillieron, Jules
192 l
Pathologie et therapeutique verbales. Paris: Champion.
PA548
548

Glasser, Eduard
1954
Review of Moritz Regula (1951) Grundlegung und Grundprobleme der Syntax. Heidelberg: Winter. Romanische Forschungen 65: 424-429.
Grinaveckiene, Elena
Lietuvitj ir slavtj kalbtj gramatinio kontaktavimo reiskiniai pietryciq Lietuvoje. In: Vyt1969
autas Ambrazas (ed.), Lietuvi11 kalbos gramatikos tyrinejimai, 219- 229 . Vilnius: Lietuvos TSR Moksltj Akademija.
Halle, Morris and Alec Marantz
1993
Distributed morphology and the pieces of inflection. In: Kenneth L. Hale and Samuel J.
Keyser (eds.), The View from Building 20. Essays in Linguistics in Honor of Sylvain
Bromberger, 111-17 6. Cambridge, MA: MIT Press.
Harris, Alice C.
2008
Explaining exuberant agreement. In: Th6rhallur Eyth6rsson (ed.), Grammatical Change
and Linguistic Theory. The Rosendal Papers, 265-283 . Amsterdam/Philadelphia: Benjamins.
Harris, A lice C.
2009
Exuberant exponence in Batsbi. Natural Language and Linguistic Theory 27(2): 267-303.
Haspelmath, Martin
1993
The diachronic externalization of inflection. Linguistics 3 1(2): 279- 309.
Herzog, Marvin and Vera Baviskar (eds.)
2000
The Eastern Yiddish - Western Yiddish Continuum. Tiibingen: Niemeyer.
Hinzelin, Marc-Olivier
2009
Neuter pronouns in Ibero-Romance: D iscourse reference, expletives and beyond. In:
Georg A. Kaiser and Eva-Maria Rem berger (eds.), Proceedings of the Workshop "Nullsubjects, expletives, and locatives in Romance", 1-25. Konstanz: Fachbereich Sprachwissenschaft Universitat Konstanz.
Hock, Hans Henrich
1986
Principles of Historical Linguistics. Berlin/New York: Mouton de Gruyter.
Horn, Wilhelm
1939
Neue Wege der Sprachforschung. M arburg: Elwert.
Hyrtl, Joseph
1880
Onomatologia anatomica. Geschichte und Kritik der anatomischen Sprache der Gegenwart, mit besonderer Beriicksichtigung ihrer Barbarismen, Widersinnigkeiten, Tropen,
und grammatikalischen Fehler. Wien: Braumiiller.
Imre, Samu
1972
Hungarian dialects. In: Benko Lorand and Samu Imre (eds .), The Hungarian Language,
299-326. The Hague: Mouto n.
Jacobs, Neil G.
2005
Yiddish. A Linguistic Introduction . Cambridge: Cambridge University Press.
Kolb, Herbert
1980
Uber verbale Interferenzsuffixe. In: Hans D. Bork, Artur Greive and Dieter Woll (eds.),
Romanica europea et americana. Festschriftfiir Harri Meier. 8. Januar 1980, 282- 292.
Bonn: Bouvier.
Lehmann, Christian
2005
Pleonasm and hypercharacterisation. In: Geert E . Booij and Jaap v. Marie (eds.), Yearbook of morphology 2005, 119- 154. Dordrecht: Springer.
Lewis, Geoffrey
1967
Turkish Grammar. Oxford: Clarendon Press.
MacKay, Donald G .
1979
Lexical insertion, inflection, and derivation: Creative processes m word production.
Journal of Psycholinguistic Research 8(5): 477-498.
Malkiel, Yakov
Diachronic hypercharacterization in Romance. Archivum Linguisticum 9( I): 79-1 13.
1957
PA549
30. Affix pleonasm
Matthews, Peter H.
Inflectional Morphology. A theoretical study based on aspects of Latin verb conjugation.
1972
Cambridge: Cambridge University Press.
Historische Grammatik der franzosischen Sprache. Vol. 2. Heidelberg: Winter.
192 1
1966
Historische Grammatik der Jranzosischen Sprache. Vol. 2. 2nd ed. Heidelberg: Winter.
Migliorini, Bruno
Sulla tendenza a evitare ii cumulo dei suffiss i nella formazione degli aggettivi. In: Ar1943
nold Steiger (ed.), Sache, Ort und Wort. Jakob Jud zum 60. Geburtstag, 12. Januar
1942, 442 - 452. Geneve: Droz.
Muller, Pete r 0.
2005
Einfiihrung. In: Peter 0 . Muller (ed.), Fremdwortbildung. Th eorie und Praxis in Geschichte und Gegenwart, 11- 45. Frankfurt/M.: Lang.
Newman, Paul
1979
Explaining Hausa feminines. Studies in African Linguistics 10(2): 197- 226.
Niedermann, Max
1953
Review of Jules Marouzeau (1951) Lexique de la terminologie linguistique fran<;ais,
allemand, anglais, italien. 3rd ed. Paris: Geuthner. Vox Romanica 13: 103-113.
Nyrop, Kristoffer
1908
Grammaire historique de la langue Jranr;aise . Vol. 3: Formation des mots. Copenhague:
Gyldendal.
Okamoto, Junji
2002
Particle-bound directions in German particle verb constructions. In: Hidekazu Suzuki
(ed.), Report of the Special Research-Project for the Typological Investigation of Languages and Cultures of the East and West. Part 11, 4 15-431. lbaraki: University of
Tsukuba.
Olsen, Susan
1996
Pleonastische Direktionale. In: G isela Harras and Manfred B ierwisch (edls.), Wenn die
Semantik arbeitet. Klaus Baumgartner zum 65. Geburtstag, 303-329. Tilbingen: Niemeyer.
Ortmann, Albert
1999
Affix repetition and non-redundancy in inflectional morphology. Zeitschriftfur Sprachwissenschaft 18( I): 7 6- 120.
Pantelidis, N ikolaos
20 I0
The reshaping of the mediopassive endings: Evidence from Modern Greek varieties . In:
Angela Ralli, Brian D. Joseph, Mark Janse and Athanas ios Karasimos (eds.), On-line
Proceedings of the Forth International Conference of Modern Greek Dialects and Linguistic Theory (MGDLT4), Chios, 11 - 14 June 2009, 3 15- 328. Patras: U niversity of
Patras.
Paul, Hermann
1880
Prinzipien der Sprachgeschichte. Halle/S.: Niemeyer.
Paul, Hermann
1920
Prinzipien der Sprachgeschichte. 5 111 ed. Halle/S.: Niemeyer.
Plank, Frans
198 1 Morphologische (Ir-)Regularitiiten. Aspekte der Wortstrukturtheorie. Tubingen: Narr.
Plank, Frans
On the reapplication of morphological rules after phonological rules and other resolutions
1985
of functional conflicts between morphology and phonology. Linguistics 23: 45- 82.
Prince, Ellen F.
1993
On the discourse functions of syntactic form in Yiddish: Expletive es and subject-postposing. In: David Goldberg, Marvin I. Herzog, Barbara Kirshenblatt-Gimblett and Dan
549
PA550
550

Miron (eds.), The Field of Yiddish. Studies in Yiddish language, folklore, and literature,
59-86. Evanston, IL: Northwestern University Press.
Rainer, Franz
Spanische Wortbildungslehre. Ti.ibingen: Niemeyer.
1993
Rehbein, Ines and Josef Genabith
German particle verbs and pleonastic prepositions. In: Boban Arsenijevic, Timothy
2006
Ba ldwin and Beata Trawinski (eds.), Proceedings of the Third ACL-SJGSEM Workshop
on Prepositions, 57- 64. Trento: Association for Computational Linguistics.
Rowlett, Pau l
1998
Sentential Negation in French . Oxford/New York: Oxford University Press.
Schwyzer, Eduard
1941
Sprachliche Hypercharakterisierung. Berlin: Verlag der Akademie der Wissenschaften.
Stolz, Thomas
2007
Word-internal agreement. Language Typology and Universals 60(3): 219-251.
Stolz, Thomas
2010
Pleonastic morphology dies hard: Change and variation of definiteness inflection in
Lithuanian. In: Franz Rainer, Wolfgang U. Dressler, Dieter Kastovsky and Hans Christian Luschi.itzky (eds.), Variation and Change in Morphology. Selected Papers from the
13 1" International Morphology Meeting, Vienna, February 2008, 2 17-244. Amsterdam/
Philadelphia: Benjamins.
Stotz, Peter
Handbuch zur Lateinischen Sprache des Mittelalters. Vol. 2: Bedeutungswandel und
2000
Wortbildung. Mi.inchen: Beck.
Stump, Gregory T.
1989
A note on Breton pluralization and the Elsewhere Condition. Natural Language and
Linguistic Theory 7(2): 261 - 273.
Thomason, Sarah
1988
Double marking in morphological change. In: Ann Miller and Joyce Powers (eds.),
ESCOL 87. Proceedings of the Fourth Eastern States Conference on Linguistics, 296305. Columbus, Ohio: Ohio State University.
Tovar, Antonio
Review of Eduard Schwyzer (1939) Die Parenthese im engern und im weitern Sinne;
1942
(1940) Syntaktische Archaismen des Attischen; (1941) Sprachliche Hypercharakterisierung. Berl in: de Gruyter. Emerita 10: 185- 188.
Wiemer, Bjorn
2009
Zu entlehnten Verbprafixen und anderen morphosyntaktischen Slavismen in litauischen
Insel- und Grenzmundarten. In: Lenka Scholze and Bjorn Wiemer (eds.), Von Zustiinden,
Dy namik und Veriinderung bei Pygmiien und Giganten. Festschrift fur Walter Breu zu
seinem 60. Geburtstag, 347- 390. Bochum: Brockmeyer.
Wierzbicka, Anna
2003
Cross-Cultural Pragmatics. The Semantics of Human Interaction. 2nd ed. Berlin: Mouton de Gruyter.
Wlodarczyk, Matylda
"More strenger and mightier": Some remarks on double comparison in Middle English.
2007
Studia Anglica Posnaniensia 43: 195- 2 16.
Wohlgemuth, Jan
A Typology of Verbal Borrowings. Berlin/New York: Mouton de Gruyter.
2009
Zwicky, Arnold
2012
Excrescent 's. http://arnoldzwicky.wordpress.com/2012/06/08/excrescent-s/ [last access
30 Sept 2012].
Francesco Gardani, Vienna (Austria)
PA552
552
2.1. The presuffixal interfix

The starting point is Malkiel 's definition of the interfijo:
[ ... ]el segmento, siempre atono y falto de significado propio, entre el radical y el sufijo de
ciertos derivados, p. ej. el elemento -ar- en hum-ar-eda, polv-ar-eda, palabras que no es
licito descomponer en humar- y polvar-eda, por no existir ni haber existido nunca, que
sepamos, las fases intermedias *humar, *po/var come formaciones independientes.
[[ ... ] the element, always unstressed and devoid of proper meaning, between the radical and
the suffix of some derivatives, e.g., the string -ar- in hum-ar-eda, polv-ar-eda, words that
cannot be broken down into humar- y polvar-eda since the intermediate stages *humar,
*po/var do not exist and have never existed as independent formations, as far as we know.]
(Malkiel 1958: 107)
Before Malkiel, other words had named the interfix itself or the combination of interfix
plus suffix (and some are still preferred by linguists who do not acknowledge the notion
of interfix). Darmesteter (1890), for instance, used "suffixes intercalaires" [interposed
suffixes] for such examples as Fr. gant-el-et 'gauntlet' +-- gant 'glove'; Prati (1942),
"antisuffissi" [presuffixes] (e.g., It. acqu-er-ello 'watercolour ' +-- acqua 'water '); Pottier
(1953) and Badia Margarit (1962), "infixes" (e.g., Pg. sant-arr-ii.o 'zealot' +-- santo
'saint', Cat. branqu-ill-6 'twig' +-- branca 'branch'); Adams (1913), "compound suffixes" (e.g., Oc. pan-at-ier 'baker' +--pan 'bread'); Graur (1969), "suffixes elargis" [enlarged suffixes] (e.g., Rom. lung-ar-et 'long-D IM' +-- lung 'long'); etc.
Most of these denominations are ambiguous and confusing. Adams distinguishes
"compound suffixes", which are the combination of an interfix and a suffix, from "real
double suffixes" (which are the result of two consecutive derivations), but he has no
specific term for the interfix itself. Pottier calls " infixes modificateurs" the evaluative
suffixes themselves as well as the interfixes which precede some of them. In the French
tradition, "formes elargies" can designate variants of a suffix as well as the combination
of a suffix and an interfix. The first interest of Malkiel 's founding paper was to identify
a specific morphological object and name it.
Unfortunately, the word inter.fix has thereafter been used for other inserted elements:
the non-etymological velar segment in Spanish verbs such as tengo (Dworkin 1995), for
instance, or the marker -i-1-esc- in the so-called inchoative conjugation (Allen 1977). It
is widely spread, inside and outside Romance, for the linking element of compounds. In
the meantime, for the interfix as defined by Malkiel, the word did spread in studies on
Ibero-Romance, but the notion itself was bitterly discussed. As regards Italian, the term
and its content were adopted by the followers of natural morphology (Dressler and
Merlini Barbaresi 1989; Crocco Galeas 1991; Merlini Barbaresi 2004), and rejected or
ignored by most other morphologists. In Gallo-Romance, constructions featuring the
equivalent of a Spanish interjijo had not really been examined until our recent studies.
The etymologies of the Tresor de la langue fi'aru;aise (TLF) exhibit this deficiency. For
instance, Fr. briquetier 'brickmaker', pelletier 'furrier', malletier 'trunk-maker', which
are obviously constructed after the same pattern (the bases are the simplices brique
'brick', peau 'pelt', malle 'trunk') receive different analyses: briquetier as derived from
brique with the suffix -ier and the intercalation of the consonant -t-; pelletier from OFr.
pel with the extended suffix -(et)ier; malletier from mallette ' briefcase'.
PA554
554

a. porta
b.
'door, gate'
~
~
portina
portinaio
'small door'
'doorman, gatekeeper'
Italian
Brickmakers do not make only (or especially) small bricks and the name of the Big Bad
Wolf cannot be constructed on that of the cubs. (3a) and (3b) are parallel derivations
from the same base.
In (4a), 0 gouttelle and 0 gouttette would separately be plausible (cf. coupelle 'small
cup'+--- coupe 'cup'; chafnette 'small chain'+--- chafne 'chain'), but a two-step derivation
goutte ~ 0 gouttelle ~ gouttelette is highly improbable since the meaning of the actual
derivative gouttelette is the same as that of 0 gouttelle and 0 gouttette.
(4)
a.
goutte
'drop'
~
~
~
b.
gos 'dog'
0
0
gouttelle
gouttette
gouttelette
gosset
' puppy'
'droplet'
'droplet'
'droplet'
~
gosseton
French
'puppy-DIM'
Occitan
The consecution of two evaluative derivations, which is frequent in meridional Romance

(4b), is unusual in French and gouttelette does not correspond to this pattern.
The crucial point in Malkiel 's definition is not "por no existir" - a form could exist
without being presently attested - but "falto de significado propio". IIll spite of its form,
the interfix is not a suffix which could act as such in a derivational chain. It is an empty
morph.
2.3. The presuffixa l interfix is not an infix

It is generally assumed that an infix is an affix which is inserted into a lexeme. Regarding
only the derivative, presuffixal interfixes could be viewed, superficially, as infixes. In
fact, they differ from infixes in several ways. They are not inserted inside the root, but
between the root and a suffix. While infixes, like prefixes and suffixes, have a form of
their own and a derivational or inflectional value, interfixes do not: they are always
neutral elements and, in most cases, have the form of a suffix.
The word infix is no longer used systematically for interjix, but both terms can still
be found in the analysis of various constructions for which neither is really appropriate.
In our view, there are some cases of infixed elements in Romance, such as the often
scrutinized Spanish "infixed diminutives" (e.g., Victitor +--- Victor), but no infixes properly speaking. French trompinette 'small trumpet' +--- trompette 'trumpet' and poissillon
'small fish' +--- poisson 'fish', for instance, exhibit another kind of anomalous infixation,
due to the frequency of the endings -inette and -illon in the lexicon, but the string -inin trompinette is the same affix as the suffix -ine in trompettine, which is also attested,
and the string -ill- in poissillon is the same affix as the suffix -ille infaucille 'sickle' +faux 'scythe '. They cannot be analysed as interfixes, in spite of their position before a
suffix-like ending, since they retain their full evaluative value. In Spanish, Fabregas
(2006: 17) calls "infix" the string -et- in corretear 'to scamper' +--- correr 'to run' while
Portoles ( 1999: 5066) lists it among interfixes and Lazaro Mora ( 1999: 4649) among
suffixes. The question of verbal suffixation is too intricate to be discussed here (cf.
PA555
31. Interfixes in Romance
SSS
article 84 on word-formation and lexical aspect: deverbal verbs in Italian), let us just
say that we prefer to view the full string -etea(r) as a suffix, since it bears the evaluative
value of the derivation. But in despernancarse 'to spread one's legs' +-- pierna ' leg', for
instance, -anc- has no proper meaning and can be considered as an interfix.
2.4. lnterfi x plus suffix vs. suffix va ri ants

Since the base of Fr. briquetier, for instance, is not briquette but brique (cf. examples
(3)), an easy solution would be to consider -etier as a variant of the suffix -fer. The
principal argument against this analysis, since Malkiel, has been descriptive economy:
each suffix would have too many variants. Owing to the multiplicity of combinations
between interfixes and suffixes, it is simpler to list them separately.
An inventory of interfixes in the different Romance languages would exceed the
length of this article, a fortiori an inventory of the combinations between interfixes and
suffixes (for a partial one, Roche 200S). Their numbers are indeed considerable. In
Spanish, a conjunction of Portoles' (1988, 1999) and Rainer 's (1993) lists - which do
not coincide - comes to 473 combinations. The sole -ar- can precede 28 different suffixes (Malkiel l 9S8: 178- 184) and -6n be preceded by 4 1 different interfixes (Portoles
1999: S044). In French, where interfixation is far less frequent, a dozen different interfixes can be found before the suffix -on (Roche 2003a), a dozen before -ette in contemporary attestations (Plenat 200S), a score before -ierl-iere (Plenat and Roche 2004). All of
them also appear before other suffixes. Nevertheless, as pointed out by Rainer (1993:
l S4 ), the argument of economy is not decisive. If there were good reasons to regard
each combination of interfix plus suffix as a variant of a suffix, morphology ought to do
so, whatever their numbers. But more important than the numbers themselves is the
versatility of interfixes. They have a broad independence from suffixes within the apparatus of word construction, whereas variants are specific to each suffix and closely associated to its history.
Most variants originate in a reanalysis - a coalescence of two suffixes or a miscut
between radical and suffix. Fr. -erie and -eraie typically exemplify two stages in the
development of variants: the former superseded -ie long ago, the latter is rivaling -aie
in a growing proportion. Originally, the ending -eraie, like -erie, was used in derivatives
constructed on bases ending in -ier (Sb), the string -er- being a part of the radical. A
miscut leads to adding it to the suffix, as in (Sc) (Roche 2011).
(S)
a. chene 'oak tree' ___.. chenaie 'oak grove'

b. [rose 'rose' ___..] rosier 'rosebush' ~ roseraie 'rose garden'
c. bambou 'bamboo' ___.. bambouseraie 'bamboo grove'
French
French
French
The string -er- is used as an interfix in other derivatives (e.g., puceron 'aphid' +--puce
'flea', paquerette 'daisy' +-- Paques 'Easter', hachereau 'hatchet' +-- hache 'axe', secheresse 'drought' +-- sec 'dry', etc.) but there are two important differences: (i) before
-on, for example, -er- is one interfix among others (e.g., bottillon '(short) boot' +-- botte
'boot', palichon 'peaky' +-- pale ' pale', gu,euleton 'feast' +-- gu,eule 'mouth', etc.),
whereas -eraie is the only variant of -aie (as -erie is the only variant of -ie); (ii) interfixes
PA556
556

appear after monosyllabic bases only (cf. section 3.2), whereas -eraie (as other suffixal
variants) may follow any radical (for other examples, Roche 2003 b, 2009).
The process which leads to the diffusion of interfixes (cf. section 3.4) is similar to
the one which generates suffix variants. It is natural that the distinction between the two
kinds of items is not clear-cut. The recurrence of the same interfix before the same suffix
leads to the formation of a variant. The Romanian evaluative -ef, for instance, is preceded
by -ul- in an overwhelming majority of derivatives (Graur 1969: 330), so that -ulef seems
to have become not only a variant of -ef but the standard form of tlhe suffix. But we
shall see that there are also good reasons to regard the interfix as an extension of the
stem. Even though many cases are subject to discussion, the combinations of interfix
plus suffix cannot be analysed merely as suffix variants.
2.5. Base + interfix vs. base allomorphy and epenthesis

Symmetrically, the combination of base plus interfix could be assimilated to base allomorphy. At first sight, the morphophonological accidents in Fr. bombinette 'bomb-DIM'
- bombe 'bomb' and Fr. criminel 'criminal' - crime 'crime' are similar: the string -inwhich appears in the derivative is not present in the free form of the base. Historically,
the difference is obvious: the allomorph crimin- is inherited from La6n and is the only
one for this lexeme, whereas a diminutive of bombe can be coined with another interfix
(bombelette and bombounette are also attested). Synchronically, we shall see that both
crimin- and bombin- can be regarded as the stems on which criminel and bombinette are
constructed. So the boundary between interfixation and stem allomorphy is not always
clear-cut. But an important difference remains: inherited allomorphy can add a syllable
to the stem regardless of its length (cf. Fr. longitudinal - longitude, for instance),
whereas an interfix is added to monosyllables (and a few disyllables) only (cf. section
3.2).
Epenthesis is another means of modifying a stem, and epenthetic consonants are
listed by Malkiel and his followers among interfixes. They have in common with (other)
interfixes that they are semantically devoid of meaning. But they differ from them in an
important point: they do not add a syllable to the stem. In most cases, their raison d'etre
is the same as that of latent consonants: to provide an onset to the suffix, in order to
satisfy the antihiatic constraint. The only differences between the three examples in (6)
lie in the fact that the final consonant of the stem is graphical and etymological in (6a),
graphical and non -etymological in (6b), neither graphical nor etymological in (6c).
(6)
/pol - /p'Jt/
'pot'
a. pot
b. canot /kano/ - /kan'Jt/ 'canoe'
c. echo /eko/ - /ek'Jt/
'gossip'
~
~
potier
'potter'
canotier 'boater'
echotier 'gossip columnist'
French
French
French
This is the reason why epenthetic consonants, unlike presuffixal interfixes, are inserted
irrespective of the length of the base (e.g., numeroter 'to number ' - numero 'number').
However, in spite of the differences between the two phenomena, a continuity between epenthesis and interfixation can be observed in some of the objects in which they
are embodied. The choice of /el as an antihiatic consonant in (7a) and its insertion after
a sonorant in (7b) are closely akin to the choice of -ec- as syllabic interfix in (7c).
PA557
(7)
a.
b.
c.
cafe 'coffee'
rat6n 'mouse'
tren 'train'
557
--
cafecito 'coffee-DIM'
ratoncito 'mouse-DIM'
trenecito ' train-DIM'
Spanish
Spanish
Spanish
Being nothing else than phonological material, epenthetic consonants need no other label
than epenthetic consonants. To call them interfixes is a matter of metalinguistic conventions, but in all cases they must be di stinguished from syllabic interfixes.
This survey of different constructions leads to the conclusion that to be or not to be
an interfi x is neither a matter of presence (or absence) in a list of morphological objects
nor a matter of position in the derivative: it depends on the role of the item in a given
construction.
3. I nterfixes: What? Why? Where?

Having distinguished the presuffixal interfix from other morphological objects, we shall
now try to clarify what it is in itself and what role it plays in the economy of word
construction.
3.1. The morphemic approach

The Spanish controversy (Lazaro Carreter 1980 [ 1972]; Martinez Celdran 1978; Montes
1985; Dressler 1986; Portoles 1988, 1999; Martin Camacho 2002, etc.) has focused for
a great part on the morphemic quality of the interfix. To be recognized as an entity of
its own, the interfix was supposed to be a morpheme. And consequently, to have the
counterpart of its signifiant, in an orthodox Saussurean view: a signifie. Or, at least,
another function which could legitimate its morphemic status.
Portoles (1999) devotes a section to the meaning of Spanish interfixes ("El significado de los interfijos"), but acknow ledges that this meaning can be di scerned in some of
the interfixes only and is due to "un impreciso sema que se cunfunde con el que aporta
el sufij o" [a vague seme which merges with that of the suffix] (p. 5058). For Dressler
and Merlini Barbaresi ( 1989: 246), Spanish interfixes have no meaning, but Italian ones
may have connotative undertones ("sfumature connotative"). Actually, it is somewhat
contradictory to admit Malkiel 's defi nition of the interfix as devoid of mean ing - all
authors start from these premises - and intend to ascribe a meaning to it. The strings
which look like interfixes and have a semantic value of their own occur in other kinds
of constructions. For instance, Oc. femnassier 'womanizer' (+-- femna 'woman') is both
agentive and pejorative, the agentive value being borne by -ier, the pejorative one by
-ass-. A two-step derivation (femna - femnassa 'fat and unattractive woman' - femnassier) would not correspond to the denotatum: the pejoration does not regard the
chased woman but her chaser. Thus both suffixes, -as and -ier, are full suffixes but act
simultaneously, as a suffix cluster. Fr. noblaillon ( +-- noble 'noble'), chef!aillon ( +-- chef
'superior, manager'), poetaillon ( +-- poete ' poet'), etc., which are all three pejorative
(through -ail/-) and diminutive (through -on), could receive the same analysis (Roche
PA558
558

2009: 165- J 67). These are indeed marginal, anomalous formations, but it would not be
coherent to see interfixes in them.
Among the other functions attributed to interfixes is the distinction of possible homonyms, if bo,t h derivatives were constructed without an interfix: Sp. llamada 'call' +llamar 'to call' I llamarada 'blaze' +--llama 'flame'; Oc. porquier 'pig-keeper' I porcatier 'pig-merchant' +--pore 'pig'; It. volpino 'spitz' I volpicino 'fox cub' +-- volpe 'fox'.
But the contrast in meaning is not connected to the choice of a particular interfix, and
much more numerous are the examples of synonyms constructed with and without interfix (e.g., Oc. gatiera I gatoniera 'cat flap' +-- gat 'cat') or with different interfixes (e.g.,
Oc. vimenier I vimotier 'osier grove' +-- vim 'osier'). The distinction of homonyms is a
side effect of the presence of an interfix, not the reason of its insertion.
More convincing is the idea that an interfix could adapt a base for a derivation in
which the base normally does not enter. Portoles (1999: 5056) gives the example of the
Spanish suffixation with -6n, which usually selects feminine nominal bases: interfixation
enables it to select masculine or adjectival ones (e.g., Sp. grande 'tall' ---+ grandull6n,
-ona 'tall person'). In several Romance languages, the interfix -and-I-end- helps a verbal
base to receive a suffix which is usually dedicated to nominal bases: Sp. curandero
'witch doctor' +-- curar 'to cure'; Oc. teissendier 'weaver' +-- teisser 'to weave'; Fr.
lavandiere 'washerwoman' +--!aver 'to wash ' ; It.filandaia 'spinster'+-- filare 'to spin'.
However, the same suffixes can be added to verbal bases with other interfixes, or with
no interfix at all, and the same interfix -and-I-end- be used in quite different contexts.
These "functions" would concern, furthermore, a minority of the interfixes. As a
whole, the d ispute about the morphemic status of the interfix leads to a stalemate. The
linguists who stick to a strict definition of the morpheme (e.g., Tekavcic 1968: 77: "Un
segmento 0 morfema 0 non lo e" [either a segment is a morpheme, or it is not]) reject
the notion of interfix, but they can hardly account for the constructions where others see
an interfix. And the defenders of the interfix have to place it in a particular subcategory "morfema marginal" [marginal morpheme], "morfema residual" [residual morpheme]
(Malkiel 1958: 185, between double quotes); "morfemas vacios" [empty morphemes]
(Dressler 1986: 384) - which makes the notion inconsistent. It is better to recognize
interfixes plainly as empty morphs. As pointed out by Maiden (1999), autonomous morphology has cast a new light on such morphological objects and is more apt to acknowledge their role, if they are put back in the phonological and lexical dimensions of wordformation.
3.2. The pho nologica l approach

In its very definition, Malkiel ( 19 5 8: 107) notes that the Spanish interfix is always
unstressed ("siempre atono"), add ing a few years later (1970: 75): "As interfixes are, by
definition, pretonic [ ... ] they offer a strong [ ... ] appeal to the speakers' sense of rhythm
(syllabic count) [ ... ]". In an interfixed derivative, a secondary accent is kept on the
syllable which was stressed in the base, and would otherwise be totally unstressed (e.g.,
Sp. tos 'cough'---+ tosegoso l ,tose'gosol 'having a chronic cough' vs. 0 tososo l to'sosol). In
some particular cases, the interfixation blocks an a lteration of the base and facilitates its
recognition (e.g., Sp. tierno 'tender' ---+ tiernecillo 'tender-DIM' vs. 0 ternillo, cf. ternura
PA559
559

'tenderness'). As pointed by Allen (1976) in the title of his artic le, " [i]nterfixes preserve
syllables and word roots". However, this "function" of the interfix (Dressler 1986; Portoles 1999) must not be overestimated. The gain in "naturalness" is paid w ith a loss of
faithfulness due to the introduction of a meaningless element.
More important are the observations that Spanish interfixes "sirven para agrandar la
pieza radical de una palabra antes del sufijo" [are used to lengthen the radical part of a
word before the suffix] (Dressler 1986: 388) or that in Italian "parole trisillabiche evitano
l' inserzione di interfissi" [trisyllabic words preclude the insertion of an interfix] (Dressler and Merlini Barbaresi 1989: 245). The examples which follow these assertions show
that thanks to the interfi x, derivations based on disyllables in (8a) can match the prosodic
pattern of derivations on trisyllables in (8b).
(8)
a. padre
b. compadre
'father '
'friend'
a. serpe
b. serpente
'snake'
'snake'
---
padrecito
compadrito
'father-DIM'
'friend-DIM'
serpicella
serpentello
'small snake'
'small snake '
Spanish
Spanish
Italian
Italian
Detailed investigations of extensive lexical databases have led to more systematic conclusions for Occitan and French. In a corpus of 1503 derivatives with Oc. -ier/-iera, 207
are constructed with an interfix, among which 199 have a monosyllabic base (or a disyllable whose unstressed final vowel has been regularly deleted), 8 a disyllabic one, and
none a longer one (Plenat and Roche 2004). Among the 8 disyllabic bases, 7 have a
vocalic initial. This particularity confirms that interfixation is closely related to the prosodic constitution of the base: the only di syllables which receive an interfix are not, in
fact, full disyllables by virtue of being deprived of their first onset (Plenat 1997).
Table 31. l: Prosodic conditioning of interfixation in Occitan and French
suffixation
with ...
base
total
I syllable
interfixes
base
2 syllables
base 2: 3 syllables
total
interfixes
total
interfixes
Oc. -ierl-iera
695
199
29%
673
1%
135
0%
Fr. -on
393
133
34%
188
2%
14
0%
Fr. -ette
1382
434
31 %
984
26
3%
247
2%
The results are similar in a corpus of 595 derivatives with Fr. -on from dictionaries
(Roche 2003a) and in a corpus of 2613 derivatives with Fr. -ette collected on the Web,
most of them neologisms (Plenat 2005) (Table 31. 1). The only interfixed derivations
with bases longer than one syllable are particular cases: hypocoristic person names, bases
with a vocalic initial, or stems w ith a final /t / before -ette.
It is difficult to determine whether the prosodic template concerns the stem (two
syllables) or the derivative (three syllables in French), since the result would be the
same. According to Lazaro Mora (1999: 4665), the presence of the interfix -ec- (vs. -cor no interfi x) in Spanish derivatives with -ito (cf. examples (7)) is due to "el control
de las dimensiones silabicas de los derivados diminutivos" [the control of the syllabic
PA563
563
Fr. fourniture +-- fournir 'to supply'; OFr. garneture, Fr. garniture +-- garnir 'to fill';
etc.), and on the other to the formation of the suffix variant -ature (in Fr. musculature
+-- muscle 'muscle', for instance).
Let us add that, more generally, the derivational series play a part in the above mentioned prosodic constraint. The disyllabic optimum of the stem is probably a consequence of the predominant structure of derivatives in the lexicon, which tends to be
generalized. Horcajada (1987-88) argues that the form of Sp. panecito ' bread-DIM' +pan 'bread', for instance, is due to historical accidents, not to a prosodic pattern. But
this does not rule out Lazaro Mora's generalization about the interfix -ec- (cf. section
3.2): panecito (after the more ancient panecillo) may have been one of the leader words
to have started a sub-series. The majority of derivatives ending in -ec-ito do not have a
Latin origin. To say that an interfix is added to fit a prosodic pattern is not a teleologic
explanation - an approach opportunely put into question by Rainer (1993: 156) - but
an observation about the lexical dynamics of derivation.
The other lexical paradigms which play a part in interfixation are the derivational
families. Frequently, the interfix is selected among the suffixes which have already been
used after the same base (cf. examples (3)). We said that Oc. lobat 'wolf cub' cannot be
the base of lobatas 'wolf-AUG' . But the choice of -at- as an interfix in lobatas, as well
as in lobatiera 'wolf den ' , is most probably related to the presence of lobat in the
derivational family of lop 'wolf'. In the family of Sp. casa 'house', the augmentative
caser6n owes its interfix to the adjective casero, which normally plays no part in its
formation, while symmetrically, in the family of calle 'street', the adjective callej ero
owes its interfix to the diminutive calleja, with no more semantic reason. The process
is less visible when the first derivative has disappeared. According to Malkiel (1949),
the origin of the rare interfix -egin Sp. p edregoso ' stony', p edregal 'stony area' (+piedra 'stone') is to be found in the remote Latin derivative 0 p etricare 'to pave'.
When the meaning of a previous derivative is close to that of the primitive, both can
be (lla), or possibly could be (11 b ), accepted as the base.
(11)
a.
b.
brer; I brer;ol 'craddle' ---+ brer;oliera 'nanny'

