English Abstract Nouns As Conceptual Shells, From Corpus To Cognition

English Abstract Nouns as Conceptual Shells
WDE
G
Topics in English Linguistics
34
Editors
Bernd Kortmann
Elizabeth Closs Traugott
Mouton de Gruyter
Berlin New York
English Abstract Nouns
as Conceptual Shells
From Corpus to Cognition
Hans-Jrg Schmid
W
Mouton de Gruyter
DE
G Berlin New York 2000

Mouton de Gruyter (formerly Mouton, The Hague)
is a Division of Walter de Gruyter G m b H & Co. K G , Berlin.
Printed on acid-free paper which falls within the guidelines of the

A N S I to ensure permanence and durability.
Als Habilitationsschrift auf Empfehlung der Philosophischen Fakultt fr

Sprach- und Literaturwissenschaft I der Ludwig-Maximilians-Universitt Mnchen gedruckt
mit Untersttzung der Deutschen Forschungsgemeinschaft.
Die Deutsche Bibliothek CIP-Einheitsaufnahme
Schmid, Hans-Jrg:
English abstract nouns as conceptual shells : from corpus to
cognition / Hans-Jrg Schmid. - Berlin ; New York : M o u t o n
de Gruyter, 2000
(Topics in English linguistics ; 34)
ISBN 3-11-016767-0
Copyright 2000 by Walter de Gruyter G m b H & Co. K G , D-10785 Berlin

All rights reserved, including those of translation into foreign languages. N o part of this
book may be reproduced or transmitted in any form or by any means, electronic or
mechanical, including photocopy, recording, or any information storage and retrieval
system, without permission in writing from the publisher.
Cover design: Christopher Schneider, Berlin.
Printing: Druckerei Gerike, Berlin - Binding: Lderitz & Bauer G m b H , Berlin
Printed in Germany
For Susanne,
David, Luis and Quirin
Acknowledgements
This book would not be what it is without the support of a number of

friends, colleagues and institutions. To all these I am extremely grateful.
In particular, I would like to thank the Deutsche Forschungsgemein-
schaft (DFG) for giving me a one-year grant which enabled me to carry out
research in Birmingham and Oxford, and what is more, to have my family
with me during this exciting period of time. I am grateful to the colleagues
at COBUILD, Birmingham, especially Jeremy Clear, Gill Francis and
Ramesh Krishnamurthy, for letting me work with their wonderful corpus
and take out large amounts of data, for their practical help in this process
and their patience with me. I would also like to thank Oxford University for
giving me access to the Bodleian and other libraries. Malcolm Coulthard
(Birmingham), Geoffrey Leech (Lancaster) and Len Lipka (Munich) kindly
gave me the opportunity to present my ideas in their staff seminars and
linguistic colloquia, and I should like to thank them and the members of
these seminars for their interest in my work.
It is much more difficult to appreciate and acknowledge the enormous
debt of gratitude I owe to Len Lipka and Geoffrey Leech, as well as Jean
Aitchison (Oxford) and Wolfgang Falkner (Munich), for their involvement
in the early stages of the development of this study. Each of them in their
own characteristic way contributed immensely to the formation and formu-
lation of the ideas presented here. Len Lipka and Wolfgang Falkner also
read and commented on the entire manuscript of an earlier version of this
book, which was submitted to Munich University as a Habilitationsschrift,
and so did Nick Jacob-Flynn, Dick Janney, Wolfgang Schulze, Dietmar
Zaefferer (all Munich), Thomas Herbst (Erlangen), Bernd Kortmann and
Lieselotte Anderwald (both Freiburg). Walter Hofstetter (Munich) helped
me to come to terms with some of the inconsistencies in grammatical termi-
nology. I would like to thank all these colleagues for their invaluable com-
ments and suggestions for improvement. And I am grateful to Claudia
Hommers (Bochum) and Anja Gebert (Bayreuth) for their help with the
tedious task of revising the earlier manuscript, Anja Gebert also for proof-
reading the final one. Finally, I would like to thank my family, to whom this
book is dedicated, for staying with me all the same.
Table of contents
Part I Foundations: Theory, terminology and methodology 1
1. Introduction 3
2. Approaching shell nouns 10
2.1 The term shell noun 10
2.2 Defining shell nouns and shell-content complexes in
functional terms 13
2.3 A brief note on the theoretical stance 20
3. The links between shell nouns and contents 21
3.1 Triggering co-interpretation 21
3.1.1 Lexico-grammatical patterns of shell-noun uses 21
3.1.2 From identity of reference to experiential identity 27
3.2 The semantic contributions of different types of complements:
a survey of the evidence from verbal complementation 31
3.3 Basic functions of shell-noun typical patterns 36
4. The systematic investigation of shell nouns 38
4.1 The From-Corpus-to-Cognition Principle 38
4.2 Data retrieval 40
4.3 Cleaning up the data 48
4.4 Systematic misses of the corpus inquiry 51
4.5 A survey of the results of the corpus inquiry 53
5. Semantic prerequisites 63
5.1 Abstractness 63
5.1.1 Extensional abstractness and classes of abstract entities 63
5.1.2 Stylistic abstractness and grammatical metaphor 70
5.2 Unspecificity and structure-inherent semantic gaps 73
5 .3 Summary of Part I 80
Part II The use of shell nouns 83
6. Describing shell-noun uses 85

6.1 Degrees of typicality 85
Table of contents
6.2 Explaining the meanings of shell-noun uses: features

and frames 87
7. Factual uses 92
7.1 Introduction 92
7.2 Neutral uses 93
7.3 Causal uses 101
7.4 Evidential uses 110
7.5 Comparative uses 113
7.6 Partitive uses 116
7.7 Attitudinal factual uses 120
8. Linguistic uses 131
8.1 Introduction 131
8.2 Prepositional uses 139
8.3 Illocutionary uses 147
8.3.1 Assertive uses 153
8.3.2 Rogativeuses 166
8.3.3 Directive uses 170
8.3.4 Commissive uses 176
8.3.5 Expressive uses 181
9. Mental uses 184
9.2 Conceptual uses 188
9.3 Psychological-state uses 195
9.3.1 Creditive uses 195
9.3.2 Dubitative uses 208
9.3.3 Volitional uses 209
9.3.4 Emotive uses 226
10. Modal uses 231
10.2 Epistemic uses 235
10.3 Deontic uses 244
10.4 Dynamic uses 251
11. Eventiveuses 261
11.2 General eventive uses 262
11.3 Specific eventive uses 266
11.4 Attitudinal eventive uses 270
Table of contents xi
12. Circumstantial uses 275

12.2 General circumstantial uses 277
12.3 Specific circumstantial uses 279
13. Summary of Part II 292
Partili Functions of shell nouns 301
14. Introduction to Part III 303

15. Semantic functions 308
15.1 The characterizing potential inherent in shell nouns 308
15.2 Characterization expressed by premodifiers 317
16. Pragmatic, rhetorical and textual functions 329
16.1 Focusing and topicalizing 329
16.2 Linking 339
16.3 Signposting 349
17. Cognitive functions 360
17.1 Conceptual partitioning 360
17.2 Reifying and hypostatizing 363
17.3 Integrating 370
18. Conclusion and outlook 377
Appendix 381
Notes 409
References 421
Index of shell nouns 443
Index of subjects _ 453
Parti
Foundations: Theory, terminology and

methodology
Chapter 1
Introduction
This study is concerned with a class of abstract nouns and their linguistic
environments. Among the most typical and frequent examples are the nouns
case, chance, fact, idea, news, point, problem, position, reason, report,
situation and thing. From a grammatical point of view, the most striking
feature of these nouns is that they can be inserted in one or both of the two
grammatical patterns given and illustrated in (1.1):
(1.1) (a) Determiner + (Premodifier) + Noun + postnominal that-c\ms&, wh-

clause or to-infinitive
The (deplorable) fact that I have no money.
(b) Determiner + (Premodifier) + Noun + be + complementing that-
clause, wA-clause or to-infinitive
The (big) problem was that 1 had no money.
(It should be noted that the postnominal that-clause in example (1.1a) must
not be mistaken for a relative clause; the conjunction that cannot be re-
placed by the relative pronoun which). As is shown in (1.2), not all English
nouns, not even all nouns that are commonly regarded as abstract nouns,
can be used in these patterns.
(1.2) (a) *The boy that I had no money ...

*The democracy that I had no money ...
*The inflation that I had no money ...
(b) *The boy was that I had no money.
*The democracy was that I had no money.
*The inflation was that I had no money.
For reasons explained in detail in this and especially the next section, I refer
to nouns which can be used in the two types of constructions given in (1.1)
as shell nouns. To give the reader a first rough idea of what shell nouns are,
I have collected some of the most typical and frequent examples in Table
1.1. As the Table shows, shell nouns can be categorized into six classes on
the basis of their meanings.
4 Introduction
Table 1.1 Examples of shell nouns
Class Examples
Factual fact, thing, point, problem, reason, difference, upshot
Linguistic news, message, rumour, report, order, proposal, question
Mental idea, notion, belief, assumption, aim, plan, decision
Modal possibility, truth, permission, obligation, need, ability
Eventive act, move, measure, reaction, attempt, tradition, trick
Circumstantial situation, context, place, area, time, way, approach
Shell nouns make up an open-ended functionally-defined class of abstract

nouns that have, to varying degrees, the potential for being used as concep-
tual shells for complex, proposition-like pieces of information. Some of
them seem to be geared for this type of usage, and can therefore be seen as
prototypes of the class, some are occasionally used that way, and some
hardly ever so. As will be shown in greater detail in Section 2.2, nouns are
not shell nouns because of some inherent property, but become shell nouns
when they are used the way described above. The term shell noun is thus
only a convenient shorthand for 'use-as-shell noun'.
Shell nouns have on the whole received far less attention from linguists
than they deserve. Among those linguists who have noticed some of the
special features of these nouns are the early modern grammarians. Both
Poutsma (1929: 619-620) and Jespersen (1927: 24-26), for example, re-
mark that the nouns fact and circumstance can be used, as Jespersen puts it,
to "prop up the clause" when a that-cieaise functions as subject, or to evade
"the difficulty of joining an object to certain verbs", as in this could not
conceal the fact that he was growing old. Jespersen also mentions exam-
ples like their belief that and the idea that, adding that "some grammarians
here disapprove of the term 'object' and say that the clause is appositional
to the preceding substantive" (1927: 27). It is also interesting to note, espe-
cially in the light of the image of shell-content complexes used in this study,
that Jespersen calls the clauses in such constructions "content-clauses"
(1927: 23-32), a term which is also found in Huddleston (1984: 118-120,
263-264).
In Quirk et al. (1985), the nouns in question are mentioned in the con-
texts of complementation (1985: 1231) and apposition (1985: 1260-1261,
1271-1274, 1321). Quirk et al. argue that abstract nouns that are morpho-
logically related to verbs and adjectives (e.g. likelihood, warning, reminder,
Introduction 5
advice) take complements in a way which largely but not completely corre-
sponds to their verbal and adjectival counterparts. In the section on apposi-
tion, it is claimed that not only morphologically isolated abstract nouns like
fact and idea but also nouns that are related to verbs like remark or answer
occur as heads of noun phrases with appositive clauses as postmodifiers. I
will come back to these claims in Section 3.1.1.
Biber et al. (1999: 648-656) devote a whole section to "head nouns tak-
ing complementing clauses". Corpus findings on nouns like fact, idea, hope,
possibility and doubt (which take that-clauses) and chance, attempt, effort,
ability and opportunity (which take to-clauses) are provided and discussed.
In Sinclair's COBUILD grammar (1990), nouns like statement, advice,
opinion, information and decision are discussed as "nouns used with re-
ported clauses" (1990: 338), and the same nouns, as well as many others,
are described in a later chapter on Making texts as a means of "referring
back in a general way" (1990: 389-391). In another book based on the
COBUILD corpus, the Collins COBUILD English Guide 9: Linking words
(Chalker 1996), nouns like thing, case, fact, idea and situation are also
described as means of establishing links that "help to bind sentences to-
gether in a text" (1990: 94; see 82-83 and 94-115). Francis (1993), who
also works with this corpus, looks at nouns which can occur with "apposi-
tive -clause qualifiers" (1993: 148-155). She claims that there are
roughly four hundred nouns or "broad senses of nouns" (1993: 148) of this
type and divides them into six classes. Her estimate as to the number of
head nouns that can be combined with appositive âi-clauses is supported
by the present study, in which 350 lexemes were found to occur in the pat-
tern in noteworthy frequencies. Francis' classification, on the other hand,
will be considerably refined here, and a much wider range of nouns - nouns
that take appositive infinitives and w/z-clauses, as well as nouns which take
-clauses, w/j-clauses or infinitives as complements after the copula -
will be examined.
In the 1960s, the philosopher Vendler (1968: 72-82) discussed the syn-
tax and semantics of a similar group of nouns, which he called container
nouns, mainly with the aim of illuminating the philosophical distinction
between facts and events. The basis for his image of containers was that
combinations of a copula verb with nouns like fact or idea can fonction
syntactically as containers, or hosts, for âi-clauses, for example in sen-
tences like That he died is a fact (Vendler 1968: 73). Menzel (1975) took
up these ideas in a later phase of Transformational Grammar, which had of
course also influenced Vendler, and argued that such nouns as fact, propo-
6 Introduction
sition, event, action and act are abstract head nouns in the deep stucture of
clauses, which first determine the match between predicates and types of
complements and are then deleted by transformation rules. More recently,
Vendler's approach was developed further by Asher, who calls such expres-
sions as the possibility that Mary had left without John "noun complement
constructions" (1993: 30).
The most familiar term for a closely related class of nouns in the lin-
guistic literature is probably the notion of general noun (Halliday and
Hasan 1976: 274-277). Yet not all of Halliday and Hasan's examples can
be used in the patterns in (1.1), for example people, person and creature
can not. Bolinger uses the terms low-content nouns (1977: 5-6) and classi-
fiers (1977: 50-51) for a similar group of nouns. Like general noun, the
former notion mainly highlights the semantic generality or unspecificity of
the nouns in question. Yet although most of the nouns that can be used in
the patterns in (1.1) are semantically unspecific in a certain way,1 the em-
phasis on this property has unduly diverted the attention of linguists from
the interesting relation between the nouns and the postnominal or comple-
menting clauses which can frequently be found in their contexts. Bolinger's
term "low-content nouns" also evokes the image of a deficient class, and
therefore does no justice at all to the ubiquity and utility of the nouns.
Since the terminology in this area thus does not seem to be quite satis-
factory, I would like to introduce the new term shell noun (already men-
tioned above) for the particular class of abstract nouns that I am interested
in. This term is derived from the recognition that shell nouns are used by
speakers2 to create conceptual shells for complex and elaborate chunks of
information. These are expressed in clauses,3 or sometimes in longer
stretches of the neighbouring discourse. More about the motivation for this
term will be said in Section 2.1.
I think that shell nouns are worthy of a volume-sized study for a number
of reasons. For a start, many shell nouns belong to the most frequently used
nouns in the English language. For example, in a corpus of 225 million
running words of British English,4 the singular forms of the lexemes case,
fact, idea, news, point, problem, report and thing are among the one hun-
dred most frequent nouns, with frequencies of occurrence ranging from
80,013 (or 356 occurrences per million) for thing to 46,654 for idea (207
occurrences per million). In view of the finding that "by far the majority of
lexical items have a relative frequency in current English of less than 20 per
million" (Clear 1993: 274), these frequencies are indeed remarkable.
Introduction 7
One reason for the frequency of these nouns is that they are surprisingly
versatile and powerful linguistic and conceptual tools. A first impression of
this potential can be gleaned from example (1.3), an extract from a radio
programme concerned with the health policy of the British Government. The
example is taken from the BBC material collected in the COBIULD corpus
(see Section 4.2):
(1.3) The Government's aim is to make GP's more financially accountable,

