Sunteți pe pagina 1din 235

Word Formation in South American Languages

Studies in Language Companion Series (SLCS)

This series has been established as a companion series to the periodical
Studies in Language.
For an overview of all books published in this series, please see

Werner Abraham

University of Vienna /
University of Munich

Elly van Gelderen

Arizona State University

Editorial Board
Bernard Comrie

Max Planck Institute, Leipzig

and University of California, Santa Barbara

William Croft

University of New Mexico

sten Dahl

University of Stockholm

Gerrit J. Dimmendaal
University of Cologne

Ekkehard Knig

Free University of Berlin

Christian Lehmann
University of Erfurt

Marianne Mithun

University of California, Santa Barbara

Heiko Narrog

Tohuku University

Johanna L. Wood

University of Aarhus

Debra Ziegeler

University of Paris III

Volume 163
Word Formation in South American Languages
Edited by Swintha Danielsen, Katja Hannss and Fernando Ziga

Word Formation
in South American Languages
Edited by

Swintha Danielsen
University of Leipzig

Katja Hannss
University of Regensburg

Fernando Ziga
University of Bern

John Benjamins Publishing Company



The paper used in this publication meets the minimum requirements of

theAmerican National Standard for Information Sciences Permanence
of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data

Word Formation in South American Languages / Edited by Swintha Danielsen, Katja
Hannss and Fernando Ziga.
p. cm. (Studies in Language Companion Series, issn 0165-7763 ; v. 163)
Includes bibliographical references and index.
1. Indians of South America--Languages. 2. South America--Languages. 3. Grammar,
Comparative and general--Word formation. 4. Language and languages-Variation. 5. Languages in contact--South America. I. Danielsen, Swintha,
1972- editor. II. Hannss, Katja, 1977- editor. III. Ziga, Fernando. editor.
isbn 978 90 272 5928 8 (Hb ; alk. paper)
isbn 978 90 272 6966 9 (Eb)

2014 John Benjamins B.V.

No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any
other means, without written permission from the publisher.
John Benjamins Publishing Co. P.O. Box 36224 1020 me Amsterdam The Netherlands
John Benjamins North America P.O. Box 27519 Philadelphia pa 19118-0519 usa

Table of contents
Introduction and acknowledgments
Swintha Danielsen, Katja Hann, & Fernando Ziga

Nominal compounds in Mapudungun

Fernando Ziga


Towards a characterization of compounding in Mak

Temis L. Tacconi


Augmentative in Toba (Guaycuruan): Form and function

Paola Cneo


Productive compounding in Baure (Arawakan)

Femmy Admiraal & Swintha Danielsen


Two types of incorporation in Ese Ejja (Takanan)

Marine Vuillermet


Reduplication in Yurakar
Rik van Gijn


Reduplication strategies in Kallawaya

Katja Hann


Compounding in Kalapalo, a Southern Cariban language

Ellen B. Basso


Nominalization in Choln
Astrid Alexander-Bakkerus




Introduction and acknowledgments

Swintha Danielsen, Katja Hann, & Fernando Ziga
University of Leipzig / University of Cologne / University of Bern

Word formation is not a well-defined area of linguistics but encompasses a variety of

processes as different as e.g. back formation and noun incorporation. Accordingly,
Laurie Bauer notes:
The term word formation is ambiguous in modern linguistics. In one usage of
the term, it is equivalent to morphology and refers to matters such as affixation
and reduplication as used in the creation of words. In the more widely accepted
sense of the term, word formation refers to the creation of new lexemes in a given
language. Thus it specifically excludes inflectional morphology, such as that which
makes manages and managed from the base manage. At the same time, it may
include things that are only marginally considered to be morphology, or that may
be excluded from morphology. (L. Bauer 2006:632, emphasis in the original)

For the present volume, we understand word formation in the wider sense as
mentioned by Bauer. Although the word formation processes in South American
languages described in this volume are basically all of a morphological nature,
this wider approach allows us to include what may be called borderline cases
of morphology, such as e.g. clitic compounding in Kalapalo (see Basso, this volume), adjectival incorporation in Ese Ejja, resulting in attributive predicate clauses
(see Vuillermet, this volume), or nominal compounding in Mapudungun, where
semantics can influence the headedness of compounds (see Ziga, this volume).
That is, although morphology is certainly one pivot of word formation, it is not
the only one and we find it important to stress that words can be created in so
many different ways. In other approaches to the topic, the term word formation
is used synonymously to derivational morphology (see e.g. tekauer et al. 2012:1,
footnote 1), in particular for the traditional contrast to inflectional morphology. Since we follow the broader definition by Bauer quoted above, inflectional
processes are not in the focus of our approach to word formation. Furthermore,
South American languages tend to be less inflectional than derivational (see e.g.
Adelaar with Muysken 2004:467, on Choln nominalizations; see also AlexanderBakkerus, this volume). However, while this distinction may not be so relevant for

Swintha Danielsen, Katja Hann, & Fernando Ziga

the languages of this volume, contributions show that it is still worthwhile a closer
look (see Admiraal & Danielsen; see also Tacconi both this volume).
The definition of what is a word, which is also necessary when considering
word formation, is likewise not uncontroversial (see e.g. Dixon & Aikhenvald
2002; Haspelmath 2011; Schiering et al. 2010). A word is defined by phonological,
morphological and syntactic criteria. It is expected to be a phonological unit and
subject to morphological processes as one unit as well. Syntactically, we can note
its independent distribution throughout clauses. In the following, we want to give
a few examples in order to demonstrate that the definition is not as straightforward as it may be wished for.
Possibly one of the best known phenomena that illustrate the problematic
nature of the concept word are clitics. They are usually considered a challenge
to an integrated approach to wordhood that considers phonological, morphological as well as syntactic criteria, because, as Marantz (1988:253) puts it, they are
[] an independent syntactic constituent which shows up phonologically as
part of a derived word. This ambivalent status of clitics is clearly illustrated in the
paper by Basso on clitic compounding in Kalapalo (this volume). Usually, compounding is defined as the joining of two independent lexemes (see e.g. L. Bauer
2003:40). However, Basso illustrates that compounding is by no means restricted
to lexemes, but can occur with clitics, too. These compounded clitics take over
important functions in clause chaining and interclausal reference marking. It is
in particular these functions that place Kalapalo clitic compounds closer to words
than to affixes. However, while the compounded clitics are rather word-like in
their syntactic behaviour, phonologically they are dependent on a host. Moreover,
most of the Kalapalo clitics cannot occur in unbound form and have to follow a
template like affixes. The paper by Basso thus demonstrates that the boundaries of
wordhood are not always clearly defined.
Another example is provided by Tacconi in her paper on compounding in
Mak (this volume). There, the author mentions the form as child, which is used
to express diminutive notions. What is remarkable about the form as is that it can
occur as a free form, forming one of the components of a compound, which then
often denotes animal offspring, while in other cases it is a diminutive suffix. This
then can be analyzed as a borderline case between compounding and derivation.
This ambivalent use of as child shows that wordhood is at times an unstable concept, particularly in grammaticalization processes.
A final example concerns the paper by Ziga on nominal compounds in
Mapudungun (this volume). Referring to recent literature on Mapudungun
multi-stem nominal constructions, Ziga critically discusses the different
approaches to this topic, which analyze these constructions as either different
types of compounds or as complex noun phrases. Although the traditional

Introduction and acknowledgments

parameters for defining wordhood (see above), in particular phonological criteria, are applied to these multi-stem nominal constructions, the ongoing discussion and the critical re-evaluation by Ziga demonstrate that what is one
complex word and what are two words in a syntactic relationship is not so readily answered and can be analyzed from quite different angles.
It is beyond the scope of this volume to solve all the problems encountered
when dealing with wordhood or to discuss the grammaticalization paths of free
forms into bound forms, some of which may then become inflectional. However, the above quoted examples show that word and wordhood are not firmly
defined categories, but rather that we are dealing with a continuum of bound and
free forms, with frequent overlaps in between. This also entails that the definition of word and word formation process, including also the meanings that are
expressed through the creation of new lexemes, is idiosyncratic for every language.
It is in particular these idiosyncrasies that justify, from our point of view, a broad
approach to the topic of word formation.
It may be related to exactly this fuzziness of word formation, but so far, only
a handful of studies have addressed the issue of word formation as a whole, i.e.
without focusing on a particular language, language family, or a specific process of
word formation. If we leave aside classical works, such as e.g. Sapir (1921) and turn
our attention to more recent studies, Aronoff (1976) is certainly among the earlier
attempts to tackle the field of word formation, applying a generative approach.
In another publication, Fisiak (1985) is concerned with historical word formation processes, but also stresses the central role played by the distinction between
inflection and derivation, which is also outlined by Bybee (1985). The latter topic
has repeatedly been picked up in subsequent works, e.g. by ten Hacken (1994),
among many others. The various threads of word formation were brought together
in a descriptive and, as far as that is possible, theory-free overview by the work
of Anderson (1985) on Typological distinctions in word formation. In the second
edition of this work, Aikhenvald (2007) contributed the chapter on this topic. That
word formation remains a field of interest within linguistics is evidenced by the
recent work of Olsen (2010), which brings together papers that approach word
formation mainly from a theory-driven perspective, concentrating on compounding, but also dealing with nominalizations.
However, the last-mentioned literature on word formation demonstrates a
certain tendency in the research, namely a bias to focus on a particular type of
word formation. The topic of compounding has probably attracted the greatest
attention, as can be seen by numerous publications, often focusing on a particular
(and mostly major Indo-European) language. Examples abound: tekauer (2000)
on English word formation as well as L. Bauer (2004) and Plag (2003) on similar
topics, Donalies (2005) on German, Mengel (2002) on word formation in Slavic

Swintha Danielsen, Katja Hann, & Fernando Ziga

languages, and Becker (2003) on Spanish, to name just a few recent works. Compounding is so far also the only type of word formation to have merited a volume
of the Oxford Handbook series, edited by Lieber & tekauer in 2009. The first part
of the book provides a broad overview of the several aspects of compounding by
approaching the topic from different theoretical angles and also by dealing with
compounds in several subfields of linguistics. The second part of the volume is
dedicated to a typology of compounds. Although less well described languages,
such as Slave (Athapaskan) by Rice and even extinct Maipure-Yavitero (Arawakan)
by Zamponi are included, major Indo-European languages prevail.
Of course, other types of word formation are also discussed in the literature.
Well-known examples are Mithuns paper on noun incorporation (1984) as well as
the works of Moravcsik (1978), Marantz (1982), and Hurch (2005), among others,
on reduplication.
Although this brief sketch can by no means be exhaustive, what transpires
is that there is a certain bias towards centre on compounding in Indo-European
languages. Apparently, this was also felt by tekauer & Lieber 2005 and especially
in tekauer et al. (2012), who address word formation from an explicitly crosslinguistic angle in Word Formation in the Worlds Languages. As the authors rightly
point out, it is [a] pioneering book [that] fills a gap in cross-linguistic research by
being the first systematic survey of the word-formation of the worlds languages
(back cover blurb), following a theory-neutral approach to the topic (tekauer
et al. 2012:5). Certainly one motivation to dedicate an entire volume to word formation is also the editors observation that cross-linguistic research has focussed
much more on inflection than on derivation (see tekauer et al. 2012:1). This
has possibly to do with the perception of inflectional categories as more regular
and thus easier to approach in a systematic way (see Bickel & Nichols 2007:171).
Moreover, inflection is a prominent feature of Indo-European languages, which
has probably added to a certain emphasis on it. After outlining the field of word
formation, tekauer et al. proceed to an in-depth study of the various morphosyntactic means used in word formation. These include, of course, compounding and reduplication, but also comprise features that have so far received less
attention in the literature, such as e.g. word formation by tone and/or pitch and
back-formation. The data are based on questionnaires and the authors specifically
include small languages like Udihe (Altaic) (tekauer et al. 2012:11). However,
as the authors state, sufficient and satisfactory data for cross-linguistic studies of
word formation in smaller and in the literature under-represented languages are
often still lacking (tekauer et al. 2012:1ff, 7f).
We aim to contribute to filling this gap by providing a typological and
descriptive overview of several word formation processes in a number of native

Introduction and acknowledgments

South American languages which, so far, have not been in the focus of research
on word formation. In contrast to the book by tekauer et al. we do not intend
to address each possible type of word formation here, but would rather like to
highlight some and until recently mostly undescribed native South American
languages and their particular word formation processes. These may be common
ones, which are also found in other and major languages, such as e.g. nominal
compounding or reduplication (see the contributions by Admiraal & D
and van Gijn), but also less frequent and language-specific processes, such as
clitic compounding (see the contribution by Basso). Another difference to the
above mentioned work by tekauer et al. is that the data of the present volume
stem from fieldwork conducted by the author(s) of a given paper. That is, the
last ten years or so saw great efforts being made in describing smaller andoften
endangered indigenous languages of South America, resulting in a considerable
number of grammars on these languages (see e.g. Epps 2008 on Hup; Guillaume
2008 on Cavinea; Haude 2006 on Movima; Rose 2011 on Emerillon). It is now
possible to progress to more specific investigations with the data of these native
South American languages. From a long-term perspective, our single-language
approach in this volume is also meant to contribute tofurther c ross-linguistic
research on word formation phenomena in South American languages and
The idea to publish a volume dedicated to word formation in indigenous languages of South America developed on a workshop held in June 2011 at Leipzig
University. This volume partly includes articles on the presented languages of this
workshop, and in addition we motivated interested colleagues to contribute to this
study with their findings. In the following, we summarize each chapter with a special focus on the respective processes described.
1. Ziga: Nominal compounds in Mapudungun
In his paper, Ziga is concerned with nominal compounds in Mapudungun
(Chile). The author bases his findings on previous literature (Smeets 2008; Baker&
Fasola 2009) as well as his own fieldwork data, critically discussing previous
attempts to classify multi-stem nominal expressions. These can be head-initial and
head-final. Accordingly, features used to categorize these expressions are headedness, phonological properties, and the semantic relations between the elements of
the nominal expression. Ziga proposes a re-classification of the Mapudungun
multi-stem nominal expressions, postulating that non-morphological features
such as semantics can interfere with headedness.

Swintha Danielsen, Katja Hann, & Fernando Ziga

2. T
 acconi: Towards a characterization of compounding in Mak
The article on compounding in the Mak language of the Paraguayan Chaco
elaborates on different nominal compounding types and gives details about the
typical grammaticalization path from compound root to derivational element.
Compounds are also analyzed morphologically and semantically. Interestingly,
the compounds often include possessive markers and have some phrase-like characteristics. The findings are also compared to other genetically related Chaco languages, such as Wich and Chorote.
3. Cneo: Augmentative in Toba (Guaycuruan): Form and function
The article on Toba, also a Chaco language, describes derivational and compounding morphology used to construct augmentatives. There appears not to be one
single augmentative suffix, but a variety of alternative augmentative-encoding
forms that can be considered to be lexical roots. Thus, augmentative constructions
in Toba resemble nominal compounding. The functions of augmentative compounds comprise the expected intensification and affection, but they also have
some pragmatic-discursive functions and can be used pejoratively, for instance.
4. A
 dmiraal and Danielsen: Productive compounding in Baure
The Baure language has very productive compounding strategies, including the
joining of nominal and verbal roots, as well as of classifiers. This article describes
one source of morphological complexity, namely lexical root compounding that
creates complex lexical bases. The described word formation process is not simply
derivational, but seems to be on the edge towards inflectional, since it is a productive and still transparent strategy of adjusting words to their environment.
5. Vuillermet: Two types of incorporation in Ese Ejja (Takanan)
Ese Ejja has very productive incorporation of lexical roots, and it can be classified
as belonging to two types. One is the typical Amazonian kind of argument incorporation into verbs, which forms complex verb stems. The second type involves
the incorporation of lexical roots into adjectival predicates, including verbal stems

Introduction and acknowledgments

as incorporated elements. This word formation process is a common strategy in

this language to create attributive predicate clauses.
6. van Gijn: Reduplication in Yurakar (language isolate)
In his contribution, van Gijn deals with reduplication in Yurakar (central Bolivia).
The author distinguishes between complete reduplication of a root, partial prefixed reduplication, and partial suffixed reduplication. While the first is used to
form property-denoting words, the partial reduplication types express mainly
pluractionality, causation, and intensity. Furthermore, van Gijn shows that each
Yurakar reduplication type can be accounted for by the Iconicity Principle.
7. Hann: Reduplication strategies in Kallawaya (mixed language)
Another contribution on reduplication is provided by Hann, who discusses reduplication strategies in Kallawaya (Bolivia). In her paper, the author addresses features that are less typical of Andean languages, such as semantic lexicalizations
and the expression of associative qualities. The author concludes that semantic
lexicalizations and the expression of associative qualities are used to enlarge the
otherwise restricted Kallawaya lexicon.
8. Basso: Compounding in Kalapalo, a Southern Cariban language
The paper by Basso is concerned with clitic compounding in Kalapalo (Brazil).
Kalapalo has compounding that includes extended verb stems, but what the
author calls clitic clustering is much more common. Basso distinguishes four different semantic sets of clitics, and clustering can take place either within one set or
across different sets. The most important function of clitic clustering is to produce
clausal cohesion, as evidenced by interclausal reference marking and taxis morpheme clustering.
9. Alexander-Bakkerus: Nominalization in Choln (Cholonan)
Choln has a variety of nominalizers that are employed for subordinate clauses
of the following kinds: relative clauses, complements, subject and object clauses,
causal and temporal clauses. In addition to this, the nominalizing suffixes can be

Swintha Danielsen, Katja Hann, & Fernando Ziga

combined with case-markers and specify the clause types further through this,
producing a variety of adverbial clauses. This article shows that a common word
formation process like nominalization can have syntactic functions rather than
just contributing to expand the lexicon.



Ese Ejja



Cavinea Baure
North LaPaz Quechua


Central Yurakar Paunaka










600 km

languages analyzed in detail in this volume

languages mentioned in passing in this volume







Map 1. Indication of South American languages analyzed in this volume and in earlier


Introduction and acknowledgments

10. Acknowledgments
We would like to thank our authors, not only for taking part in the workshop, but
particularly for putting forward and patiently revising their contributions to this volume. We are furthermore grateful to Patricia Leplae from John Benjamins Publishing
Company for her support and patience in editing and finally producing this volume.
Our thanks also go to Michael Dunn from the Max Planck Institute for Psycholinguistics, Nijmegen, for producing the map for us. We are furthermore indebted to
Lucy Zuberbhler for proofreading our contributions. We would also like to thank
the student assistants at the University of Leipzig: Susann Schildhauer for practical
support all around the workshop and Lena Sell for correction and editing of articles.
The University of Leipzig provided the facilities for the workshop, for which we are
likewise grateful. Last, but not least, we would like to express our sincerest thanks to
our native language consultants, without whose support and enthusiasm we could
have never conducted the research that was necessary to realize this volume.
Individual support for the contributors is acknowledged in the respective

Adelaar, Willem with Muysken, Pieter. 2004. The Languages of the Andes. Cambridge: CUP.
DOI: 10.1017/cbo9780511486852.001
Aikhenvald, Alexandra Y. 2007. Typological distinctions in word-formation. In Shopen (ed),
Vol. 3, 165. DOI: 10.1017/cbo9780511618437.001
Anderson, Stephen R. 1985. Typological distinctions in word formation. In Shopen (ed), 156.
Aronoff, Mark. 1976. Word Formation in Generative Grammar. Cambridge MA: The MIT Press.
Baker, Mark & Fasola, Carlos. 2009. Araucanian: Mapudungun. In Lieber & tekauer (eds),
Bauer, Laurie. 2003. Introducing Linguistic Morphology, 2nd edn. Edinburgh: EUP.
Bauer, Laurie. 2004. English Word-Formation. Reprint. Cambridge: CUP.
Bauer, Laurie. 2006. Word formation. In Encyclopedia of Language and Linguistics, Keith Brown
(ed), 632633. Oxford: Elsevier.
978008044854 DOI: 10.1017/cbo9780511486807.011
Becker, Martin. 2003. Die Entwicklung der modernen Wortbildung im Spanischen: der politischsoziale Wortschatz seit 1896. Frankfurt: Peter Lang.
Bickel, Balthasar & Nichols, Johanna. 2007. Inflectional morphology. In Shopen (ed), Vol. 3,
169240. DOI: 10.1007/s11525-007-9117-7
Bybee, Joan L. 1985. Morphology: A Study of the Relation between Meaning and Form [Typological Studies in Language 9]. Amsterdam: John Benjamins.
Dixon, Robert M. W. & Aikhenvald, Alexandra Y. 2002. Word: A Cross- linguistic Typology.
Cambridge: CUP.
Donalies, Elke. 2005. Die Wortbildung des Deutschen: Ein berblick. Tbingen: Narr. DOI:

Swintha Danielsen, Katja Hann, & Fernando Ziga

Epps, Patience. 2008. A Grammar of Hup. Berlin: Mouton de Gruyter. DOI: 10.1515/9783110199079
Fisiak, Jacek. 1985. Historical Semantics, Historical Word Formation. Berlin: Mouton de Gruyter.
DOI: 10.1515/9783110850178
Guillaume, Antoine. 2008. A Grammar of Caviena. Berlin: Mouton de Gruyter.
Haspelmath, Martin. 2011. The indeterminacy of word segmentation and the nature of morphology and syntax. Folia Linguistica 45(1): 3180. DOI: 10.1515/flin.2011.002
Haude, Katharina. 2006. A Grammar of Movima. PhD dissertation, Radboud University
Hurch, Bernhard (ed). 2005. Studies on Reduplication [Empirical Approaches to Language
Typology 28]. Berlin: Mouton de Gruyter.
Lieber, Rochelle & tekauer, Pavol. 2009. The Oxford Handbook of Compounding. Oxford: OUP.
Marantz, Alec. 1982. Re reduplication. Linguistic Inquiry 13(3): 435482.
Marantz, Alec. 1988. Clitics, morphological merger and the mapping to phonological structure. In Theoretical Morphology. Approaches in Modern Linguistics, Michael Hammond &
Michael P. Noonan (eds), 253270. San Diego CA: Academic Press.
Mengel, Swetlana. 2002. Slavische Wortbildung: Semantik und Kombinatorik. Materialien der 5.
Internationalen Konferenz der Kommission fr Slavische Wortbildung beim Internationalen
Slavistenkommitee; Lutherstadt Wittenberg, 2025 September 2001. Mnster: Lit.
Mithun, Marianne. 1984. The evolution of noun incorporation. Language 60(4): 847894. DOI:
Moravcsik, Edith A. 1978. Reduplicative constructions. In Universals of Human Language,
Vol.3: Word Structure, Joseph H. Greenberg, Charles A. Ferguson & Edith A. Moravcsik
(eds), 297334. Stanford CA: Stanford University Press.
Olsen, Susan. 2010. New Impulses in Word Formation. Hamburg: Buske.
Plag, Ingo. 2003. Word-Formation in English. Cambridge: CUP. DOI: 10.1017/cbo9780511841323
Rice, Keren. 2009. Athapaskan: Slave. In Lieber & tekauer (eds), 542563.
Rose, Franoise. 2011. Grammaire de lmrillon teko, une langue tupi-guarani de Guayane.
Louvain: Peeters.
Sapir, Edward. 1921. Language. An Introduction to the Study of Speech. New York NY: Harcourt,
Brace & World. DOI: 10.1111/modl.1921.6.issue-3
Schiering, Ren, Hildebrandt, Kristine A. & Bickel, Balthasar. 2010. The prosodic word
is not universal, but emergent. Journal of Linguistics 46(3): 657709. DOI: 10.1017/
Shopen, Timothy. 2007. Language Typology and Syntactic Description, 2nd edn, Vol. 13.
Cambridge: CUP.
Smeets, Ineke. 2008. A Grammar of Mapuche. Berlin: Mouton de Gruyter.
tekauer, Pavol. 2000. English Word-Formation. A History of Research (19601995). Tbingen:
tekauer, Pavol & Lieber, Rochelle. 2005. Handbook of Word-Formation. Dordrecht: Springer.
DOI: 10.1007/1-4020-3596-9
tekauer, Pavol, Valera, Salvador & Krtvlyessy, Lvia. 2012. Word-Formation in the Worlds
Languages. A Typological Survey. Cambridge: CUP.
ten Hacken, Pius. 1994. Defining Morphology: A Principled Approach to Determining the Boundaries of Compounding, Derivation, and Inflection. Hildesheim: Olms.
Zamponi, Raoul. 2009. Arawakan: Maipure-Yavitero. In Lieber & tekauer (eds), 584593.

Nominal compounds in Mapudungun

Fernando Ziga
University of Bern*

It is perhaps unsurprising that the rich agglutinative-polysynthetic verb

morphology of Mapudungun has drawn most attention in linguistic studies. So
far unnoticed in the literature are Mapudungun complex noun phrases, which
show a puzzling distribution in terms of the internal structure they display.
Somecomplex NPs are head-final (mapu-che people of the land). Others
are head-initial, and of these a subset appears to be less lexicalized. In some
cases, all three possibilities are found with the same components: mamll-che
wood people, che-mamll people made of wood, and che mamll wood-like
people. The present paper reviews the comparatively modest literature on these
phenomena, deals with them in an account based on semantic factors, and places
them in a broader typological context.
Keywords: Mapudungun; nominal compound; head-final; head-initial, nonhead;
complex noun phrases; modification; subordination

1. Introduction
Mapudungun is a linguistic isolate spoken by approximately 250,000 people
in south-central Chile and Argentina. Excepting the southernmost Huilliche
variety reportedly the most divergent one , the dialects of the language share
most morphological and syntactic features and differ basically as to comparatively few phonological and lexical features. The variety addressed here is Central Mapudungun, originally spoken in the Araucana region in Chile.
Mapudungun phonology is relatively simple, but its agglutinating morphology is rather complex, especially so in the verbal domain, with respect to which
the term polysynthetic is appropriate both as an impressionistic label and as

* I am indebted to Swintha Danielsen and Katja Hann for their numerous comments on
a previous version of this article, as well as to the Swiss National Science Foundation for
supporting this spin-off research (through EuroBABEL-related grant no. 10BA13-125811).

Fernando Ziga

a technical description. According to the traditional accounts (Augusta 1903;

Harmelink 1996; Salas 2006; to some extent, also Smeets 2008), finite verbs can
mark tense, aspect, modality, evidentiality, spatial direction, and voice, as well as
person and number of up to two arguments; transitive verbs can also incorporate
whole NPs into the verbal complex. Some recent accounts of Mapudungun morphosyntax have also argued in favour of postulating the category of direction/
inversion (Arnold 1996; Ziga 2001, 2006). By contrast, the nominal domain
is rather unspectacular and has received less attention in the literature. A systematic analysis of one particular facet of nominal morphology, viz. nominal
compounding, is the object of the present study. I have based the presentation
and the analytical claims not only on the previous literature but also on my own
fieldwork conducted with native speakers in January 2011 in the Cautn province
in south-central Chile. Data presented here come from my notes unless otherwise
Section 2 introduces the basics of Mapudungun noun phrase structure.
Section 3 is devoted to the issue of compounding and, more generally, complex
NPs (i.e. NPs that may consist of one or more words but consist of more than one
root) in contexts other than possession. I will claim here that there are two distinct
kinds of nominal compounds in the language, as well as three different types of
multi-stem complex NPs thereby following the spirit, but not the letter, of two
previous studies. More importantly, the data unmistakably show that there is a
systematic correlation between the semantic relationship between the constitutive nouns and their relative order in the complex expression. Section 4 closes the
paper with some general and typological remarks.
2. Basics of Mapudungun NP structure
The structure of lexical NPs in Mapudungun is schematically represented in (1):1

(dem) (art) (possessive) (pl/numeral) (adj) N

Lexical NPs can consist of merely a noun, which is minimally a root, e.g. wentru
man, kura stone, and ko water. Plurimorphemic monoradical nouns can be
derived via suffixation of elements like -ntu, -we, and -wen, e.g. kura-ntu stony
land, milla-we place where there is gold, form milla gold, and fotm-wen father

1. Pronominal NPs consist of a demonstrative or personal pronoun and will not be a ddressed
any further here.

Nominal compounds in Mapudungun

and son, from fotm son (of man). Roots can cooccur with demonstratives, e.g.
tfachi wentru this man and feychi kura that stone.2 An attributive adjective
immediately precedes its head noun: kme wentru good man, pichi kura small
stone.3 Plural number is obligatorily marked on humans/animates by adding pu
(e.g. pu wentru men) or by suffixing -ke to the adjective (e.g. motri waka fat cow
vs. motri-ke waka fat cows).4 Count roots can appear with numerals, e.g. epu
wentru two men and meli kura four stones. In addition, there are other quantifiers, like pichi-n (nonfinite form of pichi- (be) little, small), e.g. pichin mapu a
little, some land, fill every and kom all, among others; these occupy the slot of
the article.
Salas (2006:85) notes that there are two kinds of articles, viz. indefinite kie
one, a and definite chi; the latter can be reinforced with the empty particle ta
and/or the demonstrative fey.5 Thus, there is a three-way opposition in Salass view
between, e.g. mansun (an) ox, kie mansun one/ an ox, and chi mansun the ox,
and the latter NP could also appear as ta chi mansun, fey chi mansun, or even fey ta
chi mansun, apparently without any clear difference in meaning.
The situation is more complicated than this, however; there is also an element
ti that stands in a somewhat elusive kind of opposition to chi, since numerous
speakers use them interchangeably in some cases (e.g. when producing translational equivalents of Spanish NPs including the definite article el/la without
a context) but as clearly distinct in others (e.g. in specific occurrences in narrative texts).6 Smeets (2006:86) labels ta and ti anaphoric pronouns and describes
their yield based on the minimal pair shown in (2). In her view, the reference of
2. The demonstratives are proximal tfa, medial (t)fey, distal tie, and remote distal (t)ye; in
order to modify nouns, they appear attributivized with chi.
3. Adjectives can be compounded with nouns as well, as in German kleines Kind little child
vs. Kleinkind infant, toddler, e.g. weche youngster (cf. we young, new and che person).
4. I have treated pu as a nominal plural marker and ke as an adjectival nonsingular marker
(i.e. covering both dual and plural) in other studies (Ziga 2001, 2006). Smeets (2008:72,
111f) proposes a different analysis, according to which pu is a collective marker and ke signifies distributive number.
5. Salas (2006) gives the form ta for this element throughout (i.e. with interdental t instead
of dento-alveolar t). I follow other sources and my own field notes here in identifying this
morpheme as ta.
6. Note that the articles are indefinite kie and definite ti in Harmelinks (1996:51) account.
The latter author explicitly says that the functions of ti go beyond marking definiteness, and
he notes that it stands in opposition to ta, tati, nga, and ngati. Cf. also the main body of text for
more on such multimorphemic elements.

Fernando Ziga

the doctor is said to be determined contextually in (2)a, while it is determined

by shared knowledge in (2)b. Observe that the demonstrative fey is pronominal
here and means s/he, i.e. it is NP-external, and therefore has no attributivizer
-chi, since the whole utterance is a predication rather than merely a referential



Fey ta
dem art1 doctor
He is the doctor (the one we talked about).


Fey ti
dem art2 doctor
He is the (generally known profession of) doctor.

Furthermore, ta and ti ~ chi (Smeets treats the latter two as alternants) often
cooccur in the compound tati. (Smeets says that in such cases reference is determined by resorting to the text and shared knowledge simultaneously.) Some
speakers regularly use chi and tachi (rather than ta), and tichi is arguably rare
but attested. Despite Smeetss detailed exposition of how these elements work
(2006:8695), I think more work is needed on these challenging phenomena
in order to settle matters. I will limit myself here to noting the existence of a
number of article-like elements that may occur in the corresponding (complex)
slot in (1) above, whose exact relation to definiteness and referentiality is not
entirely clear yet.
Attributive possession is expressed by means of possessive elements
transparently related to personal pronouns that occupy the position between
the articles and the plural/numerals in (1) above, and seem to preclude the
appearance of demonstratives.7 These possessive elements often occur with
the element ta, but the conditions under which this is obligatory have not
been studied in detail yet. Lexical or pronominal NPs making the reference of
the possessor explicit or particularly prominent precede the expression of the
possessum. Note that the 1st person singular possessive and the 3rd person
possessive are homophonous:



(ta -)mi
pu che
art-2sg.psr pl person
your (sg) people

7. Possessed NPs are not only found with nominal roots but also with most nonfinite verb
forms, which mark their subjects via verb-external possessive markers instead of suffixed
person markers. The study of possession in the context of nonfinite verb forms lies outside the
scope of the present study.

Nominal compounds in Mapudungun


Juan i
Juan 3sg.psr house
Juans house


eymu ta-mu
2du art-2du.psr mother
your (du) mother


iche i
chaw i
pu kme weny
1sg 1sg.psr father 3.psr pl good friend
my fathers good friends

(Smeets 2008:133)8

3. Nominal compounds
A note on orthography is in order at this point: Mapudungun is only seldom
written and there are several different conventions used by writers, linguists, and
anthropologists on both sides of the Andes; none of them is universally used.
These conventions reflect regional differences in pronunciation as well, and most
debates centre around how to represent particular phonemes. The important
consequence of such a situation for our present purposes is the fact that there is
significant variation with respect to how people write compounds: some are usually written as one simple word (e.g. mapuche Mapuche), while others appear as
separate words (e.g. longko kachilla ear of wheat). In Sections 3.1 and 3.2, which
deal with the basics of nominal compounds and with some interesting problems
respectively, I will retain the one-word and two-word spellings used in the sources.
I will return to the issue of wordhood in Section 3.3, which centres on the question
of appositional constructions.
3.1 The basics of nominal compounding
Harmelinks (1996:49) brief presentation may be used to introduce the basics of
nominal compounding in Mapudungun: complex expressions consisting of at least
two nouns are either head-final (4) or head-initial (5). While in the former group
the first element characterizes the second in an unspecified way, the latter group
is more restricted in semantic terms, viz. it consists of part-whole relationships

8. I have minimally adapted the orthography in examples from other sources in order to
be consistent with the rest of the study (i.e. by underlining interdental consonants where my
informants pronounce them as such).

Fernando Ziga

(5)ac or of those in which the former noun is an element of the latter (my
translation; (5)d):


Head-final nominal compounds

(Harmelink 1996:49)
a. kna ruka (straw house) house made of straw9
b. rme ruka (rush house) house made of rush10


Head-initial nominal compounds

a. kal ufisa
(wool sheep)
b. mta mansun (horn ox)
c. tapl koyam (leaf roble.beech)11
d. ilo kawellu
(meat horse)

(Harmelink 1996:49)
sheeps wool
oxs horn
roble beech leaf
horse meat

Harmelink also mentions the contrast between two possible expressions built
upon che person and mamll wood: che mamll wooden statue refers to a piece
of wood carved into human form, whereas mamll che may signify a person
made of wood or a person which, for some reason, is defined by wood (1996:49,
my translation). In his account, both expressions are head-final compounds and
belong to the series illustrated in (4) above.
Baker and Fasolas (2009) analysis of Mapudungun compounding accounts
for different kinds of noun-noun compounds and verb-noun compounds. In both
cases there are head-final and head-initial structures; head-final verb-noun compounds (6) correspond to the series in (4) and are denominal verbs formed via
compounding with the verbal root tu- take (in the alternative analysis favoured
by Baker and Fasola, the common and productive verbalizing affix -tu) while

9. Casa hecha de totora in the original. Mapudungun trome is normally used for Spanish
totora (cf. Quechua tutura) when representing Typha domingensis or Scirpus californicus
(Ziga 2006:333); trome in Augusta (1916:236) is totora (Cyperus vegetus; a species of
sedge that superficially resembles grass or rush). Augustas (1916:105) kna reads cualquiera [sic] paja con que se techa la casa (Hernndez and Ramos 2005 give paja, un tipo de
totora for kna); in addition, there are ichu thick straw, pinu thin straw, and kolm corn/
wheat siftings.
10. Casa hecha de junquillo in the original. Standard Spanish junco means rush, reed;
its derivative junquillo is sometimes found in dictionaries as representing English jonquil
(Narcissus jonquilla; an old world daffodil), but also English rattan (any subspecies of the tribe
Calameae; palms). In Augusta, rme (1916:204) is Juncus procerus (Meyer); different speakers
appear to use the term with varying degrees of specificity.
11. Koyam is Mapudungun for the Nothofagus obliqua tree in its adult state. In Spanish, the
word roble is normally used (in the northern hemisphere, trees belonging to the genus Quercus
are called robles), but the loanwords coyn and hualle are found as well (cf. Mapudungun walle,
the name of the tree in its young state). Occasionally, roble chileno or roble pelln are also used
in Southern Cone Spanish for koyam.

Nominal compounds in Mapudungun

head-initial verb noun-compounds (7) correspond to the series in (5) and are
instances of noun (phrase) incorporation:12


Head-final verb-noun compounds

a. kofke-tu-n
b. kitra-tu-n
c. tralka-tu-n (gun-tu-inf)
d. mamll-tu-n (wood-tu-inf)

(Baker & Fasola 2009:597)

eat bread
smoke a pipe
shoot a gun
fetch wood


Head-initial verb-noun compounds

a. kintu-waka-n (seek-cow-inf)
b. ngilla-kofke-n (buy-bread-inf)
c. nentu-ant-n (take.out-day-inf)
d. pto-ko-n

(Baker & Fasola 2009:595)

look for (the) cows
buy (the) bread
fix a date
drink water

According to Baker & Fasolas account, head-final compounds originate in the

lexicon while head-initial ones are created by syntactic rules. In the former case,
the nonhead element is interpreted as a modifier with an unspecified semantic
relationship to the head while in the latter, the nonhead is interpreted as an argument that bears a thematic relation to the head (2009:600). In this view, headinitial nominal compounds might be considered the first known cases of noun
incorporation into a noun something of particular theoretical and typological
interest for these authors in the light of particular claims made in Baker (1988).
Let us take a look at most of Baker and Fasolas examples for both kinds of
nominal compounds, listed in (8) and (9):13


Head-final nominal compounds

a. mapu-che
b. mapu-dungun (land-speech)
c. ilo-kor
d. mamll-wangku (wood-chair)
e. kna-ruka
f. wingka-kofke
g. kutran-che
h. wariya-che
i. pulku-fotilla

(Baker & Fasola 2009:598)

Mapuche person
the language of the Mapuche
soup containing meat
wooden chair
straw house
European-style bread
sick person
person living in a town/city
wine bottle (bottle type)

12. The reader is referred to Harmelink (1992), Golluscio (1997), and Baker et al. (2005) for
more on nominal incorporation in Mapudungun.
13. I have slightly adapted the translations and even some glosses (e.g. che is sometimes
glossed as people and sometimes as person in Baker and Fasola 2009, but mapuche, wariyache
and the like can be used in the singular as well); kna-ruka is erroneously given as birdhouse and birdhouse in the original cf. (4)a. Mp wing and mollf blood appear as lpi
and molif respectively in the original.

Fernando Ziga


Head-initial nominal compounds

a. nge-trewa
b. saku-kachilla
c. longko-waka
d. mp-achawall (wing-chicken)
e. longko-kachilla (head-wheat)
f. namun-mesa
g. ilo-trewa
h. lichi-waka
i. mollf-che
j. fotilla-pulku

(Baker & Fasola 2009:598)

dogs eye
bag of wheat
cows head
chicken wing
head/ear of wheat
table leg
dog meat
cows milk
human blood
bottle of wine
(bottle containing wine)

Baker & Fasola (2009:599) say that head-initial compounds involve body parts or
some other kind of part-whole relationship, while others involve a relationship
between a container and a substance (9)j or between a substance and an entity
extracted from it (9)gi. They point out that there are revealing minimal pairs like
(8)i/(9)j: head-final pulku-fotilla wine bottle illustrate what they see as a modificational relationship while head-initial fotilla-pulku bottle of wine is an instance
of argumental relationship.
The head-initial group does indeed include numerous body part expressions;
see the additional ones listed in (10). Excepting fid-pilun (?-ear) cerumen, earwax,
both the first and the second noun are found as simplex items as well, and the
compounds meaning differs from those of its constituting elements:


Head-initial nominal compounds

a. chll-kewn (brooch-tongue)
b. chll-kw
c. chll mollf (brooch blood)
d. chll-ponon (brooch-lung)
e. kal longko
(wool head)
f. kewn kutri (tongue vagina)
g. longko moyo (head breast)
h. longko pnn (head penis)
i. mnul-longko (wrapping-head)
j. ptra-kw (stomach-hand)
k. ptra-namun (stomach-foot)
l. relmu-nge

(Hernndez & Ramos 2005)14

frenulum linguae
eponychium, loose cuticle
hair of the head
glans penis
head kerchief
palm of the hand
sole of the foot

14. Some of these words appear in the original written with an orthography suggesting a
different pronunciation: chillkewn, chulkw, and munulongko. Such variation (i ~ u ~ and
l ~ l ~ ll) is frequent in the language (Salas 1992: Chapter3).

Nominal compounds in Mapudungun

Observe that, even though the semantic relationship between the first and second
elements in (10)ad is possibly construed as part-whole, this is unlikely to be the
case in (10)i. Mnu refers to a wool ball and mnul means wrapping. If the correct analysis of this compound is wrapping for the head, the rather unexpected
expression kuykuy-dew (bridge-mouse) mouse bridge should be analyzed along
the same lines, i.e. as bridge for (the) mice.
A number of body part expressions are different from those in (10) in that the
first element can appear either on its own, as unmarked item, or with a qualifying
second element if needed. Changll in (11)a/ b is parallel to Spanish dedo finger,
toe.15 Chuu means elbow with or without kw hand as second element (11)c,
just as mellf means lip with or without wn mouth (11)d.16 Lastly, wili means
nail, claw and can be further specified, as in wili-namun toenail (11)e:


Head-initial nominal compounds

a. changll-kw (finger-hand)
b. changll-namun (finger-foot)
c. chuu kw
(elbow hand)
d. mellf-wn
e. wili-namun


(Hernndez & Ramos 2005)

3.2 Beyond the basics

This subsection addresses some details usually neglected in the literature, their
relevance for the account of compounding presented in Section 3.1, and Smeetss
(2008) analysis.
3.2.1 Potential quirks
Harmelink (1996) gives two further examples of the head-final group (introduced
in (4) above) that differ from the ones we have seen so far:


Selected rooms of the Mapuche house

(Harmelink 1996:49)
a. umaw-tu-we ruka (sleep-make-nmlz house) sleeping room
b. ktral-tu-we ruka (fire-make-nmlz house) kitchen

15. Chang actually means branch, and can appear with the meaning tree branch either on
its own, with an nonproductive final element as changki, or with mamll wood as second
element, i.e. chang (ki)-mamll. The element ll is also nonproductive.
16. Smeets (2008:117) says that the first element of the compound mellf-wn is unknown.
Even though it is possible that the ascription of the meaning lip is originally derivative (i.e.
mellf could originally have meant something else and came to mean lip only after the compound was in use), all my consultants agree with Augusta (1916) in identifying mellf(-wn)
as signifying lip.

Fernando Ziga

Even though Harmelink does not comment on them, thereby suggesting that they
are essentially similar to kna-ruka straw house and rme-ruka rush house, these
examples are formally more complex. The first noun of the two expressions in (12) is
not a nominal root but a nominalized verb form (consisting of a denominal verb plus
the nomen loci suffix -we already mentioned in Section 1 above). Since their literal
meaning is sleeping place of the house and fire-making place of the house respectively, it is not difficult to see that they are actually head-initial rather than headfinal; the first element corresponds to a part of the whole denoted by the second.
Furthermore, some expressions that look like head-initial nominal compounds include the following expressions referring to astronomical events:


Astronomical events expressions

(Hernndez & Ramos 2005)
a. la-n ant
(die-nfin sun)
solar eclipse
b. la-n kyen (die-nfin moon) lunar eclipse
c. kon-n-ant (enter-nfin-sun) sunset

There are two possible analyses for these expressions. In the first, the nonfinite verb
form is a verbal noun and functions as the head, in which case there would be an
argumental relationship between head and nonhead; the literal meanings would
then be dying/death of the sun, dying/death of the moon, and entering/entrance
of the sun respectively. Alternatively, the nonfinite verb form is a verbal adjective or participle and functions as an attribute of the head, in which case dying
sun, dying moon, and entering sun would be the literal meanings of the above
expressions and Baker & Fasolas generalization would not apply here, because
these would not be nominal compounds at all.
Superficially similar cases are the following:



rangi-ant (middle-day) noon

rangi-pun (middle-night) midnight

The part-whole relationship is what one would expect if these expressions were
head-initial nominal compounds, but there are a couple of details that make a
straightforward answer somewhat difficult in this case. The noun rangi means
centre, middle, half but can also occur as an adjective (central) and a preposition
(in the middle of ); e.g. rangi che can mean either mediocre person or amidst
people, and my consultants consistently rejected the reading the centre/middle
of a person. Moreover, some speakers prefer rangi for the noun in isolation (or
other compounds) and have rangi only as adjective or preposition (as suggested by
Augustas 1916 entries on p. 193).
Another interesting expression in this context is kawi-kura (feast-stone),
which refers to the sound made by small objects placed within the kultrung shamanic drum when they collide against each other; the expression is translated

Nominal compounds in Mapudungun

in the anthropological literature dealing with Mapuche culture as either stone

feast (head-initial) or feasting stones (head-final). The latter analysis, however,
is rather implausible on semantic grounds (kawikura is actually the event and
the associated sound, not the stones involved in it) although it might simply be
a case of metonymic shift. In addition, the correct structure of the construction
should then be kawi-n kura (feast (v.)-nfin stone), which is possible, since the
verb kawi(-tu)-n feast is still in use, but this NP does not refer to the objects
inside the shamanic drum; it is parallel to its English translation in semanticopragmatic terms and requires a rather particular context to be felicitously used. If
the expression is a head-initial nominal compound, on the other hand, its formfunction correspondence is anomalous.
Head-initial expressions including ngen owner introduce an interesting
nuance to the above account of the semantic relationship between head and nonhead. Such NPs range from cases in which the second element refers to a concrete
physical entity (15)ab, through cases in which the nonhead is abstract (15)cd,
to those where the second element refers to a quality or state (15)ef; the latter
expressions are typically found as predications and require the copula/verbalizer
nge- be). Note that the notion of ngen is not limited to ownership sensu stricto but
covers potential/possible control and responsibility of some sort; ngen la (owner
dead) denotes someone in charge of disposing of a deceased person (typically a


Head-initial expressions with ngen(Augusta 1916:55)

a. ngen mapu
(owner land)
b. ngen weshakelu
(owner thing)
owner of the thing
c. ngen kdaw
(owner work)
d. ngen pi-n
(owner say-nfin)
orator (in ceremonies)
e. ngen kutran nge-n (owner sickness be-inf) be frail
f. ngen dumi nge-n (owner darkness be-inf) be dark/gloomy

Especially noteworthy in this context are expressions including the element ngen
that are related to distinctive traits of Mapuche culture, viz. those in which ngen
refers to a parent or parent-like person (16)ac and those in which it refers to
a guardian spiritual being occasionally adopting animal shape in charge of
particular phenomena or features of nature (16)df. Again, even with the former
group, responsibility and power rather than strict blood relationship appears to
be at issue; ngen kulme (owner orphan) denotes someone in charge of an orphan:


Head-initial expressions with ngen(Augusta 1916:55 and p.c.)

a. ngen fotm (owner
(adoptive) father (of a man)
b. ngen awe (owner (adoptive) father (of a woman)
c. ngen pe (owner child.of.woman) mother

Fernando Ziga


ngen ko
(owner water)
water spirit
ngen lawen (owner medicine) (herbal) medicine spirit
ngen m (owner bird)
bird spirit

There is an intriguing quirk related to the form/function correspondence in some

of these expressions. According to Augusta (1916:55), the parent-like items are
pluralized as though ngen were an adjective: people with male children is ngen-ke
fotm instead of the expected *pu ngen fotm. The same applies to ngen-ke awe
and ngen-ke pe, and most of my consultants accept these forms only. By contrast, some consultants use these forms as interpreted by Erize (1960:288), i.e. as
though ngen-ke pe and ngen-ke fotm meant mother of several children and
father of several male children respectively. This latter pattern is unexpected not
only because ngen is formally treated as an adjective but also because the marker
-ke pluralizes the nonhead element even though it appears on the head. Lastly, it
is possible to find the expression ngen-ke fote owners of the ship, which is comparatively very similar to its Spanish formal counterpart dueo-s del bote (ownerpl of.the boat) note that this similarity may actually be contact-induced and
differs from the situations described by Augusta and Erize. My Chilean consultants tend to accept this expression and to interpret it the same way, but this may
be merely an effect of the animacy asymmetry. It goes without saying that more
research is needed here.
3.2.2 Smeets (2008)
Smeetss (2008) analysis differs from Harmelinks (1996) and Baker & Fasolas (2009)
in several respects. Even though she treats head-final expressions like mapu-che
Mapuche and pulku-fotella wine bottle as compounds in which the first noun is in
an attributive relationship to the second, she postulates the existence of three subordinative complex noun phrases, viz. possessive, genitive, and partitive, in order
to analyze other nominal structures consisting of more than one noun (pp. 133f).
Possessive NPs are those including the possessive markers and were introduced in Section 2 above. Genitive NPs are said to be expressions in which the
first element forms part of, belongs to or is connected to the referent of the second one, which is then seen as an NP complement, as in namun mesa (leg table)
table leg, ad mapu (habit land) customs of a country, things associated with the
country, mollf che (blood person) human blood, and ilo ufisha (meat sheep)
mutton (Smeets 2008:136).
Smeets further says that there are minimal pairs like the following:


Possessive and genitive NPs

a. tfa-chi
kawellu i

dem-attr horse 3.psr ear

the ear of this horse

(Smeets 2008:136)

Nominal compounds in Mapudungun


pilun kawellu
dem-attr ear horse
this horses ear


wangku i
art-3.psr chair
3.psr colour
the colour of his/her chair


art colour art-3.psr chair
the colour of his/her chair

Whereas in (17)a the ear is portrayed as a body part belonging to a particular

horse, the ear is rather generically described in (17)b something that is in accord
with Harmelinks and Baker and Fasolas accounts of head-initial compounds as
presented in Section 3.1 above. By contrast, the second minimal pair is problematic: in addition to the semantic contrast being unclear, my consultants and Baker
and Fasolas consultant, Elisa Loncon, reject structures like (17)d, where the second nominal element is a proper name or a full NP with an explicit determiner
(Baker & Fasola 2009:599). My own data support Baker & Fasolas claim that
expressions like pilun kawellu horses ear are head-initial compounds rather than
[noun + NP complement] structures.
Other genitive NPs discussed by Smeets consist of a first element that is a
locational noun that can be used adverbially, i.e. nouns referring to spatial notions
like wenu (space) above, wlngi (in) front, wif breadth, along, and miche bottom, under, or to temporal notions like pukem (in) winter and wn (at) dawn
(2008:67, 136). This is obviously connected to the case of rangi() middle, centre,
illustrated in (14) above, and perhaps less obviously so to those referring to some
astronomical events listed in (13). According to Smeets, instances like (18)a (parallel
to those in (13) above) are not only analogous to but also different from those like
(18)b. Both expressions consist of a nonfinite verb form followed by a noun, but the
noun is subject with intransitives, as in (18)a, and object with transitives, as in (18)b:


Genitive NPs with nonfinite verb forms

(Smeets 2008:137138)
a. nag-n ant (descend-nfin sun) afternoon
b. wiri-n longko (stripe-nfin head) hair part

Smeets finally notes that nge-n (which she analyses as have-nfin) appears in precisely this position and with this function see (15) and (16) above -, and that the
nominalized form ngen means master and may take her distributive marker -ke
(cf. footnote 5), as in ngen-ke mapu land owners.17
17. Both Augustas translation of Mapudungun ngen as Spanish dueo (English owner) and
Smeetss English master (Spanish maestro) are attempts at using familiar European terms in
order to capture the different semantics of the Mapudungun term, of course.