Occitan
mule 'she-mule' I mulet 'he-mule ' ---+ (chemin) muletier 'mule track' French
mare 'sea' I marina 'sea coast', 'navy' ---+ marinaio 'seaman'
Italian
Received etymologies give muletier as derived from mulet and marinaio from marina.
But the lexical base of the former is more logically the generic term mule, and the
meaning of the latter more logically constructed on the basis of mare than on any of the
acceptions of marina. In all cases, the impo1iant point is that the choice of the stem is
triggered by the prosodic constraint, as confirmed by the examples in (12): the adjective
is constructed on a previous derivative when the primitive is monosyllabic (12a), on the
primitive itself w hen it is polysyllabic (12b).
(12) a. Jorn I jornada
mes I mesada
an I annada
b. setmana I setmanada
'day'
' month '
'year'
' week'
---
jornadier
mesadier
annadier
setmanier
'daily'
'monthly'
'annual'
'weekly'
Occitan
Slightly different is the case of Sp. condesil 'relating to a count' . The base cannot be
condesa 'countess', since the adjective is semantically connected to conde 'count' , but
PA565
- Presuffixal interfixes do not form a class of derivational affixes. They are empty
morphs which recycle the form of a suffix (or a former suffix) deprived of its semantic
or categorial instruction.
- Interfixation is not a derivational process which could be placed on a par with suffixation or prefixation. It is a means - among others - of modifying the stem presented
by the base lexeme. The interfix adds a syllable to the stem and offers a more suitable
onset to the suffix (while epenthetic consonants fu lfill the second requirement only).
- In such an approach, the notion of interfix is rid of the outdated debate on its morphemic status and can take its place in a lexeme-based morphology.
- The addition of an interfix to the stem and the choice of the interfix depend upon
phonological and lexical parameters. The former are both prosodic and segmental, the
latter are associated with both paradigmatic dimensions of derivation - derivational
series and derivational families.
- lnterfixation has no teleologic "functions" and cannot be accounted for by word formation rules. It is the consequence of two kinds of forces which come into play in
word-formation: phonological well -formedness constraints (presumably universal)
and the influence of the inherited lexicon (proper to each language).
- Interfixes are present in all Romance languages and are not only a collection of historical oddities, the flotsam and jetsam of morphological accidents. They take their place
among other stem adjustments, alongside consonantal epenthesis, truncation, substitutions, etc., in contemporary formations as in past ones. But their presence differs
remarkably from one language to another, from one period to another, from one sector
of the lexicon to another. Additional studies based on extensive data are necessary in
order to determine which interfixes are still productive and to what kinds of suffixation
they are associated.
5. References
Adams, Edward L.
1913
Word Formation in Prove111; al. New York: MacMillan.
Allen, Andrew
1976
Interfixes preserve syllables and word roots. In: Henry Thompson, Kenneth Whistler,
Vicki Edge, Jeri J. Jaeger, Ronya Javkin, Miriam Petruck, Christopher Smeall and Robert D. Yan Valin Jr. (eds.), Proceedings of the Second Annual Meeting of the Berkeley
Linguistics Society, 31-35. Berkeley: Berkeley Linguistics Society.
Allen, Andrew
1977
The interfix ilesc in Catalan and Rumanian. R omance Philology 31: 203- 211.
Aronoff, Mark
1994
Morphology by Itself. Stems and Inflectional Classes. Cambridge, MA: MIT Press .
Badia Margarit, Antoni M.
1962
Gramatica catalana. Madrid: Gredos.
Bauer, Laurie
A Glossary of Mo1phology. Edinburgh: Edinburgh University Press.
2004
Bosque, Ignacio and Violeta Demonte (eds.)
Gramatica descriptiva de la lengua espaffola. Madrid: Espasa Calpe.
1999
Crocco Galeas, Grazia
Gli etnici italiani. Studio di morfologia naturale. Padova: Unipress.
1991
565
PA566
566

Darmesteter, Arsene
1890
Traite de la formation de la langue frans:aise. ln: Adolphe Hatzfeld, Arsene Darmesteter
and Antoine Thomas (eds.), Dictionnaire general de la langue fran<;aise du commencement du XVIr siecle a nosjours. Vol. l , 1- 300. Paris: De lagrave.
Dressler, Wolfgang U .
1986
Forma y funci6n de los interfijos. Revista Espanola de Linguistica 16: 381 - 395.
Dressler, Wolfgang U . and Lav inia Merlini Barbaresi
1989
Interfissi e non-interfissi antesuffissali nell'italiano, Spagnolo e inglese. In: Societa di
Linguistica Italiana (ed.), L'italiano tra le lingue romanze, 243 - 252. Roma: Bulzoni.
Dworkin, Steven N.
1995
Two studies in Old Spanish homonymics. Hispanic Review 63: 527-542.
Fabregas, Antonio
2006
Infixes: Right in the middle. SK.ASE Journal of Theoretical Linguistics 3(3): 12- 29.
Faitelson-Weiser, Silv ia
1993
Sufijaci6n y derivaci6n sufij al: Sentido y forma. In: Soledad Varela (ed.), Laformaci6n
depalabras, 119- 161. Madrid: Taurus Universitaria .
Graur, A lexandru
1969
Suffixes roumains elargis. Revue Roumaine de Linguistique 14: 327- 332.
Horcajada, Bautista
1987-1988
Morfonologia de los diminutivos fo rmados sobre bases consonanticas monosilabas. Revista de filologia romanica 5: 55-72.
Lazaro Carreter, Fernando
1980 [1972]
Sobre el problema de los interfijos: lConsonantes a ntihiaticas en espaf\ol? Jn:
Fernando Lazaro Carreter, studios de Linguistica, 11-26. Barcelona: Critica.
Lazaro Mora, Fernando A.
1999
La derivaci6n apreciativa. In: Ignacio Bosque and Vio leta Demonte (eds.), Gramatica
Descriptiva de la Lengua Espanola. Vol. 3, 4645-4682. Madrid: Real Academia Espanola, Espasa.
Maiden, Mart in
1999
Romance historical morphology and empty affixes. In: Sheila Embleton, John E. Joseph
and Hans-Josef Niederehe (eds.), The Emergence of Modern Language Sciences. Studies
on the transition from historical-comparative to structural linguistics in honor of E. F
Konrad Koerner. Vol. 2, 189-202. Amsterdam/Philadelphia: Benjamins.
Malkiel, Yakov
1949
Studies in the Hispanic infix -eg-. Language 25: 139- 181.
Malkiel, Yakov
1958
Los interfijos hispanicos: Problema de la lingi.iistica hist6rica y es.tructural. In: Diego
Catalan (ed.), Miscelanea homenaje a Andre Martinet. Vol. 2: Estructuralismo e historia, 107- 199. La Laguna: Universidad de La Laguna.
Malkiel, Yakov
1970
Patterns of Derivational A/fixation in the Cabraniego Dialect of East-Central Asturian.
Berkeley: University of California Press.
Martin Camacho, Jose Carlos
2002
El problema linguistico de los interfijos espanoles. Caceres: Universidad de Extremadura, Servicio des Publicaciones.
Martinez Celdran, Eugenio
1978
En tomo a los conceptos de interfijo e infijo en espafiol. Revista Espanola de Linguistica
8: 447-460.
Mascaro, Joan
1986 M01fologia. Barcelona: Enciclopedia Catalana.
PA569
32. Linking elements in Germanic
1. Introduction
Linking elements are sound material or graphemes appearing at the boundary between
two parts of a word-formation product, cf. -s in German Versicherung-s-vertreter ' insurance agent'. They are usually considered to be meaningless items, but hypotheses about
possible functions exist. Linking elements are frequently found in various Germanic
languages, especially in nominal compounds. We provide a detailed sketch of the linking
system in German in section 2. Section 3 contains a comparative survey of other Germanic languages.
2.
Linking elements in German
German Versicherung-s-vertreter is a compound that is made up of two stems, just as

its English translation insurance agent is. It is worthwhile to note that the first stem does
not surface as Versicherung, but as Versicherungs. Fuhrhop (1998) introduces the term
"compositional stem form" for this form. In this sense, all words that form compounds
have compositional stem forms - in most cases these are equivalent to the base form,
i.e. they exhibit a "zero link" as in Schrank-tiir 'cabinet door'. According to miscellaneous counts, 70 % of all German compounds do not employ any kind of linking element
(Ortner et al. 199 1; Baayen et al. 2007; Kurschner 2003 ), hence about 30 % of compounds do so. It is quite possible that the same stem may exhibit different compositional
stem forms as in Mann-es-kraft, Mann-s-bild, Mann-deckung, Miinn-er-verein, the last
one including umlaut. However, although variation between -s and zero link is quite
common (cf. Hauptseminar-(s)-arbeit), the number of stems forming compositional stem
forms with different linking elements is rather small.
Recent research has shown an interest in functional questions, i.e. questions of which
individual or consistent functions linking elements fulfill. There is general agreement on
three assumptions: First, linking elements are not merely extant inflectional elements;
second, some linking elements are productive; and third, they fulfill one or more functions which are to be investigated.
In the following, we wi ll first introduce the inventory of linking elements in German,
establishing a classification (section 2.1 ). Afterwards, we will address functional questions (section 2.2).
2.1. A classification of linking elements in Germ an

2.1.1. Form and i nventory of linking elements in German
Hereafter, only linking elements of nominal first parts are considered. We find the following linking elements: -s, -es, -n, -en, -ens, -er, and -e (Augst 1975). These elements
differ in frequency of occurrence. As far as the origin of linking elements is concerned,
even in today's system, it is evident that linking elements are formally identical to inflectional suffixes in German. It therefore seems plausible to trace the phenomenon of link-
569
PA571
571
Paradigmatic and non-paradigmatic linking elements are discriminated from each other
based on the infllectional system. A linking element is paradigmatic if it is formally
identical with an inflectional ending occurring in the stem's inflectional paradigm. A
linking element is non-paradigmatic if it does not constitute an inflectional ending of the
word involved (cf. Versicherung-s-vertreter, but * Versicherungs). There is only one linking element that is both non-paradigmatic and productive, namely -s; in addition, it is
the only linking element that is not syllabic.
2.2.
Fu nction of linking elements in German
2.2.1. Linking element -s in German
a) Konig-s-mord, Teu.fel-s-kuche - Liebling-s-getrank

b) Freundscha.ft-s-dienst, Versicherung-s-vertreter, Schonheit-s-wettbewerb,. Prufling-snummer
c) Religion-s-unterricht, ldentitat-s-bescheinigung
d) ??Auto-s-verkau.fer - Plastikauto-s-verkau.fer, Kuckuck-s-uhr
e) ??Seminar-s-arbeit - Hauptseminar-s-arbeit
f) Fahrt-zeit - Ab.fahrt-s-zeit, Suchan.frage - Versuch-s-aujbau, Triebwagen - Betriebs-wirtscha.ft
Using the examples listed in a)-f) above, we would like to demonstrate the various
functions under discussion.
a) lists cases that might account for genitive constructions like Konig-s-mord - des
Konigs Mord ([Kiing-LE] -homicide 'regicide' ). However, this cannot apply to Liebling-sgetrank because it does not mean ' darling's drink' but 'favorite drink'. Hence, genitive
marking cannot be a general function of the s-link: first, not all s-links can be taken as
such (e.g., Liebling-s-getrank), second, -s is not always employed even if there is a
genitive relation inv olved, cf. Ap.fel-kern (not *Ap.fel-s-kern) 'apple core', see also section 3.2.2.
b) lists words with suffixes that commonly take linking elements. Aronoff and Fuhrhop (2002) identify these as "closing suffixes" (-heit, -ung, -ling, etc.) - suffixes that
are not open for further word-formation suffixes, e.g.,
(2)
beobachten
observe-INF
to observe
> beobacht> observe-STEM
> Beobacht-ung
> observe-NOUN
> observation
> *beobacht-ung-lich
> observe-NOUN-ADJ
In these cases, the linking element reopens closed stems for further word-formation
processes, i.e., further compounding, and to some extent also derivation, cf. .frohling-sha.ft 'springlike'.
A different account of these suffixes comes from Eisenberg and Sayatz (2004) who
outline a continuum of suffixes and demonstrate that -ling, -scha.ft, -tum, etc. are adjacent
to each other if projected on a scale of animacy. This semantic scale coincides with
prosodic properties: Midway between the poles (where the collective suffixes in question
PA572
572

are positioned), suffixes are stressable and phonologically heavy. Close to the poles, by
contrast, suffixes contain unstressable schwa (the agent noun suffix -er and the plural
suffixes). Following Eisenberg and Sayatz (2004), the form of the suffixes is not accidental. This observation can be connected to the findings of Niibling and Szczepaniak
(2009), who demon strate that linking elements are conditioned by prosodic features (see
below). Given that closing suffixes generate precisely those prosodic structures which
Niibling and Szczepaniak describe, it seems that the condition described by Aronoff and
Fuhrhop (2002) is only indirectly relevant, i.e. the meaning condlitions the (non-)
derivability as well as the prosodic weight of closing suffixes and thus the specific
behavior with respect to linking elements.
c) lists suffixes in loan words that share the characteristics of the linking elements
found in b ), cf. Niibling and Szczepaniak (2009).
d) lists cases that were highly debatable for a long time with respect to their grammaticality. While it was often argued that linking-s is impossible if the compositional stem
form is homonymous with a plural form (* Auto-s-bahn, cf., e.g., Wiese 2000), Wegener
(2003) demonstrates empirically that the s-link is not primarily conditioned by inflectional, but by phonological factors: It is prohibited if the first part of the compound ends in
a full vowel (Auto 'car', Kina 'cinema', Uhu 'owl'). This is exactly the condition in
which s-plural most often applies (cf. Eisenberg 2006: 164). Conversely, if the stem
ends in a consonant, the s-link is generally more acceptable (Kuckucksnest, -ruf, -schrei
'cuckoo's nest/cry/scream' vs. *Uhusnest, -ruf, -schrei 'owl nest/cry/scream', cf. Wegener 2003: 434). The acceptability of the -s-link is increased additionally when the first
part of the compound is complex (cf. Wegener 2003: 445):
(3)
?Auto-s-verkiiufer
?[car-LE]-salesman
[Plastik-auto-s]-verkiiufer
[plastic car-LE] salesman
e) lists doubtful cases that are covered by Niibling and Szczepaniak (2011). The first
parts bear final stress, cf. above.
f) lists cases showing that the probability of linking elements increases if the first
element is prosodically and/or morphologically complex.
Based on compounds with varying linking elements, Niibling and Szczepaniak (2008)
have demonstrated that linking elements are primarily phonologically and only secondarily morphologically conditioned. According to Niibling and Szczepaniak (2008), the
"worse" the phonological word is, the more often -s is selected. In German, good phonological words conform to the ideal foot which is formed by a trochee. In addition to this
criterion, monosyllabics also constitute good phonological words. Nevertheless, there are
words which deviate from the ideal prosodic structure, namely words in which stress is
not found on the first syllable, and words that show secondary stress in addition to
primary stress. The above-named suffixes in b) (-schaft, -heit, etc.) can carry secondary
stress, and a few " loan" suffixes such as -itiit ' -ity ' attract stress (cf. c) above). Thus,
the s-link is found regularly in these cases for prosodic reasons. With respect to the cases
in d), simple vowels in bisyllabic words (Auto, Kina, Uhu) typically do not attract stress,
hence they constitute a trochee, and no s-link is usually found. If the first part of a
compound has more than two syllables and thereby deviates from the ideal phonological
word, however, the s-links becomes acceptable. Typically, "bad" phonological words are
found in morphologically complex words (Plastikautosverkiiufer).
PA573
573

In contrast to syllabic linking elements (cf. section 2.2.2), linking-s does not improve
the word prosodically. Its function is rather that of a boundary marker because it clearly
marks the end of the phonological word. The less sonorous the preceding offset, the
more prominent the end becomes through -s.
2.2.2. Syllabic linking elements in German

The syllabic linking elements in Modern High German are -es, -e, -er, -enl-n, (-ens); -n
constitutes an allophone of -en (cf. (4)), which is why it is included. Out of these linking
elements, -en/-n is considered productive.
(4)
a.
Doktorand
doctorand-SG, NoM
Doktorand-en
doctorand-oBL
b.
Biologe
biologist-SG, NoM
Biologe-n
biologist-OBL
Typically after attaching a syllabic linking element, the first part of a compound ends in
a trochaic foot, cf. Gott-es-dienst, Kind-er-wagen, Schmerz-ens-schrei, with -n/-en being
the only element which can be productively attached to polysyllabic stems (Doktoranden-stipendium = scholarship for a doctoral student, Biologe-n-test = test for/from a biologist). The linking element -en is typical for weak masculine nouns that are usually
formed with a stressed "loan suffix" (-and, -ist, -ent, Kopcke 1995). The linking element
-n is employed if the stem ends in schwa, which applies to weak masculine nouns (Eaten-gang 'e1Tand') as well as feminine nouns (Blume-n.-stiel 'flower stalk'). Feminine
nouns ending in schwa occur without linking elements if the schwa represents a suffix
like in Frische-behiilter (5) and Schwiiche-an.fall. Obviously, suffixation "protects" the
schwa from further changes. The distribution of the linking element thus occurs complementarily with -s, i.e., it is not placed if the first element of the compound is complex.
As a consequence the linking element -n marks morphological simplicity. According to
Fuhrhop (2000), schwa occurring at a link has to be analyzed (besides its interpretation
as derivational suffix), as a verbal link, cf. Bad-e-hose 'bathing trunks' , or rare ly as
plural, cf. Stiidt-e-tag 'association of cities').
(5)
frisch
fresh-ADJ
Frisch-e
fresh-SUFFIX
fresh-NOUN
[Frisch-e]-behalter
[fresh-NOUN]-container
food container
2.2.3. "Common" functions of linking elements in German

Linking elements modify the phonological structure of the first part of the compound:
Non-trochaic structures become trochaic if the first part allows for paradigmatic linking
e lements. In " bad" phonological words, the boundary of the phonological word can be
marked more prominently by means of the s-link. Frequently, a bad prosodic word struc-
PA575
plural, in conformity with the genitive forms of the respective declension class, cf. Farnese jjall-a-lamb 'mountain lamb' (gen. pl. jjalla), Icelandic tolv-u-utskrift 'computer
print' (gen. sg. tolvu). While at first sight, these languages seem to preserve nominal
phrases at a very early stage of lexicalization, there are also signs that inflectional endings in nominal compounds are reanalyzed as linking elements, cf. Thrainsson et al.
(2004: 204-208):
(i)
In spoken Farnese, the genitive case is only infrequently used and thus has only a
weak status, i.e. genitive markers can be regarded as archaic.
(ii) A genitive marker can appear in singular or plural form. Often, the number semantics are not logically related to the semantic relation between fi rst and second part
of a compound, cf. Icelandic plot-u-buo (gen. sg. p!Otu), Farnese plat-u-handil (gen.
sg. platu) 'record store', where the plural form would be expected.
(iii) Especially -s is found with nouns which do not take -s as an inflectional marker.
For example, -s does not occur in any feminine declension class, but is found in
compounds with feminine first parts such as Icelandic keppni-s-holl 'contest hall ',
Farnese gular6t-s-pakki 'carrot pack'.
(iv) Nouns usually appear in the genitive form when the first part of a compound is a
compound itself in Faroese, such that the distribution of genitive forms is not (only)
morphosyntactically conditioned as with case marking, but governed by conditions
of word-formation.
In West Frisian (and in other Frisian as well as some Dutch and Low German dialects),
a class of constructions called genitive compounds exists. These constructions are characterized by features which suggest that they are still close to their origin in a word
group. For example, stress is not on the first part, as is usually the case in compounds,
but on the final part of the construction. The first part additionally always appears in the
genitive form, cf. keamer-s-' doar 'door of the living room'. Hoekstra (2002) does not
analyze these constructions as word-formation units, but as lexical phrases, which are
stored in the lexicon as phrasal patterns, i.e. items in between phrasal and lexical units.
The genitive markers retain a strong relation to the old case system in these units, but
are functionally limited.
3.2. Distributional and functional aspects

Generally, analogy plays an important role in the distribution of linking elements. Especially with respect to the first parts, new word-formation units are often formed in analogy to existing ones (cf. Plank 1976: 217-218; Becker 1992: 10-16), i.e. using the same
linking element, including the zero link. Based on the notion of compositional (or derivational) stem forms in the sense of Fuhrhop (1998) (see section 2), new units emerge in
analogy to existing stem forms rather than according to more general principles. Nevertheless, it is possible to formulate some tendencies (rather than rules) which show that,
in addition to linking in analogy with existing stems, there are certain independent trends
specific to the linking elements. This is especially clear with regard to linking-s (cf.
section 3.2.1), but there are also principles for other elements (cf. section 3.2.2).
575
PA577
3.2.2. Other linking elements

The most frequent linking element -s is unsyllabic. By contrast, most other linking elements are syllabic, cf. Danish -e [g], Swedish -o, Dutch -en [g], German -en, etc. Just
as described for German above, they are often still identical with inflectional suffixes.
If this inflectional suffix is actually part of the declensional paradigm of a noun, it is
interesting to ask whether the linking element is also identical to the inflectional suffix
with regard to its function, i.e. whether it also carries inflectional meaning inside the
compound. It is often assumed that plural meaning is carried by linking elements which
are identical with a plural marker. Experimental studies on Dutch and Afrikaans have
shown that linking elements are associated with plural meaning in the first part of a
compound. Schreuder et al. (1998) conducted a number of experiments using written
compounds. The first part of the compounds could be pluralized with -en. In the written
form, the linking element varied between <e> and <en>. The results showed that plural
semantics were activated when the link was written as <en>, i.e. in the form which is
homographic with the plural marker. Jansen, Schreuder and Neijt (2007) conducted an
experiment with test subjects who were native speakers of Afrikaans. In Afrikaans, the
orthographic form of the corresponding plural marker is <e>. The experiment showed
that when confronted with written Dutch compounds containing either the linking element <e> or <en>, a plural interpretation was activated when the linking element was
homographic with the Afrikaans plural suffix <e>, and linking-<en > was associated
with singular meaning. These results suggest that paradigmatic syllabic linking elements
are associated with plural semantics both in Dutch and Afrikaans. Nevertheless, in contrast with inflection, plural marking is not obligatory in the link, cf. Dutch spier+groep
'muscle group' and mens-en-ziel 'soul of a human being'.
While plural semantics are associated with linking-e(n) in Dutch and Afrikaans, this
is less clear in German (cf. Krott 2001: 224). Dressler et al. (2001) conducted a constituent naming task with German compounds showing that plural semantics play no role in
compound analysis. For German compounds in which associating linking-(e)n with plural semantics makes sense (cf. Frau-en-heim 'home for women'), the same response
latencies were found as for opposite (Schlange-n-biss 'snake bite') and unclear cases
(Vill-en-bau 'villa building'). Additional support is provided by Banga et al. (2013) who
presented novel Dutch compounds to native speakers of Dutch, Dutch-Frisian bilinguals,
and German-speaking learners. The subjects were asked to express their preference for
a form with or without linking-en. A context was provided by which either singular or
plural semantics should be evoked. The results showed that plurality played a role in all
groups of subjects, but that the effects were considerably smaller within the German
group.
While the expr ession of number by linking elements generally seems possible, there
are strong theoretical claims against case marking inside compounds (cf. section 2 .2.1
for a discussion of German examples). While number is inherently assigned to a noun,
case is a contextually licensed category, i.e. the purpose of case is to determine the
relation between a given noun and other entities in the syntactic context. Gallmann
(1998) argues that it is impossible to assign a contextually licensed category to a nonhead of a compound, and since linking elements are part of the non-heads (i.e. first
parts) of compounds, in this view no case marking is possible. Still, so-called genitive
compounds in West Frisian might serve as a counter-example (see discussion above and
577
PA578
578

in Hoekstra 200 2), and the status of compounds in Icelandic deserves further di scussion
with respect to case marking in compounds (cf. Indrioason 1999).
In addition to inflectional functions, specifically syllabic linking elements have been
claimed to have prosodic functions (cf. section 2.2.2 on German). As such, syllabic
linking elements can create trochaic feet from monosyllabic stems, cf. Danish mcelk-ekarton 'milk package'. Usage of linking elements in this manner is far from consistent,
but many syllabic linking elements are almost totally constrained to monosyllabic (or
finally stressed) stems, e.g., -e in Danish, -en in Dutch, and -e, -en, -er, and -es in
German. Neijt and Schreuder (2007) examine the rhythmic function of Dutch -en more
specifically, considering an influence from the prosodic structure not only of the first
part, but also of the second part of the compound. They fi nd out that linking-en is found
more often between the parts of compounds in which two stressed syllables clash (cf.
boom 'tree' + rij 'row' ~ bom-en-rij 'tree row') than in contexts in which this is not
the case.
Since Dutch -en is thus both semantically (plural) and prosodically (rhythm) relevant,
Hanssen et al. (2013) conducted a study into the relation between both functions in
Dutch. For spoken pseudo-compounds and existing pairs of compounds which are variably realized with or without linking-en (cf. bloem-bak and bloem-en-bak 'flower box'),
they tested experimentally whether the perceived plurality is affected by -en preventing
a stress clash or not. In fact, the results show that when a stress clash is prevented, the
perceived plurality is lower than when no stress clash is prevented.
For Danish Iinking-e (realized as [~]), a semantic distribution based on animacy has
been claimed (cf. Kurschner 2005). Native monosyllabic nouns and derivates suffixed
by -ing tend to appear with a linking-e in compounds, cf. dreng-e-streg 'boyish prank',
fisk-e-suppe 'fish soup', svamp-e-skade 'damage on mushrooms', udlcending-e-lov ' law
on foreigners'. The same tendency is also found for the formally identical plural marker
-e (cf. Hansen 1967: 112), but the use of the linking element is even more consistent
compared to the inflectional marker, cf., e .g., ven 'friend', pluralized as venner, in venne-tjeneste 'friendly turn'. For Dutch, the results of studies on the effects of countability,
animacy, and concreteness have shown that the semantic features of the first part of a
compound are relevant to the choice of linking element. Van den Toom (l 982b, cf. also
l 982a) suggests that the probability of the appearance of linking-en is higher w ith first
parts denoting countable and in particular animate entities, especially when the first part
is related to the second part as owner or initiator, cf. mugg-en-been 'mosquito leg' or
spinn-en-web 'spider's web'. Corpus analyses and experimental studies by Krott (2001:
I 03- 120) reveal that the semantics of the right part of compounds has no effect, but that
the semantics of the first part is influential: -s is more li kely to appear with abstract
nouns, while -en is more often found with concrete nouns. With respect to concrete
nouns, -en appears with highest probability when the semantics of the left part is characterized by an animate feature. Hoekstra (1 998: 39) discusses how the distribution of
linking-s vs. no linking element in subgroups of West Frisian derived nouns varies wi th
the semantics of the first part of the compound.
4. Conclusions
Various linking elements occur in the Germanic languages, and they are distributed and
functionalized in various manners. The di stribution and functionalization has been stud-
PA579
ied in corpora, both with regard to the di stribution in the current lexicon and in the
historical development of the lexicon. Additionally, experimental studies provide evidence for factors relevant in the production of linking elements, and for morphological,
prosodic, and semantic functions associated with linking elements both in production
and perception. The link between the corpus-based and the experimentally based lines of
research will need to be tightened in future research. In addition, especially the historical
development of linking elements is still far from well-studied in most of the Germanic
languages. What is more, we can expect new perspectives on linking elements from
studies concerning items which appear in the link but do not stem from inflectional
suffixes, cf. Renner 's (2013) study on -cum in English coordinate compounds such as
poet-cum-philosopher.
5. References
2002
Restricting suffix combinations in German and English: Closing suffixes and the monosuffix constraint. Natural Language and Linguistic Theory 20(3): 451 - 490.
Augst, Gerhard
1975
Ober das Fugenmorphem bei Zusammensetzungen. In: Gerhard Augst (ed.), Untersuchungen zum Morph eminventar der deutschen Gegenwartssprache, 71 - 155. Tiibingen:
Narr.
Baayen, Harald, Wolfgang Ulrich Dressler, Andrea Krott and Robert Schreuder
2007
Analogical effects on linking elements in German. Language and Cognitive Processes
22: 25-57.
Banga, Arina, Esther Hanssen, Anneke Neijt and Robert Schreuder
2013
Preference for linking element en in Dutch noun-noun compounds: Native speakers and
second language learners of Dutch. In: Morphology 23( 1): 33- 56.
Becker, Thomas
1992
Compounding in German. Rivista di Linguistica 4( 1): 5-36.
Botha, Rudolf P.
1968
The Function of the Lexicon in Transformational Generative Grammat: The Hague:
Mouton.
Delsing, Lars-Olof
2002
Svenskt foge-s. Folkmalsstudier 41: 67- 78.
Dressler, Wolfgang Ulrich, Gary Libben, Jacqueline Stark, Christiane Pons and Gonia Jarema
200 1 The processing of interfixed German compounds. In: Geert Booij and Jap van Marie
(eds.), Yearbook of Morphology 1999, 185- 220. Dordrecht: Kluwer.
Eisenberg, Peter
2006
Grundriss der deutschen Grammatik. Das Wort. 3rd ed. Stuttgart/Weimar: Metzler.
Eisenberg, Peter and Ulrike Sayatz
2004
Left of number: Animacy and plurality in German nouns. In: Gereon Muller, Lutz Gunkel and Gisela Zifonun (eds.), Explorations in Nominal Inflections, 97- 120. Berlin/New
Faarlund, Jan Terje, Svein Lie and Kjell Ivar Vannebo
2006
Norsk referansegrammatikk. Oslo: Universitetsforlaget.
Fuhrhop, Nanna
1998
Grenz/ii/le morphologischer Einheiten. Ti.ib ingen: Stauffenburg.
579
PA580
580

Fuhrhop, Nanna
2000
Zeigen Fugenelemente die Morphologisierung von Komposita? In: Rolf Thieroff, Matthias Tamrat and Nanna Fuhrhop (eds.), Deutsche Grammatik in Theorie un.d Praxis,
201 - 213. Tiib ingen: Niemeyer.
Gallrnann, Peter
1998
Fugenmorpheme als Nicht-Kasus-Suffixe. In: Matthias Butt and Nanna Fuhrhop (eds.),
Variation. un.d Stabilitiit in der Wortstruktur. Untersuchungen. zu Entwicklun.g, Erwerb
un.d Varietiiten. des Deutschen. un.d an.derer Sprachen., 177-190. Hildesheim: Olrns.
Hansen, Aage
1967
Moderne dansk fl: Sprogbeskrivelse. K0benhavn: Grafisk forlag.
Hanssen, Esther, Arina Banga, Robert Schreuder and Anneke Neijt
2013
Semantic and prosodic effects of Dutch linking elements. Morphology 23(1): 7-32.
Hoekstra, Jarich F.
1998
Fryske Wurdfoarming. Ljouwert: Fryske Akademy.
Hoekstra, Jarich F.
200 1 Comparative aspects of Frisian morphology and syntax . In: Horst Haider Munske (ed.),
Han.dbuch des Friesischen.. Handbook of Frisian. Studies, 775-786. Ti.ibingen: Niemeyer.
Hoekstra, Jarich F.
2002
Genitive compounds in Frisian as lexical phrases. Journal of Comparative Germanic
Linguistics 6: 227-259.
Humbert, Adrie nne
2006
Die Fugenelemente in den Substantivkomposita des Luxemburgischen . Unpublished MA
thesis, University of Trier.
Indrioason, Thorsteinn G.
1999
Um eigna rfallssamsetningar og adrar samsetningar i islensku. fslen.skt ma/ 21: I07-150.
Jansen, Carel, Robert Schreuder and Anneke Neijt
2007
The influence of spelling conventions on perceived plurality in compounds. Written.
Language and Literacy I0(2): I05-1 I4.
Josefsson, Gunlog
1998
Minimal Words in a Minimal Syntax. Word formation in Swedish. Amsterdam/Philadelphia: Benjamins.
Kopcke, Klaus-Michael
1995
Die Klassifikation der schwachen Maskulina in der deutschen Gegenwartssprache. Zeitschrift far Sprachwissenschaft 14(2): 159- 180.
Krott, Andrea
200 1 Analogy in Morphology. The selection of Linking elements in Dutch compounds. Wageningen: Ponsen and Looijen.
Kurschner, Sebastian
2003
Von Volk-s-musik und Sport-0-geist im Lemming-0-land - af folk-e-musik og sport-sand i Jemming-e-landet: Fugen.elemente im Deutschen un.d Diin.ischen - eine kontrastive
Studie zu einem Grenz/all der Motphologie. Freiburg: FreiDok. http://www.freidok.unifreiburg.de/volltexte/ 1256/
2005
Verfugung-s-nutzung kontrastiv: Zur Funktion der Fugenele mente im Deutschen und
Diinischen. Tijdschrift voor Scandinavistiek 26(2): l 0 1- 125.
2007
Grenzganger zwischen F lexion und Wortbildung: Zur Geschichte des danischen Fugens. ln: Wolfgang Behschnitt and Elisabeth Herrmann (eds.), Uber Grenzen. Grenzgiinge
der Skandinavistik, 349- 367. Wiirzburg: Ergon.
PA589
33. Synthetic compounds in German
words where the respective element is a noun. The prototypical example for this type of
synthetic compounds is blauiiugig, as introduced in the first section of this article. In
German, the suffix -ig is particularly productive in synthetic compounds of this type.
The transparent core of this derivational pattern can be characterized in the following
way: The suffix takes nouns as bases and forms adjectives that denote a possessive
relation. For example, the adjectives biirtig ' beardy' is based on the noun Bart 'beard ',
and it can be paraphrased as ' having a beard'. The following list shows some relevant
data, only some of which will be classified as synthetic compounds in the following
discussion:
(5)
a.
kurz+bein +ig
short+leg+ed
'short-legged'
b.
rot+biirt+ig
red+bea rd+ed
' red-bearded'
c.
bos+will+ig
bad+will+ed
'malicious'
d.
rad+form+ig
wheel+shape+ed
'wheel-shaped'
e. fremd+art+ig
fo reign+kind+ed
'foreign'
While the adjective biirtig is an existi ng word in German, the similarly structured word
iiugig only exists in technical terminology (cf. section 1) and beinig probably does not
exist as a free word at all, but only as part of a synthetic compound like in (Sa). Leser
( 1990: 86- 93) attempts to explain this peculiar behavior with reference to the concept
of inalienable possession as discussed in Ljung (1976) in regard to English adjectives
like blue-eyed and bearded. A possessive relation holds between the basic noun of the
adjective and a noun that the adjective modifies in phrasal constructions, for example in
the phrase a bearded man. Ljung (1976: 162- 163) distinguishes two degrees of inalienable possession: The first degree holds when all referents of a noun have as. an inherent
part or property the referent of the other noun (as is the case for objects and surfaces)
or when all referents of a noun are at least expected to have as an inherent part or
property the referent of the other noun (like in the relation between human beings and
body parts). The second degree holds when a subset of referents of a noun has as an
inherent part or property the referent of the other noun, as is the case for males and
beards. Besides inalienable possession, there is alienable possession. Adjectives in -ed
(as well as German adj ectives in -ig) are impossible for first degree of inalienable possession. Ljung (1976: 162) relates this property to a pragmatic conversation principle that
bans tautological expressions. Since all objects that can have legs regularly do have legs,
it is tautological to modify a noun with the adj ective legged.
589
PA592
592