in charge of their own budgets, as well as to extend the choice of the pa-
tient. Under this new scheme, family doctors are required to produce
annual reports for their patients ...
The two noun phrases that contain shell nouns (i.e. 'shell-noun phrases') are
printed in bold-face types in this example, while the 'content' of the shells,
i.e. what both shell-noun phrases relate to, is underlined. Essentially, the
noun phrase the Government's aim, which occupies the subject position in
the first sentence, does three things: first, it relates to the underlined pas-
sage, the group of propositions expressed in the two coordinated comple-
menting infinitive clauses and the appositive clause attached to the first of
them; second, it characterizes this information as an aim, i.e. as something
the British Government wants to achieve; and third, it casts this complex
piece of information into one single noun phrase.5 This is mainly achieved
by the equative relation evoked by the copula is. The speaker uses this rela-
tion between a clause and a nominal shell to help the hearer keep the gist of
the information active and to re-activate it if this should be required as the
discourse unfolds.
In the second sentence the speaker uses the noun phrase this new scheme
as a signal for precisely such a re-activation. Since the same information is
activated, the clause-initial noun phrase in subject function can thus be re-
garded as given information. However, the speaker provides it with a differ-
ent conceptual shell. By using the noun scheme, rather than just repeating
the noun aim, it is conveyed that the intended achievements have already
been outlined in a fairly concrete and detailed form. This characterization is
included in the meaning of the noun scheme. Since noun phrases allow for
the possibility of premodification, the speaker can easily and economically
add the information that the scheme is new, before he or she goes on to en-
large on some of the details that are involved. In short, the speaker uses the
anaphoric demonstrative this to link the second shell noun scheme to the
information expressed in the previous sentence, and the whole shell-noun
8 Introduction
phrase to modify and characterize it in a particular way. The fact that given
information is taken up as a starting-point for the second sentence contrib-
utes considerably to the impression that the passage is about one topic and
therefore coherent, but also helps the speaker to get on with what he or she
is trying to say about it.
It is essentially this linguistic and conceptual process that I want to
capture by calling the nouns in question shell nouns, and the noun phrases
in which they occur shell-noun phrases. Shell nouns and shell-noun phrases
can only be studied appropriately if what they link up with is taken into
account. This means that I will generally be concerned with shell-content
complexes, rather than just the nouns alone. However, since it would be
cumbersome to speak of shell-content complexes all the time, the terms
shell nouns and shell-noun phrases will be used with the tacit understand-
ing that their communicative impact always depends on their occurrence in
shell-content complexes.
It is important to emphasize even at this early stage that it is always the
speaker of an utterance who characterizes some piece of information by
choosing a particular shell noun and modifier. Instead of using the fairly
neutral nouns aim and scheme the speaker of (1.3) could have emphasized
that the government is struggling hard to introduce these changes by using
such shell nouns as endeavour or effort; he or she could have stressed the
necessity and importance of these changes by using the nouns need or obli-
gation, or could have introduced emotional aspects by using such shell
nouns as wish or desire. Another possibility would have been to use an
evaluative shell-noun phrase such as the Government's problem in order to
highlight that difficulties are involved. The ease with which different nouns
can be inserted into this context shows how shell nouns provide speakers
with powerful tools for the characterization, perspectivization, and indeed
even manipulation, of their own and other speakers' ideas. Especially politi-
cians and other people with debating experience are proficient in character-
izing their own ideas as facts, truths, advantages, important points and
central issues, while characterizing the ideas of their opponents as theories,
hypotheses, problems, questions or dangers.
Yet another reason why shell nouns deserve a thorough investigation is
the ubiquity, especially in informal spoken conversation, of such seemingly
awkward expressions as the thing is that these children for instance are
badly behaved ones usually (SPOKEN CONVERSATION, COBUILD). Similar
"utterance launchers" or "ouvertures" (Biber et al. 1999: 1073-1076) are
the problem is (that), the trouble is (that), the fact is (that) and the truth is
Introduction 9
(that). Since the noun phrases introducing such clauses are more or less
redundant from a purely propositional point of view, it is definitely worth
trying to find out what the reasons for their high frequency of occurrence
are. As will emerge in Sections 7.2 and 16.1, pragmatic, rhetorical and in-
formation-distributional aspects like focusing and topicalizing play an im-
portant role here.
I will look at shell nouns, shell-noun phrases and shell-content com-
plexes from various perspectives in this study:
a) Theoretical and methodological perspectives: How can shell nouns and

shell-content complexes be defined, and how can they be investigated
systematically?
b) Descriptive perspective: Which nouns do speakers use as shell nouns
and what types of shell-content complexes do they create?
c) Functional perspective: What are the semantic, pragmatic, rhetorical,
textual and cognitive motivations for using shell nouns, and why do we
use some of them so frequently?
This study falls into three major parts which try to provide answers for
these three questions in turn. In the remainder of Part I, I will first explain
the metaphor underlying the concept of shell noun (Section 2.1), and then
put forward a functional definition of shell nouns and shell-content com-
plexes (2.2). In Chapter 3, the types of linguistic links that are used by
speakers to trigger a co-interpretation of shell nouns and their contents are
examined. Chapter 4 contains an account of the method used for the investi-
gation of shell-content complexes. It is this method which motivates the title
of this book, From Corpus to Cognition. Part I closes with an attempt to
clarify the role played by the notions of abstractness and unspecificity,
which are seen as essential semantic prerequisites for the successful use of
shell nouns.
Part II is devoted to a detailed description of the use of shell nouns and
shell-content complexes. In Chapter 6 some necessary linguistic tools are
introduced. Chapters 7 to 12 contain the descriptive core of the study, which
is divided into six classes of shell-noun uses.
Although pragmatic aspects are taken into account throughout Parts I
and II, it is in Part III that the pragmatic perspective takes centre stage.
Semantic, pragmatic, rhetorical and textual, as well as cognitive functions
of shell-content complexes will be examined in Chapters 15 to 17.
Chapter 2
Approaching shell nouns
2.1 The term shell noun
Although it is true that abstract nouns have not been very popular as objects
of linguistic research, this is of course not the first study that is concerned
with shell nouns or similar types of nouns. Other researchers have taken an
interest in nouns which overlap with the class of shell nouns or form a sub-
group of them. Their selections of nouns and their choices of names for
them reflect their predominant interests, and it will be helpful to have a look
at these before I explain and justify my own choice of terminology in more
detail.
One group of authors already referred to have focused on the semantic
generality or unspecificity (see also Section 5.2) of such nouns as fact, idea
or thing. In addition to Bolinger (1977) and Halliday and Hasan (1976),
Winter (1992) must be mentioned here, who uses the term unspecific nouns.
Although semantic unspecificity is also highlighted by Halliday and Hasan's
term general noun (1976: 274) for such nouns as people, person, creature,
thing, object, stuff, affair, matter, move, place, question and idea, it is well
known that Halliday and Hasan's main interest is the contribution of these
nouns to the cohesion of texts.6
This aspect is also the focus of Francis' (1986) notion of anaphoric
nouns (or -nouns for short). Francis uses this term to refer to nouns which
can function as anaphoric pro-forms, can be used "metadiscursively" within
a discourse and "are presented as the given element within a clause con-
taining new information" (1986: 7). Building on work by Winter (1977: 2)
and Hoey (1979) on lexical signalling, Francis supports this function with
the image of signposts: -nouns are linguistic signposts which signal to the
reader that the specific information can be found somewhere else in the text
(1986: 2). Among the nouns that meet these criteria are nouns derived from
speech act verbs, e.g. accusation, claim, comment, conclusion, declara-
tion, judgement, report and suggestion, other nouns describing verbal ac-
tivities, e.g. controversy, critique, eulogy, implication, nonsense and para-
dox, and metalinguistic 'text' nouns such as chapter, excerpt, phrase,
The term shell noun 11
quotation and word. Also included are 'cognition' nouns (1986: 14-16), for
example analysis, concept, doubt, idea, inference, perspective, view and
viewpoint. So-called 'ownerless' nouns such as aspect, context, fact, issue
and problem are seen as borderline cases with regard to the criterion of
metadiscursivity, while cause, development, move, process and result are
not treated as -nouns because they do not fulfill this criterion. To give an
early impression of what is included in the class of shell nouns, it will be
helpful to note that with the exception of metalinguistic 'text' nouns, all
examples quoted in this paragraph can be used as shell nouns.7
In Germany, the textual functions of abstract nouns were also recognized
very early. Krenn (1985: 133-138, 212-224), for example, draws attention
to the metacommunicative and anaphoric potential of English abstract
nouns. Chiefly interested in the items this, that and it in extended reference,
she also discusses anaphoric noun phrases headed by general nouns like
thing, matter, point or question as lexikalische Verweise 'lexical refer-
ences'. The earliest account of the textual function of abstract and general
nouns that I am aware of can be found in Raible (1972). Writing about
French and German abstract nouns, Raible speaks of Wiederaufnahme auf
Abstraktionsebene 'reiteration on an abstract level' and Wiederaufnahme
auf metasprachlicher Ebene 'reiteration on a metalinguistic level' (1972:
150-151) and argues that nouns like case, process, manner and condition
should play an important role in any theory of texts. Much more recently
but also in German, Koeppel (1993) discusses what he calls satzbezogene
Verweisformen 'sentence-related forms of reference'. Koeppel's study is
interesting and illuminating, because he also takes an empirical and func-
tional approach and works with different text-types. It is less helpful for the
present study than it could be, however, because Koeppel (1993: 43) ex-
plicitly excludes cases in which shell nouns and shell contents (in my termi-
nology) are connected by structural means.
As noted in Chapter 1, Vendler exploits the container-image in his study
of nouns denoting facts and events. Nouns like fact, result, reason, cause,
axiom and idea are understood as container nouns (1968: 72-82) because
they can act as central parts of clauses which function as containers, or
hosts, for other nominal clauses (see also Vendler 1967: 122-146). Two of
Vendler's examples are "That he died is a fact" (1968: 73), where the noun
fact (together with the copula is) acts as a container for a clause, and "It is
an axiom that all men are equal" (1968: 77), a case of extraposition in-
volving an abstract noun, which Vendler sees as another variant of a con-
tainer noun structure.
12 Approaching shell nouns
Ivanic (1991), who speaks of carrier nouns, exploits a related image.