Fernando Ziga

Smeetss account of such forms is problematic on at least two grounds. First,

while nagn ant afternoon is analogous to the examples given in (13), wiri-n
(write/stripe-nfin) is not only an infinitive meaning writing, striping but also
a nomen actionis meaning line, stripe a polysemy regularly found with these
nonfinite verb forms. Thus, wiri-n longko hair part is perfectly analyzable as
head stripe/line, which follows from Harmelinks and Baker & Fasolas accounts
in a straightforward way. Second, nge-n (be-nfin) (note in passing that have is
actually nie-n, not nge-n) and ngen (owner/master) are distinct in the speech of
careful conservative speakers. But even in the speech of those speakers who do
not longer distinguish dento-aveolar segments from their interdental counterparts, there is no evidence I am aware of that would suggest that ngen < ngen and
ngen = nge-n are anything else than homophones. Therefore, I prefer to reject
Smeetss analysis of expressions like Ngenko water spirit as involving verbal
Lastly, Smeetss partitive NPs are instances in which the first noun indicates
measure, size or quantity [and] the second noun phrase refers to the object measured (2008:135). It is worth listing all of her examples:


Partitive NPs
a. fotella kme pulku
b. klko kel i
c. kongka kna
d. wikef kofke
e. pichin plata
f. fta-ke kaman mapu-che

Tunten kilo ilo ufisha?


kie-ke i pu weny

(Smeets 2008:135136)
(bottle good wine) bottle of good wine
(basket red cherry) basket of red cherries
(sheaf rush)
sheaf of rush
(piece bread)
piece of bread
(a.bit money)
a bit of money
(big-nsg quantity land-person)
large numbers of Mapuche
(how.much kilo meat sheep)
How many kilos of mutton?
(one-nsg 1sg.psr pl friend)
some of my friends

Some of these correspond to what Baker and Fasola mention as involving a

relationship between a container and a substance (19)ab (cf. (9)b and (9)j in
Section 3.1 above). The others have an initial element that indeed refers to a
quantity or measure although wikef piece might be regarded as standing in
a part-whole relationship, or in the relationship Baker and Fasola described as
between a substance and the entity extracted from it, like ilo-trewa (meat-dog)
dog meat, lichi-waka (milk-cow) cow milk, and mollf-che (blood-person)
human blood in (9) above. Interestingly enough, example (19)h includes the
element kie-ke some, one by one; sometimes, which can be used as a determiner (e.g. kieke (pu) che some people) or pronominally (e.g. kieke elufi

Nominal compounds in Mapudungun

tachi epu we-che kuifall ( he.gave.them the two young-person poor)
he gave one [of his daughters] to [each one of] the two poor youngsters, Salas
There is evidence that supports postulating a separate category for cases
involving quantities: unlike head-initial compounds including body parts, these
NPs allow, as acknowledged in Smeetss definition quoted above, complex NPs
as second elements and not only bare nouns. More research is needed here as
well, but my own data suggest that only adjectives are readily accepted within
the embedded second-position NP; my consultants reactions to expressions
like epu kilo tfa-chi ilo (two kilo dem-attr meat) two kilos of this meat were
sometimes inconsistent but basically varied. This is clearly different from what
can be seen in examples like kla tripantu awka (three year mare) three-year
old mare (Augusta 1903:283), which are parallel to their English counterparts
and suggest that the initial nonhead element of head-final NPs can itself be a
complex NP.
Finally, it is important to observe that one of the sources of variation in the
accepted order observed with some two-noun expressions apparently has to do
with this partitive category. The head-final compound lafken-lawen (sea/lakemedicine), for instance, refers to a medicinal plant found in sandy beaches called
pichoa in Spanish (Euphorbia portulacoides), and it admits only one order of its
constitutive elements. The same holds for lafken-che, the name of the coastal Mapuche in Chile, and a number of other compounds built upon lafken sea, lake. In
order to refer to seawater or lake water, however, both the expected lafken-ko and
its mirror image ko lafken are possible. Consultants usually emphasize that there
is no difference in meaning (both get translated as agua de mar/lago in Spanish),
but the latter expression is predictably more flexible in that selected adjectives can
appear in between (e.g. ko fta lafken water of a big lake). Just like both pulkufotilla wine bottle and fotilla-pulku bottle of wine, as we saw in Section 3.1 above,
are possible because they represent different relationships between the two nouns,
lafken-ko and ko lafken are arguably two different ways of describing a particular
referent: the water is portrayed as being in a nonspecific relationship with the sea
in the former case (which naturally leads to the reading seawater, as in English)
and as being extracted from the sea in the latter.
3.3 Appositional constructions
Based on what we have seen so far, the following expressions based on kutran
illness, pain employed in traditional Mapuche medicine are straightforward: as
expected, the noun-noun compound in (20)a is head-final. Note in passing the

Fernando Ziga

adjective-noun compounds in (20)bd and the adverb-noun compounds in (20)

ef, which behave alike with respect to their internal structure:

Head-final expressions
a. wenu(-mapu)-kutran
b. re-kutran
c. pichi-kutran
d. fta-kutran
e. lef-kutran
f. kuyfi-kutran

(Bacigalupo 2007:29)
(above-land-illness) spiritual illness
naturally caused illness
minor illness
major illness
recent illness
(long.ago-illness) chronic illness

The related expressions in (21), which involve body parts as nonheads, are all
predictably head-initial:

Head-final expressions
a. kutran-longko (illness-head)
b. kutran-piwke (illness-heart)
c. kutran-foro
d. kutran-ptra (illness-stomach)
e. kutran-namun (illness-foot)

(Bacigalupo 2007:29)
head illness
heart illness
tooth/bone illness
stomach illness
foot illness

That things can be more complex than this can be seen from the following examples, which refer to different kinds of powerful kura stones used either by a machi
shaman and/or a kalku witch, warlock:



shamans stones


witchs stones


stone shaman


stone witch




kura kalku
stone witch
malefic stones

stone shaman
(beneficial stones)

Nominal compounds in Mapudungun

The head-final expressions in (22)ad are comparable to the one in (20)a above,
and there is no alternative nonhead-head construal available (at least, not along
the lines of the examples in (21) above), since there is a clear animacy asymmetry between the constitutive elements and there are no body parts involved. The
expressions in (22)ab customarily refer to stones used by a shaman and a witch
respectively, while those in (22)cd denote shamans and witches that are characterized by making particularly frequent/prominent use of stones in some of their
activities. By contrast, the last two expressions are unexpectedly head-initial (only
kura kalku is really used; ??kura machi is readily interpretable but some consultants reject it while others say it is not normally used).
What distinguishes (22)ef and similar instances (e.g. trewa ngr (dog fox)
fox-like dog and m filu (bird snake) snake-like bird) from all other examples
presented hitherto is the fact that they are probably best seen as two nouns in
apposition rather than as compounds. The second noun functions as an attribute
of the one, but both are separate words. Semantically, a kura kalku is a kura stone
that is like a kalku witch in some relevant sense; the nonhead is understood as an
attribute of the head that conveys a specific similarity rather than an unspecified
Even though in-depth research is needed here, I was able to detect two alternative single-stress pronunciations of bona fide compounds like kurakalku stone
witch, e.g. [kuakalku] and [kuakalku]. By contrast, the appositional construction kura kalku malefic stone can be realized as either [kua kalku] or, arguably at
least equally often, [kuakalku]. In addition to such phonological criteria, traditional accounts attempt to distinguish compounds from phrasal structures based
on morphological (i.e. inflection and linking elements) and syntactic parameters
(Lieber & tekauer 2009). Morphology cannot help decide here, but it seems that
the second NP can be complex: it can be a compound and/or include selected
adjectives. Thus, such expressions are similar to Smeetss partitive NPs introduced
in Section 3.2 above, which also allow more prosodic autonomy than her genitive NPs and her attributive compounds. Unlike partitive expressions like futilla
(kme) pulku bottle of (good) wine, however, such appositional constructions are
both less limited in semantic terms (not only quantities, parts, or containers are
eligible as initial head element) and apparently less widely used or possibly even
The above does not mean that all other instances of two or more nouns following each other are comparable, however. Head-initial expressions like ruka
machi (house shaman) house of/for shamans, for instance, could be regarded as
relatively recent coinages that mirror the Spanish head-nonhead order (cf. the
expressions ngenke fote owners of the ship mentioned in Section 3.2 above), since
the expected machi ruka is also possible but has, according to some speakers, a

Fernando Ziga

somewhat archaic sound to it. (Note that such buildings where shamans can meet
and work for the public are of very recent date in Chile and Argentina, viz. only a
couple of decades). Be it as it may, according to most speakers, the semantic relationship between machi and ruka is fairly vague in the case of machiruka, whereas
in the case of ruka machi it appears to be constrained in a systematic way by the
pattern found with body parts and part-whole relations. Head-final expressions
like wentru machi (man shaman) male shaman, domo ngr (woman fox) vixen,
domo trewa (woman dog) bitch, on the other hand, are directly comparable to
English woman doctor and therefore differ from both kura kalku malefic stone
and ruka machi house of shamans.
4 Conclusions
With respect to the classification of Mapudungun expressions that include more
than one nominal root (or, more generally, more than one stem), the picture
emerging from Section 3 can be summarized as in Table 1. Observe that I have
written as one orthographic word only those structures that allow a single stress
(Ia and IIa) and as two words those that allow two stresses. The terminology
used by other studies (discussed further down) is given in the last four rows of
Table 1.
Class I includes two kinds of head-final expressions, viz. those that arguably
constitute one phonological word (e.g. mapuche Mapuche, person of the Land)
and those in which the head and the nonhead retain some phonological autonomy
(e.g. domo ngr vixen). Whereas in both subclasses the head must be a simple stem, the nonhead can consist of a noun and an adjective and other elements
(apparently not articles or demonstratives) in Class Ib. In the two subclasses the
semantic relationship between both elements is unspecified.
Class II includes three kinds of head-initial expressions, viz. those that appear
to be single phonological words (e.g. longkomoyo nipple) and two subclasses in
which the head and the nonhead retain some phonological autonomy (possibly
less in instances like futilla pulku bottle of wine and more in others like kura
kalku witch-like stone). As with Class I, tighter phonological units in this class
impose some syntactic restrictions on their constitutive elements as well: in Class
IIa, head and nonhead are basically only nouns, but in Classes IIb and IIc, heads
can be complex NPs, and nonheads can possibly take adjectives. In all subclasses
the semantic relationship between the constitutive elements is specified: it is an
intimate part-whole relationship in IIa (prototypically, a body part relationship), a
non-intimate part-whole relationship in IIb (typically, quantity expressions), and
a similarity relationship in IIc.

Nominal compounds in Mapudungun

Table 1. Multi-stem nominal expressions in Mapudungun

I. Head-final

I. Head-initial

Class Ia

Class Ib

Class IIa

Class IIb

Class IIc


wine bottle

domo ngr

cows head

futilla pulku
bottle of wine

kura kalku

main stress units





complex NPs


head excluded excluded





Baker & Fasola

Smeets (2008)

head ok
head ok
nonhead limited nonhead

intimate part non-intimate





(syntactic) argumental


subordinate subordinate
complex NPs complex NPs


subordinate compounds

Bisetto & Scalise subordinate


In terms of Baker and Fasolas (2009) analysis, Class Ia is their (lexical)

modificational head-final noun-noun compounds, and Class IIa and IIb are the
(unacknowledged) subclasses of their (syntactic) argumental head-initial nounincorporating compounds. In terms of Smeetss (2008) account, Class Ia corresponds to her compounds, Class IIa to her genitive subordinative complex NPs,
and Class IIb to her partitive subordinative complex NPs. I hope I have been able
to show in Section 3 that these analyses should be refined along the lines proposed
here and to some extent we can already do so, even though I have left some recalcitrant anomalies in Class IIb unexplained and have characterized Classes Ib and
IIc only sketchily. Nevertheless, not only Baker & Fasola but also Smeets have to be
commended for having pointed out the systematic semantic asymmetry, and some
of the syntactic asymmetries, between head-final and head-initial compounds.
Such a form-function correspondence pattern has not found a prominent
place in the taxonomies of compounds found in the typological literature. In
terms of the classification proposed in Bisetto and Scalise (2005), Classes Ia, IIa
and IIb are the (unacknowledged) subclasses of their subordinate class (in turn

Fernando Ziga

corresponding to what Scalise and Bisetto 2009 call ground subordinate class), like
English windmill and mushroom soup, while Class Ib corresponds to their coordinate class.18 Thus, Mapudungun presents typologists with an interesting case
of how form-function correspondences can pattern in a systematic way, suggesting as it does that different kinds of semantic relationship between the constitutive elements have to be taken into account as well. It is simply not the case
that all compounds in a language will be head-final or head-initial because some
language-specific parameter stipulates such a regularity; even synchronically, and
arguably without the intervention of contact-induced restructuring, different
ordering-meaning patterns have to be recognized and explained.

adj = adjective; art = article; attr = attributivizer; dem = demonstrative; inf =
infinitive, N = noun; nfin = nonfinite, nmlz = nominalizer, nsg = nonsingular,
pl = plural, psr = possessor, sg = singular

Arnold, Jennifer. 1996. The inverse system in Mapudungun and other languages. Revista de
Lingstica Terica y Aplicada 34: 947.
de Augusta, Flix Jos. 1903. Gramtica araucana. Valdivia: Imprenta Cenral J. Lampert.
de Augusta, Flix Jos. 1916. Diccionario araucano. Santiago: Imprenta Universitaria.
Bacigalupo, Ana Mariella. 2007. Shamans of the Foye Tree. Gender, Power, and Healing among
Chilean Mapuche. Austin TX: University of Texas Press.
Baker, Mark. 1988. Incorporation: A Theory of Grammatical Function Changing. Chicago IL:
University of Chicago Press.
Baker, Mark, Aranovich, Roberto & Golluscio, Luca. 2005. Two types of syntactic noun incorporation: noun incorporation in Mapudungun and its typological implications. Language
81(1): 138176. DOI: 10.1353/lan.2005.0003
Baker, Mark & Fasola, Carlos. 2009. Araucanian: Mapudungun. In Rochelle Lieber & Pavol
tekauer (eds), 594608.
Bisetto, Antonietta & Scalise, Sergio. 2005. The classification of compounds. Lingue e Linguaggio
4(2): 319332.

18. It is in order to note that all of the classes postulated here for Mapudungun multi-stem
nominal expressions are endocentric; exocentric compounds like Italian lavapiatti dishwasher and Spanish espantapjaros scarecrow would correspond to nominalized verb-noun
compounds. See Scalise and Bisetto (2009) for more information on the role played by the
endocentric-exocentric opposition in the classification of compounds.

Nominal compounds in Mapudungun

Erize, Esteban. 1960. Diccionario comentado mapuche-espaol. Baha Blanca: Yepun.

Golluscio, Luca. 1997. Notas sobre la incorporacin nominal en mapudungun. Actas III Jornadas de Lingstica Aborigen 3: 155167.
Harmelink, Bryan. 1992. La incorporacin nominal en el mapudungun. Lenguas Modernas 19:
Harmelink, Bryan. 1996. Manual de aprendizaje del idioma mapuche. Aspectos morfolgicos y
sintcticos. Temuco: Ediciones Universidad de la Frontera.
Hernndez, Arturo & Ramos, Nelly. 2005. Mapuche: Lengua y cultura. Mapudungun espaolingls. Santiago: Pehun.
Lieber, Rochelle & tekauer, Pavol. 2009. Introduction: Status and definition of compounding.
In Rochelle Lieber & Pavol tekauer (eds), 318.
Lieber, Rochelle & tekauer, Pavol (eds). 2009. The Oxford Handbook of Compounding. Oxford:
Salas, Adalberto. 1992. El mapuche o araucano. Madrid: MAPFRE.
Scalise, Sergio, & Bisetto, Antonietta. 2009. The classification of compounds. In Rochelle
Lieber& Pavol tekauer (eds), 3453.
Smeets, Ineke. 2008. A grammar of Mapuche. Berlin: Mouton de Gruyter. DOI:
Ziga, Fernando. 2001. Mapudungun. Munich: Lincom.
Ziga, Fernando. 2006. Mapudungun El habla mapuche. Santiago: Centro de Estudios

Towards a characterization of compounding

in Mak
Temis L. Tacconi

(University of Buenos Aires, CONICET)

This paper explores the formation of compounds in Mak and intends to
characterize these constructions in formal, semantic and typological terms.
To this end, we have considered elements that constitute compounds and
the resulting lexemes, as well as the relationship between constituents. We
also seek to contribute to the discussion about the parameters that should be
taken into account for lexemes to be considered compounds, and whether
they correspond to the features found in languages from the same family or
to a typological universal. In order to do so, we also compare the findings to
derivational strategies, and discuss the sometimes unclear boundaries between
the phenomena of composition and derivation.
Keywords: Mak; Gran Chaco region; composition; derivational morphology;

1. Introduction
In this paper we study the structure of compound formation in the Mak language
from a formal and typological point of view. We explore compounds as a whole,
as well as the lexemes that form them, in order to find the existing patterns in the

* Research for this article was part of a doctoral scholarship (20092012) granted by the
Consejo Nacional de Investigaciones Cientficas y Tcnicas, and of a more extensive project
(ANPCyT FONCYT PICT Bicentenario 0136 20102013), entitled Lenguas indgenas de
la Argentina y pases limtrofes, con especial referencia a la regin del Gran Chaco (toba,
chorote, mak, mbya y quechua boliviano). Estudios descriptivos, tipolgicos y de contacto
lingstico, supervised by Cristina Messineo.
A preliminary version of this article was presented at the XIII Congreso de la Sociedad
Argentina de Lingstica, San Luis, March 2730, 2012.

Temis L. Tacconi

In some indigenous languages of the Americas, compounding is not commonly found as a productive lexicon expanding process, for example in Quechua,
Aymara (Adelaar with Muysken 2004) and Arawakan (Zamponi 2009; but consider also the contrastive view of Admiraal and Danielsen, this volume), however,
this does not seem to be the case for Mak, as shown by previous papers on the
ethnobiological lexicon (Cneo 2008 and this volume; Gonzlez & Montani 2010;
Messineo 2009; Messineo & Cneo 2011; Spinelli 2008). This paper intends to
contribute to the discussion about the criteria used for lexemes to be considered
compounds, and whether they correspond to the features found in languages from
the same family or to typological universals. To this end, we shall also relate compositional processes to derivational processes and discuss the sometimes unclear
boundaries between both types of phenomena.
Section 2 contains a brief description of the language community and specifies the data sources and methodology of this paper. Section 3 is about the general
features of composition in Mak, while Section 4 deals with the internal structure of compounds. Finally, Section 5 compares information on Mak with that of
other languages from the same family and region.
2. Characteristics of the Mak community and language
The Mak community, who originally came from the Paraguayan Chaco, currently
lives in three locations: Mariano Roque Alonso (hereafter referred to as MRA),
Falcn and Ita Paso. The three groups have 1,500 speakers altogether (1,282 residents based on the 2002 DGEEC1 census). The chief and the largest group dwell in
MRA. There is a school in the area they inhabit,2 as well as a church, a community
centre, and a first aid room. The same facilities can also be found at the Falcn
community, which is alternatively referred to as Qemkuket. The three groups have
permanent contact and are governed by the same chief, thus they can be considered one social and political organization.
The Mak have permanent contact with speakers of Spanish and Guaran (the
official languages in Paraguay), as they trade crafts in the city centre and surrounding areas and purchase food, hygiene supplies, etc. in local stores, where the official languages are used. Despite this, children and most women only speak Mak,
1. Direccin General de Estadstica, Encuestas y Censos [General Bureau of Statistics,
Surveys and Census]
2. Lessons are exclusively in the Mak language until the fourth grade. From the fifth grade
onwards Spanish and Guaran (the official languages of Paraguay) are incorporated.

Towards a characterization of compounding in Mak

while adult men speak all three languages. Daily community conversations are
conducted in vernacular, i.e. Mak.
According to sources, the Mak people historically inhabited the Paraguayan
Chaco, between the rivers Confuso and Montelindo (Gerzenstein 1995). After the
Chaco War (19321935) the community was moved to Colonia Fray Bartolom
de las Casas, which faces Puerto Botnico (Asuncin). Later, in 1985, as a result of
frequent floods, they were relocated to the New Mak Indigenous Colony in MRA,
a 25-acre area of land, which belongs to the group.
The Mak language belongs to the Mataco-Mataguayan language family,
together with Wich, Chorote, and Nivacl. Some authors acknowledge a possible common link to Guaycuruan languages, or at least common features resulting from geographical proximity and a concurrent cultural and ecological context
(Messineo 2011; Viegas Barros 2006; among others). One of the most relevant
features for our analysis is the fact that it is an agglutinative language. The basic
word order is VS for intransitive sentences and SVO for transitive sentences. This
order can be found in noun phrases and compound formation (Tacconi 2009).
Inalienable possession is marked on the possessed noun by means of a prefix.
Mak also uses a prefix (wit- ind) as an indefinite possessor marker. Number
and gender are marked through suffixes. Allomorphs for the plural suffix are -ts,
-y, -l and -i, which adjust to morphophonological conditions (see Gerzenstein
1995:150 for further details). The marked gender is feminine, which is indicated
through a suffix with the following variants: -i, -e, -ki and -le.
The data were collected by the author on various fieldtrips between 2009 and
2011. In addition, secondary data from Gerzenstein (1999) and Arenas (1983)
were used where indicated. First-hand information was elicited in situ with the
help of questionnaires designed according to the specific needs. Furthermore, a
group of Mak speakers visited the University of Buenos Aires in 2011, where
previous questionnaires were corrected and new data were elicited.
3. Compounding in Mak: General features
Compounds are defined as complex lexical units containing two or more simple
or compound units, which syntactically act as a single phonological, morphological and semantic unit, i.e. they are indivisible and syntactically opaque (Brinton
& Traugott 2005:34; Matthews 1991). However, [phonological, morphosyntactic
and semantic] criteria which make it possible to distinguish between compounds
and syntactic phrases may vary from one language to the other (Aikhenvald 2007:
24). We believe it is possible to talk about clear compound cases in Mak, since
these units correspond to morphological and syntactic criteria specified below.

Temis L. Tacconi

First, simple attributive clauses have the following default internal structure:
Attributive Predicate + Demonstrative + Noun

The basic word order in Mak is VS for intransitive sentences and VSO for transitive ones (see Section 2). This order also applies to Attributive Predicate clauses, as
in (1), where nominal predicates with subject markers are followed by a demonstrative and a noun:


3sg-beautiful dem-f woman
The woman is beautiful.

The following are subject markers in nominal predicates:


yV- 4
V- ~ V-

1pl inclusive jV-


(Gerzenstein 1995:157)

These attributive predicates are actually nominal predicates with the same subject
markers as verbs predicating a quality of the referent.
In the case of compound nouns that replicate this structure (Attributive Predicate + Noun), the clear syntactic difference lies in the fact that these compounds
do not allow for a demonstrative to be inserted between the attributive predicate
and the noun, as in (2):


dem-f 3sg-yellow-3sg.poss-flower
Shrubby Yellowcrest (type of plant, Heimia salicifolia),
(lit. yellow its flower)

(PA6 158)

3. The column to the left corresponds to the markers for nouns with an initial consonant; the
right column corresponds to the markers for nouns with an initial vowel.
4. V stands for a vowel adhering to the vowel harmony principle: theme vowel a corresponds to a, theme vowel o to o, and theme vowel e, i or u correspond to e.
5. For the sake of uniformity I use small hyphens to divide morphs or bases in compound
lexemes throughout this paper, even though the Mak orthography is not standardized yet
and some compounds are still written as two separate words.
6. The following abbreviation is used for botanical and zoological references: PA (Arenas
1983). After the common name of the plant, follows the scientific name in Latin.

Towards a characterization of compounding in Mak

By contrast, example (3) is a predicative clause with the same attribute, but it is not
a compound:


-maajaji n-e
3sg-yellow dem-f 3sg.poss-flower
The flower is yellow.

Secondly, the possessive third person singular marker7 in the compounds, which
is also part of the compound in (2), is frozen and cannot change (as shown in
(4)). In example (5), by contrast, we see a predication, i.e. an attributive clause:


big armadillo (lit. white its ribs)


-yipjolax yi-wei
3sg -white 1sg.poss-rib
My rib is white.

From a semantic perspective, a compound is a lexical unit, as proven by the fact

that the meaning of the whole is not necessarily equivalent to the sum of each of
the compounds parts.8 Example (6) shows the way both elements in the compound (nouns referring to a bird and a part of a birds body) make reference to a
plant, due to their similar shape. It is, however, not possible to infer the designated
object from both elements in isolation.


Passion Flower (type of plant, Passiflora caerulea// P. cincinnata//P.
mooreana), (lit. how hows (a bird type) testicles)
(PA 148)

While the criteria cited above seem to work for a characterization of compound
lexemes in Mak and other languages of the world, the boundaries between derivation, composition and syntax are sometimes unclear, as claimed by Anderson
(1985:40) and others. Since it is difficult to establish clear boundaries between
these categories, we can think of a continuum instead (compare also Admiraal &
Danielsen, this volume).

7. The third person singular possessive prefix has the following allomorphs: V- when the
noun has an initial consonant (the V stands for the first vowel of the root, which adheres to
the vowel harmony principle, see footnote 5); and - when it is vowel-initial.
8. Some compounds in Mak are still semantically transparent despite their undisputed
status as compounds.

Temis L. Tacconi

In the case of Mak, we can cite two liminal cases, where it is hard to find a
specific category for the lexeme (derivative, compound or phrase). One example
is the suffix -wet place, which is always attached to verbs and nouns to indicate a
place, as in (7) and (8). This suffix does not exist as an independent lexeme in the


stage (lit. place for singing)


football pitch (lit. place for ball)

So far -wet place has been analyzed as a derivational suffix in Mak (Tacconi
2011; Tacconi & Abrach 2010). However, in other languages of the same family,
wet is an independent lexeme meaning place, and it has been used as an independent lexeme in discourse, but also as part of a compound, as for example in the
Wich equivalents of (9) beach and (10) bed, which are both formed with wet.
Apart from the analysis as lexical roots in Wich, the compounds look very similar
to the suffix constructions in Mak, given in (7) and (8).






(Spinelli 2008:8)

In Chorote wet place is also analysed as an independent lexeme, which can be

part of a compound, as in the word for uterus in (11):


uterus (lit. the place of her child)


(Carol 2012:379)

In (12), a further example from Chorote, the lexeme wet place appears by
itself within a sentence:


Juan y-i
Juan 3sg-be 3poss-place-appl
Juan is in his house.


(Carol 2012:268)

The analysis of these comparative data may lead to assumptions about the origin of
the Mak derivative suffix -wet place as an independent lexeme capable of forming compounds and grammaticalizing into a productive suffix.

Towards a characterization of compounding in Mak

A similar process, at a different stage of development, can be observed with

certain independent lexemes from the Mak language, e.g. as child, which forms
compounds such as names of animal offspring, see (13) and (14):


kitten, cat offspring (lit. cats child)


calf (lit. cows child)

The lexeme as child appears to be undergoing a grammaticalization process in

this language, since it also works as a diminutive suffix (cf. Messineo 2009; compare Cneo, this volume). The use of the term child as a diminutive is a very
common phenomenon in languages of the world (Heine & Kuteva 2002:65). In
Examples (15)b and (16)b the derivation of diminutive forms with -as child is




little ball






In (17)b and (18)b the suffix marks the diminutive of a kinship term, where the
original meaning of the lexeme (child) is not used in its literal meaning but marks



my grandfather


my grandpa (lit. my grandfather his child)


my grandmother

Temis L. Tacconi


my granny (lit. my grandmother her child)

The two studied cases (-wet place and -as child) demonstrate how difficult it
is to distinguish between derivation and compounding. The parameters to take
into account will change from one language to the other. Nevertheless, if we leave
diachronic data aside, in the specific case of Mak, -wet place can be classified as
a derivative suffix and as child as an independent lexeme that forms compounds
and also has the function of creating diminutive lexemes.
4. Internal structure of compounds
As far as their constitutive elements are concerned, compound nouns can be
formed in the following ways in Mak:

Noun + poss-Noun
Attributive Predicate + poss-Noun
poss-Noun + Attributive Predicate (unexpected order)
Existential verb + poss-Noun
Quantifier + poss-Noun

Note that the second constituent is always a possessed noun, bar one exception.
The subsections below contain a detailed analysis of each of the listed types of
Mak compounds, distinguished by the first (modifying) root.
4.1 Noun + poss-Noun
Some compounds are formed by two nouns in a possessive relationship. The first
noun refers to the possessor, while the second one, invariably marked with a predetermined third person (see Section 3), refers to the possessed. In this type of
N+N compound, the third person singular possessive prefix always refers to the
possessor of the first element, as in (19) through (21), i.e. the compound is endocentric, with the second element as the formal head:


Little Bluestem (type of plant, Schizachyrium microstachyum),
(lit. foxs tails)
(PA 178)


Monkeys Comb (type of plant, Pithecoctenium cynanchoides),
(lit. cows tongues)
(PA 172)

Towards a characterization of compounding in Mak


type of orchid (The Large Flower Sarcoglottis, Sarcoglottis grand flora),
(lit. she who lives in the palm grove)
(PA 146)

Among compounds formed with two nouns, those making reference to the use of
the denoted object are very common. Names of plants, which allude to the physiological and ecological aspects (particularly food) of animal species, see (22) and
(23), or to the ways in which the plant is used in the culture, as in (24) and (25),
often have this form.


Eryngo (type of plant, Eryngium coronatum), (lit. the deers food) (PA 136)


Hummingbird Bush (type of plant, Dicliptera tweediana),
(lit. the hummingbirds food)

(PA 145)


f oj--efitetsii
Yellow Alder (type of plant, Turnera ulmifolia), (lit. whistle root) (PA 147)


Poreleaf (type of plant, Porophyllum lanceolatum), (lit. conjunctivitis roots,
i.e. used to cure conjunctivitis)
(PA 152)

Note that this type of compound formation is very productive, proven by the fact
that we can find neologisms created by joining two nouns, as in (26) and (27). The
compound in example (26) metaphorically designates the cars wheels as the hands
of the car.


wheel (lit. car its hands)

In example (27) the compound is formed by the union of two lexemes, both
of which are complex themselves. The first noun omeas child is apparently a
lexicalized compound derived from the lexeme as child mentioned above (see
Section3). The second noun athilax sanitary towel is derived from the verbal root
athii to menstruate with the instrumental nominalizing suffix -lax that indicates
finality, as is indicated in (28). These two complex forms can therefore also be
joined in order to compose a new lexeme (27).

Temis L. Tacconi


diaper (lit. child his/her sanitary napkin)


sanitary napkin (lit. for menstruate)

4.2 Attributive Predicate + poss-Noun

A second type of compound bears the structure of an attributive phrase (see
Section 3). Here the third person singular or indefinite possessive prefix is also
attached to the head noun. Mak does not have a morphologically and syntactically distinct lexical class of adjectives (see Gerzenstein 2002; Cneo et al. 2009 for
an in-depth analysis of this matter)9, so, as we see in Section 3, these attributive
predicates are nominal predicates indicating a quality of the referent, cf. also (34)
below. This type of compound is often found with phytonyms and zoonyms, as the
following examples show:



Croton (type of plant, Croton lachnostachyus),
(lit. its leaves are soft)

(PA 142)

Wingleaf Soapberry (type of plant, Sapindus saponaria),
(lit. its fruit is black)

(PA 160)


Espinhera Santa (type of plant, Maytenus ilicifolia), (lit. its
leaves are sharp)
(PA 179)


elephant (lit. big its ears/big-eared)

9. Compare also Cneo and Admiraal & Danielsen, this volume.

10. In this particular example the subject marker corresponds to those of regular verbs
(y- for 3rd person singular), not to those of attributive predicates.

Towards a characterization of compounding in Mak

4.3 poss-Noun + Attributive Predicate (unexpected order)

Based on the pattern of this language, we would expect to find the order that
occurs in the compounds in Section 4.2 in other similar examples. This expectation is met in example (33), where the compound is formed by placing an attributive predicate before a noun:


Dayflower (Commelina erecta), (lit. blue its flower)

(PA 155)

This order should be reproduced in more complex compounds, i.e. those constructed with two complex nouns, like the one in (34). In this example, the second
component is a complex noun, consisting of an attributive element and the modified marked noun, compare to Examples (2) and (24). Thus, (34) concurs with the
expected order of Attributive Predicate + Noun within the compound.


-majaji -opom
3sg-yellow 3poss-flower
Water Poppy (Hydrocleis nymphoides), (lit. camalote yellow
its flower)

(PA 150)

Nevertheless, it is worth pointing out that we have found cases of complex compounds with a different order of the constitutive elements. Example (35), for
instance, contains a complex noun, in which the attributive predicate follows the
Noun + poss-Noun + Attributive Predicate


type of plant, Euglypha rojasiana, (lit. the rheas skin is yellow) (PA 170)

A divergence from the regular pattern can also be seen in (36), where the order of
the elements in the compound is reversed, in spite of not even being a very complex construction.
possNoun + Attributive Predicate

egg yolk

11. aquatic plant

Temis L. Tacconi

Following the pattern of the language, we would expect a compound, such as the
one in (37). However, this example represents an ungrammatical construction in
Attributive Predicate + possNoun


yellow its food

We are unable to provide an explanation for this phenomenon. However, the fact
that they are lexicalized constructions, which do not follow the expected order,
might contribute to their classification as compounds, i.e. lexical items, which
are not completely governed by the rules of syntax. In addition, more lexicalized
compounds sometimes reflect older structures. Note that neologisms would rather
adhere to the order Attributive Predicate + Noun, as illustrated in (26). Another
possible explanation for the reversed order of these compounds could be related
to the fact that most of them have more than two elements, which means that the
compound has gone through different steps in the compounding process, which is
why a different structure may be preferred.12
4.4 Negative Existential Predicate + poss-Noun
The negative existential predicate ham (without, there is/are no...) in Mak differs
from the verbal negator. Negative existential predicate clauses are formed by ham
+ Noun + Privative (-e/-ye13), as shown in (38) and (39):



neg.exs fire-priv
There is no fire.

(Gerzenstein 1995:212)

neg.exs water-priv
There is no water.

(Gerzenstein 1995:212)

Another compound structure found in Mak, though less common, is based on

the pattern Negative Existential Predicate + Noun. Again, there is a third person
possessive prefix involved in these compounds.

12. Note that also for Mapudungun there are compounds in the reverse order, as shown in
Ziga, this volume.
13. The allomorphs are: -e if the noun has a final consonant; -ye if the noun is vowel-final.

Towards a characterization of compounding in Mak


young bull (lit. There is no testicle of it [i.e. the bull].)


Milkwort (type of plant, Polygala mulluginifolia),
(lit. There are no children of her.)


(PA 148)

mute (lit. There are no words of him/her.)

In fact, the privative suffix does not always appear in compounds, as you can
observe in (41). At the present state of research we cannot explain the lack of the
suffix here. However, this type of compound is not frequent, so we do not have
enough data to reach a conclusion about the way they are formed.
4.5 Quantifier + Noun
Finally, a fourth type of compound shows the structure Quantifier + poss-Noun.
As we have seen in other compounds, the possessive prefix is either the third person, as in (43), or the indefinite possessive prefix, as in (44):


centipede (lit. many its hands)


type of plant, Melothria cucumis, (lit. little blood)14

(PA 159)

In both (43) and (44), the construction can be distinguished from a phrase in
which the predicative element quantifies the noun, as in (45), since it lacks a
demonstrative before the noun (see also Section 3):


olots n-e
many dem-f woman-pl
There are many women.

(Gerzenstein 1999:281)

As we have seen, the distinction between compounds and phrases is not always
so straightforward; nevertheless, we have been able to present some criteria that

14. This noun refers to the plant being used as a haemostatic agent to reduce menstrual flow.

Temis L. Tacconi

help to identify compounds in Mak. These criteria, as pointed out by Anderson,

might be more or less reliable in the different languages (Anderson 1985:41).
The reliable criterion in Mak is the inseparability of the elements of the compound, e.g. an element like the demonstrative cannot be inserted without changing the meaning of the construction. In addition, the meaning of most compounds
is not compositional.
5. Composition in other languages from the Chaco region
This section focuses on composition types in other languages from the MatacoMataguayan family (Wich and Chorote) and the same region (Toba) in order to
show that the described characteristics are widespread in the Chaco region.
In other languages from the same family, and in the Gran Chaco area, different types of compounds have been identified as well. Some structures found
in compounds in Mak have also been attested in compounds in Wich (Spinelli
2008:7), for example, Noun + Noun in (46); Attributive Predicate + Noun in (47)
and also the reversed order Noun + Attributive Predicate in (48):








Other compounds, which have not been found in Mak, are those formed by an
adverb and a noun, resulting in an adverb, e.g. (49), and a verb and a noun, resulting in a noun, e.g. (50). In Mak, the only verbal element that occurs in compounds is the negative existential, as shown in Section 4.4. Mak verbs may be
nominalized and then become part of a compound, see (27).





(Spinelli 2008)

Towards a characterization of compounding in Mak

As claimed by Carol (2012), in Chorote, constructions of the kind N-3poss-N

have a listed meaning, and although they have an internal structure, they are also
syntactically opaque, i.e. they do not allow for (certain) internal syntactic operations (Carol 2012:381). Nonetheless, pluralization of the first term in some of
these compounds and the presence of the possessive marking create doubts about
their status as compounds and make the author characterize them as lexicalized
phrases (Carol 2012:383). But this does not seem to be a conclusive fact, since
other languages allow syntactic operations inside the compound (for example,
Portuguese marks plural in both components (Aikhenvald 2007:26), and Spanish
pluralizes the first noun in compounds).


lamb younger than one year old


(Carol 2012:379)


uterus (repeated from (11) above)

(Carol 2012:379)

piece of bread

(Carol 2012:379)


On the other hand, Toba, a language from the Guaycuruan family, has compounds similar to those found in Mak and Wich, such as (54), (55) and (56),
formed by a Noun + Noun, a Noun + Attributive, and a Predicative Noun + Noun,


frog (lit. horse its drink)

Toba (Guaycuru)

(Cneo 2012:244)


type of bat (lit. fishing bat)

(Cneo 2012:247)

type of parrot (lit. (is) hard its head)

(Cneo 2012:250)


15. This word can also be analyzed as a compound: wolepo (wool-3sg-exist) sheep (lit.
wool exists).

Temis L. Tacconi

Even when Mak and Toba are languages from different families, they share a
common area, the Gran Chaco, and there are authors that claim that both families
(Mataguayan and Guaycuruan) are related (Viegas Barros 1993, 2006). The fact
that both languages have similar compound structures could be due to language
contact, or it might be further proof of their belonging to the same family.
6. Conclusion
The purpose of this paper is to characterize compounds in the Mak language.
In the first part, we addressed difficulties that occur when trying to make a clear
distinction between derivation, composition and syntax. Grammaticalization and
lexicalization processes play a fundamental role in this distinction, as we have seen
in Examples (7) and (8), where lexemes, that were independent before, are now
derivative morphemes. However, we have been able to establish clear criteria, such
as the inseparability of the elements meaning that no demonstrative or any other
element can be inserted between the two parts of a compound (see Examples
(2) and (3)) and identify certain lexemes as compounds in this language. Also,
the idiosyncratic and non-compositional meaning is commonly a characteristic of
compounds. Both these criteria are commonly attested in other languages of the
area. Finally, we found a recurrent structure in the studied lexemes, which is also
attested for neologisms.
We have also been able to compare the structures described for Mak to the
structures of compounds in other languages of the Chaco region. As we have
shown, Chorote, Wich, (both Mataco-Mataguayan language family, like Mak)
and Toba (Guaycuruan) have similar types of compounds. Although there may be
some obscure cases where it is difficult to establish whether certain compounds are
lexicalized, we could claim that composition is an existing and productive strategy
in the Mak language and presumably also in the whole Chaco area (Messineo &
Tacconi 2010; Cneo 2008 and this volume; Spinelli 2008; among others).

appl = applicative; attr = attributive; coll = collective; dem = demonstrative; f=
feminine; ind = indefinite; neg = negation; neg.exs = negative existential verb;
nmlz = nominalizer; pl = plural; poss = possessive; priv = privative; quant =
quantifier; s = single argument of intransitive verb; sg = singular; ?: unidentified

Towards a characterization of compounding in Mak

Adelaar, Willem F.H. with Muysken, Pieter C. 2004. The Languages of the Andes. Cambridge:
CUP. DOI: 10.1017/CBO9780511486852
Aikhenvald, Alexandra. 2007. Typological distinctions in word formation. In Language Typology and Syntactic Description, Vol. III: Grammatical Categories and the Lexicon, 2nd edn,
Timothy Shopen (ed), 165. Cambridge: CUP.
Anderson, Stephen R. 1985. Typological distinctions in word formation. In Language Typology and Syntactic description, Vol. III: Grammatical Categories and the Lexicon, 1st edn,
Timothy Shopen (ed), 356. Cambridge: CUP.
Arenas, Pastor. 1983. Nombres y usos de las plantas por los indgenas Mak del Chaco Boreal.
Parodiana 2(2): 131229.
Brinton, Laurel & Traugott, Elizabeth. 2005. Lexicalization and Language Change. Cambridge:
CUP. DOI: 10.1017/CBO9780511615962
Carol, Javier. 2012. Lengua chorote (Mataguayo). Estudio descriptivo. PhD dissertation, Universidad de Buenos Aires.
Cneo, Paola. 2008. Categorizacin lingstica: Los nombres compuestos en la denominacin
de plantas y animales en toba (guaycur). In Estudios en Lingstica y Antropologa.
Homenaje a Ana Gerzenstein, Cristina Messineo, Marisa Malvestitti & Roberto Bein (eds),
257268. Buenos Aires: Instituto de Lingstica, Facultad de Filosofa y Letras, Universidad
de Buenos Aires.
Cneo, Paola. 2012. Clasificacin nominal y formacin de palabras en toba (familia guaycur),
con especial referencia al lxico etnobiolgico. PhD dissertation, Universidad de Buenos
Cneo, Paola, Dante, Patricia & Tacconi, Temis. 2009. Tipologa lxica: Una aproximacin a
dos lenguas chaqueas: toba (guaycur) y mak (mataguaya). Cadernos de Etnolingstica
Sul-Americana 1(2). (26 November 2013).
DGEEC de Paraguay. 2002. Atlas de las Comunidades Indgenas en el Paraguay. Asuncin:
Gerzenstein, Ana. 1995. El mak. Estudio descriptivo. Buenos Aires: Facultad de Filosofa y
Letras, Universidad de Buenos Aires.
Gerzenstein, Ana. 1999. Diccionario Etnolingstico Mak-Espaol. Buenos Aires: Facultad de
Filosofa y Letras, Universidad de Buenos Aires.
Gerzenstein, Ana. 2002. Lengua indgena mak: el adjetivo, una categora autnoma o una
subclase nominal o verbal? Actas del IX Congreso de la Sociedad Argentina de Lingstica,
Salta. Edition on CD.
Gonzlez, Hebe & Montani, Cecilia. 2010. Lxico etno-botnico tapiete (tup-guaran): anlisis morfolgico preliminar. Presentation at XII Congreso de la Sociedad Argentina de
Lingstica (SAL). Facultad de Filosofa y Letras, UNCUYO.
Heine, Bernd & Kuteva, Tania. 2002. World Lexicon of Grammaticalization. Cambridge: CUP.
DOI: 10.1017/CBO9780511613463
Matthews, Peter Hugoe. 1991. Morphology. Cambridge: CUP. DOI: 10.1017/CBO9781139166485
Messineo, Cristina. 2009. Modos de clasificacin nominal en el dominio de la flora y de la fauna
en mak (familia mataco-mataguaya). UniverSOS: Revista de Lenguas Indgenas y Universos
Culturales 6: 119138.

Temis L. Tacconi
Messineo, Cristina. 2011. Aproximacin tipolgica a las lenguas indgenas del Gran Chaco.
Rasgos compartidos entre toba (familia guaycur) y mak (familia mataco-mataguayo).
Indiana 28: 183226.
Messineo, Cristina & Cneo, Paola. 2011. Ethnobiological classification in two indigenous languages of the Gran Chaco region: Toba (Guaycuruan) and Mak (Mataco-Mataguayan).
Anthropological Linguistics 53(2): 132169. DOI: 10.1353/anl.2011.0010
Messineo, Cristina & Tacconi, Temis. 2010. Recursos de formacin del lxico en mak
(mataguayo): Zoonimia y fitonimia. In Lxico, clasificacin nominal y categorizacin etnobiolgica en el Gran Chaco, Christina Messineo, Gustavo Scarpa & Florencia Tola (eds),
83116. Santa Rosa: Editorial de la Universidad Nacional de La Pampa (EDUNLAPam).
Spinelli, Silvia A. 2008. Derivacin y composicin en la lengua wich del Chaco salteo. Actas del
XI Congreso de la Sociedad Argentina de Lingstica. Santa Fe: Facultad de Humanidades y
Ciencias. Edition on CD.
Tacconi, Temis. 2009. Orden de los constituyentes en mak (mataco-mataguaya). Actas del II
Encuentro de Lenguas Indgenas Americanas (ELIA). Resistencia: Universidad Nacional del
Nordeste. Edition on CD.
Tacconi, Temis. 2011. Neologismos en mak (mataco-mataguayo). un fenmeno de resistencia frente al contacto? Paper presented at I Jornadas de Jvenes Lingistas, March 2123.
University of Buenos Aires.
Tacconi, Temis & Abrach, Luisina. 2010. Formacin del lxico y estrategias de resistencia tnica
entre los mak (mataguayo). Paper presented at VII Encuentro de Lenguas Aborgenes y
Extranjeras, July 13. Universidad Nacional de Salta.
Viegas Barros, Jos Pedro. 1993. Existe Una Relacin Gentica Entre Las Lenguas Mataguayas
y Guaycures? Variedades Diatpicas del Toba. Hacia Una Nueva Carta tnica Del Gran
Chaco, 5: 193213.
Viegas Barros, Jos Pedro. 2006. La hiptesis macro-guaicur: semejanzas gramaticales
guaicur-mataguayo. UniverSOS: Revista de Lenguas Indgenas y Universos Culturales 3:
Zamponi, Raoul. 2009. Arawakan: Maipure-Yavitero. In The Oxford Handbook of Compounding,
Rochelle Lieber & Pavol tekauer (eds), 584593. Oxford: OUP.

Augmentative in Toba (Guaycuruan)

Form and function
Paola Cneo

University of Buenos Aires, CONICET*

The purpose of this paper is to analyse the morphosyntactic and semantic
aspects of the augmentative forms in the Toba language (Guaycuruan family),
spoken in the Gran Chaco region (Argentina). The study of the forms linked
to the notion of augmentative in this language comprises both derivational
morphology and nominal composition, and embodies a great range of
meanings from big size, abundance, intensity or affection to the (generally
pejorative) notions of excess or mockery. Augmentative forms also play a
role as a source of lexical creation and as a nominal categorization device.
Furthermore, in the ethnobiological lexicon, they constitute a preferred means
in word formation for naming animals and plants, and they code meanings such
as hierarchy (more dangerous or with outstanding qualities) or anomaly
(stranger/ unknown/ unusual).
Keywords: Toba; Guaycuruan; augmentative; evaluative morphology;
kinship compounds

* Research for this article was part of a postdoctoral scholarship (20122014) by the Consejo
Nacional de Investigaciones Cientficas y Tcnicas, titled Morfologa evaluativa: aspectos
morfosintcticos, semnticos y discursivos de los aumentativos y los diminutivos en toba
(guaycur), and of a more extensive project, titled Lenguas indgenas de la Argentina y pases
limtrofes, con especial referencia a la regin del Gran Chaco (toba, chorote, mak, mbya
y quechua boliviano). Estudios descriptivos, tipolgicos y de contacto lingstico, financed
by the Agencia Nacional de Investigaciones Cientficas y Tcnicas, Argentina, and d
irected by
Dr.Cristina Messineo.
A preliminary version of this article was presented at the XIII Congreso de la Sociedad
Argentina de Lingstica, San Luis, March 2730, 2012.
I am grateful to Cristina Messineo for the enriching exchange of ideas. I also gratefully
acknowledge the very close reading of the manuscript and the helpful comments provided by
the editors of the present volume.

Paola Cneo

1. Introduction
The augmentative is an instance of what some authors call evaluative morphology (Stump 1993; Bauer 1997; Grandi 2002, among others), as it comprises size
encoding and positive/ negative emotional affection features and expresses a type
of evaluation of the linguistic referents or the audience, according to the speakers
judgments and feelings.
The various linguistic encoding mechanisms (affixes, independent lexical
items, tones, etc.) vary from language to language, and while (suffixal) derivation
has traditionally been regarded as the basic strategy to encode evaluation (hence
the term evaluative morphology), this definition can be extended to include syntactic strategies.
In the Toba language, the study of augmentative-related forms is a rich and
complex field, as it is not limited to one form or suffix but comprises both derivational morphology and compounding. It also includes a wide range of meanings, from big size, abundance, or intensity to (generally pejorative) notions of
excess, or mockery. Specifically in the ethnobiological lexicon, these forms are
the preferred resource for creating names of animals and plants, encoding meanings such as hierarchy (more dangerous or of outstanding qualities) or anomaly (rare, unknown, unusual).
The purpose of this paper is to analyse the morphosyntactic and semantic aspects
of the augmentative derivation in the Toba language, as well as its role as a source of
lexical creation and as a linguistic categorization mechanism, i.e. a resource which
makes it possible to group nouns into classes according to their semantic basis.
The article is structured as follows. The introduction offers an overview of the
Toba language (1.1), describes data sources and methodology (1.2), introduces
definitions and presents the morphological, semantic, phonetic and diachronic
properties of evaluative morphology from a typological and cross-linguistic perspective (1.3). This will enable us to frame the study of augmentative forms in Toba
in Section 2. Preliminary conclusions are then presented in Section 3.
1.1 The Toba language
The Toba language (qom laqtaqa)1 belongs to the Guaycuruan language family,
as do Pilag, Mocov, Caduveo, and extinct languages such as Abipn and Mbay.

1. The phoneme inventory of the Toba language and orthographic graphemes [in square
brackets] used in this paper are based on Messineo (2003:36). Consonants plosives: p, t, d, t
[ch], k, g, ; uvular plosives: q, G; fricatives: s, [sh], [z]; nasals: m, n, []; tap: [r/d]; laterals:
l, (palatal); glides: w, y. Vowels: i, e, a, o.

Augmentative in Toba (Guaycuruan)

The Toba group is made up of approximately 70,000 people,2 most of which dwell
in the Gran Chaco area, which encompasses the provinces of Chaco, Formosa
and Salta (Argentina), south-eastern Bolivia, and a region near the district of El
Cerrito (Paraguay). As a result of migration, the Toba people also live in permanent settlements close to big cities (Resistencia, Presidente Roque Senz Pea,
Rosario, Santa Fe, Buenos Aires and La Plata, in Argentina).
From a sociolinguistic point of view, Toba is an endangered language, being
replaced by Spanish in certain fields especially in urban settings , and its speakers are increasingly the victims of social and economic isolation.
The most important previous studies on the Toba language were conducted by
Klein (1978), Censabella (2002) and Messineo (2003). Word formation and noun
classification resources have been the focus of my own doctoral research (Cneo
2013), including the use of morphemes and lexemes encoding notions related to
the augmentative and diminutive.
The following are some of the typological characteristics of the Toba language
relevant for the present paper. Firstly, Toba is characterized by morphological
complexity, both in nouns and verbs. It tends towards polysynthesis and agglutination. The high level of synthesis and the morphophonological changes derived
thereof sometimes hinder morpheme identification. Like in all synthetic languages, Toba has a rich derivational morphology. As regards the order of constituents, the basic, though flexible, word order is AVO and VS. The language makes
a distinction between alienable and inalienably possessed nouns, and attributive
possession is encoded following a head-marking pattern. Additionally, nouns are
inflected for number and gender. Noun gender is a lexical inherent feature in this
language. Verb morphology shows the features of an active-inactive language,
involving three sets of dependent pronominal markers: active, middle and inactive (Messineo 2003; see Carpio 2007 for further reference concerning alignment).
Toba verbs do not inflect for tense or mood but show aspectual distinctions as well
as suffixes expressing direction, position, reflexive and reciprocal. Another feature
is the absence of adpositions or a separate class of adjectives (Messineo 2008) and
adverbs of manner (Cneo et al. 2009).
Like the rest of the Guaycuruan languages, Toba has a closed system of
six demonstrative classifiers placed before nouns and combining features of

2. According to the Complementary Indigenous Peoples Survey (Encuesta Complementaria

de Pueblos Indgenas, ECPI 20042005) performed by the INDEC [Instituto Nacional de
Estadstica y Censos] , out of 69,452 persons who claim to be first generation members and/
or descendants of the Toba people, only about half of them (34,949) can speak and/or understand the indigenous language.

Paola Cneo

c onfiguration (form and position) with deixis (proximity, motion and absence)
(Klein 1978; Vidal 1997; Gualdieri 1998, 2006; Messineo 2003, among others).
1.2 Data sources and methodology
Toba data presented in this paper come mostly from the dapigemlek dialect
(northwest of the province of the Chaco), and were personally collected at successive fieldworks (2002 to this date) with speakers in the province of Chaco (northwestern region: Castelli, El Colchn, El Espinillo, and the region close to the river
Bermejo), and the district of Derqui (Province of Buenos Aires), in Argentina.
These data were gathered by eliciting speech from native speakers and from a vast
and varied text corpus (advice, prayers, humoristic tales, historical accounts, etc.).
In addition, I have based this paper on a corpus of about 1,000 names of plants and
animals,3 which was the focus of my previous work on the language dealing with
ethnobiological vocabulary (Cneo 2013).
1.3 Evaluative morphology
The expression of diminution, augmentation, endearment and contempt by means
of dedicated markers is referred to as evaluative morphology (Stump 1993). Evaluative morphology has been given special attention in descriptions of particular
languages (Krtvlyessy & tekauer (eds) 2011; Corbera Mori 1998, 2002 for studies on American indigenous languages), as well as in regional perspectives and
investigations of entire language families (e.g. Grandi 2002 for Mediterranean languages; Grandi 2011 for Indo-European languages; Mufwene 1980 for the Bantu
family; etc.). Studies dealing with this phenomenon from a typological perspective
focus on the morphological component (Scalise 1986; Stump 1993; Bauer 1997;
Grandi 2002; Grandi & Montermini 2005), on polysemy and semantic complexity (Jurafsky 1996), and on its renowned tendency towards sound symbolism (see
Bauer 1996; Gregov et al. 2010; Krtvlyessy 2011).
The most commonly found semantic, morphological, phonetic and diachronic properties of evaluative morphology from a cross-linguistic perspective
are briefly presented below.
Semantic-functional attributes traditionally considered as evaluative are BIG,
SMALL, GOOD, BAD (Grandi 2002; Grandi & Montermini 2005). These can be
classified as descriptive (SMALL-BIG) and qualitative (GOOD-BAD). While the

3. This corpus was partly developed from existing sources: Buckwalter 2001 [1980]; M
Crovetto 1995.

Augmentative in Toba (Guaycuruan)

first two (BIG, SMALL) may refer to a physical, objective and verifiable attribute
of the entity, the other two (GOOD, BAD) refer to a quality of the entity, which
is subjectively perceived by the speaker. Grandi (2005:190) arranges the possible
evaluative semantic attributes based on these two perspectives. The descriptive perspective comprises the following attributes: young/ old X, short/ long X (temporal dimension); bodily/ physically/ spatially small/ big (physical and spatial
dimension); small/ big quantity of X (quantitative dimension). The qualitative
perspective encompasses the speakers feelings dear X (expressing appraisal and
affection), bad X (expressing contempt), good-/ bad-quality X and attributes of
the referent X to a high degree (intensification); X to a low degree (attenuation).
According to Grandi (2005:191), languages that give formal expression to the
whole range of the possible semantic sub-divisions of evaluation are extremely
rare. On the contrary, languages tend to encode only some of the attributes in the
list. In this paper, we shall observe evaluative attributes encoded by means of augmentative derivation and compounding in Toba.
One of the prototypical properties of evaluative forms is that they alter the
semantics of the bases they are attached to. In this sense, the meaning of evaluative suffixes is attributive (relative to one of the four above-cited basic semantic
attributes) and non-referential (i.e. the basic word and the derived word usually have the same referent, Grandi 2005:193). Furthermore, the application of
evaluative suffixes is usually subjected to semantic restrictions, as they are preferably attached to concrete and count nouns, but rarely to abstract or mass nouns
(Grandi 2005:194).
Finally, another semantic property attributed to evaluative morphology is the
existence of a hyponymy relation between words with an evaluative suffix and the
base word. This means that there might be an implicational relation: saying that, in
Italian, X is a cagnolino, small dog, implies that X is a cane, dog (Grandi 2005:192,
195). As we shall see, this hyponymy relation is not always strictly adhered to in
the Toba language.
In general, evaluative forms are attached to lexically autonomous bases,
although there might be some exceptions (Grandi 2005:188; 2009:47), as in some
of the Toba examples that will be analysed. From a typological standpoint, different central morphosyntactic features typical of evaluative morphology have
been suggested, and they are described below (based on Stump 1993; Bauer 1997;
Grandi 2002, 2005).
a. It has been claimed that evaluative affixes can be attached to words belonging
to different syntactic categories. The relationship between evaluative morphology and the lexical categories of the bases it is attached to has been analysed

Paola Cneo

by Bauer (1997:540).4 According to cross-linguistic evidence, this author suggests the following implicational hierarchy for bases with a tendency to be
attached to augmentative and diminutive morphemes:





Noun > Adjective, Verb > Adverb, Numeral, Pronoun,

Interjection > Determiner

Therefore, evaluative forms can be attached to bases following a productivity order: the further down the hierarchy, the less productive evaluative derivation is. Likewise, the implicational hierarchy reveals that frequency is also
reduced as we move rightwards across the hierarchy (Konstanz Universals
Archive #2009; Grandi 2009:48).
It has been claimed that evaluative affixes do not tend to alter the syntactic
category of the base word. The two attributes mentioned in (a) and (b) represent the so-called categorial neutrality (Grandi 2005:194), as evaluative suffixes may select from bases belonging to various syntactic categories with no
change to their categorial status. While this seems to be one of the most problematic properties in evaluative morphology, Bauer (1997:549; in: Grandi
2005:193) notes that it can be semantically and functionally explained.
Although evaluative suffixes preserve at least one of the morphosyntactic feature specifications of the base (Stump 1993:1213), they can change the gender of the base word or distinguish between count and mass nouns.
Evaluative suffixes are not relevant for syntax. They do not specifically trigger
any evaluative agreement and they are not required by any syntactic context.
Evaluative suffixes even those with different meanings may be successively applied, as in Spanish narig-on-cito nose-aug-dim/ big nosed (familiar
Finally, it has been claimed that evaluative morphology can be applied before
or after inflectional or derivational morphology. However, it is widely accepted
that evaluative suffixes have a more external position than other derivational
morphemes, but a more internal position than inflectional suffixes. In this
sense, evaluative suffixes are usually placed in a grey area (Grandi 2005:192)
between derivation and inflexion,5 as some of the properties cited above are
typically derivational (they change the semantics of the base, allow for iterative applications, and are not relevant to syntax), while others are typically
inflectional (the syntactic category of the base remains unchanged). In fact,

4. Based on previous claims by Ettinger (1974) and Nieuwenhuis (1985).

5. Scalise (1986) suggests the existence of an evaluative morphological subcomponent
placed between inflection and derivation (cf. also Stump 1993).