Booij, Geert
2009
Compounding and construction morphology. In: Rochelle L ieber and Pavol Stekauer
(eds.), The Oxford Handbook of Compounding, 201 - 216. Oxford: Oxford University
Press.
Botha, Rudolf P.
1981
A base rule theory of Afrikaans synthetic compounding. In: Michael Moortgat, Harry
van der Hulst a nd Teun Hoekstra (eds.), The Scope of Lexical Rules, 1- 77. Dordrecht:
Foris.
Botha, Rudolf P.
1984
Morphological Mechanisms. Lexicalist Analyses of Synthetic Compounding. Oxford:
Pergamon Press.
Bzdyga, Andrzej Z.
1999
Zusammenrilckung, -setzung, -bildung . In: Andrzej Ki:itny and Christoph Schatte (eds.),
Das Deutsche van innen und van auj3en. Ulrich Engel zum 70. Geburtstag, 9-23. Poznan : Wydawnictwo Naukowe UAM.
Donalies, E lke
2001
Zur Entri.impelung vorgeschlagen die Wortbildungsarten: Ri.ickbildung, Zusammenbildung, Zusammenri.ickung, Klammerform und Pseudomotivierung. Studia Germanica Universitatis Vesprimiensis 5(2): 129- 145.
E isenberg, Peter
2013
Grundriss der deutschen Grammatik. Vol. 1: Das Wort. 4th ed. Stuttgart/Weimar: Metzler.
E lsen, Hilke
Grundzuge der Morphologie des Deutschen. Berlin/Boston: de Gruyter.
2011
2012
Gaeta, Livio
20 IO
Synthetic compounds: With spec ial reference to German. In: Sergio Scalise and Irene
Vogel (eds.), Cross-Disciplinary Issues in Compounding, 2 19- 235. Amsterdam/Philadelphia: Benjamins.
Giegerich, He inz
2009
Compounding and lexicalism . In: Rochelle L ieber and Pavol Stekaue r (eds.), The Oxford
Handbook of Compounding, 178- 200. Oxford: Oxford University Press.
Hoekstra, Teun
1986
Deverbalization and inheritance. Linguistics 24: 549- 584.
Jackendoff, R ay
2009
Compounding in the parallel a rchitecture and conceptual semantics . In: Rochelle Lieber
and Pavol Stekauer (eds.), The Oxford Handbook of Compounding, 105- 128 . Oxford:
Leser, Martin
1990
Das Problem der ,Zusammenbildungen '. Eine lexikalistische Studie. Trier: Wissenschaftlicher Verlag Trier.
L ieber, Roche lle
2009
IE, Germanic: English. In: Rochelle Lieber and Pavol Stekauer (eds.) , The Oxford Handbook of Compounding, 357- 369. Oxford: Oxford University Press.
Ljung, Magnus
1976
-ed-adjectives revis ited. Journal of Linguistics 12: 159-168.
Maling, Joan
200 1
Dative: The heterogeneity of the mapping among morphological case, grammatical fu nctio ns, and thematic roles. Lingua 111: 41 9- 464.
PA595
34. Verbal pseudo-compounds in Gennan
Different modes of word-formation are reflected in common terminology: Composition subsumes concatenation of otherwise free forms. Derivation encompasses affixation
of a bound morpheme. Besides these concatenative operations there are certain, observationally at least, non-concatenative ones. Conversion amounts to a change in category
without any change in form. Allomorphy may also indicate a change in lexical category.
A further specific type is represented by backformation, an operation that apparently
deletes a derivational affix resulting in the (re)constitution of a potential base. The latter
operation is often invoked in the case of apparent compounds which, nevertheless, cannot
be formed by the regular rules of compounding in a given language, hence the term
pseudo-compound as an index of this class of words.
According to strict lexicalism, syntax is opaque to the specific word-formation operation an element of syntactic structure emerges from. That means word-fonnation does
not interfere with syntactic operations at all. This view, however, is challenged by a
fundamental aspect of clausal syntax in languages like German. Finite verbs may adopt
different positions within a clause, depending on clause type. These facts are modelled
by a syntactic operation commonly called verb movement. Verb movement, however,
does not apply to finite verbs in a uniform way, actually, it is affected by the morphological structure of the verb. These facts are often used to motivate a slight modification of
the strict lexicalist hypothesis. Peculiar conditions on verb movement (e.g., with particle
verbs) are associated with particular processes of word-formation. This view of a oneto-one correspondence between word-formation operations and syntactic conditions on
verb movement has also been extended to the case of verbs that cannot move at all and,
hence, only occur in subordinate clauses. The idiosyncrasy of these verbs is frequently
attributed to backformation as the morphological source of these pseudo-compounds.
Whatever appeal such an approach may have, it is not supported by the empirical evidence. There are, for instance, verbs, which, according to morphological premises, have
to be considered as pseudo-compounds but nevertheless undergo verb movement in the
canonical way. A thorough reconsideration reveals that no trivial correspondence relation
obtains between descriptively determined word-formation operations and syntactic ones
like verb movement. This asymmetry gives rise to a revision of certain word-formation
rules in a more abstract mode allowing for operations on elements of word structure
which a strict lexicalist view would restrict to syntax.
2. Verb movement and word structure in German

In German observationally three types of clauses have to be distinguished, which differ
with respect to their respective constituent structure. These three types are crucially differentiated by the position of the finite verb into verb-second, verb-fi rst and verb-final clauses.
(1)
a. Theo hat einen Aal geangelt.

Theo has an
eel angled
'Theo has caught an eel. '
b. Hat Theo einen Aal geangelt?

Has Theo an
eel angled
'Has Theo caught an eel?'
595
PA597
3. Finite verbs t hat do not move

The partition into prefix verbs and particle verbs, then, strongly suggests that any finite
morphologically complex verb can be subsumed under one of these two classes (possible
ambiguous forms set aside). Nevertheless, there is a small set of complex verbs like
urauffehren 'to premiere sth. ', bergsteigen ' to mountain climb' which, in their finite
form, are restricted to verb-final clauses.
(3)
das Ensemble die Komodie urauffehrte.

a. als
when the company the comedy first performed
'when the company premiered the comedy.'
b. *das Ensemble uraujfiihrte
die Komodie.
the company first performed the comedy
c. *das Ensemble fiihrte

die Komodie urauf
the company performed the comedy first
The phenomenon, which, by the way, is not peculiar to German (cf. Koopman 1995 on
Dutch), has inspired a debate on a technical aspect of modelling clause structure. The
fact that there are finite, i.e. inflected verbs in German which get stuck in their base
position has been called upon as evidence against the postulate that German, like for
instance English, employs a functional category inflectional phrase (IP). If, the argument
runs, there is a functional head filled by the morpheme which represents inflectional
features, the (uninflected) verb stem has to move from its base position within the VP
to this functional head-position 1. But if moved once, there would be no plausible
reason to fix the finite verb in this position and to prevent it from moving further to the
head of the CP (i.e. to the verb-first/ verb-second position), which represents the complete structure of the clause (cf. Haider 1993, 2010; Sternefeld 2006; Vikner 2005).
In addition to this peculiar aspect of clausal syntax, the matter merits attention because of its impact on the morphological analysis of word-structure. It is quite obvious
that the morphological complexity of the verb is responsible for its syntactic behaviour.
Simplex verbs are never restricted in any comparable way. Looking for morphological
factors that might restrict verb movement, two are of principle concern. The first one is
categorial in nature: does the categorial property of the prefixed element - being a Vprefix or a N-prefix, etc. - and /or its structural relation to the stem implement an obstacle
to applying syntactic operations on the verb like movement? The second one is a matter
of quantity: does the number of prefixed elements constrain the relevant syntactic options? Besides the verification of these factors, the specific morphological operation must
be determined from which these remarkable verbs emerge. Analyses which address this
phenomenon often resort to backformation as the pertinent word-formation process.
Nevertheless, in modelling the (morpho-)syntactic conditions which fix the immobile
verb in its base position, the morphological process itself from which they are derived
is in many cases of minor interest. Therefore, the term pseudo-compound often seems
to be employed as a convenient term to characterize just these verbs.
Apart from the question of how backformation should be modelled in terms of structural description and structural operation, it must be decided whether it provides a necessary as well as sufficient condition for determining the class of possible immobile verbs.
597
PA598
598

A number of proposals have been made to account for the immobility of these fin ite
complex verb forms. They differ with respect to their technical details as well as the
empirical predictions they make.
3.1. Confl icting licensing requirements on particles and prefixes

According to one line of argumentation, certain complex verbs, which are licit morphological objects, fail to move because no licit antecedent gap relation can be established
which can also license the particle left behind.
Haider (1993, 2010) argues that a mismatch in the requirements on the prefixed
elements prevents a verbal pseudo-compound from moving up to the verb-first/verbsecond position. His most prominent example is the verb urauffiihren, the result of backformation from the noun Urauffiihrung 'premiere'. This verb consists of the verbal stem
fiihren lit. 'to guide', from which the verb aujfiihren 'to perform' is derived. Ur-, on the
other hand, is a nominal prefix. Thus, the noun Urauffiihrung is formed by prefixing
the nominalization Aujfiihrung. Finally, the verb uraujfiihren derives from the noun by
truncation of the nominal izing affix -ung.
Auffiihren, a particle verb, leaves the particle au/behind when moved.
(4)
a. man fiihrte die Komodie mehrmals auf

one put
the comedy repeatedly on(stage)
'one performed the comedy repeatedly.'
b. *man auffiihrte die Komodie mehrmals.
one onput
the comedy repeatedly
A prefix like ur- instead can never be separated from its host. Moving the verb would
only be possible if the prefix were also to move. Since the ur- prefix precedes the au/particle, two contradictory demands arise, which cannot simultaneously be satisfied.
(5)
a. *man urauffehrte die Komodie im

vergangenen Jahr.
one first onput the comedy in the last
year
'one premiered the comedy last year. '
vergangenen Jahr urauf

b. *man @hrte die Komodie im
one onput the comedy in the last
year first
Koopman O995) gives an account of the Dutch variant of immobile verbs like herinvoeren 'to reintroduce'. According to her explanation, only one particle can be licensed by
the trace left by the verb under movement to the clause initial position. This leaves the
other particle, namely her- in the case of herinvoeren, unlicensed when the verbal root
voeren is moved. This explanation could easily be adopted for the case of German. But,
as Vikner (2005) notes, any account that crucially resorts to multiple prefixation of the
verbal root is insufficient since there are numerous immobile verbs with only one prefixed element like schutzimpfen 'to vaccinate' which also resist verb movement.
PA600
600

prefix or a particle, then, might be incompatible with the requirements on the interpretation of the complex verb due to its derivation from its nominal base. Syntactic operations
like verb movement, which require unequivocal [max] marking of the non-head component of the verb, hence, are inapplicable to this type of verb. If the complex verb remains
in its base position, no need arises to unambiguously determine the preverbal element(s)
with regard to that feature.
Vikner (2005) models the distinction between prefix verbs and particle verbs with
two distinct levels of projection. While both form legitimate lexical objects, their internal
structure is different. Combining a prefixed element w ith a verbal head v 0 may result
in morphologically complex, albeit syntactically simplex, V 0 . On the other hand, the
particle verb is syntactically complex. The verbal projection, which incorporates the
particle, has a (phrasal) projection level V* between the head level x 0 and the X ' level
(enclosing the position of phrasal complements) (cf. Wurmbrand 1998).
(8)
[yo zer [y ofall-]]
a.
'to decompose '

[ v auf [yo wach-]]
'to awake'
b.
Verb movement only affects x 0 categories but not the X * category. While (8) represents
two options of structure building by concatenation, the immobile verbs emerge from
nominal compounds by truncating the nominal-suffix of a compound deverbal noun:
(9)
a.
[N schutzN [N [ v impfy ] ungN ]]
b.
---+
[N schutzN [ N [ y impfy ]]]

'to vaccinate; lit. protection -vaccinate'
(9b), of course, is not a legitimate morphological object. Lacking a nominal head, the
whole structure cannot be a noun. Since the verb stem impf remains on the right-hand
side of the word, this element becomes the head of the compound projecting its categorial
property to the resulting word.
In principle, a nominal like schutz may occur as a constituent of either a v 0 or V *
category. But in the case of backformation, according to Vikner, an indeterminacy arises
with respect to the projection level of the resulting verb. The resulting verb has to fulfill
the requirements on both a v 0 and a V* category.
(10) [y / v* schutzN [v impfv ]]

0
Movement of the whole word is blocked because V* as a non-head cannot assume a

head position, namely c 0 . Movement of the verbal stem would create a trace which is
not a possible constituent within a x 0 category. Hence the complex verb is fixed in its
base position.
(11)
a.
*[c p ... [c [volv* schutzN [v impfv ]]; ... f; ]]
b.
*[c p ... [c [yo impfv ]; ... [yo/y schutzN [yo f; ]]]
PA601
3.3. Phonological constra ints

The accounts pointed out thus far rely on structural conditions and constraints within the
domains of syntax and morphology. Zeller (2001) draws additionally on a phonological
condition which indirectly restricts the options of morphologically complex verbs to
undergo verb movement. In the first place, he argues that the non-head component of
the verb is a phrasal object, like particles in general. As an immediate consequence, the
complex verb as a whole is also a phrasal object beyond the head projection level and,
hence, precluded from verb movement.
(12) ... UrrtP urauf]fiihrv-]
This leaves open the possibility of only moving the verbal head, just as in the case of
canonical particle verbs, which nonetheless is not given. Here, the phonological consideration comes into play, which, on its part, is based on the premise that immobile verbs
are derived from complex nouns by backformation. Within a nominal compound, strict
conditions on linearization obtain. The non-head constituent must precede the nominal
head; linear order cannot be reversed, for instance by syntactic operations.
(13) a.
b.
Urauffuhrung
*Fiihrungurauf
The restriction postulated by Zeller says that the derived verb has to be formally similar
to its derivational source with respect to its phonological shape. Namely, the linear order
of head and non-head has to be preserved. Obviously, this need is fulfilled only if the
verb remains in its clause final base position.
(14) a.
b.
[er ... [C' C .. . [rrtP urauf] fahrteni ]]

*[er ... [C' [fiihrtenL ... [rrtP urauf] ti]]
According to the account just sketched, immobile verbs are derived by backformation.
Zeller even strengthens his c laim to the effect that backformations are licensed only if
their non-head component is reanalysed as a particle, i.e. as a phrasal constituent of the
clause. This means that any verb derived by backformation is predicted to be immobile
at any rate.
3.4. Some essentials

Though differing in details, these analyses accord with respect to a number of aspects.
Verbs that are restricted to the clausal base position in syntax are different from
simplex verbs and prefix verbs. As for their word structure, they are on a par with
particle verbs. This correspondence refers, firstly, to the hierarchical relations between
the components of the verb. The crucial difference between particle verbs which are
accessible to verb movement and those verbs which do not allow separation of the verbal
head from the non-head element is traced back to the specific morphological operation
601
PA602
602

by which the latter are derived. The specific differences in word structure between complex nouns and complex verbs live on in the derivational result of backformation of a
verb from a nominal compound.
Irrespective of the significance of the analyses mentioned in the preceding sections
certain issues still await an explanation.
In the face of the fact that, in German, verbal compounding is not productive with
nouns, on the one hand, and the considerable number of immobile verbs with nominal
non-head elements, on the other, it is reasonable to assume a somewhat irregular derivational source like backformation.
However, a ce11ain amount of dissent should be expected with respect to the precise
determination of the derivational origin of N +V compounds. Eisenberg (1998), for instance, cons iders the verb ehebrechen ('to commit adultery') which resists movement as
a case of Univerbierung - i.e. consolidation.
Though backformation is mostly conceived of as deletion of a (putative) nominalizing
suffix, it has been argued by Eschenlohr (1999) that this term covers an ensemble of
diverging factors which may more properly be characterized by other derivational operations as for instance conversion. Furthermore, it is often impossible, to determine a
unique derivational base of a given pseudo-compound.
Apart from this reasoning, it should be noticed that there are a number of verbs like
rollem which are derived from a deverbal noun Roller not by suffix deletion. These
verbs demonstrate that verbalization of a deverbal noun is not necessarily the result of
reversing a derivational operation by deleting its morphological index, i.e. the suffix.
But even if backformation were necessarily a condition on the formation of immobile
verbs, it is not an unquestionably sufficient one. Take the nominal element Wette 'bet',
which may be part of several verbs like wettrennen 'to race', wettriisten 'to arms race',
wettlaufen ' to race' , and wetteifern 'to compete'. Since the verbs rennen 'to run' , riisten
'to arm', laufen ' to run' and eifern 'to strive' do not subcategorize for a nominal complement, incorporation is not a possible source for these complex verbs. Backformation has
to be regarded as the operation wh ich derives the verbs wettrennen from the noun Wettrennen and wettriisten from Wettriisten/ Wettrustung. But, while wettrennen is incapable
of movement (15a), wettriisten does move in the same manner as a prefix verb (15c).
However, backformation cannot be presumed a necessary condition either. The verbs
wetteifern and wettlaufen result from conversion of the respective nouns Wetteifer and
Wettlauf The former may occur in the verb-second position like a prefix verb (15d)
whereas the latter does not m ove at all (15a).
(15)
a. *Eberhard wettrennt/wettlauft mit Karl-Eugen.

Eberhard races
with Karl-Eugen
'Eberhard runs a race with Karl-Eugen.'
b. Eberhard verliert stets, wenn er wettrennt/wettlauft.
Eberhard looses always when he races
'Eberhard always looses when he races.'
c. Sie wettrusten bis zur Selbstvernichtung.
They arms race until
self-destruction
d. Eberhard wetteifert mit Karl-Eugen um die Gunst von Paula.
Eberhard competes with Karl-Eugen for the grace of Paula
PA603
Hence, we have to realize that, although nominal compounds with a deverbal head are
the derivational source of a subclass of complex verbs, terms like backformation or
pseudo-compounding alone do not sufficiently elucidate the structural conditions on their
formation and syntactic characteristics. In the face of (15) it is likewise inopportune to
specify verbal backformation as an operation with a structurally unique output.
In addition to these more general considerations, there are some specific objections
to a reasoning along the line of, for instance, Stiebels and Wunderlich (1994) or Vikner
(2005). Despite technical differences, they accord to the extent that the preverbal element
of an immobile verb evokes an indeterminacy in its status as a prefix or particle. However, it is not that clear why prefixes like ur- and erst-, which may be attached to nouns,
should in addition assume the properties of a particle if the containing noun is converted
into a verb. It seems more plausible to assume that these elements are prefixes in both
cases (cf. Haider 1993 ). But then, the question arises why verbs like ursenden, urwiihlen,
erstwahlen are restricted to verb-final positions within clauses as well.
(16)
gab es einen Skandal.

a. Als man das Horspiel ursendete,
when o,ne the radio play first broadcasted gave it a
scandal
'The first broadcast of the radio play created a scandal.'
b. *Man ursendete
das Horspiel schliejJlich doch nicht.
one first broadcasted the radio play finally
yet not
'Finally, the radio play was not yet broadcast.'
As for the constraint on the productivity of N+V compounding, the assumption of a

somewhat irregular ( prima facie non-concatenative) derivational operation like backformation/pseudo-compounding is justified in the case of immobile verbs. However, the
structural conditions governing the syntactic behaviour of these verbs are not confined
to the derivational process.
4. Meaning and structure of verbal pseudo-compounds

The interpretation of complex verbs gives a clue as to the syntactic restrictions imposed
on the subclass under consideration. Within the domain of putatively backformed verbs,
a bipartition can be found. The discriminating property is compositionality of interpretation (cf. Fortmann 2007).
One set comprises verbs like bauchlanden ('to belly-land'), ehebrechen ('to commit
adultery'), schutzimpfen (' to vaccinate'). The interpretation of these verbs is computed
compositionally from the meaning of both the non-head and the head. The non-head
either functions as an adverbial modifier, or, in a few cases, it saturates an argument of
the verbal base. These verbs regularly admit a paraphrase composed of the verbal head
and a phrasal adjunct or complement.
(17)
a.
als das Flugzeug bauchlandete.

when the airplane belly-landed
b.
als das Flugzeug auf dem Bauch landete.

when the airplane on the belly landed
603
PA606
606

According to the premises of our explanation, the nominal suffix is not deleted at all but
only left unpronounced.
(25) [[ gewahrN [[ leistv] 0N ]N]N 0v ]v
Here again, the nominal element gewahr does not c-command the verbal suffix.
While the restriction of these verbs to a non-compositional meaning can be explained
in terms of morphological structure, their behaviour under verb movement follows immediately from one and the same structure. Suffixes project X 0 -categories. Hence the projection level of the verbs in (21), (23) and (25) is v 0 . Since v 0 elements are subject to
verb movement, these complex verbs move to the clause initial C 0 -position if necessary.
5.2. Compositional meaning and verb non-movement

The structural underpinnings of the compositional interpretation of an immobile verb
like bergsteigen, which is derived from the compound noun Bergsteiger (26a) have to
be like (26b), as a first approximation.
(26) a.
b.
[ bergN [ [steigv ] -erN ]N] N

[ bergN [[[steigv ] -0 N] -0 v]v]v
Although in (26b) the verbal head formed by zero suffixation is c-commanded by the
nominal element berg, th is structure cannot be derived from the complex noun in (26a)
by successive concatenative operations. The head of (26b) is the verbalized noun steiger
alone but not the compound bergsteiger and the noun berg is directly attached to this
denominal verb: [[[steigv] -0N] -0v ]. Hence (26b) emerges as nothing other than the
result of an otherwise non -productive word-formation process N+V. In order to get the
right derivational result from the correct base, verbal suffixation has to target the compound in (26a) as a whole.
But now the verbal head in (27) is no longer c-commanded by the nominal berg because
the latter is part of the complement of the zero verbal suffix. In order to obtain the
required c-command relation to the verbal head, the nominal element berg must be raised
into an appropriate position adjoined to the verb leaving a trace in the complex Nstm cture.
The verbal zero-affix introduces an event whose content is specified as a climbing action
prototypically performed by the referent of the complex noun. In its position adjoi ned
to the derived verb, the nominal berg functions as a directional modifier specifying the
path or goal of the climbing.
The structure postulated in (28) may give rise to sceptical objections because a movement operation within morphological structure is employed in order to accomplish it.
PA608
608
6. Consequences
The presumably most controversial aspect of the reasoning in the preceding sections is
the postulate of morphological objects without phonological content - i.e. zero affixes and of the accessibility to movement of morphological objects. The measure of justification of this approach is its explanatory potential. It should be noticed that there is still
another domain of verbal morphology which justifies the present point of view.
According to lexicalist premises, morphologically complex word s resulting from a
given word-formation operation should undergo further morphological modification in a
uniform way, possibly distinct from the results of different word-formation operations.
So, for instance, the perfect participle and the zu-infintive of particle verbs in German
have in common that the particle precedes the inflectional prefix ge- as well as the
infinitive marker zu (cf. (32)). With prefix verbs, on the other hand, affixation of the
perfect marker ge- is suppressed by the prefix while the infinitive marker zu must precede it (cf. (33)).
(32)
a. Paul ist eing!!_schlaf?n.

Paul ist in -pfx-sleep-sfx
' Paul has fallen asleep.'
b. Paul vermied einzuschlaf?n.
Paul avoided in-to-sleep-sfx

' Paul avoided to fall asleep.'
(33)
a. Paul hat verschlaf?n.

Paul has pfx-sleep-sfx
' Paul has overslept.'
b. Paul vermied zu verschlafen.
Paul avoided to pfx-sleep-sfx

' Paul avoided to oversleep. '
There is, however, an array of complex verbs that are unquestionable prefix verbs judged
from movement and perfect participle formation (cf. (34)), but which behave like particle
verbs if the zu-infinitive is formed (cf. (35)).
(34) a.
D er Beklagte anerkannte seine Obliegenheiten.

' The defendant admitted his obligations. '
b. D er Beklagte hat seine Obliegenheiten anerkannt.

'The defendant has admitted his obligations. '
(35) a.
*Der Beklagte war gezwungen, seine Obliegenheiten zu anerkennen.
b. Der Beklagte war gezwungen, seine Obliegenheiten anzue rkennen.

the defendant was compelled hi s obligations
to admit
'The defendant was compelled to admit his obligations.'
Avoiding the unreasonable assumption of two distinct word-formation operations, the
explanation of the data in (35) has to acknowledge that the prefix an- has hopped over
the infinitive marker zu within the complex word (cf. Fortmann 2007).
PA610
610

Fortmann, Christian
2007
Bewegungsresistente Verben. Zeitschrift fiir Sprachwissenschaft 26: 1- 40.
Haider, Hubeti
1993
Deutsche Syntax Generativ. Tubingen: Narr.
Haider, Hubert
2010
The Syntax of German. Cambridge: Cambridge University Press.
Kastovsky, Dieter
2005
Conversion and/or zero: Word-fo rmation theory, historical linguistics, and typology. In:
Laurie Bauer and Salvador Valera (eds.), Approaches to Conversion I Zero-Derivation ,
31- 50. Berlin: Waxmann .
Koopman, H ilda
1995
On verbs that fail to undergo V-second . Linguistic Inquiry 26: 137- 163.
0 lsen, Susan
1990
Konversion als ein kombinatorischer Wortbildungsprozel3. Linguistische Berichte 127:
185-216.
Pesetsky, David
1985
Morphology and logical form. Linguistic Inquity 16: 193-246.
Rapp, Irene
2006
'Was den Besuch zum Ereignis macht': E ine outputorientierte Analyse fiir die VerbNomen-Konversion im Deutschen. Linguistische Berichte 208: 407- 430.
Stemefeld, Wolfgang
2006
Syntax. Eine merkmalbasierte generative Analyse des Deutschen . Tubingen: Stauffenburg.
Stiebels, Barbara and Dieter Wunderlich
1994
Morphology feeds syntax: The case of particle verbs. Linguistics 32: 913-968.
Vikner, Sten
2005
Immobile complex verbs in Germanic. Journal of Comparative Germanic Linguistics 8:
8J- l 15.
W iese, Richard
2002
A model of conversion in German. In: Ingrid Kaufinann and Barbara Stiebels (eds.),
More than Words. A Festschriftfor Dieter Wunderlich, 47- 57. Berlin: Akademie Verlag.
Wiese, Richard
2008
A two-level approach to morphological structure. Journal of Germanic Linguistics
20(3): 243-274.
Wurmbrand, Susanne
1998
Heads or phrases? Particles in particular. In: Wolfgang Kehrein and Richard Wiese
(eds.), Phonology and Morphology of the Germanic Languages, 267- 295. Tiibingen:
Niemeyer.
Zeller, Jochen
200 1 Particle Verbs and Local Domains. Amsterdam/Philadelphia: Benjamins.
Christian Fortmann, Tiibingen (Germany)
PA611
35. Particle verbs in Germanic

I.
2.
3.
4.
Introduction
Morpho-syntactic properties of particle verbs in the Germanic languages
Approaches to the morpho-syntax of particle verbs in the Germanic languages
References
Abstract
This article provides an overview and comparison of particle verbs in the Germanic
languages, giving examples of the construction in German, Dutch, Afrikaans, Yiddish,
English, Icelandic, Norwegian, Faroese, Swedish and Danish. In particular, we will
survey some of the basic morpho-syntactic properties ofparticle verbs, which have given
rise to varying syntactic approaches to the phenomenon.
1. Introduction
The aim of this article is to give an overview of the phenomenon of particle verbs (PVs)
in present-day Germanic including English. Particle verbs (also: phrasal verbs, separable
(complex) verbs, verb-particle combinations) are combinations of verbs and prepositionlike elements (see also article 23 on particle-verb formation). Together, verb and particle
form a close semantic unit. In some languages, notably languages in which the particle
precedes the verb in the infinitive, such as German, Dutch, Yiddish and Afrikaans, this
is reflected in the orthography. Examples of particle verbs are send OFF (German ABschicken, Dutch OPsturen, Yiddish AVEKsikn; particles in small capitals throughout), look
UP (German NACHschlagen, Dutch OPzoeken, Icelandic jletta UPP, Norwegian sla OPP,
Faroese slew UPP), ring UP (German ANrufen, Dutch oPbellen, Swedish ringa UPP), drink
UP (German Austrinken, Dutch oPdrinken, Yiddish oJstrinkan, Farnese drekka UPP) and
stand/get UP (German AUFstehen, Dutch OPstaan, Afrikaans OPstaan, Icelandic standa
UPP). Particle verbs are a characteristic feature of the Germanic languages (e.g., Harbert
2007: 366; Holmberg and Rijkhoff 1998: 85), but they are also known in other languages
or language families, for example Romance (see article 36 on particle verbs in Romance)
and Hungarian (see article 37 on particle verbs in Hungarian). Semantically, a verbparticle combination may be anywhere on a scale from transparent to opaque. Particles
may contribute directional, locative, resultative, temporal or aspectual meaning to the
complex verb meaning, or the verb-particle combination may have an idiomatic meaning.
Accordingly, a threefold distinction between compositional, idiomatic, and completive
(or aspectual) verb-particle combinations has been suggested (e.g., Emonds 1985; Jackendoff 2002). As a unit, particle verbs may be intransitive (see (1)), take a nominal
object (see (2)) or appear in syntactically more complex variants such as secondary
predication constructions or double object constructions (see (3)). (See Mcintyre 2007
and article 23 on particle-verb formation, section 2, on semantic and argument-structural
properties of particle verbs, and references given there.)
611
PA612
612

(1)
Intransitive PVs
a. Another opportunity turned
UP.
b. Der Film fangt AN.

The film catches Part
' The film is (just) starting. '
c. Ek staan nou OP.
I stand now Part
' I'm getting up now.'
(Donaldson 1993: 263)
(2)
PV s with a nominal object

a. John sparka UT hunden
John kicked Part the.dog
'John kicked out the dog.'
(Afarli 1985: 75)
b. Hann gjerdi UPP snerio.
He
made Part fishing-line.the
'He wound up the fishing line.'
(Thrainsson 2007: 142)
(3)
German
Afrikaans
Norwegian
Farnese
More complex variants

a. They made John OUT a liar.
(den Dikken 1995: 45)
b. Peir ha/a sent strakunum peningana UPP.
they have sent the boys the money Part
' They have sent the money up to the boys.'
(Collins and Thrainsson 1996: 435)
Icelandic
Unlike prefixes and prepositions, particles are prosod ically strong, i.e. in a verb-particle
combination, main stress falls on the particle and stress in this context may be contrastive. For illustration, consider the minimal pairs in (4) (underlining indicates main stress):
(4)
a.
German
umahren (prefix verb; 'to drive around ' )

(particle verb; ' to knock down ')
~ahren
b.
Dutch
omblazen (prefix verb; 'to blow around ')

oMblazen (particle verb; 'to blow down' )
(Booij 2002: 23)
Particle verbs have been dealt with continually in the linguistic literature from various perspectives, such as their morpho-syntactic behaviour and representation, their semantics, but
also their behaviour in language processing and acquisition (see, for example, Svenonius
1996b; Dehe et al. 2002; Raiden 2006 for overviews).
One major theoretical challenge arises from the fact that verb and particle are separable and have phrasal properties in the syntax but have characteristics of morphological
PA617
617
b. *AUF haben sie das Stiick gefuhrt. I Sie haben das Stuck AUFgefuhrt.
Pa1i have they the piece lead
'They performed the piece. '
(Wurmbrand 1998)
(18) Fronting
a. ARA YN iz er gekumen.
Part/ in is he come
' In, he came (not out). '
Yiddish
b. *OP iz dos ayz nit gegangen.

Pa1i is the ice not gone
'The ice hasn't thawed.'
(Diesing 1997: 384)
( 19) Fronting
a. AF maak ik mijn huiswerk niet.
Part make I my homework not
' I will not finish my homework.'
Dutch
b. *op bel ik mijn moeder niet.

Part call I my mother not
' I will not call my mother.'
(Booij 2002: 24)
2.2. English, Norwegian, Icelandic, Faroese

In Present-day English, Norwegian, Icelandic and Farnese, particles follow the verbal
stem in the infinitive (e.g., English look UP, Norwegian sla OPP, Icelandic fletta UPP,
Farnese slit.a UPP). Due to this order, these languages display contrasts between prepositional verbs and h.omophonous particle verbs (e.g., English look UP a word (in a dictionary) vs. look up the road (to see if someone's coming)). Their syntactic behaviour helps
to tell particle and prepositional verbs apart. For example, a preposition always precedes
its object, while a particle may either precede or fo llow a full NP object and must follow
an unstressed pronominal object. Moreover, a preposition but not a particle can be fronted together with its object. This is illustrated using an Icelandic example from Thrainsson
(2007: 13 9); see (20)- (21 ).
(20) Prepositional verb halda via 'to support; lit. hold with'
a. Tveir menn he/du via stigann/ hann.
two men held w ith stairs.the/it
' Two men supported the stairs/it. '
b.
*Tveir menn he/du hann via.
c.
Via stigann heldu tveir menn.
PA619
619
(27) Clausal complement

a. Jon tok FRAM {ao Maria hefoi fario}.
John tnok Part that Mary had left.
' John explicitly mentioned that Mary had left.'
b.
Icelandic
*Jon t6k [ao Maria hefoi fario J FRAM.