Apparently the term carrier has a double motivation in Ivanic's paper. On
the one hand, she argues that the nouns in question "frequently carry a spe-
cific meaning within their context in addition to their dictionary meaning"
(1991: 95; my emphasis). On the other hand, the term carrier is used to
underline the affinity to Halliday's Systemic-Functional Grammar, where
the term Carrier is used to refer to the subjects in one of two types of
clauses which express relational processes (Halliday 1994: 120-122).8
Halliday himself also deals with nouns that can function as shell nouns.
They crop up in his discussion of "projections", i.e. constructions in which
"a clause comes to function not as a direct representation of (non-linguistic)
experience but as a representation of a (linguistic) representation" (1994:
250). According to Halliday, "nouns that project belong to clearly defined
classes" (1994: 263), a rather categorical claim which does not receive sup-
port from the present study. As the quotation indicates, Halliday looks at
the way these nouns occur from the grammatical rather than the lexical
perspective. What he regards as projections involving nouns are "embedded
projections", i.e. the grammatical pattern noun + postnominal /za-clause,
infinitive clause or gerund (in traditional terminology). As a consequence of
this focus, he overlooks the fact that the particular nouns that can occur in
these constructions ("function as Things in embedded projections") share
other properties, most notably the capacity to occur in other frequently re-
curring grammatical patterns, and a number of highly interesting linguistic
and cognitive functions (see Part III).
In a second paper Francis (1994) uses the term label for a group of
nouns which largely overlaps with her earlier set of -nouns. Her new
choice of terminology reflects the recognition (already present in her 1986
study) that such nouns as argument, point or statement are used to label
stretches of discourse in a certain way. A second important property of both
-nouns and labels is the potential to "encapsulate", as Francis (1986: 36-
38, 1994: 85) calls it, stretches of discourse. The image of encapsulation,
which is taken over from Sinclair (1981: 76), corresponds to my idea that
shell nouns create conceptual boundaries by casting larger chunks of infor-
mation into nominal structures.
The image of encapsulation is also taken up by Conte (1996). Although
Conte's article is very short, it provides many fundamental insights into the
use and functions of encapsulating nouns. The fact that there is a close cor-
respondence between the points mentioned by Conte and what I have been
The term shell noun 13
able to find out about shell nouns independently before I became aware of
her work supports the findings of this study.
Looking at the collection of terms and the different images underlying
such notions as containment, signalling, pointing and encapsulating, I think
that with a little stretch of imagination the shell metaphor incorporates all
these aspects. Clearly, one of the main functions of shells in the real world
is to contain something, to act as host and shelter for things that would oth-
erwise easily be dispersed or damaged. This encapsulating function seems
to be particularly important if one starts to switch over to the linguistic
domain, which the metaphor of shells is to illuminate. Discourse without
shell nouns can be compared to an egg-and-spoon race using eggs without
shells. One would not be able to get on in discourse (and in the race), if it
were not for the encapsulating function of shell nouns (or egg shells). In
other words, shell nouns can supply propositions with conceptual shells
which allow speakers to grab them and carry them along as they move on in
discourse. Shells also act as signals for their content. Looking at various
types of shells, say an egg shell, a nutshell, a snail shell and the shell of a
mussel, one always knows what type of thing is inside. In the same way
shell nouns also function as labels for their content, as indicators of what
they contain.
2.2 Defining shell nouns and shell-content complexes in functional

terms
Analogies such as the metaphor of shells may help to explain abstract no-
tions but they must not be misused as substitutes for definitions. If one
looks at the publications by Francis (1986 and 1994) and Ivanic (1991),
who have provided the most detailed accounts so far of the phenomena in
question, one finds that both authors seem to struggle with the definition of
their subject-matter. Why are -nouns, carrier nouns and shell nouns so
hard to define? The reason is that they are not defined by inherent properties
but constitute a functional linguistic class. This means that whether a given
noun is a shell noun or not does not depend on inalienable characteristics
inherent in the noun, but on its use. A noun is turned into a shell noun when
a speaker decides to use it in a shell-content complex in the service of cer-
tain aims. The property of shell-nounhood is thus a functional property. The
right way of thinking about shell nouns is as particular types of uses of
certain nouns, rather than as shell lexemes in their own right.
This functional definition of shell nouns has a number of fundamental

consequences. First, the list of shell nouns given in Table 1.1 in Chapter 1 is
actually quite misleading because it suggests that these nouns are shell
nouns as such, whereas in fact they are only nouns that are very frequently
used as shell-nouns. Second, it is impossible to give an exhaustive list of
shell nouns because in suitable contexts, many more than the 670 nouns
discussed in this study can be found in shell noun uses. As will be explained
in Section 4.2, the choice of nouns that will be considered here is based on
objective syntactic criteria. Finally, the class of shell nouns is highly hetero-
geneous, both from a semantic point of view and with respect to how good
an example of shell-nounhood a given use of a noun is (see Section 6.1 for a
discussion of the typicality gradient in the class of shell nouns). The typi-
cality gradient of shell nouns also affects the boundaries of the class, which
are fuzzy rather than clear-cut. It will be seen in Chapter 12, for example,
that although circumstantial nouns like time, place and way can indeed
function as shell nouns, they mark a transition zone between shell noun uses
and other uses of nouns.
What, then, are the functions that define uses of nouns as shell nouns?
What do the nouns allow speakers to do? A whole array of more or less
specific functions will be identified in Part III. Three functions, however,
stand out from the rest because they can be seen to play a role in all uses of
shell-content complexes. As a consequence, these three can be used to de-
fine the functional class of shell nouns:
1. Shell nouns serve the semantic function of characterizing and per-
spectivizing complex chunks of information which are expressed in
clauses or even longer stretches of text.
2. Shell nouns serve the cognitive function of temporary concept-
formation. This means that they allow speakers to encapsulate these
complex chunks of information in temporary nominal concepts with
apparently rigid and clear-cut conceptual boundaries.
3. Shell nouns serve the textual function of linking these nominal con-
cepts with clauses or other pieces of text which contain the actual de-
tails of information, thereby instructing the hearer to interpret different
sections of a text together (see Section 3.1.2).
In view of the feet that many linguistic items have the potential to charac-
terize, form concepts and/or link pieces of text, it must be emphasized that
shell nouns fulfill these functions in a very special way. In order to demon-
Defining shell nouns in functional terms 15
strate this, it will be helpful to compare shell nouns to full-content nouns on

the one hand, which can be seen as best examples of characterizing and
concept-forming linguistic items, and to anaphoric elements such as the
personal and demonstrative pronouns on the other, which are arguably
among the best examples of nominal linking items. The idea for this com-
parison originates from Ivanic (1991), but the dimensions used as criteria
are my own. Examples of the three types of words are given in (2.1):
(2.1) (a) Full-content nouns: teacher, cat, journey

(b) Shell nouns: fact, problem, idea, aim
(c) Pronouns with anaphoric function: she, it, this, that
Characterization
Full-content nouns have an enormous potential for detailed characteriza-
tions of what speakers want to talk about. The reason is that nouns like
those listed in (2. la) have a more or less stable and rich denotation. Due to
their specific and fully-fledged meaning, full-content nouns and other open-
class items such as adjectives and verbs are the main means of describing
persons and objects, animals and plants, activities and events, and proper-
ties and circumstances.
Pronouns with anaphoric function, on the other hand, have a very limited
potential for characterization, if any at all. The personal pronouns I, you,
he, she, it, we, they, for example, characterize their referents only with re-
spect to a very small number of semantic dimensions: speaker vs. addressee
vs. other role, human vs. non-human, singular vs. plural, and male vs. fe-
male. The demonstratives this and that characterize a piece of experience
only with regard to the dimensions of spatial (and/or emotional) proximity
and singular number.
Shell nouns hold a middle position between these extremes. To a certain
extent, speakers can indeed use them to characterize a piece of experience,
say as a fact, a problem, an idea or an aim. Like full-content nouns, shell
nouns derive their potential for characterization from their denotation. The
nouns stand in a relatively stable relation to a recurrent type of experience,
just like the noun cat stands in a stable relation to a category of entities in
the concrete world. Yet, as will be shown in detail in Chapter 5, nouns that
can be used as shell nouns typically have abstract and unspecific meanings.
As a result, speakers can only use them to characterize a piece of their ex-
perience in a fairly general way, while the details of information must be
expressed as shell content in the context. For example, when a noun like
aim is used, the meaning of the noun itself includes no information about
the precise details of what somebody is aiming for. In this respect, shell
nouns are similar to anaphoric pronouns which depend on contextual infor-
mation for their interpretation.
Concept-formation
When a word is used repeatedly to refer to a certain type of experience, the
recurrent association between the linguistic form and the idea results in the
formation of a more or less stable concept. Essentially, the resulting con-
ceptual relation corresponds to Saussure's model of the sign. It is with the
process of establishing this relation in mind that Leech (1981: 32) speaks of
the "concept-forming power of the word".
Nouns denoting classes of persons, animals, organisms and concrete ob-
jects lend themselves readily to the formation of concepts. On the basis of a
naive view of the world, which corresponds to the philosophical position
called "realism" by Lyons (1977: 109-114), we tend to think that words are
no more than names for categories of things. Given the apparent similarity
of the things that belong to one category of concrete individuals, this idea
comes so naturally that the role of words in the formation of concepts is
hardly noticeable. The way in which words contribute to the formation of
concepts can be illustrated better with nouns denoting abstract entities and
with nouns denoting events, for instance a noun like journey (cf. Leisi 1975:
26). The naive view of words suggests that there is a class of experiences
which exists readily packaged somewhere out there and is simply named by
the word journey. A closer examination, however, shows that what can be
referred to by the word journey is a fairly complex matter. It can involve
many different actions such as checking in at an airport, sitting in car, on a
coach or train, walking through the jungle or hitching a lift somewhere in
the middle of nowhere. It is not even easy to define when a journey starts
and where it ends. If you travel from your home to another city, does your
journey start when you leave your house, when you step into a taxi or when
you board the train at the railway station? Despite the variety of experiences
that can be referred to as a journey and the vagueness of the boundaries of
journeys, the word journey gives us the impression that there is one neatly
bounded class of entities or experiences which we have in mind whenever
we use it. But this is of course not the case.
Nor is it the case with more abstract words such as love, inflation or
democracy. Again, the words suggest that there are things existing inde-
pendently of the human mind, which are simply named or labelled by the
words love, inflation and democracy. But yet again, this is an illusion.
What people talk about when they use the word love can be a large variety
of different types of experiences with an enormous range of different mani-
festations, and the same is true of the notions of inflation and democracy.
Nevertheless we tend to think that the words love, inflation and democracy
stand for ready-packaged, autonomous, even almost substantial entities
(witness the frequent personifications, particularly of the first item).
In short, the notion of concept-formation really captures a combination
of two illusions: first, that a word stands for one single entity which is
neatly bounded, and second, that this neatly bounded entity has a thing-like
quality with a substance of its own. It refers to the combined illusion of
encapsulation and reification.9
Although all types of open-class words contribute somehow to the for-
mation of some kind of concepts, the concept-forming power of words has
its strongest effects with nouns. The most prototypical examples of nouns,
namely nouns denoting classes of concrete entities, refer to things. As a
consequence, they lend themselves much more readily to a conceptualization
of what they stand for as 'things' and this greatly encourages the illusion of
reification. But even nouns denoting abstract entities, relations or properties
are affected by this illusion. For example, the adjective round evokes a
property of things, i.e. a relational concept, while the derived noun round-
ness seems to evoke a 'thing'. Similarly, the verb assume stands for a men-
tal process, while the derived noun assumption again seems to stand for a
'thing'. This recognition forms the basis of Langacker's cognitive concep-
tion of word classes (1987a: 183-213, 1987b), and it will emerge that it also
plays an important role for the issue of shell nouns (see Sections 5.1.2 and
17.2).
How do the three types of linguistic elements behave with regard to con-
cept-formation? Disregarding such phenomena as polysemy and vagueness,
we find that full-content nouns have a relatively constant relationship to the
experience they encapsulate as a concept. This is due to their stable denota-
tion. Although we know at least since Labov's work (1973, 1978) that the
boundaries of categories of concrete entities are fuzzy, his experiments have
also shown that such categories do have conceptual boundaries and that
they are relatively strong, too.
Deictics, on the other hand, exhibit virtually no such concept-forming ef-
fects. Personal pronouns stand in for other instantiations of concepts which
are explicitly mentioned elsewhere. And demonstratives functioning as de-
terminers with nouns specify the reference of particular expressions. Neither

contribute to the formation of a concept. When the pronouns it, this or that
are used in extended reference or text reference (Halliday and Hasan 1976:
52-53, 66-67), for example in utterances like it helped a lot or I didn't say
that, it is also impossible to regard them as instantiations of stable concepts.
Whatever it is that is being referred to is not bounded as a concept stored in
the mental lexicon. Instead, the semantic impact of such anaphora is com-
pletely context-dependent.
Again, shell nouns stand between the two opposing poles. Like full-
content nouns, they exhibit a constant conceptual relationship to a specific
recurrent type of experience, to problems, opportunities, reasons, facts and
so on. And, being nouns, they create the impression that the types of experi-
ences they encapsulate as concepts are 'things' or, more precisely, instances
of classes of 'things'. On the other hand, the concepts created by shell
nouns are also very variable. They are of a temporary nature because their
content changes with the situational and linguistic context10 in which they
are used. So the concepts created by shell nouns consist of a stable symbolic
and a variable indexical part.
Linking
Pronouns with anaphoric function are among the best examples of nominal
linking elements because they instruct hearers to interpret two groups of
linguistic elements together, as being related to and even dependent on each
other. One can observe such links in their clearest form in cases of ana-
phoric personal pronouns, which have been thought of as creating links of
referential identity or co-reference (cf. e.g. Halliday and Hasan 1976: 309).
Viewed in isolation, full-content nouns have hardly any potential to cre-
ate cohesive links (apart from the semantic relations described by Halliday
and Hasan (1976: 274) as lexical cohesion; see also Hasan 1984 and Hoey
1991). Because of their more or less specific denotation and the resulting
potential for characterization, full-content nouns are better suited for exo-
phoric reference to the world outside a text.
With regard to linking, shell nouns are more similar to anaphora than to
full-content nouns. Like anaphora, shell nouns cannot thrive without textual
links. Their interpretation crucially depends on the shell content which must
be expressed in the context, or at least be inferrable from it. I will show in
Section 3.1.1 that the links between shell nouns and their contents are usu-
ally supported by specific lexico-grammatical patterns, and in Section 5.2
that the semantics of the nouns themselves require these links and contribute
to a large degree to their establishment.
In sum, shell nouns seem to be a unique linguistic phenomenon for two
reasons. First, they combine the three functions of characterization, con-
cept-formation and linking, which are otherwise performed separately, each
by different types of linguistic elements. And second, they perform these
functions in a fine-tuned balance between conceptual stability and informa-
tional flexibility. These two aspects, the combination of functions on the one
hand, and the balance between opposing poles on the other, are illustrated in
Figure (2.1).
potential for
concept-formation
potential for
concept-formation
Figure 2.1 The converging balance of shell nouns
The figure suggests that shell nouns hold a central position on the dimen-
sions of stability of evoked concept, potential for characterization and po-
tential for linking elements in a text. These functional properties are greatly
facilitated by the type of semantic structure that is unique to shell nouns
(see Chapter 5). It is this combination of stability and flexibility that turns
shell nouns into such powerful communicative and cognitive tools.
The term shell noun is employed in this study to refer to uses of nouns
which meet the combination of criteria laid down in this section. However,
as will be shown in greater detail in Chapter 5, certain types of nouns lend