Augmentative in Toba (Guaycuruan)

evaluative morphology can be derivational in some languages (e.g. IndoEuropean languages) and inflectional in others (e.g. Bantu languages; see
Mufwene 1980 for discussion).
In addition, in noun phrases, diminution and augmentation have a borderline
position, mostly at the crossroads of gender, number and the count/ mass distinction, possibly due to their peculiar status at the cognitive and functional levels (Di
Garbo 2011:1).
Studies of evaluative morphology focus on synaesthetic sound symbolism,6
referring to the process through which certain sound segments are consistently
selected to represent the visual, tactile or proprioceptive properties of objects (e.g.
position, orientation, movement, size and shape). Synaesthetic sound symbolism
can be defined as acoustic symbolization of non-acoustic phenomena (Hinton et
al. 1994:4). This means that the phonological form or certain phonetic properties
of a word are considered iconic for some sensorial quality of the meaning or referent (Beck 2008:6). The correlation between vowel sounds and size is an example
of synaesthetic sound symbolism.
According to Payne (1997:110), there is an apparently universal iconic tendency in diminutives and augmentatives: diminutives tend to contain high front
vowels, whereas augmentatives tend to contain high back vowels (Konstanz Universals Archive #1926). Based on a genetically diverse sample of fifty languages,
Bauer (1996) discusses these claims and concludes that there is no evidence of
any universal pattern for sound symbolism in evaluative morphology; however, it
seems to be restricted to particular language families, including Indo-European.
From a diachronic perspective, there is a well-known (probably universal) tendency, which establishes a link between the expression of the parental
relation and the emergence of the diminutive meaning (Heine & Kuteva 2002;
Grandi 2005, 2011, among others). However, the relation between terms of kinship mother/ father together with the notion of augmentative as I discuss in
Toba has been given less attention.7 In a diachronic survey of some branches of
the Indo-European family, Grandi (2011) defines the development of augmentatives as a frequent but an unstable phenomenon since many different tendencies
6. Sound symbolism or phonetic iconicity is generally defined as a direct association
between the sound and meaning of a word.
7. For example in Kana (Cross River language, south-eastern Nigeria), the augmentative
prefix k- was probably derived from the word for mother (Ikoro 1996:61 in Aikhenvald
2007:58). Likewise, in many Tibeto-Burman languages and other languages from East and
Southeast Asia (Matisoff 1992), the words mother and child are also grammaticalized to
serve as augmentative and diminutive, respectively.

Paola Cneo

occur (Grandi 2011:21). The author finds a process of innovation, which has led
to the emergence of augmentative suffixes from agentive, collective, or locative
2 The augmentative in Toba
Augmentatives in the Toba language involve both derivational (i.e. suffixing) and
analytical (i.e. lexical) means. The first subsection (2.1) shows the morphological
augmentative devices: the pejorative suffix -naq (2.1.1); the attributive suffix -day
(2.1.2); the non-productive suffix -alo (2.1.3); and the temporal quantifier and
emphatic -o:m (2.1.4). In subsection (2.2), I present the analytical augmentative:
constructions with kinship nouns l-taa poss3-father and l-atee poss3-mother
(2.2.1) and derived forms (2.2.2).
2.1 Morphological devices
2.1.1 The (pejorative) augmentative suffix -naq
The suffix -naq has recently been recognized as augmentative (Messineo 2003:111).
The feminine form, -naGa, is marked by the suffix -a.8 While these suffixes derive
attributive nouns with a possible descriptive meaning, they are often used to connote pejorative meanings or jokes, and even insults. The suffix is attached to noun
bases denoting body parts in order to form attributive nouns with pejorative connotations. The gender of the suffix corresponds to the gender of the (human) referent, as shown in the following examples:




your nose


big-nosed man


big-nosed woman


his/her foot

8. Voiced velar /g/ and uvular /G/ consontants are required in non-final word position (see
Messineo 2003:54f.).

Augmentative in Toba (Guaycuruan)



big-footed man


big-footed woman


his/her hair


hairy man


hairy woman

As shown in the examples above, attributive nouns, which are formed with this
suffix, lose the inalienable possession markers (which are present in the (a)
The suffixes -naq (m) and -naGa (f) may be applied both to compound words,
as in example (5), and to Spanish loanwords, as in example (6):


bearded man


bespectacled woman

To a lesser extent, the allomorphs -saq (m) and -saGa (f) have been documented,
with exactly the same value, in nouns derived from the suffix -ek (possibly nominalizer, see Censabella 2005:200):9


potbellied man

9. The suffix occurs as -ege (voiced velar) in non-final word position (see Footnote 8). In
Mocov, a related language, only the form -saq/ -saGa has been documented (after nmlz ek)
as having an augmentative value (Gualdieri 1998:177). However, the occurrence of the nominalizer suffix requires further consideration.

Paola Cneo


big-mouthed woman

Nouns derived by suffixing -naq ~ -saq (m) or -naGa ~ -saGa (f) combine a physical dimension (physically/ spatially BIG) with a quantitative dimension (BIG
quantity of X), depending on whether they are count nouns (such as nose or foot)
or mass nouns (such as hair or beard). In both cases, an evaluation on the part of
the speaker is expressed and therefore, this suffix combines several of the attributes suggested by Grandi (2005), as was mentioned above (1.3).
The augmentative may possibly be found in nouns for fauna, although the
meaning of the bases is opaque. We only have examples in the masculine formed
with -saq and -naq:

qoige-saq lizard, Tupinambis sp.11


keedege-saq a type of lizard (climber), undet.


kyo-saq a fish similar to a bluegill, undet.


qage-saq ant, Formicidae


saashi-naq golden dorado, Salminus brasiliensis

2.1.2 The attributive suffix -day

The suffix -day (~ -ay ~ -oy)11 is widely used in the language and allows for
derivation of attributive nouns from nominalized verbal bases and, to a lesser
extent, from simple noun bases, in order to create concepts expressing the referents features, qualities or characteristics, for example yi-ako (3-fish) he fishes
> yi-ak-oG-oy-k (3-fish-nmlz-attr-masc) fisherman. Based on formal characteristics similar to those of nouns and verbs, attributive nouns have been defined
as an intermediate category (Messineo 2003:112, 2008), with the functions of

10. The scientific designations for certain botanical and zoological species are taken from
Martnez (2009) and Medrano et al. (2011), respectively. I thank both Gustavo Martnez and
Celeste Medrano for helpful information concerning botanical and zoological references.
The following abbreviations are used: undet. (undetermined species or no data available); sp.
(several species from the same type).
11. Allomorphy -day ~ -ay seems to be morpho-phonologically determined. In general, -day
occurs when attached to a noun base ending in a vocalic phoneme, and -ay occurs after velar
and uvular consonants and when attached to a nominalized verbal base; nevertheless, this
distinction needs further research. On the other hand, as a result of vowel harmony, the suffix
absorbs the features of the root vowel when this is the back vowel /o/, and occurs as -oy.

Augmentative in Toba (Guaycuruan)

prototypical adjectives.12 Attributive nouns inflect for gender the feminine is

unmarked while the masculine takes the suffix -k.
When the suffix -day(k) is attached to noun bases referring to body parts, it may
have a pejorative augmentative value, similar to the value observed in -naq/ -saq
constructions, and in some cases it may even be attached to the same bases (compare
(15) and (16) to (3) and (7),13 respectively).





big-headed man


big-headed woman


big-footed man


big-footed woman


potbellied man


potbellied woman

The suffixes -naq and -naGa seem to have a more clearly pejorative meaning (i.e.
mockery or insult), however, their frequency of use and pragmatic role in specific
contexts has not been studied in detail yet.
The suffix -day(k) may also be attached to verbal bases. If it is attached to verbal bases, the attributive suffix bearing the augmentative value allows for temporal
quantification of the action referred to by the verbal base, and highlights certain
actions or habits which society considers to be negative. In (17), kiaGay(k) big
eater refers to a person who eats all day, at any time, and who may have a disorder (caused by avoidance of the taboo of sharing food with ones sister or mother
during her period). Similarly, in (18), ochaGay(k) sleepyhead means a person

12. For a discussion of the adjective category in Toba, see Messineo (2008).
13. The semantic or pragmatic contrast between both suffixes is still unclear.

Paola Cneo

who sleeps all day (damaged as a result of shamanism or avoidance of a cultural



big eater man


sleepyhead woman

The suffix -day(k) is highly recurrent in the formation of names of animals and
plants, since it makes it possible to make reference to qualities or attributes of the
designated entities, e.g. ikiaGay (> da-iki 3-float) to name a type of water bird
(Phalacrocorax), or onaGanaGay (> do-onaGan 3-sing) with reference to the
calandra lark (Mimus saturninus).
2.1.3 The non-productive augmentative suffix -alo
The suffix -alo belongs to the synchronically unproductive morphology of the language, and speakers claim that it is an ancient form. It is used for nouns referring
to mythical entities, such as the great mythical fire in the account on the origin of
animal species:


the great (mythical) fire

When names of animals are derived by using -alo, the suffix may, in some cases,
add the notion of larger size/ dimension, as in the following examples, where
there is a clear size-based distinction between two types of cats and two types of
armadillo (the (b) examples refer to the bigger animals):



Geoffroys cat, Leopardus geoffroyi
kopayk-alo ocelot, dwarf leopard, Leopardus pardalis



napam different species of armadillo15,

Chaetophractus sp. and Euphractus sexcinctus


pam-alo giant armadillo, tatou, Priodontes maximus

14. The name napam refers to the large hairy armadillo (Chaetophractus villosus), the
screaming hairy armadillo (Chaetophractus vellerosus), or the six-banded armadillo or the
yellow armadillo (Euphractus sexcinctus).

Augmentative in Toba (Guaycuruan)

The suffix -alo is also used to name types of similar animals, where a characteristic of their behaviour or habitat (and not necessarily size) is a distinctive
feature. Semantically speaking, the suffix is similar to the terms latee his/her
mother, and ltaa his/her father, discussed in Section 2.2.1, since it highlights an
outstanding feature, as compared to another similar entity, i.e. it stresses the referents attributes. The qualitative dimension present in these augmentative attributes
respond to social assessments expressed in these lexicalized names, as in Examples
(22) and (23) below. In (22), pyoGodalo is a very strong kind of dog living in the
forest.15 Similarly, the kaayalo (23) is a horse living solely in the water, with long
manes and red eyes (see Medrano 2012:2413) that comes out when the Bermejo
River is about to rise, and it is considered a powerful being.



pyoGod-alo crab-eating raccoon, Procyon cancrivorus



kaay-alo the water horse (powerful being)

Finally, in certain nouns for birds and ophidians derived with -alo it is not possible
to identify the meaning of the base:

matak-alo flamingo, undet.


lek-alo Argentine boa or yellow anaconda, Boa occidentalis, Eunectes



sok-alo ~ tak-alo type of bird or snake


chid-alo a type of bird 

(Buckwalter 2001:175, 177)

(Buckwalter 2001:9)

2.1.4 The augmentative suffix -o:m

Finally, the form -o:m,16 involving the long back mid vowel [o:] plus a final bilabial
consonant, has an emphatic and expressive function, and vowel duration is also
based on the emphasis conveyed by the speaker.17 This suffix is attached to nominal and adverbial bases.
15. A single pyoGodalo is capable of fighting ten dogs and usually wins the fight (Medrano
et al. 2011:645).
16. According to my data, the use of this suffix seems to be limited to the western area at the
Chaco province.
17. According to Messineo (2003:56, 249), vowel lengthening has an uncertain phonemic
status in Toba, and it can be regarded as a prosodic phenomenon commonly used for
expressive purposes (to express emphasis, exhortation, surprise, exclamation, interrogation,
or other communicational features).

Paola Cneo

In the following examples, the suffix is attached to a noun expressing time

and acts as a temporal quantifier (length). In (28) the speaker expresses disagreement with the fact that her brother sleeps for many hours; the augmentative suffix
emphasizes this quality of lasting a long time:


saishet da
d-ooche-ta naaGa-lo:m18
comp 3-sleep-asp day-aug
he cannot sleep all day long

In contrast, example (29) is part of an account on an expert hunter pretending to

have been caught by a tiger so as to be able to hunt the animal. By means of the
augmentative suffix, the narrator intends to emphasize the fact that the tiger will
eat the hunter a long time afterwards (enough time for the hunter to be able to
react, catch the tiger unawares, and hunt it in the end):


keeka l-ki-aGak
3-look.for dem
poss3-eat-nmlz 3-be.comfortable later-aug
looked for [a place] where to eat comfortably some time later

In the spatial domain, the adverb kero:m far away (< ked-o:m there-aug) shows
that the suffix has been lexicalized, the lexical base not being synchronically identifiable any more.
I also documented the use of this suffix applied to nouns in order to express
the speakers emphasis and expressivity. The following example represents a conversation between two adults, one of whom has a young daughter or qaaole.
According to ethnographic studies on the stages of the lifecycle among the Toba
people (Hecht 2010:166170), a woman undergoing the stage starting with her
first period and ending with the birth of her first child regardless of her biological age is called qaaole, a term formed by the lexicalized diminutive suffix
-ole. A girl defined as a qaaole is believed to be ready to become independent,
have her own family and start a life almost like an adult (Hecht 2010:170). In our
example, the speaker in (b) answers emphatically by stressing his daughters social



aq qaa-ole
still young.woman-dim
Is she still young?

18. The suffix appears as -lo:m (Examples (28) and (29)) possibly remnants of a diminutive
morpheme. In (29) consider the synchronic use of the following adverbs: qome later, qomele
a bit later, qomelo:m some time later/ much later. See also Examples (30)(33) where the
sequence dim-aug is very clear.

Augmentative in Toba (Guaycuruan)


[Yes,] very young! (emphatic)

Example (30) see also (31)(33) below , shows that evaluative suffixes with different meanings (dim, aug) may be successively applied, as already mentioned in
Section 1.3.
The augmentative suffix -o:m is also applied to other very common words
referring to culturally prominent social relations, such as young man, elderly
man, and elderly woman. In these cases, the augmentative is applied externally
with respect to a diminutive suffix and it involves an intensifying function close to
elative forms (e.g. in Spanish).


young man


young man


extremely young man


extremely old, very old man


extremely old, very old woman

2.2 Constructions with kinship nouns

2.2.1 Compounds with ltaa his/her father and latee his/her mother
The terms l-taa poss3-father/ his/her father and l-atee poss3-mother/ his/her
mother encode notions of augmentation and similarity, resemblance or approximation when they are involved in the formation of compound nouns i.e. phonologically, morphologically and semantically autonomous units in this language
(see Cneo 2013).19 This type of compound is mainly used to refer to body parts
(34)(35) and to form neologisms (36).
19. These are some of the criteria that make it possible to differentiate compound nouns in
Toba: the presence of a single primary accent and the behaviour of velar and uvular (voiced intervocalic) consonants among constituents; the impossibility of introducing an element (e.g. a

Paola Cneo




his/her toe


l-qona- latee
his/her big toe (lit. toes mother)


his/her intestine


his/her gut (lit. intestines father)

(36) a. wataGan-aq

police officer (lit. the one who waits)


captain, sergeant (lit. police officers father)

In addition, the kinship terms are widely present in the ethnobiological domain, as
they are a preferred resource when forming names of animals and plants as we
have seen in Messineo & Cneo (2011) and Cneo (2013). In Toba, nouns that are
derived by drawing upon this resource categorize both zoological and botanical
entities, and their role is to distinguish between somewhat similar species.20

demonstrative) between the terms of the compound and exchanging the third person possessive prefix for another possessive person; the inflectional morphology marker (e.g. number)
in the compound as a unit by means of a suffix (the first constituent cannot be inflected); the
lack of semantic compositionality in many cases; and the fixed order possessor-possessed,
not prototypical for possessive NPs in this language (Cneo 2013). Furthermore, in Cneo
(2012, 2013) I have shown that kinship compound nouns may be considered class terms based
on their formal and functional properties. While they have a clear lexical origin and may also
occur as independent nouns, in many cases they have lost their original meaning and are part
of emerging grammaticalization processes.
20. From a social and cultural perspective, the terms latee his/her mother and ltaa his/her
father are permeated by a strong meaning with social relevance, as they refer to the owners,
lords, or parents of animal and plant species. These are non-human entities which dominate
and protect species and to which shamans are linked (see Miller 1979; Wright 1992, 2005; Tola

Augmentative in Toba (Guaycuruan)

In the examples below,21 the compound noun emphasizes the larger size of a
type of fish (37), or herbal plant (38).




type of fish, Chichlaurus dimerus

type of fish, Cichlasoma sp. (larger than qapoi)

(38) a. awaqpi
grass (generic)

b. awaqpi-ltaa lemon grass Cymbopogon citratus (of a considerable size)

Regardless of size (physical dimension), the kinship terms latee (f)/ ltaa (m)
form nouns referring to more dangerous species for humans, like the coral snake,
which is poisonous and lethal (39)b, or like a type of nettle which can burn and
poisen you when touched (40)b (examples are from Messineo & Cneo 2011:153):



fake coral snake, Oxyrhopus guibei, O.

rhombifer rhombifer
wizik-latee coral snake, Micrurus pyrrhocryptus





nettle, Urtica circularis and others

type of nettle

Other nouns formed by utilizing this resource refer to outstanding abilities or

qualities in the named species. In (41)b it means a type of tree, which, although
shorter than the prototype (41)a, bears fruit with a stronger smell.



type of tree undet.
kalmayk-latee type of tree undet.

(Buckwalter 2001:4)
(Buckwalter 2001:4)

Finally, some compound nouns formed with these terms of kinship refer to species
which are less numerous (more unusual or rarer) in the area or have recently been
introduced to the culture (42)b:
(42) a. teesaq type of bee, undet.

b. teesaGaltaa ~ teesaq-ltaa

domestic (queen) bee, Apis mellifera (European type, introduced)

To sum up, these terms function as derivational morphemes that categorize an

entity in association with another entity, which is conceptualized as prototypical
or more widely known. The use of these terms encodes notions of augmentation
and similarity, resemblance or approximation. In all cases, they emphasize a certain attribute of the referent. In this sense, their role is similar to that identified for
the -alo morpheme in Section 2.1.3.

21. For further examples, see Messineo & Cneo (2011:1523) and Cneo (2013:25694).

Paola Cneo

We can also observe that kinship nouns like ltaa his/her father and latee
his/her mother are part of emergent grammaticalization processes, since they
have lost their original semantics and have acquired a more functional meaning in
certain contexts. In addition, one of their functions is to indicate a change in the
lexical meaning of the stem; in this sense, compounds formed with these terms are
sometimes lexicalized.
In the rest of the Guaycuruan and Mataguayan languages, the use of evaluative
morphology in word formation has been widely recorded by Messineo & Cneo
(2011), particularly though not exclusively in the ethnobiological vocabulary.
Moreover, similar examples of evaluative morphology used to form names of animals can be found in indigenous languages of North America, such as the use of
the augmentative form of rabbit for sheep in Natchez and Tunica (both from the
Gulf language family) (Mithun 1999:468, 532; in Bauer 2004:9), or in Spanish,
where the use of the augmentative ratn mouse (from rata rat) may also be an
example of this kind.
Beyond the field of word formation, the augmentative and diminutive derivations may also serve as linguistic categorization mechanisms, due to their
evaluative function (see Cneo 2013 for Toba). In this respect, they are widely
used suffixes in the language allowing for specific lexical domain formation and
organization. In order to explain the classifying function of terms such as latee
his/her mother and ltaa his/her father in Toba, I follow the idea suggested by
Scarpa (2010:1723) to analyse botanical names in Chorote (Mataguayan). This
author suggests that terms of this kind do not denote a hierarchical relation with
the generic noun they are attached to nor do they refer to a subgeneric category
(species or type), but they form kinds with a certain family resemblance, in the
Wittgensteinian sense.
Likewise, the terms for father and mother in the ethnobiological vocabulary are used to refer to a familiar, related, similar entity with no hierarchical
relation. The compound noun formed with these terms may share similar shape
features with the prototype (referred to by the base) and be slightly different from
it regarding certain aspects of behaviour or habitat (i.e. they might be morpho
types, Celeste Medrano p.c.). Finally, apart from this notion of similarity, there is
the idea of something extra resulting from the augmentative notion (extra size,
extra power, extra colour, extra danger, and extra exotic, rare or unknown). In this
sense, these terms combine classification (approximation or similarity in relation
to a prototype) with augmentation.
2.2.2 Constructions with extended kinship nouns
The attributive nouns l-taa-day-k poss3-father-attr-m/ big (m) and la-tee-day
poss3-mother-attr/ big (f) are forms derived from the nuclear kinship terms

Augmentative in Toba (Guaycuruan)

father and mother (see 2.2.1) through the suffix -day (see 2.1.2).22 As noun
modifiers within the noun phrase, these terms preserve the grammatical gender of
the head noun. As stated above, the ability of preserving at least one of the morphosyntactic properties of the base is typical of evaluative morphology (Stump
1993:13) (see 1.3). The next examples show that there is gender agreement within
the noun phrase, both in the modifier noun adding the augmentative character
and in the demonstrative (feminine gender is marked through the prefix a-).


a-na -adoo lateeday

f-dem poss1-hat big.f
my big hat


dem tree/wood big.m
the big wood

The modifying nouns that express augmentative notions can be attached to concrete (45) or abstract nouns (46), and encode attributes for the physical and spatial
dimensions (physically/ spatially BIG), according to Grandis characterization
(Grandi 2005):


nozik ltaadayk
house big.m
big house


poss1-be.happy-nmlz big.m
my great delight (I am very delighted)

The terms ltaadayk (m) and lateeday (f) big are especially used for modifying
nouns referring to natural environments and atmospheric phenomena expressing
large spaces:


a-i lapel lateeday

f-dem lake big.f
the big lake


a-so tala lateeday

f-dem river big-f
the big river (= the Bermejo River)

22. As stated above (2.1.2), attributive nouns express adjectival concepts, they possess formal
characteristics of nouns (they inflect for gender and number), although they lack a possessive
marker. Syntactically speaking, they may function as noun modifiers and nominal predicates
(Messineo 2008).

Paola Cneo


awiaq la-tee-day
forest poss3-mother-attr
big forest

In the last two examples, (48) and (49), meanings are lexicalized as they have specific referents and become toponymic references: the Bermejo River is the most
important river in northwest Chaco. Awiaq lateeday, on the other hand, is a vast
area of native forest covering more than 40,000 km2, in the northwest of the province of the Chaco, called El Impenetrable (this region was named after its dense
vegetation, which makes it difficult to access). Similarly, in (50), the construction
may refer both to a big, abundant meal and to a social occasion when vast amounts
of food are eaten, i.e. a party:


poss3-eat-nmlz big.m
his/her big meal, his/her feast, his/her party

The use of these modifier nouns based on terms of kinship can also encode a qualitative perspective by denoting intensifying features in relation to the referents
attributes (X to a high degree), as in the following example, where the term is
applied to an abstract noun:


poss1-disease big.m
my severe/ serious disease23

The contrast between these terms and the suffix -naq (see above 2.1.1) becomes
evident when attaching nouns referring to body parts. While the modifier term
denotes big size for the noun (52)a, suffixes add pejorative connotations to the attributes of the human referent (showing an agentive derivational function)24 (52)b:



mouth big.m
big mouth


big-mouthed man


big-mouthed woman

23. A serious disease means an incurable or hard-to-cure condition, such as tuberculosis or

Chagas disease.
24. A similar and possibly related suffix -aq functions as an agentive nominalizer (see
example (36)a.

Augmentative in Toba (Guaycuruan)

The contrast between the use of the attributive nouns and the non-derived kinship
nouns in compounds makes it possible to specify differential functions. Attributives (Examples (53)(54)a) refer to augmentative descriptive features: big quantity of X for the nouns referring to rain or wind (inanimate and mass nouns).
However, examples in (53)(54)b are created using (non-derived) terms of kinship
and form compound (i.e. lexicalized) nouns to refer to a different and particular
entity, such as the Flood or the tornado:




awot ltaadayk
rain big.m
big rain


the Flood


laat ltaadayk
wind big.m
generous, abundant wind; a lot of wind



Similarly, while kos lateeday (55)a means the pig (kos) that leads the gang and has
a big body, koslatee (55)b refers to the white-lipped peccary or Tayassu pecari,
which is characterized as extremely dangerous (it is bad, fierce, unknown, with
noisy tusks, notably different from the domestic pig kos).



kos lateeday
pig big.f
leader pig


white-lipped peccary, Tayassu pecari

In short, as observed in the examples, while ltaadayk and lateeday have a

typically augmentative value (by lexical means), ltaa (m) and latee (f) have a
privileged role in lexical creation and noun categorization (as shown in 2.2.1).

3. Preliminary conclusions
This paper described the morphosyntactic and semantic aspects of augmentative
forms in Toba. Formally speaking, the augmentative in this language involves different linguistic encoding mechanisms, ranging from derivational morphology

Paola Cneo

(suffixes -naq; -day; -alo; -o:m) to analytical forms: kinship nouns such as latee
mother and ltaa father, forming compound nouns, and derived attributive
forms such as lateeday (f) and ltaadayk (m), being involved in attributive noun
Augmentative forms in Toba are most commonly and productively attached
to noun bases. However, according to our data, the suffix -day(k) may also be
attached to verbs (preceded by -aGa nmlz), and -o:m may be attached to adverbs,
thus both suffixes are cross-categorial morphemes. In terms of an implicational
hierarchy for bases tending towards attachment of augmentative and diminutive
morphemes, as proposed by Bauer (1997), augmentative forms in Toba can be
represented as in Table 1.
Table 1. Lexical categories for augmentatives in Toba
Noun >

Verb >






ltaa/ latee

ltaadayk/ lateeday

As expected (see Section 1.3), evaluative augmentative forms do not generally

alter the syntactic category of the base they are attached to. Nevertheless, the suffix -o:m may sometimes act as an adverbializer, deriving adverbs from nouns; and
-day(k) may act as an adjetivizer deriving attributive nouns.
Regarding gender (see Table 2), we have noticed that augmentative morphology is transparent to the base gender in the case of ltaa/ latee, ltaadayk/
lateeday, -alo and -o:m. Attributive suffixes -naq/-naGa and -day(k), in turn,
define grammatical gender by means of semantic agreement with the (human)
male or female referent. Unexpectedly (see Section 1.3), the meaning of these
evaluative suffixes is referential: the basic word and the derived word do not have
the same referent.
Concerning the phonetic iconicity phenomenon, we have observed that augmentative forms in Toba involve mainly back mid /o/ and central low /a/ vowels
(see Table 3). Consistent with Bauers findings (1997), we can also claim that Toba
is a language where augmentative markers do not contain the front high vowel /i/
nor palatals.

Augmentative in Toba (Guaycuruan)

Table 2. Augmentative and gender in Toba

Augmentative form





Transparent to the
base gender

Agreement with the

(human) referent



















Table 3. Phonetics of augmentatives in Toba


nasal + central low + uvular


central low + glide


central low + lateral + back mid


long back mid + bilabial


lateral + alveolar + central low + glottal + central low


lateral + central low + alveolar + front mid + glottal + front mid

We have observed that the lexical morphemes latee mother and ltaa father,
when they are part of nominal compounds, acquire grammatical functions with
augmentative attributes. In addition, the language grammaticalizes these basic
social relationships (mother, father) to serve as nominal classification devices
(mother, father > augmentative; mother, father > classifier).
Semantically speaking, the physical and spatial dimension (physically, spatially, bodily big X) is mainly expressed in an analytical way through the modifier noun lta adayk (m)/ late eday (f), which can be glossed as big. In contrast,
the attributive suffixes -naq/-naGa and -day(k) involve pejorative connotations
when attached to nouns referring to body parts or an action pervaded by cultural
assessments. The augmentative emphasizes and stresses an outstanding attribute
or habit of the referent, giving rise to the speakers feelings. These also have strong
involvement in the use of the suffix -o:m, which acts as a quantifier when applied
to nouns and adverbs expressing time (temporal dimension: long X), and as an
intensifier when applied to certain kinship terms.
Finally, the suffix -alo and the kinship nouns late e (f)/ lta a (m) are resources
involved in word formation, creating new lexemes for new concepts, which make

Paola Cneo

it possible to categorize entities relative to an outstanding feature or attribute

(intensification), as compared to a similar entity, named by the base. They serve
to distinguish among different species. In these cases, there is no strict hyponymy
relationship between the base noun and the derived noun but a similarity (nontaxonomic) relationship.25 As observed in previous studies (Messineo & Cneo
2011), the relationship among the augmentative, noun classification and ethnobiological noun formation seems to be typical of languages from the Guaycuruan
family and the Chaco area.
In summary, documented and analysed forms in Toba involve two of the most
important functions of evaluative morphology in languages of the world: intensification and affection. Augmentative attributes in Toba are summarized in Table 4.
Table 4. Meanings of augmentatives in Toba
Physical and spatial dimension

physically, spatially, bodily big X

Temporal dimension

long X


very X

New concept
Attributes of the referent

-alo and ltaa/ latee

similar to X but with a certain attribute to a high degree

Speakers feelings
Attributes of the referent
Descriptive and pejorative

-naq/-naGa and -day(k)

one who is/ has/ makes X to a high degree

It is perhaps not surprising that the augmentative category involves not only
the languages morphosyntactic and semantic levels but also pragmatic-discursive
ones, as it may either state or create discursive contexts, communicative events or
situations, or reveal aspects of the social and cultural organization. Future studies
should therefore investigate a pragmatic approach to different uses of the augmentative forms in the Toba language.

asp = aspect; attr = attributive; aug = augmentative; comp = complementizer;
dem = demonstrative; dim = diminutive; f = feminine; m = masculine; neg =
negation; nmlz = nominalizer; pl = plural; poss = possessive

25. For example, a kopaykalo ocelot is not a kind of kopayk Geoffroys cat (although this
statement needs to be supported by a more profound study on native Toba taxonomy).

Augmentative in Toba (Guaycuruan)

Aikhenvald, Alexandra. 2007. Typological distinctions in word-formation. In Language typology and syntactic description, Vol. III: Grammatical Categories and the Lexicon, 2nd edn,
Timothy Shopen (ed), 165. Cambridge: CUP.
Bauer, Laurie. 2004. A glossary of morphology. Edinburgh: EUP.
Bauer, Laurie. 1996. No phonetic iconicity in evaluative morphology. Studia Linguistica 50:
189206. DOI: 10.1111/j.1467-9582.1996.tb00349.x
Bauer, Laurie. 1997. Evaluative morphology: In search of universals. Studies in Language 21(3):
533575. DOI: 10.1075/sl.21.3.04bau
Beck, David. 2008. Ideophones, adverbs, and predicate qualification in Upper Nexaca Totonac.
International Journal of American Linguistics 74(1): 146. DOI: 10.1086/529462
Buckwalter, Alberto. 2001 [1980]. Vocabulario toba, rev. edn.Formosa IN: Equipo Menonita
Mennonite Board of Missions.
Carpio, Mara Beln. 2007. Sistemas de alineacin en toba (familia guaycur, Argentina). MA
thesis, Universidad de Sonora, Hermosillo, Mxico.
Censabella, Marisa. 2002. Descripcin funcional de un corpus en lengua toba (Familia
Guaycur, Argentina). Sistema fonolgico, clases sintcticas y derivacin. Aspectos de
sincrona dinmica. PhD dissertation, Universidad Nacional de Crdoba.
Censabella, Marisa. 2005. Nominalizacin deverbativa en toba. Revista Anclajes 9(9): 189211.
Universidad Nacional de La Pampa.
Corbera Mori, ngel H. 2002. Morfologa apreciativa: aspectos contrastivos entre aguaruna y
castellano. In Contribuciones a la lingstica aplicada en Amrica Latina, Carmen Curc
(ed), 295306. Mxico DF: UNAM, Centro de Enseanza de Lenguas Extranjeras.
Corbera Mori, ngel H. 1998. Morfologia avaliativa: Diminutivos e aumentativos em Aguaruna
(Jvaro). In Proceedings of the XLVI Seminrio do GEL, Vol. I: 134134. So Jos do Rio
Preto, So Paolo, Brazil.
Cneo, Paola. 2012. Trminos especializados en la composicin de nombres etnobiolgicos en
toba (guaycur). Revista UniverSOS: Revista de Lenguas Indgenas y Universos Culturales 9,
133146. Universitat de Valncia.
Cneo, Paola. 2013. Formacin de palabras y clasificacin nominal en el lxico etnobiolgico en
toba (guaycur). Lincom Studies in Native American Linguistics 68. Munich: Lincom.
Cneo, Paola, Dante, Patricia & Tacconi, Temis. 2009. Tipologa lxica. Una aproximacin a
dos lenguas chaqueas: Toba (guaycur) y mak (mataguaya). Cadernos de Etnolingstica
1(2). (9 February 2014)
Di Garbo, Francesca. 2011. Evaluative morphology in noun phrases: a typological study. Paper
presented at Third Conference of the Scandinavian Association for Language and Cognition (SALC III), Copenhagen, June 1416.
Ettinger, Stefan. 1974. Diminutiv- und Augmentativbildung: Regeln und Restriktionen. Tbingen:
Grandi, Nicola. 2002. Morfologie in contatto: Le costruzioni valutative nelle lingue del Mediterraneo. Milano: Franco Angeli.
Grandi, Nicola. 2005. Sardinian evaluative morphology in typological perspective. In Sardinian in typological perspective, Ignazio Putzu (ed), 188209. Bochum: Universittsverlag
Grandi, Nicola. 2009. Restrictions on Italian verbal evaluative suffixes: the role of aspect and
actionality. York Papers in Linguistics Series 2:4666. YEMM Special issue.

Paola Cneo
Grandi, Nicola. 2011. Renewal and innovation in the emergence of Indo-European evaluative morphology. In Diminutives and Augmentatives in the Languages of the World. Lexis
6:525. (9 February 2014).
Grandi, Nicola & Montermini, Fabio. 2005. Prefix-suffix neutrality in evaluative morphology. In
Morphology and Linguistic Typology. On-line Proceedings of the Fourth Mediterranean Morphology Meeting (MMM4) Catania, September 2123 2003, University of Bologna, Geert
Booij, Emiliano Guevara, Angela Ralli, Salvatore Sgroi & Sergio Scalise(eds), 143156. (9 February 2014)
Gregov, Renta, Krtvlyessy, Lvia & Zimmermann, Jlius. 2010. Phonetic iconicity in the
evaluative morphology of a sample of Indo-European, Niger-Congo and Austronesian languages. Word Structure 3(2): 156180. DOI: 10.3366/word.2010.0003
Gualdieri, Beatriz. 1998. Mocov (Guaycur). Fonologia y morfossintaxe. PhD dissertation,
Universidade Estadual de Campinas, San Pablo.
Gualdieri, Beatriz. 2006. Clasificadores guaycures: Un desafo para la lingstica. Paper presented at Actas del I Encuentro de Lenguas Indgenas Americanas (ELIA), Santa Rosa, La
Pampa, june 1517.
Hecht, Ana Carolina. 2010. Todava no se hallaron hablar en idioma: Procesos de socializacin
lingstica de los nios en el barrio toba de Derqui, Argentina [Lincom Studies in Sociolinguistics 9].Munich: Lincom.
Heine, Bernd & Kuteva, Tania. 2002. World Lexicon of Grammaticalization. Cambridge: CUP.
DOI: 10.1017/CBO9780511613463
Hinton, Leanne, Nichols, Johanna & Ohala, John (eds). 1994. Sound Symbolism. Cambridge:
Jurafsky, Daniel. 1996. Universals tendencies in the semantics of the diminutive. Language
72:53378. DOI: 10.2307/416278
Klein, Harriet. 1978. Una gramtica de la lengua toba: Morfologa verbal y nominal. Montevideo:
Universidad de la Repblica.
Konstanz Universals Archive. 19982009. Universitt Konstanz, Frans Plank Archivist General.
At: (17 February 2012)
Krtvlyessy, Lvia. 2011. A cross-linguistic research into phonetic iconicity. In Diminutives and
Augmentatives in the Languages of the World. Lexis 6:2739.
IMG/pdf/Lexis_6_-_Kortvelyessy.pdf (9 February 2014)
Krtvlyessy, Lvia & tekauer, Pavol (eds). 2011. Diminutives and Augmentatives in the Languages of the World. Lexis 6.
pdf (9 February 2014)
Martnez, Gustavo. 2009. Fitonimia de los tobas bermejeos (Chaco Central, Argentina). In
Hacia una nueva carta tnica del Gran Chaco, Jos Braunstein & Cristina Messineo (eds),
194212. Buenos Aires: Centro del Hombre Antiguo Chaqueo.
Martnez Crovetto, Ral. 1995. Zoonimia y etnozoologa de los pilag, toba, mocov, mataco y
vilela, Pedro Viegas Barros (ed). Buenos Aires: Universidad de Buenos Aires.
Matisoff, James A. 1992. The mother of all morphemes: Augmentatives and diminutives in a real
and universal perspective. In Papers from the First Annual Meeting of the Southeast Asian
Linguistics Society,Martha Ratliff & Eric Schiller (eds), 293349. Tempe AZ: Arizona State
University, Program for Southeast Asian Studies.

Augmentative in Toba (Guaycuruan)

Medrano, Celeste. 2012. Zoo-sociologa qom: de cmo los tobas y los animales trazan sus relaciones en el Gran Chaco. PhD dissertation, Universidad de Buenos Aires.
Medrano, Celeste, Maidana, Mauricio & Gmez, Cirilo. 2011. Zoologa Qom: Conocimientos
tobas sobre el mundo animal [Serie Naturaleza, Conservacin y Sociedad 3]. Santa Fe
(Argentina): Ediciones Biolgica.
Messineo, Cristina. 2003. Lengua toba (guaycur). Aspectos gramaticales y discursivos. Munich:
Messineo, Cristina. 2008. Entre nombres y verbos. Categorizacin de los conceptos de propiedad en toba. In Estudios en Lingstica y Antropologa. Homenaje a Ana Gerzenstein,
Cristina Messineo, Marisa Malvestitti & Roberto Bein (eds), 157166. Buenos Aires:
Universidad de Buenos Aires.
Messineo, Cristina & Cneo, Paola. 2011. Ethnobiological classification in two indigenous languages of the Gran Chaco Region: Toba (Guaycuruan) and Mak (Mataco-Mataguayan).
Anthropological Linguistics 53(2): 132169. DOI: 10.1353/anl.2011.0010
Miller, Elmer. 1979. Los tobas argentinos. Armona y disonancia en una sociedad. Ciudad de
Mxico: Siglo XXI.
Mithun, Marianne. 1999. The Languages of Native North America. Cambridge: CUP.
Mufwene, Salikoko S. 1980. Bantu class prefixes: Inflectional or derivational?Papers from the
annual regional meeting of the Chicago Linguistic Society16:246258.
Nieuwenhuis, Paul. 1985. Diminutives. PhD dissertation, University of Edinburgh.
Payne, Thomas. 1997. Describing Morphosyntax. Cambridge: CUP.
Scalise, Sergio. 1986. Generative Morphology. Dordrecht: Foris. DOI: 10.1515/9783110877328
Scarpa, Gustavo. 2010. Hacia una etnotaxonoma vegetal chorote II: Clasificacin de las plantas
entre las parcialidades iyojwja y iyowjwa del Chaco argentino. In Lxico y categorizacin
etnobiolgica en grupos indgenas del Gran Chaco, Cristina Messineo, Gustavo F. Scarpa &
Florencia Tola (eds), 157198. Santa Rosa: Universidad Nacional de La Pampa.
Stump, Gregory. 1993. How peculiar is evaluative morphology? Linguistics Faculty Publications,
Paper 28. (9 February 2014)
Tola, Florencia. 2010. Matres, chamanes et amants. In Ateliers du LESC [Socit dEthnologie
de Nanterre 34],Valentina Vapnarsky & Aurore Monod Becquelin (eds). http://ateliers. (9 February 2014)
Vidal, Alejandra. 1997. Noun classification in Pilag (Guaykuruan). Journal of Amazonian Languages 1(1): 58111.
Wright, Pablo. 1992. Dream, shamanism, and power among the toba of Formosa Province.
In Portal of ower: Shamanism in South America, Jean Langdon & Gerard Baen (eds),
149172.Albuquerque NM: University of New Mexico Press.
Wright, Pablo. 2005. Cosmografas. Etnografas Contemporneas 1(1): 173210.

Productive compounding in Baure (Arawakan)

Femmy Admiraal & Swintha Danielsen
University of Amsterdam / University of Leipzig*

The Baure language (Arawakan) can be described as polysynthetic, agglutinating,

head-marking, and classifying. Verbs are the most complex parts of speech,
with various levels of derivational and inflectional morphology. Baure nouns are
involved in fewer morphological processes, such as number and locative case
marking. However, productive compounding leads to additional morphological
complexity of nouns.
Our broad understanding of compounding includes bound lexical
morphemes, such as classifiers and locative stems. We argue that we are actually
dealing with one structure in Baure, based on a right head that is modified by
any component to its left, be it that the elements in the compounds are bound
or free forms. The construction is structurally also similar to noun or classifier
incorporation to verbs.
Keywords: Arawakan; compounding; incorporation; classifiers;
nominal morphology; verbal compounding

1. Introduction
The Baure language group belongs to the southern branch of the Arawakan language family (Aikhenvald 1999:6571), and it is closely related to the Moxo languages Ignaciano and Trinitario, and to the other Southern Arawakan languages
Terna and Paunaka. Originally, Baure was spoken in three different mission
towns, Baures, El Carmen and San Joaqun, and previously the varieties were called
dialects of the same language. Nowadays, the three varieties Baure, C
armelito and

* The research was carried out as part of the project The Documentation of Baure, which
received funding from the DoBeS programme (Dokumentation Bedrohter Sprachen) of the
Volkswagen Foundation ( We are deeply grateful
to the last Baure speakers in all three communities (Baures, El Carmen, and San Joaqun), and
wish to thank them for helping us understand their language and for sharing their cultural
knowledge with us. Unfortunately, in the course of the project, a number of speakers have
passed away, and we will certainly remember them.

Femmy Admiraal & Swintha Danielsen

Joaquiniano are usually referred to by their individual names, which are derived
from the places where they used to be spoken. In the course of time, the varieties
diverged substantially from each other, Joaquiniano in particular,1 and therefore,
we propose to regard them as three different languages forming the Baure language group.
Baure can be considered a critically endangered language (Krauss 2007).
Based on recent counts, the total number of speakers (ranging from fluent speakers to semi-speakers) for all three varieties amounts to 59. The speakers are all
elderly of the grandparental or great-grandparental generation, and transmission
of the language from one generation to the next has come to an end since the
mid-twentieth century. Today the speakers of Baure are all fully bilingual in Spanish, the dominant language in the region. Baure is scarcely used in daily life, and
although a few speakers do use Baure in conversations, most often there are other
people present and Spanish is preferred.
2. Baure morphology
This section introduces the Baure nominal morphology and some general facts
about compounding in Baure, which are then described in detail in subsequent
sections of the article.
2.1 Baure nominal morphology
Baure nominal morphology is fairly uncomplicated and there are only few nominal suffixes, such as plural marking, the diminutive, the augmentative and the
locative suffix. Plural marking is obligatory with nouns referring to humans; it is
optional, and often omitted, with all other nouns.
Strictly speaking, a distinction in word classes can only be made between
nouns, verbs and particles. Adjectives form a subclass of nouns and can take
all nominal morphology. Baure distinguishes between alienable and inalienable
nouns, the latter including kinship terms and body parts. Inalienable nouns are
bound roots that obligatorily take a personal proclitic indicating the possessor.
Furthermore, Baure has a rich classifier system, with over 30 classifiers. Classifiers
are found in numerals, in nominal and adjectival compounds, incorporated into
verbs, and in a number of cases these constructions are lexicalized (Terhart 2009).
1. A number of sources even claim that Joaquiniano is a separate Arawakan language (Szab
1998:21; Rivero Pinto 2006), and yet other sources assume Joaquiniano to be a Moxo dialect
(Fabre 2005:46, 71).

Productive compounding in Baure (Arawakan)

Bound nominal roots are often engaged in productive compounding and are also
frequently incorporated into verbs, as are classifiers. Given the many similarities
in morphosyntactic behaviour (see Section 3), and the existence of bound forms
that can occur both as possessed nouns as well as classifiers, we claim that bound
nominal roots including classifiers form a continuum from rather lexical to more
grammatical elements. On the outermost lexical side are bound nouns that are
most nominal in nature, resembling free nouns in that they are rarely used in
compounds or incorporated (e.g. kinship terms). On the other extreme, the classifiers are mostly grammatical elements, but nevertheless share characteristics with
bound nouns that are more to the lexical side of the continuum.
2.2 Compounding
We define a compound as a morphological unit (see Marchand [1960:11] in Lieber
& tekauer 2009:5) contain[ing] more than one stem or root (Bybee 1985:105),
thus a morphological combination of lexical material. The lexeme, however,
may be a free root or a bound one, and this is exactly the part that leads to most
discussion where do we cross the line to derivational morphology? Lieber &
tekauer (2009:4) claim that there has always been much discussion of exactly
what a compound is, and even of whether compounds exist as a distinct species
of word formation. Here we do not pick up on this discussion and therefore do
not doubt the existence of a process found in languages that adjoins prototypically lexical roots and forms a new morphological unit. One reason why defining
compounds causes conflicts is the definition of their parts as lexical versus grammatical. The compound units may be more or less transparent, i.e. the meaning of
a compound can but does not have to be predictable from its components meanings. Fabb (1998:68) describes that [a] diachronic loss of transparency (both formal and interpretive) can be seen in the process whereby a part of a compound
becomes an affix, in which case we would not call it compounding any longer. But
when are we still dealing with a bound lexical root and when with a derivational
affix? Furthermore, the less a complex noun may be viewed as a fixed unit, the less
convincing its classification as a single word is. Thus, productive compounds may
sometimes be categorized differently, for example as phrasal constructions.2
We claim that there are productively created compounds in Baure. A compound in Baure forms a unit as a phonological word, which is also mirrored by the
orthographical representation of these compounds. They prototypically consist of

2. Lieber & tekauer even state that Plag [2006] simply remains agnostic on whether there
is a distinction between noun + noun compounds and phrases (2009:14).

Femmy Admiraal & Swintha Danielsen

the combination of two or three lexical roots, most often nominal ones, which
usually also results in a nominal unit. Compounding in Baure can have various
functions, such as modification, indication of a certain spatial relation, or other
semantic relations.
Without doubt, the suggested definition gives rise to problems in classifying
concrete (plurimorphemic) units. In some cases, the parts of Baure compounds
can be analyzed as rather grammatical morphemes, e.g. the classifiers, and subsequently we would ask if we are not facing derivation instead of compounding.
For Baure we claim a concept of a lexical root that is scalar and depending on the
amount of lexical information. Because of the more transparent and less grammaticalized character of the Baure classifiers, we still categorize them as lexical
roots, which has, of course, consequences for the definition of the constructions
in which classifiers occur. It is possible that a very small subgroup of most grammaticalized classifier should be separated. Their status as a semantic and grammatical head in grammatical words is rather problematic, and they may obey even
more morphosyntactic restrictions. The question of compounding versus derivation is certainly related to this categorization of a morpheme as root versus affix.
The productivity of the construction of particularly adjectival roots with classifiers
may raise the question of whether it is compounding or a phrasal construction for
modification. Our aim in this paper is not to discuss the definition and specific
classification of each and every particular construction with different root types,
but instead we want to show the character of a widespread construction type, its
productivity, and the different root types that can participate in it. In doing so, we
do not limit ourselves to compound nouns, but we extend the description to incorporation and the combination of verbal roots, in order to point at the uniformity
of the constructions, or observe the differences.
As already mentioned in 2.1, strictly speaking, we only distinguish the word
classes nouns, verbs and particles in Baure. Particles never play a role in compounding, but nouns and verbs do. For the morphological combination of lexical
roots in Baure, we recognize three broad compounding types: (1) nominal compounding (Section 3), (2) mixed compounding of a verbal and a nominal root
and incorporation (Section 4), and finally, (3) compounding of two verbal roots
(Section 5), all of which are discussed in different sections of this article. The compounding types in Baure are
Type 1: N+N = N
Type 2: V+N = N; V+N = V
Type 3: V+V = V

The most common types are Type 1, and Type 2 V+N = V.

Productive compounding in Baure (Arawakan)

It may be surprising that incorporation is included in the discussion here,

because usually only the nominal type (Type 1) is considered when talking about
compounding, rarely verbal compounding (Type 3), as e.g. in Aikhenvald &
Muysken 2011. However, the structural similarity of nominal compounding and
noun incorporation in verbs has been pointed out before. In Adelaar with Muysken
(2004:518, citing Smeets [1989] on Mapudungun), e.g. it is stated that incorporation must be analysed as part of a general tendency of the language towards
compounding (see also Baker & Fasola 2009:595, compare also Vuillermet, this
volume). At the end of each section, we return to the discussion of the construction and its classification as compounding.

3. Nominal compounding (Type 1)

This section describes the first type of compounding lexical material, namely
nominal compounding, which was presented above as Type 1: N+N = N. The
nominal roots involved can belong to different subclasses, namely the subclasses
of free nouns and different types of bound nouns, to which not only inalienably
possessed nouns, adjectives and numerals belong, but also locative noun roots
and classifiers. How these nominal roots differ in their morphological behaviour, not only with respect to compounding, is summarized in Table 1. As noted
in Section 2, bound nouns are on different positions of the continuum with
respect to their degree of lexicality or grammaticality. However, we claim that
they are still more on the lexical side.3 Even though they all participate in nominal compounding, we have subdivided this section according to the semantics
of the nominal roots. Thus, we first show the compounding of nouns with nouns
and classifiers (3.1), then move on to locative compounds (3.2), to adjectives and
numeral compounds (3.3), to classifier compounds, which consist of two classifier roots (3.4), and finally give examples of multiple compounding in 3.5 (see
also example (2)).
The basic structure of a nominal compound is represented as follows:
[N1+N2]N N2 with relation R to N1 (see Booij 2009:201)

3. For some classifiers this may not hold, but see below in Section 3.1.2 on N-CLF compounds.

Femmy Admiraal & Swintha Danielsen

Two nominal roots are combined and result in a new complex nominal root that
can undergo nominal morphological processes as a whole. This unity is illustrated
here in (1) and (2):4


your blond head


wak + -ji + -pe
spoons made of cow horn

joro- + -poe

In (1), the unity is expressed by the derivation of a possessed form of the compound joropoe blond head, by the possessive suffix -no and the personal possessor
proclitic pi= 2sg. In (2) the complex compound consisting of three roots, wakjipe
spoon made of cow horn is pluralized, just like any simplex noun, by the suffix
N2 is the head of a compound, and N1 its modifier. Right-headed compounds,
like in Baure, are reported to be very common, and are in particular well known
thanks to their pervasiveness in Germanic languages (see Booij 2009:210). The
general semantic relation is the modification in an endocentric construction;
exceptions to this are given in 3.1.4. How productive compounding is and was
in Baure is proven by the fact that there are also compound classifiers (3.4), and
that multiple compounding is possible with some nominal roots (3.5). The basic
structure of compounds can thus be extended, as will be specified in the respective
Table 1 gives an overview of the morphological behaviour of the different nominal roots. It is most striking that some roots only occur in N1 position,
whereas others only occur in N2 position. The types of nominal roots involved in
compounding can be subdivided into free roots, right-bound roots and left-bound
roots, and this is also the order in which they are listed in Table 1.
As will be shown in the following subsections, free Ns can be used as N1, but
rarely as N2 in compounds. Thus, free Ns are also seldom bound to an adjectival
or numeral root in N2 position. Free Ns cannot be directly possessed, and they are
never incorporated into verbs. Right-bound nominal roots usually refer to plant
species. These roots are only used in N1 and never in N2 position, irrespective
4. In each example, we provide the separate lexical roots on the right-hand side, indicating
the bound character by a hyphen in the place where the lexeme is bound.

po- other

-weri house

-chipi roof;
CLF: rooflike

-api under

-pi CLF:


bound N

bound N/ CLF
(can be used as

locative N



er- plantain

right-bound N


never (except
for classifier

never (except
for complex
locative roots)



mes table

free N

free ROOT



Type of noun






Table 1. Morphological characteristics of nominal roots involved in compounding

N2 in

never N2 in


rarely N2 in

no combination
with other

never N2 in

rarely N2 in



preferred, but
not obligatory








Productive compounding in Baure (Arawakan)

Femmy Admiraal & Swintha Danielsen

of the type of root. Also, they cannot be possessed or incorporated into verbs.
Adjectival or numeral roots are a specific semantic type of right-bound roots,
but we can argue that they behave in exactly the same way. Moving on to the
left-bound roots, one type of bound Ns can be used as N1 or N2 in compounds,
but it is rarely found as N2 with adjectival or numeral roots. As a bound N, they
have obligatory possessor marking, and they are regularly incorporated into verbs.
Another type of bound Ns can also be used as a classifier and therefore it behaves
morphologically differently. These bound Ns are rarely used as N1 in compounds,
but they are generally in N2 position, together with adjectives and numerals as
the N1. Furthermore, just like the other bound Ns, they are obligatorily possessed
and incorporated into verbs. Locative Ns are not only semantically different from
other bound roots; also their morphological behaviour is more restricted. Locative
Ns are never used in N1 position, with the exception of some complex (lexicalized)
locative roots. Locative Ns occur in N2 position, however, not with adjectival or
numeral roots. Possession of a locative N is possible, but not obligatory for all the
locative roots. Locative Ns are incorporated into verbs. Classifiers are bound roots
with generally broader semantics and a more restricted morphological behaviour,
especially in contrast to bound Ns which can sometimes be used as classifiers.
Unless they are compounds themselves, classifiers do not occur in N1 position, but
they are used as N2 in all kinds of compounds. Classifier roots cannot be directly
possessed, but they are incorporated into verbs.
3.1 Nominal compounds (N-N and N-CLF and more)
This subsection of the nominal compounding description refers, in particular, to
the two types of nominal roots: free and bound. Here left-bound roots are argued
to be related to classifiers on a continuum with decreasing lexicality. We do try
to draw a line between N-N and N-CLF compounds, and have therefore devoted
one subsection to the description of endocentric compounds with nominal roots
in N2 position (3.1.1) and another one to classifiers in the same position (3.1.2).
In some cases of compounding, a linking element can be found (3.1.3). There
are also exocentric compounds in Baure, even though they are not as common
as endocentric ones (3.1.4). Different from the default pattern are reverse compounds (3.1.5) and coordinate compounds (3.1.6), but they seem to be rather
infrequent as well.
3.1.1 Endocentric compounds with nominal N2
Nominal compounds of which the first element N1 consists of a free or bound
noun root can have three different types of N2, namely bound noun roots, classifiers, and, exceptionally, free noun roots. This compounding type seems to be

Productive compounding in Baure (Arawakan)

almost always subordinate or attributive, according to the classification by Scalise

and Bisetto (2009:50). We start off with the most common construction, a free or
bound N1 with a bound N2, which is used productively to name parts and products of animals and plants.
It is striking that animal names are free nouns, whereas plant names generally
consist of a bound N1 and an N2 that identifies the part referred to, an observation
which may be a hint at a different conceptualization of the two specimens in the
Baures environment (see Terhart 2009): animals can be individualized and plants
cannot. Two contrastive examples are given in (3) and (4), which both show the
following structure: [N1+N2]N.


tiporekpoe = tiporek + -poe

chicken head
chicken head


koshii = kosh (i)- + -i

motac fruit
fruit of the motac palm (Scheela princeps)

The compound in (3) represents a whole-part relation, typical of subordinate compounds. The one in (4) is rather an attributive compound, in which the N1 determines the species and N2 singles out the referent (genus-part).5 In (5) and (6),
there are source-product (material) relations, partly with Spanish loans:6


rospan = ros + pan

rice bread
rice bread


koyoroemantek = koyoroe- + mantek

koyoroe-mantek tota
tota oil (Acromia tota)

5. Mind that in Baure, we are not dealing with these so-called cranberry-morphs (Scalise &
Guevara 2005:157), where the first element is bound, but not productive.
6. ros < arroz rice; pan bread; mantek < manteca butter.