Several factors have been suggested in the literature as contributing to the choice of one
word order over the other (for overviews focusing mostly on English, see Dehe 2002:
76- 80 and Gries 2003: chapter 2). Among these factors are phonological factors such
as word and sentence stress; morpho-syntactic factors such as pronominal vs. fulI-phrase
status of the object (see (22)-(25) above), (in)definiteness, the heaviness, length or
syntactic complexity of the nominal object (illustrated for Icelandic and English in (28)
and (29), respectively), the presence of a directional adverbial, and modification of the
particle (see (35) below); semantic and discourse factors such as the idiomaticity of the
particle-verb unit or the entire verb phrase, and focus or given vs. new information (see
(22c) above and (3 0) below); and other factors such as dialectal variation (e.g., some
Norwegian dialects prefer the particle-object order, see Svenonius 1996a, 1996b). Based
on a large set of English corpus data, Gries (2003) develops a mu ltifactorial stati stical
approach in order to account for the interaction of some of these fac tors. He finds that in
the multifactorial analysis, morpho-syntactic variables outrank semantic and discoursefunctional factors .
Icelandic
(28) NP-Heaviness
a. ?Ste/pan bar {allar storu toskurnar sem vio komum meo ur friinu} INN.
girl.the carried all big bags.the
that we brought from vacation-the PART
'The girl carried all the big bags that we brought from the vacation in. '
b.
Ste/pan bar INN {alfar storu toskurnar sem vio komum meo ur friinu}.
(29) NP-Heaviness
a. ??She sewed the sleeve with lace around the cuff on.
b. She sewed on the sleeve with lace around the cuff
(Olsen 1996: 279)
English
(30) The news value of the direct object affects its position
Michael laboriously puts DOWN the bags, pushes wide the door, picks the bags UP
again and enters,
(Dehe 2002: 130; example from Stephen Fry, Making History)
The alternation between the V-particle-object order and the V-object-particle order possible with non -pronominal objects in English, Norwegian, Icelandic and Faroese has
been referred to as particle shift (PS; e.g., Svenonius 1994, l 996a, l 996b; Thrainsson
2007). On the basis of the similarities between PS in English and Icelandic, Johnson
(1991 ) suggests that English has object shift (OS) similar to Icelandic, a llowing for
overt verb and object movement. However, as Thrainsson (2007: 141 f.) shows, there
PA620
620

are important differences between PS and Icelandic OS, among them the fact that the
presence of an auxiliary blocks the latter but not the former; see (31 ), from Thrainsson
(2007: 141). According to Thrainsson (2007), this suggests that the object moves to a
lower position in PS than in OS.
(31)
a. Eg hef aldrei jlett

UPP nofnunum.
I have never looked Part names-the
' I have never looked up the names.'
b.
Eg hef aldrei jlett nofnunum UPP.
c.
*Eg hef nofnunum aldrei flett
(no shift)
(PS only)
(OS blocked)
UPP.
In a V2 context where the subject is non-initial, verb and particle must be separate and
the particle immediately follows the subject or the object. Th is is illustrated in (32) for
Icelandic (Collins and Thrainsson 1996: 432).
(32)
a.
i gcer
sendu peir peningana
UPP.
Yesterday sent they the money Part

'Yesterday they sent the money up.'
b.
i gcer sendu peir UPP peningana.
c.
*i gcer sendu
UPP peir peningana.
In more complex constructions, such as double object particle constructions (here illustrated for two nominal objects in English and Icelandic), the following particle positions
are possible (see (33) and (34)).
*John sent OUT the stockholders a schedule.

b. John sent the stockholders OUT a schedule.
c. *John sent the stockholders a schedule ouT.
(33) a.
English
(Neeleman 2002: 141)

(34)
a. *i gcer
hafa peir sent UPP strakunum peningana.
Yesterday have they sent Part the boys the money
'Yesterday they sent the money up to the boys. '
b.
(?)i gcer hafa peir sent strakunum
c.
i gcer hafa peir sent strakunum peningana UPP.
UPP
Icelandic
peningana.
As shown in (35) for Icelandic and Norwegian, modification of the particle is possible
only in the V-object-particle order.
(35) Modification of the particle
a. i gcer
hafa peir sent peningana beint UPP.
Yesterday have they sent the.money straight Part.
'Yesterday they sent the money straight up.'
*i gcer hafa peir sent beint UPP peningana.

(Coll ins and Thrainsson 1996: 430)
Icelandic
PA629
36. Particle verbs in Romance
Giovanni mise giit ii telefono.

John
put down the telephone
'John hung up the phone.'
A quite regular minor change in argument structure is illustrated in (3). Here the particle
su in (3b) absorbs the indirect ground argument (sul fuoco) of the verbal base (cf. 3a),
whereas the direct argument (il cajfe) is not affected.
(3)
a.
Metti il cajfe sul fuoco.

'Put the coffee on the stove.'
b.
Metti su il caffe.
' Put on the coffee.'
An increase of argumental roles is less frequent than their reduction, but it is possible,
cf. (4).
(4)
a.
fl cane corre.
'The dog runs.'
b. fl
cane corre dietro ai
gatti.
The dog runs after to_the cats
'The dog runs after the cats.'
A more important and frequent argumental change is the formation of an unaccusative
PV from a verbal base admitting both transitive and/or unergative constructions. This
shift in valency is signaled in Italian by the change of the auxiliary verb: the auxiliary
essere 'to be' is used in the expression of telic events (Sa), the auxiliary avere 'to have'
in atelic ones (Sb). Auxiliary shift in the same verb distinguishes the Italian language
from the current stage of the other major Romance languages, while in their nnitial stages
all of them share these characteristics, cf. sections S and 6.
(S)
a.
fl piccione
volato via.
'The pigeon flew away. '

b.
fl piccione ha volato per tre ore.

'The pigeon flew for three hours. '
Romance PVs are characterized by morphosyntactic cohesion and fixity. Normally, only
clitics (6a) and other light constituents, e.g., adverbs or focal elements (6b), can occupy
an intermediate position between the verb and the particle.
( 6)
a.
Devi buttarla fuori.

'You must throw her out.'
b. Devi andare sempre avanti.

You must go always forward
'You must always go forward.'
629
PA631
631
Cohesion between verb and particle is particularly evident in coordinating structures, where
PVs ( 11 a, b) behave differently from apparently similar prepositional phrases (11 c, d ).
(11)
a. Irene porta su ii tavolo e Anita

le sedie.
'Irene brings up the table and Anita the chairs.'
b. *Irene porta su
ii tavolo e Anita
su le sedie.
Irene bring.3sG up the table and Anita up the chairs.
c. Irene mangia sul tavolo tondo e Anita su quello quadrato.

'Irene eats on the round table and Anita on the square one. '
d. *Irene mangia sul
Irene
tavolo Londo e Anita
eat.3so on.the table round
and Anita _
quello quadrato.
that square.
The frequency of use of PY s in final clause position and the fact that they form a unitary
prosodic constituent, in which the particle plays a prominent role (cf. Simone 1997;
lacobini 2008), may be considered further support for the strong bond between verb and
particle.
Besides the configurations in ( 1) and (2), Italian PV s display a number of other
possibilities, such as reflexive transitive constructions (l 2a), pronominal unaccusative
constructions ( l 2b), and constructions with an indirect argument that may be expressed
by a locative (l 2c) or a pronominal clitic (l 2d ). The pronominal clitic construction is
almost absent in Spanish and Catalan (cf. Calvo Rigual 2008), but it is quite common
in French (cf. Pourquier 2001 ).
(12) a.
tirar=si faori
pull= REFL out
'to get out '
b. precipitar=si fuori
rush=PRON out
'to rush out'
c.
ber=ci sopra
drink=LOC.PRT up
'to drink to forget something'
dar=ci dentro
give=LOC.PRT inside
'to go hard at, crank it up'
d.
saltar=gli addosso
JUmp=PRON on
'to jump on, pounce on'
When the referent of sentences like the one in (12d) is not expressed by a clitic, but by
a nominal autonomous form (noun or pronoun), it must be preceded by a preposition
(usually the preposition a), cf. (13).
(13)
saltare addosso al
nemico
jump on
to_the enemy
'to jump on the enemy'
PA633
(16)
entrare dentro
enter m
'to enter'
uscire fuori
exit out
' to exit'
PVs formed with stative verbs express locational meanings. This kind of construction is
often interpreted metaphorically ( 17).
( 17) stare accanto

stay next to
'to stand by, support'
stare dietro
stay behind
' to keep after, keep up, court'
covare dentro
brood inside
' to harbour'
PVs may display idiomatic meanings independently of the semantics of the verb and of
the particle. Nevertheless, polysemous PV s tend to maintain their spatial meanings along
with the new idiomatic meanings. (18) illustrates some examples of PVs that are used
most often with a non-compositional meaning, rather than a spatial meaning, and (19)
shows how they are used in contexts.
( 18) portare avanti
carry forward
'to carry forward/on/out; to hold, handle, manage, conduct'
tirare avanti
pull forward
' to get by, draw forth; to survive, resist'
tirare su
pull up
'to pull up; to raise, bring up, cheer up, hike up'
venire fuori
to come out
'to come out; to turn out, result, emerge'
( 19) questo modo di portare avanti la discussione
'this way of handling the discussion'
un po' di soldi per pater tirare avanti

'a little money in order to survive'
633
PA634
634
Ho avuto due figli, Ii ho tirati su da sofa.

'I have two kids, I brought them up by myself.'
Alla fine def conteggio e venuto [uori che .. .
'At the end of the tally it turned out that ... '
Another sign of semantic lexicalization and of the entrenchment of the construction is
the coinage of PVs from non-motion verbs. The meaning of such constructions often
results from metaphoric or metonymic interpretation, cf. (20).
(20) fare fuori
do out
' to wipe out; to kill; to eat up'
parfare dietro
speak behind
'to tittle-tattle, buzz about'
tenere sotto
hold down
'to keep down, subjugate, submit, put under pressure'
tenere su
hold up
'to keep up, cheer up'
The verbs used in PVs that are attested more frequently and combine with a higher
number of particles are generic motion verbs, deictic verbs, and verbs of putting or
removal (e.g., andare via 'to go away', mettere dentro 'to put in/away', tirare fuori ' to
pull out', venire su 'to come up/out'). Path verbs permit only particles that express the
same orientation that is already encoded in the verb root (e.g., salire su lit. 'ascend up',
scendere giu lit. 'descend down ' ). Manner verbs are less frequently used, even if they
are much less constrained by semantic compatibility with the particle (e.g., correre vial
fuorildentrolsulgiu 'to run away/out/in(to)/up/down').
Table 36.1 shows the range of spatial meanings that may be conveyed through particles in semantically transparent PVs in contemporary Italian. The data have been extracted from LIP (Lessico difrequenza dell'italiano parfato; cf. De Mauro et al. 1993), which
is a spoken corpus amounting to approximately 500,000 tokens balanced for diaphasic
and regional varieties. The left column of the table lists the particles grouped by semantic
type. The second column shows their meanings. The third column reports the percentage
of use of PVs (tokens). And the rightmost column allows a comparison between postverbal particles and verbal prefixes. Prefixes that cannot be productively used with spatial meaning are underlined; question marks (?) indicate prefixes whose productivity is
doubtful.
The most frequent particle meanings result as: the opposites 'up/down' on a vertical
axis, and the expression of distancing, separation, or removal, rendered with the particles
fuori and via. Fuori and via, considered by themselves, represent the most frequently
used particles, but the corresponding opposite meaning 'in, into' is much less frequent.
The expression of forward movement is well represented, while the corresponding back-
PA635
635
Table 3 6.1: Spatial meanings expressed by particles in semantically transparent PVs in LIP and
comparable prefixes
Italian post-verbal
particles
Spatial
m eanmgs
% in PVs
sopra, su
up, over
14
sop ra-lsovra-, sor-
giu, sotto
down, under
15
dentro
in, into
sottoin-, intro-
Juori, via
aw ay, o ff, out
37
ab-, de- (?), dis-, e-les-, estro(?), s-, se-
avanti
forward
15
ad-, pre-, pro-
dietro, indietro
back
re-lri- (?), retro-
accanto, appresso,
vicino
near, by
giusta-
oltre
across, beyo nd
per-, trans-
intorno
around, about
c1rcum.-
lontano
far
addosso
ag ainst
contro-
between
Ita lia n prefixes
fra-, inter-
ward movement particles occur only a few times. Another important feature of Italian
(as well as of other Romance languages) is the lack of particles and prepositions expressing goal attainment (e.g., Engl. into), and the almost total absence of particles expressing
inherently directional meaning (e.g., Engl. to).
Comparison of the spatial meaning expressed by PVs with those expressed by verbal
prefixes shows that the two ways of expressing direction of motion largely overlap.
However, only a small number of the prefixes can be productively used to express the
direction of motion. Moreover, only one (intro-) of the few productive prefixes listed in
Table 36. l expresses mainly spatial meaning when added to verbs. Whereas the other
prefixes that formerly had a spatial value are now productively employed with mostly
other meanings (the same can be said for the verbal prefixation in the other Romance
languages, cf. Ludtke 1996).
The most frequent PVs in LIP expressing transparent directional meanings are listed
in (21) in alphabetical order.
(21) andare avanti
'to go ahead/forward/on'
andare fuor i
to go out/outside'
andare giu
'to go down'
PA637
637
Tab. 36.2: Proportion of telic vs. atelic PVs (toke ns) in LIP formed w ith the most typically telic
and atelic particles
% PV +Tel
Particle
% PV - Tel
100
via
giit
92.7
7.3
SU
92.l
7.9
fuori
89.7
9 .2
dentro
83.8
16.2
dietro
40.0
60 .0
accanto
12.5
87.5
1.8
98.2
a van ti
intorn.o
100
via 'away' expressing telicity. When added to verbs, the particle via may stress the
completion of the process and emphasize the result of the action expressed by the verb
(cf. strofinare le macchie 'to rub the stains' and strofinare via le macchie 'to rub off, to
wipe off the stains' ). Even though Italian does not present a developed system of aspectual particles, there are nonetheless some traces of regularity. Particles that can be used
to express movement oriented towards a goal or originating from a source may come to
imply attainment of the goal (telic situations) (22a), whereas particles that express stasis,
location, or motion without a specific endpoint may contribute to indicate atelic situations (22b).
(22) a.
saltare
'to jump'
volare
'to fly'
b.
saltare dentro
'to jump in' + TEL
- TEL
volare via
'to fly away'
- TEL
tirare
'to throw, pull'
girare
'to tum'
TEL
TEL
+TEL
tirare avanti
'to get by' -TEL
girare intorno
'to go round and round; to roam, wander
around' - TEL
The results of an analysis performed on the LIP corpus (cf. Table 36.2, taken from Iacobini 2008: 11 3) confirm the high correlation between the meanings expressed by particles
and the aspectual characteri stics of the PVs they contribute to.
Perhaps the most significant difference between standard Italian and the Romance
varieties in which PVs are extensively employed is the more frequent and systematic
use of aspectual post-verbal particles. In Rhaeto-Romance and in the dialects spoken in
the North-East of Italy PVs are currently used to encode telicity (23a) - beside spatial
PA638
638

information - , and they can also express durative (23b), progressive (23c), or inchoative
values (23d), with both motion and non-motion verbs (cf. section 7.3).
(23)
a. comedar fora
repa1r
out
'to repair'
Trentino dialects
ligar su
'to tie up'
b.
vardar le carte
'to look at the documents'
Trentino dialects
vs.
vardar fora le carte

lit . look out
'to examine the documents'
(Cordin 2008a: 184)
c. L'
~
qua.Se dre
ca
la
seca.
3sG.F be.3sG almost behind coMPL 3sG.F dry3sG
'[The plant] is dying.'
Bergamasque dialect
s
e
metit
dre
a locia.
3PL.M REFL be.3sG put.PART.PAST.PL behind to cry.INF
'[The people] started to cry.'
(Bernini 2008: 152-153)
d. I
Bergamasque
dialect
5. Particle verbs in Medieval Romance languages

The current uneven distribution of PVs tilted toward Rhaeto-Romance, N orthem dialects
of Italy, and standard Italian does not reflect the situation in the oldest stages of the
Romance languages. There was a period in which PVs were largely used in all the
Romance languages.
The transition from Late Latin to Romance languages does not coincide with an
abrupt shift from the satellite- to the verb-framed type. Both the univerbation of prefixed
verbs and the emergence of PVs are evident in the initial stage of Romance languages.
The formation of new path verbs (resulting from the blurring of the boundary between
stem and prefix of Latin verbs (24a), or from nominal or adjectival b ases referring to
entities that have an intrinsic direction reference (24b), or from directional prepositions
(24c)) and PVs in early stages of Romance languages (e.g., Old Catalan pujar sus 'to
ascend', Old Provenc;al venir foras ' to come out') may be interpreted as two different
strategies emerging at the same time and sharing the same purpose: the encoding of
motion events.
(24) a.
Old Sp. exir, Cat. eixir, Old Fr. eissir, It. uscire, Rom. a
bessire < Lat. ex-ire 'to exit'
ie~i,
Sard. bessiri,
PA641
par/er
I aparler
'to speak; to talk' I 'to address'
(cf. Martin 2001: 311)
(31)
II
s' entrecommencent a regarder et semont li uns
Old French
they CL.ENTRE.begin
to look
and ask
the one
I' autre de parle avant.
the other to talk forward
' They begin to look at each other and ask each other to start talking.'
(Artu, p. 13, quoted from Burnett and Tremblay 2009: 32)
According to Troberg and Burnett (2011 ), in Middle French texts there is clear evidence
of strong adjectival resultatives, cf. (32).
Middle French
(32) que tricherie abat jus plate
'that deception beats down flat'
(Christine de Pisan, Livre de la Mutacion de Fortune, t. 2, 29; DMF 2009)
6. Va riation amo ng co ntemporary Ro mance pa rt icle verbs

In addition to PVs, contemporary French, together with the majority of Romance languages, has strongly reduced or completely lost the above mentioned satellite-framed
type features (cf. Old Fr. Des que cil furent fort issu - Brut, Wace, 12th century 'As
soon as these people had come out; lit. had exited outside' vs. Fr. des qu'ilsfurent sortis).
Nevertheless, a lot of variation can still be observed among the Romance languages.
Currently, among the major Romance languages, auxiliary shift signaling the distinction between transitive and unergative from unaccusative constructions is allowed only
in Italian. In French, auxiliary selection responds to the same basic requirements as in
Italian, but auxiliary shift in the same verb is very limited. In contemporary Portuguese,
Spanish, Catalan, and Romanian the use of the auxiliary 'to have' is generalized to all
predicates, replacing the old contrast in auxiliary selection (cf. Loporcaro 2007, which,
looking at nonstandard (Italo-)Romance varieties, describes different patterns of auxiliary
verb variation that are unknown in standard languages).
One of the major obstacles to the use of PY s is the fact that nowadays major Romance
languages do not have pairs of prepositions or particles distinguishing locative from
directional meanings (cf. English in/into, on/onto). Inherently directional prepositions
and particles are very few in number (e.g., Sp. hacia, Fr. vers, It. verso 'towards'),
neither of them can unambiguously express the attainment of a goal. A result location
interpretation can be univocally expressed only by complex prepositions (e.g., Fr. jusque
dans, It. fin dentro 'up to the inside of'), but, among Romance languages, this shategy
seems to be commonly used in post-verbal position only in Rhaeto-Romance varieties.
Since the Romance languages have lost most satellites (e.g., directional prefixes) or
other morphosyntactic devices (e.g., Lat. in + accusative case), which clearly express
goal-oriented motion, they tend to express path in the main verb. If manner cannot be
omitted, it is typically expressed in an adjunct, as in (33), that represents a "heavier"
construction if compared with Eng. to run into or Latin incurro.
641
PA645
emigrate', par antes 'to advance',por contra ' to oppose',p6r fora 'to put off, exclude',
ter dentro 'to include, contain' (cf. Batoreo Jakubowicz 2000: 789-791).
Among the few studies on diatopic variation in the lexicalization of motion events
there are: Rojo (1974) on Galician dialects, and the careful survey of Aragonese varieties
by Hij azo-Gasc6n and lbarretxe-Antufiano (20 I 0).
7.2. French
It is generally agreed that French at its cmTent stage is the Romance language that is
farthest away from the satellite-framed type, and in particular the Romance language
that uses PVs least.
In publications dealing with the typological shift from the satellite-framed to the verbframed pattern and with the actual typological intricacy of path expression in French,
Kopecka (2009a, forthc.) focuses her attention on prefixation rather than on PVs. The
a lmost complete absence of PVs in contemporary French is also confirmed by Herslund
(2005) and Hijazo-Gasc6n and Ibarretxe-Antufi.ano (20 13).
We may disti ngui sh two kinds of PVs used in contemporary French. In the first one,
the path verb is followed by a directional particle that reinforces the expression of direction (e.g., descendre en bas lit. 'to descend down', entrer dedans lit. ' to enter into',
monler en haul lit. 'to mount up', sorlir dehors lit. ' to exit out' ). The other, formed from
a short list of manner verbs and directional particles, is characterized by the presence of
a clitic pronoun (which can only refer to animate referents) and by the dominant metaphorical meaning next to the spatial one (je lui cours apres I him/her run after ' I run
after/chase/court him/her '; ii m'a saute dessus ' he jumped on me '; j e vais lui rentrer
dedans ' I am going into him/her ' ). PV s of the last kind are common especially in nonformal use, although they are confined to the spoken modality; cf. Pourquier (2001 ,
2003) for an analysis of their characteristics and a list of verbs and particles employed.
In areas where the contact between French and the Germanic languages is more
direct, we can identify some PVs that are not present in standard French, cf. Kramer
(198 1) about French spoken in Brussels (e.g., lamber bas, sauler bas) and King (2000)
on French spoken in Canada. Both authors argue that the spread of such constructions
is not due to the calque of individual Flemish or English expressions, rather to the
Germanic languages that acted as an enhancing factor favoring the use of a strategy
present at earlier stages of French. Treffers-Daller (201 2) comes to different conclusions,
stressing the importance of individual calques of Brussels French (e.g., recevoir dehors
'to get out') derived from the regional variety of Dutch.
7.3. Rhaeto-Rom ance varieties and dialects of Northern Italy

Rhaeto-Romance varieties spoken in Italy and in Switzerland, and the dialects of the
North-East of Italy (Venetian, Lombard, and Emilian dialects) are the Romance languages that make a wider use of PVs.
In northern dialects such a mode of expression is even more used that in Tuscany, cf.
Mi lanese dafora to give out 'to spend', da gio to give down 'to go down', di su to say up
645
PA647
647
Friulian does not have a synthetic form for the expression of upward/downward direction, while the verbs Jessi ' to exit' and jentrd 'to enter' are used very sparcely and have
a defective paradigm limited to the participle and the infinitive forms. Likewise, according to Bernini (2008: 143-144), the Bergamasque dialect uses PVs exclusively for the
first three meanings, cf. (38), while 'to go into' can be expressed both by a PV (inda de
det) and a path verb (entra).
(38) inda de 'n so
'to go up'
da de 'n S6
'to go down'
inda de/a
' to go out'
In all these varieties PVs represent the most natural and frequent way to express spatial
meanings. The base verbs are mainly generic and causative motion verbs, but also a
number of manner and path verbs are used (39).
(39) oanociarse Jo
'to kneel down'
Trentino dialects
nar via
'to go away'
rampegarse su
' to climb up'
sentarse z6
'to sit down'
cu/ete ju
'to tumble down'
Ladin (Gardenese variety)
ficeltuce ite
' to stuff into'
scaneju
' to climb down'
PVs can also be formed from non-motion verbs (e.g., Trent. star su ' to stand upright').
A consequence of the spread of the construction is the high number of PVs with metaphorical meanings from both motion and non-motion verbs (40).
(40) entra so
enter up
' to understand'
butarse
via
throw oneself away
' to despair '
Bergamasque dialect
Trentino dialects
PA652
652

interference between local dialects and colloquial regional Italian varieties, but these
speakers refrained from using PVs in more formal contexts because these were considered to be dialectally overconnotated. Italian speakers with less or no competence in a
dialect and fewer socio-cultural inhibitions acquired the PVs circulating in the regional
communities and extended their usage to more fonnal contexts. The strong presence of
PVs in regional varieties of Northern Italian has played a positive role in the diffusion
of these constructions in Italian, thanks to the migration (and the subsequent linguistic
integration) of millions of people from Southern Italy to the industrial towns of No11hern
Italy after World War II, and to the fact that Northern varieties of Italian have gained in
prestige. No,wadays, PVs have become a lexical resource of standard Italian that is almost
independent of a regional distribution.
Contemporary Italian PVs do not originate only from the Northern dialects. Although
they are less frequent than in Northern dialects, they have been quite common in Tuscan
dialects since the 131h century at least (cf. Meyer-Lubke 1899: 370; Masini 2006). PVs
have also been documented from the first texts in the other dialects of the central regions,
and (although less used) in the insular and Southern dialects, cf. Rohlfs (1969: 916,
918), Tekavci6 (1972: 634- 694), Loi Corvetto (1982), Cini (2002: 147- 148), Amenta
(2008), Simone (2008: 27- 28), lacobini (2009a).
The presence of PVs in medieval Tuscan texts (as well as in other central and insular
dialects) shifts back to in the Early and High-Middle Ages the presumed Germanic
influence on Italian. It is an influence that can be regarded as a supporting effect to the
diffusion and entrenchment of a construction resulting from an autonomous development
of Romance varieties. This reveals itself as a catalyzer of processes that were already in
use in Late Latin.
7.5. Romanian
The two reference works on Romani an PVs are Vicari o (1995) and Papahagi (2009),
on the expression of motion in Romanian, see also Dragan (2012). The first contains
bibliographical indications and an examination of dictionaries and grammars, the latter
adopts a typological cross-linguistic perspective and is enriched by data extracted from
corpus analysis.
Currently, after Italian Romanian is the standard language in which PVs are most
numerous and most commonly used. However, according to Papahagi (2009), the proportion of path verbs used to encode the trajectory confirms the classification of Romanian
as a verb-framed language.
The list of Romanian post-verbal particles is qu ite rich: inauntru 'inside', afara 'outside', inainte 'forth, ahead', inapoi, indiiriit 'back', dincolo 'beyond', sus (in sus) 'up',
jos (injos) ' down ', drept 'straight', la vale 'downstream', la deal 'upstream', la dreapta
'to the right', la stanga 'to the left'. Romanian seems to be the only Romance language
to have deictic particles (incoace 'hither', incolo 'thither') and an interrogative directional post-verbal particle (incotro 'whereto ' ). The two denominal particles (aJii thread
'straight ', roata wheel 'around') are probably due to a calque from Hungarian, a fact
confirmed by their limited geographical distribution.
PVs may result from both transi tive (44a) and intransitive motion verbs (44b).
PA655
Cadierno, Teresa and Peter Robinson
2009
Language typology, task complexity and the development of L2 lexicalization patterns
for describing motion events. Annual Review of Cognitive Linguistics 7: 245- 276.
Calvo Rigual, Cesareo
2008
I verbi sintagmatici italiani: Appunti contrastivi con lo spagnolo e ii catalano. In: Carmen Gonzalez Royo and Pedro Mogorr6n (eds.), studios y analisis de fraseologia
contrastiva. Lexicografia y traducci6n, 47- 66. Alicante: Universidad de Alicante.
Calvo R igual, Cesareo
20 IO
Trattamento nella lessicografia monolingue (italiana) e bilingue (italiano- spagnolo e
catalano) dei verbi sintagmatici: Panorama attuale e proposte di futuro. In: Maria Iliescu,
Heidi S iller-Runggaldier and Paul Danler (eds.), Actes du xxve Congres International
de Linguistique et Philologie Romanes. Vol. 7, 375- 383. Berlin/New York: de Gruyter.
C ifuentes-Ferez, Paula
2009
A Crosslinguistic Study on the Semantics of Motion Verbs in English and Spanish.
Mtinchen: LINCOM Europa.
C ini, Monica
2002
I verbi sintagmatici negli etnotesti dell'ALEPO. In: Giovanna Marcato (ed.), La dia/ettologia oltre ii 2001, 143- 150. Padua : Unipress.
C ini, Monica (ed.)
2008
I verbi sintagmatici in italiano e nelle varieta dia/etta/i. Stato dell' arte e prospettive di
ricerca. Frankfurt/M.: Lang.
Cordin, Patrizia
Su e giu modificatori del verbo in alcune varieta dell ' italiano . In: N icola Grandi and
2006
Gabriele Iannaccaro (eds.), Zhi. Scritti in onore di Emanuele Banfi in occasione del suo
60 comp/eanno, 2 15-225. Cesena/Roma: Caissa Italia.
Cordin, Patrizia
2008a L'espressione di tratti aspettuali nei verbi analitici dei dialetti trentini. In: Monica C ini
(ed.), I verbi sintagmatici in ita/iano e nelle varieta dialettali. Stato dell' a rte e prospettive di ricerca, 175-192 . Frankfurt/M.: Lang.
Cordin, Patrizia
2008b Spazio fisico e spazio figurato nelle collocazioni verbo piu locativo in italiano e in
alcune sue varieta. In: Gerhard Bernhard and Heidi Si ller-Runggaldier (eds.), Sprache
im Raum - Raum in der Sprache, 3- 20 . Frankfurt/M.: Lang.
Cordin, Patrizia
20 11
Le costruzioni verbo-locativo in area romanza. Dalla spazio all' aspetto. Berl in/New
York : de Gruyter.
De Mauro, Tullio, Fede rico Mancini, Massimo Vedovelli and Mirian Voghera (eds.)
1993
Lessico difrequenza dell'italiano parlato. Milano: Etas Libri.
Diez, Friedrich
1882
Grammatik der romanischen Sprachen . 5th ed. Bonn: Weber.
Dragan, Ruxandra
20 12
Aspects of Lexical Structure. Verbs in Locative Constructions in English and Romanian.
Bucure~ti: Editura Univers itatii d in Bucure~ti .
Dufresne, Monique, Fernande Dupuis and Marie-Catherine Longtin
2001
Un changement clans la diachronie du frarn;:ais: La perte de la prefixation aspectuelle en
a-. Revue quebecoise de /inguistique 29: 33- 54.
Dufresne, Monique, Fernande Dupuis and Mireille Tremblay
2003
Preverbs and particles in Old French. In: Geert Booij and Jaap van Marie (eds.), Yearbook of Morphology 2003, 33- 60. Dordrecht: Kluwer.
Gonzalez Fernandez, Maria Jesus
1997
Sohre la motivaci6n semantica de las expresiones pleonasticas de movimiento: subir
arriba, bajar abajo, entrar adentro y salir afuera. In: Concepcion Company Company
655
PA658
658

Nikitina, Tatiana
2008
Pragmatic Factors and Variation in the Expression of Spatial Goals: The Case of into
vs. in. In: Anna Asbury, Jakub Dotlacil, Berit Gehrke and Rick Nouwen (eds.), Syntax
and Semantics of Spatial P, 175- 209. Amsterdam/ Philadelphia: Benjamins.
Papahagi, Cristiana
2009
La place du roumain clans la typologie motion event: Esqu isse de description. In: Gavril
Neami , $tefan Gencarau and Adrian Ch ircu (eds.), Limba romanii. Abordiiri tradifionale .$i moderne, 397-408. Cluj-Napoca: Presa Universitara Clujeana.
Porquier, Remy
200 1
'Tl m'a saute dessus', 'je lui ai couru apres': Un cas de postposition en fran9ais. Journal
of French Language Studies 11: 123-134.
Porquier, Remy
2003
'G Ii corro dietro/J e lu i cours a pres': A propos d'une construction verbale specifique en
italien et en fran9ais. In: Mathee G iacomo-Marcellesi and A lvaro Roccbetti (eds.), fl
verbo italiano. Studi diacronici, sincronici, contrastivi, didattici, 49 1-500. Roma: Bulzont.
Quaglia, Stefano
2012
On the Syntax of some Apparent Spatial Particles in Italian. In: Miriam Butt and Tracy
Ho lloway King (eds.), Proceedings of the LFG12 Conference. Standford: CSU. http://
web.stanford.edu/group/cslipublications/cslipubl ications/LFG/ 17/papers/
lfg l 2quaglia.pdf [last access 1 Oct 2014].
Rohlfs, Gerhard
1969
Grammatica storica delta lingua italiana e dei suoi dialetti. Turin: E inaudi [= revised
version of Historische Grammatik der italienischen Sprache und ihre Mundarten. Bern:
Francke 1949- 1954].
Rojo, Gu illermo
1974
Perifrasis verbales en el gallego actual. Vigo: Universidad de Santiago de Compostela.
Sch0sler, Lene
2008
L'expression des traits maniere et direction des verbes de mouvement: Perspectives
diachroniques et typologiques. In: Elisabeth Stark, Roland Schmidt-Riese and Eva Stoll
(eds.), Romanische Syntax im Wandel, 113- 132. Tiibingen: Narr.
Schwarze, Christoph
1985
"Uscire" e "andare fuori": Struttura sintattica e sernantica lessicale. In: Annalisa Franchi
de Bellis and Leonardo Maria Savoia (eds.), Sintassi e morfologia della lingua italiana
d'uso. Teorie e applicazioni descrittive, 355- 37 1. Roma: Bu lzoni.
Schwarze, Christoph
2008
I verbi sintagmatici: Prospettive di ricerca. In: Monica Cini (ed.), I verbi sintagmatici
in italiano e nelle varieta dialettali. Stato dell' arte e prospettive di ricerca, 209- 224.
Frankfurt/M.: Lang.
Siller-Runggaldier, Heidi
2009
Review of: Raphael Bertele Ort und Weg. Berlin/New York: de Gruyter, 2006. Zeitschrift
far romanische Philologie 125: 145-153.
Simone, Raffaele
1997
Esistono verbi sintagmatici in italiano? In: Tullio De Mauro and Vincenzo Lo Cascio
(eds.), Lessico e grammatica. Teorie linguistiche e applicazioni lessicografiche, 155170. Roma: Bulzoni.
S imone, Raffaele
2008
I verb i sintagmatic i come costruzione e come categoria. In: Monica Cini (ed.), I verbi
sintagmatici in italiano e ne/le varieta dialettali. Stato dell' arte e prospettive di ricerca,
13- 30. Frankfurt/M.: Lang.
PA661
37. Particle verbs in Hungarian
verbal particles, form a phonological and semantic unit with their verb. H owever, the
verbal particle differs from the other verbal modifiers in that, unlike them, a) it cannot
satisfy any of the basic verb's arguments, i.e. it cannot be governed by the verb, and b)
the particle verb can be the input into a morphological rule, whereas complexes formed
by the other verbal modifiers cannot take part in secondary word-formation (cf. Kiefer
2003: 177-178, 181-185).
Hungarian particle verbs are referred to as both verbal prefixes and preverbs in the
English linguistic literature. With the term verbal prefix it is the morphological character
of the particle verb's formation and the lexical unity of the verbal particle with the verb
that is stressed, while the term preverb emphasizes the canonical preverbal position of
the verbal particle.
2. Defining the verbal particle, and the verbal particle stock

There is a partial overlap between the category of verbal particles and the category of
adverbs, because the bulk of the verbal particles have grammaticalized from directional
adverbs and have kept their original directional meanings as polysemantic units. The
same lexical unit after the verb, i.e. in a non-canonical position, can often be understood
as both verbal particle and adverb; in the case of some preverbal adverbial elements
syntactically behaving like verbal modifiers the question arises as to whether or not they
should be classified as new verbal particles. This is the chief reason why the transitional
character of the category of verbal particle was emphasized in 20 1h-century Hungarian
linguistics.
It remains to be seen on what grounds an element can be grouped under the category
of verbal particles. Beothy and Altmann (1985) evidently start from a wider syntactic
category of verbal modifier, assuming that Hungarian has around 100 verbal particles
(in their terminology: verbal prefixes), whereas Jakab (1982: 65- 66), on the basis of
lative-type directional meaning and lexical/textual density, excludes 39 out of 75 verbalparticle-like elements chosen from special literature. Kiefer and Ladanyi (2000: 463465, 481) define the so-called productive verbal particle stock considering the following
three criteria: a) non-argument status; b) marking of aspectuality/perfectivity (see section
4.1) and c) occurrence in productive patterns of particle verbs (this requirement was
mainly drawn into the research in connection with defining the newer, controversial
elements as verbal particles on the basis of Ladanyi 2000, see section 3).
On the strength of the above three guidelines, according to Kiefer and Ladanyi (2000:
482), and also D. Matai (2007: 86), the stock of Hungarian verbal particles consists of
the following:
a) verbal particles originating from adverbs - tovabb 'fmther, on(ward)', ujra 're-',
vegig 'to the (very) end, throughout', vissza 'back', at 'across, through" kereszti1l
'across, through ', tul 'over';
b) verbal particles also used with possessive endings, which originated in parallel with
postpositions - ala 'under(neath)', ele 'before, in front of',Jole 'above, over', melle
' next to, quite near', moge ' behind', utana 'after' as well as verbal particles that are
closely related to case-endings: hozza 'to(wards)', neld 'to, at', ra 'on(to)';
661
PA667
37. Particle verbs in Hungarian
667
20 12), but its presence or absence in the sentence is not directly connected to the aspectual value of the sentence, because it is influenced by other factors as well (see section
4.1). However, it is only in perfective sentences that productively formed particle verbs
with a canonically placed (preverbal) verbal particle can occur. An exception could be
provided by the so-called locative verbal particles mentioned in E. Kiss (2005, 2006),
but the elements belonging to this group are not considered as verbal particles in other
Hungarian sources, because - while their syntactic behavior corresponds to that of the
verbal particles - they do not form a semantic unit with the verb, cannot be repeated
unlike the verbal particles and play no role in aspect-forming, either - cf. Kiefer (2003:
185- 186); see also section 2.
E. Kiss (2005, 2006) does not regard verbal particles as aspectual operators, and
ascribes their syntactic function chiefly to the fact that they contribute to the event
structure of complex verbal predicates as secondary predicates referring a) to the state
of the thing or person that experiences the change as a result of the verbal activity resultative verbal particles, e.g., meg:
(2)
vaj
meg-olvad-t.
DET butter megPERF-melt-PAST

'The butter melted.'
b) to the destination of the shifting thing's or person's movement - terminative verbal
particles, e.g., be ' into':
(3)
Janos be-gurit-ott-a
labda-t
kapu-ba.
John roll-PAST-3SG(DEF) DET ball-ACC DET gate-into

'John rolled the ball into the net.'
or c) relates to the existence/spatial place of the thing or person in question - locative
verbal particles, e.g.,fent 'up' :
(4)
Janos Jent alsz-ik
az
emelet-en.
John up sleep-3SG DET storey-on

'John sleeps upstairs.'
According to E. Kiss (2005: 82), the presence or absence of a verbal particle in a sentence is mostly predictable on the basis of whether the verbal particle functions as a
secondary predicate: a verbal particle usually occurs in the sentence in case the event
structure of the predicate is complex. The verbal particle can only be omitted if the
canonical preverbal slot is occupied by a resultative or terminative phrase functioning,
like the verbal particles do, as a secondary predicate, yet contrary to them, having a
descriptive content (cf. E. Kiss 2005: 61):
(5)
Mari zold-re
fest-ett-e
[be]
kerites-t.
Mary green-on paint-PAST-3SG(DEF) [bePERv] DET fence-ACC

'Mary painted the fence green.'
PA676
676

(IO) (Tl) Un nCEud-papillon est un nCEud.
'A knot butterfly (bow tie) is a knot.'
(T2) *Un nCEud-papillon est un papillon.