themselves more readily and regularly to such uses than others. I will use
the term shell noun with an intended and systematic ambiguity: for nouns
(qua lexemes) which have the systemic, langue-related potential to fulfill
the specific combination of functions described here, as well as for actual
uses of these nouns in these functions, i.e. parole phenomena. The fact that
shell nouns always occur in the functional units of shell noun phrases,
which may even include the shell contents as postmodifying clauses, is
taken for granted. I will also use the derived verbal expression of shelling a
piece of information. Thus by saying that the noun phrase the Govern-
ment's aim in example (1.3) shells the information expressed in the com-
plementing infinitive clause, I want to convey that the noun phrase is linked
to this clause, characterizes the information given in it as an aim, and
achieves that this information is temporarily turned into a context-dependent
concept with a thing-like quality.
2.3 A brief note on the theoretical stance
This study is not consistently set within a single theoretical framework. As

has already emerged from the considerations and arguments put forward so
far, the underlying view of language is cognitive and pragmatic.
On a very general level, this means that I assume that questions like
"Can the use of shell nouns be explained on the basis of general cognitive
abilities?", "How are expressions involving shell nouns processed?" and
"Why are expressions involving shell nouns used?" are interesting and
worth pursuing, and that arguments like "shell nouns and shell contents
activate components of one cognitive model" or "we use shell nouns be-
cause they help us to draw attention to certain aspects of events and states
of affairs" are legitimate. The theoretical stance is eclectic rather than
monolithic, mainly in order to avoid lengthy theoretical introductions and
justifications which would be indispensable if one theory were chosen as a
basis. The method builds on the ideas and achievements of corpus linguists.
The grammatical framework and terminology are taken from 'traditional'
descriptive grammar unless Cognitive Grammar (see Langacker 1987a,
1991) or Systemic-Functional Grammar (see Halliday 1994) provide more
appropriate concepts or terms. The semantic descriptive apparatus consists
of semantic features similar to those used in structural semantics and of
frames as used in cognitive semantics.
Chapter 3
The links between shell nouns and contents
This chapter looks more closely at a number of syntactic, semantic and

pragmatic aspects of the links between shell nouns and shell contents. In
Section 3.1.2,1 will argue that a relation which I call experiential identity
constitutes the semantic and cognitive basis of the four major linguistic
means of linking shell nouns to their contents. Since one of these links, the
combination of a shell noun with a postnominal clause representing the shell
content, exhibits highly conspicuous similarities to the complementation of
verbs and adjectives, the linguistic literature on the latter issue will be
sieved for relevant findings in Section 3.2. Finally, Section 3.3 provides a
brief account of the basic functions that can be attributed to the links de-
scribed in Section 3.1.1. Although functional considerations will take centre
stage in Part III, this basic account will prove to be useful in the descriptive
second part as well.
3.1 Triggering co-interpretation
3.1.1 Lexico-grammatical patterns of shell-noun uses
It is vital for the communicative success of shell nouns that they are inter-
preted together with their content. Speakers trigger such a co-
interpretationn by means of a fairly small number of linguistic devices. The
lexico-grammatical patterns they use to link shell nouns to their contents
and the semantic relations underlying them will be discussed in this section.
My previous research into the use of one typical example of a shell
noun, the noun idea (Schmid 1993: 165-219, 1997) and my long-standing
interest in other shell nouns suggest that these nouns are mainly used in four
types of lexico-grammatical patterns. These are given in Figure 3.1 and
illustrated by short examples taken from the COBUILD corpus. In the middle
column of the table, abbreviations both for the four general patterns and
their more specific variants are introduced, which will be used in the rest of
the study.
22 The links between shell nouns and contents
Pattern Abbreviation Example of the general pattern
Shell noun + postnominal N-cl (3.1) Mr Bush said Iraq's leaders

clause had to face the fact that the rest
Variants: that-clause -that of the world was against them.
to infinitive-clause -to (BBC)
w/z-clause -wh
Shell NP + be + comple- N-be-cl (3.2) The advantage is that
menting clause there is a huge audience that can
Variants: that-clause N-be-that hear other things you may have
to infinitive-clause -be-to t o s a y , (PAPERS)
wA-clause N-be-wh
Referring item + (premod) + th- (3.3) (Mr Ash was in the clearest
shell noun possible terms labelling my
clients as anti-semitic.) I hope it
is unnecessary to say that this
accusation is also completely
unjustified, (PAPERS)
Referring item as subject + th-be-N (3.4) (I won the freshmen's
be + shell noun (phrase) cross-country. - Mm.) That was
a great achievement wasn't it?
(SPOKEN)
Note: The abbreviations of the corpus sections, i.e. BBC, PAPERS and SPOKEN, which are
also used in all further examples, are explained in Section 4.2.
Figure 3.1 Lexico-grammatical patterns favoured by speakers for the use of

shell nouns
For the last three of these patterns, the syntactic structures and the map-
pings of clause constituents on shell nouns and shell contents are fairly
straightforward. In the pattern N-e-cl, the shell-noun phrase occurs as
subject in a SVC-clause with the linking verb be, in which the shell content
is embedded as a that-, wh- or infinitive clause functioning as subject com-
plement (the respective abbreviations are -be-that, -be-wh and -be-to).
In the pattern th-N, the link between shell noun and shell content is created
by the potential of a number of linguistic elements, mainly the, this, that,
other, same and such, to establish demonstrative or comparative anaphoric
reference (Halliday and Hasan 1976: 57). In this pattern, the shell-noun
phrase can fulfill various syntactic functions. In the pattern th-be-N, the
Triggering co-interpretation 23
link between shell noun and shell content extends over three groups of ele-
ments. The pronouns this, that or it mediate between the passage of text
which actually expresses the shell content and the shell noun. These pro-
nouns occur in the subject position at the beginnings of SVC-clauses and
refer back to the shell contents by means of what Halliday and Hasan call
"extended reference", "text reference" or "reference to fact" (1976: 52-53,
66-67). They transfer this reference via the linking verb be to the shell noun
phrase which functions as subject complement in the clause structure. In a
way, then, the pattern th-be-N is a blend of the copular type N-e-cl and the
anaphoric type th-N.
The syntax of the pattern N-cl, i.e. the combination of an abstract head
noun and a following that-clause (N-that), to-infinitive (N-to) or w/j-clause
(N-w/j), is much less straightforward. What is clear is that the postnominal
clauses express the shell contents in these patterns. The syntactic relations
between these clauses and the head nouns, however, and the syntactic status
of the clauses in particular, are all but clear. Their analysis depends on two
factors: the grammatical framework applied and the type of noun that oc-
curs as head noun. As already mentioned in Chapter 1, Quirk et al., for
instance, regard both noun complements and appositive postmodifiers as
possible functions of the clauses. Thus they give the likelihood that Joan
will get married (1985: 1231) as an example of "noun complementation"
and the highly similar expression the belief that no one is infallible (1985:
1260) as involving an appositive postmodifying clause. This may seem
somewhat paradoxical, especially if one thinks of complements as giving
necessary, and of appositions as giving non-necessary, additional informa-
tion, as many linguists have traditionally done. It must be added, however,
that Quirk et al. (1985) allow for cases of restrictive appositive clauses.
While Biber et al. (1999: VIII) explicitly state that they have borrowed
"the grammatical framework of concepts and terminology" from Quirk et
al. (1985), they do not mention the possibility of treating the clauses fol-
lowing abstract head nouns as appositive postmodifiers. Instead they con-
sider all examples of the type the idea that he was completely cold and
unemotional and a chance to do the right thing (Biber et al. 1999: 575) to
be complement clauses, which, according to them, "are distinct from post-
modifiers in structure and meaning" (ibid.). They argue that "complement
clauses differ from postmodifying clauses in that they do not have a gap
corresponding in meaning to the head noun" (1999: 645), which is for ex-
ample filled by relative pronouns in relative clauses. As a result, comple-
ment clauses can stand on their own as independent clauses, while post-
modifying clauses can not. However, this apparently simple test runs into
difficulties with to-clauses following abstract head nouns, because these, as
Biber et al. (1999: 645) admit, have missing subjects just like postmodify-
ing -clauses and can only be separated from the latter on semantic
grounds. According to Herbst (1988: 269), who, like Hudson (1984: 263-
264) and linguists from the transformational paradigm (see e.g. Radford
1997), also works with the notion of noun complementation, there are two
criteria for determining the status of complements: the existence of co-
occurrence restrictions between the complement and the noun, and the de-
pendence of the form of the complement on the noun. My own impression is
that even when these criteria are applied, a strict general separation between
postmodifiers and complements is impossible if one considers the whole
range of possible sequences of abstract nouns followed by clauses. There-
fore I use the neutral term postnominal clause in this study.
A brief overview of the possible types of postnominal clauses will be in
order here. To start with, there is a superficially similar construction in-
volving the so-called extraposition of a clausal subject and the insertion of
an anticipatory subject it. This is illustrated in example (3.5):
(3.5) But it is a good idea to stop and think about it. (PAPERS)
In such sentences, there can be no doubt that the clauses representing the
shell contents do not belong to the noun phrases which are headed by the
shell nouns. Instead they make up the notional subjects of the clauses, which
are moved to the end of the sentences. This can be demonstrated by trans-
forming (3.5) into (3.5'):
(3.5') To stop and think about it is a good idea.
For many shell nouns, these "canonical" patterns (Quirk et al. 1985: 1392)
are less common than the postponed ones. What the transformation in (3.5')
shows is that such uses are similar to the patterns N-e-cl and th-be- in
that they also revolve around the copula be. In feet, they can be seen as
variants of the pattern th-be-N, in which the shell contents are not taken up
by an anaphoric pronoun but mentioned explicitly in the same clause.
A related pattern, especially from a pragmatic and rhetorical perspective,
is the existential-i/zere construction, which is illustrated in (3.6):
(3.6) Agnelli is due to step down soon as head of Fiat and already there is
speculation that he might move into politics, (MAGS)
Such sentences can also be traced back to canonical intensive clauses but,
as in this particular example, transformations are usually only possible if
one introduces a definite article. The resulting paraphrase is given in (3.6'):
(3.6') The speculation is that he might move into politics.
These examples are not particularly good examples of shell-content com-

plexes because indefinite noun phrases do not create as strong conceptual
boundaries as the definite noun phrases in which shell nouns tend to occur.
Since these existential-êre constructions typically involve nouns that are
morphologically related to verbs, there is a second parallel, the parallel to
verbs with complement clauses. A paraphrase of (3.6) along these lines is
given in (3.6"):
(3.6") ... and people are already speculating that he might move into politics.
Similar affinities to simple verbs can be observed with occurrences of

nouns in so-called "expanded predicates" (Algeo 1993), which consist of
function verbs like have or make and nouns. Typical collocations of this
type are for example have the idea/ feeling/impression that, have the job/
task/duty to, but also the more specific express relief/regret/concern that.
While these combinations of expanded predicates with that-clauses and
infinitives obviously resemble the pattern N-cl and thus support the idea
that the clauses following them must be seen as complements, they differ in
one important respect. The effect of the temporary formation into a nominal
concept is not very marked here because the whole expanded predicates
function as verbs and can normally even be replaced by simple verbs. Often
the motive for their use is only a stylistic one (see Section 5.1.2). These uses
can therefore not be regarded as good examples of shell nouns either.
In combinations of nouns with adjacent infinitive clauses, the analysis of
the clauses crucially depends on the nouns. With modal nouns like ability,
chance, intention, need, opportunity or willingness the analysis of the
postnominal clauses as complements, rather than postmodifiers, is most
convincing since the nouns seem to determine the fom of the complements.
Difficulties arise with nouns which have no inherently modal meaning, for
example with temporal or locative nouns. Example (3.7) is a case in point.
(3.7) Britain is a great place to live and work in ... (MAGS)

Traditionally the infinitive clauses in such examples as (3.7) have been

traced back to relative clauses, as is reflected in the paraphrase given in
(3.7').
(3.7') a great place in which/where it is possible to live and work
Examples of this type are therefore also treated as highly marginal instances
of shell nouns (see Sections 6.1 and 12.1).
Finally, there is a transition zone between complementing or appositive,
relative and adverbial postnominal clauses in examples like (3.8) and (3.9),
which consist of nouns with circumstantial meanings and postnominal wh-
clauses:
(3.8) The lists of Indonesian communists were compiled by the CIA and State
Department over two years, at a time when Washington believed that
there was a real threat that southeast Asia would fall tinder communist
rule, (BBC)
(3.9) Is there any place where you can go and play snooker or anything like
that? (SPOKEN)
Cases of this type tend to be analysed as variants of relative clauses as well.