Femmy Admiraal & Swintha Danielsen

Examples like (5) with two free nominal roots borrowed from Spanish are certainly very rare.7 This example also shows how structurally different Baure is from
Spanish, where we usually have complex phrasal NPs, like pan de arroz rice bread.
However, examples with free N2 roots are extremely rare in Baure. This is presumably because the great majority of noun roots in Baure are bound, and it then does
not come as a surprise that these nouns tend to bind another nominal root to their
left, in the same slot where the possessor marker would be.
Bound roots display different degrees of lexicality, as has been mentioned
above. There are truncated noun roots in N2 position that are only found in compounds, like -ap bone (< nop bone) or -an egg (< na egg).8 Some reduced
forms are also used as classifiers, i.e. attached to adjectival and numeral roots,
like -esh meat (< nesh meat).9 Potentially, we could say that any N2 root could
be transfunctionalized as a classifier, be it truncated or not, and result in semantic
extension.The compounds in (7) and (8) are examples of reduced N2 roots that
can be categorized as being halfway between noun and classifier root.


simoriesh = simori + nesh



siporan = sipor + na
rhea egg
rhea egg

3.1.2 Endocentric compounds with classifiers

It is an expected development of the N2 in a compound construction to be subject
to greater phonological and morphological modification. Considering the expansion of the semantic reference of classifier roots in N2 position, we can actually
observe a process from lexical compounding to grammatical derivation. The latter case is most prominent with a subclass of fairly grammaticalized classifiers,
7. See also Example (9)h from Mapudungun lichi-waka cows milk, Ziga, this volume.
8. Vowel alternation between o and a is common in Baure. However, an alternative analysis
cannot be excluded: -a being the linking suffix and -p(o) the element designating the bone.
Inthe case of -an < na egg, we can additionally think of metathesis of the lexeme.
9. It is remarkable that some words start in the nasal n- and are then truncated as if it were
not part of the root. This could indeed be the case: there is a (classifying) root n- that is part
of many lexemes in Bsiro, a Macro-G language that Baure may have had intensive contact
with in history, just like the Moxo languages.

Productive compounding in Baure (Arawakan)

where it becomes difficult to argue that the classifiers function as the semantic
heads in the constructions because of their broad semantic reference. Their status
as morphological heads also depends on their categorization in terms of morphological class. If we assume that classifiers are nominal roots, they can be identified
as morphological heads, just like other nominal roots. But we do still not analyse
the construction as essentially distinct from the basic construction presented in
the preceding Section 3.1.1. Even though in some cases, the head character of the
classifier root may be questionable, we consider it to be the general head of the
construction in the majority of cases, leaving open a possibility for an alternative
analysis of grammaticalized classifiers as derivational suffixes and thus not as lexical roots and semantic heads. In fact, the most common kind of compound in
Baure has such a classifier root in N2 position, thus:

Similar to compounds for the plant parts, (9) is an example of a whole-part relation, where the classifier refers to the seed, and (10) illustrates a material-product



moeishop = moeish +
papaya seed


kajaropi = kajar- + -pi

cotton thread

The classifier -po clf:multitude in (9) is one of the grammaticalized and productive classifiers in Baure. It mostly refers to powders, seeds, or materials consisting
of small grains, and to insects, fish, and other animals that apparently come in
swarms. The classifier can derive such nouns in compounds, as shown in (9). Some
nouns are represented as lexicalized units, and the word is not decomposed, if the
N1 root is unproductive or unknown. Instances of lexicalization that were probably derived by compounding are shep manioc meal, jajap sand, yomoep bee,
kosip shell, and rop mosquito larva.
The classifier -pi in (10) is used with long and thin items, and it is also
lexicalized in some animal names, such as jorombi eel and shipi worm. In his
reconstruction of a large cognate set of Proto-Arawakan, Payne (1991) mentions that where some languages use lexical nouns relating to a cognate form,
other languages have a classifier as reflex of the same cognate, in particular with
respect to this classifier to mark long, slender and snake-like (Payne 1991:383).

Femmy Admiraal & Swintha Danielsen

He reconstructed the Proto-Arawakan lexeme *aphi snake. We may also have

had cases of compounding in Proto-Arawakan with certain morphemes, which
then grammaticalized into classifiers in some of the languages. Thus, compounding was presumably the original construction producing the classifier system in
Arawakan languages (Danielsen 2007:137).10 The grammaticalization process
would have started with a compound N2 root, which suffered phonetic truncation or semantic bleaching (or both) and was then subsequently used to refer to a
broader class of items as a nominal classifier. In some cases, the more lexical and
the classifier root even still co-exist, as in -wok tree and clf:place and -chipi
back, roof and clf:roof-like.
3.1.3 Compounds with a linking morpheme
Baure has a linking suffix -a lk applied in compounding. It occurs most often
with plant compounds, where the N1 is generally bound. Two examples with the
linking morpheme are given in (11) and (12). Note that (11) and (13) contrast in
that both are deriving a kind of leaf, but once with and once without the linker.


erapon er-a + -pon

er-a-pon plantain lk
plantain leaf


-a + -po
chor-a-po maize (cob) lk
maize meal


sonipon soni+ -pon

soni-pon tobacco
tobacco leaf

It is not clear which compounds appear with the linker -a. It can be traced back
to the irrealis marker -a (during the course of the centuries, Baure lost the final
vowels -o and -a; see Danielsen 2007 for details). According to current analyses, in some cases, the vowel -a may be analysed as belonging to the first root.
In other cases, it may be the remnants of an irrealis marking in the derivational

10. We could also imagine that an already existing classifier system that marked nouns in
final position in a compound-like constructions (as N2), was taken as the model on the basis
of which new classifiers were derived from (bound) noun roots occurring in this N2 position.

Productive compounding in Baure (Arawakan)

process, which is not transparent to the authors.11 There are also words in which
the linker has already been lexicalized, such as in the first lexicalized compound
part siriatot-, including the lexicalized linking suffix -a (bold in example (14)) of
siriatotawok chonta palm tree (Astrocaryum tucuma).


chonta palm tree (Astrocaryum tucuma)

3.1.4 Exocentric noun compounds

Concerning the semantic relation between its components, almost all compounds
with this productive pattern are endocentric, the head noun thus referring to the
denoted element, either in a more concrete way by a nominal root, or in a more
abstract sense by a classifier. In addition to this, we also find a few attributive compounds (called ATAP in Scalise & Bisetto 2009:50), animal names in particular,
which have an exocentric relation, as in Examples (15) and (16):


tiyowkopasiri tiyowko + -pasiri

tiyowko-pasiri spoon


tir- + -ijos

As typical of ATAPs, there is a metonymic relation to the denoted element, in (15)

focussing on the beak, and in (16) on the tail of the animal.
3.1.5 Reverse compounds
We do not want to neglect the few examples of idiosyncratic compounds, which
have the opposite semantic coding of the roots, namely the N1 being the head and
the N2 the modifier, like in example (17):


-chipi + -poiy
ni=chipi-poiy back
dorsum of the foot

11. The irrealis marker or another morpheme -a was applied in certain nominalization
types and other constructions in historical Baure documented in the 18th century (Magio
1880[1749]), but the sources do not present enough evidence for a clear picture.

Femmy Admiraal & Swintha Danielsen

These kinds of expressions seem to be lexicalized in Baure. They may be an indication of a formerly different compounding construction with initial head. However,
nowadays, these compounds should be analysed as constructional idioms (Booij
3.1.6 Coordinate compounds
Exceptional in Baure are coordinate compounds12, where both nouns can equally
be the head in the construction (see Fabb 1998:67; Scalise & Bisetto 2009:50). We
cannot determine any distinction between a coordinate compound and a close
apposition. Appositions are multi-functional and frequent in Baure (see Danielsen
2007:381, 411412), and in most cases, we should regard the appositions as
phrasal. Nonetheless, the biological gender marking of animals by the separate
noun jir man/male, as in (18), is a fixed expression and pronounced as a phonological unit, which may be analysed as one single compound noun.

to jir tiporek
jir tiporek
art man chicken
a cock

3.2 Locative compounds

One of the specific types of compounds found in Baure is the locative compound,
which consists of an N1 plus a locative noun root in the N2 position13, and it is
obligatorily followed by the general locative marker -ye loc. The basic structure
of the locative compound is[[N1-N2]-loc].
The N1 is usually, but not necessarily, a free noun (compare Examples (19) and
(20)). The locative noun roots that appear in N2 position are often phonologically
reduced forms of bound or free nouns. They indicate a certain region of the N1,
which may be part of the N1, but can also extend beyond the physical boundaries
of the N1.14

12. These are the ones also called dvandva compounds, going back to Sanskrit (Fabb 1998:67).
13. However, not all locative noun roots can occur in the N2 position of a locative compound.
The most notable exception is the root ani, sky, which is used for above, and always occurs
as a free noun.
14. Locative compounds basically indicate a static topological relation between a Figure and
the Ground noun, which forms part of the compound (N1). However, since the general locative marker is semantically neutral, the compounds may get SOURCE, GOAL, or PATH readings, depending on the context and/or the verb phrase (see Admiraal 2013).

Productive compounding in Baure (Arawakan)

The locative noun stems can be bound nouns, as is the case with the body
parts terms -chipi, back, and -imir, face, which are both used in locative compounds. The body parts are mapped onto the objects denoted by the N1, either
on the basis of the intrinsic sides of the particular object as in (19) and (20), or
based on the current, but possibly temporal and atypical, position of the object,
as in (21).
(19) kwore noiy resiamiri-ye
kwore noiy resia-imir-ye

exist.3sgm there church-face-loc
He is there in front of the church.

resia + -imir
church face

(20) roshim nan rowermiri-ye.

-weri + -imir

nan ro=wer-imir-ye
house face
3sgm=arrive there 3sgm=house-face-loc
He arrives there in front of his house.15
(21) to tawe kwore yashorchipi-ye.16
yashor + -chipi

to tawe kwore
yashor-chipi-ye canoe back

art ball exist.3sgm canoe-back-loc
The ball is on top of the canoe. (note: the canoe is lying upside down on

The N2 of a locative compound is not necessarily a bound form, as is the case with
the noun root api, bottom. In example (22) api is used in the N2 position of a
locative compound, and in example (23) it is used as a free form.
(22) ver vishim ne ewokoeapi-ye.

ne ewokoe-api-ye
already 1pl=arrive here tree-bottom-loc
We already arrived here under the tree.


ewokoe + api
tree bottom

kwore api-ye.
exist.3sgm bottom-loc
It is at the bottom.

Additionally, some of the locative components of a compound are compounds

themselves. This is discussed in more detail in Section 3.5.

15. Note that -chipi, back is used in locative constructions for on top of , based on the position of the back of a quadruped animal.

Femmy Admiraal & Swintha Danielsen

3.3 Adjectival and numeral compounds

Adjectives and numerals are right-bound nominal roots, similar to those used
for species, and they have to get a (bound) nominal root or classifier attached.
Numerals are subsumed under the same category here, even though their use in
the language may be a bit more restricted for semantic reasons. Since the head in
the compound is always nominal, an adjectival compound is structurally a noun
as well: it can have nominal morphology attached, such as plural (except for the
numerals) or locative; the adjectival compound can also form more complex compounds with another root attached to its right (see Section 3.5); and it can be used
as a modifier as well as the head of an NP. The surface structures of these specific
kinds of compounds are:

Examples are given in (24) through (26). Not all of these right-bound roots that we
may call adjectives can also have a neutrally derived form, such as monik pretty,
derived by the absolute suffix -ko, which is also used to derive free from bound
nouns, like -toer < toerok field.


big head


nice drink


two hands

cho- + -poe

moni- + -aro

mapi- + -wojis

Adjectival compounds can be used as a modifier, as in (27), or as a head in an

(27) kwe jim chopcha.

kwe jim cho-po-cha
exist fish big-clf:multitude-aug
There are big fish.

Productive compounding in Baure (Arawakan)

(28) ndi nimon pash to chachanev.

ndi ni=imono po-a-sh
1sg 1sg=buy one-clf:body-one
to cho-a-cha-nev

art big-clf:body-aug-pl
I buy one of the big ones (here: pigs)

In (27), the adjectival compound modifies the preceding noun jim fish and is part
of the subject NP in the clause. In (28), there is a complex object NP (a partitive
apposition), consisting of two nominal constituents: the first one being the numeral
compound, the second one is an adjectival compound. There is no additional noun
in this clause, since the reference is marked by the classifier -a clf:body.
There is no structural difference between nominal compounds with a rightbound N1 to adjectival compounds, except for the semantics. Adjectival roots are
more general than those N1s referring to species; they generally denote a dimension, size or other attribute. In Baure, however, most attributes are expressed
through verbal roots, which may then also be derived and used as modifiers by the
nominalizer -no nmlz (see incorporation in Section 4.2).
3.4 Classifier compounds
Since the head-final compounding construction is so productive in Baure, it does
not come as a surprise that there are even some compound classifiers. The outcome
of compounding two classifiers is a complex classifier with more restricted usage
than that of the simple classifier (also pointed out in Danielsen, forthcomingb).
The general pattern of compound classifiers is:

Compound classifiers are based on the most common relation with the right head
and a left modifier in order to specify the reference, i.e. clf2 is more general and clf1
more specific, relatively speaking. The majority of classifier compounds consist of
combinations with -ki clf:enclosed (clf2), referring to the contents of the object
referred to by clf1. One example is -seki clf:oval.enclosed, being composed of
-se clf:oval, referring to an oval object, and -ki clf:enclosed. More examples are
given below in Table 2. The compound classifier can be attached to numerals and
adjectives, just like the simple classifiers, as compared in (31):


three (e.g. bottles)


three (e.g. fields)

Femmy Admiraal & Swintha Danielsen


mbo-seki =
three (e.g. glasses full)

+ -ki
clf:oval clf:enclosed

The degree of lexicalization of each of the compound classifiers (Table 2) differs,

and we still find newly created forms, indicating the productivity of the formation
of complex classifiers. While the classifiers -seki clf:oval.enclosed, -ajaki clf:pot.
enclosed, and -chiki clf:round.enclosed, seem to be relatively transparent, this is
not true for -poki clf:hammock. The classifier compound -aki clf:body.enclosed
can refer to the contents of a (human) body this being its transparent meaning -,
but also to a lagoon, which shows its lexicalized status as a unit (see Terhart 2009).16
Table 2. Compound classifiers
Simple clf1







































Only the most lexicalized ones of these compound classifiers can also be found
as N2 in compound constructions and incorporated as a whole, as demonstrated
with the forms -peki clf:pit (33) and -aki clf:body.enclosed, (32) and (38):


paraki 17
room (lit. inside of the house)

16. There are probably even more -ki examples which are more lexicalized, such as -soki
seed, kernel, which could derive from -so clf:wood and -ki clf:enclosed, since these seeds
are often in a wooden shell.
17. This is one of the very few examples where the N1 is actually truncated, here with the
dropping of the vowel i; this only shows the lexicalized status of the whole compound.

Productive compounding in Baure (Arawakan)


The tub is empty.

Example (38) in 3.5 illustrates that a complex classifier may even be part of a multiple compound, if the degree of lexicalization is high.
3.5 Multiple root compounds
There are also numerous examples of multiple compounds in Baure. However, the
actual limit of productivity is three lexical roots, with the exception of lexicalized
complex roots, so that we may come across four nominal roots combined into one
word. The layering within the compounds is always successive, a compound of
N1 and N2 forming a compound with another N3 to their right, taking N1 and N2
already as a lexical unit:

general structure:

specific root types:

[[N1 + N2] +N3]N

1. [[N +N] + N]N

2. [[N+N] + clf]N
3. [[N+ clf] + clf]N
4. [[ADJ + N] + N]N
5. [[ADJ + N] + clf]N

The semantic relation is generally that N3 refers to the product of the compound
N1+N2, and these first two are most of the time body or plant parts, where N1
refers to the kind of animal or plant and N2 specifies the part (see Section 3.1.1).
We present three examples here.


wakchomoeki wak -chom -iki

wak-chomo-iki cow skin clf:container
cowhide bag


kotieshaj koti- nesh -aj

koti-esh-aj new meat soup
soup made with fresh meat


era- -po-aro
era-po-ar plantain-clf:multitude-clf:liquid
plantain chicha

Femmy Admiraal & Swintha Danielsen

There are some lexicalized locative noun compounds with the second element
-wani place. If taken as multiple compounds, the layering is different from number 1 in the list of multiple compounds:
[N + [N +N]]N -loc


siyshiriwani-ye siy
siy-shiriwani-ye chair behind
behind the chair

Finally, we have a very complex compound in example (38), where, however, the
final classifier compound can be considered to be lexicalized. The word in (38)
belongs to the number 3 of the multiple compounds listed in this section. The
structure of the compound parts and the layering is indicated as follows:
or even: [[[N+clf]N +clf]N+[clf+clf]CLF]N19


woven clothes (with a hammock style of weaving)

3.6 Discussion of Type 1 compounds

In Sections 3.1 through 3.5, we have shown the structure of Type 1 compounds in
Baure. There are some common patterns, to which we will return here, and there
are exceptions. The structural exceptions for compounds consist of reverse compounds, as presented in 3.1.5, where the head is not in N2 position; and the very
rare coordinate compounds in 3.1.6. Semantically, we have mainly subordinate or
attributive compounds in Baure, but exceptions to this are a few exocentric compounds for animal names, where a metonymic relation to the denoted object can
be argued. All the other compound constructions with different kinds of nominal
roots show the same basic structure with the head in N2 position and a modifier
18. The marking as (*) means that it is at the present state of the language doubtful that this
bound root exists at all as a separate one.
19. In fact, the first noun wotoki hammock is presumably also a lexicalized compound noun
of a noun and the classifier -ki, which would then sum up into five lexical roots in one nominal
word. But of course, we do have to consider certain units as lexicalized and should refrain
from artificial decomposition here.

Productive compounding in Baure (Arawakan)

in N1 position. The relation is endocentric. The presence or absence of a linking

morpheme does not play a role for the categorization of these compounds. The
compounds are distinct from phrases, since they form single phonological words.
This is proven by any further applied nominal morphology (3.1), e.g. plural, possessive, or locative (3.2) marking. The composition of phrases is much more complex in Baure and is shown in contrast at the end of this section.
We now return to the theoretical discussion and address the possible problematic issues. When describing another Arawakan language group with respect
to compounding, Maipure-Yavitero, Zamponi (2009:586) claims the absence of
(productive) compounding in this northern Arawakan language group. He also
concludes that a more or less strong resistance to forming compounds was also
observed in other Arawakan languages (Zamponi 2009:586). What is the reason
for this extreme contrast to our findings? As we will show below, the characters
of the languages are not complete opposites, but the classification of certain processes as compounding was considered on the basis of different criteria. It was
already mentioned in Lieber and tekauer (2009:4) that assigning a construction
type to compounding depends on the particular definition of what compounding is and the definition of its parts. Of course, Baure does not have compounding
of many free nominal roots, but this is not surprising, since the language does not
have many free lexical roots to begin with. The majority of lexical roots in Baure
are bound, albeit in different ways (as shown in Table 1): left-bound nominal roots
only need the possessor marking and can be considered relatively independent,
apart from that; the other bound roots are functional lexical elements, which are
not complete without additional lexical information, and these are adjectives,
numerals, classifiers, and locative roots. Now the question arises where to draw the
line between bound lexical roots and derivational affixes (see Lieber & tekauer
2009:5). The most difficult case is that of classifiers in Baure. Classifier roots are
attached to other lexical material, where they may be viewed as derivational, but
they also occur in classifier compounds, which reflect nominal compounds on a
more generic level (see 3.4). The position of classifiers in compounds is the same
as of other bound roots as N2 (head); their function as the head in compounds is
diminished because of the semantic bleaching; however, they do still determine
the kind of referent of the compound, as e.g. a kind of animal (mammal, bird, fish/
insect, etc.), kind of plant part (leaf, seed, stem, branch, fruit, etc.), and so on. But
whether these classifiers are still referential is difficult to answer. We should say
some are and some are not anymore, depending on their place in the continuum of
degree of lexicality. All Baure classifiers are somewhere on the continuum between
fully (i.e. specific) lexical, generic lexical, and grammatical. For many roots that
function as a classifier the root may still be referential, but for a class of highly
grammaticalized classifiers in Baure, we may see more grammatical-functional

Femmy Admiraal & Swintha Danielsen

than lexical qualities of the root; thus, cases of compounding with a grammaticalized classifier are like grammatical derivation; we do not want to make any arbitrary
distinction here, since from a formal descriptive perspective the constructions do
not differ. Therefore we do not exclude compounds with classifiers and join these
kinds of compounds with derivational processes, such as nominalization; if they
need to be distinguished from lexical compounding, then compounds where the
N2 is a grammaticalized classifier (be it with nominal, adjectival, numeral root in
N1 position), should be called lexical derivation versus strict grammatical derivation, and this is in between a lexical and a grammatical process:





combined NPs




ADJ+clf N



Figure 1. The decreasing degree of lexicality in the combination of lexical material

Even though he is describing a language of the same linguistic family, Zamponi

comes to completely different conclusions for Maipure-Yavitero. This is because he
has a very strict view of what a compound is: he only accepts free roots in compounds (2009:586), and he argues that the meaning of the few compounds is not
predictable (2009:587). In spite of this first assumption, Zamponi presents cases of
compounds similar to the ones described for Baure, e.g. with a bound nominal root
in N2 position meaning excrement, see (39) from Yavitero, Baniva, and Maipure,
respectively (Zamponi 2009:588589). See also the compound found in Baure,
where the N2 position is filled by a classifier root, translated as honey in (40):
(39) mhatsia (yvt), mpatsi (bvv), mapaik (qij)20

bee-excrement bee-excrement bee-excrement21

20. Maipure does not have an ISO-code, but only a local code used in MultiTree http://
21. Zamponi (2009:589) translates mpa in Maipure as honey, but this is supposedly
wrong, since the same form is translated as bee in the other two languages. Apparently all
of the forms include the classifier -pa, which is cognate with Baure -po in yomoep(o) bee in
Example (40). It is also possible that the form in Maipure was ambiguous.

Productive compounding in Baure (Arawakan)




The language group described by Zamponi also forms compounds of numerals

and classifiers, only that he would not categorize these forms as compounds. Furthermore, we can suspect that most patterns shown for Baure also exist in these
three languages to some extent: nominal compounds with species names in N1
position, classifiers in N2 position, compounds of adjectival and numeral roots in
N1 position with classifiers and other bound roots in N2 position, and finally, also
locative compounds. For more details, the respective grammatical descriptions of
these and other Arawakan languages of that region can be considered, for which
we have no room here.
As for the contrast of compounds to phrases, we want to give four examples
for Baure: an apposition phrase of two nouns (41), a possessive NP (42), a partitive
NP (43), and a modified NP (44):
(41) to roper kove

to ro=per

art 3sgm=domesticated.animal dog
his dog

The apposition in (41) is similar to the coordinate compound in example (18) in

3.1.6. The construction is needed because no animal can be directly possessed. We
could also refer to the noun -per domesticated animal as a classificatory noun
(see Fabre 2007). The pronunciation is that of one phonological phrase, but the
noun -per can also be used on its own. We could argue that this phrase is in fact a
borderline case and could also be classified as a coordinate compound.
The construction in (42) is a frequently occurring possessive phrase, consisting of two full NPs, in this case a possessed NP coordinated with a demonstrative
NP. The difference to compounds is quite clear: the two nouns belong to two
separate NPs, which can include determiners or individual possessive marking,
as well as separate nominal morphology, such as locative, diminutive, or plural
(42) riavinon tich eton

3sgf=husband dem2.f woman
this womans husband

The partitive construction in (43)a is the way quantifiers are juxtaposed to other
NPs, similar to the possessive NP in (42). Example (43)b shows that there may also

Femmy Admiraal & Swintha Danielsen

be one simple numeral compound with the loan word mes table. We can contrast
the compound construction to phrasal constructions with a numeral, as in (43)a.
(43) a.

apiamok to mes
in contrast to: b. apimes
to mes
two-clf:flat&raised art table
two tables
two tables

In example (44), you can observe how modifiers are coordinated in one complex
NP. In such a case, the compound numeral mbowok three trees and the compound adjective chowkocha very big tree co-occur with the noun ewokoe tree
within one NP. We could analyse the marking of the noun or classifier -wok tree
as agreement on the numeral three as well as the adjective big in (44). In such a
case, the argument that there are three compounds in one NP is probably difficult
to hold. But we believe that the fact that each of the marked components can also
be used as an NP on its own proves their otherwise independent nominal character. The marking has not been fully grammaticalized as agreement yet, in Baure.
(44) mbowok ewokoe chowkocha

mbo-wok ewokoe cho-wok-cha
three-tree tree
three very big trees

The complex functions of phrases in Baure show a contrast to compounds in that

complex phrases can be distinguished from compounds which are single words.
4. Mixed compounds and incorporation (Type 2)
In this section we describe compounds of two different lexical roots, namely a
verbal in first position compounded with a nominal in the second position, which
we address as Type 2 compounds in 2.2. There are two different results of having
these mixed compounds: the least common are those that actually result in nouns,
i.e. they are free nominal bases (V+N = N), briefly illustrated in Section 4.1. The
far more common mixed compounding process is the incorporation of nominal
roots into verbs (V+N = V), which does not change the status of the word base as
verbal, shown in Section 4.2.
4.1 Mixed compounds
There are a few nouns that consist of a verbal and a nominal root, which still have
a semantic relation, i.e. the noun in the construction is generally an incorporated

Productive compounding in Baure (Arawakan)

object of the verb. The compounds are typically exocentric. They are used as nominal bases without further derivation. Examples are given in (45) and (46):





Mixed compounds can also consist of a verb and a classifier, as given in (47)


ant (sp.)


a. konoes
alternatively: b. konoropi

write-clf:stick write-der-clf:long&thin

The exocentric compound in (47) is used to name a specific kind of ant. This kind
of deriving names for animals and plant parts was already shown to be very productive with compounds of Type 1 above (3.1.4), the only difference here being
that the modification is done by a verbal root. Note that example (48)a, a relatively
recently created compound for pen, also has an alternative derived name in (48)b,
based on the same verbal root and a different classifier, bound to the other root
by a derivational morpheme -ro. The mixed compounds given here in 4.1 are all
subordinate compounds.
4.2 Incorporation of nouns and classifiers
In Baure, the incorporation of nouns and classifiers is a productive process,
in which transitive as well as intransitive verbs are involved.22 Baure verbs

22. Baure thus shows the incorporation type 3, as categorized in Mithun (1984).

Femmy Admiraal & Swintha Danielsen

incorporate instruments, locatives and goals, equatives, possessa, and, to a limited extent, subjects, just as described of languages that incorporate in general
(Sadock 2006:585). The incorporated element in Baure, whether a noun or a
classifier, directly follows the verb root, and together they form the verb stem,
to which additional verbal morphology can be attached. The incorporated element may be repeated explicitly for emphasis (51), similar to the co-occurrence
in complex NPs with marked adjectives (see Example (43)). The basic structure
of incorporation is:
[[V]root +N/clf]stem

Examples of incorporation are given in (49) through (51). The kind of nominal
roots that can be incorporated are indicated in Table 1: these are all left-bound
roots (49), including bound nouns, locative roots and classifiers (50). The types
of elements that are incorporated are the same ones as the forms found in the
N2 position of a nominal compound, and they may also be truncated noun
roots (51).


I am scratching my foot


I am sitting in the water

(51) nijirishekow to eshe.

ni=jiri-(e)she-ko-wo to eshe
1sg=sit-trunk-abs-cop art trunk
I am sitting on the trunk

In some cases, the verbs and the incorporated elements are highly lexicalized
units, and although the incorporated element is still clearly recognizable, the verb
does not occur without this or another incorporated element. For example, the
verb -eposopoek- fall down clearly contains the locative noun root -poe down, but
-eposo- is hardly found as a verb root in isolation. It is used, however, in another
possibly lexicalized form -eposowakia- applaud < -waki palm of the hand. The
segmentation as in example (52) is thus somewhat artificial.


He tumbled/fell down.

Productive compounding in Baure (Arawakan)

One of the possible relationships between the verb and the incorporated noun is
that of predicate and argument. When this is the case, intransitive verbs incorporate the S argument, as in example (53). Transitive verbs incorporate the O argument, and when the incorporated noun is inalienably possessed, the possessor
may be raised and marked on the verb in the object marking slot, as shown by =ro
3sgm in example (54).
(53) raromosewapa to yashor.

to yashor
3sg=sink-clf:oval-cos art boat
The boat has sunk.
(54) nopiripoekier ach noemer eshe-ye.

ach no=im=ro
eshe -ye and 3pl=put=3sgm trunk-loc
They cut off his head and put it on a trunk.

Intransitive verbs that incorporate nouns are, however, expectedly unaccusative

(i.e. their S argument is a theme or patient), and incorporation is particularly productive with stative verbs. Furthermore, incorporation in Baure is not restricted
to direct objects, but oblique objects may be incorporated as well, of which some
authors claim that it is a rarer phenomenon cross-linguistically (Gerdts 1998:87).
In Baure, this happens, for example, when locative noun roots are incorporated.
An example is given in (55):


He is walking in front.

When oblique arguments are incorporated, they are often followed by the absolute
suffix -ko as can be noted in the examples above, or by the linking morpheme, as
in (56):
(56) ropanshiriaw to simori.

to simori
3sgm=follow-behind-lk-cop art pig
He is following behind the pig.

The presence or absence of the absolute morpheme or the linking suffix after
incorporated elements seems to be related to the verb semantics and the relation
between the verb and the incorporated element, but at present the function of the
absolute morpheme is not entirely clear yet.
The incorporation of a noun root into a transitive verb does not necessarily
cause the valency of the resulting verb to decrease. Example (51) already showed

Femmy Admiraal & Swintha Danielsen

that the full noun corresponding to the truncated form that was incorporated can
be repeated. When a complement of the verb is incorporated, it is less definite
and less individuated, and when the incorporated element is repeated as a full
NP, this is done for contrastive focus. Another example of an incorporated classifier is given in (57), where it is clear that the repetition of the full noun serves to
specify the referent, where the incorporated classifier fails to do so. The classifier
-po clf:multitude is also part of the lexicalized noun shep (*she- manioc + -po


ronipap shep.

3sgm-eat-clf:multitude-lk-pfv manioc.meal
He ate chive.24

4.3 Discussion of Type 2 compounds

We have shown in Section 4.1 that subordinate compounds consisting of a verbal
nexus (Scalise & Bisetto 2009:50) are exocentric, but they are just as rare as exocentric compounds are in general in Baure (see 3.1.4). Nominal root incorporation
is much more common. As Bybee (1985:108) confirms, [l]anguages which allow
incorporation are not common, but they are of considerable theoretical interest,
since they allow the fused expression of a greater variety of semantic notions than
other languages. She settles incorporation on a continuum of morphological
Compounding, incorporation, derivation, and inflection are on a continuum, in
which compounding is the freest, involves the largest (indeed an open) class of
items, with the richest and most specific meanings, and inflection is the most
constrained, involves the smallest classes of items with the most abstract and
general of meanings. (Bybee 1985:108).

In the Baure case, we have demonstrated that the same left-bound nominal
roots can be incorporated as there are compounded in N2 position in nominal
compounds. This means that here nominal compounding and mixed compounding show striking similarities, which is not always the case in languages that use
incorporation (see Bybee 1985:105). Gerdts (1998:99) claims that [t]rue noun
23. We can also argue that the repetition or explicit mention of a referent N within NPs with
adjectives is more frequently observed in cases where the reference is marked by a classifier.
24. Chiv is the Spanish word for toasted manioc meal, which is used in lowland Bolivia as
a type of condiment.

Productive compounding in Baure (Arawakan)

incorporation, where a noun stem compounds with a verb, is a typologically rare

phenomenon, thus Baure does not represent a frequently encountered language
In the nominal compounds, we generally call the compounded root to the
right (N2) the head of the construction. From a syntactic point of view, in incorporation, the verb root is the head of the construction. Nonetheless, in mixed
compounds with a verbal modifier in the first position, the incorporated noun
is considered the head. This does not mean that the structures look so different
morphologically, but whereas a mixed compound is a full noun, the outcome of
verbal incorporation is generally a verb base, which is bound and needs further
verbal morphology; and this proves the headedness of the verb. In fact, exocentric
compounds with a verbal root, like in 4.1 are so rare that more commonly a noun
involving a compounded verbal and nominal root has to be derived by the nominalizer -no nmlz. This is particularly common with colour terms, which derive
from a verbal root in Baure, as shown in Examples (58) and (59):
(58) te nia nikotipawapa.


dem3m 1sg=body 1sg=white-clf:body-cos
I am white in my body.
(59) kotipan

white (e.g. dog)

The verby character of the colour predicate in (58) is marked by the verbal suffix
-wapa change of state and the person cross-referencing of the subject in the proclitic slot, as obligatory for verbs and opposed to non-verbal predicates.
Incorporation is an extremely productive process in Baure. When considering
that either core arguments or obliques are incorporated, we see that we are actually
facing a process at the limits between morphology and syntax: Mithun [1984:847]
points out that noun incorporation is perhaps the most nearly syntactic of all morphological processes and Katamba [1993:287] adds that it obscures the boundaries between syntax and morphology (tekauer, Valera & Krtvlyessy 2012:44).
When a nominal root is incorporated in Baure, the verb alone can reference the
argument the incorporated element denotes. However, it is not ungrammatical
to have the same nominal root outside the verb again. This tendency holds much
stronger for classifier and locative roots, which are more general, but the observation is also an indication of the less lexicalized and grammaticalized character of
incorporation as a morphological or syntactic process.

Femmy Admiraal & Swintha Danielsen

5. Verbal compounds (Type 3)

This section refers to verbal compounds, the Type 3 in Section 2.2 (V+V = V).
Verbal compounding is only a limited productive process in Baure today and produces semi-grammaticalized constructions. In the first construction type the second verb root is more or less grammaticalized; in the second construction type, it
is rather the first verbal root that is almost grammaticalized in its position.
5.1 Verbal compounds leading to grammaticalization of suffixation
The verb base -pik- come occurs rarely as a simple verb root in modern Baure
data one example being (60) , whereas historical data provide many such examples (Magio 1880 [1749]). Already in historical data, this verb root is used in verb
compounds in second position, functioning as a directional (intentional) suffix
-pik all, which is its predominant use today. In recent Baure data, the allative
construction can be contrasted with the intentional/ablative marking with the suffix -pa, as is also done in the utterance of (61).25
(60) ndi nipikow pon sorati-ye.

ndi ni=pik-wo
1sg 1sg=come-cop other-clf:general village-loc
I come from another village.
(61) nokope piyonpik ach neriki piyonpa noiy.

nokope pi=yon-pik ach neriki pi=yon-pa
yesterday 2sg=walk-all and now 2sg=walk-intl there
Yesterday you walked here (to visit me) and now you walk there.

5.2 Verbal compounds with empty verb root

In the other semi-grammaticalized construction, the first verb root is always the same,
and only the second verb root varies: The first verb, -ke- do, say, go, is semantically
rather empty. It is used in the manner construction do how, either with the general
root -toeri- do how (62), or less often, with a specific verb root in the second position
(63). The nominalization in the examples below marks the interrogative construction.
(62) pikietoerin to chichorop?

to chichorop art bean
How will you do (prepare) the beans?

25. The verbal suffix -pa, however, does not derive from a verb root.

Productive compounding in Baure (Arawakan)

(63) pikevesachowon?

How did you read it (What do you read it with)?

In the constructions expressing manner, the first verb root developed into a prefixlike element.
5.3 Discussion of Type 3 compounds
It is very well possible that verbal compounding used to be more common in the
past. We may possibly find more verbal affixes that can be traced back to verbal
roots.26 For suffixes, we do not have any other clear evidence than the example of
-pik all in 5.1. For prefixes, on the other hand, there are possibly more examples
than the one presented in 5.2, after a thorough analysis of all Baure verbs, especially considering the so-called lexical prefixes (see Danielsen 2007:225) veundo, ja- off , ej- out, and epo- back, down, which may be lexicalized verb roots.
Verbal compounding is similar to the other compounding types in that two
lexical roots are simply juxtaposed, and since both roots are verbal, the product is
also a verb. The issue of headedness, however, is difficult to determine and may be
different in the two contrastive cases of 5.1 and 5.2. Verbal compounding can be
well distinguished from serial constructions, which are very frequent in Baure. In
any serial verb or complement construction, the verb has to be a complete word
with at least subject marking. The serial constructions we find in Baure today do
not seem to be the source of the verbal compounding described above. At least the
latest grammaticalization process of verb roots into free preverbal particles shows
a different tendency (see Danielsen 2014). Compare the following two examples
of a serial construction with -kach- go in (64) and a particle construction with the
same verb grammaticalized in (65):
(64) ndi nikachpikow nijinoekpikovi.

ndi ni=kach-pik-wo ni=jinoek-pik-wo=pi
1sg 1sg=go-all-cop
Me, I came to see you.
(65) kach rojinoekpa tekow to ndorinev.

kach ro=jinoek-pa
tek-wo to ni=tori-nev

intl all-cop art 1sg=friend-pl
He went to see all our friends.

26. Note that currently verbs tend to grammaticalize into preverbal particles, creating a
more isolating language than Baure used to be (see Danielsen 2014).

Femmy Admiraal & Swintha Danielsen

6. Conclusions
In this article we have shown that Baure has three compounding types, one
where two nominal roots are combined (Type 1), mixed compounds with a verbal root combined with a nominal root (Type 2), and verbal compounds with
two verbal roots (Type 3). We could observe that verbal compounding, even
though present in the Baure language, is distinct from the two other compounding types involving nominal roots. Verbal compounding is rare and has in both
demonstrated cases almost grammaticalized into affixation. We can therefore
only presume that verbal compounding was a once more common word formation process, even though it may have been more productive, as a number
of complex verbs derived by a so-called lexical prefix indicate. Nonetheless,
verbal lexemes rather tend to combine with either nominal roots or only affixes
and clitics.
Apart from verbal compounding, it is verbal morphology that gives the Baure
language its polysynthetic character. Nominal morphology has been argued to be
less complex than verbal morphology. However, both nominal and verbal morphology share one morphological strategy, namely the lexical derivation with a bound
nominal root to the right of the base lexeme. The productivity of the processes of
nominal compounding and incorporation is typical of Baure and illustrates the
connection between morphology and syntax. Since we claim that the bound roots
are always to be analysed as lexical roots rather than grammatical (derivational)
morphemes, we categorize the processes as compounding (see Figure 1). Being
aware of the fact that we are referring to a continuum where compounding is on
the most lexical side, followed by incorporation, derivation, and finally inflection,
we note that in Baure, compounding and incorporation seem to be almost on the
same place in the continuum, but both processes are possibly less lexical than in
many other languages, and may even be viewed as derivational in the examples
with classifier elements in the N2 slot. In Baure we also have a vast variety of highly
grammatical operations, which can be contrasted with these lexical processes, on
the one hand, and we have shown how we define phrasal constructions, which
are also distinct from productive compounding, on the other hand. As a whole,
Baure does not have much inflection either, which may be relevant for the general
characterization as well.
When authors like Zamponi (2009) claim that Arawakan languages are
typically not compounding, then we are in fact facing a definitional problem
(Lieber & tekauer 2009:17), where we draw a different line between lexical and
grammatical morpheme on the side of the components, and between a phrasal
and a word formation construction, on the side of the construction type.

Productive compounding in Baure (Arawakan)

abs = absolute; ADJ = adjective; all = allative; appl = applicative; art = article;
aug = augmentative; clf = classifier; cop = copula; cos = change of state; dem
= demonstrative (1, 2, 3 different pragmatic types); der = derivational suffix; f =
feminine; intl = intentional; lk = linking morpheme; loc = locative; m = masculine; N = noun; nmlz = nominalizer; NUM = numeral; pfv = perfective; pl =
plural; poss = possessive; sg = singular; V = verb

Adelaar, Willem F.H. with Muysken, Pieter C. 2004. The Languages of the Andes. Cambridge:
CUP. DOI: 10.1017/CBO9780511486852
Admiraal, Femmy. 2013. Baure motion events: Exploring the grammatical resources. In Expresin de nociones espaciales en lenguas amaznicas, Ana Mara Ospina Bozzi (ed), 6184.
Bogot: Instituto Caro y Cuervo.
Aikhenvald, Alexandra Y. 1999. The Arawak language family. In The Amazonian Languages,
Robert W.M. Dixon & Alexandra Y. Aikhenvald (eds), 65106. Cambridge: CUP.
Aikhenvald, Alexandra & Muysken, Pieter C. (eds). 2011. Multi-verb Constructions. Leiden:
Brill. DOI: 10.1163/ej.9789004194526.i-313
Baker, Mark C. & Fasola, Carlos A. 2009. Araucanian: Mapudungun. In Rochelle Lieber & Pavol
tekauer (eds), 594608.
Booij, Geert. 2009. Compounding and construction morphology. In Rochelle Lieber & Pavol
tekauer (eds), 201216.
Bybee, Joan. 1985. Morphology: A Study of the Relation between Meaning and Form [Typological
Studies in Language 9]. Amsterdam: John Benjamins. DOI: 10.1075/tsl.9
Danielsen, Swintha. 2007. Baure: An Arawak Language of Bolivia. Leiden: CNWS.
Danielsen, Swintha. 2014. Observing the evolution of TMA particles replacing agglutinating strategies. In Taming the TAME Systems [Cahiers Chronos 27], Emmanuelle Labeau &
Qiaochao Zhang(eds). Amsterdam: Rodopi.
Danielsen, Swintha. forthcoming b. Afijos compuestos y su lexicalizacin en el baure (arawak).
In Serie Volmenes Temticos de la Sociedad Argentina de Lingstica, Marisa Censabella,
Cristina Messineo, Hebe A. Gonzlez & Beatriz Gualdieri (eds), Buenos Aires: Sociedad
Argentina de Lingstica.
Fabb, Nigel 1998. Compounding. In Andrew Spencer & Arnold M. Zwicky (eds), 6683.
Fabre, Alain. 2005. Diccionario etnolingstico y gua bibliogrfica de los pueblos indgenas
sudamericanos: Arawak.
pdf (16 December 2011)
Fabre, Alain. 2007. Morfosintaxis de los clasificadores posesivos en las lenguas del Gran Chaco
(Argentina, Bolivia y Paraguay). UniverSOS: Revista de Lenguas Indgenas y Universos
Culturales 4:6785.
Gerdts, Donna B. 1998. Incorporation. In Andrew Spencer & Arnold M. Zwicky (eds), 84100.

Femmy Admiraal & Swintha Danielsen

Krauss, Michael. 2007. Classification and terminology for degrees of language endangerment. In
Language Diversity Endangered, Matthias Brenzinger (ed), 18. Berlin: Mouton de Gruyter.
Lieber, Rochelle & tekauer, Pavol (eds). 2009. The Oxford Handbook of Compounding. Oxford:
Lieber, Rochelle & tekauer, Pavol. 2009. Introduction. In Rochelle Lieber & Pavol tekauer
(eds), 318.
Magio, Antonio P. 1880[1749]. Gramtica de la lengua de los indios Baures de la provincia de
Majos. In Arte de la lengua de los indios baures de la provincia Moxos, conforme al manuscrito original del padre Antonio Magio, Lucien Adam & Charles Leclerc (eds), 153. Paris:
Maissonneuve & Cia, Libreros Editores.
Mithun, Marianne. 1984. The evolution of noun incorporation. Language 60 (4): 847894. DOI:
Payne, David L. 1991. A classification of Maipuran (Arawakan) languages based on shared lexical retentions. In Handbook of Amazonian Languages, Volume 3, Desmond C. Derbyshire
& Geoffrey K. Pullum(eds), 355499. Berlin, New York, Amsterdam: Mouton de Gruyter.
Rivero Pinto, Wigberto. 2006. Pueblos indgenas de Bolivia: joaquiniano. At: http://amazonia.
bo/mas_detalle_proi.php?id_contenido=15 (16 December 2011)
Sadock, Jerry M. 2006. Incorporation. In Encyclopedia of Language and Linguistics, Vol. VIII,
2nd ed., Keith Brown, Anne H. Anderson, Laurie Bauer, Margie Berns, Graeme Hirst, Jim
Miller (eds), 584587. Boston: Elsevier.
Scalise, Sergio & Guevara, Emiliano. 2005. The lexicalist approach to word-formation and the
notion of the lexicon. In Rochelle Lieber & Pavol tekauer (eds), 147188.
Scalise, Sergio & Bisetto, Antonietta. 2009. The classification of compounds. In Rochelle Lieber
& Pavol tekauer (eds), 3453.
Spencer, Andrew & Zwicky, Arnold M. (eds). 1998. Handbook of Morphology. Oxford: B
tekauer, Pavol & Lieber, Rochelle (eds). 2005. Handbook of Word-Formation. Dordrecht:
Springer. DOI: 10.1007/1-4020-3596-9
tekauer, Pavol, Valera, Salvador & Krtvlyessy, Lvia (eds). 2012. Word-Formation in the
Worlds Languages: A Typological Survey. Cambridge, New York: Cambridge University
Press. DOI: 10.1017/CBO9780511895005
Szab, Henriette va. 1998. Demanda Territorial Baure. Trinidad und La Paz: Viceministerio de
Asuntos Indgenas y Pueblos Originarios. Ms.
Terhart, Lena. 2009. Klassifikatoren im Baure. MA thesis, Free University of Berlin.
Zamponi, Raoul. 2009. Arawakan: Maipure-Yavitero. In Rochelle Lieber & Pavol tekauer (eds),

Two types of incorporation in Ese Ejja

Marine Vuillermet

(University of California, Berkeley)*

Ese Ejja is an Amazonian language that displays two types of noun incorporation.
The first type is typical of the Amazonian area: it occurs within verb predicates, is
restricted to inalienable nouns and has no influence on the valency of the verbal
predicate. The second type is unusual in that it involves adjectival predicates and
results in adjectives, semantically comparable to the English derived adjectives
blue-eyed or red-haired. However, the phenomenon is far more productive in
Ese Ejja: adjectives do not only incorporate body-part terms but also possessed
nouns, and even NPs and verbal roots. Their modifying function compensates for
the quasi absence of attributive adjectives in the language. The analyses presented
here are based on first-hand data collected in a Bolivian community.
Keywords Ese Ejja; incorporation; adjectives; adjectival predicates;
verbal compounding

1. Introduction
Ese Ejja [eseexa], a Takanan language spoken in the Bolivian and Peruvian lowlands, displays two types of noun incorporation. The first type of noun incorporation (NI), the regular type of noun incorporation, meets Mithuns (1984:847)
definition: a noun stem is compounded with a verb stem to yield a larger, derived

* I am truly and deeply indebted to the Ese Ejja people, especially those living in Portachuelo
Bajo, and I warmly thank all of them here. I am particularly grateful to my main consultants
Kanono Calixto Calla Casirno and to Sooai Florentina Calla Varga, to the participants
of the workshop on adjectives in 2007, and to my host family, Inotawa Mara Calla Santa
Cruz, Ernesto Monje Racua and their relatives. Fieldwork was made possible thanks to various
grants (HRELP-SOAS 2007, ELF 2008, AALLED 20082009). I also wish to thank Swintha
Danielsen, Katja Hannss and Fernando Ziga for their offer to participate in this publication
and their helpful comments on earlier drafts. I am grateful to Lucy Zuberbhler to have proofread my paper. Finally, I have benefited from valuable comments made by Zachary OHagan
and Antoine Guillaume. I am responsible for any omissions and errors.

Marine Vuillermet

verb stem (my emphasis). The second type of NI is more remarkable in that it
does not involve a noun and a verb stem, but (mostly) a noun and an adjective
stem. Consequently, the resulting compound is not a larger verb stem but a larger
adjective stem.
The two types are illustrated in (1) and (2) respectively. In (1), the nominal
root -jyoxi foot is compounded with the verb root jeyo- tie.

A'a kwichi jyoxi-jeyo-naje?

(e-jyoxi npf-foot)1
int pig.abs foot-tie-pst

Did (you) tie up the foot of the pig (lit. did you foot-tie the pig)?
{volunteered (jeyo).006}

The incorporated noun root and the verb root, together with the past marker
(obligatorily associated with the verb root), form a single word on phonological and morphological grounds. Syntactically, the incorporated noun -jyoxi foot
vacates its position of core argument, and the possessor of the foot (kwichi pig)
can thus fill the vacated position. This manipulation of case role corresponds to
Type II of Mithuns (1984:856) typology: NI advances an oblique argument into
the case position vacated by the incorporated noun.
NI into adjectives is illustrated in (2): the borrowed nominal root chinela flipflop is compounded with the adjective root kya-beje (adjectival prefix)-thin.

2sg.abs apf-flip_flop-thin-very~red
Your flip-flops are very thin. (lit. you are very thin-flip-flopped) {beje.006}

The two types of NIs are similar in many ways. In (2), the two roots (plus the other
morphemes attached to the adjectival root) form a single word on phonological
and morphological grounds. The possessor miya 2sg fills the position vacated by
the incorporated noun. NI into adjectives (henceforth NIA) and into verbs (henceforth NIV) thus share the following properties: morphologically, both result in
one-word compounds. Syntactically, both manipulate case roles. Semantically,
both involve a possessor-possessee relation between the notional and the new P
(or S) argument.2 Discursively, the salient human argument is thus foregrounded.
The shared properties argue for an analysis in term of noun incorporation in both

1. npf stands for Noun PreFix. This prefix characterizes the e-nouns in their citation form
(see Section 2.2.a below for more details).
2. S is the single argument of an intransitive verb, while A is the agent-like argument and P
the patient-like argument of a transitive verb.

Two types of incorporation in Ese Ejja (Takanan)

The debate on the specific functions of noun incorporation was opened by

Mithun (1984). In her reference paper, she demonstrates that NI is widespread
around the world and accross language families and that different languages display distinct degrees of cohesion (usually reflecting the morphological properties
of the language). More importantly, she also establishes four distinct types corresponding to four distinct functions. She therefore argues that NI is not an obligatory syntactic mechanism, but has strong functional motivations. This position
has never been seriously questioned (see the discussion between Sadock (1986)
and Mithun (1986), or Bakers (1999:295) own judgement revision) and was confirmed by various studies (see esp. Rose 2008 for mrillon or Velzquez-Castillo
1996, 1999 for Guaran).
The two NIs in Ese Ejja are no exception and are in fact motivated. NIV in Ese
Ejja can be analyzed as a case of external possession, i.e. a construction in which
a semantic possessor-possessum relation is expressed by coding the possessor as
a core grammatical relation of the verb and in a constituent separate from that
which contains the possessum (Payne & Barshi 1999:3). In numerous Amazonian
languages, NIV is frequently (or mostly) restricted to body-part terms, inalienable
or dependent nouns (see for example Dixon & Aikhenvald (1999:45), Aikhenvald (2012:194), and also Rose (2008) for the Tupi-Guaran languages or Michael
(2006) for the Arawakan language Nanti). Sapir already stated that noun incorporation tends to involve body-part terms or inalienable nouns in the Americas
(Sapir 1911:253), and this particularity is attested cross-linguistically (see e.g. the
Australian, non-Pama-Nyungan language Murrinh-Patha (Walsh 1995). By contrast, NIV in Ese Ejja does not only involve body-part terms of the inalienable
e-noun class, but any noun of the semantic class of possessed nouns. This specificity highlights the important functional role of NI.
NIA is much less common and has only been reported for a handful of languages. It seems to display a distinct function of (noun) modification, due to the
adjectival nature of its incorporating predicate. Furthermore, it often cumulates
with the expression of (external) possession. The importance of the modifying
function is revealed by the variety of incorporated lexemes: not only possessed
items can be incorporated in NIA, but any entity that requires modification. I
argue that this incorporating construction is a response to the very small (and
unproductive) class of attributive adjectives in the language, a strategy allowing
the predicative adjectives to modify the incorporated lexemes.
The present paper is organized as follows: the second section (Section 2)
presents the speakers and the main typological features of the language. It also
examines the three word classes involved in the two types of NI, namely nouns,
verbs and predicative adjectives. The third and fourth sections describe the morphological and functional properties of NIV (Section 3) and NIA (Section 4).

Marine Vuillermet

The fifth section (Section 5) explores the function of the two NI and focuses on
NIA as a functional response to the virtual absence of attributive adjectives in
Ese Ejja.
2. The Ese Ejja people and their language
The Ese Ejja live in nine communities situated along the Madre de Dios and the
Beni rivers and their tributaries. The language vitality is relatively well preserved
in the Bolivian communities, where the language is still transmitted to younger
generations and spoken in everyday conversations. The situation is different in
Peru, where language loss is evident (Alexiades, p.c. by email, May 6, 2012). See
Vuillermet (2012:69ff.) for a detailed account of the sociolinguistic situation in
each community.
Ese Ejja belongs to the Takanan family, together with Araona, Takana, Reyesano/Maropa and Cavinea, all of which are spoken solely in Bolivia. The Takanan
languages are further divided into three branches according to Girard (1971:41):
the takanik branch (Araona, Takana and Reyesano/Maropa), the kavinik branch
(Cavinea), and the chamik branch (Ese Ejja).
2.1 Grammatical features of Ese Ejja
Ese Ejja is best regarded as a verb final language, although word order is fairly
flexible in simple declarative clauses. The language displays ergative alignment,
both in case marking and in very limited verb indexation, as illustrated in (3). The
word for child eshoi, as an S argument in (3)a, or as a P argument in (3)b, takes
the absolutive. Also, in both sentences, this absolutive argument is not indexed on
the verb. By contrast, the A argument dokweiai big stag in (3)b takes the ergative
clitics =a and is indexed on the verb by -ka 3A.
(3) a. Intransitive clause


npf-child=abs shout-prs

The child is shouting. {SoFWA.021}3

3. The curly brackets after each free translation indicates the type of data used. When data
come from narratives, the two first letters stand for the name of the consultant, the following
three letters for the name of the story, and the three digit number for the number of the sentence in the Toolbox file. Data also come from a workshop on adjectives I organized in 2007
with a dozen Ese Ejja. In order to investigate the morphological possibilities of adjectives, I

Two types of incorporation in Ese Ejja (Takanan)


Transitive clause
Dokwei-ai=aA e-shoi=P
stag-big=erg npf-child=abs throw-put_down-depr-3A-prs
The big stag throws the child away. {BaFWA18.001}

Transitivity is very straightforward in Ese Ejja. This characteristic is of importance

for this paper, because I argue that NI in Ese Ejja manipulates roles but has no
influence on the valence of the verb. Ese Ejja verbs are either transitive or intransitive (only 3 ambitransitive verbs were found). As a consequence, the valency
status of the predicate is never ambiguous, even though Ese Ejja is a pro-drop
language. For example, kwiji- bark at X is transitive in Ese Ejja. In (4)a, P is not
expressed but the sentence remains transitive and must be translated as the dog,
it barks at him (and not *the dog, it barks). In order to obtain the intransitive
counterpart, antipassive reduplication must be used, as in (4)b where the root is
fully reduplicated.

kwiji- bark at X vs. kwiji~kwiji bark

a. Owaya iawewa=a kwiji-ka-ani
3.erg dog=erg bark_at-3A-prs

The dog, it barks at (him).