'A knot butterfly is a butterfly.'
Units that satisfy neither test are secondary exocentric compounds, and units that satisfy
both tests are coordinative - but see below for apparently ambiguous cases.
3.2.1. Attributive units

Test T3 tests the attribution of features from the nonhead's semantic representation to
the head:
(11)
(T3) Une tente-igloo est une tente qui est (comme) un igloo.
'A tent igloo is a tent that is (like) an igloo.'
The analogy manifested in comme rests on a subset of the semantic features in the
representation of the denotatum of N2, namely the function and the hemispherical shape
in the above example, while other features are suppressed. Analogy, however, is less
perceptible in cases like voiture-belier 'car-ram (a car used by thieves to shatter a shop
window)' where N2 is independently metaphori zed in the lexicon, and it is absent from
a unit like date-limite 'date limit (deadline)':
(12) (T3) Une date-limite est une date qui est(@) une limite.
'A date limit is a date which is (0) a limit.'
In such a case, both Tl and T2 give positive results, so we are apparently close to the
category of coordinative compounds. It is possible, however, using various tests which
cannot be detailed here, to ascertain that a unit like date-limite is semantically leftheaded in the same way as tente-igloo is.
The attributive category is productive, and hapax occmTences can be found in the
media, like:
(13) des moutons-tondeuses embauches au fort de Dardilly.

'sheep lawnmowers (lawnmower sheep) hired at the fort of Dardilly'
(with analogical attribution)
A particular subcategory of attribution is to be found in a limited set of units like thonalbacore 'tuna albacore', mesange nonette 'tit little-nun (marsh tit)'. N2 is the name of
a species while Nl is that of a genus, and N1N2 has the same denotation as N2:
(14) (T4) Le than albacore et !'albacore, c'est la meme chose.

'Tuna albacore and albacore, it is the same thing. '
PA679
38. Noun-noun compounds in French
return identical results on differe nt units and show a notable amount of disagreement
among informants. It seems wiser to conclude that these units occupy various positions
along a scale between attributives and coordinatives. In some cases, however, like that
of bebe (as in bebe-phoque 'baby seal'), the N 1 has acquired some of the characteristics
of a prefix (Amiot and van Goethem 2010).
3.5. Exceptions
" In many languages there are exceptions or fuzzy transitions to non-compounding"
(Dressler 2006: 24). The most numerous exceptions are semantically and syntactically
right-headed units like auto-ecole 'auto school (driving school)' or radio-balise 'radio
beacon', where Nl results from the reanalysis of a neo-classical component and may be
considered an "affixoid" (Booij 2009). In bidonvillemasc 'can town (shanty town) ', the
syntactic head is bidonmasc and the semantic head villefem Other exceptions, like the
graphically fused secondary exocentrics chiendent 'dog tooth (couch grass)' and chevrefeuille 'goat leaf (honeysuckle)', are opaci fied remnants of earlier patterns. The next
section is devoted to non-exceptional units only.
4. Grammat ica l charact eristics

This section presents a survey of the grammatical characteristics of French [NN]N units.
Two characteristics of French complicate the analysis of compounding: French does
not have lexical stress, and noun plurals are phonetically identical to the singulars, except
for a few "irregular" nouns rarely present in compounds anyway, so it is often necessary
to rely on spellings and then caution is necessary with Web occurrences. When possible,
examples with audible plurals have been selected (these are in bold type).
a) Syntactic headedness
French [NNJN units are syntactically left-headed with NI communicating its gender
to the compound, as in the following examples from different categories described in
section 3:
(24) [chap eaumasc-clocheremJmasc
' hat bell (cloche hat)'
[sty/omasc bil/efemJmasc
'pen ball (ballpoint pen)'
rJemme fem-commissairemascJfem
'woman police-commissioner'
[pointmasc-virgulefem]masc
' point comma (semicolon)'
679
PA684
684

The morphologist viewpoint is represented by Corbin (1992), who notes that timbreposte 'stamp post (postage stamp)' exhibits a "syntactic rupture", i.e. the lack of a determinant of its N2, which differentiates it from a lexicalized phrase. In Corbin (1997), she
also criticizes Zwanenburg (1992), who compared timbre-paste and hamme-grenauille
'man frog (frogman)' with the phrases le prajet Delars 'the project Delors' and un avacat
ami 'a lawyer friend' and concluded that the former also had syntactic structure but, in
having been used for the naming of entities, they have become lexicalized. Corbin notes
that paste and grenauille have a categorizing effect resulting in NI being the hyperonym
of N 1N2, while modifiers like Delors and ami respectively have an identifying and a
qualifying effect on N 1. The fact that the categorizing effect is not found in the syntactic
constructions leads Corbin to the conclusion that [NNJN units like timbre-paste and
hamme-grenauille result from different rules, i.e. word-formation rules. Fradin in turn
applies the principle that "compounds may not be built by syntax" (Fradin 2009). Like
Corbin, he considers attributives like paissan-lune 'fish moon ', requin-marteau 'shark
hammer (hammerhead shark)' as morphologically formed, but unlike her he claims that
examples like langage auteur ' language author', impot secheresse 'tax drought (a farm
relief tax)' are formed syntactically (Fradin 2003: 195). He later (Fradin 2009) reiterates
this distinction with more explicit arguments: according to him, the perceptible properties
of the compound's denotatum do not cogently produce a subcategory of the head, "N2
never introduces a semantic predicate that N 1 would be an argument of', and these units
exist with prepositional variants. The characteristics described above in section 4 weaken
these arguments: for instance, as we have seen, many relationals are w ithout a safely
attested prepositional variant.
It has become progressively apparent that compounds are problematic in models of
grammar with segregated morphology and syntax. In particular, English "phrasal" compounds show that morphology may take in syntactic constructions (Lieber 1992: 14).
Montermini (2008), examining Italian units like raccolta rifiuti ingombranti 'collection
refuseplur cumbersomeplur (cumbersome refuse collection)', sees in them objects that include syntactic sequences, but are in other respects morphological. This kind of consideration has led to the construction of models where the separation is less strict, for
instance in the framework proposed by Di Sciullo (2005) where morphology and syntax
can exchange word sequences: English compounds are derived in the morphological
space, but French compounds are derived in the syntactic space before being transferred
to the morphology. All categories of compounds are derived by the application of identical rules. As the derivation of French compounds is syntactic, it involves a functional
projection realized as an operator that provides the semantic relation between the two
nouns. In the case of coordinatives, this operator is AND!oR and in the case of relationals,
it is the semantically vague SORT, which may surface in prepositional units (epargne de
precaution 'savings of precaution') or not be realized phonologically (epargne logement
'savings housing') [my examples].
In construction grammar, morphology and syntax are on a continuum, which solves
the disputes over their delineation. Research on compounds in cognitive linguistics has
been focused mainly on semantics (Heyvaert 2009), however, and to the best of my
knowledge, no construction grammar study of French [NN]N units has been published
at the time of writing, but we can refer to an article by Masini (2009), who examines
Italian N+prep+N and V+and+V constructions. In this study [NN]N units like ejfetta
serra 'effect greenhouse' figure among phrasal nouns, an intermediate category, together
PA696
696

Nevertheless, the morphological option for VN formations still seems us to fit better
the data. Perhaps the strongest argument against the syntactic option, put forward by
Corbin (1992: 48), Rainer (1993: 275) and many others, is the virtual absence of V[DP]
compounds (i.e. compounds including the determiner). If VNCs were to be synchronically derived from VPs structures, they should freely allow for, or indeed prefer, a V[DP]
pattern (*porta-la-carta-igienica and the like), which is definitely not the case. Instances
of lexicalized V-Det-N formations are in fact attested (e.g., Fr. trompe-l'oeil id.; lit.
'deceive-the-eye', It. battiloro beat-the-gold 'gold-beater', apart from family names like
Fr. Boileau = It. Bevilacqua lit. 'drink-the-water', It. Cantalamessa lit. 'sing-the-mass',
etc.), but they are exceedingly rare and are best treated, as Gather (2001: 22- 23) argues,
as a separate process, i.e. as idiosyncratic instances of univerbation from frozen clausal
chunks, not qualitatively different from the types It. nontiscordardime 'forget-me-not',
Sp. hazmerreir make-me-laugh ' laughing stock'.
A further strong argument in favour of the essentially morphological nature of VN
formations (the same point is raised, e.g., by Rainer 1993: 75) is given by the interesting
syntactic and semantic restrictions involving both V and N, which will be discussed
below in section 5. They should not be expected if the VN formations were the output
of the same syntactic rule which forms VPs.
5. Constraints on V and N and "peri pheral" fo rmations

Concerning phonology, no categorical restriction appears to hold for VN compounding,
although in all Romance languages a tendency may be detected which favours "short"
V bases. Being "short" translates as bisyllabic in Italian and Spanish, and monosyllabic
in French, due to the different phonotactics of the languages in question. So Rainer
(1993: 269) reports an overwhelming majority (around 80 %) of compounds with bisyllabic V bases in his corpus, and a similar percentage (around 75 %) is found by Ricca
(2010: 244) among over 2,200 Italian VNCs occmTing in a large newspaper corpus. As
for French, Villoing (2009: 194) obtains very similar percentages (80 %) for monosyllabic bases. Percentages become still higher if one considers that many trisyllabic bases
(bisyllabic for French) begin with a vowel, which often undergoes resyllabification with
the preceding word - e.g., an article - and therefore may be considered "extra-metrical".
Methodologically, this can be a tricky issue, since a prevailing number of short bases
might be just a side effect of a more general tendency to avoid words that are too long.
Ricca (2010: 244- 245) checked out this possibility, showing that no similar preference
is displayed among deverbal derivatives of equal length in the same corpus. Thus the
preference for two-syllable verbal bases appears to be specific to VN compounding, at
least for Italian. It is obviously just a tendency, which does not rule out perfectly legitimate (and attested) V bases with up to five syllables at least, like It. macchina distribuisci-biglietti ' ticket-dispensing machine'.
A similar tendency concerning the length of N seems harder to prove, despite some
proposals along these lines (convincingly dismissed by Gather 2001: 16- 17).
As another general limiting factor on VNC formation, it has been often observed that
only a very small set of potential verb bases really gives rise to a significant number of
VNCs. Gather (2001: 11 - 12) quotes some lists of the most frequent bases for French,
PA700
700

cies do exist, which are difficult to deal w ith in a short survey. They seem to concern
essentially specific points in the build-up of forms, rather than semantic features and
general restrictions. In the following, such points of detail will be briefly illustrated in a
contrastive perspective between Italian and Spanish, which happen to differ extensively
in these respects. French wou ld be harder to evaluate at any rate, due to the great difference in relevance of inflectional markers between the graphematic and the spoken code
(the former shou ld not be fully disregarded in this particular domain, given that several
new formations plausibly appear first in the written medium; for some discussion see
Villoing 2009).
6.1. The V stem

The first and most hotly debated issue concerns the precise character of the V element.
In this domain care should be taken in keeping the synchronic and diachronic perspectives well separated.
The three positions on the nature of V, as surveyed for instance in Rainer (1993:
265- 268) and Gather (2001: 87- 92), are: (i) V is an imperative 2nd person form; (ii) V
is an indicative 3rd person form; (iii) V is a bound verbal stem. However, if formulated
as above the three alternatives make sense chiefly as a diachronic issue. Clearly, this
was also the main focus of the debate until the first half of 20th century. The imperative
proposal was most energetically defended by Darmesteter (1894) with respect to French
and, according, e.g., to Rainer (1993: 265), essentially won the contest unti l the revival
of the indicative alternative, prompted especially by Tollemache's (1945) study of Italian.
Not only historical, but also semantic argumen ts were put forward in defending the
different options. For instance Tollemache's plea for the indicative, despite its formal
implausibilty for Italian (see below), rested on the paraphrasability of most VNCs as
'entity who does the action V on N'.
From a contemporary and synchronic perspective however, the option that the V
occurring in VNCs is a kind of verbal stem seems hardly questionable, as stressed, e.g.,
by Villoing (2009: 176- 179) for French. Whatever form V takes in these compounds, it
carries just its lexical content, not the meaning associated with any specific verbal inflectional marker: neither the imperative nor the indicative, and much less so a specific
person value. Consequently, whatever the ultimate origin of the model(s) which triggered
the morphological template (quite plausibly the imperative: for a partial re-evaluation of
its role even in synchronic word-formation, see Floricic 2008), the present-day productive formation rule should inevitably be formulated in terms of a verbal stem (a morphome in the sense of Aronoff 1994), and not of verbal inflectional forms. The question
remains, however, how such a morphome is formed, and some cross-linguistic diver gence is found in this respect.
In Italian , the V stem in VNCs formally coincides with the 2nd person imperative for
all verbs. There is no possibility of relating it to yd person indicative, because the two
forms differ for all verbs in 2nd and 3rd conjugations (indicative regg~ 'holds', apr~
'opens' vs. imperative regg{, aprD. In Spanish, the opposite pattern holds: while there is
a nearly complete overlap between the forms of 2nd imperative and 3ra ind icative for all
three conjugations, in the rare cases where they contrast (e.g., with tenir 'keep' and
derivatives) the few instances of lexicalized compounds show the yd s ingular form, and
speakers agree that possible new formations would do the same (Val Alvaro 1999: 4789).
PA702
702

The pattern in Spanish is different. Singular VNCs display the -s plural marker on
N in the great majority of cases. The -s has extended even to unique referents (cf.
trotamundo_ 'globetrotter' vs. It. giramondo above) and to mass nouns (cf. Sp. quitanieve_ 'snow plough', rompehielo~ 'ice-breaker' vs. It. spazzaneve, rompighiaccio), although exceptions remain (a list in Val Alvaro 1999: 4798), not only within the "stableN" type, but sometimes even among "variable-N" compounds (Am. Sp. chupaflor suckflower 'hummingbird'). Such an extension implies a substantial change in the role of
the -s marker. As already noted by Rainer and Varela (1992: 130), the -s in Spanish
VNCs has mainly taken the role of marking the composition process itself, with no
consistent semantic motivation. That is, it has come a long way to becoming a morpheme, similar to the linking elements familiar from the compositional processes in
Gennan and other languages (cf. article 32 on linking elements in Germanic) with the
interesting difference that the -s marker occurs at the word's end and not at the border
between the lexical morphemes. However, the generalization of -s in VNCs is surely not
a concluded process, and probably has di fferent relevance across the Spanish-speaking
world: for a variety of Argentinian Spanish, Rainer (1993: 272) even reports the opposite
tendency of omitting the -s in the singular only.
Catalan seems to take an intermediate position between Italian and Spanish: according
to Gracia (2002: 798-99), the plural N is the normal choice for countable Ns, but does
not extend to unique referents (rodam6n vs. Sp. trotamundo~) or mass nouns (llevaneu
vs. Sp. quitanieve.f).
6.3. The plural of the w hole compou nd

The above variation in the marking of the internal N plural has obvious consequences
for the possibility of marking external plurals, i.e. the plural of the whole compound.
Nowhere in Romance languages is there an independent slot available for this function.
Consequently, all VNCs with internal plural Ns are necessari ly invariable. Very marginal
exceptions may at most occur if the compound is so deeply entrenched in the mental
lexicon that it may be perceived as non-analyzed unit. For instance, the Italian plural i
paracadut 'the:M:PL parachutes' is not found in dictionaries, but on the Internet it is
about as frequent as the expected i paracadut~. This deviant form necessarily implies a
reanalysis of the (F):PL -e ending, occurring in the singular ii paracadute stop-fall:PL
'parachute', as the (M):SG ending of the -e/-i inflectional class (as in can-e 'dog'),
which in turn cannot happen w ithout a substantial opacization of the compound.
A different state of affairs occurs with VNCs displaying a singular N in the singular.
Since these are a minority in Spanish, Italian is more interesting in this respect. As a
general rule, it appears that the possibility of marking the whole VNC as plural by
pluralizing the internal N obeys plausible semantic distinctions. Both "variable-N" and
"stable-N" compounds may display a plural N when they have plural reference, but there
is great oscillation and in most cases both forms are possible and largely attested: i
cacciavit?/ thrust-screw(s) 'screwdrivers', i coprilettQ/ cover-bed(s) 'bed covers'. This
makes sense semantically, since for both subclasses a plurality of VNC referents normally implies a p lurality of Ns as well. Therefore, there is no need to ana lyse the plural -i
in spazzacaminf sweep-chimney:PL 'chimney sweeps' and the like - as done by Scalise
(1994: 139) - as an instance of external plural implying that the compound is treated as
PA703
39. Verb-noun compounds in Romance
an unanalysable whole, like paracaduti above. The -i is still an internal plural compatible
with the inflectional class of N, and quite naturally licenses the plural interpretation of
the whole compound.
The latter analysis is supported by the fact that when there is a gender clash between
the N and the VNC, the compound is mostly invariable: i portabandierQ carry-flag (f.)
'standard-bearers (m .)'. The clash may be overcome (a plural portabandier~ 'the:M.PL
carry-flags' is possible, as witnessed by the many occurrences on the Internet), but
always in favour of the in ternal plural (here the (F):PL ending -e of bandiera): a form
like *i [portabandier]-i, with the -i modelled on the inflectional class of masculines in
-a (like poetQlpoet) is unacceptable.
6.4. Interaction with other morpho logica l domains

Traditional lexicalist models usually order derivation before compos1t10n (cf. Scalise
1984: 115- 122), which would imply that VNCs may freely take derived Ns as input,
but would rule out the inverse case, apart from idiosyncratic instances of very opaque,
essentially unanalyzed compounds. The first hypothesis is easily confirmed by the data,
as stated also by Gather (2001: 142). To give just a single example, cf. It. copri[teiera] = Fr. couvre[the-iere] = Sp. cubre[te-teras] cover-teapot 'tea cosy'. Maybe the high
frequency items are not very numerous, but in corpora many more can be found, although for Spanish the possibility of internal derivations is rejected in NGLE (2009:
747).
The further derivability of VNCs is indeed less straightforward. At least for Italian,
data suggest a neat distinction between prefixal and suffixal derivation, which is probably
to be expected, given the recognized higher autonomy of prefixes with respect to their
bases. Obvious examples of prefixes freely combining w ith VNCs are anti- 'anti-,
against', which attaches without problems to the w ide subclass of [+animate] derogatory
VNCs (e.g., anti-ficcanaso 'anti-meddler'), and ex- 'former', which simi larly freely applies to the equa!Iy wide domain of job nouns (e.g., ex-cantastorie 'former storyteller ' ).
Both types have very solid attestations on the Web.
As for suffixation, the picture seems different. Surely in every language some well
established examples can be found, but often they involve partially opaque VNCs (this
is the case for It. paracadut-ista = Sp. paracaid-ista 'paratrooper', It. guardaroba guardthing 'wardrobe ' ---+ guardarob-iera 'wardrobe maid', Sp. paraguas protect-water 'umbrella' ---+ paragii-ero 'umbrella maker', as also Gather 2001: 143-144 suggests). Similar
considerations hold for N ---+ V conversions: among the few Italian instances one could
list paracadutare ' to parachute', ficcanasare 'to nose around', rendicontare 'to report'
from the highly lexicalized paracadute, ficcanaso ' meddler ' and rendiconto give(ac)count 'report (N)'.
Such a restriction may be weakened, however, for some specific suffixes which more
generally display practically no structural limitation on their input: the best example in
Italian is -ismo '-ism' . So the model of qual ity nouns from [+animate] derogatory VNCs
(like rompiball-ismo, from rompiballe break-balls 'bore') is presumably productive,
since it is widely attested, at least on the Web, with many VNC bases. The situation is
probably similar in Spanish: the only example of a derived VNC cited in the NGLE
(2009: 771) is chupamedismo, from chupamedias suck-socks 'flatterer', and several
703
PA704
704

quality noun s parallel to the Italian ones are abundantly present on the Web, with some
even attested in dictionaries, e.g., papanatismo 'dumbness'.
Finally, as for the interaction of VNCs with evaluative morphology, we do not expect
to find significant restrictions against diminutive/augmentative Ns serving as input for
VN compounding, neither semantically nor theoretically, as long as the Ns in question
have some referential autonomy, and this is certainly the case in Italian (cf. porta[telefonino] carry-telephone:DIM ' mobile holder ', guarda[port-one] guard-door:AUGM 'doorkeeper'; examples could be multiplied). Such internal diminutivization is however ruled
out in NGLE (2009: 74 7) for Spanish, similarly to derivation in general.
The possibility of applying evaluative morphology to the whole compound is more
interesting a nd again displays some degree of contrast between Italian and Spanish.
Whi le the option seems rather restricted in both languages, its formal realizations differ.
Spanish prefers the diminutive - when acceptable - to attach at the N, although the
semantics involves the whole compound, e.g., cortauii.as ~ cortauiiitas cut-nails:DIM
' [little [nail clipper]]' (NGLE 2009: 770). However, the picture is made complex by the
possibility of interpreting the above formations as infixed diminutives according to the
pattern Carlos ~ Carl-it-os (Gather 2001: 150). In Italian, on the contrary, diminutives
referring to the whole compound are not as marginal as Gather (200 l :150) would suggest, but they overwhelmingly apply externally : so paracadute ~ [paracadut]-ino ' little
parachute (e.g., on hang gliders)' vs. *paracadutine, portafogli carry-sheets 'wallet' ~
portafoglino/portafoglietto 'little wallet' (not *portafoglini), guardaroba ~ guardarobino 'child wardrobe'. This holds surely for the denotative uses of evaluatives, while
for the pragmatic use, the "landing site" may vary: so for the meaning ' little bore m./f',
rompiball-ino/a '[break-ball(s)]-DIM:M/F:SG' alternates on the Web with rompiballine
' break-[ball:DIM:(F):PL]'.
7. References
Aronoff, Mark
1994
Morphology by Itself. Stems and Inflectional Classes. Cambridge, MA: MIT Press .
Baayen, Harald
200 1
Word-Frequency Distributions. Dordrecht: Kluwer.
Bauer, Laurie
The typology of exocentric compounding. In: Sergio Scalise and Irene Vogel (eds.),
2010
Cross-Disciplinary Issues in Compounding, 167-175. Amsterdam/Philadelphia: Benjamms .
Booij, Geert
1994
Against split morphology. In: Geert Booij and Jaap van Marie (eds.), Yearbook of Morphology 1993, 27- 49 . Dordrecht: Kluwer.
Bisetto, Antonietta
1999
Note sui composti VN de ll'italiano. In: Paola Beninca, A lberto Mioni and Laura Vanelli
(eds.), Fonologia e morfologia dell'italiano e dei dialetti d'Italia. Atti de/ XXXI Congresso internazionale di studi de/la Societa di Linguistica Jtaliana, 505-538. Roma:
Bu lzoni.
Bork, Hans Dieter
1990
Die lateinisch-romanischen Zusammensetzungen Nomen + Verb und der Ursprung der
romanischen Verb-Erganzung-Komposita. Bonn: Romanistischer Verlag.
PA713
40. Co-compounds
Whether there is a co-compound-specific double marker does not depend only on the
semantic subtype of the co-compound. Not all generalizing co-compounds in Mordvin
are marked with -n'ek/nek (Walchli 2005: 140). What is important to note here is that
for many languages there is no single general rule that can produce the form of all cocompounds and not even of one semantic subtype of co-compounds.
The morphologically most complex co-compounds in Europe are probably found in
Northwest Caucasian languages which are known for their high degree of polysynthesis.
In Abkhaz it is no problem to have co-compounds with pa1is consisting of more than
two words. However, there is usually one part repeated and one part varied as in (7).
The co-compound 'lf-axa ='l(a-me 'two-day=two-night' is even interesting from a semantic point of view; while 'night and day' is natural coordination, be it generalizing as in
(6) or additive as in Latvian vien-u dien-nakt-i 'one-Acc:sG day-night-Acc:sG > one day
(period of 24 hours) ', (7) from Abkhaz is a rather unusual kind of additive co-compound.
It would be expected that ' two' should have scope over the whole co-compound, but it
is part of each member.
(7)
'y-axa= 'ya-m~
'a-ff aq""a- 'ra j-a-'[e' a-n
Abkhaz
two-night=two-day its-nick-MASDAR it-it-be.in-FIN
' [Since it (the woodpecker) knew that, if he (the hero Abrskj'yl) went to sleep, he
would actually not so quickly wake up, it perched calmly beside his staff and
spent] 2 days and nights chipping away at it.'
(Hewitt 2005: 242)
In (8) the prefixes na- 'thither' and a:- ' hither' yield a typical generalizing co-compound
(' > in all directions'). Similar generalizing co-compounds with alternating prefixes can
also be found in K artvelian languages (see Walchli 2005: 201).
(8)
a-'xw()ada-'kwa na-pa=a:-p'pa-rta-n
Abkhaz
the-hillock-PL thither-look=hither-look-place-FrN
' [And it is known why Abrskj'yl would choose hillocks as hiding-place:] hillocks
were places with an all-round view; [as soon as his enemies emerged, they at once
fell under his sights.]'
(Hewitt 2005: 222)
The morphologically complex co-compounds in Northwestern Caucasian are a morphological variant of discontinuous co-compounds which are not characteristic of European
languages, but occur in many languages of Asia (e.g., Hmong, Khasi, Karen), Mesoamerica (e.g., Mixe, Chinantec) and New Guinea (e.g., Toaripi). In discontinuous cocompounds, the parts of the compound A and B are interrupted by a repeated element
C, often a word, according to the formula CACB or ACBC. An example from Hmong
Daw where discontinuous co-compounds are very frequent is the synonymic co-compound teb chaws ' land land > land' as in kuw lub teb lub chaw ' I CL land C L land, my
land' or Yawm Pus teb Yawm Pus chaw 'Yau Pu 's land'; Bisang 1988: 36, 56, 37;
Walchli 2005: 102).
A further peculiar feature of co-compounds in some Caucasian languages is that they
can contain affixes for 'and', which can be observed in the two synonymic co-com-
713
PA714
714

pounds in Abkhaz in (8). Suffixes for 'and' in co-compounds can also be found in some
Nakh-Dagestanian languages.
(9)
wa-ra-j= 'saxia-j
xidzo-j= 'p~a-j
Abkhaz
look-ABST-and=face-and name-and=reputation-and
@-z- 'ga-ma-z
'pn waz-ba
ssajr-k'
they-whom-be.lacking-not-NFIN woman-young lovely-a
' [Very long ago there was in Abkhazia] a wonderful maiden who lacked neither
looks nor reputation.'
(Hewitt 2005: 209- 21 0)
The examples from Abkhaz show again that the form of co-compounds cannot be generated by a single rule. Some co-compounds have overt markers of coordination, some do
not. While prefixes are usually repeated in co-compounds, there is a more varied picture
as far as suffixes are concerned. In some verbal co-compounds suffixes occur on both
parts, in others only on the second part as in (8).
3. Co-compounds in Eu rope and elsewhere

Co-compounds are characteristic of languages of three macro-areas: Eurasia, New Guinea and Mesoamerica. In Eurasia the highest density of co-compounds is found in the
easternmost continental languages (e.g., Mandarin, Vietnamese, Hmong) and the frequency of co-compounding decreases towards the west. Europe is the westernmost part
of this continuum, there is no special European kind of co-compounding. Co-compounds
in the Caucasus, for instance, have about the same frequency level in European and
Asian Caucasian languages at least as far as non-Indo-European languages are concerned. European co-compounds are simply Eurasian co-compounds in languages with
moderate, low or very low levels of co-compounding. A first question to address then
is whether there are any characteristic ways in which Eurasian co-compounds differ from
co-compounds in New Guinea and Mesoamerica.
For at least some languages of Mesoamerica, metaphorical co-compounds are highly
characteristic. In Classical Nahuatl there are almost only metaphorical co-compounds,
e.g., yn j-petla-tzin in i-cpla-tzin 'DEF hi s-mat-HON DEF his-seat-HON > throne, government' and they are characteristic devices of elaborate style (Lehmann and Kutscher
1949). Metaphorical co-compounds do occur in European languages (see, e.g., example
(24)), but they are never as dominant as they can be in a Mesoamerican language.
Languages in New Guinea with co-compounds including Tok Pisin typically have a
co-compound for 'people' consisting of the parts 'man-woman' and this form tends to
be highly frequent in texts, e.g., Tok Pisin man-meri 'man-woman > people', Abau uwrsa ' man-woman > people', East Kewa ona-aa 'woman-man > people', Kobon nibi bi
'woman man > people'. The same also occurs in some Australian languages with cocompounds: Wik Mungkan pam wanch 'man woman > people'. In all the languages
mentioned here this co-compound occurs on average in about every seventh verse or
more of the translation of the New Testament. Therefore, a useful simple strategy for
finding co-compounds in New Guinea is to consider the expression for 'people'. This
test does not work at all in Eurasia.
PA715
40. Co-compounds
715
It is not easy to decide which European languages have and which ones lack cocompounds. Co-compounding is a continuous, not a discrete feature. Co-compounds are
not evenly distributed across different registers and styles in a language. There are, for
instance, huge differences between newspaper, fiction and traditional folklore in Erzya
Mordvin (Walchli 2005: 220). In Russian, most texts in the literary language lack cocompounds almost completely, while they are quite frequent in the epic Byliny (see
example (2) above) and are common even in colloquial style in Modem Russian. Tkacenko (1979) observes an areal continuum across East Slavic varieties in the frequency of
the co-compound zili=byli 'lived=were > (once upon a time there) were' in the beginning
of fairy tales (even though zili=byli ' lived=were' originally might derive from a pluperfect, see Kiparsky 1967: 230 for discussion). For cross-linguistic comparison it is most
convenient to consider parallel texts which largely represent the same register and style
(Walchli 2005, ch. 6).
The most reliable single word indicator for co-compounds in Eurasia is 'parents'.
Nominal additive co-compounds are a central group of co-compounds and within them
kinship terms tend to be salient and the domain 'parents' is paiiicularly important for
natural coordination of kinship terms. However, not even the 'parent' -test divides Eurasian languages into two neat groups. There are some languages such as Georgian and
Turkish where 'parents' can, but need not, be expressed by a co-compound and in Latvian there is usually a co-compound for 'parents' only in folk songs (in the dainas) but
not elsewhere. This is opposed to languages such as Avar, Mordvin and Chuvash where
the co-compound is the only lexicalized option for 'parents' (see Map 6.2 in Walchli
2005: 217).
The areal east-west distribution of co-compounds is salient for instance in FinnoU gric and Turkic languages. Pitkanen and Heikkila (see article 176 on Finnish) claim
that dvandva compounds are the oldest type of compounds in Finnish, but the Finnie
languages and Hungarian have a considerably lower frequency of co-compounds than
Mordvin and Komi. Uotila (1980) observes parallels in the extent of co-compounding
in Finnie and Baltic languages.
In Turkic languages across Eurasia co-compounds form an areal cline. As a tendency
they are more frequent in Eastern than in Western Turkic languages irrespective of the
genealogic affiliation within Turkic. It is no coincidence that co-compounds are not
mentioned for Karaim in Lithuania which is strongly influenced by Slavic languages.
Example (10) shows that Karaim prefers a Slavic loanword for 'parents'.
juv' -d' a.
( 10) bu =mo dz' ed-lar Jax
DEM=Q parent-PL NEG.EX house-Loe
'Are the parents not at home?'
(Kowalski 1929: 112)
Karaim
An extremely simplistic and therefore inaccurate approximation to describing the occurrence of co-compounds in Europe would be to say that co-compounds are lacking in
Indo-European and Semitic and are present in all other languages of Europe. In fact, if
you browse through the articles in the final two volumes of this handbook you will find
co-compounds described for most non-Indo European languages but not for most lndoEuropean languages of Europe. Notable exceptions for Indo-European include Tat and
Modern Greek (and rural Russian for which there is no article).
PA716
716