Biber et al. (1999: 626-630), for example, treat them as "head nouns taking
relative clauses with adverbial gaps". I will return to the syntactic analysis
of such examples when I look at circumstantial uses of shell nouns (see
Sections 12.1 and 12.3).
This closes the discussion of the pattern N-cl. A fifth pattern that is not
included in Figure 3.1 is the combination of abstract head nouns with post-
modifying o/-prepositional phases, as in the problem of raising money, the
idea of going out or the question of where to go. For two reasons, this fifth
pattern is not treated on a par with the other four. For one thing, it is re-
stricted to a much smaller group of nouns than the other four patterns. And
secondly, as will be shown in Section 4.2, this pattern does not lend itself as
readily as the others to the computer-aided systematic retrieval of linguistic
data from a large corpus. The pattern will therefore be taken into consid-
eration and included in the descriptions of the nouns which occur in it, but
not examined quantitatively.
It is quite remarkable that a fairly large and heterogeneous set of nouns
should favour such a small set of lexico-grammatical patterns.12 And what
is perhaps even more extraordinary is the fact that these patterns do not
seem to share a syntactic or semantic basis which could be used to explain

why the nouns are used predominantly in them. On closer inspection, how-
ever, such a common basis can be found. I will try to show in the next sec-
tion that a relation of what I call experiential identity between shell noun
and shell content is evoked by most uses of shell-content complexes.
3.1.2 From identity of reference to experiential identity
The idea that identity, i.e. the notion that the shell noun and the shell content
express ideas about the same thing, actually plays a role here is perhaps
most convincing for instantiations of the pattern N-Ae-cl, because here shell
noun and shell content are linked by a form of the verb be. Equative expres-
sions of the type A is clearly suggest that A and are identical. This also
holds true when the first element of the equation is expressed by a noun
phrase such as the advantage, as in example (3.2) in Figure 3.1, and the
second by a f/zai-clause. That SVC-clauses with linking verbs have to do
with identity is acknowledged in one way or another in many schools of
grammatical thought. Quirk et al. (1985: 741) describe the semantic role of
subject complements as that of "ATTRIBUTES", which can either identify or
characterize the subject. In the construction N-e-cl, the semantic relation is
identification. In Halliday's Systemic-Functional Grammar such clauses are
analysed as intensive relational processes in the "identifying mode" on the
level of clause as representation (Halliday 1994: 122-124). The experiential
structure of these relations consists of two elements, the IDENTIFIED and the
IDENTIFIER, and the identifying relation between them. Although shell-
content complexes are a special case of identifying relations because the
IDENTIFIER is a rank-shifted clause, this does not affect the status of the two
components involved and the relation between them. To mention just one
further example, Langacker, within his framework of Cognitive Grammar,
calls such sentences "equational" and also regards identity as the underlying
semantic relationship (Langacker 1987b: 77).
Anaphoric links (pattern f/?-N) such as those triggered by the demon-
strative determiner this in example (3.3) in Figure 3.1 have also been inter-
preted as being based on a relation of identity, namely identity of reference
(Halliday and Hasan 1976: 308).13 The central idea, which is dubbed the
"substitution view" of anaphora by Brown and Yule (1983: 201), is that
anaphoric personal pronouns and noun phrases with anaphoric determiners
simply replace other groups of linguistic elements which "refer to the same
thing" (Halliday and Hasan 1976: 314). Although especially for cases of
anaphoric personal pronouns this idea is intuitively convincing, one cannot
apply it directly to shell-content complexes of the type th-N. From the
standpoint of the traditional philosophical and semantic view of reference,14
the main problem with anaphoric expressions containing shell nouns is that
the antecedents of the anaphora are normally not referring expressions but
clauses, extended stretches of discourse, or even pieces of information
which must be inferred from the context, as in example (3.3) in Figure 3.1.
If, as in this case, one of the two expressions involved does not have the
potential for reference at all, identity of reference is out of the question and
the substitution view runs into serious difficulties. 15
The same problems affect the pattern th-be-N, since it is implied in the
definition of extended reference and text reference that the target of the ana-
phoric item must not be a referring expression. In example (3.4) in Figure
3.1 for example, the demonstrative that refers to the event that the first
speaker has won a particular cross-country race. (It should be noted in
passing here that in the pattern th-be-N, the semantic relation between the
subject and the subject complement is not, as in the pattern N-e-cl, an
identification but a characterization).
From a philosophical and logical point of view, then, identity of refer-
ence cannot be accepted as the relation underlying the link between shell
noun and shell content in the patterns th- and th-be-. Strangely enough,
both anaphoric patterns nevertheless strongly suggest that some sort of
identity between shell noun and shell content is involved after all. This in-
tuition has presumably influenced Halliday and Hasan's thinking and it also
reflects the unspoiled view of the naive language user, which should be the
basis for a genuinely cognitive view of language. In order to be able to ac-
count for this intuition, however, Halliday and Hasan's (1976: 31-37) rigid
distinction between exophoric and endophoric reference, which lies at the
heart of the whole dilemma, must be replaced with a more cognitively-
oriented view of reference. According to such a view, items with referring
potential are seen as being related neither to the text itself nor to the world
outside the text but to the cognitive models that are created in the minds of
language users. All these items contribute to the activation or reactivation of
components of such models.16
It should be mentioned here, if only in passing, that such a step has far-
reaching consequences from a philosophical point of view, because all
questions concerning the truth of propositions become virtually irrelevant. It
is no longer important whether some state of affairs holds true in 'objective'
reality, but only whether it is represented in the cognitive models of situa-

tions that participants activate or create.17 From a linguistic point of view it
must be emphasized that this integrative cognitive view of reference levels
out a number of traditional distinctions, e.g. the contrast between anaphora
and deixis and the special status of text or discourse deixis. However, this
does not mean that the insights linguists have gained concerning these ques-
tions are no longer useful. They are only transferred onto a more finely-
grained level of linguistic and cognitive analysis.18
Various types of cognitive models related to and evoked by texts have
been postulated in the fields of semantics, text-linguistics, discourse analy-
sis, psycholinguistics and other cognitive sciences. Kallmeyer et al.'s (1974:
23 et passim) notion of Wirklichkeitsmodell has already been mentioned
above in note 16. Other examples are the notions of universe-of-discourse,
textual world, discourse representation, mental model, situation model,
mental space, frame, script and schema}9 I will stick to the more general
term cognitive model here (see Ungerer and Schmid 1996: 45-55) because
all the terms just mentioned invoke certain theoretical or even ideological
implications which would complicate the issue unnecessarily at this stage.
In accordance with a more or less explicit consensus in the linguistic and
psycholinguistic literature I assume that cognitive models contain three
basic types of information, namely concepts or components, attributes, and
relations (see e.g. de Beaugrande and Dressier 1981: 84-90, Prince 1981:
235, van Dijk and Kintsch 1983: 344-346). The cognitive model evoked by
a text is understood as a mental representation of all the people, organisms
and objects, as well as the events, states of affairs, settings and other rela-
tions involving them, which are either mentioned in a text or suggested to be
inferred from the information given in it.
Subsuming reference and anaphora as well as deixis under the idea of
activation of components of a cognitive model solves the problem inherent
in the claim that shell nouns and the linguistic elements expressing the shell
content have to do with the same thing. On this highly general descriptive
level the link between shell nouns and shell contents is that they activate
identical or closely related components of a cognitive model. This co-
activation is the cognitive counterpart to the pragmatic concept of co-
interpretation and it is experienced by language users as experiential iden-
tity. In plain terms, experiential identity means that two or more separate
linguistic elements contribute to the formation of one thought.
So far, I have been able to show that experiential identity is the relation
holding between shell nouns and shell contents in the equative pattern -be-
cl and the anaphoric patterns th- and th-be-N (see Figure 3.1). This leaves
me with the pattern N-cl still to account for.
All clear and typical cases of shell nouns in the pattern N-cl also evoke
the impression that the nouns and the postnominal clauses are about the
same 'thing' or state of affairs. Intuitively, experiential identity is again at
work, then. This is particularly convincing when one regards the postnomi-
nal clauses as appositions.20 As argued by Quirk et al. (1985: 1300-1302),
the relation between linguistic units in apposition is identity of reference.
My previous discussion therefore also applies to the relation between ab-
stract head nouns and appositive postmodifying clauses. The relations of
apposition and experiential identity between the noun and the postnominal
clause can be tested by checking whether a matching form of the verb be
can be inserted between them (cf. Quirk et al. 1985: 1261) without distort-
ing the semantic relation. This is possible in example (3.1) in Figure 3.1
above: the paraphrase the fact is that the rest of the world was against
them is compatible with the original version the fact that the rest of the
world was against them.
Interestingly, Langacker explains such examples as (3.1) in terms of a
notion called "referential linkage" (1991: 432). This term is used to account
for constructions whose components "are construed as being identical"
(1991: 430). This clearly supports my view of the relation between shell
nouns and postnominal clauses. I prefer my own term experiential identity
to Langacker's referential linkage, however, because, as outlined above, I
find it problematic when the notion of reference is used to link noun phrases
and clauses.
The notion of experiential identity thus serves well to explain the links to
all clear examples of appositive postnominal clauses (which are treated as
complements by some scholars, as we have seen in 3.1.1). For some of the
cases outlined in Section 3.1.1, where the syntactic ties between nouns and
clauses are less strong, the notion of experiential identity becomes also less
clear. For example, while it can be argued that a great place in (3.7) above
refers to the same location as where it is possible to work, such an inter-
pretation constitutes a considerable extension of the notion of experiential
identity. Examples of this type are therefore treated as highly marginal in-
stances of shell nouns (see Sections 6.1 and 12.1). In (3.8), on the other
hand, at a time and when Washington ... clearly seems to refer the same
point in time, and in (3.9) any place and where you can go ... seems to refer
to the same (unknown) location.
Let me sum up briefly. I have started this section by explaining the four
lexico-grammatical patterns in which shell nouns are predominantly found. I
have then claimed that the links between shell nouns and their contents can
be traced back to the perception that the linguistic elements used to express
the two components are about one and the same piece of experience. For the
two patterns in which the link is based on anaphoric reference (th- and th-
be-N), it is impossible to account for this intuition on the basis of formal,
logical or language-immanent views of the notions of anaphora and refer-
ence. In contrast, a pragmatic and cognitive view of the relation is neces-
sary, and it is on this basis that I have argued for an underlying relation
called experiential identity. In the two patterns N-cl and N-e-cl, the link of
experiential identity between the shell nouns and their contents are created
by syntactic structures with identifying or equating meanings.
3.2 The semantic contributions of different types of complements:

a survey of the evidence from verbal complementation
Not just words, but also grammatical constructions have the systematic
potential to activate meanings. Accordingly, not just the shell nouns them-
selves and the words making up the shell content contribute to the combi-
natorial meaning of the shell-content complex, but also the types of clauses
through which the shell content is expressed in the patterns N-cl and N-6e-
cl. Although intuition suggests that that-clauses, infinitive clauses and wh-
clauses are not chosen at random by speakers but used to convey different
grammatical meanings, it is quite a challenge to isolate and identify these.
An enormous amount of linguistic research has gone into the study of the
forms and meanings of the clause types that are used as complements of
English verbs. Although the three types of clauses concerned in this study
of nouns also feature prominently in the verbal domain, one cannot take it
for granted that their usage is completely identical. For one thing, gerunds
play an important role in the complementation of verbs, in addition to that-
clauses and infinitives, but only a marginal one as prepositional comple-
ments in o/prepositional phrases in the complementation of nouns. This
means that the system of available options is fundamentally different. This
means that, as Quirk et al. (1985: 1231) put it, "the assumption of corre-
spondence cannot be automatic, for it may fail in both directions". Two of
their examples are given as (3.10) and (3.11) here:
(3.10) Joan is likely to get married.

Joan's likelihood to get married.
(3.11) *It is likely of Joan's getting married ...
the likelihood of Joan's getting married ...
Erring on the side of caution, I will therefore start out from what could be
called the zero-hypothesis that there is no direct correspondence between
the meanings of the three types of clauses in combination with verbs and in
combination with nouns. Nevertheless, it would be ridiculous to pretend that
nothing is known about the grammatical meanings of the complements. So I
will have a look at some findings that may help to understand the comple-
mentation of nouns - a highly selective look, since there is a veritable flood
of literature on the complementation of English verbs.21 In what follows, I
will concentrate on /a-clauses and infinitives because w/?-clauses can be
credited with two types of meaning without much further ado: they refer
either to unknown information (by virtue of their relation to direct ques-
tions) or to circumstantial information (by virtue of their relation to adver-
bial clauses).
While traditional grammarians like Poutsma (1929: 607-632, 763-992)
and Jespersen (1940) were of course also interested in the issue of the com-
plementation of verbs and adjectives, it was in the early days of Transfor-
mational Generative Grammar that this issue really started to attract the
attention of linguists. The major challenge for generative grammarians was
to construct rules which, depending on the higher predicate, generated the
matching "complementizers" (Rosenbaum 1967: 24). For example, the
grammar should be able to generate sentences like I think that John will be
late, but exclude sentences like I think John's being late (Rosenbaum 1967:
29). When attempts to solve such problems with a mixture of (so-called)
syntactic features attributed to the predicates and ever more sophisticated
transformations ran into more and more serious difficulties, it transpired
that complementation is "in part determined by semantic considerations"
(Menzel 1975: 35). Important steps on the way to this recognition were the
work of Vendler on events and facts (see Chapter 1 and Section 2.1), and
the articles by the Kiparskys on factive and emotive predicates (1971) and
by Karttunen on factive verbs (1971a) and implicative verbs (1971b). Al-
though many of the insights gained in the generative paradigm are illumi-
nating and will be helpful for the description of shell-noun uses in Part II,
the relevance of this research for the present study is limited because of the
authors' preoccupation with technicalities of Transformational Grammar. It
The semantic contributions of complements 33
is symptomatic, for example, that in MenzePs (1975) study of the semantics