{SoFWA.076} *the dog barks

b. Oya ma
iawewa kwiji~kwiji-ani.
3.abs dem2 dog.abs bark_at~red-prs

This dog, it barks. {SoFWAVc.019}

Morphologically, the third person Agent is only indexed on the transitive verb in
(4)a. In addition, the subject takes an ergative in (4)a and an absolutive in (4)b
(owaya 3.erg and iawewa=a dog=erg vs. oya 3.abs and iawewa dog.abs,
2.2 Main word classes
The description of the specificity of each class is relevant to understanding the
functioning of NI in Ese Ejja. Main word classes include nouns, verbs, and two
distinct classes of adjectives, which are described in turn.

prepared a list of a hundred adjectives found in Wyma and Pitkin de Wymas (1962) lexicon
and in my own corpus. The Ese Ejja volunteered various morphological combinations with
each adjective of the list; they had to provide contexts, and the volunteered sentences had to
be validated by the whole group. These data are annotated as {volunteered (adjective)}. Fieldnotes (always re-checked with consultants) and elicitation (based on natural data o
elsewhere) are other sources for examples; I refer to them as {fieldnotes} and {elicited}

Marine Vuillermet

Nouns can be subdivided into independent nous, e-nouns, kinship nouns and
proper nouns (Vuillermet 2012:298ff). NI can involve any of these subclasses,
though there is only one example of NI with a proper noun. By contrast, e-nouns
are very frequently incorporated and are therefore of specific interest for this
The e-noun class corresponds to the inalienable noun class in other languages.5 It consists of one hundred items which mostly refer to body parts or
parts of plants (71 items), but also include landscape and natural elements
(7), spatial (7) and kinship relations (6), and other less specific domains (12).
These semantics are fairly common cross-linguistically (see again Chappell &
McGregor (1995) for the semantics found in inalienable classes).
Morphologically, e-nouns are bound roots which require a nominal prefix enpf to appear as independent NPs, as in (5)a. However, the root appears without
e- prefix in a number of morphological processes, including NI. The next examples
contrast the e-noun -sapa head in its citation form (with the e-prefix) in (5)a, as
an adjective with the suffix -m priv (without the e-prefix) in (5)b, and incorporated into the verb ixya-kaa- tighten (eat-strong) in (5)c.

citation form
e-sapa (npf-head) head


derivation (into a predicative adjective)

sapa-m (head-priv) without head, scatterbrain




(it) tightens (its) head

head-eat-strong-3A-pst (lit. head-eated-strong)

The absence of the e-prefix in NI is used as a morphological argument for the

dependence of the e-noun on the verb and the tightness of the two lexical roots
nominal and verbal (Section 3.1)/ adjectival (Section 4.1).
Kinship nouns are a small class consisting of eight kinship terms. They behave
like the e-nouns, but the affix which turns a kinship root into an independent NP
is the circumfix e--mese. An independent noun is any noun that does not belong

4. The e-nouns are typical for the whole Takanan family. See Guillaume (2008:450ff) for
Cavinea, Emkow (2006:207ff) for Araona, Ottaviano & Ottaviano (1989), cited in Guillaume
(2008:409) for Takana, and Guillaume (2012) for Maropa. Cross-linguistically, the inalienable
noun class is often the only one from which nouns can be incorporated.
5. Also called obligatorily possessed or dependent nouns, as e.g. in mrillon (Couchili et
al. 2002).

Two types of incorporation in Ese Ejja (Takanan)

to the three other classes, especially borrowed nouns; they do not display specific
b. Verbs
Ese Ejja verbs can be inflecting or non-inflecting.6 Only the inflecting verbs are
relevant for this study, and their predicate structure is given in table 1. Although
verbal predicates may be fairly complex (13 slots), only verbal indexation (Slot +4)
and tense/mood (Slot 3, +6 or +9) are obligatorily marked (in bold in Table 1).










(incorpor. noun)




Aktionsart + AM










Table 1. Basic structure for inflecting verbs

Examples (6)ab display two transitive verbs, with the minimal obligatory morphology in bold. (See also (3)a and (4)ab above).



Akwi pojo-je.
tree divide-fut
I will chop wood.{elicited}


akwi jaja-ka-ani.
man-white=erg tree cut-3A-prs
Ernesto (lit. white-man) is cutting wood.{elicited}

The lexical base of the verb is either monomorphemic, as in most previous

examples, or bimorphemic, when the lexical verb is a compound. Examples
(7)ab illustrate the two monomorphemic verbs jaja- cut and taxa- squeeze,
while (7)c exemplifies how the two verbs can be grouped together to form a
bimorphemic stem (see Vuillermet (2012:394ff, forthcoming) for more details
on compounds).

6. Non-inflecting verbs require a copula to bear the obligatory tense and mood morphology.
This verb class consists of loanverbs and derived inflecting verbs (see Vuillermet 2012:381ff.
for more details). For the sake of clarity, this study does not address cases of NI into noninflecting verbs, although it is possible (Vuillermet 2012:518).

Marine Vuillermet



[jaj]-ka-ani he/they cut (s) {elicited}


b. [tax]-ka-ani he/they squeeze (s) {elicited}



[jaj-taxa]-ka-ani he/they cut (s)/chop (s){elicited}


In either case (mono- or bimorphemic root), the incorporated noun precedes the
(first) verb root: it always occurs in Slot 1 (framed in table 1). Example (8) shows
that the incorporated noun doxoxanixi rib can only appear before the two verb
roots kwya-saja- as in (a), not in between (b).



npf-child mid-rib-press-break_in_two-mid-pst
The child broke his rib. {elicited}



Ese Ejja verbs are either transitive or intransitive, and both types allow noun incorporation, as highlighted below in Section 3.
c. Adjectives
Ese Ejja has two distinct classes of adjectives, both labeled after their main function: attributive adjectives and predicative adjectives. Attributive adjectives consist
of a very small class of bound roots that always form compounds. When modifiying a noun, they attach to its right and form a single word with it, as in (3)b,
dokwei-ai (stag-big) big stag.7 However, the resulting noun compound should
often no longer be analyzed as a noun plus an adjective, but rather as a lexical
entry, as is the case of e-me-ai (npf-hand-big) thumb and e-me-sisi (npf-handsmall) finger (except thumb). In comparison with predicative adjectives, attributives adjectives play a very small role.
Predicative adjectives constitute the largest class of adjectives. They cover
most semantic categories established by Dixon (1982), except for similarity and
age: dimension, value, color, physical property, qualification, human propensity,
position, speed, quantification and difficulty. As their name suggests, predicative adjectives are mostly used in a predicative function in copula clauses, with
possible noun incorporation. A copula (as in (9)a) or a posture verb in a copula

7. As Fernando Ziga suggested, these bound roots could alternatively be analyzed as


Two types of incorporation in Ese Ejja (Takanan)

function (as in (9)b) may specify the tense/mood, but the tense/mood value can
also remain implicit.



ekwana po-ani.
apf-frightened 1excl.abs be-prs
We are afraid. {SoCre.018}


npf-child apf-angry-very~red sit.prs
The child is (lit: sits) very angry. {EpFWA.014}

The copula subject is encoded in the absolutive case. Predicative adjectives also
occur in adverbial function or in verb compounds to a lesser extent.
NI only occurs in the class of predicative adjectives, and more specifically only
in the subclass of non-derived adjectives, called the kya-adjectives after their form.
Morphologically, the kya-adjectives are bound roots which require one of four
affixes to form independent words, as illustrated in (10)ad. The adjectival prefix (henceforth apf) kya- is semantically neutral (a), while the other three affixes
negate (b), attenuate (c), or question the adjectival root (d).

kya-kemo big


kemo-ama not big





ache-kemo how big



not quite big

When incorporating a root, they require any affix, as illustrated below with the
interrogative prefix ache- how and the negative prefix -ama priv.


Ache-bakwa-wiso miya? (e-bakwa child)

how-child-many 2sg.abs
How many children do you have?{fieldnotes}

b. Ma kwiixi wanase-jya-ama. (e-wanase wife)

this man wife-lovely-priv

This man does not like (his) wife. (lit. he is not lovely-wifed)
{volunteered (jya).009}

These predicative adjectives also participate in verb compounds (Vuillermet

forthcoming). In this case, adjectives occur without affixes, as illustrated with
kyachwe wet.

Marine Vuillermet



kya-chwe wet


[kwya-chwe]- wet someone (by squeezing/hitting him

when one is wet)

Predicative adjectives may also participate in noun compounding. Such compounds correspond to a highly specific function, as they are only used to tease or
insult people, or to name people as in the following examples.



Me-woo hand-red
Ino-tawa thread-green
Sapa-siyo head-shiny (better translated as bald face)

This very specific morphological process of noun compounding is productive, but

functionally highly restricted. More importantly, it is distinct from NIA. As will be
demonstrated, NIA is highly productive and transparent in meaning.
These noun-adjective compounds may actually be picked up as names by parents for their child, so as to let people know that they know about a gossip or insult.
Interestingly, Peluso (2004:104) cites the words of an Ese Ejja explaining how the
Ese Ejja do not choose names because they like them, but because certain names
have reasons to be. Names are originally semantically motivated and the meaning of the morphemes that compose them can be retrieved in most cases (Peluso
2003:300ff). On the other hand, in my fieldwork experience, every time I tried
to decompose a name, the Ese Ejja insisted on the fact that proper names should
not be translated, as they were only names, i.e. that the compounds are lexicalized and their names should not be decomposed, because they merely became a
proper name. Once they are given, I believe, Ese Ejja regard their names as units.
See Peluso (2003:291ff) for a detailed account of the naming process in Ese Ejja.
3 Noun incorporation into verbs (NIV)
NIV is listed as a characteristic areal feature of the Amazon by Dixon & A
(1999:9). At first sight, NIV in Ese Ejja virtually corresponds to the prototypical
NI for the Amazon as described by them and in Aikhenvald (2012:194):
the incorporated noun precedes the verb root;
verbs incorporate body-part terms (but not only);
verbs need not be transitive, i.e. they may be intransitive.
This section addresses in turn the morphology of NIV in Ese Ejja (Section 3.1),
the types of nouns incorporated into verbs (Section 3.2), the argument structure
(Section 3.3) and the type of verbs which can incorporate nouns (Section 3.4).

Two types of incorporation in Ese Ejja (Takanan)

3.1 Morphology of the NIV

In Ese Ejja, the incorporated noun forms a single word with the verb stem it
is compounded with. The incorporated noun in Ese Ejja immediately precedes
the verb root or stem, i.e. it occurs in Slot 1 (see Table 1 above for the verbal
structure). This position in the Ese Ejja verb structure is easily exemplified with
the presence of the middle marker in the preceding Slot 2 (xa--ki mid in
(14)a), or with that of the jussive marker in Slot 3, as (ka--awa) juss in



Ena pwa=jo
(e-sapa head)
water be=tmp.DS mid-head-cover-mid-prs
When it rains, one covers ones head.{elicited}


Nawoo ka-see-jya-ka-awa.
(e-see bowels)
Let him gut the fish! {elicited}

How tight the resulting compounds are can be seen by considering their stress patterns. In Ese Ejja, independent words contain one stressed syllable each. Similarly,
an incorporated noun and the new derived verb stem contain only one stressed
syllable betweem them, i.e. they form a single phonological word. Compare the
two sentences in (15): in (a), the P argument akwi-je tree-skin is an independent
(compound) noun and has a stressed syllable of its own, just like the predicate
jajxojaa I cut off. In (b), the root -jee skin is incorporated into the verb predicate, which still contains only one stressed syllable. The monomorphemic P argument kwi tree also has a stressed syllable of its own.



akwi-je jaj-xoja-aa.
(e-jee skin)
1sg.erg tree-skin cut-peel-prs.A1/2
I cut off the tree-bark (lit. tree-skin). {elicited}


kwi je-jaja-xoja-aa.
(e-jee skin)
1sg.erg tree skin-cut-peel-prs.A1/2

I cut off the bark of the tree (lit. I skin-cut-peel the tree). {elicited}

The last argument in favor of an analysis as a single word comes from the
e-noun class, the class which most incorporated nouns belong to. As mentioned earlier (Section 2.2.a), when an e-noun is incorporated, only the root
is incorporated. Were the e-noun to stand as an independent noun next to the
verb, it would require an e- prefix. The citation form of the e-nouns is given
on the right of the examples (as in (1), (11)ab, (14)ab, (15)ab and in the
examples below).

Marine Vuillermet

3.2 Nouns that can be incorporated into verbs

Only body parts are subject to incorporation in most Amazonian languages, as
was briefly mentioned in the introduction. Body-part terms are the most frequent
nouns to be incorporated in Ese Ejja (the feet of the pig in (1), the heads of people
in general in (14)a, and the bowels of the fish in (14)b), but any noun which is possessed can in fact be incorporated. In (16)a, the incorporated noun is daki clothes.
It belongs to the independent noun class, and does not refer to a body part, but to
an item possessed by the subject (the ancestors).8


...ajyo onaya
Q.loc indf.abs mid-clothes-get-mid-sit/ipfv-pst
(to know) what they made their clothes out of
(lit. they clothes-get-themselves). {KaBab.004}

Interestingly, Warray (non-Pama-Nyungan, Gunwinjguan) displays the exact

reverse situation (Walsh 1995:150): while the word for clothes belongs to the
inalienable class morphologically (same class prefix as body-part terms), this word
is analyzed as syntactically alienable and cannot be incorporated, unlike the bodypart terms of the inalienable class.
Here, possession is understood loosely as any item belonging to the referents
sphere. In (17) below, the e-root -jyoxi trail is incorporated, and is conceived of as
belonging to the personal sphere of the people going along it.

Jamaxeya esea
owaya jyoxi-kea-ka-ani.
therefore 1incl.abs 3.erg trail-block-3A-prs
Therefore they (vipers) block our trail. {KaPey.032}

Nouns referring to part-whole relationships can also be incorporated. In (18), the

wick of the kerosene lamp is incorporated. Note that the term for wick is a semantic extension of nose, and one could alternatively analyzed it as an example of
body-part incorporation.


(e-wi nose)
kerosene_lamp nose-heat-short-pst
The kerosene lamp got its wick shortened by heating
(lit. the lamp nose-heat-shortened). {chia.004}

8. Interestingly, Rose (2008) mentions that in mrillon, only the dependent nouns can be
incorporated, and manufactured objects do belong to this class. However, the next example
demonstrates that non-manufactured possessed items can also be incorporated in Ese Ejja.

Two types of incorporation in Ese Ejja (Takanan)

Similar cases are not rare cross-linguistically. In Nichols (1988:572, 1992:160)

hierarchy, kinship and body-part terms are the most likely to be inalienable,
immediately followed by part-whole (and spatial relation) terms.9 The nouns
incorporated in Nanti (Arawak), for example, include body parts and also plant
parts (Michael 2006:2).
Finally, compound nouns, and not only single nominal roots, can also be
incorporated into verbs. In (19), the only instance of this type, (e)-naba-na mouth
blood is a compound of two e-noun roots, -na blood and -naba mouth.


Ache emanomeexi=a=shai
which illness=erg=on_earth 3.abs
Which illness on earth made him spit blood
(lit. make him mouth-blood-go out)? {KaPey.073}

When possessed nouns are incorporated, their possessors fill their case role, as is
described in the next subsection.
3.3 Argument structure
The impact of NI on the argument structure is described by Mithun (1984) for the
NIs of Type II: the valence of the verb is preserved, but the case role vacated by the
incorporated noun is filled by an oblique argument. In Ese Ejja, it is the possessor
that occupies the vacated case role. The next example contrasts incorporation and
its syntactic periphrastic counterpart. In (20)a, the NP [iawewaja sapa] dogs
head is the P argument of the predicate: -sapa head is the NP head and iawewa
dog is its genitive modifier.10 In (20)b, the original NP head -sapa is incorporated
and the original genitive modifier iawewa (ja) is now a core argument of the

9. Ewe and Mandarin invalidate Nichols hierarchy in that spatial relations in these languages are the most inalienable category (see Ameka (1995) and Chappel & Thompson
(1992) respectively, cited in Chappell & McGregor (1995, 8)). However, Nichols based her
hierarchy on several North American languages plus languages from other regions and families, confirming that such semantics groupings are not arbitrary but correspond to a cognitive reality.
10. Although iawewaja sapa the head of the dog is written in two separate words, they
form a single phonological word with a single stress syllable.

Marine Vuillermet



Tawoo=a iawewa=ja sapa

bottle=erg dog=gen
head[abs] eat-strong-3A-pst
The bottle tightens on the dogs head.{elicited}

b. Tawoo=a iawewa sapa-ixya-kaa-ka-naje.

bottle=erg dog[abs] head-eat-strong-3A-pst

The bottle tightens on the dogs head (lit. it head-eat-strong the dog).

The verb remains transitive, as the ergative marking of the subject tawoo=a
bottle=erg and the 3A person indexation demonstrate.
Incorporation into intransitive predicates is less frequent and no syntactic
periphrastic counterparts are attested. However, the structure is similar. In (21),
the possessor of the (incorporated) skin -jee is encoded as the unique core argument of joka- dry: eya stands for a 1st person personal pronoun in the absolutive
case, i.e it is the subject of the predicate.


jee-joka-yo-naje. (e-jee skin)
bathe-tmp.SS 1sg.abs skin-dry-tel-pst
After bathing, my skin dried completely (lit. I skin-dried). {joka.001}

3.4 Verbs attested in NI

NI into transitive verbs is the most basic type of incorporation cross-linguistically
(Mithun 1984:875) and it has been extensively exemplified above. In Ese Ejja however, NI occurs with both transitive and intransitive verbs. Example (22), a repetition of (18), shows how the verb kiyo- heat can incoporate the root -wi wick


(e-wi nose)
kerosene_lamp.abs nose-heat-short-pst
The kerosene lamp got its wick shortened by heating
(lit. the lamp nose-heat-shortened). {chia.004}

See also (21) with the intransitive verb joka- dry.

Incorporation into intransitive predicates, however, is not rare. According
to Aikhenvald (2012:195), it exists in most Amazonian languages.11 VelzquezCastillo (1996, 1999) also describes incorporation into descriptive predicates in
Guaran (Tupi-Guaran), a language spoken in Lowland South America, and Sapir
(1911:258; 270ff) mentions it in several North-American languages (in the Iroquoian languages, in Pawnee (Caddoan) and in Yana (north-central California)).

11. Cf. also Admiraal & Danielsen, this volume.

Two types of incorporation in Ese Ejja (Takanan)

The predicates involved are usually unaccusative, descriptive/stative predicates

like be sick, be tall, hurt, in which the subject is not agentive.
Posture verbs can also incorporate body-part terms. To my knowledge, this
has not been reported elsewhere.12 In (23)ab, the (body part) -me hand is incorporated into neki- stand and bae- hang.13

Incorporation into posture verb

a. (e)-me hand + neki stand

Iawewa e-pi
neki akwi=jo me-neki.
res-straight stand tree=loc hand-stand.prs
The dog is standing straight, with its paw on the tree
(lit. it hand-stands). {KaFWA.089}

b. (e)-me hand + bae float, hang

E-shoi oja e-naba=jo
npf-child 3.gen npf-mouth=loc hand-hang.prs
(Sitting leaning against a big tree), he has his hand (lit. he hand-hangs)
on his mouth. {KaFWA.137}

The possibility for posture verbs to incorporate nouns must be r e-contextualized

within the grammar of Ese Ejja: most linguistic phenomena related to posture verbs, that are found cross-linguistically, can be observed in this language
(Vuillermet 2009, 2012:603 ff/Chapter 14). Ese Ejja posture verbs:
are present in no less than four distinct constructions as main verbs (basic
locative, existential, possessive and copula constructions);
are grammaticalized as two separate tense and aspect markers (present and
may express the source in a periphrastic construction and undergo several
semantic extensions.
Note that the productivity of this subtype of NIV requires verification, as it is only
attested with hand and eye, i.e. only with two body-part terms. Furthermore, the
incorporation of this latter body part, koxa-neki watch (lit. eye-stand) may be
better analyzed as a case of lexicalization, as discussed in Vuillermet (2012:631).

12. Iroquoian languages do incorporate nouns into positional verbs, but it is mostly used
with immovable entities which cannot appear as independent nouns on their own. The
posture verbs then serve as a kind of classifier (Mithun, p.c., March 2013).
13. The verbs do not seem to be marked for tense, but this is a specificity of the posture verbs
in the present tense, due to the origin of the present markers: they are (grammaticalized)
posture verbs.

Marine Vuillermet

NI into the path verb dobi- go in is also attested in Ese Ejja. This possibility is
remarkable, as path verbs cannot be described as stative. However, examples like
these are rare. The phenomenon is attested in only three examples, with just one
of the five path verbs. In (24)ab, dobi- go in incorporates the two e-nouns -wi
nose and -jyoxi foot.14

Incorporation into path verbs

a. e-wi nose + dobi- go in

Iawewa botella=asixe wi-dobi-ki-yo-naje.

bottle=all nose-go_in-go_to_do-tel-pst

The dog put its nose (lit. nose-entered) into the bottle. {EpFWA.011}


(e)-jyoxi foot + dobi- go in

foot-go_in-go_to_do-pst tree=all
She got her foot stuck in the tree. {fieldnotes}

Note that an identical case is described for Warray (non-Pama-Nyungan,

Gunwinjguan) by Harvey (1995:1423), where a participant enters his hand into
a log. Future investigations should test if the other path verbs can also incorporate
nouns, and whether or not the incorporated nouns need to be body parts.
To summarize this section, NIV in Ese Ejja is typical of the Amazonian area
in several respects: the two lexical stems are tightly bound into a single phonological word and the incorporated noun precedes the verb stem, the most frequently
incorporated nouns refer to body-part terms, and intransitive verbs are good
incorporating candidates. However, Ese Ejja is less typical in that any possessed
nouns may be included in noun incorporation, as opposed to only inalienable
nouns. It is also atypical in permitting the unergative posture and path verbs to
incorporate nouns. The next section examines noun incorporation into adjectives.
4 Noun incorporation into adjectives (NIA)
NIA is rarely mentioned in the literature. Morphologically distinct constructions
with semantics comparable to the English derived adjectives red-haired and
blue-eyed are attested however.
The first subsection briefly presents the morphology of NIA (Section 4.1) while
the second one lists the expected and unexpected candidates for incorporation
into adjectives (Section 4.2). The third section explores the argument structure of
14. A third example displays the e-noun -to upper leg incorporated into the same path verb
dobi- go in. However, the predicate in that third example is uninflecting and presents specific
morphology which would require further comments. It is therefore not presented here.

Two types of incorporation in Ese Ejja (Takanan)

the incorporating predicates (Section 4.3), and the last section examines NIA and
functionnally comparable constructions cross-linguistically (Section 4.4).
4.1 Morphology of the NIA
NIA only occurs with kya-adjectives. As mentionned in Section 2.2c, the adjectives of this subclass of predicative adjectives require an affix to form independent
words, and this affix occurs obligatorily in NI, too. Example (25)ab illustrates
NIA with the neutral prefix kya- apf and the interrogative prefix ache- how.
a. Kya-bkwa-wiso-nee~nee taxakaka. (e-bakwa child)

apf-child-many-very~red frog

The frog has very many children (lit: the frog is very many
childrened). {EpFWA.062}
b. Ach-bakwa-wiso miya? (e-bakwa child)
how-child-many 2sg.abs

How many children do you have?{fieldnotes}

These adjectives form a single phonological word together with their incorporated
nouns, as shown by the occurrence of one stressed syllable per example (noted
with the acute stress).
Incorporation into adjectives is also available with the negative adjectival suffix -ama. Example (26)ab contrasts a sentence with NI into the adjective jyaama
not lovely (b) and its syntactic paraphrase (a).15 In this example, the presence or
absence of a nominal prefix on the e-noun -wanase wife is further proof of the
phonological boundedness of the two incorporated nouns and their adjectival host.

Ma kwiixi=ja e-wanase jya-ama.

this man=gen npf-wife pleasant-priv
This man does not like (his) wife. {volunteered (jya).008}
(lit. the wife is not pleasant to this man)

b. Ma kwiixi wanase-jya-ama.

this man wife-lovely-priv

This man does not like (his) wife.{volunteered (jya).009}
(lit. this man is not-lovely-wifed?)

Phonologically, the noun incorporated into an adjective forms a single word with
the adjective (and its morphemes). Example (27) contrasts noun incorporation
into the adjective kyapoji bald with its corresponding syntactic paraphrase.
In (27)a, ewo tail and kyapji bald contain one stressed syllable each, while
kyawopoji tail-bald in (27)b has only one stressed syllable in total.
15. The man in sentence (a) is encoded as a genitive and is the experiencer of the feeling
expressed by the adjective. The wife is the entity to which the feeling applies.

Marine Vuillermet



E-wo kya-pji d=ja.

npf-tail apf-bald red.howler=gen
The tail of the red howler (monkey sp.) is bald. {KoAni.137}


red.howler=also apf-tail-bald
The red howler also has a bald tail. {KoAni.138}

Also note the absence of the nominal prefix when the e-noun -wao tail is
4.2 What can be incorporated into adjectives
Just like in NIVs, e-nouns are the most frequently incorporated nouns, but any
possessed items can be incorporated. Example (28) exemplifies NIA with the
e-nouns -bakwa child (a) and -ami body (b) and with the independent nouns
daki clothes (c).


a. Kya-bakwa-wiso-nee~nee taxakaka. (e-bakwa child)

apf-child-many-very~red frog

The frog has very many children (lit: the frog is very many-
childrened). {EpFWA.062}


(e-(y)ami flesh)
its body (of a parrot) really stinks (lit: it is smelly-bodied). {KoAni.199}


kya-daki-chwe neki.
1sg.abs apf-clothes-wet stand.prs
(After I get wet with the rain), my clothes are wet. {chwe.005}

Loanwords can be incorporated like any other nouns. Examples (29)ac illustrate
this case with zapato shoe, chinela flip-flop and ameba amoeba.


a. Kya-zapato-jeyo.


(My) shoes are slippery. (lit: I am slippery-shoed). {volunteered
I have many amoebas. (lit: I am many-amoebaed) {fieldnotes}


c. Miya

2sg.abs apf-flip_flop-thin-very~red
Your flip-flops are very thin. (lit. you are very thin-flip-flopped)

Two types of incorporation in Ese Ejja (Takanan)

However, NIA also accepts non-possessed items: in (30), beni wind is incorporated into (kya)kemo big.

There is lots of wind. {elicited}

Such incorporation results in an impersonal construction, as there is no possessor to fill the vacated case role. This is discussed in the next subsection
(Section 4.3).
More surprisingly, lexemes incorporated into adjectives need not be simple nominal roots: they can be NPs already marked with an attributive adjective (31)a, NPs marked for plural (31)b, genitive phrases (31)cd, and even
pronouns (31)e.



(e-)noun + attributive (e-bakwa child)


Riberalta=loc apf-child-big-many

At Riberalta there are lots of big children. {volunteered}


noun + plural
There are lots of different sorts of snakes. {volunteered}

c. Genetive Phrase (1)

There are lots of old peoples stories/ ancient stories.

d. Genitive Phrase (2)
1sg.abs apf-1sg.gen=country-sad

I miss my country (lit: I am sad-my countried).

{volunteered (yeno).008}1617

e. pronoun

Oya iby kya-ese-sei.17
3.abs jaguar apf-1incl-fond_of/and_of

The jaguar, it is fond of us! (i.e. it likes human flesh) {KoAni.247}

16. Note that the copula subject would then be redundant here.
17. Note that the pronominal form ese- cannot appear in main clauses as an independent
form. It needs the -ya(dummy) focus marker.

Marine Vuillermet

The incorporation of such modified NPs (and pronouns) is infrequent in the language: most examples were volunteered during the workshop on adjectives, and
only the last example with pronouns comes from a spontaneous text. Note that the
NP in (31)c, etiikyana=ja-sowi the ancients stories, may be considered a lexical
unit, but the same does not hold for the other incorporated NPs. Such a situation
is highly unexpected for incorporation, as Mithun (1984:859) observed: incorporated nouns lose their syntactic status as arguments of the clause; and they are
unmarked for definiteness, number, or case. The variety of incorporated lexemes
represent an argument against the analysis as noun incorporation or a modification of the typology.
Finally, non-finite verb stems can also be incorporated into the kya-adjectives,
so that the term noun incorporation would be problematic. Being non-finite verb
stems, they are neither marked for tense nor indexation, i.e. they are lexical roots.
Examples (32)ad illustrate incorporation of a simple (verb) root (a, b), a compound and a derived stem (c, d).
(32) a.

Simple verb root (1)

kya-mimi-pame poxyama.
teeth-less=reason.SSS apf-talk-good neg
I dont talk well because I am teethless. {fieldnotes}


Simple verb root (2)

Ibya kya-besa-kwajipashi-nee~nee.
tiger apf-cross-quickly-very~red
Tigers cross (rivers) very quickly.
{volunteered (kwejipashi).005}

c. Compound verb stem (2)

Makwa e-wanase=a jya-ka-je

Sixto npf-wife=erg throw-3A-fut

Sixtos wife will leave him because he walks shakily.18 {volunteered


Derived verb stem

Dirigente=kwana kya-xa-shawa-ba-ki-tai
apf-mid-soul-see-mid-bad stand-prs
Leaders think (lit. they soul-see themselves) badly/do not
think right. {fieldnotes}

18. During the workshop, they liked making fun of the people, who were participating. This
statement is only a joke.

Two types of incorporation in Ese Ejja (Takanan)

Such cases are rare in my corpus, and should definitely be considered very atypical. Nevertheless, two occurrences (a and d) come from my fieldnotes and are
spontaneous. Note that only intransitive roots or stems are attested. Verb root
incorporation also exists in the related language Cavinea (Guillaume 2008:378
ff), and both intransitive and transitive verbs are candidates for the process.
The types of incorporated elements into adjectives differ from the ones that
can enter verbs to such an extent that the (noun) incorporation terminology is
questionable. Such differences have an impact on argument structure.
4.3 Argument structure
The argument structure of the prototypical NIA (involving possessed nouns) differs from the one of the atypical ones (involving unpossessed and modified nouns,
or verb stems). When possessed nouns are incorporated, the situation is similar to
the one described for NIV: the adjectival predicate remains monovalent, whether
incorporation takes place or not. When the incorporation of a possessed object takes
place, the possessor of its original copula subject becomes the syntactic subject. The
examples in (33) are taken from a narrative, where a blind speaker was talking about
the salient physical characteristics of several animals. The two sentences follow each
other in the text and the rephrasing is natural. They illustrate nicely how the original
genitive modifier (possessor) in (a) is advanced to subject status in (b).


possessed (abs)
possessor (gen)
a. E-mekishe kya-ao-nee~nee

npf-nail apf-tall-very~red 3.gen

The nails of it (the porcupine) are long. {KoAni.123}

incorporated possessed possessor (abs)
b. Kya-mekishe-ao-nee~nee oya.

It has very long nails (lit. it is very long-nailed). {KoAni.122}
See (27)ab for a similar spontaneous pair.

The subject status of the new subject is proven by the possibility of having the
same-subject subordinate marker -maxe tmp.SS in the temporal subordinate
clause in (34). The argument that is wet is the person whose clothes are wet (see
Vuillermet to appear).


kya-daki-chwe neki.
water=loc be_wet-tmp.SS 1sg.abs apf-clothes-wet stand.prs
After Ii get wet with the rain, myi clothes are wet. (lit. I stand wet-clothed)
{volunteered (chwe).005}19

19. Indices are used here for convenience and comparability with the next example.

Marine Vuillermet

This co-reference marker is not available in situations where the incorporated

noun is not possessed, as will be shown below.
As for non-prototypical incorporated nouns, the situation is different. When
non-possessed nouns are incorporated, there is obviously no possessor to advance
to subject status. The resulting sentence is thus subjectless, as the presence of the
different-subject subordinate marker -jojo DS shows in (35).


[kya-meshi-jeyo=jojo] ()
1sg.abs apf-earth-slippery=reason.DS fall-pst
II fell [because thej/*myi ground (was) slippery/??I was slipperygrounded]. {volunteered (pii).002}

Example (35) with the different-subject marker contrasts with the same-subject
marker in (34), where the incorporated noun has a possessor encoded as a subject,
which is then co-referencial with the subject of the main clause.
When verbs are incorporated, the notional subject of the incorporated stem
isthe copula subject, and there is no case role manipulation. Example (36) contrasts the verb stem incorporated into an adjective (in (36)a, a repetition of (32)d)
with the corresponding syntactic structure, when another speaker commented
on the first sentence at my request. In both sentences, dirigentekwana the leaders
is the subject.


a. Dirigente=kwana kya-xa-shawa-ba-ki-tai

apf-mid-soul-see-mid-bad stand-prs
Leaders think (lit. they soul-see themselves) badly/do not think right.

Dirigente=kwana kya-tai bexo=jo=nei

apf-bad money=loc=very

xa-shawa-ba-ki-ani, bexo=jo=tii, bexo

mid-soul-see-mid-prs money=loc=ints money
Leaders think badly, about money only, only about money, only about
(how) to make money. {fieldnotes}

In Cavinea, where transitive verbs can also be incorporated (Guillaume 2008:

378 ff), the notional subject of a transitive verb is the copula subject of the adjective.
4.4 NIA cross-linguistically
NIA is described for a few languages from the Amazon and the Australian
Northern Territory, as well as for Sanskrit. Unfortunately, most descriptions do

Two types of incorporation in Ese Ejja (Takanan)

not mention how the newly derived adjective is used in context. Therefore, the
function of these incorporations is difficult to assess.
Ese Ejjas sister language Cavinea displays a similar process of NIA,
described by Guillaume (2008:375 ff) as compounded nouns and verbs. In
noun-adjective compounds, only e-nouns can be incorporated. In verb-adjective
compounds, the verb cannot have its affixes (even though reduplication is
attested). Incorporation of any possessed noun or genitive phrases as in Ese
Ejja is not reported.
Michael (2008:331) briefly describes noun incorporation in adjectives in
Nanti (an Arawakan language of Southern Peru). The phenomenon in Nanti
mostly appears with deverbal adjectives (i.e. the derivation into an adjective could
be analyzed as following the incorporation into the original verb root, (Michael,
p.c., February 2013; see also Admiraal & Danielsen, this volume). Instances of
incorporation into true adjectives (i.e. the small class of non-derived adjectives) are limited (2008:312). The incorporating adjective illustrating the authors
description is kite-gonake-ri (be.yellow-elbow-adjectivizer) yellow-elbowed. It
can modify a noun like a crab species, but is not used as a predicate. Note that
incorporation also occurs into numerals.
Interestingly, Fleck (2006) and Zariquiey and Fleck (2012) describe body part
prefixes that attach to nouns, adjectives and verbs for two Panoan languages, Matses and Kashibo-Kakataibo, respectively. The prefixes often have a corresponding
body part term, but not always. Fleck (2006) refutes an analysis of these prefixes
in terms of NI. The role of the prefixes is roughly to give a locative meaning to the
lexeme they attach to. Adjectives involved in prefixation tend to refer to colour
and physical properties, even though there are no cases of unproductively prefixed adjective roots in Kashibo-Kakataibo (Fleck 2006:395). Prefixed adjectives
result is adjectives like in-piu red-tailed (in Matses) or chi-ushin red-butted (in
Kashibo-Kakataibo). The two authors suggest two possible diachronical paths, one
of which is an evolution from a (synchronically not productive) noun incorporation system.20
At least three Aboriginal Australian languages exhibit types of NIA, referred
to by the authors as both lexical compounding and incorporation into adjectives.
All three languages are non-Pama-Nyungan languages of the Northern Territory

20. Whitney (1950:481) calls secondary adjective compounds a type of compounding, the
value of which is not given by a simple resolution into their component parts, but which,
though having as final member a noun, are themselves adjectives () and conveys the idea
of possessing. However, there are no further examples in the grammar.

Marine Vuillermet

of Australia. Murrinh-Patha displays a type of NIA very similar to that described

for Ese Ejja, though only body-part terms can be incorporated (Walsh 1995:342ff).
A striking similarity between the system found in Ese Ejja and in Murrinh-Patha
is that the non-incorporating equivalent corresponds to teasing and cursing (see
the Ese Ejja noun compounds used for teasing and insults in 2.2.c), while incorporation constructions express statements of facts, which allow one to focus on
purported physical deformities (Walsh 1995:346) (see below the excerpts listing
the physical attributes of the animals).
Warray (non-Pama-Nyungan, Gunwinjguan) also has compounds involving
body parts (Harvey 1995:122ff). The author states that the compounds are either
nouns or adjectives, depending on the incorporating root. The adjective examples
given are out of context, and the translation suggests that they are referential NPs.
The examples include adumjaminyu a man with bad eyes (lit. the bad-eyed one),
abamgubamu brown snake (lit. the big-headed one), angarndibitu Eucalyptus
miniata (lit. the colour-throated one), (a)ganimbaliwu frill-necked lizard (lit.
the big-eared one) and dumdingdingu fish species (lit. the sticky-eyed one).
Velsquez-Castillo (1999:124) mentions that adjectival synechdochic compounds
describing natural species tend to be reanalyzed as nouns.
In Anindilyakwa, another Gunwinjguan language, Leeding (1995) describes
a similar phenomenon. Morphologically, these nominal compoundings are either
nouns or adjective (depending on the nature of the incorporating lexeme). Nevertheless, the author highlights that semantically, they are regarded as names of a
particular item unless otherwise identified by another referent (Leeding 1995:234).
This phenomenon is distinct from the NIA described for Ese Ejja: NIA is highly
productive and transparent, and it involves predicative adjectives.
Noun incorporation into adjectives also exists in Sanskrit under the term
Bahuvrhi. Note that this terminology is often used to describe nouns, though the
(initial) phenomenon in Sanskrit refers to compounds that are always adjectives
(Gonda 1966:83). The final member of the compound is always a noun. Also in
this elementary grammar of the Sanskrit language, the author does not give the
context in which the incorporating adjectives appear. However, the author qualifies the bahuvrhi compounds as possessive compounds, which probably suggests
the same semantics of possessive relations as in Ese Ejja.
Several languages display productive morphology to express similar concepts. English derived adjectives such as blue-eyed or red-haired are essentially
productive with colours associated to body parts, but other types of derived
adjectives also occur (see Tsunoda 1995:614ff) citing examples from Hirtle (1970
and H
udson 1975, like a rough-faced youth, a hard-featured girl, a white-coated
attendant, a white-hatted cabman, etc.).

Two types of incorporation in Ese Ejja (Takanan)

NIA is thus reported for several languages. However, a closer look at the data
shows that none seems to be as clearcut and as productive as the incorporation
into adjectives described for Ese Ejja. Interestingly, several languages use derivation to reach similar functional results.

5. Function (s) of the two NI types

5.1 Manipulating case roles and much more
As mentioned earlier, when a possessed entity is incorporated in Ese Ejja, the
notional oblique possessor can fill the role vacated by this incorporated entity.
This phenomenon has been carefully described by Mithun (1984) as type II in her
hierarchy, the manipulation of case role. In Ese Ejja, the only obliques that can fill
the vacated role are possessors. It is largely attested in other languages and may
also be referred to as a phenomenon of external possession.
It is interesting to examine the expression of possession in Ese Ejja. The
language has no possession verbs such as have or belong. A construction
with posture verbs is one of the strategies to express possession. The possessor
appears in the dative/genitive, and the literal translation of a spontaneous sentence expressing the possession of children would be the children stand to me/
mine. Another strategy is to derive a predicative adjective from a noun with
a proprietive marker. I have a child can also be said with bakwa-xi, lit. I am
However, I am certainly not claiming that the main function of NI in Ese Ejja
is to express possession. Here, I follow the refutation of the following hypothesis
by Velzquez-Castillo (1999), who gives a cognitive/functional analysis of incorporation in Guaran (and external possession in Spanish):
Entities with lesser saliency [i.e. the possessum here] are accessed mentally
via their reference points [i.e. the possessor here] (). In a possessive relation, for
example, there is always an asymmetric relation between the possessor and the
possessum, where the possessor is conceptually more prominent and serves as the
reference point for the possessum. (Velzquez-Castillo 1999:80).
In Ese Ejja NI is mostly motivated by discursive factors. Velzquez-Castillo
also emphasizes the necessity to keep up discursively with the same topic, i.e. the
highly topical possessor. Example (37) is an excerpt from the Frog story (Mayer
1969), where the child finally discovers the frog by the pond, with its big family.
The consultant first asserts the possession with the proprietive marker bakwa-xi
child-propr, and then comments on the fact that (it does not only have children,

Marine Vuillermet

but) it has many children. The frog remains the topic in both sentences in (37), as
it was already in the three preceding sentences.


no, kya-bakwa-wiso-nee~nee
and child-propr no apf-child-many-very~red
And (the frog) has children, hasnt it, it has very many children.

Finally, Velzquez-Castillo also emphasizes the subjective construal embodied

by external possession constructions. This function of incorporation is especially
evident with adjectives, where several pairs of periphrastic vs. incorporating
sentences follow one another. Examples (38) and (39) illustrate this subjectivity: in a narrative about the habits of animals and their physical properties, the
(blind!) speaker reviews the specific attributes of several animals. The choice of
alternative expressions (incorporation or not) is a matter of discourse choice of
topic, whether pointing to a specific body part (non-incorporating predicates,
underlined in the text and in the translation) or emphasizing that the body part
mentioned is an attribute of that particular animal (incorporating constuctions
in bold).



Ojaya e-mekishe kya-ao-nee~nee,

3.gen npf-nail apf-long-very~red
(preceding context: and sloths go up (the trees) just like
porcupines go up.) Their nails are very long, {KoAni.121}

b. kya-mekishe-ao-nee~nee oya,

apf-nail-long-very~red 3.abs
they have very long nails, {KoAni.122}


e-mekishe kya-ao-nee~nee
ojaya jikyo,
npf-nail apf-long-very~red 3.gen dem
their nails, this (pointing gesture), theirs are very long, {KoAni.123}


apf-long npf-nail
their nails are long, {KoAni.124}

e. kya-mekishe-ao.

they have long nails. {KoAni.125}


apf-nose-thin&long-very~red 3.abs
(preceding context: They also have a tail; their nose is small, very small).
They have a very small nose. {KoAni.129}(following context: their nose is
thin; as for their tail, it is big).

Two types of incorporation in Ese Ejja (Takanan)

Having established the habits of an animal, the consultant focuses on physical

specificities of three body parts of this animal: its nails, its nose and its tail. The
three body parts (underlined in the examples and translations) represent what is
focused on; each body part is introduced as an independent noun, which is copula
subject of the predicative adjective. In (38)c, a (pointing) demonstrative is used to
clear any ambiguity about the topic. These independent instances of the body parts
contrast with the incorporated ones, when the speaker focus again on her main
line, the sloth she was first describing.
5.2 Modifying NPs
As previously highlighted, the rarity of attributive adjective contrasts with the
importance of the predicative class. In order to modify adjectives, the productive
strategy of the language is to use incorporation into adjectives rather than have
independent nouns modified by an attributive adjective. Example (37) (repeated
below (40) about the many children the frog had) does not only demonstrate that
possession is primarily expressed with the proprietive markers, but it also shows
how the speaker has to use another construction to qualify the possessed entity.


no, kya-bakwa-wiso-nee~nee
and child-with no apf-child-many-very~red
And (the frog) has children, hasnt it, it has very many children.

In English for instance, the same construction can be used with or without a
The importance of this function of incorporation into adjectives is reflected
in the possibility for non-possessed items to also be candidates for incorporation,
subsequently yielding subjectless existential constructions such as there is much
wind, as in (30), or there are slippery trails, as in (35).
6. Concluding remarks
The two types of incorporation that are found in Ese Ejja are similar in some
ways but are also distinct from one another at all levels. Morphologically, the
two incorporation types can be described as a general process where, paraphrasing Mithun (1984), a lexical stem is compounded with a predicative stem
to yield a larger, derived predicative stem. The main differences lie in the fact
that the two processes involve a different word class with regard to the incorporating lexemes but also with regard to what can be exceptionally incorporated
into adjectives.

Marine Vuillermet

Syntactically, both types of incorporation manipulate roles and allow a

notional oblique possessor to be encoded as a core argument of the predicate.
Nevertheless, this case role manipulation is systematic with NIVs, while it is not so
with NIAs, where it can leave the subject of the predicate undetermined, yielding
an impersonal construction. This can only happen in situations where the incorporated entity is not possessed.
Functionally, the two sorts of incorporation may be used to background the
body part or a possessed item and foreground its possessor, though NIA may only
background the incorporated item without the concomittant foregrounding of
another entity.
This paper may encourage other linguists to describe similar phenomena with
adjectives from a functional-typological perspective, so as to refine the role of such
devices within the languages they are found in.

1 = first person; 2 = second person; 3 = third person; A = agent-like argument
of a prototypical transitive verb; abs = absolutive; all = allative; apf = adjective prefix; caus = causative; dem1 = demonstrative 1; dem2 = demonstrative 2;
depr = depreciative; deg = degree; ds = different subject; erg = ergative; excl =
exclusive; fut = future; gen = genitive; incl = inclusive; indf = indefinite; int =
interrogative particle; ints = intensifier; ipfv = imperfective; loc = locative; mid
= middle; npf = noun prefix; P = patient-like argument of a prototypical transitive
verb; pst = past; pl = plural; priv = privative; propr = proprietive; prs = present;
q = question word; red = reduplication; res = resultative; S = single argument of
an intransitive verb; sg = singular; SS = same subject; SSS = same subject (single to
single) co-reference; tel = telic; tmp = temporal

Aikhenvald, Alexandra Y. 2012. Languages of the Amazon. Oxford: OUP. DOI: 10.1093/acprof:
Baker, Mark. 1999. External possession in Mohawk. In Doris L. Payne & Immanuel Barshi (eds),
Chappell, Hilary & McGregor, William. 1995. The Grammar of Inalienability: A Typological Perspective on Body Part Terms and the Part-whole Relation. Berlin: Mouton de Gruyter.
Couchili, Tiiwan, Maurel, Didier & Queixals, Francesc. 2002. Classes de lexmes en mrillon.
Amerindia 2627:173208.

Two types of incorporation in Ese Ejja (Takanan)

Dixon, Robert M.W. 1982. Where Have All the Adjectives Gone ? And Other Essays in Semantics
and Syntax, 162.Berlin: Mouton. DOI: 10.1515/9783110822939.1
Dixon, Robert M.W. & Aikhenvald, Alexandra Y. 1999. Introduction. In The Amazonian
Languages, Robert M.W. Dixon & Alexandra Y. Aikhenvald (eds), 122. Cambridge: CUP.
Emkow, Carola. 2006. A Grammar of Araona, an Amazonian Language of Northwestern Bolivia.
PhD dissertation, La Trobe University.
Fleck, David W. 2006. Bodypart prefixes in matses: Derivation or noun incorporation? International Journal of American Linguistics 72(1): 5996. DOI: 10.1086/505279
Girard, Victor. 1971. Proto-Takanan Phonology. Berkeley CA: University of California Press.
Gonda, Jan. 1966. A Concise Elementary Grammar of the Sanskrit Language. Translated by
Gordon B. Ford Jr., 2nd edn. Tuscaloosa AL: University Alabama Press.
Guillaume, Antoine. Forthcoming. Sistemas complejos de movimiento asociado en las lenguas
tacana y pano: Perspectivas descriptiva, comparativa y tipolgica. In Antoine Guillaume &
Pilar M. Valenzuela (eds).
Guillaume, Antoine. 2008. A Grammar of Cavinea. Berlin: Mouton de Gruyter. DOI:
Guillaume, Antoine. 2012. Maropa (reyesano). In Lenguas de Bolivia, Vol. 2: Amazona, Mily
Crevels & Pieter Muysken (eds), 191229. La Paz: Plural Editores.
Guillaume, Antoine & Valenzuela, Pilar M. (eds). Forthcoming. Estudios sincrnicos y diacrnicos sobre lenguas pano y takana: Fonologa, morfologa y sintaxis. Amerindia.
Harvey, Mark. 1995. Body parts in Warray. In Hilary Chappell & William McGregor (eds),
Leeding, Velma J. 1995. Body parts and possession in Anindilyakwa. In Hilary Chappell &
William McGregor, 193249.
Mayer, Mercer. 1969. Frog, Where Are You? New York NY: Dial Books for Young Readers.
Michael, Lev. 2006. Noun incorporation and verbal classifiers in Nanti (Kampa, Arawak). In
Proceedings of the Second Conference on the Indigenous Languages of Latin America. http://, 127 (2 February 2014).
Michael, Lev. 2008. Nanti Evidential Practice: Language, Knowledge, and Social Action in an
Amazonian Society. PhD dissertation, University of Texas at Austin.
Mithun, Marianne. 1984. The evolution of noun incorporation. Language 60(4): 847894. DOI:
Mithun, Marianne. 1986. On the nature on noun incorporation. Language 62(1): 3237. DOI:
Payne, Doris L. & Barshi, Immanuel. 1999. External possession: What, where, how, and why.
In External Possession [Typological Studies in Language 39], Doris L.Payne & Immanuel
Barshi (eds), 329. Amsterdam: John Benjamins. DOI: 10.1075/tsl.39.03pay
Peluso, Daniela. 2003. Ese Eja Epona: Womans Social Power in Multiple and Hybrid Worlds.
PhD dissertation, Columbia University.
Peluso, Daniela. 2004. Variabilidad y cambio en los nombres personales en una sociedad indgena amaznica. Amazona Peruana 2829:103123.
Rose, Franoise. 2008. L incorporation nominale en merillon: Une approche lexicale et discursive. Amerindia 31:87112.
Sadock, Jerrold M. 1986. Some notes on noun incorporation. Language 62(1): 1931. DOI:

Marine Vuillermet
Sapir, Edward. 1911. The problem of noun incorporation in American languages. American
Anthropologist 13:250282. DOI: 10.1525/aa.1911.13.2.02a00060
Tsunoda, Tasaku. 1995. The possession cline in Japanese and other languages. In Hilary Chappell
& William McGregor, 565630.
Velzquez-Castillo, Maura. 1996. The Grammar of Possession: Inalienability, Incorporation and
Possessor Ascension in Guaran [Studies in Language Companion Series 33]. Amsterdam:
John Benjamins. DOI: 10.1075/slcs.33
Velzquez-Castillo, Maura. 1999. Body-part EP constructions: A cognitive-functional analysis.
In External Possession [Typological Studies in Language 39],Doris L. Payne & Immanuel
Barshi (eds), 77109.Amsterdam: John Benjamins. DOI: 10.1075/tsl.39.08vel
Vuillermet, Marine. 2009. Los verbos de postura Ese Ejja no se quedan inmviles. Y a veces se
desvan. In Conference on Indigenous Languages of Latin America-IV. Austin TX. http:// (2 February 2014).
Vuillermet, Marine. 2012. A Grammar of Ese Ejja, a Takanan Language of the Bolivian Amazon.
PhD dissertation, Universit Lumire Lyon 2.
Vuillermet, Marine. Forthcoming. Verb compounds in Ese Ejja. In Guillaume & Valenzuela
Vuillermet, Marine. 2014. The multiple co-reference systems in the Ese Ejja subordinate clauses.
In Information Structure and Reference Tracking in Complex Sentences, R
ik van Gijn,
Jeremy Hammond, Dejan Mati, Saskia van Putten & Ana Vilacy Galucio (eds), 341372.
Amsterdam: John Benjamins.
Walsh, Michael. 1995. Body parts in Murrinh-Patha: Incorporation, grammar and metaphor. In
Hilary Chappell & William McGregor, 327380.
Whitney, William Dwight. 1950. Sanskrit Grammar. Cambridge: Harvard University Press.
Wyma, Richard & Pitkin de Wyma, Lucille. 1962. Eseejja y Castellano. Vocabularios Bolivianos, No. 3.Cochabamba, Bolivia: Instituto Lingstico de Verano, Ministerio de Asuntos
Campesinos, Ministerio de Educacin y Bellas Artes Oficila Mayor de Cultura.
Zariquiey Biondi, Roberto & Fleck, David W. 2012. Body-part prefixation in K
synchronic or diachronic derivation? International Journal of American Linguistics 78(3):
385409. DOI: 10.1086/665918

Reduplication in Yurakar
Rik van Gijn

University of Zrich
Yurakar, an isolate language spoken in central Bolivia, makes extensive use of
reduplication to form words. Three different types of morphological reduplication
can be distinguished on formal grounds: complete root reduplication, partial
prefixed reduplication, and partial suffixed reduplication, each associated with
different functions. I describe each of the three types, and discuss their formal
and functional characteristics. I furthermore connect the three reduplication
types to the Iconicity Principle, showing that each reduplication type represents a
different way of applying the Iconicity Principle that more form represents more
meaning, thus maximally exploiting the iconic connection between form and
meaning and on the other hand maintaining interpretative transparency.
Keywords: Yurakar; reduplication; verbal morphology; Iconicity principle

1. Introduction
Reduplication is an important word formation process in Yurakar, an isolate
language spoken in central Bolivia. Three basic types of reduplication can be
discerned on formal grounds: complete root reduplication, suffixed partial reduplication, and prefixed partial reduplication.
In complete root reduplication, the entire (mostly nominal) root is copied. The
resulting form is interpreted as a descriptive adjective.1 Although this type is best
considered complete reduplication, in some cases phonological processes change
the root forms or the resulting form (this would count as instances of complex

* First and foremost, I thank the Yurakar people for patiently teaching me their language
in all its richness. Furthermore, various funding agencies have contributed to the collection
and processing of data that has made this article possible: the support of NWO, KNAW, Volkswagen Stiftung, and SNSF is greatly appreciated. Finally, I thank the editors for their useful
suggestions for improvement of the paper, all remaining errors are mine.
1. The following word classes are distinguished in Yurakar: nouns, verbs, adjectives, adverbs,
ideophones, and interjections.