Co-compounds in Greek are an innovation. Accord ing to Browning (1983: 99- 100)
dvandva compounds which first appear in late Hellenistic Greek become particularly
frequent in the Turkish period. However, this does not mean that Greek co-compounds
are generally due to language contacts with Turkish. Especially their form is very different. Greek co-compounds are exceptional in representing phonological words. This may
be because at least some co-compound s have developed from exocentric compounds. In
Classical Greek andr6gzmos ' hermaphroditic' is an adjective with a nominalized masculine form andr6gunos 'coward, hermaphrodite'. According to Debrunner (1917) there is
only one attested case in Classical Greek where, in an epigram, the adjective can be
interpreted as a co-compound: andr6guna (PL:N) loutra 'baths for men and women ' and
there it is an adj ective. It may therefore be assumed that Modern Greek andr6-guno
'man-woman >couple' originates from an exocentric compound and not from coordination diachronically. Turkish co-compounds, however, often look very much like sequences of two words, especially in verbal co-compounds where the first part is a converb.
Yet, there are some close lexical parallels between Modern Greek and Turkish, consider,
e.g., (11) and (12) from the New Testament.
(11)
6so
pern-i
na
anigo-klisi
to
as.much take-PRs3sG in.order.to open-close:sUBJ3sG DEF:sG:N
mati
eye
' in the twinkling of an eye'
(1 st Corinthians 15:52)
(12)
goz ~
kapa-yan-a
dek ...
eye open-cNV close-PTCP-DAT as.far.as
' in the twinkling of an eye'
(1st Corinthians 15:52)
Modern Greek
Turkish
In translations into Turkic languages it is common to express the passage with two verbs
' to open' and 'to close' where the first is a converb (e.g., Kazan Tatar, Uzbek). In Kirgiz
and Turkmen the collocation is hyphenated, marking in orthography that it is a cocompound. While most other translations use different strategies, "opening and closing
of an eye" is not completely restricted to Turkic and Greek. In the Balkans we can add
Albanian hap e mbyll syte 'open and close eye:PL' and outside areal reach Portuguese
num abrir e fechar de olhos, Spanish en un abrir y cerrar de ojos (only in the modern
translation en Lenguage Sencillo) and in Africa Wolof ci xef=ak=xippi 'PREP blink=
with=close.eyes'. The example shows that it is difficult to delimit co-compounds from
other looser coordination strategies expressing natural coordination and that it is difficult
to track down language contact effects beyond any doubt.
The most notable exception to the areal east-west co-compound cline in Europe is
Basque w hich has far too many co-compounds for its western location. But if we now
turn to lndo-European western languages, not even these lack co-compounds altogether
as can be seen in examples (13) and (14) from French and Spanish. Additive co-compounds of adjacent periods of times such as days of the week are not unusual in Western
Europe as, for instance, in (13) from French.
PA718
718
that they are context-dependent and context-renewing, to use a common slogan from
conversation analysis (Atkinson and Heritage 1984). In order to illustrate this point, I
will discuss selected co-compounds from one particular text in one European language
Erzya Mordvin, the story Vit'a Kil'd'az'even' er'amo 'The life ofVitya Kil ' djazev ' (first
part) by Vasilij Kudaskin published in the Erzya literary journal S'atko (Kudaskin 1996).
It is a story of fourteen short chapters wh ich is representative of Literary Erzya Mordvin.
Vitya is a naughty schoolboy in a Mordvin village during the time of Stalinism before
World War I I.
The most obvious discourse structuring aspect of co-compounds is their use with
proper nouns. A co-compound consisting of two names expresses that there is a close
relationship between two persons; most often this is used for siblings, spouses or other
pairs of relatives. In Kudaskin (1996) there are co-compounds of Vitya together with as
many as four different other schoolboys and these work to structure the text into episodes
of different adventures w ith different companions. Typically the other boy is first introduced as a simple name before the additive co-compound of proper names occurs. In
(15) Mikita is Vitya's desk neighbor with whom he quarrels which is why the teacher
makes the two of them stand in the comer. The co-compound illustrates the double plural
marking mentioned in section 2 above.
(15)
Vit'a-t=Mildta-t
pr'a-n'
komavto-z'
Erzya Mordvin
Vit'a-NoM:PL=Mikita-NoM:PL head-GEN bow-PTC
sirga-s-t'
ugol-s
set.off-PST-3PL corner-ILL
'Vitya and Mikita went to the corner bowing their heads.'
(Kudaskin 1996: 7)
The co-compounds of proper names depend on the previous context and create groups
which are locally valid in episodes, not in the language as a whole, not even in one text
as a whole.
Another context-dependent use of co-compounds is distributivity. Example ( 16) describes how Vitya as an older boy walks a long way to the school in town under d ifficult
circumstances "through snow and dirt, through rain and storm" which is expressed by
two co-compounds. While rain and storm often co-occur, snow and dirt can occasionally
be combined, but more often they occur on different occasions. There is snow in winter
and dirt in autumn and spring. The co-compound thus summarizes different difficult
road conditions which apply on different occasions. The expression "wade through snow
and dirt" does not make much sense if considered in isolation of this distributive context,
marked here with the adverb s'ejet'ste 'often'.
(16)
... s'ejet'ste kel'e-ms lov-ga=rudaz-ga,

often
wade-TNF snow-PROL=dirt-PROL
Erzya Mordvin
piz'eme-n' =davol-on' pack ...

rain-GEN=storm-GEN through
' [He studied in the second shift. He had to walk home in the dark and] often to
wade through snow and dirt, through rain and storm.'
(Kudaskin 1996: 46)
PA720
720

contextually synonymic co-compound in the context of a sitting accommodation, is hardly ever used other than in combination with mujems 'to find' in negation. In other texts
in Literary Erzya Mordvin it sometimes has the metaphorical use of "not having found
yet the right place in life". This co-compound is not listed in the dictionary however.
Kudaskin (1996) contains two rather specific uses of it which demonstrate that the meaning of the idiom is much less fixed than the formal collocation of the co-compound with
a lexical verb and negation. In the first passage it is the first day in school for Vitya, but
he is so excited about the apples on the apple trees outside that he cannot concentrate
on anything in the first lesson. He does not find "place-bench" (tarka =ez'em e-z' muje
'place-bench NEG-PST3sG find') which is meant entirely figuratively. It does not mean
that he cannot find a seat. In the second passage the teacher is putting on stage an
atheistic play at the end of which an actor puts a horse collar on the neck of an orthodox
priest. The mother of the boy who is supposed to act the communist's role takes her son
home saying that it is not good to mock people of God. This is now where the teacher
does not find "place-bench": tarka=ez'em e-z' mu-ksno 'place-bench NEG-PST3SG find FREQ' . He is put off his stride for a moment and does not know what to do; until he
sees Vitya and so Vitya unexpectedly becomes an artist. What is interesting about this
co-compound from a formal point of view is that it is never used with a suffix, there are
no plural markers as is characteristic for most otherwise unmarked co-compounds and
there is never a possessive suffix. This testifies to the idiomatic character of the cocompound. However, the semantics of the idiom is much less fixed than its form. The
only thing that is clearly fixed is that the meaning must be metaphorical; it can mean
anything ranging from 'not being concentrated' to 'being put off stride' to 'not having
found one's vocation in life yet'. All these uses have a family resemblance, but the
actual meaning of the co-compound can only be determined in a given context.
Preferred collocations are not restricted to figurative co-compounds, however. The
common verbal co-compound lovnoms=s'ormadoms 'read=write' hardly ever occurs in
any other context than that of learning or being able to read and write. In Kudaskin
(1996: 39) it is used to state that Vitya's father learned to read and write himself without
going to school.
On the opposite pole of least context-dependent lexical units we have the fully lexicalized example t'et'a-t=ava-t 'father-PL=mother-PL > parents' which is not only used for
pairs of parents, but even in t'et'a-n'=ava-n' promks-so 'father-GEN=mother-GEN meeting-INE > in parents' meeting' (Kudaskin 1996: 36). However, some co-compounds of
kinship terms are quite contextual. The girl Masha which replaces Mikita as Vitya's
bench neighbor after the event in (15) is very quiet; she goes to school to relax from
housework and her eight brothers and sisters : kavkso sazor-tne-d'e=jalaks-tne-d'e 'eight
younger.sister-PL:DEF-ABL=younger.brother-PL:DEF-ABL'. The co-compound tells us not
only that Masha is older than all her brothers and sisters, but also that the younger sisters
are somehow more prominent than the younger brothers (perhaps there are more sisters
than brothers or Masha has to take care more of her sisters than of her brothers). In
conventionalized co-compounds of kinship terms in Erzya the order is male-female.
Masha's quiet character has the effect that her name never enters a co-compound together
with Vitya, she is not a companion in any of his adventures.
PA721
40. Co-compounds
721
5. Co-compounds and construction morphology

Here we will consider to what extent construction morphology (see article 12 on word formation in construction grammar) might be useful to deal with co-compounds. At least
at first glance, construction grammar seems very useful for co-compounds, given the
relevance of both semantic properties (natural coordination, superordinate-level concepts) and formal properties (juxtaposition of two words) for co-compounds. Construction morphology allows for holistic properties of word structure which is appealing for
co-compounds. We might therefore try to formulate a general schema for co-compounds
such as (19):
(19) <[XiXj]xk ~ [Natural coordination of Xi and Xj
superordinate level concept X1ch >
---+
One difficulty is to determine whether the category labels for the parts should all be of
the same category or of three different categories <X, Y, Z >. In a majority of cases the
word class is the same, thus (19) is at least appropriate for prototypical co-compounds.
However, it seems nearly impossible to reduce co-compounds and sub-compounds (cf.
example (4) in article 12) to one constructional schema which is well-in-line w ith Walchli 's (2005, ch. 4) claim that co-compounds are a construction type of their own and not
a sub-type of compounds.
A highly problematic aspect of construction morphology, however, is that word-formation processes are viewed as parts of a hierarchical lexicon. This presupposes that
word-formation is always a relation between parts and wholes which are types rather
than tokens. Above I have argued that many co-compounds are not listed i111 the lexicon
and that the parts of productive co-compounds are rather tokens than types, which means
that many co-compounds are not fonned in the lexicon. In order to become a useful
framework for the description of co-compounds therefore, construction morphology must
countenance the composition of tokens in particular given contexts.
Booij (in article 12, section 2) points out that in sub-compounds parts "may have
lexicalized, yet productive meanings". This phenomenon is much more restricted with
co-compounds. There are some schemas with productive parts. However, it is not clear
whether they should be considered true co-compounds. In Erzya Mordvin there are nominal compounds of the type N-t=mezt' 'N -PL=what:PL > N and the like' and verbal
compounds of the type V -ms=t'eje-ms 'V-INF=do-INF > V for a short while ' . The former
have a function similar to echo words in South Asia and m-doublets in Turkic and
contact languages (see Walchli 2005: 167-169 for a survey and further literature). Mordvin compounds with mez'e 'what' share a number of features with co-compounds. Generalizing examples and examples in comitative function have the suffix -n 'ek on both parts
which is characteristic for generalizing co-compounds in Mordvin: poks kudo-st sadnek=mez-n 'ek 'big house-Poss3PL garden-coM=what-coM > their big house with garden
and everything' (Kudaskin 1996: 22). Like co-compounds, N -t=mezt' 'N-PL=what:PL >
N and the like' typically has a nominative indefinite plural marker on both parts if the
form would be in the unmarked nominative indefin ite singular otherwise. However, what
seems to be peculiar to this type is that there can be wide scope semantically over a
noun phrase with an adjective as in (20):
PA727
41. Multi -word units in French
727
Propp, Vladimir Jakovlevic and Boris N ikolaevic Putilov

1958
Byliny. Vol. 2. Moskva: Gosudarstvennoe izdatel'stvo xudofostvennoj literatu ry.
Rushd ie, Salman
198 1 Midnight's Children. London: Vintage.
Saronov, Aleksandr Markovic
1994
Mastorava. Saransk : Mordovijan' kn'igan' izdat'el'stvas'.
Sketan, M.
199 1
Cumyren lukmo ojporo. Kokymfo tom. Ojlymai, myskara, novella, ocerk, statJa, korrespondencij, fel'eton -vlak. Joskar-Ola: Marij kn'iga izdate l'stvo.
Tirnusev, Dmitrij A . (ed.)
Obrazcy komi-zyrjanskoj reci. Syktyvkar: Akademija Nauk SSSR.
197 1
Tkacenko, Orest B.
1979
Sopostavitel' no-istoriceskaja frazeologija slavjanskich i finno-ugrorskich jazykov. Kiev:
Naukova dumka.
Uotila, Eeva
1980
Asyndeton in the baltic and finn ic languages: An archaic construction in its typological
periphery. Journal of Baltic Studies 11(1): 86-9 l.
Walchli, Bernhard
2005
Co-Compounds and Natural Coordination . Oxford: Oxford University Press.
Walchli, Bernhard
2007a Lexical classes: A functional approach to "word-formation". In: Matti Miestamo and
Bernhard Walchli (eds.), New Challenges in Typology. Broadening the Horizons and
Redefining the Foundations, 153-175. Berlin: Mouton de Gruyter.
Walchli, Bernhard
2007 b Ko-Komposita (im Vergleich m it Parallelismus und Reduplikation). In: A ndreas Ammann and Aina Urdze (eds .), Wiederholung, Parallelismus, Reduplikation. Strategien
der multiplen Strukturanwendung, 81-107. Bochum : B rockmeyer.
Bernhard Walchli, Stockholm (Sweden)
41 . Multi-word units in French

l.
2.
3.
4.
5.
Introduction
Functions of multi-word units and domains of description
C lasses of multi-word units
Multi-word seque nces at the crossroads of linguistic preoccupations
References
Abstract
After pointing out the emergence of the concept of multi-word units in French linguistics,
we focus on the principal functions assigned by language to multi-word sequences (referentiation, predication, "argumentalisation ", actualisation, etc.) and on the domains of
description. Then we provide a sample of the descriptions devoted to these units classified by parts of speech.
PA730
730

to a series of common names which exist alongside the learned ones: tete-de-chat head
of cat 'pieces of rocks', <ii-de-chat eye of cat 'variety of quartz', etc.
What has interested linguists most in referentiation is primarily the link established
between the new terms and the domains of the lexicon from which material is drawn for
the creation of a multi-word unit, as all quoted examples illustrate that use parts of the
body and animals. Behind this motivation one can raise questions about the mechanisms
of abstraction, conceptualisation and categorisation (Mejri 2005 and Petit 2008). These
questions find a great deal of material in multi-word units to exploit in order to better
grasp not only the operations by which complex designations are constructed but also
the patterns (nominal, verbal, etc). Take for instance the French verbal sequence if a un
poi! dans la main lit. 'he has a hair in his hand' used for naming the character of
someone who is very lazy. Such a designation reveals the stereotypes shared by the
members of the linguistic community and anchored in the lexicon.
In this way linguists have gradually allowed multi-word units to fu lfil the three primary functions that single-word units do, namely, playing the role of predicates (action of
pred icating, predication), of predicate arguments ("argumentalisation", relative to argument) and actualisers (all the elements necessary to ensure the grammatical character of
an utterance: time, aspect, gender, number, person ... ). A sentence as common as La
plupart des directeurs adjoints ant fait l'objet d'une contre-enquete avant-hier 'Most
deputy directors were the object of a counter-enquiry the day before yesterday' nicely
illustrates both the omnipresence of multi -word sequences and the coverage they ensure
at the level of the three primary functions. It can be analysed as follows:
(2)
a.
nominal predicate
contre-enquete
'counter-enquiry',
b.
argument
directeurs adjoints
'deputy directors',
c.
actualisers
a.. of the predicate: ant fait l' objet de 'were the object of' (light verb expressing both perfect and passive),
avant-hier 'day before yesterday' (adverb expressing the
tense: past),
~ of the argument: la plupart de 'most of' (complex degree modifier/proportional modifier expressing the category of number and of
indefiniteness).
As we can see, all components of this sentence, except for the indefinite une 'a' which
actualizes contre-enquete, are multi-word un its. One can continue the exercise and add
for example an agent: La plupart des directeurs adjoints ant fait l'objet d'une contreenquete de la part de la cour des comptes ' Most deputy directors were the object of a
counter-enquiry on behalf of the Court of Auditors'. This sentence contains cour des
comptes 'Court of Auditors' as an additional argument and de la part de 'on behalf of',
an agentive argument specific to the passive construction.
While the functions of referentiation and predication, with what they entail, such as
"argumentalisation" and actualisation, are already part of the landscape of research on
PA732
732

b.
Pierre a de la perseverance (est perseverant). Grace a quoi il a bien reussi.

'Pierre has perseverance (is persevering). Thanks to which he succeeded well.'
At the level of intersentential relations multi-word units also occupy a place of choice
concerning the structuring of discourse; this includes all the connectors of the type
d' abord 'initially', to which one can add all multi-word sequences playing the role of
anaphora, including proverbial utterances.
To finish with the functions assigned by language to multi-word units, we must mention an area which has been studied very little, but which has begun to interest linguists,
that of stylistic marking, which can find its expression inter alia in the recognition of
specialised discourses and in the stylistic characteristics of authors. It has been demonstrated, thanks to the frequency of complex denominations in specialised discourses, that
the frequency of multi-word units is so great that it can be used as a reliable indication
for automatic domain text classification. Other studies have managed to demonstrate
how certain authors use multi-word units as stylistic markers. Balzac for example commonly resorts to the de-fi xing or diversion of proverbs (Navarro Dominguez 2000). It is
a very promising field in which linguists are investing more and more effort.
2.2. Domains of description

This stylistic dimension is for the moment rather marginal compared to the linguistic
domains that make multi-words a privileged object of study, namely syntax and semantics. Multi-word units have primarily been studied from a syntactic point of view. Faced
with the rather widespread idea that multi-word units are supposed to be completely
fixed, linguists were initially interested in verifying this working hypothesis. The abundance of works in keeping with this position have made it possible to posit the following
characteristics:
(4)
a.
with the exception of certain productive syntactic patterns especially for noun
compounds and archaic constructions (of the type advienne que pourra 'come
what may'), all multi-word units are grammatically well formed;
b.
multi-word sequences that manifest absolute fixedness, i.e. do not accept any
variation of whatever nature, are by no means common: although precise numbers are not available, their volume is estimated at around 10 %;
c.
the majority of fixed sequences show variations which place them in a continuum that reaches from the most constrained to the least constrained;
d.
these variations can correspond to the transformations allowed by the internal

combinatory of units (cf. section 3) or to the existence of variants of the same
sequence as is the case of emporter (en/ever) le morceau to carry away the
piece 'to carry the day'.
The challenge is obviously the measurement of the degree of fixedness of multi-word

sequences according to the number of variations (transformations) allowed by each se-
PA733
41. Multi-word units in French
quence. This can only be done by relying on the syntax that is specific to the internal
structure of the sequence (cf section 3).
On the semantic level, the following topics need to be investigated:
(5)
a.
the compositionality/non-compositionality of the meaning of polylexical units:

multi-word units are, in general, ambiguous between a compositional and noncompositional meaning. There is a tendency, in particular for the non-native
speaker, to understand the expression in its compositional (literal) meaning.
This has prompted linguists to contrast the compositional meaning of multiword expressions with their non-compositional meaning. In the first case, although there is a global meaning, one is able to make sense of the components
of the sequence. Consider the following examples: Examiner quelque chose a
la loupe 'to examine something with the magnifying glass'. Although figurative, this expression does not pose particular problems of interpretation as long
as the implications of the image are obvious. Mettre quelque chose en surete
'to put something in safety' is even more transparent. Consider now, however,
en boucher un coin lit. 'to plug a corner': nothing in this sequence suggests
the meaning 'fill with wonder';
b.
semantic transparency/opacity: A corollary of the first opposition, this dichotomy focuses on the decoding of sequences. Opacity, like non-compositionality,
is of scalar nature: the more opacifying mechanisms are used in the sequence,
the greater the opacity is. In the sequence, (faire, jouer) la mouche du cache
(to do, to play) the fly of the coach 'backseat driver', two mechanisms come
into play, figuration (metaphor) and exocentricity (nothing in that sequence
indicates that it refers to a person). This is the reason why the meaning of this
expression needs to be memorized as it is, "a person who is overactive while
claiming to give invaluable help" (Rey and Chantreau 1989). On the other
hand, one can find examples of transparent sequences like Beaucoup de bruit
pour rien A lot of noise for nothing 'Much ado about nothing';
c.
the role of context in "removing" opacity (Gross and Massoussi 2011 ): Since
the meaning of a unit can only be determined within the sentence frame, taking
context into account can de-opacify such sequences. Speaking about tete de
Maure (Moorish head) out of a sentence context does not help to access the
meaning of the sequence. But to say on nous a servi entre le plat principal et
le dessert une tete de Maure ('we were served a Moorish head between the
main dish and the desert') makes it possible to infer that tete de Maure is
something edible, and that it probably refers to a variety of cheese, considering
the order in which it is served during a meal;
d.
space precludes the explicit demonstration of a number of other relevant semantic aspects, as for instance stereotypy, semantic unfolding, referentiation,
semantic globalisation, conceptualisation, etc.
733
PA737
b.
On etait servi a la carte (adv.)

' We were served a la carte'
3.4. Adverbial sequences

We refer to previous studies of Guimier (1996) for the characteri stics of adverbs especially those ending in -ment -, linked diachronically to multi-words, and those of
M. Gross (1986) and G. Gross (1996) for multi-word adverbs. We have already mentioned that the most productive structures are those of the pattern Prep (Det) N and that
adverbs can play a role both in the structuring of sentences and in intersentential sequences. As we have shown previously, their construction can bring them closer to adjectival
usages. But they can also be grouped together with prepositional or conjunctive phrases
as connectors, what we call connecting adverbs:
( 11)
a.
sentential usage
Il parle a tort et a travers
' He talks a lot of nonsense'
b.
intersentential usage
[ ... ] ses doigts etaient tourmentes d'un tremblement febr ile; en outre, il transpirait abondamment des paumes
[ ... ] ' his fingers were tormented of a nervous trembling; moreover~ he sweated
abundantly from his palms' .
(Duhamel, cited by the Grand Robert 2001)
Certain multi-word adverbs can appear in the form of discontinuous units, thus making
it possible to structure entire passages:
(1 2) a.
d'un cote ... de l'autre (cote)

' on the one hand' ... 'on the other (hand)'
b.
d' abord ... ensuite ...en.fin

'initially' ... 'then ' ... 'finally'
c.
certes ... mais

'certainly' ... 'but', etc.
(cf. inter alia Siepmann 2008)
3.5. Prepositional and conju nctive sequences

Relegated to the rank of simple markers of syntactic functions, in particularly circumstantial complements, these sequences have, thanks to the works of Gross and Prandi
(2004), seen their status change completely. These authors were able to demonstrate that
a prepositional sequence as cause de 'because of' is on the contrary the principal
element in a sentence like:
737
PA738
738

(13) Paul n 'a pas pu monter son projet a cause de ses difficultes avec Jes concurrents.
'Paul could not put together his project because o/his difficulties with the competitors.'
The predicative analysis shows that there are three predicates in this sentence:
(14) Paul n 'a pu monter son projet (B)
' Paul could not put together his project',
Paul a des difficultes avec les concurrents (A)
'Paul has difficulties with the competitors',
Ceci (A) a cause cela (B)
'This caused that',
and that the first two predicates are actually the arguments of the prepositional phrase a
cause de. This is why it is considered as a predicate of second order.
Two points are to be kept in mind on this subject: some linguists do not consider
such sequences as fixed because they admit several variations (dans le but de in the
goal/target of 'with the aim to', dans ce but 'to this end', dans le seul but 'in the only
goal', etc.), but they omit that fixedness here is of a functional nature, i.e. the grammatical function assigned to it by language. The rest is a question of degree of fixedness: if
we compare cause de 'because of' and grace 'thanks to', we notice that the second
is more fixed than the first. The second point, related to the first, relates to grammaticalisation: all the works agree on the fact that fixedness is the only multi-word process that
equips languages with grammatical tools.
3.6. Determinative sequences

Buvet (2012) describes predeterminers in French, specifying that their number reaches
4,000 of which 3,000 have a nominal core (for example, un tas de livres 'a lot (lit. heap)
of books'), and 2,000 determinative fixed sequences that imply a predeterminer and a
modifier (for example, une memo ire d' elephant 'a memory of an elephant'). As actualisers of the noun, he studies them according to whether they determine predicates (Il a
pris une kyrielle de decisions ' He made a string of decisions') or arguments (11 a lu un
tas de livres 'He read a lot a/books').
3.7. Multi-word interjections

Whi le the subject of interjections has not yet been add ressed by many researchers, analysing them as predicates could indicate some possible avenues for future research: as
"phrasoi'des" (synthetic phrases), they could felicitously accept a predicative analysis.
To be convinced it is enough to compare the long form to the truncated one: Va au
diable ! 'Go to hell!', Au diable ! 'To hell!'. From the very strong links that they have
PA739
to the situation of utterance, they generally function as pragmatemes whose significance
is fi rm ly anchored in verbal exchange.
3.8. Other types of unit

Until now no special reference has been made to multi-word pronouns, which appear in
language in small numbers since they belong to what is traditionally called "closed
paradigms''. It should be kept in mind however that besides multi-word pronouns as
demonstratives (ceux-cilla, celles-cila ... 'this/that one') or relatives (a laquelle 'to
which '), grammars classify a whole series of multi-word pronouns as inde finites: l'un
'the one', quelqu'un 'somebody', etc. (cf. Riegel, Pellat and Rioul 2009 [1994]).
Descriptive multi-word sequences could also be included in the list of multi-word
sequences. Oddly enough, they can have all kinds of functions (titles of paint ings, names
of bars, restaurants, etc.) as well as a w ide variety of structures (cf. Bosredon 1998).
Finally hybrid sequences which do not belong to a part of speech nor are sentences,
as for instance La balle est dans le camp de X ' the ball is in X's court', could also be
mentioned here.
4. M ult i-word seq uences at t he crossroads of linguistic

preoccupations
The centrality of fixedness and its massive presence both in language and in discourse
raises a number of questions relati ng to:
(1 5) a.
the theoretical dimension regarding the status of multi-word units: Are they
words? If not, what status should be granted to them?
b.
the automatic recognition imposed by the development of corpus research:

How are multi-word units to be dissociated from the corresponding syntactic
units w ith the same signifier?
c.
the construction of a lemmatiser that takes into account all the degrees of
fixedness and the multiple variations of the sequences imposed consequently
in discourse;
d.
the lexicographic description of multi-word units imposed by the great vacuum

left by traditional lexicographic practices: Is it necessary to compile dictionaries specific to multi-word units? If so, what needs to be taken into account?
Syntax? Semantics? Variations in discourse? Stereotypes? Inferences?, etc.
e.
the combinatory profile (Blumenthal 2008) of the single-w ord units as the
basis of the most recurring collocations: Should we see this phenomenon, if
it is systematic, as the beginnings of fixedness which leads, in the long run,
to the constitution of new multi-word units?
739
PA742
742
42. Multi-word expressions and univerbation

in Slavic
l.
2.
3.
4.
5.
Introduction
General overview
Univerbation in contemporary Czech, Slovak and Polish
Multiverbation
References
Abstract
Univerbation can be generally defined as the formation of a single-word unit based on a
semantically equivalent multi-word denomination. Jn Slavic linguistics, however, the term
has undergone certain specifications: in contemporary studies, it most often comprises the
transition ofmulti-word denominations into one-word units (formally accompanied by ellipsis ofthe head and conversion or suffixation ofthe modifier). The activity ofsimilar processes can be partially explained by the lower productivity of composition in Slavic languages. Differences in the interpretation ofuniverbation can be explained by the individual
conceptions ofword-formation as word-based or based on transformations ofcomplex expressions. The introduction ofthis article deals with terminological issues. The second section briefly outlines the research background in individual Slavic languages, the late 195 Os
and early 1960s being selected as starting point in Czech linguistics. Section 3 focuses on
univerbation in contemporary Czech, Slovak and Polish and its linguistic description. In
section 4, a correlative process, i.e. multiverbation, is mentioned.
1. Introduction
One of the most common definitions of ''univerbation" in the Slavic languages, describing the concept of the phenomenon from a genetic and procedural point of view, is given
by the mult ilingual terminological Slovnik slovanske lingvisticke terminologie. Vol. 1
[Dictionary of Slavonic Linguistic Terminology. Vol. 1] (Jedlicka 1977:: 184): "Univerbation - formation of a single-word denomination based on a synonymic multi-word denomination, e.g., Cz. plenarni schuze ['plenary meeting'] ~ plenarka [ 'id.']" (translated
from Czech). What we can observe here is the replacement of a multi-word expression
by a shorter synthetic form. This is the way in which the term is used in most articles on
Slavic languages in this handbook, in contrast to other articles where the term refers to
a procedure and the result of word-formation consisting in the contraction of the words
in a fixed word combination into a single word (cf. E. anytime, whenever, everyday).
However, in the literature on Slavic word-formation, univerbation is not only based
on derivation as the above-mentioned Czech example might imply. There are also other
types of denominations regarded as univerbation where the head of the semantically
identical multi-word expression is formally deleted, but its meaning is preserved in the
univerbate (cf. Russ. vychodnoj 'day off' ~ vychodnoj den' ' id.; lit. go out-ADJ day',
PA743
42. Multi-word expressions and univerbation in Slavic
Cz. trvala 'perm' +-- trvala ondulace ' permanent wave; I it. permanent undulation '), or
synecdoche (e.g., Slk. ird' 'pole vault; lit. pole' +-- skok o irdi 'pole vau lt; lit. jump with
pole'), blends, etc. As will be shown in this article, a narrower or broader definition of
univerbation can also be connected with individual traditions of the description of wordformation as word-based or based on syntagms.
The above-mentioned examples are formed from nominal word combinations. The
univerbate is identical with the modifier and represents the same meaning as the complex
denomination (cf. Russ. vychodnoj). In other cases, such as Cz. plenarka, the modifier
becomes the basis of a derivative which is semantically identical with the complex base.
Interestingly, the formal deletion of the head which is specific to numerous results of
univerbation in Slavic, has only in some cases served as criterion for the definition of
univerbates. Since Isacenko 's works of the 1950s which also include special types of
compounds (see section 2.1), terms such as "semantic condensation'', "compression" (of
multi-word units) have been applied as hyperonyms of the different univerbation processes, only some of which can be regarded as strncturally based on the ellipsis of one
of the components. Keeping to this tradition, the term "ellipsis" will be used in this
article in accordance with the terminology of the respective authors (e.g., Isacenko
l 958b; Kucerova l 973a, 1974; Buttler 1981 ). With this term, the authors refer to the
ellipsis of the head and the nominalization of the modifier of the corresponding multiword expression, sometimes also to the ellipsis of the modifier of multi -word units. In
Slavic studies on word-formation, the term "ellipsis" is usually not applied to the univerbation of multi-word expressions by means of suffixation with the modifier becoming
the derivational base. For this reason, we use the term "ellipsis" in a broader sense
exclusively with the objective of an additional explanation (as laid out in the abstract of
this article) and a cross-reference (section 3.3).
Terminological issues also concern the naming of the process and the results of uni verbation. In Czech, Slovak, Polish, Bulgarian, and Ukrainian linguistics, the term univerbization (e.g., Cz. univerbizace, Pol. uniwerbizacja) is used instead of univerbation.
The term (in either variant) is ambiguous. In Czech linguistics, for example, it is mainly
used to designate a) a process of denomination consisting in the transition of a multiword denomination into a single-word unit, b) a way to form words (i.e. in relation to
procedures such as derivation, composition, etc., leading to univerbation). The descriptive nature of the term covers the phenomenon from a genetic and procedural point of
view, but much less from a strnctural viewpoint that focuses on the result of univerbation,
etc. (see section 2.1 and 3.1).
For univerbized denominations the term univerbate (Cz. univerbat, Pol. uniwerbat)
is used (recently preferred in West Slavic languages), the term univerb, but also univerbat, can be found in Bulgarian and Russian word-formation. In Russian it refers either
to suffixal transformations of attributive constructions of the A+N type (see Nescimenko
2003: 293), or it is also used in a more general meaning.
2.
General overview
2.1. Czech studies

The origins of research on univerbation in Czech linguistics go back to the Russian
linguist A. V. Isacenko who lived in former Czechoslovakia. Isacenko 's papers (1956,
743
PA744
744