and syntax of complementation, which is one of the most semantically-
oriented approaches in the generative paradigm, nouns like fact, proposi-
tion, event, process, act and action mainly feature as elements in the deep
stucture of clauses which are deleted by transformation rules.
Closer to the pragmatic-cognitive approach taken in this study are at-
tempts to explain the choice of complements on a predominantly semantic
basis. The study by Ransom (1986) is such an attempt. Ransom tries to
capture the meanings and forms of complements with a system of 16 types
of so-called 'Combined Modalities', which result from the interaction of
four 'Information Modalities' {Truth, Future Truth, Occurrence and Ac-
tion) with four 'Evaluation Modalities' (Predetermined, Determined, Un-
determined and Indeterminate). Depending on their propositional contents
and their higher-level predicates, complements are credited with one of these
modalities. In addition, modalities affecting the higher-level predicates like
tense, aspect or negation are also taken taken into account. Interestingly,
Ransom mentions in a note (1986: 29, note 2) that clauses can occur not
only as subjects and objects of predicates, but also as "predicate nouns,
appositive to nouns, and objects of prepositions". With the sensible justifi-
cation that these sentence types derive their cooccurrence restrictions from
the nouns rather than the higher predicates, she dismisses these structures as
posing a problem outside her immediate field of interest.
What is nevertheless intriguing about Ransom's approach is that she
manages to disentangle the potential meanings of the types of complements
from the highly complex interplay of higher-level predicates and the mo-
dalities affecting these. Her matrix-like approach will therefore be taken up
in Chapter 13. The price for this achievement is high, however, because it is
only possible at the cost of introducing the somewhat idiosyncratic types of
modalities mentioned above. Despite her claim that her analysis of modality
meanings and their forms "developed out of previous linguistic theories"
(1986: 29), Ransom indicates only very briefly (1986: 31, 57) that the first
three of her Information Modalities correspond to what has traditionally
been referred to as epistemic modality and the fourth to root modality. An-
other problem with Ransom's study is that highly frequent and characteris-
tic combinations of predicates with complements are treated on a par with
extremely rare ones. This has to do with her explicit reliance on the TG
method of inventing examples and having them judged by native speakers
(1986: 3), rather than on authentic data. As a result, everyday examples like
I remembered to play chess tomorrow or They watched Bo receive the
awards are given without additional comments alongside with such exotic
combinations as I remembered whether to play chess tomorrow or It took
place that Bo received the award (1986: 18, 38). It is in comparison to
such approaches that the advantage of corpus studies and controlled data on
frequencies of occurrences (see Chapter 4) manifests itself most clearly.
Wierzbicka's (1988: 23-168) chapter on the semantics of English com-
plementation is also not based on corpus data. Yet it is much closer to my
concerns because Wierzbicka's approach is radically semantic and based on
a cognitive view of language. It is her explicit aim to show "that ALL con-
trasts between TO, ING and THAT can be accounted for in terms of
meaning" (1988: 26; emphasis original). Though far from being simplistic,
Wierzbicka's account is less complicated than Ransom's because she tries
to reduce the semantic basis of the complements to the smallest possible
common cognitive denominators. Her method of extracting meaning from
grammatical constructions is similar to mine; as in the present study, Wierz-
bicka compares the combinatorial possibilities of verbs and complements in
order to derive the meanings of //jctf-clauses and infinitives from them. Her
findings are thus particularly well suited to provide the default assumptions
that are needed as a starting-point for this study of the complementation of
nouns, while keeping in mind that they may have to be modified at the end
of the descriptive part.
Unlike most other linguists, Wierzbicka does not work with abstract
metalinguistic terms like actuality, potentiality, certainty or reification (see
e.g. Bolinger 1968: 124, Ney 1981: 129, Frajzyngier and Jasperson 1991:
138), but with simple semantic primitives like 'know', 'say' or 'want',
whose significance and use are justified in her previous work (cf. Wierz-
bicka 1972, 1980, 1985). Since it is difficult to report Wierzbicka's ideas
outside her framework without introducing distorting abstract terms, I will
first quote her verbatim and then explain how I understand her. Wierzbicka
claims that
THAT complements can be shown to be derived from either SAY clauses

or KNOW clauses. I have argued that of these two types the KNOW type
is more basic, and SAY clauses can be reduced to the KNOW type.
(Wierzbicka 1988: 163)
The infinitive complement, on the other hand
is associated with a personal, subjective, first-person mode: want',

think', or know' [...] TO complements are compatible with the element
The semantic contributions of complements 35
'know', but only in the subjective first person mode of know' (which is
sometimes reflected in the so-called 'coreferentiality constraint'); by con-
trast, THAT complements introduce an 'objective', impersonal, 'one can
know' perspective. [...] In most types of TO complements which have been
discussed here there is also a clear future orientation ('this will happen'),
and there are reasons to think that this feature, too, should perhaps be re-
garded as part of the semantic invariant of all TO complement construc-
tions. (Wierzbicka 1988: 164-165)
Translated into more common, but also more abstract, terms, this means
that ^//-complements belong to the epistemic and the linguistic domain.
Infinitives are more difficult to describe. The straightforward part of Wierz-
bicka's account is the relation of infinitives to volition and future events.
Less transparent is her reference to what she calls the "first person mode of
know' ". I interpret this as Wierzbicka's way of saying that infinitives
incorporate an element of subjectivity. This element is used by her to ex-
plain the phenomenon, which was accounted for by the TG grammarians
with the help of the so-called EQUI-NP-DEL transformation,22 that the
subject of the subordinate clause is not expressed on the linguistic surface in
infinitive clauses because it is co-referential with the subject of the matrix
clause.
On the whole, Wierzbicka's account is confirmed by more recent work
on complementation by other authors. Langacker, for example, explicitly
states that "despite a rather different approach to semantic description, her
[i.e. Wierzbicka's, HJS] analyses are roughly compatible with the ones
sketched below" (1991: 439). Givn talks of a "systematic isomorphism
that exists between the semantics of the complement-taking verbs, and the
syntax of verb-plus-complement constructions" (1990: 515). He correlates
âi-clauses with "cognition-utterance verbs" and infinitives with so-called
"manipulative verbs" (1990: 517-561). With cognition-utterance verbs, the
main clauses contain verbs of perception, cognition, mental attitude or ver-
bal utterance, and the complement clauses express propositions. With ma-
nipulative verbs, the main clauses encode manipulations by one agent of
another potential agent, and the complement clauses encode the actions
perfomed by the manipulated agent. Clearly, this account is also compatible
with Wierzbicka's claims. The same can also be said of Frajzyngier and
Jasperson's (1991) proposal, who argue that that-clauses belong to the de
dicto domain and infinitives to the domain of de re. The de dicto domain is
understood by the two authors as including all statements that can be true or
not. Even propositions that strictly speaking do not represent the contents of
utterances fall under this definition.
Supported as it is, then, by the concurring views of other functionalist
and cognitive scholars, Wierzbicka's view can provide the default assump-
tions for the examination of ^/-clauses and infinitives in co-occurrence
with nouns. It will be assumed that /-clauses convey meanings related to
the epistemic and linguistic domains, and to-clauses meanings related to the
domains of volition and future orientation.
3.3 Basic functions of shell-noun typical patterns
In this section, the account of the links between shell nouns and shell con-
tents will be rounded off by examining the basic functional properties of the
four types of lexico-grammatical patterns introduced in Section 3.1.1. The
whole of Part III of this study is devoted to the functional perspective, but
the descriptive part (Part II) will be more illuminating if we already have an
idea of what speakers do with these patterns. Four short but typical exam-
ples are given as reference points in (3.12) to (3.15):
(3.12) Pattern N-cl: This week he repeated his belief that intervention
could draw the UN into a Balkan Vietnam, (ECON)
(3.13) Pattern N-e-cl: ... the eventual aim is to set up a new discipline
from a fusion of two or more old ones, (NEWSCI)
(3.14) Pattern th-H: But what does it all mean? Anyone who claims to
know the Ml answer to that question is either
bragging or lying, (TODAY)
(3.15) Pattern th-be-N : And furthermore it has [pausel erm introduced the
idea that people ought to be compensated for it. -
Aha. - Don't you think that's a crucial point.
(SPOKEN)
For the patterns th- and th-be-N, two of the three functions that define the
class of shell nouns (see Section 2.1) can be seen to dominate. The use of
the pattern this mainly motivated by the linking function. Speakers em-
ploy it to refer to information that has been mentioned before in a text or
conversation (see 3.14). The other two functions are welcome side-effects of
the use of noun phrases here. Although anaphoric reference also plays a role
in the pattern th-be-N, the use of this pattern is mainly motivated by the
Basic functions of shell-noun typical patterns 37
wish to characterize a piece of information in a certain way, in (3.15), for

example, as a crucial point.
The pattern N-e-cl can be related to the third central function, tempo-
rary concept-formation, but in a less direct way. The shell-content complex
in this pattern is not just used to form a temporary concept, but to explicitly
identify the details at the same time. It has already been noted in Section
3.1.1 that the semantic role of the clauses functioning as subject comple-
ments in the pattern N-e-cl is that of an identifier, and it is in the service of
this function that the whole pattern is used.
Both from a semantic and from a functional point of view, the pattern N-
cl (cf. 3.12) is the most versatile of the four. This is largely due to the fact
that the shell-noun phrases are syntactically more flexible in this pattern
than in others. The shell-noun phrases in the patterns N-e-cl and th-be-
always function as subjects and subject complements respectively in their
clauses. In contrast, the shell-noun phrases in the pattern N-cl can fulfill all
syntactic functions except predicates: subjects, objects, subject or object
complements, and, as part of prepositional phrases, adverbials. As a result,
it is impossible to attribute a single dominant function to the pattern N-cl.
(Note that the shell-noun phrases in the pattern th- are syntactically also
variable; this, however, does not affect the dominance of the linking func-
tion.)
It must also be noted here that the pattern N-e-cl is remarkable from the
point of view of information distribution (see Section 16.1 for more details).
It belongs to a group of syntactic constructions with focusing effects, whose
best-known representatives are probably the two variants of cleft sentences
(see e.g. Dik 1980: 210-229, Quirk et al. 1985: 1383-1389). The pattern N-
be-cl shares with these constructions, in particular with the so-called
pseudo-clefl or wh-cleft, that it allows speakers to front certain pieces of
information as given, while highlighting other pieces as new and therefore
particularly noteworthy. In the expression the eventual aim is to ... in
(3.13), for example, the shell-noun phrase is construed as given information
and the shell content as new information and marked for special attention in
the prominent end-position. In the Prague School terminology taken up by
Halliday (1994: 37-67) and other linguists (see Chapter 14 for more de-
tails), the shell-noun phrase in this pattern is theme or starting-point of the
utterance, and the shell content the rheme, i.e. what the utterance is leading
up to.
Chapter 4
The systematic investigation of shell nouns
How can the nouns that are used as shell nouns be identified in a large com-
puterized corpus? The answer to this central methodological question will
be provided in this chapter. In section 4.2, it will be shown that the problem
can be solved by retrieving the lexico-grammatical patterns discussed in
Section 3.1.1. Since these patterns show a clear tendency for including shell
nouns in the syntactic slot reserved for nouns, the nouns emerge from these
corpus queries automatically, 'for free' as it were. The procedure will be
described in considerable detail in this chapter because it cannot be taken
for granted that all readers are familiar with the methods that are used in
corpus studies. A detailed description is also called for since the results of
the corpus investigation also determine what nouns are included in this
study, and since the statistical data constitutes the basis for the functional
interpretation in Part III. To start with, however, it will be briefly explained
in Section 4.1 why it is so important to use the corpus method, given that
one issue of The Times, or for that matter, the Sun or the Daily Mail, would
presumably yield enough examples for a book-size study.
4.1 The From-Corpus-to-Cognition Principle
Why use a large corpus? For a start, there are of course the general benefits
of the corpus method to consider. These have been outlined by many schol-
ars, 23 so it will suffice to review the most important points:
- The material collected in large computerized corpora represents
authentic rather than invented language.
- Computers can process enormous amounts of data.
- The method of retrieving the data is objective rather than intuitive,
which implies that studies can be replicated by other researchers using
the same or different corpora.
- Specific corpora selected from particular types of texts allow for com-
parisons of the use and frequency of certain features in different text-
types, provided that the corpora are large enough.
The From-Corpus-to-Cognition Principle 39
Finally, for non-native students of a language a corpus acts both as

compensation for native-speaker intuition and as protection against
preconceptions and idiosyncratic near-native-speaker intuitions.
To some extent, all these general benefits of the corpus method can be ex-
ploited for the purposes of this study. The crucial advantage of retrieving
the material from a large computer corpus, however, as compared to read-
ing the paper and taking down examples, has to do with their frequency of
occurrence and the relation between frequency and representativity. As
Halliday has put it in a highly illuminating paper on Quantitative studies
and probabilities in grammar, "frequency in text instantiate^] probability
in the system" (1993: 3). The basic idea is that observed frequencies in the
actual use of a language correlate with degrees of preferences in the lin-
guistic system. The importance of a phenomenon in a given language can be
extrapolated from an analysis of its frequency in a large corpus. To quote
one of Halliday's examples (1993: 14-16), if in a corpus of 18 million
words 961,646 clauses are found to be positive with regard to polarity, and
64,391 negative, this seems to be a strong indication that the preference for
expressing ideas by means of positive rather than negative clauses is indeed
a property of the system of the English language.
Taking Halliday's example one step beyond his systemic view to a cog-
nitive view of language, it can be argued that there is also a tendency in our
cognitive system to think in positive rather than negative terms. This meth-
odological transfer from frequency in actual use, first to preference in the
linguistic system, and then to entrenchment24 in the cognitive system, is not
very daring here, given that a negative statement (e.g. Mother isn't at home)
leaves open an infinity of other options and is therefore logically speaking
not as informative as a positive statement. Always provided that the corpus
is large enough, it can also be claimed, however, that such a cognitive re-
interpretation of Halliday's frequency principle is feasible for other linguis-
tic features as well. Thus it seems reasonable to postulate the following
'From-Corpus-to-Cognition Principle':
Frequency in text instantiates entrenchment