Rik van Gijn

reduplication in terms of Rubino 2005). For instance, both degemination (both

examples) and gemination (1)b occur.



patpata thorny


tbttb red

The function of this reduplication type can at least diachronically be connected to

one of the iconic functions of reduplication, distributivity, given that the resulting
property words refer to distributed instances of the referent of the n
form over a surface or space.
Suffixed partial reduplication copies the onset and nucleus of the final syllable
of the root. This type of reduplication has two possible functions: pluractionality
or causation. The pluractional reading occurs almost always in combination with
the perfective prefix i-.



i-bobo~bo hit repetitively


i-bb~b search in various places

Causative reduplication is cross-linguistically rarely associated with reduplication, and is diachronically probably posterior to the more iconic pluractionality
In Yurakar this reduplication effect is restricted to ideophonic roots, or roots that
have an ideophonic origin.



go in
yupa~pa cause to go in (= put in)



boil (intr)
cause to boil

Prefixed partial reduplication, with an added emphatic [h] sound, spelled j, indicates intensity, and is applied mostly to property words, but also to some action



shuj~shudyul very beautiful




very big

go (singular)
go really far, for a long time (singular)

Reduplication in Yurakar

The formal properties of this type of reduplication are not always predictable, as it
sometimes copies only the onset and nucleus of the first syllable, and sometimes
of the first two syllables.

adyaj~adyajta very fast

Intensity reduplication can apparently apply after inflection. In (6) the base for
reduplication includes the object agreement marker m-.


What really hurts you?

In this paper, I describe the formal and functional regularities and irregularities of
the three types of reduplication in Yurakar, and discuss them from the perspective of the Iconicity Principle more of form is more of content (Lakoff & Johnson
2003:127). In Section 2, I start by giving some background information about the
accent system of the language, which plays an important role in some of the reduplicative patterns. Section 3 describes the three reduplication patterns, focusing
on their formal and semantic properties. In Section 4, I discuss the Iconicity Principle, and how this plays out in the three reduplication types of Yurakar. Section
5 summarizes the major points of the paper.
2. The accent system of Yurakar
The accent system of Yurakar is complex in the sense that there are many layers
involved. In what follows I do not give a complete account of the system (which
involves intonation and context-dependent pragmatic accent), but rather only the
lexical layer of the accentuation system.
In the normal case, starting from the left edge of a word, every second syllable is accented; the rightmost of these stressed syllables gets primary accent. Put
differently: a regular word in Yurakar is built up out of binary feet with a right
dominant node (iambic). Primary accent falls on the rightmost dominant node
of a word. The word-final syllable is extrametrical, and normally2 not accented,
even if a strict application of the rule given above would lead to an accented final
syllable. In the following examples, primary accent is indicated with an acute diacritic, secondary accent with a grave diacritic extrametricality is indicated with
fish hooks, and syllable boundaries with a period between syllables.
2. It can be accented for pragmatic reasons; this is not discussed here, as it has no bearing
on the topic of the paper.

Rik van Gijn


po.j. re
my canoe re my future canoe

The examples in (7)show that three-syllable words have a primary accent on the
second (penultimate) syllable, according to the iambic, left-to-right pattern. Foursyllable words also have their primary accent on the second (in this case antepenultimate) syllable, because accentuation of the final syllable is blocked due to its
extrametricality. Five-syllable words, finally, have a secondary accent on the second syllable, and a primary accent on the fourth (penultimate) syllable, according
to the iambic pattern.
Two major interfering factors in this basic accentuation pattern are syllabic
weight and what can be called accent-attracting morphemes or syllables. To
start with the first, it is a common characteristic of languages with a basic iambic
parameter that they are sensitive to syllabic weight (see e.g. Van de Vijver 1998:2).
In Yurakar, trisyllabic words that start with a heavy (closed) syllable normally
have their accent on the first (penultimate) syllable.



py.da. ra
kk.ku. l
yn.ka. la ta
shm.pi. ru

mythological jaguar
wherever; whatever
cultivated plant; plant seed

The general analysis for this phenomenon is to say that the iambic pattern functions on the basis of moras, heavy syllables being bimoraic.
Apart from this sensitivity to syllable weight, certain prefixes seem to attract
stress, thereby overruling the general pattern outlined above. As an example, consider the noun sibb.3



ti.s.b my house our house

The form tisb in (9)a carries the prefix ti- 1sg.4 Accentuation takes place according to the normal iambic pattern, with the accent falling on the penultimate

3. Like many disyllabic nouns, sibb has a geminate middle consonant when it does not carry
any affixes. In this way, the iambic pattern is protected, because the first syllable of the root
has two moras, and therefore offers a template for an iambic foot, even if the final syllable is
extrametrical. I consider the non-geminate form (in this case sib) to be the underlying form.
4. Most personal prefixes have identical forms whether they mark possession or (different
kinds of) objects. I gloss these markers with their person and number values only, and do not
include their relational function.

Reduplication in Yurakar

s yllable. The form in (9)b, however, has a deviant accentuation pattern, with the
accent falling on the antepenultimate syllable, even though the trisyllabic structure is expected to yield accentuation of the penultimate syllable. This deviation
is due to the prefix ta- our, which behaves like a heavy syllable in that it attracts
accentuation, thereby overruling the basic iambic rhythm. The prefix ta- is part of
a small group of diacritically marked or heavy prefixes, that attract accentuation
when applied to roots, which are given in Table 1.
Table 1. Accent-attracting prefixes



3rd person singular


1st person plural


2nd person plural


3rd person plural


delimiter, location




1st person singular voluntary comitative


2nd person singular voluntary comitative


3rd person singular voluntary comitative


1st person plural voluntary comitative


2nd person plural voluntary comitative


3rd person plural voluntary comitative

The morphemes in Table 1 are meaning-bearing units, but syllables that form
part of roots can also be accent-attracting.
There are more complications to the accentuation system involving word
groups, but a description of that goes beyond the scope of the present paper.
3. Three reduplication patterns
As mentioned in the introduction, there are three major reduplication patterns
in Yurakar: complete reduplication, partial suffixed reduplication, and partial
prefixed reduplication. I discuss each of these in turn, outlining their formal and
semantic characteristics.
3.1. Complete reduplication
Complete reduplication derives descriptive property-denoting words, often with a
distributive meaning, X all over.

Rik van Gijn




full of holes
having many twigs

(lit.: hole~hole)
(lit.: fur~fur)
(lit.: bald~bald)
(lit.: arm~arm)

Formally, the basic form of this type of reduplication is the simplest one in that the
original root and its copy together form a new word, pronounced under a single
lexical accent contour. The forms in (10)a and (10)b have secondary accent on
the first syllable and primary accent on the penultimate syllable, because of the
influence of syllable weight. The forms in (10)c and (10)d do not have any heavy
syllables, so they follow the normal iambic pattern with primary accent on the
antepenultimate syllable (given that the final syllable is extrametrical).
Although the basic pattern for this type of reduplication is to repeat the root,
some forms deviate slightly from this pattern, as transpired from some of the
examples given above. Some forms geminate the first consonant of the rightmost
formant, and there is a minor pattern of reduplications which have an epenthetic
/n/ with allophone [m] is inserted (12) .



(tbb blood)6
numannuma dawn, daybreak (numma night)
samassama watery
(samma water)



bolembole low, short (bole = ?)

darandara thin
(dara = ?)

The addition of phonetic material may be due to the fact that heavy syllables facilitate accentuation (in all these cases the accented syllable is made heavy by adding
a coda consonant to it).
The most common pattern is the reduplication of a nominal root, but
adjective roots and ideophonic roots can also be reduplicated. Moreover, many
property-denoting words have lexicalized reduplication patterns in which the
parts of the reduplication do not have any meaning synchronically.




low, short

5. As was mentioned above, many disyllabic noun roots have a geminated consonant in their
citation form, which is, however not part of their underlying root form.

Reduplication in Yurakar

In terms of semantics, the basic and most straightforward effect of complete reduplication is distribution over space, the reduplication in that case is an iconic (and
cross-linguistically common see e.g. Rubino 2005:1920) way to encode multiple instantiations.



patapata thorny
(patta thorn)
meremere ribbed
(merre ridge)
segmented (tilli knot, joint)

However, more abstract relations between the root and the reduplicative derivation also exist, like similative, e.g. tbttb blood-like (see tbb blood) in (11)a,
metonymical association, as in uliwliw naked (see ulliw penis), temporal, e.g.
numannuma dawn, daybreak in (11)b (see numma night).
A number of colour terms seem to have been derived from a root by means
of complete reduplication, sometimes additionally carrying the similative marker
-shi. In a few cases, the non-reduplicated source is still used synchronically, in others not anymore. I briefly come back to colour terms in Section 4 below .



worew (o)reshi

purple/dark blue (ynn genipa)

red (tbb blood)
grey, dim

As an extension of the descriptive function, there are also many animal, plant and
tree names that have a reduplicated form of this type, often but not always referring to the sound they make.7




capybara (rodent sp.)

russet-backed oropendala (bird sp.)
kapok (tree sp.)
tropical screech owl

Summarizing, complete reduplication in its most basic or simple form is the

repetition of the entire root, and the entire word is pronounced under a single
lexical accentuation contour. Semantically, the most iconic function is to mark

6. Formally, this could also be an instance of intensity reduplication. It is analysed as an

instance of complete reduplication, however, because of the analogy with other colour terms
and because of the fact that the prefixed partial reduplication of disyllabic roots is usually
7. Identifications are taken from Hirtzel & van Gijn (in prep).

Rik van Gijn

distribution of referents or qualities over space. Some forms are lexicalized and
are semantically no longer transparent.
3.2. Suffixed partial reduplication
Suffixed partial reduplication marks either pluractionality or causation. In this
type of reduplication, it is normally the onset and nucleus of the initial syllable of
the root that is reduplicated. Reduplication of the final syllable also occurs marginally (only with the causative function). When the reduplication marks verbal distributive, or pluractionality, it mostly combines with the prefix i- perfective, and
the reduplicated suffix itself is in complementary distribution with the suffix -uma,
distributive. Whether pluractionality can be marked by means of reduplication or
not is lexically determined, and not very widespread.






send repeatedly
search everywhere
hit repeatedly
see repeatedly

Causation marked by reduplication stands in contrast with the middle marker-tA.9

There may have been a consistent semantic difference between the two derivations, but synchronically this is no longer clear.




hide (intr)
grow up
untie (intr)
shattered, torn


hide (tr)
untie (tr)
shatter, tear (tr)

In a few cases, reduplication with a pluractional meaning also contrasts with the
middle marker-ta. These instances are also often formally deviant.






kick repeatedly
move repeatedly

8. It is not clear why the vowel of the reduplicated suffix here is /e/ rather than /a/. Reduplicated forms are generally transparent in Yurakar, but there are some exceptions. Bjtub also
has a deviant form, as the root is bjta see.
9. In some circumstances, the vowel of this suffix adapts to the preceding vowel.

Reduplication in Yurakar

In the case of yupa/yupata go in sg/pl, the situation is somewhat more complex.



go in.caus

These instances of lexicalization, in combination with the formal overlap, suggest

a common origin for the causative and pluractional reduplication. Whereas pluractionality is readily and iconically connected to repetition, causation is much less
commonly marked by reduplication. I come back to this issue in Section 4.
3.3. Prefixed partial reduplication
The probably most productive reduplication strategy is the intensity prefix. Semantically, this type of reduplication is regular when it applies to property words, but
it may have slightly different effects when applied to stems denoting actions and
processes, see the examples in (21). The latter effects can be seen in (22)b, (22)c,
and (22)d, where process-verbs that have an inherent endpoint are shown with
the intensification prefix, and where the effect is that the process is very close to
its culmination point. For actions without an inherent endpoint, like mala go in
(21)d, the effect of the prefixed reduplication is that the action extends continuously over time.




really small (uujul small)

at the very end (limeye last)
really strong (binta strong)
go really far (mala go singular subject)

The reduplicated prefixes in (21) copy the first or the first two syllables of the
stem they attach to, followed by j, which should perhaps be interpreted as voiceless lengthening of the vowel. If the syllable is closed, only the onset (if present)
and the nucleus of first syllable are copied, and the coda consonant is not copied,
as can be seen in the case of bijbinta in (21)b. In most other cases the first two syllables are copied with the same coda-deletion rule for the second syllable of the
reduplication as for the single-syllable reduplication.



what-entity ints~2sg-mal-frighten-hab-3 mother
What really frightens you, mother?


They really would have died.

Rik van Gijn

c. shwj~shw-ishta-=ti

When it was really getting dark
d. malaj~ma-la-wita-=ya

They really had the urge [to urinate]: i.e.: [the urine] really arrived

with them.
e. meyej~meyeye-=w

They were really disobedient.

Although there are a few exceptions, the rule that seems to be able to explain
almost all cases is that the first potential foot is copied.
The accent system, as was described in Section 2, is built on iambic feet, with
a sensitivity to syllable weight and certain diacritically marked syllables and with
an extrametrical final syllable. In practice this means for the prefixed reduplication rule that the first two syllables are copied, unless the first syllable is either
heavy or diacritically marked, or if the base is a disyllabic word. The monosyllabic reduplication of the examples in (21) can be explained in this way. uujul
and limeye have a diacritically marked first syllable, which requires (secondary)
accenting; binta has a heavy first syllable, and both binta and mala consist of only
two syllables. With the extrametrical final syllable; this leaves only one syllable as
a potential foot.
Note that potential foot is quite distinct from the actual accent pattern of the
word, as in the examples in (22)b and (22)c, the first two syllables do not have any
accent (this is the result of a process of refooting the details of which need not
concern us here). This also shows that the prefixed reduplications, although based
on foot structure, do not interfere with the overall accentuation pattern of a word
and seem to form small prosodic units in their own right.
4. Reduplication in Yurakar and iconicity
It has long been recognized11 that reduplication often has iconic functions like iteration, pluralization, continuation, intensification, etc. that can mostly be c aptured

10. The functions of this marker and its counterpart =ja go beyond marking mere (lack of)
co-reference between subjects of linked clauses (see van Gijn 2011), but for the sake of convenience I gloss them ds different subject and ss same subject, respectively.
11. For example, Sapir (1921) contains a list of iconic meanings of reduplication.

Reduplication in Yurakar

by the Iconicity Principle (Lakoff & Johnson 2003:127), adapted to account for
reduplication by Kouwenberg & LaCharit (2005:534) as more of the same form
stands for more of the same meaning. The way this general principle applies, however, is not always clear.12 In this section, I argue that the three types of reduplication described above represent three different ways in which the Iconicity Principle
is applied.
Scholars have proposed additions to the general principle of iconicity to
account for the rather great variety of functions replication can have in the languages of the world. Kouwenberg and LaCharit (2005) divide the interpretation
of reduplication constructions in Carribbean and Creole languages into two types:
continuous occurrence versus discontinuous occurrence. Which of the two readings reduplication receives depends on the inherent semantics of the root, i.e.
whether it represents referents that can be individuated or not.
Abraham (2005) adds another parameter to the (dis)continuity distinction
based on individuation that is necessary to account for the cross-linguistic variety of functions of reduplication: intensional versus extensional value. Extensional
value increase is the classic iconic principle connected to reduplication: increase
of the number of instances of the referent (e.g. pluralization, distribution), whereas
intensional value increase relates to meaning components of the referent (e.g.
intensity, augmentative, continuative). Extensional and intensional value increase
are two different instantiations of the Iconicity Principle.
The three types of reduplication in Yurakar discussed above can be described
by these two parameters if we distinguish between the semantics of the input stem,
general effect of the reduplication process, and output of the reduplication process.
I discuss each of the three reduplication patterns from the perspective of the Iconicity Principle.
4.1. Full reduplication: Extension over space
Full reduplication encodes extensional meaning increase, and can be applied to
both inherently continuous (non-individuatable) and non-continuous (individuatable) referents, with different semantic effects. Apart from this basic function, diachronic changes seem to have yielded some deviations from this general pattern.
Some examples of the extensional function of discontinuous, individuatable
referents are given in (23). The key extensional element is extension over space.

12. One apparent puzzle which I will not address directly here because it does not apply to
Yurakar is the fact that reduplication is often associated with diminution. See e.g. Kouwenberg & LaCharit (2005) and Abraham (2005) for a discussion and proposals for a reconciliation with the Iconicity Principle.

Rik van Gijn

(23) a. boto~boto

buttony, having many buttons
b. buu~buu

lumpy, having many lumps
c. mere~mere

ribbed, having many ridges

Continuous, non-individuatable nominal roots that undergo full reduplication

can have the same extension-over-space interpretation that non-continuous referents have, but since they are non-individuatable the extension over a surface
is continuous.
(24) a. pujshi~pujshi

furry (all over)
b. samassama

watery, soaked (wet all over)
This is the basic function of complete reduplication, but there are a number of
instances of complete reduplication that are not properly described by extension
over space, like the following:
(25) a. muju~muju

center, middle

b. uli~uli


c. numa-n-numa


dawn, daybreak
It is likely that these are extensions on the basis of the basic semantic function of
complete reduplication, although the paths along which these reduplications are
formed semantically are not always entirely clear. Mujju is used to refer to bits of
land in the river that stick out of the water. Possibly the reduplication serves to
stress the fact that the location referred to is surrounded on all (or at least two)
sides by different matter or other objects, with respect to which it is positioned.
Uli~uli can be said to be built on the basis of examples like sm~sm (naked all
over, lit. bald~bald), the root form being replaced by the most salient feature of

Reduplication in Yurakar

male13 nudity. Numannuma dawn, daybreak (based on numma night), finally,

may be an example of the well-known transfer of spatial to temporal properties,
on the basis of continuous extensional meaning.
Another group of reduplicated forms that do not entirely fit the extension over
space function are certain colour terms, which in some cases can still be connected
to referents, as in the case of tbttb red (tbb blood) and ynjyn purple,
dark blue (ynn genipa) mentioned above. In (26) some other color terms are
given that have clearly been formed by reduplication, even though the root basis
no longer has any meaning in the language.14
(26) kulakkula pink

bajabaja grey, dull

woreworeshi black

If we base ourselves on the colour terms that still have an observable relation to
their root referent, this relation is at first sight not one on the basis of distribution
over space. The most likely path for these colour terms is that the roots on which
they are built referred to continuous referents. This is clear in the case of tbb
blood, and in the case of ynjyn this can be supported by the fact that the root
form ynn can refer either to the shrub, to the fruit, or to the colour produced
by the fruit. The latter interpretation is one of a continuous (non-individuatable)
referent, and so the interpretation is like the examples in (24).
A final group I want to mention here that is deviant from the extension in
space are animal names probably built on onomatopoeic sounds, like yutiyuti
capybara, sarajsara hummingbird, sp. (flapping of the wings), tubuytubu horned
screamer, and kochokkocho russet-backed oropendola. This group is interesting,
because the repetition of sound is of course a repetition in time, and not space.
This makes this group, although formally more similar to the complete reduplication group, functionally closer to the suffixed partial reduplication discussed in
the next section, and may therefore have to be regarded as a separate group.
4.2. Suffixed partial reduplication: Extension over time/participants
Suffixed partial reduplication represents extensional meaning increase either
over time (iterativity) or participants (multiple actions). In addition, partial
suffixed reduplication can indicate causation, as is shown above. The pluractional function can straightforwardly be connected to the Iconicity Principle.
13. There is also a female variant siribi~siribi (vagina~vagina).
14. The English glosses are approximations of the referents of these colour terms, see van
Gijn et al. (2010) for a detailed study of the Yurakar colour terminology.

Rik van Gijn

The most likely structure rise to have given rise to this type of reduplication is
the repetition of predicates to mean repeated action. Example (27) illustrates a
commonly found pattern in Yurakar where the repetition of words or strings
of words indicates repetition


ltta ense- mala- ltta ense- mala- chajtiya

one drink-3 one drink-3 always
One drank and went, another one drank and went, always.

Reduplication with a causative function is cross-linguistically uncommon, and

cannot be straightforwardly connected to the Iconicity Principle. Nevertheless, the
causative reading may be related to the Iconicity Principle if we consider the probable diachrony of this type of reduplication. The causative reading of the suffixed
reduplication is limited to ideophonic bases or to bases where there are traces of
an ideophonic past.
Ideophones are a common and salient feature of Yurakar grammar. They
extend well beyond sound imitations, and can refer to much more abstract sensory experiences (see van Gijn 2011 for more details). In their bare forms, they
must combine with predicates. The predicate directly follows the ideophone. The
examples in (28) illustrate this pattern for the ideophones sh bland and bana
arms stretched.



emejeme-sh sh
really-adv ideo:bland pv-bow-3=rep
He made his bow really bland.


ideo:arms.stretched lie.(down).sg-hab-3
It [i.e. a chameleon] would lie down with its legs stretched up.

Repetition of ideophones is also commonly used. Example (29) comes from a traditional narrative where a forefather of the Yurakar, called Aysa, is chased by a
giant snake. To escape from the snake, Aysa climbs into a tree, where he is followed
by one of the whiskers of the snake, which can be made to grow, spiraling up the
tree. Once the whisker reaches Aysa at the top of the tree, he cuts it and it unties,
spiraling down again, which is depicted by the repetition of the ideophone were
untie, be untied, get untied.


ideo:untie ideo:untie ideo:untie fall-hab-3=rep
ele=pchi earth=towards
were were were its whisker fell to the ground.

Reduplication in Yurakar

Crucially, ideophones are neutral with respect to whether they depict caused
events or not. This difference is made by the predicate that follows the ideophone,
as can be seen in (30) for the ideophone bak flat, even.



tt- ti-mesa
ideo:flat be-3 1sg-table
My table is flat.


ideo:flat grate:1sg.sbj 1sg-table
I sanded my table flat.

Although the diachronic path leading to causative reduplication cannot be established with certainty, examples such as (29) and (30) show that ideophones are a)
readily repeated, and b) ambivalent with respect to whether the scene they depict
was externally provoked or not.
This gives us the possibility to connect the causative reading to the Iconicity
Principle, because repetition with causative meaning may be interpreted as representing the two phases of a complex caused event, the event of causation and the
event of effect.
The major problem with this hypothesis is that, at least synchronically, there is
no evidence of a construction where two ideophones are repeated and they receive
a causative interpretation. Nevertheless, the following sequences do exist.


na aysa=ja
dem Aysa=thm 3sg.poss-machete=ins


Aysa, with his machete shj tom cut off [the tail of an agouti].15

In (31), there is a complex event where the protagonist Aysa cuts off the tail
of an agouti that is trying to steal pumpkins from his plantation, the two subevents (the moving of the blade of the machete through the air, and the snapping off of the tail) is depicted by two adjacent ideophones. Moreover, these
two ideophones are in a cause-effect relation to each other. Through this path
of the abstract ideophone-ideophone construction, the reduplicated causative
may have arisen. The internal complexity of these events can be connected to the
Iconicity Principle, because they indicate that in fact two events take place that
together form a complex, causative event.

15. According to this traditional story, the agouti used to have a tail like a squirrel.

Rik van Gijn

4.3. Prefixed partial reduplication: intensional

Prefixed partial reduplication marks intensity, which is also a function commonly
observed for reduplication. However, as is argued by Abraham (2005) it cannot
straightforwardly be connected to the Iconicity Principle mentioned above. Abraham (ibid.) proposes a distinction between extensional and intensional meaning
increase, or rather these are two levels at which meaning increase through reduplication can operate (p. 559):
Extension is the set of objects which RL [Reduplicated Lexicals RvG] refer to in
a world. It is its reference. The intension of the RL, on the other hand, is the set of
semantic features which denotate the entire meaning of RL in an exhaustive way.
(Abraham 2005:559)

As was discussed above, the two other types of reduplication in Yurakar are connected to extensional meaning increase, even though not necessarily in terms of
the number of referents, but also with respect to the spatial extension and the
internal complexity of events. Prefixed reduplication is related to intensional
meaning increase. As was the case with the two other reduplication types, the
specific way in which this general description applies depends on meaning aspects
of the input. I repeat three instances of prefixed reduplication mentioned above to
illustrate this point.
(32) a. meyej~meyeye-=w
ints ~disobedient-3=pl

They were really disobedient.
b. shwj~shw-ishta-=ti

When it was really getting dark


He walked really far.

The example in (32)a is a classic instance of intensional meaning increase: the

semantic features exhaustively describing meyeye apply to a greater extent in relation to the referent to which it applies, hence the translation really disobedient.
Example (32)b is slightly less straightforward, but can nevertheless be connected
to intensional meaning increase if we consider the fact that shw can either be
a state verb be dark or a process verb, become dark. The intensive reduplicated
prefix (and the future marker -(i)shta) force the process reading towards an endpoint. The closer one gets to this culmination point, the greater the extent to which
the semantic features describing the process apply. Intensive reduplication then
receives the interpretation of being very close to its endpoint. In (32)c, finally,

Reduplication in Yurakar

the verb mala go has no inherent endpoint, so the intensional meaning increase
results in the interpretation that the agent walked a very long time, or really far.
It must be mentioned that mala is the only activity verb that I have seen with the
intensifying reduplication prefix, so it is not a very common effect.
It might seem plausible at first sight that the reduplicated colour terms
are in fact instantiations of the intensity reduplication. Most of these colour
terms have either undergone gemination, which may be a reflex of the /h/ of
the intensity prefix, and ynjyn even has an added /h/ in the middle. Since
the roots of the reduplicated colour terms are disyllabic, it may be the case that
complete reduplication is not distinguishable from prefixed partial reduplication. However, in the first place, it is more common for a disyllabic root to have
only its first syllable reduplicated rather than both of its syllables (especially if
no syllabic material is added to the base). Second, in terms of semantics this
diachronic scenario would have to stipulate that the intensional reduplication
forced a comparative scale reading, whereby colours can be more or less like
the colour of the objects the root refers to, e.g. tbttb very much like blood,
ynjyn very much like genipa, etc. Because colour terms are different from
intensity reduplication both in formal and in semantic terms, connecting them
to intensity reduplication is far-fetched, and the analysis proposed in the previous seciton is more likely.
Intensional meaning increase also seems to be applicable at the propositional
level. If question words are reduplicated with the intensification prefix, the exactness of the answer is stressed.


ints~where lost-3=frus
Where exactly did he get lost?

When the predicate is preceded by the negator nish, the intensional meaning
increase takes this negation in its scope:


nish wilij~wilita-shta-=ti
neg ints~return-fut-3=ds
When he was never ever going to return...

Abraham (2005:560) mentions that a remaining problem for his proposal is that
it does not tell us when exactly the intensional validation has to be checked as
opposed to the extensional one. The solution Yurakar offers to this problem is
to reserve a formally clearly distinguishable reduplication pattern for intensional
meaning increase.
It is also possible that the intensive reduplication has followed a different diachronic path from the other two reduplication patterns, which basically seem to be

Rik van Gijn

grammaticalizations of repetitions of stems. Synchronically, speakers still use the

following construction alongside the reduplication prefix:


layshu bache-=ja kuj-mala-=ya
ref-dem squirrel send-3=ss then
When he had sent off this squirrel, he followed it for a long time.

The form kujmalaya has a very salient intonational pattern: the syllable consisting of the prefix ku- third person voluntary comitative and an added /h/ sound
is stretched for a long time, the /h/ taking up almost the entire lengthening
[kuh::malaja], just as is the case with the intensive reduplication. However, rather
than restarting the word with the same prefix ku-, the speaker continues with the
root mala. So there is no reduplication, but the structure is still very similar both
in form and meaning to the prefixed reduplication pattern. The prolongation of
the initial syllable (s) can sometimes be very long, so that it is actually rather functional to restart the word in order not to lose track of the meaning of the whole
word. Unfortunately, the proposed diachronic path cannot be proven because of
the lack of historical data, but it seems a reasonable assumption to consider structures such as the one in (35) as having given rise to the intensity reduplication
5. Conclusion
In this paper I have described three formally distinguishable reduplication patterns in the isolate language Yurakar, spoken in Bolivia, and I have argued that
each of the reduplication types can be connected to the Iconicity Principle in different ways. This is captured in Table 2.16
Table 2. Overview of reduplication processes in Yurakar
Reduplication type


Iconicity effect



entities, properties


distribution/extension over space

suffixed partial

events, states


distribution over time/causation

prefixed partial

properties, events



. I have chosen semantic characterizations for the input column, because semantics
rather than lexical class membership seems to determine the possibilities. But prototypical
associations are nouns for entities, adjectives or ideophones for properties, and verbs or ideophones for events/states.

Reduplication in Yurakar

The three reduplication types differ from each other in the kind of input they
take, the iconicity effect of the reduplication operation and as a function of the
former two the result of the reduplication operation.
In this way, Yurakar maximally exploits the Iconicity Principle, and on the
other hand is maximally transparent as to which type of iconicity is intended in
each case.

adv = adverbializer; caus = causative; de = discontinued event; del = delimiter;
dem = demonstrative; ds = different subject; ep = epenthetic; frust = frustrative;
fut = future tense; hab = habitual; ideo = ideophone; ins = instrument; intr =
intransitive; ints = intensifier; loc = locative; mal = malefactive; neg = negation;
poss = possessive; pl = plural; pv = possessive verbalizer; ref = referential; rep=
reportative; s = subject; sg = singular; ss = same subject; thm = theme; tr = transitive; vc = voluntary comitative

Abraham, Werner. 2005. Intensity and diminution triggered by reduplicating morphology:
Janus-faced iconicity. In Bernhard Hurch (ed), 547568.
Hirtzel, Vincent & van Gijn, Rik. In prep. Diccionario yurakar-espaol, espaol-yurakar. Ms.
Hurch, Bernhard (ed.). 2005. Studies on Reduplication. Berlin: Mouton de Gruyter. DOI:
Kouwenberg, Sylvia & LaCharit, Darlene. 2005. Less is more: Evidence from diminutive reduplication in Caribbean Creole languages. In Bernhard Hurch (ed.), 533546.
Lakoff, George & Johnson, Mark. 2003. Metaphors We Live By, 2nd edn. Chicago IL: The
University of Chicago Press. DOI: 10.7208/chicago/9780226470993.001.0001
Rubino, Carl. 2005. Reduplication: form, function and distribution. In Hurch (ed.), 1130.
Sapir, Edward. 1921. Language. New York NY: Harcourt, Brace & Co.
van de Vijver, Ruben. 1998. The Iambic Issue: Iambs as a Result of Constraint Interaction. HIL
Dissertations 37. Amsterdam: Holland Institute of Generative Linguistics.
van Gijn, Rik. 2010. Middle voice and ideophones, a diachronic connection: the case of Yurakar.
Studies in Language 34(2): 273297. DOI: 10.1075/sl.34.2.02gij
van Gijn, Rik. 2011. Semantic and grammatical integration in Yurakar subordination. In Subordination in Native South American Languages [Typological Studies in Language 97], Rikvan
Gijn, Katharina Haude & Pieter Muysken (eds), 169192. Amsterdam: John Benjamins.
DOI: 10.1075/tsl.97.07van
van Gijn, Rik, Hirtzel, Vincent & Gipper, Sonja. 2010. Updating and loss of color terminology
in Yurakar. Language & Communication 3: 240264. DOI: 10.1016/j.langcom.2010.02.002

Reduplication strategies in Kallawaya

Katja Hann

University of Cologne*
Reduplication is not a major word formation process in Kallawaya, a mixed
and secret language of Bolivia. Most reduplicative constructions resemble
compounding in that two (nominal) stems are linked. However, Kallawaya also
has a number of (in the Andean sphere) less common formal and functional
reduplication features, such as e.g. phrasal reduplication and the expression
of associative qualities. In the present paper, I will describe full and partial
reduplication in Kallawaya, arguing that the latter is possibly grammaticalized.
Further, I will also discuss iconic and less iconic functions of reduplication in
Keywords: mixed language; secret language; reduplication; associative qualities;
semantic lexicalizations

1. Introduction
Kallawaya is a mixed and secret language spoken in the Bolivian highlands near
Lake Titicaca. As a secret language, Kallawaya is used in curing ceremonies by
professional herbalists, mostly men,1 and as such it is not acquired as a native
language but only during the training to become a Kallawaya healer. The native
language of the Kallawaya is Quechua, but Spanish is becoming increasingly
important. Today, Kallawaya is severely endangered (see Adelaar 2007:19) and the
exact number of speakers and their competence are unknown. Kallawaya is also a
mixed language, recruiting its lexicon and grammar from different and genetically

* I am indebted to Swintha Danielsen and Fernando Ziga for their various and helpful
comments on this paper. The lexicon and grammar of Kallawaya is currently investigated by
the author of this paper and more definite results on the composition of Kallawaya will be
available soon. The research on Kallawaya is funded by the Deutsche Forschungsgemeinschaft
(DFG; German Science Foundation; Research Grant HA 6340/2-1).
1. According to Oblitas Poblete (1978:509; quoted in: Muysken 2009:148) there are few
exceptions to this.

Katja Hann

unrelated languages: the lexifier language of Kallawaya was Pukina (), while Quechua2 was the grammar-providing language. As a result, Kallawaya resembles the
typological profile of Quechua, being an agglutinative and almost entirely suffixing language. Apart from Pukina and Quechua, other languages, in particular
Aymara and Spanish, also contributed to the lexicon and grammar of Kallawaya
(for an overview of Kallawaya, see also Muysken 2009). Below is an example that
demonstrates the lexicon-grammar split (see Bakker & Mous 1994:4f.) observable
in Kallawaya: the underlined lexical items are of Pukina origin, while the grammatical material is provided by Quechua.
(1) nisi ika-chka-ni
komu-ntin sockei-ta
huane itila-pi

I carry-prog-1sg all-soc
medicine-do new capacho-loc
I am carrying all the medicine in a new capacho (medicine bag, KH).

(Mondaca n.d.)

Like most secret languages, Kallawaya is used only in a special domain, i.e. healing, and thus has a very reduced lexicon of approximately 1,000 basic lexical items
Given that Kallawaya is a secret language, we have an astonishing number
of sources on the language. However, most of these contain only lexical material, although some do so in abundance (see Table 1); phrases and/or clauses in
Kallawaya are rare. No ritual text, i.e. prayers and/or songs used in healing, have
been collected3 and for research on Kallawaya we therefore have to rely mostly on
dictionaries and wordlists.
Table 1 summarizes the sources on Kallawaya, which also form the database
for the present paper (see also Muysken 2009:151).
Apart from the authors current investigation, recent research on Kallawaya
has been carried out by Muysken (1994, 1997, 2009) as well as by Greg Anderson
and David Harrison (The Living Tongues Institute for Endangered Languages).4 In
his papers, Muysken analyses the available Kallawaya data with respect to grammar, while Anderson and Harrison made some recordings of Kallawaya, which are
available online. These are, however, only brief excerpts of possibly longer recordings and thus less suitable for an in-depth analysis of Kallawaya. In the following,
I will rely on the sources as listed in Table 1.

2. The Quechua grammar attested in Kallawaya is of a south-central variety, i.e. Ayakucho or

Cuzco Quechua (see Stark 1972:224, 226).
3. Note that the texts of the Kallawaya collected by Rsing (see e.g. 2008 and earlier) are
exclusively in Quechua.
4. see:

Reduplication strategies in Kallawaya

Table 1. Sources and database on Kallawaya

Author and date of publication

Type of material

Ponce Sanjins (n.d.)

field notes, approximately 50 words

Mondaca (n.d.)

19 phrases, approximately 130 words, grammatical


Soria Lens (1951)

approximately 50 words

Oblitas Poblete (1968)

about 15 phrases, approximately 12,000 words,

grammatical observations

Stark (1972)

approximately 130 words, grammatical analysis

Girault (1984)

plant names and description of their use

Gifford & Lancaster (ca. 1988)

general observations, 10 words

Girault (1989)

ca. 50 phrases, approximately 4,500 words, grammatical

observations, some recordings

Torero (2002)

field notes


ca. 84 phrases, ca. 17,190 words

The paper is organized as follows: after providing a definition of the term

reduplication in Section 2, I will turn to formal properties of Kallawaya reduplication in Section 3. I will begin with a brief presentation of full reduplication in
3.1 and an outline of continuous and discontinuous reduplication in 3.1.1, before
proceeding to partial reduplication in 3.2. In Section 4, I will present the functional properties of reduplication in Kallawaya, describing iconic functions, such
as distributivity (4.1) and intensification (4.2) as well as so-called associative qualities (4.3). In 4.4, semantic lexicalizations are discussed, followed by a brief presentation of botanical and zoological denominations formed by reduplication (4.5). I
will conclude with a discussion and a summary of my findings in Section 5.
2. Definition of reduplication
Right at the beginning it must be said that reduplication is not one of the major
word formation processes in Kallawaya; instead, non-reduplicative compounding appears to be more frequent and more important than reduplication. Nonetheless, reduplication in Kallawaya shows some interesting formal and functional
features that sets the language apart from the neighbouring, though unrelated
languages Quechua and Aymara, for which reason I consider it to merit a closer
Full (nominal) reduplication in Kallawaya (like in many other languages,
too) resembles compounding with respect to form and function and can thus be

Katja Hann

considered to form a subtype of compounding (for a similar view, see e.g. Inkelas 2006).5 However, it must be clearly said that Kallawaya also has reduplicative
structures that pose a challenge to this straightforward interpretation (see also
Sections 3.1.1 and 3.2). In order to include these interesting, though possibly not
prototypical instances of reduplication, I choose a rather broad definition of reduplication for the present paper: Reduplication is a morphological word formation
process in which some portion of a word is doubled (Inkelas 2006:417).
3. Formal properties
In what follows, I will discuss the formal properties of reduplication in Kallawaya.
I first provide a brief description of full reduplication in Section 3.1, which will be
followed by a presentation of continuous and discontinuous reduplication in Section 3.1.1. Partial reduplication will be addressed in Section 3.2.
3.1 Full reduplication
Full reduplication can be defined as the repetition of an entire word, word stem
(root with one or more affixes), or root [...] (Rubino 2005:11). In Kallawaya,
nouns (which include adjectives, quantifiers, and numerals) and verbs can be
fully reduplicated, although it must be pointed out that full reduplication of verbal
bases is less frequent than full nominal reduplication. Stems as well as entire words
can be fully reduplicated. In the case that a stem is repeated, the reduplication is
mostly continuous, while in the case of a reduplicated word, the construction can
be discontinuous.
3.1.1 Continuous and discontinuous reduplication
In continuous reduplication, the lexical base and its reduplicant follow each other
immediately without any intervening phonological6 and/or grammatical material,
see Examples (2) and (3). In Example (2), an adjectival stem is reduplicated, while
in Example (3) the verbal stem jalla- die undergoes reduplication.



(Oblitas Poblete 1968:137)

5. For an interpretation of reduplication as affixation, see e.g. Marantz (1982).

6. Note that in some languages phonological material can come in between the lexical base
and the reduplicant. This phonological material does not necessarily correspond to any lexical
or grammatical morpheme in the given language (see Rubino 2005:17).

Reduplication strategies in Kallawaya



(Girault 1989:109)

Note that continuous reduplication does not exclude the possibility of grammatical marking on the final component of the reduplication. This is the case in
example (3), where the entire reduplication is nominalized by the Quechua action
nominalizer -y.
In discontinuous reduplication, some phonological and/or grammatical material comes in between the lexical base and the reduplicant. In Kallawaya, grammatical material like inflectional or derivational affixes can interrupt the sequence
of lexical base and reduplicant, as shown in Examples (4) and (5). The intervening grammatical markers can be identical (4) or different (5). It must be noted,
though, that there are no cases where only the first component receives affixation,
while the second is unmarked, i.e. either both components of a reduplicative construction show affixation, or only the last member carries a grammatical marker
(3), or both are unmarked (2).



above-3sg.poss above-3sg.poss
katu-q7 katu-n
great-ag great-3sg.poss
greatest; highest

(Oblitas Poblete 1968:137)

(Oblitas Poblete 1968:137)

Both (4) and (5) are constructed as possessive phrases, which becomes particularly
evident if a literal translation of Example (5) is provided: the greatest its greatest,
i.e. the greatest of the greatest.
Constructions such as Examples (4) and (5) lead to another phenomenon of
Kallawaya reduplication, namely nominal reduplication as part of a verb phrase.
It appears that in some cases the reduplication itself is not sufficient to express a
certain concept, but a verb is required in addition. Consider Examples (6) and (7).
(6) ujsi-man ujsi-man jechka-na

one-dir one-dir throw-inf

to distribute
(lit.: to throw to one, to one)
(Oblitas Poblete 1968:70)
7. Originally spelt as katu-j. According to the normalized spelling of Quechua (see e.g.
Plaza Martnez 2009:227, 278), the agentive marker in this and the following examples is
represented as q instead of j. The same applies to the hispanicized spelling of c instead

Katja Hann


ujsi-manta ujsi-q pate-q

one-ag run-ag
(lit.: one runner from somewhere)

(Oblitas Poblete 1968:142)

Especially when the literal translations of Examples (6) and (7) are considered, it
becomes obvious that the particular reading of the given expression results only
from a combination of the reduplication with the specific verb. That is, the expressions of ujsi-man ujsi-man to one, to one and ujsi-manta ujsi-q one from somewhere alone would probably not yield the reading of to distribute and turncoat,
respectively, but the semantics of the given verb are required for that particular
Given these observations, one might question whether such phrase-like constructions are truly reduplications or whether they are better analysed as some
kind of (syntactic) repetition or another recursive device.8 Although at first glance
these constructions appear to be rather different from constructions as shown in
e.g. (2) and (3), I propose that they are nonetheless best interpreted as instances of
reduplication (though admittedly less prototypical ones). They differ from repetitions and recursive patterns, such as e.g. drink drink! or very, very good9 in that
they do not express the same meaning twice but convey an altogether new meaning, i.e. a new expression is created, in which the meaning of the individual components is not reflected in the resulting expression any more (see also Section 4.4).
Furthermore, these reduplicative constructions lack the illocutionary force that
comes in such recursive or repetitive patterns such as drink drink!.10 Constructions such as (6) and (7) are possibly not prototypical examples of reduplication,
but they still represent instances of this morphological process. Structures such
as these may be tentatively called phrasal reduplications (in analogy to the term
phrasal compounds; see e.g. Scalise & Bisetto 2009:47).
At this point it should be mentioned that Kallawaya does not have what
Rubino (2005:18) calls automatic reduplication, which is reduplication that is

8. For an overview of repetition and recursiveness, see e.g.:

9. Examples taken from:
10. Note that the illocutionary force in expressions such as drink drink! is conveyed by the
speech act, here: imperative, rather than the reduplication alone. An utterance such as drink!
would bear the same illocutionary force without being reduplicative. Although we lack Kallawaya texts, we may assume that a similar invitation in Kallawaya would be expressed via the
Quechua imperative marker y, yielding something like tajchi-y drink!. Thus, it is likely that
the illocutionary force of an invitation or a command would be expressed by bound affixes in
Kallawaya, not by reduplication.

Reduplication strategies in Kallawaya

obligatory in combination with another affix, and which does not add meaning
by itself to the overall construction; the affix and reduplicated matter together
are monomorphemic [...] (Rubino 2005:18). Each of the grammatical affixes
involved in Kallawaya reduplication can occur outside reduplication and does not
trigger reduplication by itself. Furthermore, the suffixes attached to reduplicative
components in Kallawaya clearly have a grammatical function and do not just
add phonological material to the reduplication (see Examples (4) through (7)). I
therefore propose that constructions, such as Examples (4) through (7), are not
interpreted as monomorphemic.
3.2 Partial reduplication
As Inkelas (2006:417) puts it, partial reduplication doubles some phonologically
characterized subpart of the word or stem [...]. The evidence for partial reduplication in Kallawaya is scarce and it is not entirely certain whether the language
has partial reduplication at all. The database yields (so far) only two instances
that could be interpreted as partial reduplication. Consider Examples (8) and (9),
where the (a) examples show the assumed lexical base and the (b) examples their
(8) a. okha-ri-na


to buy
b. okha~kha-na


to sell
(Soria Lens 1951:33)11
(9) a. uke-ku-na


to collide
b. eke-ke-y11

(Girault 1989:89)
Should these examples be indeed instances of partial reduplication, then the final
CV-sequence of the lexical base is copied and follows the base immediately. One
major problem arising with both examples is that the function of reduplication

11. The vowel change from /u/ to /e/ is attested elsewhere in the Kallawaya data and may be
analysed as a lexical manipulation (see Mous 2003). If this interpretation is correct, then the
resulting construction could be described as a complex construction (see Rubino 2005:15;
Hann & Muysken 2014).

Katja Hann

remains unclear. According to Soria Lens (1951:33), the partial reduplication in

example (8)b appears to mark the difference between to sell and to buy, which,
however, cannot be verified or falsified as the only other source that lists both
verbs Oblitas Poblete (1968) expresses the difference between these two events
by affixation (compare Examples (10)a, b). All other sources (see Table 1) provide
only the verb buy, which is nowhere derived by reduplication.
(10) a. uka-ja-na


to sell
b. uka-ja-ri-ku-na


to buy
(Oblitas Poblete 1968:146, 60)
A similar observation is true for example (9)b, and Girault (1989) is the only author
who provides the form eke-ke-y for collision. Again, it cannot be verified whether
collision is regularly built by reduplication or whether the form in example (9)b is
a single instance. Furthermore, the impact the reduplication in example (9)b has
iseven less clear than it is in example (8)b.12
Given these observations, one might question whether the instances in Examples (8)b and (9)b can be analysed as partial reduplications, particularly if we take
into consideration that partial reduplication does not seem to have any effect on the
lexical base. However, a meaningful interpretation of the sequences -kha (8)b and
-ke (9)b as regular suffixes of Kallawaya also appears to be less suitable, at least at the
current state of research. Furthermore, even if we assume that they are suffixes (and
not partial reduplications), their meaning still remains unclear. Finally, the forms
-kha and -ke look suspiciously like partial reduplications of the final syllable of the
lexical base. Therefore, I tentatively suggest that Kallawaya had partial reduplication,
which, however, had already been grammaticalized when the data were collected.
Here, I would like to point to Bybee, Perkins & Pagliucia (1994), who view partial
reduplication as the result of grammaticalization, arising from full reduplication.13
Ifollow the proposal forwarded by Bybee et al. (1994), suggesting that partial reduplication in Kallawaya developed from instances of full reduplication, where the
final component became phonologically truncated and semantically bleached.
These remnants of partial reduplication in Kallawaya may well be derived from
its grammar-providing language, Quechua. Although today Quechua shows partial
reduplication only in lexicalized onomatopoeic expressions and ideophones, such

12. Note that the nominalization in Example (9)b is due to the Quechua action nominalizer -y.
13. For a critical discussion of Bybee, Perkins & Pagliucia (1994), see e.g. Hurch & Mattes

Reduplication strategies in Kallawaya

as rapha-pa-pa-y to flutter or ala-l-w how cold! (Academia Mayor de la Lengua

Quechua 2005:7, 513), these Quechua forms suggest that partial reduplication once
played a greater role in Quechua than it does today (van de Kerke, p.c.). In addition,
these Quechua examples also show partial reduplication of the final CV-sequence
of the lexical base, thus being formally identical to Kallawaya partial reduplication. It is therefore quite possible that Kallawaya adopted this word-formation
process from its grammar-providing language. Whether this happened at a time
when partial reduplication was still productive in Quechua or whether Kallawaya
extended the use of partial reduplication beyond onomatopoeic expressions and
ideophones on its own still requires further research.
4. Functional properties
Although the functions of reduplication show cross-linguistic variation, most can
be described as being iconic. One seminal definition of iconicity is provided by
the Iconic Principle of Reduplication, stating that [m]ore of the same form stands
for more of the same meaning (Lakoff & Johnson 1980:128; quoted in Kouwenberg & LaCharit 2005:534; see also van Gijn, this volume). I adopt this definition
of iconicity for the following presentation of functional properties of Kallawaya
Kallawaya shows two of the most common and most iconic functions of reduplication, namely distributivity and intensification, while a third function, the
expression of associative qualities, is possibly derived from an intensifying reading
and therefore subsumed here under iconic functions of reduplication. I will first
discuss distributivity and intensification in Sections 4.1 and 4.2, before turning to
a description of associative qualities in Section 4.3. In 4.4 follows a discussion of
semantic lexicalizations, while in Section 4.5 plant and animal names created by
reduplication are presented.
4.1 Distributivity
The term distributivity refers to the individuation or specific grouping of entities or events/ actions.14 From the available data it seems that in Kallawaya only
nominal stems partake in distributive reduplication. Consider Examples (11)
and (12).

14. Note that a difference is made here between distributivity and distribution. While the
latter term refers rather generally to the distribution of something and is not necessarily
restricted to linguistic categories and/or notions, distributivity is a linguistic term with the
functions noted above. I owe thanks to Fernando Ziga for pointing this difference out to me.

Katja Hann




(Oblitas Poblete 1968:70)

(Girault 1989:38)

Even when the resulting expression is verbal, as is the case with example (13)
(example (6) repeated for convenience from 3.1.1), it is actually the numeral ujsi
one that is reduplicated.


ujsi-man ujsi-man jechka-na

one-dir one-dir throw-inf
to distribute

(Oblitas Poblete 1968:70)

4.2 Intensification
Intensification increases the notion expressed by the noun or the event designated
by the verb, respectively. The evidence for reduplication of verbal stems in Kallawaya expressing intensification is rare (but cf. example (15)), and most instances
of intensification by reduplication are expressed by adjectival stems.



(Oblitas Poblete 1968:81)

The following example (15) is one of the two instances of verbal stem reduplication
with an intensifying meaning. It is furthermore remarkable that the resulting expression is translated as a verb despite the nominalization (in contrast to Examples (16)
and (17) which have a similar structure; see Section 4.3 below). The infinitive marker
-na on the second component probably represents the citation form.


see-nmlz-lim see-inf
to contemplate

(Oblitas Poblete 1968:62)

4.3 Associative qualities

Closely related to intensification is the expression of associative qualities. This
function refers to a quality and/or some behavior that is typically associated

15. Note the difference in meaning between the discontinuous Example (5) and the continuous reduplication in Example (14). It appears that the possessive marking in Example (5)
contributes to the intensified reading of the greatest.

Reduplication strategies in Kallawaya

with aperson or an entity (see Rubino 2005:21). Note that the structure of the
two examples given below is identical, i.e. the first member of the reduplication
is nominalized and carries a limitative marker -lla, while the second component
receives the agentive marker -q.



eat- nmlz-lim eat-ag

(Oblitas Poblete 1968:84)

sleep-nmlz-lim sleep-ag

(Oblitas Poblete 1968:71)

I suggest that the expression of associative qualities developed from an intensifying reading; i.e. a person who sleeps a lot is a sleepy person and can thus be
described as a sleepyhead (see also Hann & Muysken 2014).16 Accordingly, the
expression of associative qualities is grouped together here with other iconic functions of reduplication.
Before proceeding to a discussion of semantic lexicalization, I would like to
raise briefly the issue of functional distinction. It is claimed by Rubino (2005:12)
that languages with full and partial reduplication also display a functional distinction between these two reduplication types. For Kallawaya, however, no meaningful functional distinction can be made, as the few instances of partial reduplication
observable in the Kallawaya corpus are probably grammaticalized and the impact
partial reduplication might have had on the construction remains unknown.
Therefore, no functional distinction between full and partial reduplication in
Kallawaya is proposed.17
4.4 Semantic lexicalization
The lexical items to be discussed in the following differ from iconic reduplications
presented in Sections 4.1 to 4.3 and are thus treated under a different heading.
In contrast to iconic functions of reduplication, the impact reduplication has on

16. Apparently, the term associative qualities goes back to Newman (1944, on Yokuts). The
observation, however, that intensification can lead to attributing (permanent) qualities to an
entity is made elsewhere, though without necessarily applying the term associative qualities.
Thus, in Yami (Philippines) the reduplication of lavi cry results in ka-lavi-lav-in crybaby
(Rau & Dong 2005:75). Although labelled intensification by the authors (ibid.), the process is
comparable to what is termed the expression of associative qualities here.
17. Note that this does not preclude that when partial reduplication was still productive in
Kallawaya, there was a clear functional distinction between these two reduplication types.

Katja Hann

semantic lexicalizations is unpredictable, i.e. semantic lexicalizations cannot be

described straightforwardly as resulting from intensification, distributivity, or any
other iconic process that would yield a predictable meaning. This type of lexicalization is also referred to as [s]emantic lexicalization [which] has occurred when,
for example, a compound has developed a meaning that is unpredictable given the
meaning of its parts (Bakken 2006:107).18
Kallawaya has a number of expressions that arise from semantic lexicalization as defined above. Below are some examples where the meaning of the lexical
base and that of the reduplication are not related by any iconic function; instead,
the semantics of the reduplication are largely unpredictable from the meaning of
the base. Note that most instances of semantic lexicalization are produced by full,
nominal, continuous reduplication.



trap, pitfall

(Oblitas Poblete 1968:101)

(Girault 1989:41)

In Example (18) the relationship between the lexical base and its reduplication
cannot be adequately captured by iconic notions and, accordingly, the meaning ofthe reduplication is not directly deducible from the lexical base (see also
Hann & Muysken 2014). Similar considerations apply to Example (19), since
there is no iconic relationship between the lexical base and its reduplicant, and the
meaning of kilin~kilin fork cannot be inferred from the semantic content of the
lexical base.
Although we lack sufficient data on most Quechua and Aymara varieties, it
seems that this particular function of reduplication is unattested for these Andean
languages.20 It thus appears to be a peculiarity of Kallawaya and I propose that
the creation of semantic lexicalizations by reduplication can be explained by the
severely restricted lexicon of this language and its need to expand it in order to
serve various communicative requirements (see also Hann & Muysken 2014).
As the contents of what is said in Kallawaya should furthermore remain largely
18. What has been called phrasal reduplication here (see Section 3.1.1) resembles these
semantic lexicalizations, since in phrasal reduplication, the meaning of the reduplicated
expression cannot be immediately inferred from the lexical base, either.
19. In Oblitas Poblete spelt as ithi iti.
20. However, Shuar seems to have a similar function of reduplication (see Hann &

Reduplication strategies in Kallawaya

nintelligible to outsiders, semantic lexicalizations, whose meanings

be inferred from their lexical bases, are well suited for this purpose (see e.g.
Mous 2003:220; Matras 2009:291, on secrecy as a constituting feature of secret
4.5 Plant and animal names
In the following, I will briefly present plant and animal names created by reduplication. In contrast to the instances of reduplication discussed so far, no lexical bases for these botanical and zoological denominations can be identified (any
more) and the function of reduplication thus remains unknown. These plant and
animal names thus represent a high degree of lexicalization. However, it must be
mentioned that referring to plants and animals by reduplicative expressions is no
peculiarity of Kallawaya, but is found in Andean and lowland languages, too (see
Hann & Muysken 2014; see also van Gijn and Tacconi, both this v olume). Below
are some examples of plant and animal names created by reduplication. It is particularly plants that are referred to by reduplication, while animals are less commonly denoted by reduplicative expressions. As with semantic lexicalizations (see
Section 4.4.), these botanical and zoological names are created by full, continuous
(20) llalwa~llalwa a type of aster (Achyrocline alata)22 

(Girault 1984:472)

(21) m
 ojlle~mojlle a type of fern (Cheilanthes incarum)23 

(Girault 1984:113)

That reduplicative expressions are a common means to refer to plants and animals
can be seen by the fact that Aymara and Quechua, too, designate fern, for instance,
by reduplication: chusi~chusi (Dryopteris palacea C. Christ) in Aymara (DeLucca
1983:715), raki~raki (not further specified) in Quechua (Academia Mayor de la
lengua quechua 2005:848).
Example (22) shows that not only plants and animals are referred to by reduplicative constructions, but apparently also figures from folk stories.

jenka~jenka ogre  (Oblitas Poblete 1968:109)24

21. Whether the lexical base is nominal or verbal has to remain an open issue, as we have no
information on the lexical base.
22. see: (21 October 2012).
23. see: (21 October 2012).
24. Note that Oblitas Poblete (1968:105) lists a word llenka grimace, which might be the
lexical base of jenka~jenka ogre.