1958a, 1958b) had a sign ificant influence on the works of Dokulil (1962), He Icl (1963),
and Jedlicka (1965, 1969, 1973) with respect to theoretical problems ofuniverbation, its
manifestations, formation processes and tendencies. Because of their importance for
Czech linguistics and their reception and influence on research of univerbation in other
Slavic languages we will discuss these works in more detail.
Influenced by the word-formation theory of Jan Rozwadowski (1904), Isacenko starts
from the thesis of the contradiction between the formally binominal denominations and
the unity of their meaning. According to Isacenko, a solution to this contradiction can
be found in the loss of formal and semantic segmentation, which is one of the rules of
lexical development. lsacenko considers different processes related to the loss of formal
and semantic segmentation as "semantic condensation" and supposes that condensation
processes can best be researched on so-called complex (i.e. multi-word) denominations.
"Complex denominations, consisting of several lexical units have a clear tendency to
univerbation, i.e. to the compression of the semantic content into one word." (Isacenko
1958b: 340; translated from Russian). Isacenko distinguishes the following processes of
"semantic condensation" (which manifest themselves in various types of denomination):
1. Composition: Russ. mesto zitel'stva 'place of domicile; lit. place domicile-GEN' ~
mestozitel'stvo 'id.; lit. place-domicile-NOM', Slk. svetovy nazor 'world view; lit.
world-REL.ADJ view' ~ svetonazor 'id.';
2. Juxtaposition (merger): Russ. uma lisennyj lit. 'mind-GEN deprived of'~ umalisennyj
'mad, insane', Cz. pravde podobny lit. ' truth-DAT similar' ~ pravdepodobny 'probable'. (Juxtaposition (merger) is the only procedure regarded as univerbation in English, German and French linguistics.);
3. Ellipsis:
a) ellipsis of the head, i.e. nominalization of an adjective: e.g., Russ. mostovaja doroga ' paved way' ~ mostovaja 'pavement', Cz. tanecnf hodiny lit. ' dance-REL.ADJ
hours' ~ tanecnf 'dancing classes';
b) ellipsis of the modifier: Russ. grammofonnaja plastinka lit. 'grammophoneREL.ADJ disc' ~ plastinka 'record', Cz. zelezna draha 'railway; lit. iron-REL.ADJ
way' ~ draha 'id.';
4. Affixal derivation: Russ. setcataja obolocka 'retina; lit. net-ADJ membrane ' ~ setcatka 'id.', Cz. ielezna draha 'railway; lit. iron-REL.ADJ way' ~ ielezn-ice 'id.';
5. Formation of compounds (with an abbreviated component) and abbreviations: Russ.
medicinskaja sestra 'nurse; lit. medicine-REL.ADJ sister' ~ medsestra 'id.'; Moskovskij_gosudarstvennyj 'l!:_niversitet 'Moscow State university; lit. Moscow-REL.ADJ stateREL.ADJ university' ~ MGU 'id.'.
Dokulil (1962; the following quotations are translated from Czech) explicitly addresses
the de 1i mi tat ion of the term univerbation. From his point of view, univerbation is a
"transformation of already existing multi-word denominations into single-word denominations" (p. 21 ); from the perspective of designation it is "a naming procedure altering
an existing denomination" instead of forming a new one (p. 27). Dokulil also adapts
Isacenko's original conception of univerbation from the functional and structural points
of view. According to Dokulil, the distinctive feature of a univerbized denomination is
the coexistence of a synonymous multi-word denomination, which should - a further
criterion emphasized by Dokulil - be a "real'', i.e. a fixed, established two-word denomi -
PA751
The typology of univerbation procedures proposed by Hladka (2002) seems to be similar.
Bosak (1987: 23 2) provides the following typology for Slovak. It includes:
l. Formation of single-word denominations on the basis of motivational multi-word
denominations by:
a) derivation (Slk. presilovka ' power play' ~ presilova hra 'id.; lit. power-REL.ADJ.
play');
b) nominalization (Slk. triedna 'class teacher f.; lit. class-REL.ADJ-FEM' ~ triedna
uCitel'ka 'id .; lit. class-REL.ADJ-FEM teacher-FEM') [the modifier is nominalized
after the ellipsis of the head];
c) synecdoche (S lk. ird' 'pole vault; lit. pole' ~ skok o irdi 'id.');
d) composition (Slk. prvoligista 'first-league player; lit. first-league-PERS' ~ hrac
prvej ligy 'id.; lit. player first-GEN league-GEN' );
2. Simplification of multi-word denominations, e.g.:
a) contextual (Slk. vlada 'the government [of the CSSR]' ~ vlada CSSR ' id.';
b) by abbreviation, acronymization (Slk. SAV 'Slovak Academy of Sciences' ~ Slovenska akademia ved ' id. ' ). Cf. also Bosak (1987: 235 f.; translated from Slovak):
"Sometimes all procedures of abbreviation ( .. . ] are also ranked as univerbation.
However, in such cases it is not univerbation in the genuine sense of the term but
a mechanical process which is above all dependent on the context (the situation
respectively). For this reason, we consider this process as simplification."
3.4. Univerbation by derivation and composition

The most widespread and frequent procedure of univerbation in Czech, Slovak and Polish is univerbation by derivation (Jedlicka 1965; Satkiewicz 1969; Bosak 1987; Pancikova 1993; Szczepanska 1994; Jadacka 200 1). The univerbates are fonned from nominal
two -word combinations whose (adjectival or nominal) attribute becomes the basis of
derivation. In their formation a narrow range of semantically non-specific and frequent
suffixes are used: in Czech -ka (mineral-k-a ' mineral water' ~ mineralni voda 'id.; lit.
mineral-REL.ADJ water'), -ak (trest11-ak 'penalty kick' ~ trestny kop ' id.; lit. penaltyREL.ADJ kick'), sometimes also -afl-ar (oc-ar 'oculist' ~ ocni lekar 'id.; lit. eye-REL.ADJ
doctor '). (In slangs the means of univerbation are more diversified.) The situation is
similar in Slovak: most often the suffix -ka (presilov-k-a ' power play', see section 3.3)
is used, -arl-iar with masculine personal nouns (oc-iar 'oculist'), in slang expressions
a lso -ak/-iak (zim-ak 'ice-ri nk' ~ zimny stadi6n ' id.; lit. winter-REL.ADJ stadium ') (Bosak 1987). In Polish, the suffixes -(6v.)ka (podstaw6w-k-a 'elementary school' ~ szkola
podstawowa 'id.; lit. school basis/foundation -REL.ADJ'), -ak (poprawcz-ak ' penitentiary,
community home' ~ dom poprawczy 'id.; lit. house (to) better-/improve-REL.ADJ'),
-(ow)iec (odrzutowiec 'jet' ~ samolot odrzutowy ' id.; lit. aircraft jet-propelled ') are
typically used. Due to the broader understanding of univerbation in Polish linguistics,
the suffixes include also -ista (polemista 'author of polemic works; lit. polemic-PERS'
~ autor utwor6w polemicznych ' id .'), -(ow)iec, a lso for personal nouns (radiowiec 'employee of a broadcasting company; lit. radio-PERS' ~ pracownik radia 'id.; lit. employee
751
PA753
4. Multiverbation
In studies on univerbation it is often pointed out that univerbation requires the formation
of new multi-word units as a permanent source of future univerbation (see, e.g., Jedlicka
1969: 94). Within this context, attention is also paid to multiverbation. The problems of
multiverbation are a border phenomenon which is involved either in the area of vocabulary or in the area of syntax. Multiverbation becomes relevant in relation to research in
analytical constructions, including phrasemes, and their syntactic and stylistic specifics.
Hereafter, we will shortly introduce the conception elaborated in Czech linguistics by
Jedlicka (1969), which has been continued by Slovak and Polish researchers (Ondrejovic
1989; Buttler 1978, 1981; Mietla 1998).
Multiverbation is defined as the formation of analytical multi-word denominations in
parallel to synonymous one-word denominations (e.g., Cz. Ctenafska obec 'the reading
public' - Ctenafi 'readers',provadet vyzkum 'to do research' - zkoumat 'to research').
Although multiverbation and univerbation are processes which are typical for the contemporary language, the former is not a counterpart of the latter: multiverbized units
coexist with synonymous one-word denominations, but they are not considered to be
formed out of single-word denominations. They also differ from univerbized units with
respect to use. Very often they are limited to the scope of literary language, the sphere
of public communication and expert discourse, etc.
Multiverbation units can also be described under formal, semantic and stylistic aspects. Based on their formal aspects, they encompass several types:
a) Verb + dependent noun (Cz. provest (pf.) I provadet (ipf.) vyzkum 'to do research' prozkoumat (pf.) I zkoumat (ipf.) 'to research');
b) Noun + dependent adjective (Cz. podnikatelska obec ' business community; lit. entrepreneur-REL.ADJ community' - podnikatele 'business people, entrepreneur-PLUR' );
c) Noun + dependent adjective, motivationally coexisting with an adverb (Cz. zajimavym zpusobem ' in an interesting way; lit. interesting- rNSTR way-rNSTR' - zajimave
' interestingly');
d) Noun + dependent noun (Cz. otazka vychovy 'a matter of education; lit. question
education-GEN' - vychova 'education' ).
There are differences in the manifestation of multiverbation in the individual Slavic
languages based on the structure of the respective denomination models and according
to their frequency and distribution.
5. References
Avramova, Cvetanka
Slovoobrazuvatelni tendencii pri sastestvitelnite imena v balgarskija i ceskija ezik v
2003
kraja na XX vek. Sofija: Heron Press.
Avramova, Cvetanka
2007
Univerbaty v lexikatnim systemu jazyka. Bohemistyka 7(3): 190-206.
Avramova, Cvetanka
2008
Univerbatite v leksika lnata sistema na ezika. Sapostavitelno ezikoznanie 33( 1): 99- l J5.
753
PA757
43. Compounds and multi -word expressions in Slavic
Zemskaja, Elena A .
Russkaja razgovornaja rec'. Moskva: Nauka.
1973
Zemskaja, Elena A .
1997
Aktivnye tendencii slovoproizvodstva. In: Evgenij N. Sirjaev (ed.), Russkij jazyk. Najnowsze dzieje }?zyk6w slowianskich, 169-20 l. Opole: Uniwersytet Opolski - Instytut
Filologii Polskiej.
Zemskaja, Elena A .
2000
Aktivnye processy sovremennogo slovoproizvodstva. In: Elena A. Zemskaja (ed.), Russkij jazyk konca XX stoletija (1985-1995), 90- 141. 2nd ed. Moskva: Jazyki russkoj
kul'tury.
Zemskaja, Elena A ., Margarita V. Kitajgorodskaja and Evgenij N. Sirjaev
198 1 Russkaja razgovornaja ree'. ObSCie voprosy. Slovoobrazovanie. Sintaksis. Moskva: Nauka.
Olga Martincow:i, Prague (Czech Republic)
43. Compounds and multi-word expressions 1n Slavic

1.
2.
3.
4.
5.
6.
7.
Introduction
Relations between modifier and head in the traditional Slavic compounds
Relations between modifier and head in combinations of relational adjectives and nouns
Types of compounds without a linking vowel and their adjectival competitors
Derivatives from multi-word expressions (RA+N) and denominal derivatives
Conclusion
References
Abstract
Slavic languages, as compared to English or German, show significant limitations with
respect to the formation of nominal compounds. These limitations are often compensated
for by complex designations consisting of a relational adjective and noun or by other
types of multi-word expressions. To a significant extent the increase of new vocabulary
in modem Slavic languages feeds on borrowings and loan translations or hybrid words,
a large part of which consists of different types of compounds and compound elements
offoreign origin. Most of the new compounds continue to exhibit parallel designations
in terms of multi-word expressions consisting of a relational adjective+noun or
noun+noungenlprep.case Their occurrence is often more frequent than that of compounds
and thus represents the continuation of typical preferences in designation.
1. Introduction
In a study on the structural typology of the vocabulary of various Slavic languages, Isacenko (1958) addressed the interaction of different methods of designation in modern Slavic
757
PA759
759
For Slavic languages there is no investigation in which compounds and designations

consisting of a relational adjective and noun have been jointly considered as "complex
nominals" on the basis of a corresponding inventory of relations between modifier and
head, as was unde1taken for English by Levi (1978). This may be due to the fact that
the formation types of nominal determinative compounds are limited in terms of possible
relations between head and modifier.
The focus of the relationship between composition and multi-word designation - (i)
N+N including a nominal modifier in the genitive case or (ii) a prepositional case or
else (iii) RA+N - is therefore rather reserved for individual contrastive studies (cf. Gladrow 1998 on Russian and German, Szymanek 2009 on the comparison of Polish and
English, Rainer 2013 also including evidence from Slavic languages). When considering
the examples provided by Szymanek (2009: 466) alongside Russian and Czech equivalents we can see that indeed they do not have compounds - just as their Polish equivalents. There is, however, also no consistent structural correspondence between the equivalen ts in the closely related Polish and Czech languages, and the frequency of use of the
examples can vary strongly when it comes to formal correspondences (for Polish and
Czech determined via Google, for Russian via Yandex; the same holds for the following
Tables), cf. Table 43.2:
Tab. 43.2: English compounds and their equivalents in Polish, Czech and Russian (adapted after
Szymanek 2009: 466)
E. telephone number
a)
N +N gen
P. i. numer telefon-u
Cz. [cislo telefon-u]
R. namer telefon-a
(33 m.)
N +N prep.case
P. ii. *numer do telefon-u
Cz. -
N+RA/RA+N
P. iii. *numer telefon-iczn-y
Cz. telefon-n-i cislo
R. telefon-n-yj namer
(3 m.)
E. computer paper
b)
N+Ngen
P. i. *papier komputer-a
N+Nprep.case
P. ii. papier do komputer-a

(5 m.)
Cz. [papir pro poCf-
R. bumaga dlja komp-
tacl
jutera
( less frequent)
N +RA/RA+N
ow-y
Cz. pocitac-ov-y
papir
( 124,000)
( 11 ,000)
P. iii. papier komputer-
R. komp'juter-n-aja
bumaga
E. tooth paste
c)
N +N gen
P. i. *pasta Zfb-6w
N +N prep.case
P. ii. pasta do Zfb6w
[Cz. pasta na zuby ]
R. pasta dlja zubov

(20,000)
N+RA/RA+N
P. iii. *pasta Zfb-ow-a
Cz. zub-n-i pasta
R. zub-n-aja pasta
(3 m.)
PA760
760

Szymanek (2009: 467) simply states: "[ ... ] certain functions that are served by compounding in other languages tend to be realized by syntactic, inflectional and/or derivational means in Polish". At the same time he points out that the designations which he
classified as "fixed nominal phrases" like, for instance, drukarka laserowa (N+RA) ' laser
printer', are , according to other Polish authors (e.g., Jadacka 2006), considered as juxtapositions (P. zestawienia), which is "a special type of a generally conceived category of
compounding".
Are there certain regularities regarding the choice of the designation method depending on the relationships that may exist between head and modifier?
On the basis of the examples presented in Table 43.2, we can first understand Isacenko 's differentiation between Polish as "genitival type" and "prepositional type" and
Czech as "adjectival type" (Isacenko 1958: 340). Russian, however, does not clearly
appear as "genitival type" in the above examples.
As will be demonstrated later on, also in Russian or Czech, general (nonspecific)
relations or purposive relations, as in b. and c., are likewise not expressed by compounds.
But what then about the Polish and Czech equivalents corresponding to the Russian
examples, provided in Table 43.1 , which additionally exhibit " object relations" between
the constituents? Equally, the Polish equivalent here is never a compound: designations
of the type N+N gen probably represent the only equivalents (e.g., budowa statk6w 'ship
building' ), and in those cases where a purposive relation is expressed and the formation
of N+RA is possible, this is less frequently attested, e.g., mechanik statk6w (11 ,000) vs.
mechanik statkowy (3, 150) 'ship mechanic'. Similarly there are no compounds in Czech,
but RA+N-equivalents (lodni mechanik 'ship mechanic', lodni motor 'ship engine') and
besides N+Ngen.pI (stavba lodi 'ship building') a synonymous abstract noun: lod'af-stvi
( +--- lod'af 'shipbuilder' +--- lad' 'ship').
This doe s not, however, imply that there are no compounds at all: the documented
types nevertheless need to be analysed with respect to the relationships realized between
the constituents. And in light of Szymanek 's statement: "However, there is a wealth of
evidence to demonstrate that nominal compounding is, relatively speaking, a living process in Polish, too", it remains to be examined which tendencies emerge in the formation
of neologisms compared to the established compounds. This question should also be
considered in regard to Czech. As a basis for the fo llowing discussion we will focus our
attention on Russian which, in adapting more recent borrowings, appears to be especially
open to compounds.
2. Relations betwee n modifier and head

compounds
in
the traditional Slavic
In the following we will confine ourselves to a brief outline of the formation of determinative noun-noun compounds (with and w ithout the linking vowel ale, particularly in
recent calques), because only here can we establish parallels to word combinations consisting of RA+N (possibly also N+Ngen or N+Nprep.case)
Referring to English nominal compounds, Plag (2009: 148) presents different possibilities of classifying semantic relations between constituents, as, e.g., ' location', 'cause',
' possessor', ' material', ' instrument', whilst considering such attempts as "somewhat fu-
PA761
tile". One should rather ask the question as to which interpretations are possible, since
N+N-compounds are basically ambiguous and can take on different meanings depending
on the context (cf. contributions to the topic in Lieber and Stekauer 2009). This evaluation of English might also be true of German, yet cannot be applied to compounding in
Slavic languages. Plag's observation concerning the description of isolated compounds
in English (i.e. without context) - which are usually interpreted in terms of a typical
relationship between the constituents - is, however, more likely to apply to Slavic compounds. Slavic grammars exclusively proceed from isolated compounds, provided this
question is at all subject to discussion. This relates to the restrictions on the formation
of compounds as well as to the limited, if not impossible, textually dependent semantic
interpretations. Consequently, none of the German compounds with a nominal modifier
presented in article 59 on schemata and semantic roles in word-formation in Table 59. l
"Types of noun-noun and verb-noun compounds in German" can have a compound
equivalent in Russian, Polish or Czech.
In Slavic languages the formation of compounds including a verbal modifier is impossible.
2.1. N+N-compou n ds
In its description of determinative N+N-compounds (e.g., R. zvuk-a-reiisser 'sound operator', most formations including a deverbal head noun, e.g., R. dac-e-vladelec ' dacha
owner', tepl-a-atdaca 'heat emission'), the "Russian Grammar" (Russkaja grammatika
1980: 242 ff.) completely ignores the semantic relations between the constituents.
In "pure compounds" of the N+N-type, the " Czech Grammar" (Mluvnice cestiny
1986: 456 ff.) distinguishes between:
a) compounds whose heads do not denote actions, e.g., Cz. vad-a-znak 'watermark',
dfev-a-prumysl 'wood(working) industry', including numerous technical terms as
well as older loan translations (-a- is the linking vowel). What is very revealing in
this context is that some compounds marked as archaic have been replaced by RA+N
(par-a-straj ' steam engine' > parni straj 'id.; lit. steam-RA machine ' ), by a prepositional phrase (kav-a-mlynek 'coffee mill' > mlynek na kavu 'id.; lit. mill-DIM for
coffee') or by a derivative (par-a-lad' 'steamboat' [> parni lod' lit. 'steam-RA boat']
> parnik 'steamer'), see also section 5;
b) compounds whose heads denote an action (mostly action nouns), including calques,
e.g., Cz. jazyk-a-veda 'linguistics; lit. language-a -science', tel-a-vychava 'physical
education; lit. body-a-education', vad-a-lecba 'water cure', beside vadni lecba lit.
'water-RA cure'.
Complex words comparable to Czech formations belonging to the group described under
b) are, according to the "Grammar of Modern Polish" (Gramatyka 1998: 457 ff.), among
the most frequently attested patterns of determinative compounds in Polish. The syntactic
relations between the constituents are taken as a basis for the description of compounds
representing the structure "N(Nl+N2)" (with the linking vowel -a-), e.g.: "NI denotes
the object of the action performed by N2" (P. mit-a-tw6rca ' myth-a -creator ', prajekt-adawca 'project developer; lit. project-a-giver')"; "Nl = medium, instrument of the action
761
PA763
43. Compounds and multi-word expressions in Slavic
performed by N2" (par-a-statek 'steam ship'), "Nl = location of N2" (lqd-a-lad '(continental) ice sheet; lit. (main)land ice'), etc. At the same time the dominant role of such
formations in terminology is pointed out.
The previous explanations indicate fundamental systemic parallels, which, however,
do not necessarily result in a structural correspondence between the equivalents. In
anticipation of section 3, we have already included designations consisting of relational
adjectives, which parallel the examples provided in the "Russian Grammar", as well as
the Polish and Czech equivalents, cf. Table 43.3 (above).
2.2. Synthetic compounds

"Synthetic compounds are complex words that contain at least three morphemes, with
neither the combination of the first two nor of the last two existing as free words" (see
article 33 on synthetic compounds in German). In Slavic languages formations belonging
to the group of synthetic compounds are traditionally more productive and can partly
compensate for the lower productivity of "pure" word-/root-compounds, but rarely express relations between the constituents that go beyond "pure" compounds (see the articles 155-168 on Slavic languages.)
The "Grammar of Modem Polish" (Gramatyka 1995: 463 f.) points out the following
relations between the constituents expressed by the structure "N(N l +V)": Nl denotes
the object, the instrument or the place of the action (V), by means of the suffix (including
the zero-suffix) categorial roles like 'agent' (also non-personal nouns), 'instrument',
'place', etc., are expressed, e.g., P. list-a-nasz-@ 'postman; lit. mail-a-carrier', wad-aaczyszcz-acz 'water purifier', kwas-a-ryt-{i} 'copperplate print; lit. acid-a -etch-0 ', dam-akrqi-c-a 'peddler,. door-to-door salesman; lit. house-a -back and forth go-er'.
Similar relationships can be found in Czech and Russian formations:
Cz. vin-a-bra-ni 'vintage; lit. grape-a -taking ', drev-a-rub-ec 'woodcutter', the calque
tepl-a-mer- ' thermometer; lit. warmth-a-measur-er' (however, cf. Kavka 2009: 19, who
proposes to view recursive constituents such as Cz. -mer ' -meter', etc. "as suffixes[ ... ]
rather than genuine compound elements"); R. ryb-a-lav-stv-a 'fishing; lit. fish-a-catchABSTR', kanat-a-chad-ec ' ropewalker; lit. rope-a -go-AGENT', sud-a-pad" em-nik 'ship elevator; lit. ship-a -elevate-INSTR', piv-a-var-n-ja 'brewery; lit. beer-a -brew-Loe '.
3.
Relations between modifier and head in combination s of

relational adjectives and nouns
3.1. General overview

As was demonstrated in section 2.1, the relations between constituents which in Russian,
Polish and Czech can be expressed by means of "pure" N+N-compounds are limited in
number.
a) If the heads are non-derived nouns, final relations between head and modifier dominate.
763
PA766
766

quence of that, avto- should not be treated as "analytical adjective" (see section 4.3).
But a RA+N-combination avtomobil'nyj prodavec (152) is also documented. Both occurrences are still less frequent thanprodavec avtomobilej lit. 'seller carsgen' (27,000). Even
today, the existence of R. *furgonnyj voditel' cannot be attested. There is, however, a
neologism R. avtomobil'nyj voditel' meaning 'driving simulator; lit. car-RA driver '. Also
cf. R. kniznaja prodaza (RA+N) ' book selling' (16,000) as opposed to prodaia knig
(N+Ngen) ' id. ' (195,000), and lmiinyj prodavec (750) marked with an asterisk by Mezhevich; the compound knigoprodavec is obsolete, while the genitive construction prodavec knig (14,000) again dominates. Yet occurrences of the type RA+N in place of a
compound or N +N gen can be found more often in Russian online forums. This could
quite possibly be indicative of a certain tendency within the Russian vernacular, such as
was already noted by Zemskaja (1992), to re-activate the use of possessive adjectives
(e.g., Tanina jubka ' Tanja's skirt' ) instead of genitive constructions (jubka Tani 'id.; lit.
skirt Tanjagen').
While - apart from studies of the parallels between word combinations consisting of
RA+N and compounds with clipped initial components mainly of foreign origin (see
section 4.2) - there are hardly any articles in the respective Slavic literature dealing
with the relation between compounds and RA+N-structures, the number of investigations
dedicated to the relation between RA+N and N+Ngen/prep.case is much higher.
On the one hand there are the contributions to historical grammar and language history. Buslaev (2009 [1862- 63]: 173), for example, points out that already in earliest translations from Greek into Old Church Slavonic relation al adjectives (in particular: possessive adjectives) were frequently used in place of the genitive or prepositional case, e.g.,
strach bozij 'fear of God; lit. fear God-Poss.ADJ'.
The frequent use of adj ective instead of case constructions is also characteristic of
Old Russian texts, e.g., vodnaja iazda 'thirst for water; lit. water-RA thirst' instead of
iazda vody lit. 'thirst watergen', smertnaja pamjat' 'memory of death; lit. death-RA memory' vs. contemporary Russian: pamjat' o smerti lit. 'memory about death', strel'naja
rana 'arrow wound; lit. arrow-RA wound' vs. rana ot strely 'id.; lit. wound from arrowgen', strach tatarskb 'fear of the Tatars; lit. fear Tatar-RA' vs. strach ot tatar 'fear of the
Tatars' .
According to Buslaev's investigations, combinations including possessive adjectives
still occurred in Russian literature of the 18th century, e.g., Kartiny Del'-Sartovy, Venera
Ticianova i procich masterov (Fonvizin) 'paintings of del Sarto, the Venus of Tizian and
of other masters; lit. paintings del Sarto-poss.ADJ, the Venus Tizian-Poss.ADJ and other
mastersgen' (particularly interesting in this context is the parallel use of possessive adj ectives in naming individual painters and the genitive (plural) when generally referring to
further masters).
In his history of the Russian language in the 18th and 19th century, Vinogradov ( 1982 :
70) points out a detail of the Petrine period which proves relevant for the topic of our
discussion: tables of direct borrowings from German were found to include - solely for
the purpose of clarification - additional translations, e.g., buchgalter 'bookkeeper' (or
knigoderiatel' lit. ' book-a -holder'), kamer-junker 'groom of the bedchamber' (or komnatnyj dvorjanin lit. 'room-RA noble man'). In the first case we also find a compound
in Russian, which corresponds to the typical expression of the argument (object) by
means of the modifier of a deverbal head, in the second case the equivalent of the
German nominal modifier is a relational adj ective.
PA767
In examining the changes of the usage of relational adjectives and the increasing
distribution of functions between Russian RA+N and N+N 0 b1.case in the 18th and 19th
century, Belosapkova and Zemskaja (1962) reached the following conclusions:
a) In cases, where a relational adjective continues to be used, no individual reference
can be established (as opposed to the genitive), e.g., okonnoe steklo 'window glass'.
b) Specific semantic relations such as object relations are less and less frequently expressed by relational adjectives but instead by use of the genitive case; also cf. ORuss.
plat'janoe myt' e lit. 'dress-RA washing' vs. contemporary myt' e plat'ja lit. 'washing
dressgen', krestn.oe celovanie vs. celovanie kresta 'kissing of the cross' (p. 8 f.).
c) At the end of the 181h and beginning of the 19th century, we can still encounter R.
vinnaja prodaza 'wine-sale; lit. wine-RA sale', notnyj perepisCik 'copyist of notes; lit.
note-RA copyist', but in the second half of the 19th century the number of occurrences
decreases, especially in combination w ith person nouns used as head, e.g., baletnyj
ljubitel' ' ballett lover' (p. 12).
Combinations of relational adjectives and (also non-deverbal) nouns are active to the
present day, e.g., designations of firms in which object and purposive relations interact,
R. obuvnajafabrika 'shoe factory', R. kirpicnyj zavod 'brickyard'. Similarly also instrumental relations are exploited, e.g., R. televizionnaja peredaca (besides telepredaca)
'telecast' (p. 14).
For synonymous and competitive occurrences between RA+N and N+Nobl.case in contemporary Russian see Zibrova (1984), where we can also prove the lack of expression
of object relations by means of relational adjectives in normative language use.
4.
Types of compounds without a linking vowel

and t heir adjectiva l competitors
4.1. Compou nds without a linking vowel

As shown in Table 43.3, apart from several Russian compounds which follow the tradi tional formation types (with a linking vowel and a deverbal head), there are (possibly)
also word combinations of RA+N. The same applies to (borrowed) compounds without
a linking vowel.
With regard to these formations the "Russian grammar" (Russkaja grammatika 1980:
253) still confines itself to a list of examples. In most cases they are direct borrowings
or calques, whose use is frequently abandoned in favor of the RA+N combination. Cf.
Table 43.4.
Neologisms show an activation of this formation type. While, for example, the Russian borrowing komp'juter 'computer ' was made available for further designations or
rather calques in the 1980s and 90s by means of the relational adjective (cf.
komp'juternaja technika 'computer technology'), in more recent designations with internet as modifier the compound without a linking vowel dominates: R. internet-opros
(also cf. P. internet-ankieta) ' Internet survey', internet-reklama 'Internet advertising'. In
Russian, parallel formations with the relational adj ective internetovskijlinternetnyj occur,
unlike in Polish (e.g., reklama internetowa), only marginally.
767
PA768
768

Tab. 43 .4: N+N compounds without linking vowel
Russian
Compound
N+N
Polish, Czech
RA+N
Compound
N +RA/RA+N
R. dizel' -motor
(16,000)
'diesel e ngine'
R . dizel' nyj
motor
(88,000)
[P. diesel (metonym)]

Cz. Dieseluv
(poss. adj.) motor
R. vakuum-kamera (3,000)
'vacuum chamber'
R. vakuumnaja
kamera
(24,000)
P. komora pr6iniowa ( f - pr6inia

'vacuum')
Cz. vakuova*
komora
R. diaz-orkestr
(90,000)
'jazz orchestra'
R . diazovyj
orkestr
(34,000)
* Jn Czech de rivatives
P. jazz-orkiestra
[Cz. jazz-orchestr]
P. orkiestra
jazzowa
Cz.jazzovy
orchestr
from foreign words in -um this ele ment is deleted.
As part of the internationalization of vocabulary that has taken place during the last
two decades, such formations (in the form of direct borrowings or hybrid calques) have
also significantly increased in other Slavic languages. Consider neologisms including
frequent constituents like R. bisnes, P. biznes, Cz. byznys 'business', documented in all
three languages, e.g., in connection with R. proekt, P./Cz. projekt and R./P. plan/Cz.
plan 'business project/plan', but R. biznes-otnofanija ' business relations' vs. P. stosunki
biznesowe (N+RA) 'id.', R. biznes-pravo 'business law' vs. P. prawo biznesowe ' id. ' .
New compounds of this type, which consist exclusively of native components and are
not loan translations, however, have not been documented . Since both constituents are
known as single words, examples of the above kind can also be perceived as independent
Russian formations. Unlike parallel designations such as R. dizajner-narkotiki and dizajnerskie narkotiki 'designer drugs' (cf. also the examples below) there are also fewer
synonymous combinations with a relational adjective competing with compounds with
biznes as modifier: R. biznesnye/biznesovye otnosenija have been documented with approximately 800 occurrences, including combinations with adverbs (Cisto biznesovye
otnoSenija ' mere business relations', but *Cislo biznes-otnosenija). The compound is
furthermore more frequent in combi nations with an additional, adj ectival modifier : politiceskie i biznes-interesy 'political and business interests' .
The occurrences that have been cited so far and many further examples show that the
modem Slavic languages (and in this context particularly Russian) are not so much
enriched by new types of compounds as they are influenced by a strong activation of
those (already existent) compounding methods that are most suited to the integration of
new borrowings/internationalisms. Several choices of designation may thereby be available (also beyond word-formation), see below.
As was established on the basis of an analysis of the "Russian Newspaper Corpus"
of the 1990s, the number of new noun compounds exceeded the number of the remaining
nominal neologisms by more than 10 % (Kukuskina, Polikarpov and Toktonov 2007).
PA769
769

Yet, with respect to such evaluations, we have to consider that the activation of
specific patterns - as reflected in the entries of neologism dictionaries (cf. Uluchanow
and Belentschikow 2007; Worbs, Markowski and Meger 2007; Martincova 1998,
2004) - does not necessarily have to go hand in hand with an increased occmTence
frequency of formations based on these patterns.
Besides new boITowed compounds, we can often find variants in the form of prepositional constructions and/or in the form of (fixed) word combinations on the basis of
RA+N, cf. Table 43.5.
Tab. 43.5: Loan compound, prepositional word combination, RA+N
Compound
Prepositional word combination
RA+N/N+RA
a)
diip-safari
safari na dzipach
diipovoe safari
'jeep safari'
(474,000)
'id.; lit. safari on jeeps'

( 43 ,000)
'id.; lit. jeep-RA safari'

( l)
P.
diip safari (233)
safari na diipach (7)
Cz.
diip safari (30,000)
safari na dzipech ( 1,360)
diipove safari (9)
serfing-kostjum (272)
kostjum(y) dlja serfinga (600)
serfingovyj kostjum ( 11)
'surfing clothes'
lit. 'clothes for surfing'
'id.'
P.
str6j do serfingu
str6j serfingowy
Cz.
obleceni na surfing
surfingove obleceni
demping-ceny
dempingovye ceny (20,000)
R.
b)
R.
c)
R.
(3,000)
'dumping prices'
'id.'
P.
dumping-ceny
ceny dumpingowe
Cz.
dumping-ceny
dumpingo ve ceny
In a), the loan compound is also more frequently documented in Polish and Czech.
In many other cases we can, however, encounter a higher frequency of synonymous
multi-word expressions, cf. b). In Czech and Polish an analogous compound could not
be documented. As is representative of the expression of final relations, combinations
consisting of N+Nprep.case dominate (in this case also in Russian), besides individual
occurrences of RA+N: P. str6j serfingowy, Cz. surfingove obleceni. The reasons for a
preference of RA+N/N+RA inc) are to be found in the enhanced syntactic availability.
In oblique singular and plural cases the ratio of the borrowed compound is considerably
lower, e.g.: R. dative plural demping cenam (59) vs. dempingovym cenam (59,000), and
the genitive plural R. demping cen, P./Cz. dumping cen is obviously entirely avoided
due to its homonymy with N+Ngen.pl 'price dumping'.
Finally we can recognize tendencies towards the use of the compound as a constituent
of proper names (e.g., R. ini iniring kompanija 'engineering company ' in business names
PA771
771
Tab. 43.6: Compounds with clipped modifiers and RA+N/N+RA combinations

Polish, Czech
Russian
avto(avtomobil'nyj RA/
avtomobil' N)
'car'
Compound
RA+N
Compound
N+RA, RA+N;
N+Ngen
avtodviienie
(34,000)
'car traffic'
avtomobil' noe
dviienie
(13 m.)
Cz. autoprovoz
P. ruch samochodowy
avtorynok
(878,000)
' car market'
avtomobil' nyj
rynok
(530,000)
P. aut0tynek*
Cz. autotrh
P. rynek samochodowy
avtoskola
(2 m.)
'driving
school'
avtomobil' naja skola

(78,000)
P. autoszkola,
auto-szkola *
P. ofrodek szkolenia kierowc6w

lit. 'centre for the
training of drivers'
Cz. autoskola
tele(televizionnyj RA!
televidenie N)
'TV '
telereportai
(47,000)
'TV report'
televizionnyj
reportai
(14,000)
P. telereportai
Cz. telereportaz
P. reportai telewizyjny
Cz. televizni reportaz
telezritel'
( 142,000)
'TV viewer'
televizionnyj
zritel' (52 1)
P. telewidz
Cz. teledivak
(occasionally;
coll.)
P. widz telewizyjny
Cz. televizni divak
* Frequently as part of proper names.
4.3.
Mu lti-word expression or compound?
4.3.l. On the notion of "analytical adjectives" in Russian studies