in the cognitive system.
Figure 4.1 The From-Corpus-to-Cognition Principle
40 The systematic investigation of shell nouns
The upshot of this principle, which provides the methodological foundation

of this study, is that the corpus method is useful and legitimate for investi-
gations not just of linguistic preferences, but also of cognitive functions and
processes.
What remains to be explained in this chapter is how the corpus method
can be applied in this particular case to obtain
a) an objectively controlled set of data, which consists of

b) the frequent, and therefore linguistically preferred and cognitively more
entrenched, shell nouns rather than exotic albeit interesting ones.
4.2 Data retrieval
Shell nouns are not directly accessible to the corpus method since only those
items can be used as search targets whose capacity to function as shell
nouns is known. Ideally, however, the corpus should not just function as a
source of data on familiar shell nouns but as a heuristic for all shell nouns
that are in frequent usage. The following strategy was developed to retrieve
an objective and representative selection of shell nouns which was to include
previously unknown items as well as familiar ones. The procedure is based
on the recognition (discussed in detail in Section 3.1.1) that speakers fre-
quently rely on a small number of lexico-grammatical patterns when they
use shell nouns. These patterns allow them to establish the linguistic links
between shell nouns and contents which are necessary to trigger the co-
activation of the two groups of linguistic elements. The crucial point from a
corpus-analytical perspective is that all four patterns introduced in Section
3.1.1 have a clearly defined slot where the shell noun can be expected.
Instead of nouns, it were these patterns which in a first step were re-
trieved from a corpus so that the nouns which occur in them could be in-
spected. In the application of this procedure it had to be taken into account
that there is a crucial difference between the patterns N-cl and N-e-cl on
the one hand, and th- and th-be- on the other. Only the patterns N-cl and
N-e-cl more or less guarantee that the noun in the nominal slot is actually
a shell noun.25 One can be fairly sure about this because what they are
linked up with is by definition expressed by means of a whole clause, and
when a clause rather than a noun phrase is used to describe some piece of
experience, this is normally done because this piece of experience is too
complex to be rendered by means of a noun phrase. Experiences of this type
Data retrieval 41
do not just involve persons, organisms and objects, i.e. first-order entities in
Lyons' tripartite taxonomy (1977: 442; see Section 5.1.1), but second-order
states, processes and events, as well as abstract states of affairs or proposi-
tions, which are third-order entities in Lyons' system. All these complex
types of experiences need to be turned into temporary concepts when they
are to continue to play a role in a text or discourse. So the fact that clauses
rather than noun phrases are the targets of reference in the patterns N-cl and
N-e-cl is a fairly reliable indicator that temporary concept-formation is
actually involved.
Unlike the patterns N-cl and N-e-cl, the patterns th- and th-be- are
also used to establish links where no temporary concept-formation is in-
volved. For example, when speakers use expressions such as this book, that
girl, the same blackboard, or this is a famous painting, that is an awful
man, this is a beautiful house, they create anaphoric links with pieces of
experience which are already pre-packaged as concepts. There is clearly no
need, and indeed no room, for temporary concept-formation here. Nouns
such as book, girl and blackboard represent semantically rich lexicalized
concepts which are stable across different contexts of situations (see Section
2.2). Therefore they do not qualify for the status of shell nouns. This is also
reflected in their inability to occur in the patterns N-cl and N-e-cl. Note for
example the inacceptability of such utterances as *The house was to raise
money for a good cause as opposed to The idea was to raise money for a
good cause (see the Introduction in Chapter 1, and Section 5.2 for more
details).
As a methodological consequence of this difference between the patterns
N-cl and N-e-cl on the one hand, and th-H and th-be- on the other, the
retrieval procedure was divided into two steps. In the first step, only in-
stances of the patterns N-cl and N-e-cl were retrieved and counted. The
aim was to arrive at a set of nouns which could fairly reliably be regarded
as shell nouns. This resulted in a set of 670 shell nouns which are listed in
the Appendix. Had occurrences in the patterns th-N and th-be- been
counted in the same way, an enormous number of unwanted concrete nouns
would have found their way into the data. In order to prevent this, the nouns
occurring in the patterns th-N and th-be- were not retrieved with the same
method. Instead, the frequencies of occurrences of only those nouns which
had before been found to occur in the patterns N-cl or N-e-cl were counted
in the second step.
This procedure also has an impact on how the functional definition of
shell nouns given in Section 2.2 is operationalised for the study of the cor-
pus. All nouns (with the exceptions discussed in Section 4.3) which oc-
curred in noteworthy numbers26 in the pattern N-cl and/or N-e-cl were
accepted as potential shell nouns. Nouns which were only found to occur in
the patterns th- and th-be- were not regarded as shell nouns, even if they
seemed to have the semantic and grammatical capacity to occur in the pat-
terns N-cl or N-e-cl. Although this is to a certain extent an arbitrary deci-
sion, it is a necessary step to focus on the most representative examples of
shell nouns and to keep the amount of data in controllable proportions. And
it is objective because it leaves the choice of the nouns which are included in
the material to the use of language as reflected in a large corpus.
The material for this study was retrieved from the British section of the
so-called Bank of English, a large corpus compiled and stored at COBUILD
in Birmingham. When the data for this study was retrieved, the whole cor-
pus, which includes texts from American and Australian sources, amounted
to 320 million words of running text. For this study only the British section
of the corpus was used, since differences between national and regional
varieties were not intended to be an issue here. This section amounted to
225 million words at the time of retrieval. Its composition is summarized in
Table 4.1.
Table 4.1 Composition of the British section of COBUILD'S Bank ofEnglish27
Subcorpus Size in million words Dates

SPOKEN 20.18 post-1990
Recordings of mainly spontaneous, informal conversation from all parts of Brit-
ain
EPHEM 4.72 post-1990

Junk mail, brochures, leaflets, newsletters, etc.; also personal letters
BBC 18.52 1990-1991

Daily transcripts from broadcasts of the BBC World Service, London
BOOKS 42.13 post-1990

384 non-fiction, 188 fiction; 300 male authors, 189 female, 27 joint male-
female, 56 other
TODAY 26.61 1991-1995

Issues of Today newspaper
Data retrieval 43
MAGS 30.14 1992-1993

Issues of 66 different magazines and periodicals; general and special interests
PAPERS
Issues of The Guardian 24.26 1995
Issues of The Independent 19.45 1990 and 1995
Issues of The Times 20.95 1995-1996
ECON 12.13 1990-1991

Issues of The Economist
NEWSCI 6.09 1992-1995

Issues of the New Scientist
The subcorpora are arranged from spoken to written medium and from
more or less colloquial to formal style in Table 4.1. With its size of 225
million words this corpus is substantial enough to meet the essential re-
quirement for the methodological link between corpus, linguistic system and
cognition to hold. Attention should be drawn to the fact that texts from a
fairly large variety of sources, most notably also from spontaneous conver-
sation and from "ephemeral" sources such as leaflets, brochures and letters,
are included. On the other hand, it must be noted that the corpus is clearly
skewed towards the type of language typically used by the media.28 With a
sum of over 65 million words, the three quality newspapers The Guardian,
The Independent and The Times alone account for more than a quarter of
the material. Adding to this the tabloid Today and the weekly The Econo-
mist (which make up about 10 per cent each), the magazines with more than
13 per cent, and finally the material from the BBC (8 per cent), one finds
that the linguistic products from media sources account for more than two
thirds of the whole material. This skewage is understandable because the
output of the media is most readily available in computer-readable format.
While I do not think that this is terribly problematic, it should be borne in
mind when the results of the corpus investigation are assessed, because the
corpus may turn out to be more representative of the English language used
by the media than of other areas of language use. The reason why the Bank
of English rather than, for example, the more balanced British National
Corpus (BNC; see Leech 1993 for a short description) was used for this
study is that the British section of the Bank of English alone contains over
twice as many words as the BNC. Size was given preference over the bal-
ance of sources. If one takes into account that even in the large COBUILD
corpus some shell nouns were not found more often than two or three times,
this choice seems to be justified.
The queries that were used to retrieve examples of shell nouns from this
corpus are listed in Table 4.2. In the query language used at COBUILD target
words are given in lower-case letters, while grammatical tags are indicated
by upper case. Tags can be used either in combination with lexical items or
alone to specify the grammatical form of an empty slot. In the former case
the combination is marked by an oblique ("/"). Strings of words and/or tags
have to be concatenated by "+"-signs.
Table 4.2 List of queries used to retrieve data from the COBUILD corpus
Step 1: Retrieval of instances of the patterns N-cl and N-e-cl
Query statement Number of matching lines

in the 225m corpus
Pattern N-cl with various types of adjacent clauses:
NN+that/CS (NN = noun, CS = conjunction) 280,217

NN+to+VB (VB = base form of verbs) 560,148
NN+where 64,029
NN+when 133,163
NN+why 14,192
Pattern -be-to
NN+is+to 28,463
NN+was+to 12,728
NN+has+been+to 962
NN+will+be+to 960
NN+would+be+to 1,421
NN+would+have+been+to 133
Pattern N-be-that
NN+is+that 37,155
NN+was+that 9,104
NN+has+been+that 433
NN+will+be+that 178
NN+would+be+that 264
NN+would+have+been+that 19
Data retrieval 45
Pattern N-Ae-wA
NN+is+WH (WH = wA-word) 5,450

NN+was+WH 1,849
NN+has+been+WH 35
NN+will+be+WH 123
NN+would+be+WH 86
NN+would+have+been+WH 7
Step 2: Frequency count of nouns in the patterns th-N and th-be-fi
Nouns in the pattern fA-N
this|that/DT+(JJ)+0MM (DT = deter-miner, scores ranging from 0 to 27,631

JJ = adjective) (for the noun timej
Nouns in the pattern th-be-N
this|that/DT+is|was+(JJ)+/7owrt scores ranging from 0 to 1,711

(for the noun way)
The query statements listed in this table reflect my effort to retrieve as wide
a range as possible of the shell nouns that occur in the patterns N-cl and N-
be-cl. In addition to the forms of the copula be that are given in the list, a
number of modal expressions such as must have been and should have been
were also tested. However, the yield was very low and the items retrieved
promised no additional insights.
Two other types of queries had suggested themselves from my previous
work on idea and from the unsystematic observation of the use of shell
nouns (see also Section 3.1.1): NN+of+Ving (as in the problem of raising
money) and NN+of+WH (the question of where to go). These were also
tested, but the results of these queries were disappointing. While typical
shell nouns such as way, chance, possibility, idea, question and problem
topped the frequency lists for these queries, a number of unwanted expres-
sions like cost of living and per cent of which were also frequent, and al-
most all of the less frequent items were unwanted hits. In technical terms,
these queries proved to be too imprecise.29 These queries were therefore
only taken into account as means of retrieving additional examples of shell
nouns, but not used for the quantitative analysis.
Because the query language requires at least one lexical item in a well-
formed query statement, it was not possible to retrieve nouns with postno-
minai w/i-clauses wholesale, as it were. Therefore the most frequent wh-

elements when, where and why were used as target words in these queries. It
should also be mentioned that generic query statements for all forms of the
copula be (e.g. NN+@be+to) would also have been possible. Yet statements
of this type were not used, since differences between the results for the dif-
ferent tenses and modalities are much easier to detect when the data is sepa-
rated.
As explained above, the occurrence of nouns in the patterns th- and th-
be- was counted directly, using the nouns found in the patterns N-cl and
N-e-cl as cues. In addition, the queries listed in Table 4.3 below were used
to obtain authentic data on these and related types of usages, although they
were not included in the quantitative analysis. For all these query state-
ments, concordances consisting of large numbers of examples from all sub-
corpora were retrieved.
Table 4.3 Queries carried out for the patterns th- and th-be-
Pattern th-N:
Referring item + shell noun + optional intervening adjective
this+l,lJJ+NN (1,1 = exactly one intervening word)
that/DT+NN
these+NNS (NNS = plural noun)
these+JJ+NNS
those+NNS
those+JJ+NNS
another+NN
another+JJ+NN
same+NOUN (NOUN = singular or plural noun)
such+NOUN
such+,
such+l,lJJ+NN
Pattern th-be-fi:
Referring item + copula + determiner + optional premodifier + shell noun
this+is+1,1RB|JJ+JJ+NN (RB = adverb)
this+would+be+1, INN
this+would+be+l,lJJ+NN
that/DT+was+1, INN
that/DT+was+1,1JJ+NN
Data retrieval 47
that/DT+would+be+1, INN
that/DT+would+be+1,1 JJ+NN
that/DT+s+1,1 NN
that/DT+s+1,1 JJ+NN
that/DT+s+l,lRB|JJ+JJ+NN
that/DT+s+been+1,1 NN
that/DT+s+been+1,1 JJ+NN
Obvious with such masses of examples as those listed in Table 4.2, each
example cannot be looked at individually. Although browsing through con-
cordance lines is illuminating and greatly enhances the researcher's feeling
and sensitivity for his subject-matter, data of these proportions cannot be
managed by this method. This is the point where the From-Corpus-To-
Cognition Principle comes into play (see Section 4.1). The principle de-
mands that, despite the indisputable charm of rare or exotic examples, one
should mainly be interested in frequent and therefore systemically and cog-
nitively more important items.
The computer system at COBUILD provides users with two very powerful
and economical ways of finding the most frequent instances of a particular
pattern one has queried. One is a computer program whose output is re-
ferred to as a picture. As is explained in detail in Clear (1993), a picture
gives a list of the fifty most frequent words in each slot within a span of
between 1 and 6 items to the right and the left of the node. Pictures are an
enormous help to get an idea of the typical linguistic environment of a cer-
tain word or string of words. By means of this program, the 50 most fre-
quent shell nouns themselves could be identified for each pattern and sub-
corpus, as well as their most frequent premodifiers and many other aspects
of the immediate lexical environment.
A second COBUILD computer program30 is designed to focus on the item
immediately to the left of the node. The advantage of this program is that its
output is a list of not just the fifty, but the 1,000 most frequent items in this
slot. Since the nouns in the patterns N-cl and N-e-cl are in precisely this
position relative to the node of the query,31 this program supplies a list of
those 1,000 nouns which occur most frequently in the slot reserved for shell
nouns. For most queries given in Table 4.2 the frequencies of occurrence of
the nouns at the bottom end of these lists were between 1 and 5. The noun
manoeuvre, for example, occurs three times in the pattern noun phrase +
was + to, and the noun ruse three times in the pattern noun phrase + is + to
and twice in the pattern noun phrase + was + to. These low scores suggest
that nouns that are not included among the 1,000 most frequent ones are so
rare that their role as shell nouns is limited and that they can be neglected
here.32
4.3 Cleaning up the data
No automatic corpus retrieval procedure that consists of more than a single