Katja Hann

There are no ideophones in the Kallawaya database, which are very common in
both Aymara and Quechua, expressing e.g. pain or referring to cold (see also
Section 3.2). We may assume that these spontaneous and in part emotional exclamations would be uttered in the Kallawayas native Quechua, as Kallawaya is
acquired only as a second language.
5. Discussion and conclusions
In the preceding sections, I have discussed formal and functional features of reduplication in Kallawaya. While full reduplication, especially that of nominal stems,
is productive and can be continuous or discontinuous, it has been argued that
partial reduplication in Kallawaya is no longer productive, the instances of partial
reduplication that we find in the corpus being possibly grammaticalized. Apart
from full, continuous reduplication Kallawaya also shows reduplicative expressions whose particular meanings develop from a combination of the reduplication
with the semantics of a given verb. For these constructions the term phrasal reduplication has been proposed.
Most functions of reduplication in Kallawaya can be said to be iconic and
convey notions of intensification and distributivity. The expression of associative
qualities is likewise subsumed under the heading of iconicity, as this particular
function probably developed from an intensifying reading. I have further discussed instances of so-called semantic lexicalization (see e.g. Bakken 2006), where
full, continuous reduplication (of nominal stems) results in meanings unpredictable from those of the respective lexical bases. It must be noted that this function
is possibly not attested for Aymara and Quechua. Finally, I have listed examples
of names of plants and animals formed by reduplication. These plant and animal names are highly lexicalized with no lexical base being available (anymore).
Table 2 contrasts the findings on reduplication in Kallawaya with reduplication
patterns in Quechua.25
Although Kallawaya and its grammar-providing language Quechua share
some features of reduplication, Kallawaya also shows some features (possibly)
unattested for Quechua (cf. also Matras & Bakker 2003:5, on Media Lengua and
Quechua). These are grammaticalized partial reduplication with verbs (other than
onomatopoeic verbs), the expression of associative qualities as well as semantic
lexicalizations of the type kilin~kilin fork (see Example (19) in Section 4.4). At

25. Referring to Cuzco Quechua as one of the two most likely Quechua varieties that
provided the grammatical template for Kallawaya.

Reduplication strategies in Kallawaya

Table 2. Reduplication in Kallawaya and Quechua

Formal and functional features



productive full (nominal) reduplication

productive partial reduplication

grammaticalized partial reduplication

with verbs
with nouns and adjectives
with ideophones

possibly not


associative qualities


semantic lexicalization

plant and animal names built by reduplication

least for the latter two we may assume that these are independent innovations
of Kallawaya. These particular functions of Kallawaya reduplication can probably be explained by the reduced Kallawaya lexicon and the resulting need of its
speakers to expand the lexicon of their language to fulfil various communicative
requirements. As has been outlined in Section 3.2, partial reduplication of verbs
in Kallawaya, nowadays grammaticalized, has possibly been taken from Quechua,
where today partial reduplication is grammaticalized, too, and occurs only in onomatopoeic verbs and ideophones. However, for the time being, it remains uncertain whether Kallawaya speakers adopted this process from Quechua when partial
reduplication was still productive in Quechua or whether Kallawaya extended
the use of partial reduplication on its own. In the latter case, partial reduplication (of non-onomatopoeic verbs) would be another instance of an independent
Kallawaya innovation. Finally, the use of partial reduplication for ideophones is
another feature in which Kallawaya differs from Quechua, since so far no ideophones have been found in the Kallawaya database. One explanation for this lack
is that Kallawaya speakers would rather use their native language Quechua for
these emotional exclamations.
In sum, there are two features that can be described as independent innovations of Kallawaya (the expression of associative qualities and semantic lexicalizations), while for a third feature (partial reduplication with non-onomatopoeic
verbs) it is uncertain whether Kallawaya is more innovative or more conservative
than its grammar-providing language Quechua.
To conclude, it can be argued that reduplication is not the primary means
of word formation in Kallawaya, as instances of reduplication are comparatively
rare. Furthermore, reduplication is mostly iconic, and it is not used to express

Katja Hann

g rammatical categories, at least not in a systematic way that would suggest a

certain degree of grammaticalization. In this, Kallawaya equals its grammarproviding language Quechua, where reduplication is often stylistic, as has also
been remarked by Adelaar (2004:1455). One possible explanation for this rather
emphatic use of reduplication in Quechua may be that most grammatical categories which in other languages are conveyed by reduplication are expressed by
the rich nominal and verbal morphology in Quechua; hence, reduplication is not,
at least not regularly, employed to express grammatical categories, such as e.g. a
progressive aspect. Since Quechua is the grammar-providing language for Kallawaya, this language, too, encodes grammatical categories by affixation and relies
on reduplication mostly for stylistic reasons (see also Hann & Muysken 2014).
However, as has been shown above, Kallawaya is not a complete grammatical
copy of its grammar-providing language, but has developed some new features of
reduplication on its own.

1sg = first person singular; 3sg.poss = possessive third person singular; ag =
agentivizer; dir = directional; do = direct object; inch = inchoative; inf = infinitive; lim = limitative; loc = locative; nmlz = nominalization; prog = progressive;
red = reduplication; sep = separative; soc = sociative; ? = unknown element; ~ =

Academia Mayor de la Lengua Quechua. 2005. Diccionario Quechua-Espaol-Quechua. QheswaEspaol-Qheswa. Simi Taqe, 2nd edn. Cusco: Gobierno Regional Cusco.
Adelaar, Willem. 2004. Quechua (Quechua). In Morphologie/Morphology, Vol. II, Geert E. Booij,
Christian Lehmann & Joachim Mugdan (eds), 14531464. Berlin: Walter de Gruyter.
Adelaar, Willem. 2007. Threatened languages in Hispanic South America. In Language Diversity
Endangered, Matthias Brenzinger (ed.), 928. Berlin: Mouton de Gruyter.
Bakken, Kristin. 2006. Lexicalization. Encyclopedia of Language and Linguistics, 2nd edn, 106198.
Oxford: Elsevier.
(1 November 2012)
Bakker, Peter & Mous, Maarten (eds). 1994. Mixed Languages: 15 Case Studies in Language
Intertwining. Amsterdam: IFOTT.
Bakker, Peter & Mous, Maarten. 1994. Introduction. In Peter Bakker & Maarten Mous (eds),
Bybee, Joan, Perkins, Revere D. & Pagliuca, William. 1994. The Evolution of Grammar: Tense,
Aspect and Modality in the Languages of the World. Chicago IL: University of Chicago Press.

Reduplication strategies in Kallawaya

Crevels, Milly & Muysken, Pieter (eds). 2009. Lenguas de Bolivia: mbito andino, Vol. I. La Paz:
Plurales editores.
DeLucca, Manuel. 1983. Diccionario Aymara-Castellano. Castellano-Aymara. La Paz: Comisin
de Alfabetizacin y Literatura en Aymara (CALA).
Gifford, Douglas J. & Lancaster, Elizabeth. 1988. Tradition and Change among the Grass Roots
of Callawaya Indigenous Medicine [Working Paper No. 18]. St. Andrews: Centre for Latin
American Linguistic Studies, University of St. Andrews.
Girault, Louis. 1984. Kallawaya, gurisseurs itinrants des Andes. Paris: ditions de lORSTOM.
Girault, Louis. 1989. Kallawaya: El idioma secreto de los Incas: Diccionario. La Paz: UNICEF,
Hann, Katja & Muyken, Pieter. 2014. Reduplication in Andean languages. In Reduplication in
Indigenous Languages of South America, Gale Goodwin Gmez & Hein van der Voort (eds),
3976. Leiden: Brill.
Hurch, Bernhard & Mattes, Veronika. 2004. ber die Entstehung partieller Reduplikation. http:// (28 October 2012)
Hurch, Bernhard & Mattes, Veronika (eds). 2005. Studies on Reduplication. Berlin: Mouton de
Gruyter. DOI: 10.1515/9783110911466
Inkelas, Sharon. 2006. Reduplication. In Encyclopaedia of Language and Linguistics, 2nd edn,
Keith Brown (ed.), 417419. Oxford: Elsevier.
referenceworks/9780080448541 (1 November 2012)
Kouwenberg, Silvia & LaCharit, Darlene. 2005. Less is more: Evidence from diminutive reduplication in Caribbean Creole languages. In Bernhard Hurch & Veronika Mattes (eds),
Lakoff, George & Johnson, Mark. 1980. Metaphors We Live By. Chicago IL: University of Chicago
Matras, Yaron. 2009. Language Contact. Cambridge: CUP. DOI: 10.1017/CBO9780511809873
Matras, Yaron & Bakker, Peter (eds). 2003. The Mixed Language Debate: Theoretical and Empirical Advances. Berlin: Mouton de Gruyter. DOI: 10.1515/9783110197242
Matras, Yaron & Bakker, Peter (eds). 2003. The study of mixed languages. In Yaron Matras &
Peter Bakker (eds), 120.
Marantz, Alec. 1982. Re reduplication. Linguistic Inquiry 13(3): 435482.
Mondaca, Jaime. n.d. La lengua callawaya: Apuntes de un cuaderno de campo [Working Paper No.
19]. St. Andrews: Centre for Latin American Linguistic Studies, University of St.Andrews.
Mous, Maarten. 2003. The linguistic properties of lexical manipulation and its relevance for
Ma. In Yaron Matras & Peter Bakker (eds), 209235.
Muysken, Pieter. 1994. Callahuaya. In Peter Bakker & Maarten Mous (eds), 207211.
Muysken, Pieter. 1997. Callahuaya. In Contact Languages. A Wider Perspective [Creole Language Library 18], Sarah G. Thomason (ed.), 427448. Amsterdam: John Benjamins. DOI:
Muysken, Pieter. 2009. Kallawaya. In Milly Crevels & Pieter Muysken (eds), 147167.
Newman, Stanley. 1944. The Yokuts Language of California. New York NY: Johnson Reprint Co.
Oblitas Poblete, Enrique. 1968. El idioma secreto de los incas (vocabulario castellano-callahuaya).
La Paz: no publisher.
Oblitas Poblete, Enrique. 1978[1960]. Cultura Callahuaya. La Paz: Ediciones Populares

Katja Hann
Plaza Martnez, Pedro. 2009. Quechua. In Milly Crevels & Pieter Muysken (eds), 215284.
Ponce Sanjins, Carlos. n.d. Unpublished field notes.
Rau, Der-Hwa Victoria & Dong, Maa-Neu. 2005. Yami reduplication. In Concentric: Studies in
Linguistics 31(2): 5787. (6 January 2013)
Rsing, Ina. 2008. Religion, Ritual und Alltag in den Anden: Die zehn Geschlechter von Amarete
[Zweiter ANKARI-Zyklus: Kollektivrituale der Kallawaya-Region in den Anden B
Berlin: Reimer.
Rubino, Carl. 2005. Reduplication: Form, function and distribution. In Hurch & Mattes (eds),
Scalise, Sergio & Bisetto, Antonietta. 2009. The classification of compounds. In The Oxford
Handbook of Compounding, Rochelle Lieber & Pavol tekauer (eds), 3453. Oxford: OUP.
Soria Lens, Luis. 1951. Pequeo vocabulario callawaya. Boletn de la Sociedad Geogrfica de La
Paz, 3235.
Stark, Louisa R. 1972. Machaj-Juyai: Secret language of the Callahuayas. Papers in Andean
Linguistics 1(2): 199228.
Torero, Alfredo. 2002. Idiomas de los Andes. Lingstica e historia. Lima: IFEA and Editorial

Compounding in Kalapalo, a Southern

Cariban language
Ellen B. Basso

(University of Arizona)*
Compounds in the Southern Cariban language Kalapalo include (a) extended
verb stems formed by nominal roots prefixed to basic verb stems and more
commonly (b) clitic clustering involving two or more members of one of the four
different semantic types of this word class. Clitic compounds have important
syntactic and semantic functions in the context of clause chaining and interclausal
reference marking.
Keywords: compounding; extended verb stems; clitic clustering;
interclausalreference marking; switch-reference

1. Introduction
In this paper, I discuss compounding in Kalapalo, a highly agglutinative clause
chaining Southern Cariban language spoken by approximately 600 people in
the Alto Xingu region of Mato Grosso, Brazil. Compounds are constructions
that are formed from more than one unit that may operate independently, but
which together manifest as a grammatical word having distinctive syntactic and
semantic functions. All types of Kalapalo compounding processes may be clearly
distinguished morphologically and functionally from the numerous inflectional
and derivational prefixing and suffixing processes in the language. Compounds
are also often unusual phonological words. Kalapalo compounds include two
subtypes: a) extended verb stems formed by nominal roots prefixed to basic verb

* This paper makes use of data collected during field research in several Kalapalo communities between 1966 and 1998. I am grateful as always for the interest shown by the residents
to my work on their language. Several scholars have kindly taken the time to review some of
my grammatical analysis of the language. I am particularly grateful to Patience Epps, and the
editors of this volume. Research was supported by the National Science Foundation, WennerGren Foundation for Anthropological Research, Inc., John Simon Guggenheim Foundation,
and the University of Arizona.

Ellen B. Basso

stems, and more commonly b) clitic compounding involving the clustering of two
or more members of the four different semantic types of this word class. Clitic
compounding both within a clitic semantic set and involving members of more
than one set is a functionally important feature of Kalapalo clause chaining, and
is particularly evident in interclausal reference marking that links clause chains in
extended narrative discourse.
I first introduce the language with a typological overview in Section 2.
In Section 3, I discuss formation of extended verb stem compounds. Section 4
includes a review of the inventory and general use of clitics and related particles. In
Section 5, I examine clitic clustering across semantic sets. In Section 6, I describe
in more detail taxis morpheme clustering, and in Section 7, interclausal reference
marking compounds. Section 8 is a discussion of compounding features in Kalapalo with reference to the work by Mithun (2000, 2010) and Aikhenvald (2002)
and Section 8 is a summary and review of the value of clitic compounding in
assisting clausal cohesion in this chaining language.
2. Typological overview
Kalapalo is a head marking highly agglutinative language1 with some fusional
features and numerous suffixes and prefixes. It is a clause chaining language with
a complex set of interclausal reference markers, clausal coordinators which manifest syntactic, semantic and narrative discourse pragmatic functions. There is a
clear distinction between main versus dependent clauses in the language. The
boundaries between the major word classes (noun, verb, to a lesser extent, adverb)
are not strict. Stem formation processes often involve pre-categorial roots to
which derivational radicals are suffixed; these radicals specify verbal valency
(intransitive, transitive, or ambitransitive), and nominal or adverbial word class.
The resultant root+radical stems are followed by both optional (e.g. causal, possessive) and required (plural concordance and aspectual) suffixes. Extended verb
stems are formed with nominal roots (some of which are followed by adverbial
elements) prefixed to basic verb stems that may otherwise stand alone. Word class
changing derivations are pervasive, involving all three major word classes. There
are eight types of main and independent verbal and n
ominal clause constructions,

1. Descriptions of the closely related Kuikuro language may be found in Franchetto (1990,
2006 and 2010). Kalapalo data verify her findings (2006) regarding the realization of tokens of
word classes in syntax and grammar.

Compounding in Kalapalo, a Southern Cariban language

distinguished by semantic versus copula verb stem or nominalization, word order,

and the treatment of core arguments by prefixed and suffixed inflections. There
are both valence increasing constructions, including several types of applicatives,
and valence decreasing constructions that include the agent-backgrounding deergative prefix originally described for K
uikuro by Franchetto (1990). The most
common alignments2 are a) plain intransitive; b) extended intransitive; c) transitive (as with indicatives) with ergative marking of A and absolutive marking
of P; d) transitive (as with imperatives) with S nominative marking of controlling arguments (as pronominal prefix) and accusative marking of O (as full noun
phrase or pronominal prefix); e) de-verbal nominalizations; f) copular be verb
constructions; g) the valence decreasing de-ergative; h) thetic non-finite. All
finite verb stems are inflected with an absolutive personal prefix (this may be zero
for third person) and an indicative, imperative, or irrealis (hypothetical) suffix.
There is no tense marking morphology; both verbs and nouns show aspect marking but the morphological features are generally different for each class. Absolutive argument marking prefixes and clitics may be used with members of more
than one word class.
2.1 Derivational formation involving verb stems
In this section, I begin with a brief description of how basic verb stems may be
formed from a proper noun attached to a valency specifying verbal radical suffix. In these cases, rather than a pre-categorial root that may occur in a verbal,
nominal, or adverbial stem, a nominal element functions as a root for a basic verb
stem. Examples (1)(2) illustrate the formation of intransitive and transitive verbal stems as well as a few of the many inflectional possibilities such stems may
host. A common stem derivation involves the suffixing of the transitivizer -te to
a proper noun. The stem is then inflected with a required argument prefix and an
aspectual suffix. The transitive verb stem derived by -te in example (1)b (boldfaced) is inflected with the zero third person absolutive prefix and the continuous
indicative suffix -ga, and is followed by the ergative argument. The final ergative
construction with pronominal prefix conveys the nominal agentive argument.
The taxis or phasal clitic =lefa mt indicates that there is a part-whole relation
between the event described in this clause (1)b and that of the earlier clause in

2. I use the following terms for syntactic arguments in these descriptions: S for the single or
controlling non-ergative argument; A for the controlling ergative marked argument; P for the
absolutive argument controlled by A; O for the absolutive argument controlled by S.

Ellen B. Basso

(1)a, which describes the sun moving to the west, spoken while shown with an
arm movement.3



inde Giti ata-ni,

here sun eqa-ant
As the sun moved over this way,


3-skewer-pau-tr-cont 3-erg-pl=mt
they skewered a few pieces, one by one.

De-verbal nominalization is particularly important in Kalapalo as it is often used

in subordination. Example (2) shows a construction where the verb derived by
-te has been renominalized with an aspectual nominal suffix, -pg, end result of
a voluntary process (vpe). The polyfunctional ergative marker -feke is used as an
applicative here in regards to a geographic setting that controls the action of the
main participant.
ifaki-la aa-feke is-uw i-dya-te-pg
manner=mt far-neg path-erg 3-father 3-rope-tr-vpe
And this way, not far from the path where her father had hung his

Note that there is a metaphorical usage in the construction, because 3-dya his
hammock rope forms the root of the derived verb.
3. Verb stem compounding: the formation of extended verb stems
There are some uncommon instances in the data of noun incorporation, the combination of a noun and a basic verb stem to form a larger or extended verb stem.
Following Mithun (2010:37), I consider these instances of noun incorporation to
be compounds, because like Mithuns examples, they differ in several ways from
single NP bases and derivational constructions as in (2). First, there are omissions
of expected inflections that would usually occur with a nominal non-compounded
argument. Second, some kinds of phonological changes (which do not occur outside compounding) occur in the nominal element. Third, the nouns are no longer

3. The following is a list of phonological symbols used in this paper; where my symbols are
different from IPA conventions, the latter are placed in brackets. Consonants: p, t, k, kw, d, g,
dy [dy], mb [mb], G (voiced uvular flap), f [], s, h, ts, (voiced velar affricate), m, n, [], , l,
w, y. Vowels: e, i, [], a, u, o [].

Compounding in Kalapalo, a Southern Cariban language

syntactic arguments and some may have metaphorical functions distinctive to the
verbal construction. Example (3) illustrates such a use of a nominal root prefixed
to an underived verb stem, to form an extended verb stem.


2>1-head-uncover-tr-imp= imm
May you uncover my head right away
(i.e. help me take off my carrying basket).

Example (3) illustrates both morphological and phonological changes from the
nominal word that make the use of the morpheme as a noun different from
the compounded use. In this construction, the proper noun uGitugu my head
becomes the prefixed u-ki- (2>1-head) in the verbalizing imperative construction.
The uvular flap /G/ becomes a velar stop /k/. Also, the derivational nominalizer
suffix -tu and the possessive inflectional suffix -gu that would occur in the nominal
word are omitted. The resultant extended verb stem is suffixed with the imperative
suffix -ke imp and the immediate taxis suffix =fofo imm.
Another example of an extended verb stem compound (boldfaced) is (4), in
which a body part noun is used with a verb root -fi, blow a spell on. Here too a
normally required possessive suffix is omitted.


lepe tindis
lepe t-indis
next refl-daughter refl-hands-blow.spell.on-tr-cont
ifeke igi kefegesuki
i-feke igi kefege-su-ki
3-erg bee spell-poss-ins
Next she blew all over her daughters hands that little black bees spell.

Examples (5) and (6) show the use of affective verb stems with the malefactive suffix -al. In example (5) -al is suffixed to the transitive verb stem formed from
an extended (noun+noun) root (igi-fes head-ugly); the resultant transitive verbal
morpheme is used to express the idea of bother, or trouble someone.


e-mugu igi-fes-i-al
2-son head-ugly-ep-mal 1-erg
I never bothered your son.

In example (6), -al is used on the causative transitive verb faa-tu-Gi- insult
(lit. ear-hurt-cause, make ones ear hurt) incorporating the noun faa, ear +
adverb tu- hurt. Here the malefactive has an applicative effect in adding a second predicate argument (bracketed) to the victim of the insult (marked by the first
person prefix u- on the verb) with the ergative-marked A.

Ellen B. Basso


3-similar-prec-unlike=negq=pej adem
Youll soon realize this is nothing at all like that stuff
2-older.sister-husband-erg 1-ear-hurt-caus-mal
with which your older sisters husband insults me.

Finally, in (7), there are two copular clauses with the same plural copular subjects, headed by the existential verb its- exist (EX) in continuous indicative
inflection.4 The second clause (in square brackets) begins with an extended
noun copular complement (boldfaced). This clause is an instrumental applicative construction using the -ki instrumental suffix (ins), with connotation of
fault. The instrumental suffix appears first on the complement and then on the
final argument NP; as this is a copula clause, there is no ergative marking on
the argument.5


ui-u-fg its-a
u-dyi-mo ex-cont 1-child-coll
My children who had been lying,
snake-refl-stomach-bad-appl ex-cont 3-child-coll-ins
have given me a stomach ache, the fault of the children.

To review the stem compounding effects shown in these examples of noun incorporation, we see a number of features in extended stems that are associated with
compounding in Mithun (2000): a) phonological shifts; b) the absence of otherwise

4. In Kalapalo, there are four morphemes with copular functions belonging to two grammatical sets: the class inclusion copula suffix -i and the three be roots a- stative equative
(EQS), at- active equative (EQA), and its- existential (EX). While these morphemes are
copular insofar as they are able to link a copular complement to a copular subject, in addition
to their copular functions copular roots have other non-copular uses. The three roots can be
inflected as finite intransitive verbs, with one S argument, using indicative aspectual or irrealis
verbal suffixes; with these inflections there is no linking of a copular complement to a copular
subject. Additionally, when in intransitive or copular verbal construction these roots can also
become nominalized or adverbialized with derivational suffixes.
5. The claim of applicativation here is provisional as there are only three examples in the
data of the use of ins in this way. In such constructions, there is no ergative marking on the
argument, which hosts the instrumental suffix. There are similar grammatical asymmetries
between base and applicative object (see Example 10).

Compounding in Kalapalo, a Southern Cariban language

expected nominal inflections (such as possessive suffixes and possessor prefixes)

and derivational (word type radical) suffixes; c) extended nominal (noun+noun)
roots that precede the verbal base (Examples (7), (8)); d) body part nouns serving
metaphorically as instruments of feelings as in Examples (6)(7).

4. The inventory and use of clitics and particles

4.1 Definition and types of clitics
In Kalapalo there is a rich variety of clitics, while a smaller subset of morphological identical particles appear in unbound form.6 These clitics generally take
the second position after the first non-permeable constituent, which is normally a nominal or verbal construction. There are four semantic sets of enclitics containing numerous members (see Table 1 for examples from these sets):
1)affective (8members), 2) epistemic (27 members), 3) taxis7 or phasal markers
(14 members), and 4) interclausal reference markers (IRM) with clause coordination functions (17 members). Insofar as Kalapalo IRM interact with taxis and
adverbial operators and often have discourse and clause chaining functions added
to argument references, they function more than to coordinate or switch argument reference. A different term is needed that will capture these clause chain level
features, hence I follow Sparing-Chavez (1998) in her use of the label interclausal
reference marking (IRM). See Tables 1 through 3.

6. Like clitics, particles are postposed to adverbial, nominal, or verbal constructions and do
not host any inflectional or derivational features. The term particle seems justified by two
features that distinguish them from clitic status. First is the fact that these morphemes sometimes show stress on the first syllable. This is especially true of the class of affective particles;
those of one syllable have strong stress and vocalic extension. This does not, however, change
the phonological shape of the preceding phonological word. Nonetheless although they are
independent words, particles are still in a constrained position, following the immediately
preceding grammatical word. In this regard, they are very similar to clitics. The second distinguishing feature is that particles may appear several times in an utterance, following several
different (and different kinds of) clausal constituents.
7. Following Roman Jakobson, I use the Greek term taxis for markers of phasal relations
between events that characterize the narrated event in relation to another narrated event and
without reference to the speech event. (1971:135). Cognate forms have been described for
Northern Cariban languages (Carlin 2004; Derbyshire 1985, 1999; Hoff 1990; Meira 1999).

Ellen B. Basso

Table 1. Examples of epistemic clitics and particles




A. Focus on object, speaker is inside the description of object or is foregrounded

1. wke


Strong, assertive

3. nika



5. ma


Knowledge lacking,

usually as particle,
distant past
usually hosted by
question word or
negative adverbial

B. Focus on object, speaker is outside the description of object or not foregrounded

8. apa


Strong assertive

10. fna


Possible but weak

11. koh


Knowledge lacking

usually as particle
sometimes as particle

C. Focus on person in subjectivity

12. tafa


1st person confirmation


13. kafa


2nd person asked to

align, assertive

14. kato


Speaker aligns with

3rd person, weak

15. tiki


Speaker denies alignment sometimes as particle

D. Focus on changing subjectivity; correctives: speaker volunteers alignment with listener

16. maki


Rejection, doubt
becomes agreement
with proposition

17. makina


becomes agreement with
negative proposition

E. Contraspective
19. =kii;= kii-ni (plural)

Negative (regretful)

20. mukwe


Positive (hopeful,
wishful), in vain

sometimes as particle

F. Intersubjectivity: Alignments; appeals to another with grades of confidence

21. aka: Existing positive


Asserted, indicative

25. nipa: Possible alignment EM25

Distant past, imperative

when in dispute

27. nifa: Necessary

alignment, must occur


Imperative mood

28. (p)apa: encouraging



Imperative, hortative

often as particle

Compounding in Kalapalo, a Southern Cariban language

Table 2. Examples of taxis marking in Kalapalo clauses

Taxis clitic


Implicit S/A Explicit phasal referent; other semantic function









coordinate, part-whole; cause-effect, completive

metonymic relations between events; very high
frequencyof use; ends C or (if present D) level chain







repeated, persistive still, also

=ale clauses



uninterrupted, perpetual, always; high frequency of use




new, overlapping; clause chain boundary




immediately following, urgent


again same

subsequent recurrent event

=tale clauses dt


unexpected, contrastive or special event





*does not mark S/A referent as same or different.

Table 3. Affective clitics and particles (full set)






pejorative, compulsive



evitative, avoid



negative mirative



neutral or positive mirative






inappropriate, wrong







Clitics are one to three-syllable non-stressed or lightly stressed grammatical words with no morphological categories of their own, that is, no inflection
or derivation is possible. Exceptions to this rule are the (very occasional) presence on some affective clitics of the class inclusion copular suffix -i cop, the
unique suffix -a u and the primary topical participant marker -fa ptp, any of
which must always appear in final position. Kalapalo clitics never stand alone
and must be hosted by one of a variety of possible syntactic units. All of the latter constitute a phonological word in themselves to which the clitic is attached
without changing the stress pattern of the word. Some are fixed position clitics
insofar as they must occur as the final clitic in a group (the taxis =(h)fale no

Ellen B. Basso

and epistemic wke particle em1). Kalapalo IRM clitics manifest syntactic word
functions as they have switch-reference clausal argument marking functions,
and often reference larger clause chain syntactic units involving subordinate
clauses and multiple arguments. This is especially true of the IRM compounds
described in Sections 7 and 8.
4.2 Functions
Each of the four semantic sub-sets of clitics manifest compounding, involving
either two or more members of their own clitic sub-type or compounded with
members of other sub-types. Affective (1) and epistemic (2) clitics are basically
distinguished semantically as morphemes that are highly relevant to stance practices, appearing frequently in descriptions of conversations between key figures
in historical narratives (Basso 1995). Taxis (3) morphemes normally involve one
kind of phasal marking of clauses within reference clause chains. These clitics
bind with various clause constituents, are important for semantic linking between
dominant and subordinate clauses, and play an important role in clause chaining
processes. The most frequent taxis morphemes are the clitics =lefa metonymic or
contingent event (mt); and =(h)fale new, overlapping event (no); and the word
gehale, recurrent. IRM (4) also have clause coordinating functions and often compound with (3) as well as adverbial morphs.
In Example (8), I illustrate the use of two different, but still uncompounded
clitics, the epistemic clitic =maki (em16) and the metonymic taxis clitic =lefa (mt).
The epistemic morpheme indexes a kind of correction in which the speaker reevaluates a previous uncertainty or denial utterance, which at the time of utterance, is recognized to be correct. The speaker in this example realizes that there is
actually a river in a certain place, something he had been uncertain about earlier.
In this example, =maki is in second position, following the location demonstrative ande; the epistemic morpheme has scope over the following noun
phrase, fanguiga=lefa river=mt.


fanguiga=lefa -ng
i-feke. river=mt 3-erg
Ok, now I see that the river is here after all, he said to (him).

Example (9) is taken from a historical narrative, in which a man reminds his
younger brother of an event involving a third relative, which took place when the
brothers were children.


wke] ku-piano-feke
wke ukw-oto
refl-see-ep-neg=em3 em1 em1 dual-parent

Compounding in Kalapalo, a Southern Cariban language

3-head-sal 3-throw away-vpe em1

Can it be that you dont remember what happened long ago when our
older brother threw away our fathers head?

In Example (9), the speaker uses wke in the initial consideration complement
clause (in square brackets), in the ergative argument NP, and in the final main clause.
The particle, repeated several times, marks strong confidence in first-hand evidence
from the past that no longer exists. Secondly, the initial complement clause features
the epistemic clitic =nika appealing to the listeners experience. Like the first em,
the scope of =nika actually falls over the entire utterance. But so does the repeated
use of the second em, wke, which links the consideration complement construction to the first-hand but no longer existing visual evidence in the main clause.
Third, the speaker emphasizes the transitive subject (A) NP by left-displacing it
inthe main clause (the structure would normally be P+V+A).
A few epistemic clitics are free morphemes, attaching themselves to various
different kinds of constituents and to more than one phrase within a single utterance, and in some pragmatically defined instances, even to every clause and NP in
the utterance. In a few cases, an enclitic may have a non-clitic particle alternant,
which acts as an independent word. The most interesting member of this set is
wke (em1) which has both epistemic and temporal reference, marking strong and
justifiable conviction about an event occurring in the distant past. Considering
the syntactic functions of wke, in all examples, the particle (like other em) always
modifies the first impermeable (phrasal) constituent. However, in many especially
strong assertions (as in contexts of anger, shock or disappointment) wke appears
at the end of each phrasal constituent. In this regard, the syntactic functions of
such repeated scoping seem to suggest that wke must have an important function
in an affective style. The syntactic, grammatical and semantic functions of wke
are thus more complex than most of the other epistemic morphemes
Turning to the interclausal reference markers (IRM), these may also have
important switch-reference function in their clause chain expressions. Example
(10) illustrates the use of the switch-reference IRM =dye same argument (sa),
which makes clear the predicates in the reference and marking clauses have the
same argument.



atapa atapa]
bee-exist-cont-ptp 3-grandfather bee bee.
Their grandfather was Bee. Bee, Bee.


bee.hive=sa-exist-cont yes
Who lives in a hive. Yes.

Ellen B. Basso


[uele i-pefe-g-ts
ana 3-pollen-poss-mod store.up-cont 3-erg
Those are the pollen caches that he stores up.

The marked clause precedes the reference clause in Example (10). Here the nominal predicate is the same as the copula argument and final transitive argument
(A). Atapa, bee (a species word used as a proper name here), is first introduced as
Sof the equative copular verb (10)a, and the speaker uses =dye same actor in the
relative clause that follows, which causes anticipation of Bees agentivity in (10)c
with regard to the same event of living in a hive. The reference clause (in (10)c)
immediately follows the marked clause in (10)b.
Example (11) illustrates the use of the simultaneous event clitic =fata sim in
(11)b which occurs hosted by a verbal construct describing an event in the marked
clause. The clitic references the relation between this marked event and another
unexpected event occurring in a preceding clause (marked by the IRM =dya da
in (11)a).



t=ge-pa=dya -ka-i-tsu-
refl-it-on=da groan-ep-intr-res refl-shoot-ps eqa-ant
Having been repeatedly shot at, it was groaning


3-groan-intr=sim groan-intr=sim
it kept groaning, it kept groaning.

5. Clitic clustering across semantic sets

In this section I turn to clitic clustering, or the use of more than one clitic or
particle belonging to different semantic sets of clitics. Where particles are present,
they typically occur at the end of such a grouping. Affective clitics may cluster
with IRM types (12) or em morphemes (13). Epistemic clitics may also cluster
together, as shown in Example (15). Taxis clustering is discussed in the following separate section, as there are more complicated uses of this construction in
embedded clauses.
Example (12) involves the affective particle nile beware which expresses the
sense of needing to avoid something because it is inappropriate or dangerous. The
example shows the affective particle following the interclausal reference marking
clitic =mbe same event (se).


oko=mbe nile
exp=se beware
Beware of that (event) and avoid it.

Compounding in Kalapalo, a Southern Cariban language

In some complex constructions, more than one epistemic morpheme is used, as

for example in (12), where the contra-spective =mukwe hope in vain, em20 is
followed by =nifa, the strong requirement for alignment, em27 in an adverbial
construction beginning with taloki, useless, no reason. The locution is part of a
ceremonial speech addressed to ritual messengers (eoku).


ah taloki=mukwe=nifa timbe-a-ke
exp useless=em20=em27 arrive-caus-imp messenger(s)
Surely you must agree theres no reason in vain, enable your arrival

Finally, as Example (14) illustrates, as many as two different clitics may occur with
wke. A woman accuses her husband of killing her mother in order to use body
parts as decorations. This example includes the mirative =seku, =apa strong
probability, em8 and wke strong conviction about distant past event, em1.


ege=seku=apa wke ukw-oto

wke tis-ia
you=mir=em8 em1 dual-parent kill-nmlz.neg em1 1+3-ben/dat
uege wke
you em1

I was shocked to realize that the person who was the killer of our parent,
we figured out it was you.

6. Taxis morpheme clustering

Taxis morphemes and interclausal reference markers often occur together. Members
of no more than three (or in the case of the presence of interclausal reference markers,
four) different enclitic and particle sub-classes may follow upon other word classes.
Where combinations of clitics and particles complement adverbials, their scope covers the clause. With phasal and interclausal reference markers, the scope encompasses a partial or full clause chain, depending upon the morpheme in question.
Taxis clitics may co-occur if there is semantic embedding, that is, when an
IRM marked clause describes an event that is part of another larger event. In
Example (15), a taxis cluster is hosted by the main clause ergative argument. This
grouping consists of the different taxis clitic dt =tale followed by the new overlapping event taxis marker =fale no with a following demonstrative that marks
this NP as the argument of the following verbal clause.

angikogo-feke=tale=fale fegei i-fu-pte-ga-ko

fierce.people-erg=dt=no adem 3-run-tr-cont-pl again
But this time it was from the fierce people that they were running away
from again.

Ellen B. Basso

Note that the ergativity in this example references a causal stimulus, but not an
active agent though the morpheme may of course do so in other instances. The
dt marker here makes reference to events in the immediately prior chain.
In the rhetorical, yes-no question Example (16), both inde, here, a local
adverbial demonstrative, and the following main verb are cliticized with the fixed
position =hale new overlapping, =no taxis marker which always occurs at the
end of a cluster. The habitual taxis morpheme =tiga hab precedes no and reappears on the following subordinate verb clause (bracketed).


inde=hale ti-tsa=tiga=hale
rq=em5 here=no refl-exist=hab=no 3-tell-cont=hab 2-ben/dat-pl
Could it be that someone who stayed here came from time to time wanting
to talk with you?

In Example (17)b, the metonymic taxis marker =lefa links a negative adverbial
adjunct and the main clause that follows. The new, overlapping event clitic =hale
no also occurs in (17)b.



ige-funde i=ge,
bottle.gourd pdem-adv think.see-it

b. ial=lefa=hale

no.longer=mt=no dis-throw-des-ep=mt 1+3-ben/dat

Draped around the bottle gourd, but now, consider that its not any
longer (used) because weve decided to throw it away.

A speaker often uses taxis clitics to link clauses in a chain with only one finite verb,
appearing at the end of the chain. In Example (18), the chain illustrates how taxis
clitics can cluster and supplement one another.


a. [i-dyo-gu-ko

3-uncle-poss-pl 3-exist=no
[tsiu tsiu bok tsiuk]

(sounds of repeated cutting and throwing down of items)

But their uncle being different (cutting them down slowly),



i-tsa=fale la=gele=fale]
like.that 3-ex=no like.that=r=no
still (working) differently (from the others) like that,
negative 3-know.about-cons 3-erg
(since) he didnt know about it.

Compounding in Kalapalo, a Southern Cariban language

The speaker is contrasting the slow work of the untrained uncle (an ordinary
person and the villain in the story) with the quick and efficient harvesting of
arrow cane by warriors, who need to be always ready to enter into battle with
enemies. In (18) there are three different taxis clitics, the new, sequential event
taxis form =fale no being repeated three times. This form references the uncles
antithetical act of slowly (not quickly, as do his skillful relatives) cutting down the
cane. In (18)b, there is a compounded clitic; =gele marks the repeated action R of
the uncle cutting arrow cane, while the =fale new overlapping taxis (no) prepares
us for the new finite verb (the stative verb -fu- know.about) at the end of the chain
(18)c: ial funm=i-feke, he didnt know about it. (The uncle as actor is ignorant
of something contextually important.)
7. Interclausal reference marking compounds
Related to the clitic clusters discussed above, but more clearly operating as compounds with special syntactic and semantic functions, is a sub-set of the interclausal reference markers. This subset includes switch-reference markers and
operates as clausal coordinators involving both syntactic argument and semantic
function. There are five semantic parameters (one graded, the others contrastive)
referenced by these Kalapalo IRM: 1) same-different; 2) anticipated-unexpected
outcomes; 3) unique-repeated (with three grades of repetitive types including
iterative, repeated and recurrent); 4) separate-overlapping; 5) partial-totality. In
this section I show that clitic clustering enables this semantic flexibility and that
Kalapalo switch-reference operates in more ways than just the marking of same
or different subject, an argument in much recent work on the topic (Huang 2000;
Jacobsen 1983; Mithun 1993; Stirling 1993; Pustet 2013; Watkins 1993).
The compound IRM (called Set 2) are constructed from the four basic
switch-reference forms (called Set 1) together with a taxis (phasal) clitic and/or
an adverbial suffix. As the resultant compounds are particularly important in having clausal coordination functions in extended narrative discourse they are rarely
if ever heard in ordinary conversation. Table 4a (see Appendix) shows the full
Set1 and Table 4b some examples of the many Set 2 morphemes I have found in a
collection of 90 narratives recorded by ten speakers. The tables provide a summary
orientation to the important clause coordinating syntactic and semantic features
of these morphemes which often coordinate larger discourse units.
As shown in Tables 4a and 4b the foundation of this IRM system is the set of
four switch-reference clitics in two contrasting pairs (Set 1). These operators reference cross-clausal continuities or discontinuities of arguments and predicates,
coordinating preceding and following clauses most often in a longer narrative
chain. While these four clitics have some deictic functions, they do not have

Ellen B. Basso

underlying full forms (pronouns or demonstratives). The first contrastive pair

has event marking functions that fall into the first (same-different) parameter:
same event (=mbe se) or different event (=mba de). The second contrastive
pair references resultative argument continuity (=dye same argument, sa) or
discontinuity (=dya different argument, da). Included in Set 2 is a related pair
of compounds of two of these otherwise independent morphemes, referencing
same syntactic argument, same prior event (=mbedye, or different syntactic argument, same, successive event (=mbedya se.da).
Other members of Set 2 are compounded of the components from Set 1 and
morphemes which are otherwise independent. These compounds all begin with
=mbe or =mba, which host one taxis clitic or an adverbial suffix. With some members of this set, =dye or =dya ends the construction. These Set 2 constructions
have clause coordinating functions involving switch-referencing of arguments of
the reference and marking clauses (the latter bear the IRM morpheme, as well as
semantic references falling into one or more of the five semantic parameters).
Kalapalo marking clauses include almost any kind of syntactically dependent,
complement clause, including non-finite subordinate clauses, intransitive or transitive clauses, quotative clauses, quotations, ideophonic clauses, de-ergative agent
backgrounding clauses, as well as the narrative segment marking anaphoric topic
referent ana clause. An IRM may also occur on nominal or adverbial adjuncts,
and hortative and lexical clausal complements, de-verbal nominalizations, and
nominalized clefted phrases derived from verbs. The latter also often carry their
own event coordination taxis clitic (usually =lefa, but also =(h)fale). This suggests
that, as with epistemic and affective clitics, Kalapalo IRMs have broad, even interclausal scope regardless of which constituent is their host.
Example (19) is taken from a story in which a shaman comes to cure a jaguar
who has been shot with a warriors arrow no one else can see. In this pair of clause
chains there are three IRM. In (19b), the compound =mbedya, same event, different argument, se.da (Parameter 1) references the argument of events described
in (19a) as different from those of (19b). In (19c), =mba, different argument, da
references a new event with an unexpected outcome marked in (19d) by =mbale,
dres (Parameter 2). The latter is a compound showing regressive vocalic elision.
It involves de+ut, the basic =mba form different event (de) together with the
uninterrupted, perpetual taxis clitic =ale ut. Scope of this compound falls over
lines ((19)c and d), a three clause chain.
iil ifeke



i-feke tiki
3-see-pncl 3-erg (sound of entry)
He looked at him, tiki: as (the shaman) came inside.

Compounding in Kalapalo, a Southern Cariban language


inde=mbedya-fa igei ata-ni

here=se.da.ptp idem eqa-ant ex-cont-cop
Someone else is here now to work on whats happening (to him).


-ii-l=mba=lefa=hale t-iu-gu-ki=lefa,
3-see-pncl=de=mt=no refl-eye-poss-ins=mt
He told him that he saw it, with his own eyes.

d. ai=mbale a-ng ai a-ng.

exist=dres eqs-ps exist eqs-ps

Theres something here (as a result of what happened), theres
something here.

Example (20)c illustrates the use of another resultative, =mbele the justified or
expected permanent effect, eres (Parameter 2). This compound involves se+ut
and contrasts with the unanticipated remaining =mbale dres. In (20), =mbele
marks the final outcome involving the undergoer (the dead mans arm). There are
two reference clauses in this example. The first clause in (20)a describes people
who are trying to eat a dead mans arm. In (20)b the contrastive resultative compound =mbekudya logn but for that same reason (Parameter 2) occurs. Also, a
deontic change occurs from the reference clause (20)a to the marked clause (20)b;
people found the arm was too bitter to eat, even though they thought they could
do so and so insist that it be thrown away.



lepe -tee-fo-
i-feke-ne, tatsi-i
next 3-eat.meat-hyp-res 3-erg-pl bitter-nmlz.neg completely
Next they tried to eat it, but it was bitter as could be.



aft=mbekudya ake-tsae ila=mbe

deo.s-deo manner=se ex-ant
But for that reason, we should leave it be,


thrown away for good.

Example (21) illustrates use of IRM =mbekudye logp, that was indeed why. This
use contrasts with that of =mbekudya logn, but for that reason seen in (21). The
pair are two of only a few attested compounds involving an adverb, in this case the
-ku intensive, int suffix.



refl-approach-th refl-face-on-adv-pl
Coming closer, peering closely at them,

Ellen B. Basso


the he ali=mbe
nice negation=se prickle-adv-pncl-cop
Beautiful, (what he saw) it wasnt something prickly.


teh heh heh=kugu=mbekudye

Because of that they were all very nice indeed!

(21) is a good example of clause-skipping. Observe the marking clause in (21)c.

The first three syllables constitute an ideophone referring to the beautiful, clean
appearance of the grass. This ideophone is followed by -kugu, the intensive adverbial suffix, and finally the IRM =mbekudye. The whole sentence is uttered in an
elevated voice by the narrator. He wants his listeners to know how excited and
happy people were to see thatching grass for the first time. As the listeners have
learned from the reference clause in (21)b, (what he saw) wasnt something
prickly. Note that the only explicit verbal construction in this chain is the initial
non-finite thetic alignment in (21)a.
Example (22) shows the use of the iterative resultative compound =mbegedya
despite it all, nevertheless, With this IRM, the consequence of the iterated event described in the preceding clause is marked as unexpected or unwanted.
The semantic parameters are same-different (1) and expected-unexpected (2).



i-feke 3-erg
Even though he kept trying to hold on to him he couldnt do it.


i-feke 3-erg
Even though he kept on squeezing him, he couldnt do it.


(sound of pushing away) throw-pncl=mt 3-erg

Afasa throw-pncl=mt
Bok, he threw him away, he threw Afasa away.

As in the example, =mbe is often followed by the iterative taxis =ge. Less commonly =mbe and =dya are linked with this same taxis marker, when =dya leaps
over =ge. Example (22) shows both phenomena in (22)a and (22)b, as well as an
additional new overlapping event (=hale taxis no). The finite transitive reference clause suffixed with =lefa mt appears in (22)c at the end of the chain.
In Example (23) there is a more complex iterative compound =mbembegedya,
pda referencing a negative consequence of many iterated actions that have more
than one component, with the sense of: but even after doing all that.

Compounding in Kalapalo, a Southern Cariban language


a. i-i-lu-ia=mbege
ete-al ben/ go.away-rep

Hei would go away to look again for himj, hei kept doing (the
b. s-i-nal=mbembegedya=fale
(and even after doing that over and over), this time when hej came
tohimi (this time the result was),

i-feke=lefa Taugi
3-weaken-soft-caus-rev 3-erg=mt Taugi

3-weaken-soft-caus-rev 3-erg=mt

hej had made himj weak, hej made Taugi weak when hej tried to take
hold of him.

The IRM in Example (23) is one of two Set 2 compounds in which mbe is reduplicated (the other is mbembege, perpetuity, again and again, pe). In (23) the main
character (Taugi) is trying to draw his childs killer to him with a charm, a pot of
boiling water containing some of the boys hair. But even so, when he tries to take
hold of him, Taugi is weakened by a spell his enemy has put on his hands.8
8. Discussion of Compounded IRM (Set 2)
The compounded structures in Set 2 illustrate a number of features described
by Mithun (2000) and Aikhenvald (2002). First, these are multi-clitic strings
(Aikhenvald 2002:43, 52) or chunks (as Mithun (2000) describes them). These
are processed as a single unit yet manifest semantic independence, but the compounds are not necessarily equivalent in meaning to the sum of the meanings of
their parts (Mithun 2000:39). Second, in some cases there is also greater phonological fusion (Mithun 2000:40) than in instances of suffix sequencing. As
Aikhenvald (2002:49) notes regarding the characteristics of clitical phonological
words in various languages, in Kalapalo there are several unusual features associated with clitical compounds. The compounds mbele and mbale illustrate (apparently uniquely) the process of regressive vocalic elision. The compounded features
of =mbele and =mbale are not as clearly represented in the syllabic structure due to
these vocalic elision processes. In addition, an occasional feature of theKalapalo
IRM compounds is reduplication of the initial =mbe se which occurs in two
8. Subscripts i and j reference the two different third person arguments.

Ellen B. Basso

instances (see Table 4a and Example (23)). Reduplication is an unusual process

in the language and it is particularly interesting to see it with a switch-reference
While there are a few IRM that are clitic-only words (Aikhenvald 2002:51),
Kalapalo interclausal reference markers interact with other morphology. The ones
used in this paper as examples are the =ale unending, =f(h)ale, new, overlapping, =tale, different, =fata, simultaneous and =ge, repeated taxis clitics and the
intensive adverbial infix -ku- and the totality adverb -tufugu. These morphemes
take on special meanings as elements of IRM compounds or as complements to the
hosts of IRM. Kalapalo IRM clitic compounds are thus also grammatical words.
While they do not have morphological categories of their own, they do allow the
use of taxis, and adverbial suffixes in more common use with non-clitical words.
Specifically, these features are marking both the phasal quality of the event in relation to another event, repeated and intensive qualities, and also the consequence
of the events on participants, particularly undergoers. Most of these clustered clitics begin with =mbe se; in my data only one (mbale dres) begins with =mba
de. Two complex clitics =mbedye and =mbedya se.da compound two
of the Type I basic forms. Compounds mbekudye logp and mbekudya logn
compound two basic forms with the linking adverbial suffix -ku intensive, int.
Forms =mbembege pe, =mbembegedya pde, =mbele eres, and =mbale dres are
compounds of =mbe or less frequently =mba forms with taxis markers; the forms
of the =mbembege group reference a scalar set of repeated subjects.
Considering IRM compounds from the perspective of recent linguistic study
of switch-reference, resultative semantic functions that can be subsumed under
Stirlings (1993) idea of continuity of eventuality must be taken into account.
These functions involve consequences that are expected or anticipated, and also
unexpected results involving locational, temporal, epistemic, and deontic features,
as well as switch- reference continuities or discontinuities which explain a Kalapalo speakers use of Set 1 forms. Some of these compounds reference special logical cause-effect, result, and reason relationships among chained clauses, similar in
function to conjunctions like English because, if , so, and, and but. Both positive and negative results are represented.
At this level of syllabic complication, it is possible to see a gradual transformation from clitic to lexeme. It is not a complete transformation yet. The switchreference function is preserved by the Set 1 operator used in the compound, but
switch-reference is attenuated in the presence of a semantic function. It is not so
unusual that a marking clause operator can have this function (see Huang 2000
for references; also discussion of lexical developments from other linguistic elements in Hopper & Traugott 2003:140211) but it is interesting that in Kalapalo,
a language with few adjectives and many derivational processes, switch-reference
operators have this secondary adverbial-like function. This is helped by the very

Compounding in Kalapalo, a Southern Cariban language

commonly used ideophone words. As in other Amazonian languages, complex

ideophones readily substitute for Aktionsart features (see Nuckolls 1996). Kalapalo descriptive properties are most commonly found in noun phrases. But since
many nouns are actually derivations from verbs, could there be a connection
between the adverbial-like tendencies of these compound IRMs?
9. Summary
Kalapalo shows stem incorporation and clitic compounding word formation
structures that involve the use of more than one phonological word. Extended
verb stemformation and clitical compounding are given special status, because
they differ syntactically, morphologically, and semantically from the numerous
inflectional and derivational processes in the language. The first type of compound discussed in Section 3 involves the formation of derived verbal stems
from nominal elements preceding the base stems. Kalapalo noun incorporation
in verbal compounding illustrates the many resultant morphological, phonological, and occasionally semantic features shown by Mithun (2000, 2010) to be
especially characteristic of compounding. A second type of compounding, syntactically related to verbal compounding, involves extended noun stem formation which involves phonological and morphological changes as well as special
semantic effects. Finally, clitic clustering involving IRM, the most highly productive form of compounding, has distinct syntactic and semantic functions in the
context of clause chaining and interclausal reference. The use of this latter type
of compounding provides further evidence for understanding how the products
of syntactic operations add to our knowledge of the morphology-syntax interface
discussed in Mithun (2010). Finally, the occurrence of Kalapalo IRM is hard to
predict from argument or predicate reference alone, as IRM interact with other
features. These include both positive and negative logical relations between events
(cause-effect, cumulative effect, iteration, repetition, and recurrence, reason, and
result); symmetric or asymmetric aspectual relations between clauses; as well as
epistemic, mirative, and deontic, as well as locational continuities or changes in
outcomes. These data support Stirlings proposal (1993:2) that interclausal continuity of eventuality must be assessed to account for the use of these systems.
Stirlings review of switch-reference in a large number of the worlds languages
demonstrates that switch-reference morphemes often have functions different
from argument reference. These functions are sometimes called secondary,
sometimes extended, suggesting that theoretically linguists consider the basic
function to be syntactic (Van Valin Jr. & La Polla 1997). However, Stirlings work
and that of Watkins (1993,) Mithun (1993) and most recently Pustet (2013) demonstrate that semantic features of interclausal relations are crucially important,

Ellen B. Basso

and indicate that switch-reference involves agreement or disagreement between

clauses (or even larger discourse units; Woodbury 1983). These linguists understand switch-reference as functioning along various parameters contributing to
discourse continuity. Kalapalo parameters include not only same-different contrasts, but also various kinds of contrasts of consequentiality. (For example, discontinuity of predicate or argument not only means difference but can also mean
unexpected.) Of much importance is the fact that the Kalapalo system manifests
these secondary non-syntactic semantic functions that have been the finding
most important for reconceptualizing the nature of switch-reference in the worlds
languages. Kalapalo data strongly support this expanded view of switch-reference
morphemes functioning in clausal cohesion. In the Kalapalo case, clitic clustering as a means of creating clausal coordinators plays an important role in this

2>1 = subject second person > object first person; 1 = first person; 1+2 = inclusive
we; 1+3 = exclusive we; 2 = second person; 3 = third person; A = transitive subject; adem = adnominal demonstrative; adv = adverbializer; ana = anaphoric;
ant = anticipation; appl = applicative; ben/dat = benefactive/dative; caus =
causative; coll = collective; cons = concessive; cont = continuous indicative;
cop = copula; da = different argument; de = different event; deo = deontic; deo.s
= deontic subject; des = desiderative; dis = dislocative; dres = unexpected result;
dt= different taxis; em = epistemic (see Table 1 for numbers 1, 3, 5, 8, 10, 11,
16,17, 20, 21, 25, 27, 28); ep = epenthetic; eqa = active equative; eqs = stative
equative; eres = anticipated result; erg = ergative; evit = evitative; ex = existential; exp = expletive; hab = habitual; hyp = hypothetical; idem = identificational
demonstrative; imm = immediate; imp = imperative; inc = inclusive adverb; ins
= instrumental; int = intensive; intr = intransitive suffix; it = iterative; logn =
negative logical association; logp = positive logical association; mal = malefactive; mir = mirative; mirn = negative mirative; mod = modifier; mt = metonymic
taxis; neg = negation; neg.em = epistemic negative (regretful); = plural regretful; negq = negation of existent qualities adverb; nmlz.neg = negative nominalizer; no = new overlapping; O = transitive object; P = patient; pa =
pathetic; pau = paucal; pda = repeated plural events leading to new result; pde=
repeated events leading to multiple consequences; pdem = personal demonstrative; pe = perpetuity; pej = pejorative; pl = plural; pncl = punctual indicative;
poss = possessive; prec = precise; ps = passing state; ptp = primary topical participant; r = repeated; rcont = same continued event, same argument; refl =
reflexive; rep = repetitive; res = resultative; rev = reversative; rq = rhetorical

Compounding in Kalapalo, a Southern Cariban language

question; S = intransitive subject; sa = same argument; sal = salient; se = same

event; = same event, new resultative location; se.da = same event, new or
different argument; = iterated event; = iterated event, new or different argument; = same event, same argument; se.tot = resultative totality;
sim = simultaneous; th = thetic; tr = transitive verb suffix; u = unique; ut =
uninterrupted; vpe = voluntary resultative nominal

Aikhenvald, Alexandra. 2002. Typological parameters for the study of clitics, with special reference to Tariana. In Word: A Cross-linguistic Typology, Robert M.W. Dixon & Alexandra Y.
Aikhenvald (eds), 4278. Cambridge: CUP.
Basso, Ellen B. 1995. The Last Cannibals. Austin TX: University of Texas Press.
Carlin, Eithne. 2004. A Grammar of Trio: A Cariban Language of Suriname [Duisburg Papers on
Research in Language and Culture 55]. Frankfurt: Peter Lang.
Derbyshire, Desmond. 1985. Hixkaryana and Linguistic Typology. Arlington TX: Summer Institute of Linguistics.
Derbyshire, Desmond. 1999. Carib. In The Amazonian Languages, Robert M.W. Dixon &
Alexandra Y. Aikhenvald (eds), 2361. Cambridge: CUP.
Franchetto, Bruna. 1990. Ergativity and nominativity in Kuikuro and other Carib languages. In
Amazonian Linguistics: Studies in Lowland South American Languages, Doris L. Payne (ed),
407427. Austin TX: University of Texas Press.
Franchetto, Bruna. 2006. Are Kuikuro roots lexical categories? In Polivalency of Roots, Ximena
Lois & Valentina Vapnarski (eds), 121158. Amsterdam: John Benjamins.
Franchetto, Bruna. 2010. The ergativity effect in Kuikuro (Southern Carib, Brazil). In Ergativity
in Amazonia [Typological Studies in Language 89], Spike Gildea & Francesc Queixals
(eds), 121158. Amsterdam: John Benjamins. DOI: 10.1075/tsl.89.06fra
Hoff, Berend J. 1990. The non-modal particles of the Carib language in Suriname and their
influence on constituent order. In Amazonian Linguistics: Studies in South American Languages, Doris L. Payne (ed), 495542. Austin TX: University of Texas Press.
Hopper, Paul J. & Traugott, Elizabeth Closs. 2003. Grammaticalization, 2nd edn. Cambridge:
CUP. DOI: 10.1017/CBO9781139165525
Huang, Yan. 2000. Anaphora: A Cross-linguistic Study. Oxford: OUP.
Jacobsen, Jr., William H. 1983. Typological and genetic notes on switch-reference systems in
North American Indian Languages. In Switch-reference and Universal Grammar [Typological Studies in Language 2], John Haiman & Pamela Munro (eds), 151184. Amsterdam:
John Benjamins. DOI: 10.1075/tsl.2.10jac
Jakobson, Roman. 1971. Shifters, verbal categories and the Russian verb. In Selected Writings,
Vol. 2, 130147. Berlin: Walter de Gruyter.
Meira, Sergio. 1999. A Grammar of Tiriyo. Unpublished PhD dissertation, Rice University.
Mithun, Marianne. 1993. Switch-reference: Clause combining in Central Pomo. International
Journal of American Linguistics 59: 11734. DOI: 10.1086/466192
Mithun, Marianne. 2000. The reordering of morphemes. In Reconstructing Gammar. Comparative Linguistics and Grammaticalization [Typological Studies in Language 43], Spike Gildea
(ed), 231255. Amsterdam: John Benjamins. DOI: 10.1075/tsl.43.09mit

Ellen B. Basso
Mithun, Marianne. 2010. Constraints on compounding and incorporation. In Cross-Disciplinary Issues in Compounding [Current Issues in Linguistic Theory 311], Sergio Scalise & Irene
Vogel (eds), 3756. Amsterdam: John Benjamins. DOI: 10.1075/cilt.311.05mit
Nuckolls, Janis. 1996. Sounds Like Life. Oxford: OUP.
Pustet, Regina. 2013. Switch-reference or coordination? A quantitative approach to clause linkage
in Lakota. International Journal of American Linguistics 79: 153188. DOI: 10.1086/669627
Sparing-Chvez, Margarethe W. 1998. Interclausal reference in Amahuaca. In Handbook of
Amazonian Languages, Vol. 4, Desmond C. Derbyshire & Gregory K. Pullum, (eds),
443485. Mouton de Gruyer.
Stirling, Lesley. 1993. Switch-reference and Discourse Representation. Cambridge: CUP. DOI:
Van Valin Jr., Robert & LaPolla, Randy J. 1997. Syntax: Structure, Meaning and Function.
Cambridge: CUP. DOI: 10.1017/CBO9781139166799
Watkins, Laurel J. 1993. The discourse functions of Kiowa shift-reference. International Journal
of American Linguistics 59: 137641. DOI: 10.1086/466193
Woodbury, Anthony C. 1983. Switch-reference, syntactic organization, and rhetorical structure in Central Yupik Eskimo. In Switch-reference and Universal Grammar [Typological
Studies in Language 2], John Haiman & Pamela Munro (eds), 291315.Amsterdam: John
Benjamins. DOI: 10.1075/tsl.2.16woo

Table 4a. Set 1 of interclausal reference marking structures, simple forms
Form Gloss

Position of marking Marking clause (MC) host

clause in relation to constituent in examples
reference clause (RC)

Semantic properties
and translations

=mbe se: same


following; may occur

across conversational

adverbial RC clause
continuity of
adjunct; copular S; adjective eventuality
complement of RC; S arg.