Particularly in Russian studies, some authors consider elements like avto-, tele-, etc.
mentioned in section 4.2 (which are described by the " Russian grammar" as clipped
compound elements going back to an N or a RA) as "analytical adjectives". They hence
agree with Panov (2004 [1971] amongst others) who treated components which mainly
stem from borrowings as "indeclinable adjectives" or "indeclinable modifiers'', respectively, e.g., R. radio- i televizionnye centry (Panov 2004 [ 1971 ]: 13 7) lit. 'radio- and television-RA centres'. In elements of thi s kind Panov saw an increase in analytic traits in Russian
morphology, i.e. grammatical categories such as gender, number and case (when establishing the congruity between an adjective and a noun) are not expressed in endings but exclusively contextually, the formation of compounds lacks linking vowels, etc.
Panov (2004 [1971]:138ff.) did not only classify the initial constituents in formations
such as partbilet ' party membership book' (+- partijnyj bilet RA+N), kinoscenarij
'script; lit. film-script', telepostanovka ' TV production' as analytical adjectives. Although we have the noun pressa, press-officer 'military press officer' would not be
PA773
43. Compounds and multi-word expressions in Slavic
able) adjective (Cz. bio celozrnna mouka 'organic wholemeal flour; lit. bio wholemeal-RA
flour'), and, on the other hand, (under bio-) as initial part of compounds. Consequently the
obvious variants Cz. bio brambory und biobrambory 'organic potatoes', which are distinct
only in spelling, receive a different structural explanation. Cf. also Gester (2001); Bozdechova (1997) on on line as an adjective in designations like Cz. on line system, besides
onlinovy (RA); on Polish: Worbs, Markowski and Meger (2007).
A broad conception of indeclinable adjectives is represented by Doleschal (2000),
where she gives special weight to the criterion of separate spelling (e.g., nonstop in Cz.
nonstop automobilova pout' 'non-stop automobile pilgrimage'). Among other things, she
points out that "a peculiar type of pseudo-compound structure", whose initial component
had to be considered as "uncommon conversion of a noun into an indeclinable adjective",
developed on the basis of loan translations (p. 288). So-called indeclinable adjectives
were for the most part business or brand names, their occurrence was strictly confined
to advertising texts, e.g., Slk. Crisan protilupinove vlasove tonikum 'Crisan Anti-Dandruff Hair Tonic; lit. Crisan antidandruff-RA hair-RA tonic' (p. 289 ff.). Doleschal (p.
295) regards the expansion of uninflectable adjectives as the "at the moment probably
most interesting question of grammatical change in the Slavic languages'', which is "advanced by intensive borrowing processes". Cf. also Hinrichs (2000) on analytic tendencies in the languages of Eastern and Southeastern Europe; Ohnheiser (2004).
b) Identification as compound (reinforcement of compound patterns)
According to Selimski (2003) a demarcation between word combinations and compounds in neologisms is difficult to achieve, since the relation between the constituents
is not expressed formally as is usually the case in Slavic languages (i.e. absence of the
linking vowel in compounds or of an adjectival suffix in an agreeing or, respectively, of
the case marker in a non-agreeing adjunct). The older formation with a linking vowel
Bulg. gazq_stancija lit. 'petrol/gas-a-station' is contrasted with, e.g., the more recent compound gazstancija (as English gas station) (p. 120). Nevertheless, Zelimski holds that a
new type of compound is thus emerging in the Slavic languages, in which the relations
between the constituents are only expressed by their position, independently of whether
the formations are spelled separately, with a hyphen or as one word (cf. Sln. Slovenijales,
Slovenija-les, Slovenija les lit. 'Slovenia-wood/timber'; name of a Slovene timber company (p. 121)). This kind of composition would be particularly suitable for formations
whose modifier is an abbreviation (e.g., Sln. TV-studio, -chronika 'TV studio, chronicle'), but also in other cases of more recent designations, compounds are the preferred
structure, e.g., Bulg. baskettitla 'basket title ' (instead of RA +N basketbolna titla), basketfinal ' basket final'. Such phenomena are, according to Selimski, especially representative of the "analytic languages" Bulgarian and Macedonian, which are also close to
Serbian, but less frequently found in other South Slavic languages or in Czech. (Cf.
however also Cz. diskohudba 'disco music', diskoodev 'disco clothes', etc.)
Jadacka (2001: 93 ff.) describes Polish formations with constituents such as euro-,
eko-, narko-, porno-, seks-, tele-, wideo- as "zrosty", i.e. a special type of compound, and
makes reference to proper names like P. Sopotfestiwal 'Sopot festival' and Jnwestbank
(= Bank Jnwestycyjny N+RA) 'investment bank' (p. 143) as an additional formation type
among neologisms, lacking fonnal markers of composition (linking vowel). (See also
Waszakowa 2005 and article 96 on foreign word-formation in Polish.)
For typological tendencies in modern Slavic languages cf. also Gutschmidt (2003).
773
PA775
for individual semantic relations. In addition to that there are several cases of ellipsis of
the head noun and nominalization of the relational adjective.
Factors pertaining to the economy of language such as brevity and easier syntactic
availability as opposed to multi-word expressions enhance the productivity of synonymous derivatives, primarily in technical jargon but also in colloquial language; the frequency of use might lead to their being stylistically neutralized (see article 42 on multiword expressions and univerbation in Slavic). The following recent formations however
might still exhibit a colloquial tinge, e.g., R. okonnik 'window air conditioner' (+----oknonnyj -RA kondicioner 'id. ' ), vidik 'video player' (+----- videokassetnyj plej erlvideoplej er
' id.' ; with clipping of the modifier), Cz. digitalky 'digital watch' (+----- digitalni hodinky
' id .'), P. mikrofal6wka ' microwave' (+----- ku.chenka mikrofalowa 'id.').
Derivatives ( +----- RA) formed on the basis of multi-word expressions (RA+N) are not,
however, always synonymous with their source, cf., for example, personal nouns such
as R. neftjanik 'oil worker ' (+----- neftjanoj 'oil-RA', semantically referring to neftjanaja
promyslennost' 'oil industry ' ), or P. naftowiec ' id. ' (+----- naftowylprzemysl naftowy ).
As could be learned from the above descriptions and tables, in Slavic languages not
only combinations of RA+N or N+N0 b1.case but also derivatives can be found as equivalents of English (or German) compounds. In some cases there are certain regularities,
e.g., in the word-formation category of pl ace nouns (implying final relations), which
is also why there are fewer parallel RA+N-combinations, e.g., E. cowshed - R. korovnik/
Cz. kravin (+----- korovalkrava 'cow'), E. vineyard - R. vinogradnik (+----- vinograd 'vine,
grape'), P. winnica/Cz. vinice (+-----wino/vino 'vine, grapevine' ); E. teapot, teakettle - R.
cajnik/Cz. cajnik (+----- caj 'tea'), etc.
Denominal person a 1 nouns also have to be mentioned in this context, whose
implicit actional meaning is explicated in their English compound-equivalents, e.g., R.
parketCik, P. parkieciarz, Cz. parketaf 'parquet floor layer' (+----- parket/parkiet 'parquet'),
R. spletnik and P. plotkarz 'scandalmonger' (+----- spletnja/plotka 'gossip'). While in English we can find both designation types (footballer and football player) in certain names
(e.g., those of athletes), Slavic languages here too usually have derivatives, be it on the
basis of borrowings (Cz. fotbalista, volejbalista), loan translations or calques (P. pilkarz
'footballer' +----- pilka noina 'football; lit. ball foot-RA', siatkarz 'volleyball player' +----siatka 'net' respectively pilka siatkowa 'volleyball; lit. ball net-RA').
6. Conclusion
To a significant extent the increase of new vocabulary in the modern Slavic languages
feeds on borrowings and loan translations or hybrid formations, a large part of which
consist of compound patterns and compound elements of foreign origin. Direct borrowings or loan transfations are not only to be found in the domains of technology, economy,
finances, etc., but also increasingly in colloquial vocabulary.
New compounds, which consist exclusively of native components and do not represent loan translations, are, however, rarely documented or never. Yet the foreignness of
neologisms is mitigated in so far as one or both components are usually already known
as borrowings; the compound can consequently also be perceived as an autonomous
formation.
775
PA783
44. Paradigmatically determined allomorphy
deverbal verbs:
783
desiderative l st conjugation verbs in -urio

intensive 1st conjugation verbs without any overt
derivational affix
iterative l st conjugation verbs in -ito
Examples of derived lexemes based on the third stem of the verbs CAPI<5 'to take, seize'
(third stem capt-) and SCRIBO 'to write' (third stem script-) are listed in (3):
(3)
a.
captor 'hunter', captio 'catching; deceit', captura 'catching of animals', capto

'to strive to seize', captito 'to strive eagerly after'
b. scriptor 'writer', scriptio 'writing', scriptura 'writing', scripturio 'to desire to

write', scriptito 'to write often'
Assuming the third stem as the base of all these processes involves the hypothesis that
all the overt suffixes that appear in the various formations are vowel-initial (-or, -io,
-ura, -urio, -ito); these suffixes are suffixed to the third stem, yielding derivatives such
as amator, cubitor, scriptor, possessor, plausor, tonsor (I use agent nouns as examples).
Under this analysis, no phonological adjustment is necessary; the simple agglutination
of the vowel-initial suffix to the third stem yields the attested form of the derivatives,
as shown in (4).
(4)
Base verb's stems

ama-, amau-, amatcuba-, cubu-, cubit-
'to love'
'to lie down'
scrib-, scrips-, scriptposside-, passed-, possessplaud-, plaus-, plaustonde-, totond-, tons-
'to
'to
'to
'to
write'
possess'
clap'
shear, to shave'
Agent noun
am a tor
' lover'
cubitor
' one who lies
down'
scriptor
'writer '
possessor 'possessor'
'applauder'
plausor
tons or
'barber'
2.2. Traditiona I ana lyses for these processes

Traditional analyses (as represented, for instance, in Leumann 1977, or Kuhner and
Holzweissig 1912) assume instead that the processes in (2) are not based on the third
stem, but on the verb's root or present stem (ama-, cub-, scrib-, plaud-) or a substring
thereof (passed-, tond-) followed by a /ti-initial suffix; apparent allomorphies are accounted for by appeal to regular phonetic laws (somewhat reminiscent of Aronoff's
"phonological tours de force"). For example, scriptor shows regressive voicing assimi lation (scrib-tor > scrip-tor); possessor, plausor and tonsor result from passed-tor, plaudtor and tond-tor by a series of events, which are assumed to be the following: voicing
assimilation (/d+t/ > /t+t/); Isl epenthesis to break the /t+t/ sequence; assimilation of the
/t/s on both sides to the epenthetic Isl; degemination of Issi in certain phonotactic contexts (after sonorants, long vowels and diphthongs). Synchronically a number of facts
remain unclear under this analysis (even in the cases for which it would be possible to
give a diachronic account): for example, the contrast between cubitor and scriptor (why
PA784
784

not *scribitor or *cuptor?). Furthermore, the form of the verb chosen by one and the
same suffix appears to vary between a stem with thematic vowel (ama-), the bare root
(scrib-, etc.), and a string lacking a predesinential -e- which appears in the present tense
forms (tond- vs. tonde-; in the case of passed- vs. posside-, the stem that is assumed to
be the base of the -tor derivative differs from the one that surfaces in present tense forms
also in the quality of a stem-internal vowel). All these problems disappear if one assumes
synchronic derivation of the agent nouns from third stems, by way of a vowel initial
suffix -or (and likewise for all the other processes).
2.3. Other processes possibly based on the third stem

Further derivational processes that might be argued to be based, at least in some cases,
on the third stem are those listed in (5):
(5)
4 111 declension masc. nominalizations without

overt affix
feminine 3rd declension agent nouns in -r'fx
Adjectives in -'ivus, -a, -um
Adjectives in -orius, -a, -um
Adverbs in -im
Adjectives in -ibilis
sensus 'feeling'
victrix 'conqueress'
activus 'active'
decretorius 'decisive'
cursim 'quickly, hastily',
corruptibilis ' liable to decay'
In addition, through a process whose exact origin has been fiercely debated (see Georges
1968 for a critical review), deverbal action nouns homophonous with the femini ne form
of the past participle, such as expensa, remissa, developed in early Romance, and are
continued in Italian (cf. section 3.11).
Each of these derivational categories deserves comment.
2.4.
4th
declension masculine nominalizations
The productivity of 4th declension masculine nominalizations based on the third stem
and lacking an overt derivational affix in Classical Latin is debated (Georges 1968: 3 76380; Panagl 1987); in Italian some of these nouns are continued (senso, frutto, atto .. .),
but they are not homophonous with the past participles of the corresponding verbs (sentito vs. senso from sentire 'to feel', agito vs. atto from agire 'to act').
2.5. Feminine agent nouns in -rTx

Feminine agent nouns in -r'ix (about 350 in Lewis and Short 1879) are sometimes considered to be derived from the corresponding masculine noun by suffix substitution. In most
cases, such an analysis is empirically undi stinguishable from one assuming that -rix is
PA785
44. Paradigmatically determined allomorphy
suffixed to the third stem: substitution of -(t)rix for -(t)or or suffixation of -rix to the
third stem give the same result in cases such as victorlvictrix 'conqueror/conqueress' .
Three derivatives appear to contain a stem which is phonologically slightly distinct
from the base verb's third stem: genetrix ' mother' vs. genit- (gigno ' to beget, bear, bring
forth, produce'), meretrix 'she who earns money; hence, a prostitute ' vs. merit- (mereo
'to deserve' ), obstetrix 'midwife' vs. obstat- (obsto 'to stand before or against sth. ') or
obstit- (obsisto ' to place one's self before sth.' ). Although the by-forms genitrix and
obstitrix are also attested, albeit less frequently than genetrix and obstetrix, meretrix has
no *meritrix counterpart. These three forms would also be problematic for an analysis
assuming that the masculine agent noun is the base, as there is no corresponding masculine for meretrix and obstetrix (i.e. "bei nur weiblichen Berufen" [with jobs that are only
feminine] according to Leumann 1977: 329.3), and the masculine noun corresponding
to genetrix is genitor. It must be observed that a genet- stem appears also in genetivus
'genitive ' (which has a less frequent by-form genitivus). Whatever the reason for this
irregularity, it appears in three nouns that designate almost exhaustively the roles traditionally assigned to women in an ancient society, thereby forming a close network.
A particularly interesting subset of nouns in this class is represented by a handful of
forms ending in -strix, such as tonstrix 'female hair-cutter or barber', plaustrix 'female
applauder'. Unfortunately most of these nouns are attested only by grammarians or in
glossaries, and are therefore of a dubious nature; however, at least tonstrix, suppostrix
'she that fraudulently exchanges or substitutes', ambestrix 'female consumer' and persuastrix 'female persuader ' are accepted lectiones in the works of Plautus (although,
since Plautus is well known for the humorous character of his language, Latinists usually
dismiss these forms as insignificant). Several analyses can be offered for these formations. Those who assume the suffix to be -trix observe that when It/ is followed by Ir/ it
doesn' t assimilate to the epenthetic Isl that arises between a stem-final /d/ or /t/ and the
/tr/ cluster: therefore, tond-trix regularly develops to tonstrix (presumably by way of the
following stages: tond-trix > tont-trix > tont-s-trix, tons-s-trix > tonstrix). Assuming
instead that the su ffix is -rix and the base is the third stem requires an explanation for
the appearance of /ti between base and suffix. Such an explanation can in fact be found
in Priscian 's (1999) Partitiones. Priscian assumes that the suffix is -rix, and that the
feminine agent nouns are derived by substituting -rix for the -or suffix found in the
masculine nouns, in his account derived from the past participle; if the masculine ends
in -sor, a feminine derivative is better avoided "propter asperitatem pronuntiationis"
[because of the harshness of pronunciation]. If it is absolutely necessary to form a -rix
derivative from a stem ending in /s/, the /sr/ cluster is broken by inserting a /t/:
o mnia partic ipia praeteriti temporis us in or conuertentia faciunt nomen uerbale in omni
coniugatione masculinum ex quo iterum or in rix mutantes fac imus femininum, nisi euphonia, id est sonus, prohibeat, quod euenit in illis quae in sor desinunt ut pransor cursor
tonsor. nemo enim dicit pransrix cursrix tonsrix propter asperitatem pronuntiationis. unde et
Terentius tonstrina dixit euphoniae causa adde ns contra regulam t. sicut e nim a doctore
doctrina consonantes eas habuit quas suum primitiuum, sic debuit etiam tonstrina absque t
esse nisi sonoritas coegisset. defenstrix quoque C icero in Timaeo protulit addita t.
(Keil III, 463 , 11-20; Passalacqua 1999, 51 , 13-22)
[All perfect pa1ticiples, by converting -us to -or, ma ke a masculine verbal noun in every
conj ugation, from which in turn, by c hanging -or to -rix, we make a fem inine [noun], unless
euphony, i.e. sound prevents it, which happens in those nouns which end in -sor, such as
785
PA786
786

pransor, cursor, tonsor. For no one says pransrix, cursrix, tonsrix on account of the harshness of [their] pronunciation. This is why Terence says tonstrina, adding -t- aga inst the rule
for the sake of euphony. For just as doctrina has gotten from doctor those consonants w hich
its base has, so ought tonstrina also have been without a -t- were it not for the fact that
sound had made it necessary. Also by adding a -t- Cicero produced the word defenstrix in
the Timaeus.]
Indeed, /sr/ is a universally di sfavoured sequence (Passino 2007), and almost unattested
in Latin (the only words containing it listed by Lewis and Short 1879 are disraro,
disrumpo, Cisrhenanus, Transrhenanus and Israel). Insertion of an intrusive stop to
enhance syilable contact is a well-attested repair strategy (e.g., in Old French essre
[< Latin *essere] developed to estre 'to be' [Modern French etre]) . This repair strategy
is only very marginally available in Italian, as will be shown in section 3.10.
2.6. Adjectives in -Tvus

Adjectives in -'fvus (about 280 in Lewis and Short 1879) can take as bases nouns, verb
roots, or verb third stems: examples are aest'fvus ' pertaining to summer ', internec'fvus
'deadly, mm derous', and actfvus 'active' . The maj ority of derivatives are from third
stems, particularly from 1st conjugation verbs (125 items in -at'fvus); many derivatives
are metalinguistic terms; often the adjectives have been nominalized after ellipsis of a
head noun, as in the names of cases (genet'fvus, dat'fvus, accusat'fvus, ablatzvus, all attested from Quintilian onwards).
2.7. Adj ectives in -orius

Adj ectives ending in -orius (about 350 in Lewis and Short 1879) are for the most part
derived from agent nouns in -or by way of the suffix -ius, as in the cases of praetorius,
quaestorius, censorius, etc. However, in some cases a corresponding agent noun in -or
is not attested, and derivation from a verb's third stem is the only possible analysis, as
in the cases of decretorius 'decisive' ( +- decerno, decret- 'to decide'), meritorius 'that
brings in money' ( +- mereo, merit- ' to earn'). In other cases, even if a corresponding
agent noun in -or is attested, an analysis that has the -orius adjective based on the verb
is semantically more plausible. This is often the case with nominalized adj ectives resulting in neuter nouns in -orium that designate places where an activity is carried out, such
as deversorium ' inn, lodging place', more likely from deverto, devers- 'to lodge' than
from the hapax deversor 'lodger '.
2.8. Adverbs in -im

Adverbs in -im have been extensively studied (cf. Ricca 2010 and references therein,
particularly Schaffner-Rimann 1958). It is commonly accepted that the suffix originated
through the reanalysis of an originally inflectional ending (considered to be an accusative
PA788
788

refer only to arguments that corresponding !-participles refer to: e.g., ducibilis means
both 'who can lead' and 'leadable' while ductibilis can only mean 'leadable'" (Steriade
forthc. 5.2). Steriade's hypothesis is that the relation "between pairs like ductus-ductibilis, passus-passibilis is the default relation between any base and its derivative: derivatives inherit both the phonology and the syntactic and semantic structure of the bases
they contain" (Steriade forthc. 5). In other words, she rejects the hypothesis that semantically void, purely morphomic, bases operate in derivation.
2.10. An alternative analysis

Steriade's account, by which the semantics of -bi/is derivatives from participial stems is
a function of the passive syntactic content of their bases, of course begs the question of
why we don 't find similar semantic properties in other derivatives from participial stems,
such as agent nouns in -(t)or. Her answer is the following: the derivatives in (2) are not
formed by concatenating a verb's third stem with a vowel-initial suffix, as in Aronoff's
( 1994) analysis; rather, they are formed by selecting a verb's root or infectum I present
stem and a !-initial suffix, in which this initial consonant is a "stem extension". The
structure of a Latin deverbal derivative, in Steriade's analysis, is the following: [[[[root]
+ buffer or theme vowel] + first C of a derivational suffix] + the rest of the derivational
suffix]; an example of th is structure is auditor, to be analyzed as [[[[aud) + 1] t] + or].
The string up to and including the initial consonant of the suffix (e.g., [[[aud] + 1) t]) is
called a "minimal stem". A derivative such as pulsar would have the structure [[[[pell]
+ i] t] + or]. But why, then, doesn't it surface as *pellitor? According to Steriade, this
is due to the workings of a series of correspondence constraints, by which, if two minimal stems headed by the same verb root end in homorganic [a sonorant] segments, they
stand in correspondence, and must merge in order to surface as identical. Min imal stems
such as puls- (the participial stem) and pellit- (the stem of the derivatives w ith !-initial
suffixes), both from the verb pello 'to push, drive away' (i.e., "headed by the same verb
root"), meet these criteria as they both end in a coronal obstruent; pellit- is then changed
to puts- to obey the correspondence requirement. The change doesn't happen the other
way around (i.e. we don't get *pellitus as a past participle) because "the direction of the
merger is determined by correspondence to a listed word form"( 4.1), in this case puts-.
Listed word forms are assumed to be those "deviating unpredictably from regular trends"
( 4.1).
Such an account explains the lack of inheritance of the passive past participle's syntactic features in the derivatives in (2): the appearance of the participial I third I t- stem
in these derivatives is not due to its selection by lexeme-formation rules, but "is the
consequence of a phonological chain of events taking place, presumably, in a postsyntactic component" (Steriade forthc. 4.5).
Steriade's analysis shifts the arbitrariness from the morphological component to the
phonological one: the specification of which entities stand in correspondence to one
another, and under which conditions this occurs, is as arbitrary and language-specific as
the specification of which lexeme-formation processes select a given morphomic stem.
Further level-internal (phonological or morphological) regular "laws" operate in each
case. Steriade's account is preferable only in models that assume no separate morpholog-
PA792
792
3.5. Comparison of analyses

The main arguments from Thornton's and Bisetto's analyses will be recapitu lated below.
For derivatives from 1st conjugation verbs and most 3rd conjugation verbs, the two
competing analyses (derivational or infinitival stem + -zione vs. past participle stem +
ione) are not distinguishable on empirical grounds: both predict the correct output (e.g.,
amministra + zione and amministrato + ione with the readjustments in (6) above both
yield amministrazione). But for most other verbs, the analysis that assumes the Italian
past paiticiple stem as base makes wrong predictions, as shown in Table 44. l.
An elaboration of the data in Thornton (1989) shows that over a third (33.8 %) of the
278 -ione derivatives in her corpus are not generated from the Italian past participle stem
by way of Scalise's series of rules. Bisetto (1999: 51-53) observes that among the -tore
derivatives from about 150 verbs with an irregular past participle, only about 40 % can
be generated by attaching -ore to the past participle, while the rest can only be analyzed
as derived by attach ing -tore to the infinitival stem (with readjustment). The data in
group A in Table 44. l are not generated by Scalise's analysis, while they would be
generated correctly by Thornton's and Bisetto's analyses. The data in group Bare incompatible with Scalise's analysis but compatible both with Thornton's and Bisetto's analyses and w ith the hypothesis that they contain a reflex of the Latin third stem. Finally,
the data in group C can be generated by neither Scalise's nor Thornton's and Bisetto's
analyses; they appear to contain a reflex of the Latin third stem.
Derivatives in -ione and -ore which are not analyzable as based on the Italian past
participle stem are either formed in Italian by attaching -zione or -tore to a verb stem
different from the past partic iple (Thornton's derivational stem, or Bisetto's infinitival
stem with a readjustment rule raising /e/ to Iii), or borrowings from Latin (and therefore
related to the form of the Latin third stem), as the examples C in Table 44.1. In Thornton's (1989) corpus, Latin borrowings represent 82.2 % of the -ione derivatives (Thornton 1989: 228).
A further argument brought forward in Thornton (1989: 230-231, 1990: 201-202) to
support the hypothesis that the productive lexeme-formation rule in Ita lian attaches -zione to verb stems was that in her corpus all the -ione derivatives which were not borrowings were analysable as formed in this way, and some were not equally analysable as
formed from the past participle (e.g., comparizione 'appearance' +--- comparire ' to appear ' vs. PST.PTCP comparso ~ *comparsione, volizione 'volition' +--- volere to ' want'
vs. PST.PTCP voluto ~ *voluzione).
Further evidence can now be coll ected by consulting electroni c resources which were
not available at the time of Thornton's and Bisetto 's studies. This evidence partly quali fies the conclusions reached by Thornton and Bisetto in the Nineties.
PA793
~
Tab. 44.1: Some Italian derivatives in -ione and -ore under several analyses
Italian Past Partic iple
Expected derivat ives under

Scalise's analysis
Italian derivationa l stem (or readjusted infinitival

stem)
Expected derivaAttested derivatives under Tho rn- tives

ton's and Bisetto's
analysis
Latin third stem
bevuto
voluto
scoperto
vinto
perduto I perso
bevivoliscoprivinciperdi-
bevitore
volizione
scopritore
vincitore
perdizion.e
bevitore
volizione
scopritore
vincitore
perdizione
vendere
aggredire
venduto
aggredito
vendiaggredi-
attribuire
convenire
costituire
costruire
attribuito
convenuto
costituito
costruito
distribuire
distribuito
istruire
istruito
possedere
posseduto
pre mere
restituire
premuto
restituito
retribuito
sostituito
venditore
*aggredizione,
*aggreditore
*attribuizione
*convenizion.e
*costituizion.e
*costruizione,
*costruitore
*distribuizione,
*distribuitore
*istruizione,
*istruitore
*possedizione,
*posseditore
*premizione
*restituizione,
*restituitore
*retribuizione
*sostituizione
venditore
aggressione,
aggressore
attribuzion.e
convenzione
costituzione
costruzione,
costruttore
distribuzione,
distributore
istruzione,
istruttore
possessione,
possessore
pressione
restituzione
retribuire
sostituire
*bevutore
*voluzione
*scopertore
*vintore
*perduzione I
*persione
*vendutore
*aggredizione,
*aggreditore
*attribuizione
*convenuzione
*costituizion.e
*costruizione,
*costruitore
*distribuizion.e,
*distribuitore
*istruizione,
*istruitore
*posseduzione,
*possedutore
*premuzzone
.
*restituizione,
*restituitore
*retribuizione
*sostituizione
bibit-
bere
volere
scoprire
vincere
perdere
Verb
"'d
.,Cl
Cl
0.
(coopert-)
victperdit-
-a
0.
(1)
.,
( 1)
attribuiconvenicostituicostruidistribuiistruipossedipremirestituiretribuisostitui-
retribuzione
sostituzione
venditaggressattributconventconstitutconstruct-
(1)
0.
e.
a
0
.::r
distributinstructpossesspressrestitutretributsostitut-
-...J
'-0
VJ
PA794
794
3.6. Recent -ione deri vatives

Examination of all the lexemes ending in -sione considered by GDU to have entered
usage after 1900 reveals that most of them are derived from older lexemes in -sione by
prefixation or compounding, and do not offer evidence as to the form of the base in a
productive suffixation rule. However, a handful of new forms that are not analysable as
formed from the verb stem + -zione are listed. Corrasione, evorsione and retrogressione
are technical terms of geology, whose base is considered by GDU to be a Latin form:
no Italian verb exists from which these lexemes could be derived; suffosione, another
technical term of geology, is claimed by GDU to derive from sujfondere, but derivation
from this verb's past participle would yield *sujfusione, as would direct borrowing of
Latin SUFFus10; the lo/ of sujfosione is hard to explain, and there are no other derivatives
ending in -fosione in Italian; this isolated form doesn't militate in favour of any of the
competing analyses. Five forms can be analyzed as derived from an Ita ~ ian past participle
stem, and are not consistent with the hypothesis that the formation rule attaches -zione
to the verb 's derivational or infinitival stem: erasione, dismossione, rimessione, scommessione, seclusione. Of these, only rimessione ' remittal ' is well attested: it is a legal
technical term, and probably the best counter-example to the claim that new Italian
derivatives are always formed on the verb stem (that would yield *rimettizione). It must
be observed that Italian legal language is well-known for making use of many Latin
borrowings and Latinate formations; since the true Latin learned borrowing remissione
was already in use as a legal technical tenn with a different meaning ('remission'),
coinage of rimessione was probably the best option for retaining a Latin model and still
coining a new technical term. Erasione and seclusione could be expla ined by appeal to
analogy with more frequent pre-existing lexemes such as abrasione (a Latin borrowing)
and a number of Latin borrowings in -clusione, related to verbs in -cludere, such as
conclusione, esclusione, occlusione, inclusione, reclusione. Dismossione is almost a
ghost-word (I could locate only 7 tokens in a Google search on 23/2/2011), and again a
technical term; scommessione is also very rare. The existence of these forms shows that,
in the competence of cultivated Italians, the possibility of forming a -ione derivative
from an Italian or Latin past participle stem is not completely excluded (contrary to what
was claimed by Thornton 1989, 1990- 91). However, derivatives based on the derivational stem continue to be formed, even if a synonymous Latin borrowing already exists or
would be possible (as in the cases of gestizione 1987 vs. gestione 1305, and statuizione
1950 vs. Latin statutio and Italian costituzione, istituzione, etc.). The overwhelming majority of neologisms are from 1 st conjugation verbs and are analyzable as formed by
concatenating the derivational stem and -zione (see the examples in Iacobini and Thornton 1992: 38- 39).
3.7. Recent derivatives in -ore, -ivo and -orio

Concerning other derivatives to which the two competing analyses were applicable, Rainer (2001: 387) has observed that "some relatively recent neologistic agent/instrument
nouns such as estorsore, interditore, manutentore, predittore or solutore [ ... ] are formally
related neither to the default stem nor to the past participle, but to the action noun in
PA802
802

Ricca, Davide
2004
Aggettivi deverbali. In: Maria Grossmann and Franz Rainer (eds.), La formazione de/le
parole in italiano, 419- 444. Ti.ibingen: Niemeyer.
Ricca, Davide
2010
Adverbs. In: Philip Baldi and Pierlu igi Cuzzolin (eds.), New Perspectives on Historical
Latin Syntax. Vol. 2: Constituent Syntax: Adverbial Phrases, Adverbs, Mood, Tense,
109-191. Berlin/New York: Mouton de Gruyter.
Scalise, Sergio
1983
M01fologia lessicale. Padova: Clesp.
Schaffner-Rimann, Judith
Die lateinischen Adverbien auf-tim. W interthur: Keller.
1958
Steriade, Donca
forthc.
T he morphome as constraint interaction: Latin t-stem derivatives. In: Ana R. Luis
and Ricardo Bermudez-Otero (eds.), Th e Morphom e Debate. Diagnosing and analysing
morphomic patterns. Oxford: Oxford University Press.
Thornton, Anna M.
1989
Sui nomina actionis in italiano. Ph.D. dissertation, Dipartimento di linguistica, Universita di P isa.
Thornton, Anna M .
1990- 9 1
Sui deverbali italiani in -mento e -zione. Archivio Glottologico Italiano 75: 169207 and 76: 79- 102.
Thornton, Anna M .
2004
Mozione. In: Maria Grossmann and Franz Rainer (eds.), La formazione delle parole in
italiano, 218- 227. Ti.ibingen: Niemeyer.
Thornton, Anna M.
2005
M01fologia. Roma: Caracc i.
Thornton, Anna M.
2007a Is there a partition in the present indicative of Italian regular verbs? Annali online delta
Jaco/ta di Lettere e Filosofia dell' Universita di Ferrara, !l. 2: 43- 61. http://annali .unife.it/lettere/article/view/ 127 /78 [last access 19 Sept 20 14).
Thornton, Anna M .
2007b Phenomenes de reduction en italien. In: Elisabeth Delais-Roussarie and Laurence Labrune (eds.), Des sons et des sens. Donnees et mode/es en phonologie et en morphologie,
241-268. Paris: Hermes Sc ience/Lavois ier.
Thornton, Anna M.
2012
Quando parlare de lie donne e un problema. In: Anna M . Thornton aind Miriam Voghera
(eds.), Per Tullio De Mauro. Studi offerti dalle allieve in occasione de! suo 80 compleanno, 301- 3 16. Roma: Aracne.
Valla, Lorenzo
1577
De linguae latinae elegantia libri sex. Coloniae Agrippinae: apud Ioannem Gymnicum.
Anna M Thornton, L'Aquila (Italy)

Word Formation

Încărcat de

Informații document

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Word Formation

Încărcat de

Drepturi de autor:

Formate disponibile

PR1

H erausgegeben von I Ed ited by

Library of Congress Cataloging-in-Publication Data

Bibliographic information published by the Deutsche Nationalbibliothek

11 1. Units and processes in wo rd-f o rm atio n II: Specia l cases

Affective palatalization in Basque Jose Ignacio Hualde

Rules, patterns and schemata in word-formation Heike Baeskow

Rules and rest rictions in word-formation 11 :

Argument-structural restrictions on word-formation patterns Holden Hartl

VI. Semantics and pragmatics 1n word-formation I:

Motivation, compositionality, idiomatization Daniela Marzo

VII. Semantics and pragmatics in word-formation II:

Noun-noun compounds Christina L. Gagne and Thomas L. Spalding

IX. Foreign word-formation, language planning and purism II:

Foreign word-formation in German Peter 0. Muller

Historical word-formation I: General aspects

102. Mechanisms and motives of change in word-formation Franz Rainer

XI. Historica l word-formation 11: Special cases

XI I. H istorica I word-formation 111: Language sketches

Historical word-formation in German Peter 0. Muller

XI 11. Word-formation in language acquisition and aphasia

XIV. Word-formation and language use

text Anja Seiffert

XV. Tools in word-formation research

134. German Irmhild Barz

Dutch Geert Booij

Portuguese Bernhard Poll

152. Breton Elmar Ternes

Upper Sorbian Anja Pohontsch

169. Lithuanian Bonifacas Stundzia

17 1. Albanian Monica Genesin and Joachim Matzinger

172. Greek Angela Ralli

173. Ossetic David Erschler

Nenets Beata Wagner-Nagy

183 . Maltese Joseph Brincat and Manwel M ifsud

184. Turkish Jens Wilkens

Crimean Tatar Henryk Jankowski

191. K almylk D anara Suseeva

192. Abkhaz Viacheslav A. Chirikba

Rutul Mikhail Alekseyev

I. Word-formation as a linguistic discipline

I. Word-formation as a linguistic discipline

adverb-forming suffix -ly as in really or elegantly should be treated as a derivational,

2. Morphological building blocks and the internal structures

I. Word-formation as a linguistic discipline

1. The scope of word-formation research

Fig. 1.2: Survey of word-formation patterns

1. The scope of word-formation research

exemp lified for English deverbal nominalization in

a) general morphological properties of

c) formal and semantic characteristics

- formal description: [V + -(a)tion ]N

- stress movement and vowel change, e.g.,

e) restrictions on the nature of the input

- generally not applicable to English words of

f) degree of productivity (see article 47)

- fu lly productive, especially in formal contexts

4. Approaches to word -formation research

5. Levels of analysis and description in word-formation resea rch

I. Word-formation as a linguistic discipline

6. Theoretical models of wo rd-formation

"Rule of negative un#

semantics (roughly) un#X = not X

Forms of the base