=mba de:


MC verb stem; quotative

discontinuity of
complement; copular S;
eventuality; different
quotative; adverbial adjunct subject; successive;


sa: same


Imperative complement
clause; existential be S

continuity of result;



copular S; oblique
argument of VP; adverbial
adjunct of EQS be verb;
free personal pronoun;
anaphoric pronominal
demonstrative; proper


Compounding in Kalapalo, a Southern Cariban language

Table 4b. Set 2 of interclausal reference compound marking structures



Position of marking clause Marking clause (MC) host

in relation to reference
constituent in examples
clause (RC)

Semantic properties and


=mbedye same event, same



following event and

preceding argument,
reference clauses

quotative, quotation VP,
ADV adjunct of
de-verbal nominalization;
extended argument of MV;
intransitive verb


se.da: same event,

different argument


following event and

preceding argument,
reference clauses

intransitive VP;
de-verbal nominalization;
MV intransitive clause;
existential verb.

over-lapping; successive

=mbege iterated event



transitive clause

repeated several times on

same occasion; even as/while

=mbegedya iterated event,

different argument



transitive clause

negative cause-effect
association of repeated
events, despite that, so that;


pe: repeated complex

(multi-featured) iterated


intransitive clause; S

perpetuity, again and again

=mbembegedya pda: repeated complex

iterated event, new
resultative argument

se~se+it+da preceding

interjection; intransitive

repeated events leading to

new result; and even after all
that happened again



expressive +adverb

positive, and for just that

reason; thats why

logp: same argument,

same inclusive event



Ellen B. Basso
Table 4b. (Continued)



Position of marking clause Marking clause (MC) host Semantic properties and
constituent in examples
in relation to reference
clause (RC)


logn: same event,

negative argument




rcont: same continued

event, same argument

se+cont +se following


eres: anticipated
resultative event



dres: unanticipated
resultative outcome


adverbial adjunct to NP or
verbal clause

negative, none of it; but for

just that reason continual;
still as before


transitive clause; MV
intransitive verb; adverbial
adjunct of interjection;
case marker; de-verbal

and in the end/after all of

that, as expected.


preceding; a clause chain


adverbial adjunct of A;

but as a result, unexpectedly same event, new

resultative location



de-verbal NP

subsequently; so afterwards

se.tot: resultative



lexical NP

happened/was done to all of


Nominalization in Choln
Astrid Alexander-Bakkerus
(University of Amsterdam)*

In Amerindian languages and in many other agglutinative languages,

subordination is often a matter of nominalization. In Choln, a language spoken
in North-Peru, this is certainly the case: nominalized forms coincide with
subordinate clauses. In this language, a nominalized verb form can also be used as
a main predicate. In this paper we study the different subordinate clauses that are
formed with nominalizations. We then find out which nominalizations are part of
a main predicate, and when this is the case.
Keywords: Choln; subordination; nominalized main predicates;
syntacticalnominalization; relative clauses

1. Introduction
The aim of this paper is to show that in Choln1 an agglutinative North Peruvian
SOV language with prefixes and suffixes nominalized forms can be used as subordinate clauses, and that a number of them can also be used as main predicates.
Traditionally, subordination is considered to be a construction in which two (or
more) clauses are combined with each other in such way that one clause (the subordinate clause) is a constituent of the other (the main clause) (van Gijn, Haude &
Muysken 2011:3). In Choln, subordination is mainly a matter of nominalization.
The language has one subordinating morpheme -pit, with which clauses can be
co-ordinated. The co-ordinator -pit can also be used to express concession, when
suffixed to an imperative or to a nominalized form. In addition to the conjunction
-pit, the language has two subordinators at its disposal, namely the suffixes -e
purposive and -hu switch reference, and one discourse marker, the morpheme

* I would like to thank the editors of this volume for their corrections, remarks and questions,
and I thank Willem Adelaar for his most useful comments and observations.
1. Choln belongs to a small language family, called Cholnan. Hbito is the other member.
Both languages are extinct.

Astrid Alexander-Bakkerus

-(w)a topic, to construct a subordinate clause. The subordinator -e (not to be

confounded with the nominalizer -(e) the fact that) is used to form a purposive
or a resultative clause; the morpheme -hu indicates that the subjects of the subordinate and the main clause are not the same person; the topic marker -(w)a can
also be attached to nominalized forms and it can be employed to construct a conditional clause. In all other cases, including the cases in which the co-ordinator -pit
and the topicalizer -(w)a are suffixed to a nominalized form, subordinate clauses
make use of nominalized forms.
Nominalizations are verb forms with nominal properties. Like nouns, they
can be followed by case markers, and they can function as subjects, direct complement clauses, nominal predicates, and adnominal modifiers. Lexical nominalizations behave like ordinary nouns: they can be pluralized and receive case marking,
but they cannot receive subject, object, tense and aspect marking. Choln nominalized forms, on the other hand, have some significant verbal properties and are
therefore rather syntactic. They can be marked for subject, object, tense, aspect
and negation, but they cannot be pluralized. The only nominal property that they
have, showing that the form at issue is a nominalization after all, is the property of
case marking. Syntactic nominalizations are in fact nominalized verb forms which
can function as the head (nominalized predicate) of a subordinate clause. Choln
nominalizations can be characterized as clauses in disguise.
Although Choln nominalizations cannot be pluralized, a number of them
can be noun-like, or, rather, more noun-phrase-like, such as the agentive nominalizations a-kole-wu (1sg.o-love-ag)2 the one who loves me = my lover and
tsap-u (3sg.o.catch-ag) the one who catches (fish) = the catcher (of fish); the
modifying nominalization ki--ikno-o (1pl.a-3sg.o-fear-fut.nmlz2) that
what is having to be feared by us = something frightening = frightening; the
objective nominalization -tsamo- (3sg.a-3sg.o.know-fac) the fact that he is
knowing something = wisdom; and the state nominalizations kot-u (be-ag) the
one who is = a human being and a-kot-lam (1sg.s-be-fut.nmlz1) my (future)
The difference between an English lexical nominalization, such as lover,
derived from the verb love, and a syntactic, noun-phrase-like, nominalization in
Choln can be illustrated by comparing the lexical form mentioned above with
syntactical, noun-phrase-like Choln forms. The canonical structure of the lexical

2. In this paper, the person cross-referencing markers are specified for their roles in the
glosses (e.g. 1sg.o), even though they are often homophonous at least the ones that mark the
speech act participants, and can mark the subject of an intransitive verb (s), the subject of a
transitive verb (a), the object of a transitive verb (o), and the possessor (poss).

Nominalization in Choln

construction is as follows: stem-nmlz(-pl)(-gen). Syntactic, noun phrase-like

nominalized forms can have the following structures in Choln:
(s-)stem-nmlz when the stem is intransitive;
(a-)o-stem-nmlz when the stem is transitive;
(a-)3sg.o.stem-nmlz when a third person object is indicated by means of an
initial stem alternation.
Besides being noun phrase-like, Choln nominalizations can also be clause-like.
Clause-like nominalizations can not only be marked for agent/subject and object,
like the noun-phrase-like ones, they can also be marked for aspect, tense and
negation, and they can be followed by case markers. Aspect and tense are only
marked when the nominalizing element is a demonstrative or the negation -pits-o.
A clause-like construction may thus have the following structures:
s-stem/3sg.s.stem-(aspect/tense)-(neg)nmlz(-cm) when the stem is intransitive;
a-o-stem-(aspect/tense)-(neg)nmlz(-cm) when the stem is transitive;
a-3sg.o.stem-(aspect/tense)-(neg)nmlz(-cm) when a third person object is
indicated by means of an initial stem alternation.
The fact that Choln nominalizations, occurring as the nucleus of subordinate
clauses, can be marked for tense, aspect, and person agreement, somewhat goes
counter to Cristofaros frequency hierarchy (Cristofaro 2003:289), according to
which the lack of TAM distinction and of person agreement are the most frequent
parameters for the morphosyntactic coding of subordination.
The concepts lexical nominalization, agentive nominalization, action nominalization, syntactic nominalization, noun-phrase-like, clause-like, and clauses
in disguise, stem from Muysken (1999:248252). The Choln examples are taken
from Alexander-Bakkerus 2005 and 2011.
In this paper we first discuss the different kinds of nominalizations, whether
or not followed by case markers that are used in subordinate clauses (Section 2).
Section 3 focuses on the cases in which a nominalization functions as the head of
a main predicate. Section 4 closes the paper with some concluding remarks.
2. Nominalized subordinate clauses
The Choln language has different morphemes that can nominalize a verb. All,
except for the agentive suffix -(w)u, can be followed by case markers. The nominalizations and the subordinate clauses formed without case marking are treated
in Section 2.1, those formed with case markers in Section 2.2.

Astrid Alexander-Bakkerus

2.1 Nominalized subordinates formed without case markers

Choln has the following nominalizers: -(e) factivizer and -(k)te infinitive
(Section 2.1.1), -(k)he simultaneity and -(n)ap sequentiality (Section 2.1.2),
-lam future nominalizer 1, -()o future nominalizer 2, the negative nominalizing suffixes -pakna and -pits-o (Section, and finally -(w)u agentive
(2.1.4). They are regularly attached to the verb stem, with the exception of -pits-o,
which is suffixed to a past tense marker. The marker -pits-o consists of the negator -pits and the future nominalizer -()o. A table presenting an overview of the
different types of subordinates constructed with these nominalizers can be found
in Section 2.1.5.
In addition to the nominalizing morphemes mentioned, the language uses
the deictics -ko this (one) and -iko that (one) to nominalize finite (fully
inflected) forms.3 Finite forms nominalized by -iko are never case-marked,
those by -ko can be followed by the causal case -(ak-)pat because and the
ablative case combination -man-ap, denoting instead of , (rather) than.
Nominalized finite forms marked by these deictics usually function as relative

a-kt-a-ko (1sg.s-be-ipfv-dem) I who is


a-meo-w-iko (1sg.a-3sg.o.want-pst-dem) I who loved him

I do not address the nominalizing demonstratives -ko and -iko in this paper.
2.1.1 Nominalizations formed with -(e) and -(k)te
Factivizer -(e) (-e after a consonant, - after a vowel) expresses the fact that.
Constructions with -(e) can function as a subject clause (3), a direct complement
clause (4), and a causal clause (5):


I spend the day spinning.
(lit. The fact that I am spinning takes the whole day.).

3. The phenomenon of a demonstrative functioning as a nominaliser is not an extraordinary

feature in Amerindian languages. In Jebero, for instance, the demonstrative asu this (one) also
has a nominalizing function when used as a suffix after verb stems (see Alexander-Bakkerus,

Nominalization in Choln


one-clf:group-ins 3pl.s-go-fac 1sg.a-3pl.o-see-pst
I saw them going in a troop.


3sg.o.miss-3pl.a-fac 3pl.a-3pl.o-whip-ipfv
They whip them, because they miss it.

The nominalizing suffix -(k)te infinitive (-kte after a vowel and -te after a consonant) is frequently employed as a complementizer:
(6) a-a-kte

1sg.s-go-inf 1sg.a-3sg.o.want-ipfv
I want to go.

2.1.2 Nominalizations formed with -(k)he and -(n)ap

The meanings of the morphemes -(k)he (-he after a consonant and -khe after a
vowel) simultaneity and -(n)ap (-ap after a consonant and -nap after a vowel)
sequentiality (after) supplement each other: the simultaneity marker -(k)he indicates that the event expressed by the verb takes place simultaneously with the
event expressed by the main verb o to do/make (7), whereas the sequentiality
marker -(n)ap indicates that the event takes place before another event indicated
by the main verb ny(e) to sleep (see below) (8):
(7) an-tsel
Fariseo Jesus mua i-l-o-w,
one-clf:elongated Pharisee Jesus prayer 3sg.a-3sg.o-do-pfv

1sg.poss-company 2sg.a-eat-imp-quot say-sim

A Pharisee prayed to Jesus, saying: Eat with me!
(8) elefante-wa a-tsel
elephant-top one-clf:elongated tree face-ad
pilmoh-no-nap --a
lean.on-refl-seq 3sg.s-sleep-ipfv
The elephant sleeps rested on a tree.
(lit. The elephant sleeps after having leant against a tree.)

Forms including the suffixes -(k)he and in -(n)ap have in common that they need
not be marked for person, when the subject of the verb to which they are suffixed
is identical (co-referential) to that of the main verbs o to do/make and ny(e) to
sleep, as in (7) and (8), respectively.

Astrid Alexander-Bakkerus

Nominalizations constructed with -(k)he may contain a subject clause (9), a

direct complement clause (10), a nominal predicate clause (11), and they can also
have causal meaning (12):
subject clause:
(9) me-t-iy
te putam makhay
2sg.s-weaken-pst 2sg.s-lame-pst high village joy
kuha-ma mi-esteh-he
ama -kot-a [...]
life-ines 2sg.s-enter-sim first 3sg.s-be-ipfv
It is better for you to enter into paradise weak and lame [...].
direct complement clause:
(10) paow me-kt-e i-m-pas-a-sim

good 2sg.s-be-purp 3sg.a-2sg.o-advise-ipfv-emp

He advises you, so that you will be good, not to abhor you.
nominal predicate clause:


aya up-he i-to
two-clf:stony-all meat eat-sim 3pl.s-be.ipfv
Both are eating meat.

causal clause:
(12) paow kot-he tsamo a-kot-t-a

good be-sim learned 1sg.s-be-fut-ipfv

Because I am good, I shall be taught.

The sequentiality marker -(n)ap after is employed to construct temporal clauses,

see (8) above, that can also be interpreted as causal clauses:


kama a-ki-a
rain-ad-abl 3sg.a-1sg.o-wet-pst 3sg.s-be-seq illness 1sg.s-be-ipfv
I am ill, because/after the rain wet me.

2.1.3 Nominalizations formed with -lam, -()o, -pakna and -pits-o

The nominalizers -lam future nominalizer 1 and -()o future nominalizer 2
(-o after a vowel, -o after a consonant) have the following features in common:
(a), both refer to a future event and to an obligation or a possibility; (b), they are
more noun phrase-like than clause-like: they can be used to derive a noun from a
verb (see Section 1), and, as nouns, they can modify a head noun and function as
relative clauses ((14)ab), and they can also function as subjects ((15)ab), direct

Nominalization in Choln

complement clauses ((16)ab), and nominal predicates ((17)ab). This is in line

with what Comrie & Thompson (1985:393) claim about nominalizations: It is
commonplace that a nominalization can occur wherever a noun-phrase is called
for. Thus, it is most natural for nominalizations to occur as subjects or objects [...].
Examples of a relative clause, a subject, a direct complement clause and a
nominal predicate clause coded by -lam and -()o are as follows:
relative clause:



1pl.poss-foot-ins 1pl.s-walk-fut.nmlz1 road
the road which we have to walk by our feet = a footpath


3sg.a-1pl.o-life-vb-fut.nmlz2 medicine
a medicine which gives us life = a vivifying medicine

subject clause:



morning one-clf:day-ant church-ad

1pl.s-go-fut.nmlz1 3sg-be-ipfv

The day after tomorrow we have to go to church.


paow ki-kot-o
good 1pl.s-be-fut.nmlz2 3sg.s-be-ipfv
We have to be good.

object clause:



ohom-a l-o-k
2sg.a-3sg.o-do-fut.nmlz1 quick-adv 3sg.o-do-imp
Do quickly what you have to do.


apo i-l-o-o
3sg.a-3sg.o-do-fut.nmlz2 3sg.s-3sg.o.know-ipfv
He knows [how] to make pans.

nominal predicate clause:



1pl.s-go-fut.nmlz1 1pl.s-be-ipfv

We can go.


mek hayu ki-kol-o

men 1pl.s-die-fut.nmlz2 1pl.s-be-ipfv
All men are mortal.

Astrid Alexander-Bakkerus

The negative nominalizer -pakna occurs in nominalizations encoding a subject

clause (18), and a nominal predicate clause (19):


which-cmp 1pl.a-3sg.o-do-neg.nmlz 3sg.s-be-ipfv
There is no remedy [for that]. (= It cannot be done by us.)


1sg.s-be-neg.nmlz 1sg.s-be-ipfv
I cannot be.

The suffix -pits-o has only been found in nominal predicate clauses:


1sg.a-3sg.o.know-neg-fut.nmlz2 1sg.s-be-ipfv
I did not know it.

Both negative nominalizers express impossibility, and both may have been
derived from the negative derivational suffix -p(e). The first one may be a
combination of -p(e), an unidentified element -ak- and a nominalizing element -na; the latter may be a combination of -p(e), the derivational suffix -its,
indicating a passive voice, and the nominalizing morpheme -()o. However,
notwithstanding these correspondences, -pakna and -pits-o differ in usage:
-pakna directly follows the verb stem, and can be used to form a nominalized
subject clause as well as a nominalized predicate clause, whereas -pits-o is preceded by a past tense marker, and is only used for the construction of a nominal
predicate clause.
2.1.4 Nominalizations formed with -(w)u
The suffix -(w)u (-u after a consonant and -wu after a vowel) is an agentive
marker. Nominalizations ending in -(w)u differ from the other nominalized
forms, in that they are never followed by a case marker, and that the subject or the
agent of the verb is not marked by a person prefix, but by the suffix indicating a
third person singular agent or subject. On the other hand, -(w)u nominalizations
can receive object marking.
As said, nominalizations with -(w)u are noun phrase-like, and, as nouns, they
could be used to form a subject clause, a nominal predicate, an object complement
clause, and a relative clause. In the data, we only have instances in which a -(w)
u nominalization is used in a subject clause (21), and instances showing that the
morpheme -(w)u is used to derive an adjective from a verb (22):
(21) pi-man-ap-sim
kamatsin ki-o-u
2sg.s-ines-abl-emp order 1pl.poss-do-ag

Nominalization in Choln

a-putam Israel
3sg.s.come.out-fut-1sg.a 1sg.poss-people Israel

From you will come out the ruler who will reign over my people of Israel.


a. kewak ole-wu
b. tup-u


a charitable person/merciful
(lit. the one who loves poor people)

It is obvious that -(w)u nominalizations are used to modify a head noun and can
be considered to be adjectives.
2.1.5 Overview of nominalized subordinates formed without case marking
The different subordinate clause types that can be constructed with the nominalizing suffixes described in this section are shown in the overview in Table 1:
Table 1. Subordinate nominalizations formed without case markers

Subject Complement Nominal Relative Causal Temporal

clause clause
predicate clause clause clause

-(e) factivizer

-(k)he simultaneity

-(k)te infinitive

-lam future nmlz 1

-(n)ap sequentiality

-()o future nmlz 2

-pakna negative nmlz 1

-pits-o negative-future nmlz 2

-wu agentive

It thus appears that subject complement clauses, direct complement clauses

and nominal predicate clauses can be formed by most of the nominalizers (five out
of eight). Causal clauses can be constructed by three nominalizers, and relative and
temporal clauses by two out of eight. The most productive nominalizer appears
to be the simultaneity marker -(k)he. It can be used to construct five different
clauses: subject, direct complement clause, nominal predicate, causal and temporal clauses. In terms of frequency, the simultaneity marker is followed by the future
nominalizers -lam and -()o. Both are used in four different clause types: subject,

Astrid Alexander-Bakkerus

direct complement clause, nominal predicate and relative clauses. The factivizer
-(e) occurs in the formation of three types of clauses: subject, object and causal
clauses. The sequentiality marker -(n)ap and the negative nominalizer -pakna are
employed in two types of clauses: -(n)ap: causal and temporal clauses, -pakna:
subject and nominal predicate clauses. The agentive marker -(w)u is used to construct a subject clause. The nominalizers -(k)te and -pits-o are the least productive
ones. They occur in one kind of clause: the former in complement clauses, the
latter in nominal predicate clauses, and they can thus be considered more specific.
2.2 Nominalized subordinates formed with case markers
As stated, the nominalizations formed with the suffixes -(e) factivizer, -(k)he
simultaneity, -(k)te infinitive, -lam future nominalizer1, -(n)ap sequentiality,
-()o future nominalizer 2, -pakna negative nominalizer, and -pits-o negationfuture nominalizer 2 can be case-marked. Case markers are used to derive adverbial subordinates, see Section 2.2.1. A special type of (subordinate) clause is the
hypothetical or irrealis clause, expressing a non-realised event, see Section 2.2.2.
2.2.1 Nominalized adverbial clauses
Several nominalized adverbial clauses, such as causal, comparative, contrastive,
instrumental, purposive, spatial and temporal clauses, can be formed by means of
different case markers and combinations of the same. For instance, a causal clause
can be constructed with the case marker sequentiality -tu-p indicating source. The
complex suffix is a combination of the personal adessive case -tu and the ablative
(23) into--a a-m-pako-kt-a
which-cmp-q 1sg.a-2sg.o-know-fut-ipfv

a-m-ya-pakna -kot-o-tu-p
1sg.a-2sg.o-see-neg.nmlz 3sg.s-be-fut.nmlz2-ad-abl
How shall I know you, for I cannot see you.

A comparative clause can be constructed by means of the suffix combination

-man-ap, consisting of the inessive case marker -man followed by the ablative -(a)
(24) a-n-utsa a-l-o-kte-man-ap
1sg.poss-rfm-sin 1sg.a-3sg.o-do-inf-ines-abl

a-kol-o a-kt-a
1sg.s-die-fut.nmlz1 1sg.s-be-ipfv
I shall die rather than commit a sin.

Nominalization in Choln

Choln uses the suffix -nake via in contrastive clauses:

(25) an-tsel a-tsap-te-na
one-clf:elongated 1sg.a-3sg.o.catch-fut-quot

-sepeh-he-nake an-tsel
3sg.s-claim-sim-per one-clf:elongated 3sg.a-3sg.o.catch-ipfv
Instead of catching one, he catches another one.
(lit. While claiming: I shall catch one, he catches another one.)

The instrumental case marker -pat can be used in an instrumental clause:


baptismo-te hayu -maso-kiah-he-pat

baptism-ad man 3sg.s-be.born-it-sim-ins1pl.poss-God-top

Our God was not satisfied with [the fact that] man was reborn in baptism.

The only example of a spatial clause found in the data is the following subordinate
clause. It is formed by means of the non-personal4 adessive case -te:
(27) ki-y-iy--te
where we slept

The benefactive case -he is used in a purposive clause:

(28) pana-nayme Soledad-te ki-ante-kte-he,
road-prol Soledad-ad 1pl.s-arrive-inf-ben
Huaylillas-nayme ki-pa-a
In order to arrive at Soledad by road, we pass Huaylillas.

A temporal clause can include the ablative -(a)p/-nap:


ama mi-l-o-ki
2sg.s-go-neg-fac-abl this first 2sg.a-3sg.o-do-imp
Before you go, do this first.

Examples (23) through (29) show that the case markers and case marker
sequences -tu-p source, -man-ap ablative, -nake via, -he benefactive, -te adessive, and -(a)p/-nap ablative can be used for the construction of a causal, a
4. There is a difference between -te non-personal adessive and -tu personal adessive.

Astrid Alexander-Bakkerus

comparative, a contrastive, an instrumental, a purposive, a spatial and a temporal clause, respectively.

2.2.2 Hypothetical or irrealis subordinate clauses
Hypothetical or irrealis clauses are formed by attaching the separative case marker
-(k)e, functioning as an anteriority or nominal past marker, to the nominalizing
suffixes -(k)te infinitive, -()o future nominalizer 2, -pakna negative nominalizer
1 and -pits-o negation-future nominalizer 2.
Example (30) below is an instance of a nominalized indirect complement
clause, constructed with the suffix combination -(k)te-ke (consisting of infinitive -(k)te followed by anteriority -(k)e), and Example (31) that of a nominalized
object clause ending in -pakna-ke (consisting of the negative nominalizer 1 and
the anteriority -ke). The former creates a frustrative, the latter refers to a nonrealised event:


1sg.poss-rfm-enemy 1sg.a-3sg.o.catch-inf-ant
You prevented me from catching my enemy.


mi-ha-wa mi-l-o-ha-pakna-ke
2-pl-top 2sg.a-3sg.o-do-pl-neg.nmlz-ant 1sg-per
I do what you (pl) have not been able to do.

The sequences -()o-ke future nominalizer 1-anteriority (32), and -pits-o-ke

negation-future nominalizer 2-anteriority (33), occur in conditional clauses
formed by means of the topicalizer -(w)a:


paow mi-kot-o-ke-wa
good 2sg.s-be-fut.nmlz2-ant-top 1sg.a-2sg.o-want-inf-ant
If you had been good, I would have wanted you.


alkalde a-kt-iy-pits-o-ke-wa
mayor 1sg.s-be-pst-neg-fut.nmlz2-ant-top
If I had not been mayor, they would not have helped you.

Nominalization in Choln

3. Nominalized main predicates

We may distinguish three groups of nominalized main predicates: nominalizations derived by the future nominalizers -lam and -()o, nominalizations containing the negative nominalizers -pakna and -pits-o, and a group constructed with the
irrealis markers -(k)te-ke, -()o-ke, and -pakna-ke. They are treated in Sections 3.1,
3.2, and 3.3, respectively.
3.1 Main predicates formed by -lam and -()o
As noted in Section 2.1.3, nominalizations constructed with the morphemes -lam
future nominalizer 1 and -()o future nominalizer 2 have in common that they
refer to a future event and to an obligation or a possibility and that they can modify a head noun and function as relative clauses ((14)ab), as subject complement
clauses ((15)ab), as object complement clauses ((16)ab), and as nominal predicate clauses ((17)ab). In the cases in which the nominalized form functions as a
subject or as a nominal predicate clause, the main verb is a copulative verb, see the
Examples (15)ab and (17)ab, respectively. However, in (15)ab the subject of the
copula is an impersonal third person singular, whereas in (17)ab the subject of
the copula is coreferential with that of the subordinate verb. It appears that, when
the copulative verb is omitted, forms with -lam and -()o may correspond to main
predicates. In (34) below, for instance, the main verb, the finite copula -kot-a he/
it is (< kot be), is omitted after the nominalizations mits-pale-kiah-lam and u-kupeo-lam. By the omission of -kot-a he/it is, the nominalizations with -lam take
over its function and meaning, so that they resemble main predicates:5


kapak mi-kot-nake
kas male-pale-kia-
power 2sg.poss-being-per wind
tsi mits-pale-kiah-lam
rain fire-cor

By your power, the wind blows from all parts, the rain falls down from
allparts, and the fire warms us from all parts.

5. Without the copula -kot-a he is, the forms mits-pale-kiah-lam and uku-peo-lam literally mean his falling down from all parts and his warming us from all parts, respectively; the
forms mits-pale-kiah-lam -kot-a and uku-peo-lam -kot-a, with the copula, are translated as he is falling down from all parts and he is warming us from all parts, respectively.

Astrid Alexander-Bakkerus

The same counts for the following example. In (35), the copula a-kt-a I am
is omitted after the nominalized verb a-meo-o:6
(35) a-pa ol-iy -kot-a-ko-man-ap-a
1sg.poss-father 3sg.s.die-pst 3sg.s-be-ipfv-dem-ines-abl-top
ok-e a-kl-iy
1sg-ant 1sg.s-die-pst 1sg.s-be-inf-ben

I would have preferred to have died myself than that my father had died!

Nominalized main predicates with -()o often occur in exclamations (see also (35)
above), as shown in example (36), repeated from above for convenience:
(36) a Dios into--am a-m-pako-o

intj God which-cmp-q 1sg.a-2sg.o-know-fut.nmlz2

a-kot-o a-m-ya-pakna
1sg.s-be-fut.nmlz2 1sg.a-2sg.o-see-neg.nmlz

O God, how shall I be able to know you, for I cannot see you!

3.2 Main predicates formed by -pakna and -pits-o

Nominalizations with negative -pakna occur as subject clauses, see (18), and as
nominal predicates (19), and the ones with negative -pits-o as nominal predicates
(20). In all these cases the main verb is a copula. When the copula is missing, the
subject and nominal predicate clauses with -pakna and the ones constructed with
-pits-o appear to take the place of the copula and resemble main predicates, like
those with -lam and -()o do when the copulative verb is absent, see Section 3.1


hayu-pit misa i-sinah-pakna-le
which-dem man-indf Mass 3sg.a-3sg.o.hear-neg.nmlz-q
Is there anyone not attending Mass?

6. a-meo-o, without the finite form a-kt-a I am, literally means my wanting it; a-meo-o
a-kt-a with the finite copula form, means I am wanting it, I want it. The form a-kot-te-he my
being preceding the nominalised form ameo-o is a complement of ameo-o (akt-a) I
want it, where it refers to a-kot-te-he.

Nominalization in Choln


I did not hear it.7

3.3 Main predicates formed by -(k)te-ke, -()o-ke, -pakna-ke

In Section 2.2.2 we have seen that the nominalizations with -(k)te-ke inf-ant,
-()o-ke fut.nmlz2-ant, and -pakna-ke neg.nmlz-ant are used to form a
subordinate clause. However, like those constructed with -lam, -()o, -pakna and
-pits-o, they can also be used for the construction of a main predicate. The nominalized forms with -lam, -()o, -pakna and -pits-o look like a main predicate when
a copulative verb is omitted. I therefore assume that, analogically, the forms constructed with -(k)te-ke, -()o-ke, -pakna-ke also resemble main predicates when a
copulative verb is absented. In Section 2.2.2 we have also seen that nominalized
forms with -(k)te-ke, -()o-ke, and -pakna-ke express non-realised or hypothetical events. The latter is also the case when the -(k)te-ke and -()o-ke nominalizations are similar to main predicates. The fact is, both -(k)te-ke and -()o-ke consist
of an element indicating future: -(k)te to do and -()o future nominalizer 2,
respectively, and, at the same time, of a morpheme indicating past: the anteriority
-ke. Both -(k)te-ke and -()o-ke thus refer to future in the past and mark events
that could or would take/have taken place, hence their occurrence in main clauses
expressing a wish:


inaham Dios a-ole-kte-ke

wishfully God
I wished I had loved God. (lit. Wishfully, I had loved God)


inaham paow a-kot-o-ke

wishfully good 1sg.s-be-fut.nmlz2-ant
I wished I had been good. (lit. Wishfully, I had been good)

The use of nominalizations formed by -pakna-ke partly differs from those formed
with -(k)te-ke and -()o-ke. The -pakna-ke nominalizations are negative by nature
and indicate an event that could not happen. Therefore, they are not suitable to
express a wishful thinking. On the other hand, nominalized forms constructed
with -pakna-ke do occur in main clauses following a conditional if clause (42), as
do the nominalizations constructed with -(k)te-ke (41):

7. a-sinah-pits-o literally means my not having to hear it. The non-nominalised counterpart
of a-sinah-pits-o I did not hear it would be a-sinah-pe-y 1sg.a-3sg.o.hear-neg-pst.

Astrid Alexander-Bakkerus

(41) Juan paow -kot-o-ke-wa

Juan good 3sg.s-be-fut.nmlz2-ant-top 3sg.poss.son-cor
paow -kot-te-ke
good 3sg.s-be-inf-ant
If Juan had been good, his son would also have been good.


yel -kot-iy-pits-o-ke-wa
salt 3sg.s-be-pst-neg-fut.nmlz2-ant-top meat

If there had not been salt, we could not salt the meat.

Both (41) and (42) are complex sentences consisting of a conditional clause: Juan
paow -kot-o-ke-wa if Juan had been good and yel -kot-iy-pits-o-ke-wa if
there had not been salt, respectively, followed by a main clause expressing the
implication of the condition. Note also that in Choln the clause in which the condition is stipulated regularly precedes the main clause, describing the fulfilment of
the condition, and that both the subordinated if-clause and the main clause are
nominalized forms.
4. Concluding remarks
In Section 1 we have seen that Choln nominalizations are more noun phrase-like
or more clause-like, and that they can be considered clauses in disguise. Besides
a nominalizing suffix, attached to the verb stem, a nominalization may contain
morphemes referring to a subject, if the stem is intransitive, or to an agent and
an object, when the stem is transitive. The nominalized form can furthermore
contain a negation, and a past marker, if the negator is -pits-o.
Choln has only four affixes to form a subordinate clause, to wit: the purposive -(e), the switch reference -hu, the coordinator -pit, and the topic marker
-(w)a. The vast majority of subordinate clauses are formed by nominalizations
constructed with the following morphemes: the factivizer -(e), the simultaneity
marker -(k)he, the infinitive -(k)te, the future nominalizer -lam, the sequentiality
-(n)ap, the future -()o, and the negative nominalizers -pakna and -pits-o. These
nominalizers are suffixed to a verb stem and used to form the following subordinates: subject, complement, nominal predicate, relative, causal and temporal
clauses. See the overview of nominalizers and the different subordinate clauses
they can create in Table 1, Section 2.1.5.
The nominalized forms mentioned above can be followed by case markers,
which underlines the nominal status of these forms, since only nominal stems can
be case marked. Examples of case marking suffixes encountered with nominalized

Nominalization in Choln

forms are -(a)p/-nap ablative, -he benefactive, -man-ap ablative, -nake perlative, -pat instrumental, -te non-personal adessive, -tu-p source. They are used to
form the following adverbial subordinate clauses: causal, comparative, contrastive,
instrumental, purposive, spatial and temporal clauses, respectively.
A special type of nominalized subordinates is the irrealis or hypothetical
clause. Irrealis clauses are constructed with the nominalizations with -(k)te, -()o,
-pakna, -pits-o, followed by the nominal past marker -(k)e, see Section 2.2.2.
In Amerindian languages, it is not uncommon for a nominalized form to
function as the head of a subordinate clause. In Choln, interestingly, the nominalizations with -lam, -()o, -pakna, -pits-o, and in the irrealis construction in -(k)
te-ke, -()o-ke, -pakna-ke, and -pits-o-ke are used as the head of both a subordinate
clause and a main clause. They normally function as a subordinate clause head, but
they become a main predicate when, as a subject clause or as a nominal predicate
clause, the verb they are related to by subordination is a copulative verb that is
omitted. In these cases the nominalized form can substitute the copula.
Nominalizations constructed with -()o future nominalizer 2 often occur
as main predicates in exclamations, see (35) and (36), those with -(k)te-ke infant and -()o-ke fut.nmlz2-ant in wishes (39) and (40), and the ones with
-(k)te-ke inf-ant and -pakna-ke neg.nmlz-ant in sentences in which a condition is made, see (41) and (42). In such if... then sentences, the conditional
if -clause contains a nominalization formed by -()o-ke fut.nmlz2-ant and
-pits-o-ke neg-fut.nmlz2-ant, and the corresponding main clause a nominalization formed by -(k)te-ke inf-ant and -pakna-ke neg.nmlz-ant, respectively.
It is never the other way round. The canonical order in if... then sentences seems
to be: -()o-ke fut.nmlz2-ant... -(k)te-ke inf-ant, and -pits-o-ke neg-fut.
nmlz2-ant... -pakna-ke neg.nmlz-ant.
So, as regards nominalizations resembling main clauses, in conclusion we can
say that only the nominalizations formed by means of the morphemes -lam, -()
o, -pakna, -pits-o, -(k)te-ke, -()o-ke, -pakna-ke, and -pits-o-ke can look like main
predicates, when: (a), they originally functioned as a subject or as a nominal predicate clause; (b), the main verb they were subordinated to is a copulative verb; and
(c), the copulative verb is suppressed. These nominalizations thus function as the
head of a main predicate, provided that a copula is understood.

Abbreviations and symbols

1 = first person; 2 = second person; 3 = third person; a = agent (subject of a transitive verb); abl = ablative; ad = adessive; adv = adverbial; ag = agentive; all =
allative; ant = anteriority; ben = benefactive; clf = classifier; cm = case marker;
cmp = comparative; cor = co-ordinator; dem = demonstrative; emp = emphasis;

Astrid Alexander-Bakkerus

fac = factivizer; fut = future; fut.nmlz1 = future nominalizer expressing obligation; fut.nmlz2 = future nominalizer expressing obligation/possibility; gen =
genitive; imp = imperative; indf = indefinite; ines = inessive; inf = infinitive; ins
= instrumental; intj = interjection; ipfv = imperfective; it = iterative; neg = negation; nmlz = nominalizer; o = object; per = perlative; pfv = perfective; pl = plural;
poss = possessive; prol = prolative; pst = past; purp = purpose; q = question;
quot = quotative; refl = reflexive; rfm = relational form marker; sg = singular;
s = subject of an intransitive verb; seq = sequentiality; sim = simultaneity; top =
topicalizer; vb = verbalizer

Alexander-Bakkerus, Astrid. 2005. Eighteenth Century Choln. Utrecht: LOT.
Alexander-Bakkerus, Astrid. 2011. Subordination in Choln. In Subordination in Native South
American languges [Typological Studies in Language 97], Rik van Gijn, Katharina Haude &
Pieter C. Muysken (eds), 193219. Amsterdam: John Benjamins. DOI: 10.1075/tsl.97.08ale
Comrie, Bernard & Thompson, Sandra A. 1985. Lexical nominalization. In Language Typology
and Syntactic Description, Vol. 3: Grammatical Categories and the Lexicon, 1st edn, Timothy Shopen (ed.), 156. Cambridge: CUP.
Cristofaro, Sonia. 2003. Subordination. Oxford: OUP.
Muysken, Pieter C. 1999. Nominalizations. In Concise Encyclopedia of Grammatical Categories,
Keith Brown & Jim Miller (eds), 248252. Amsterdam: Elsevier.
van Gijn, Rik, Haude, Katharina & Muysken, Pieter. 2011. Subordination in South America: An
overview. In Subordination in Native South American Languages [Typological Studies in
Language 97], Rik van Gijn, Katharina Haude & Pieter Muysken (eds), 123. Amsterdam:
John Benjamins.

abstract nouns 69
action nominalization 209
adjectival compounds 80,
adjectival incorporation 1
adjectival predicates 6, 113
adjective 13, 20, 22, 28, 30, 56,
61, 102, 111, 114, 117118, 120,
129, 131, 134137, 139, 143, 148,
204, 214
adjective-noun compounds
26, 122, 140
adverbializer 72, 161, 202
adverb-noun compounds 26
affection 5152
agentive nominalization 209
agglutinative language 35,
53, 79
Amazonian languages 115, 124,
126, 201, 203204
animal names (zoonyms) 39,
42, 5152, 54, 60, 62, 87, 89,
antipassive reduplication 117
apposition 27, 92, 95, 101
Arawakan 4, 6, 34, 79112,
115, 135
argument incorporation 6
associative qualities 7, 163, 165,
171173, 176177
atmospheric phenomena 69
attributive adjective 13, 131, 139
attributive clauses 36
attributive compounds 27,
91, 98
attributive nouns 5861,
6869, 7172
attributive phrase 42
attributive predicate 1, 7, 36,
40, 4244, 46
augmentative 6, 5177, 80,
111, 153
augmentative compounds 6
automatic reduplication 168
Aymara 34, 164165, 174176

body parts 18, 2528, 58, 61,
65, 70, 73, 80, 93, 114, 118,
124125, 128, 136, 139, 193
botanical names (see plant
bound nominal roots 81, 84,
94, 99, 106
Cariban 7, 181206
case, manipulation of case
roles 114, 117, 137
categorial neutrality 56
categorization mechanism 52
causative 140, 150, 156157, 161,
185, 202
causative reduplication 144,
Central Mapudungun 11
Chaco language (languages of
the Chaco area) 6, 3334,
4648, 5153
Choln 1, 7, 207224
Chorote 6, 33, 35, 38, 4648,
51, 68
classifier 73, 79112, 127, 223
classifier compounds (see
compound classifiers) 83,
85, 95, 99
classifier incorporation 79
clitic clustering 7, 181182, 192,
195, 201202
clitic compounds 2, 181, 200
cognitive/functional analysis
of incorporation 137
colour terms 107, 135136, 149,
155, 159
complete reduplication (see
also full reduplication) 7,
143, 147, 149, 153155, 159,
165166, 170, 176
complex classifiers (see
compound classifiers) 96
complex compounds 43, 94
complex construction 43, 169

complex noun phrases 2, 11, 22

complex reduplication 143144
composition (see
compound classifiers 84, 95,
compound nouns 36, 40,
6567, 72, 82, 125
compounding 17, 12, 1516,
19, 3350, 52, 55, 79113, 122,
135, 163, 165166, 181206
compounding strategies 6
compounding, the distinction
between compounding and
phrase 4547, 48, 81, 99,
concrete nouns 55, 69
constructional idioms 92
continuation 152
continuous reduplication
166167, 172, 174176
coordinate compounds 86,
92, 98
count nouns (see count/mass
count/mass distinction 5557,
60, 71
cross-categorial morphemes
derivation 24, 33, 37, 3940,
48, 52, 5556, 60, 82, 84, 88,
100, 103, 106, 110, 118, 135,
137, 149, 183, 189
derivation distinction
between derivation and
compounding 13, 3334,
37, 40, 48, 52, 8182, 99, 110
derivation distinction
between derivation and
inflection 1, 3, 6, 5657, 79,
106, 110
diachronic data 40, 52, 54,
57, 109, 135, 144, 156, 157,

diminutive 2, 3940, 53, 5657,
6465, 68, 72, 80, 101
reduplication 165167
distribution 2, 11, 149150, 153,
155, 160, 171
distributivity 144, 165, 171, 174,
elative 65
emphatic function 58, 6365,
67, 144
endangered language 53, 80
endearment 39, 54
endocentric 30, 40, 84, 86, 88,
91, 99
ergative alignment 116117
Ese Ejja 1, 6, 113142
ethnobiological lexicon
(animal names and plant
names) 34, 51, 52, 54, 62, 66,
68, 87, 103, 176
evaluative morphology 5152,
5457, 6869, 74
evaluative semantic attributes
exocentric 30, 86, 91, 98, 103,
expressivity 64
extended verb stem 184185,
extended verb stem
compound(s) 182, 185
extensional value 153
external possession 115,
137138, 141142
father (lexeme used in
evaluative morphology) 12,
15, 2122, 5773, 184, 191, 220
fauna terminology (see animal
flora terminology (see plant
foregrounding 114, 140
frequency 56, 61, 189, 209, 215
frozen (morphology) 35, 37,
47, 60
full reduplication (see
complete reduplication)

gender 30, 35, 53, 5658, 61, 69,
7273, 92
generic nouns 68
grammaticalization 23, 6,
3839, 4849, 66, 68, 76, 90,
102, 108109, 127, 170, 178
Gran Chaco region 33, 34,
4648, 5153
Guaran 34, 115, 126, 137, 142
Guaycuruan 6, 35, 4748,
head 1131, 40, 53, 79, 82, 84,
89, 9199, 107, 118, 125, 182,
208209, 212, 215, 219, 223
head noun 13, 42, 69, 91, 212,
215, 219
head-final compounds 1617
head-initial compounds 18,
23, 25, 29
hierarchical relation 52, 68,
hyponymy relation 55, 74
iconic reduplications 173
iconicity 57, 72, 143, 152153,
160161, 171, 176
Iconicity Principle 7, 143161
idiosyncratic compounds 91
implicational hierarchy 56, 72
inalienable possession 35,
59, 105
incorporation 1, 4, 6, 17, 79,
8283, 85, 95, 102107, 110,
113142, 184, 186, 201
incorporation of compounds
96, 125
incorporation of phrases 17,
128, 131
incorporation of verbal
roots 132134
independent lexeme 38, 40
individuation 87, 106, 153154,
inflection 34, 10, 27, 56, 106,
110, 145, 186, 189
inflectional (see inflection) 1,
34, 6, 5657, 66, 79, 167, 181,
183, 185, 187, 201

inflection (distinction between

inflection and derivation)
inflectional categories 4
innovation 58, 177
instrumental nominalization
41, 216
insult 61, 122, 185
intensification 6, 55, 74,
151152, 165, 171174,
intensional reduplication 159
intensional value 153
intensity 7, 5152, 144145, 149,
151, 153, 158160
intensity prefix 151, 159
intensity reduplication 145,
149, 159160
interclausal reference
marking 2, 7, 181206
interclausal reference marking
compounds (= IRM
compounds) 182, 190, 195,
internal structure (of
compounds) 34, 40
irrealis 9091, 183, 186, 216,
218219, 223
isolate language (see language
iterative 56, 189, 195, 198
Kalapalo 12, 7, 181206
Kallawaya 7, 163180
kinship compounds 39, 51,
kinship terms 39, 57, 6573,
8081, 118
language contact 48, 179
language isolate 7, 143, 160
left-bound roots 84, 86, 104
lexemes creation of new
lexemes 1, 3, 5152, 71, 73
lexical derivation 100, 110
lexical domain 68
lexical nominalization
lexical root compounding 6
lexicalization 48, 89, 9697,
127, 151, 173177

lexicalized compound 41, 64,
91, 98
lexicalized names 63
lexicalized phrases 47
lexicalized reduplication 148
linking element 86, 9091,
98, 105
locative 58, 7980, 83, 8586,
9294, 9899, 101, 104105,
107, 111, 127, 135, 140, 161,
locative compounds 83, 9293,
main predicate 207, 209, 221,
Mak 2, 6, 3350, 51
manipulation of case role 114,
117, 137, 140
Mapudungun 12, 5, 1131, 44,
83, 88
mass nouns (see count/mass
Mataco-Mataguayan 6, 35, 46,
48, 68
material-product relation 89
metaphor 41, 184185
metonymy 21, 91, 98, 149, 189,
191, 194
mixed compounding 82,
mixed language 7, 163
mockery 5152, 61
modification 11, 82, 84, 88, 103,
115, 132
modifier nouns 6970
modifying nominalization 208
modifying root 40, 43
morphotypes 68
mother (lexeme used in
evaluative morphology) 15,
2122, 5773, 151, 193
multiple compounds 9798
negative existential 44, 46, 48
neologisms 41, 44, 48, 65
Nivacl 35
nominal categorization 51
nominal compounding 12,
56, 1112, 1520, 52, 73,

8283, 86, 95, 99, 101,

106107, 110
nominal incorporation 17
nominal morphology 12,
7980, 94, 99, 101, 110
nominal predicates 36, 42, 69,
208, 213, 220
nominalization 78, 91,
100, 108, 170, 172, 183184,
205206, 207224
nominalized main
predicates 207, 219220
nonhead 11, 17, 2022, 25,
nonhead-head 27
non-individuatable 155
non-inflecting verbs 119
North-American indigenous
languages 68, 126
noun classification 53, 74
noun incorporation (NI) 1, 4,
17, 107, 113142, 184, 186, 201
noun incorporation into
adjectives (NIA) 128
noun incorporation into verbs
(NIV) 122
noun-adjective compounds
122, 135
compounds 29
noun-noun compounds 16,
numeral compounds 83, 86,
9496, 101102
objective nominalization 208
oblique arguments 105, 107,
114, 125, 137
opaque forms (see frozen
parental relations 57
part-whole relation 124
partial prefixed
reduplication 7, 143144,
147, 149, 151, 158159
partial reduplication 7, 165166,
169171, 173, 176177
partial suffixed reduplication
7, 143144, 147, 150, 155

partitive construction 101

parts of plants 118, 125
pejorative 5152, 58, 61, 70,
7374, 189, 202
person marking 36
personal clitics 80, 84
phrasal compounds 168
physical dimension 60, 67,
phytonyms (see plant names)
plant names 4142, 5152, 54,
58, 60, 62, 68, 8687, 103,
pluractional reduplication 151
pluractionality 7, 144, 150151
pluralization 47, 152153
plurimorphemic 12, 82
polysynthesis 53, 79, 109
possessed noun 35, 40, 113,
possessive relationship 40
possessive third person
singular marker (in
compounds) 37, 40
possessor marking 35, 80, 84,
86, 99
possessor raising 105
possessor-possessee relation
relation 115, 137
posture verbs 120121, 127,
predetermined third person
(possessor) 40
predicate clause 212214, 219,
predicative adjective 118, 137,
predicative noun 47
prefixed partial reduplication
(see partial prefixed
preposition 20
privative 44, 45
productive compounding 6,
3839, 41, 48, 79, 81, 84, 96,
99, 110
productive incorporation 6
productive suffix 38, 58
proper names 122
prototype 6768, 81, 122, 133

qualitative dimension 63
quantifier in compound
construction 40, 45
quantitative dimension 55, 60
Quechua 16, 3334, 51, 163180
reduplicant 166167, 174
reduplication 1, 45, 7, 117, 135,
143161, 163180, 199200
reduplication prefix 159160
reduplicative derivation 149
referential 14, 55, 72, 99, 136,
161, 211
relative clauses 7, 207, 210, 212,
216, 219
resemblance or approximation
65, 67
reverse compounds 86, 9192,
reversed syntactic order in
compounds 43, 44, 46,
right-bound roots 84, 86, 94
right-headed compounds 84
secret language 163164,
semantic bleaching 90, 99
semantic extension 88, 124,
semantic head 89

semantic lexicalization
173174, 176177
social relations 6566, 7374
Spanish grammar 22, 27, 56,
65, 68, 88, 137
Spanish loanwords 15, 23,
59, 8788, 102, 118119,
spatial dimension 55, 7374
spatial relations 82, 118, 125
state nominalizations 208
subject markers (see person
subordinate compounds 29,
87, 103, 106
subordination 11, 184, 207,
209, 223
suffixed partial reduplication
(see partial suffixed
switch-reference 181, 190191,
195, 200202
synchronically unproductive
morphology 62, 64
synesthetic sound symbolism
syntactical nominalization (see
syntactical nominalization)
syntactic nominalization 207,
Takanan 6, 113, 116, 118, 141
temporal quantification 61, 64

third person singular

possessive prefix (in
compounds) 37, 40, 42, 45
Toba 6, 33, 4648, 50, 5277
transfunctionalization 88
transparency of semantics 22,
81, 143
truncation of nominal roots
90, 104105
typological universals 33, 34
universals 10, 3334, 5657
verbal compounding 83,
108110, 113, 201
verb-noun compounds
1617, 30
verbal morphology 104, 107,
110, 143, 178
vowel harmony 3637, 60
whole-part relation 87, 89
Wich 6, 35, 38, 4648
word order 3536, 53, 116, 183
wordhood 23, 15
Yurakar 7, 143161
zoonyms (see animal names)