word yields only those examples that the researcher has actually wanted it
to retrieve. Unless corpus queries are formulated with a strong bias for pre-
cision, i.e. in a more narrow way than is actually in the interest of a high
recall, unwanted matches are endemic. This is also true of the queries for
the patterns N-cl and N-e-cl in this study. Since unwanted matches do not
contribute to the understanding of shell nouns and distort the relative fre-
quencies that will be examined for these two patterns, they have been re-
moved from the data.
The unwanted matches that have found their way into the raw material
are chiefly caused by two types of interference:
a) insertions of linguistic material between a shell noun and the postnomi-

nal or complementing clause, and
b) formal coincidences between the patterns N-cl and N-e-cl and other
grammatical constructions.
Below, I will discuss the most frequent manifestations of the two types for
the patterns examined.
Insertions
For a first idea of what I regard as insertions consider example (4.1):
(4.1) The energy policy of the government is to have no energy policy at all.
(PAPERS).
In this sentence the noun government occupies the slot to the left of the
words is to where the shell noun would normally be expected to occur.
However, the shell noun in this sentence is clearly policy, whereas the
prepositional phrase of the government functions as postmodifier to the
noun phrase headed by this shell noun. Government occurs so frequently in
Cleaning up the data 49
this position that it is in fact among the top ten in the list of the most fre-
quent nouns in this pattern. However, since these are clear cases of un-
wanted hits, this noun has been removed from the list of the 1,000 most
frequent nouns. Other nouns which occur frequently in the same position
and with the same function of identifying an institution or body that initiates
a policy, scheme, project or the like are party, council, commission and
community. These nouns and all others that behave in a similar way have
also been discarded from the list.
The most frequent insertions in the same pattern with complementing
f/jaf-clauses are less uniform. There are nouns that occur as postmodifiers
in a way similar to government and the other nouns, for example system,
book, word and school. Adverbials, especially time adverbials, are a second
source of insertions. The temporal nouns moment, night, week, time and
year and the adverbial today were among the fifty most frequent nouns in
the queries NN+z's+iAa/ and NN+wfli+zAotf. In addition, the noun course, as
the main element of the fixed expression of course, is also very frequently
inserted as an adverbial between a shell-noun phrase functioning as subject
and the copula in intensive clauses (e.g. the problem of course is that...).
In general, insertions are more frequent with the pattern N-e-cl than
with N-cl. The connection between a noun and its appositive postmodifier
or postnominal complement (see Section 3.1.1) is stronger and allows for
insertions less easily than the clausal connection between subjects and
clauses functioning as subject complements. One type of insertion which is
possible, however, although it is not very common, is again the insertion of
a temporal phrase, as in example (4.2):
(4.2) ... responding to a demand last year that Liberia pay off some seven
million Dollars worth of arrears on its debts to the US. (BBC)
In this example, demand is the shell noun and last year an inserted noun
phrase.
Formal coincidences
What is meant by formal coincidences emerges most clearly from a set of
examples of the pattern N-to. Especially in texts from media sources, noun
phrases headed by the nouns government and party are frequently used as
objects to causative verbs like force, urge, order, encourage, allow or
press and followed by a subordinate to-infinitive clause (e.g. he urges the
government to work more closely with the unions, PAPERS). Obviously this
pattern answers the query search NN+io+VB. Nevertheless, such instances
are clearly not examples of shell nouns but coincide only superficially and
are therefore unwanted matches.
With the noun home, there is a similar coincidental match between the
same query (NN+to+VB) and a large number of examples in which the
noun home indicates a location and the following -infinitive clause an
activity or intention to be carried out at this location. Example (4.3), again
taken from a newspaper source, provides an illustration:
(4.3) ... we would then return home to consume our delicious treats.
The most frequent type of formal coincidence with the pattern -that is
brought about by insertions. It occurs when linguistic material ending with a
noun is inserted between speech act, cognition or evidential verbs and the
/-clauses complementing them. See for example the utterance in (4.4)
taken from the SPOKEN material:
(4.4) I just wish we I'd known from the start that you know I'd only got six
thousand pounds to play with.
Here the -clause is related to the cognition verb know. The prepositional
phrase from the start is inserted in such a way that the noun start comes to
hold the potential shell-noun position.
Despite the fact that the tag CS (indicating conjunction) was included in
the query +/CS, examples of nouns with postmodifying relative
clauses introduced by that also found their way into the concordances and
frequency lists. An example is (4.5):
(4.5) ... they exposed it, with the same merciless treatment that Leeds reserved
for a tolerant Hughes, (BBC)
The corpus has been tagged automatically with a probabilistic tagging pro-
gram. It seems that these hits testify to the error margin of the tagging pro-
cedure. They should not give rise to great worries because nouns that do not
qualify for shell nouns can easily be spotted. For the true shell nouns which
were not discarded, the relative proportion of relative clauses that have
made their way into the frequency scores should be identical. Quantitative
comparisons between these nouns should thus not be affected dramatically.
There are also formal coincidences with fixed expressions. The noun or-
der is a case in point. This noun is the second most frequent noun in the
pattern N-e-fo with more than 15,000 occurrences. However, most of these
are due to occurrences of the complex subordinating conjunction in order
to, which answers my query pattern NN+to+VB. It seems clear enough that
order should not be considered a shell noun when it is used in this fixed
expression.
The pattern -be-to formally coincides with various temporal and modal
meanings of the same pattern (see Palmer 1990: 164-166). Compare exam-
ple (4.6), where reference to future intentions play a role:
(4.6) three long-term tasks need attention if the economy is to remain com-
petitive. (ECON)
As a general rule the patterns of both type N-cl and N-e-cl with w/z-clauses
give rise to similar types of interference as the patterns with -clauses
and infinitive clauses. A specific source of interference in the w/2-domain
are appositive locative and temporal clauses introduced by where or when,
as in example (4.7):
(4.7) The scenes in Beijing in the week or so after October 12th, when the
Chinese Communist Party opens itsfive-yearlycongress,... (ECON)
Most of these unwanted matches can be discarded automatically after re-

trieval because where and when are usually preceded by commas in these
patterns. A further advantage with wA-patterns is that unwanted matches
are even easier to spot in the frequency lists because the range of possible
candidates of shell nouns is much more limited for these patterns (see Sec-
tion 4.5).
All nouns which could be identified as unwanted items on the basis of
the insertions and coincidences discussed in this section were removed from
the frequency lists.
4.4 Systematic misses of the corpus inquiry
Unwanted matches are hits of search queries that do not comply with the
researchers' expectations. The reverse side of the coin are examples of shell
nouns that are not found in the corpus because they do not match the search
queries. Again, there are a fairly small number of systematic reasons for
such failures.
First, insertions are not only a cause for the inclusion of unwanted
matches in the frequency lists but also for the omission of relevant ones.
Methodologically, the effect of such examples as (4.1) above is not just that
government will wrongly score one example as a shell noun, but also that
policy will miss one legitimate chance to be registered as a shell noun in the
data. My detailed analyses of the noun idea (Schmid 1993: ch. 5, Schmid
1997) and the many hours that I have spent browsing through concordances
suggest that for some nouns as many as 30 to 40 percent of uses as shell
nouns are not included in the frequency scores here because of insertions of
other linguistic material. Most affected by such systematic misses are shell
nouns with distinct relational meanings such as part, feature, aspect or
basis (see Section 7.6). These nouns are normally accompanied by a post-
modifier in which the other element of the relation is expressed as a refer-
ence-point. An example is given in (4.8):
(4.8) The next part of the project is to go back and to identify where these
products come from ... (BBC)
As far as misses of tokens of other shell nouns are concerned, the extent to
which the scores are affected is similar for most nouns. So the relative
scores, i.e. the difference between frequencies of nouns, are not skewed by
these misses.
Besides insertions, variations of the patterns constitute a second common
source of misses. Thus, the query NN+is+that/CS unfortunately misses out
when the conjunction that is deleted or when an additional is is added.33
Both are notorious phenomena in spoken English, as is demonstrated by the
familiar ring of the four examples in (4.9), which are all taken from the
SPOKEN data:
(4.9) (a) The thing is I just wonder whether I'm gonna be bloody going.
But the thing is it's not as if I've forced my friends on him ...
(b) I understand what you're saying John I really do but the thing is is
that I think that Mrs Thatcher er was very ...
... always turn to the Bible to slag gay people off. Well the thing is is
that these Christians of whatever denomination ...
The omission of such examples from the quantitatively controlled data is of

course a great pity, but it is the price one has to pay when one is using
automatic corpus retrieval methods, which can never be as flexible as a

manual data collection. When specific collocations are known, the deficit
can be compensated for by looking for examples of the particular patterns,
as was done in order to retrieve the examples given in (4.9). However, it is
not possible to use the corpus as a heuristic device for finding patterns
which do not strictly answer the precise queries.
Third, the shell-noun typical patterns NN+of+Ving and NN+of+WH did
not supply data for the quantitative analysis. As mentioned in Section 4.2
above, the reason is that results were too unreliable.
Fourth, potential plural forms of shell nouns are systematically neglected
by my search queries. Good candidates for shell-nounhood are the items
circumstances and conditions, as well as the plural forms of high-frequency
nouns like time, place and way. As plural forms do not tend to be embedded
in the patterns of the types N-cl and N-e-cl, they do not lend themselves to
the kind of analysis carried out here. The environment in which they prevail
is the pattern th-N, where they are accompanied by elements like these,
those, same and such.
On the whole, these systematic misses of the search queries should be no
cause for too much methodological concern. Their effect is simply a whole-
sale reduction of the amount of examples in relation to what is actually
hidden in the whole corpus. Although this does not distort the relative dif-
ferences between the frequency scores of different shell nouns, it should be
kept in mind. When the statistical measures of attraction, reliance and
compiled reliance are now introduced to account for the dependency of a
noun on shell-noun typical patterns, low scores for these values should not
come as a surprise.
4.5 A survey of the results of the corpus inquiry
In this section a first glimpse of the results of the corpus inquiry will be
given. In line with the logic outlined in Section 4.1, the frequent nouns will
be focussed on here. Providing such an overview is not as straightforward a
task as it may seem. It is not sufficient to simply list the results of the cor-
pus queries in order of the frequencies of occurrence of different nouns in
the four main patterns. The problem with such a strategy is that this score
depends on the frequency of occurrence of a noun in the whole corpus.
Purely on the grounds of probability, highly frequent nouns like fact or
thing are of course much more likely to occur in any context than rare
nouns such as premise or misapprehension. As a consequence, frequent

nouns will occur more often in the patterns of interest here, even though less
frequent nouns may be more dependent on a certain pattern.
In order to avoid absolute scores without complicating things too much,
two very simple statistical measures will be introduced which relate the raw
frequencies of occurrence to the frequency of occurrence of the pattern on
the one hand, and to the frequency of occurrence of a certain noun in the
overall corpus on the other. 34
I call the first measure attraction. It is calculated by dividing the fre-
quency of occurrence of a noun in a pattern by the total sum of the occur-
rences of this pattern in the corpus. The result of this calculation measures
the degree to which a certain lexico-grammatical pattern attracts a particu-
lar noun. Since the denominator of the fraction is the same for all nouns
which occur in a particular pattern (see Figure 4.2), the scores for this value
are directly proportional to the raw frequencies of occurrences of nouns.
What the measure allows is to compare the importance of nouns across the
different patterns.
Attraction: degree to which a lexico-grammatical pattern attracts a certain noun
frequency of a noun in a pattern

attraction =
total frequency of the pattern
Reliance: degree to which a noun depends on a particular pattern
frequency of a noun in a pattern

reliance =
total frequency of the noun in the corpus
Figure 4.2 The statistical measures of attraction and reliance
The second measure is called reliance. It is calculated by dividing the fre-

quency of occurrence of a noun in a pattern by its frequency of occurrence
in the overall corpus. This simple measure captures the degree to which a
particular noun relies, or depends, on a lexico-grammatical pattern for its
occurrence. It is important to note that this measure is arithmetically contin-
gent upon the frequency of a noun in the overall corpus, and therefore the
importance of lexico-grammatical patterns for rare nouns emerges as well.
A survey of the results 55
The definitions and calculations of the two measures are summarized in

Figure 4.2.
That it is important to look at the relation between nouns and the pat-
terns from both sides emerges from a look at Table 4.4, which provides a
section of the results for the pattern -that. The table contains two charts.
The list on the left-hand side gives the fifty nouns with the highest scores for
attraction (see the third column in the table). As mentioned above, these
scores are directly proportional to the scores for frequency in the pattern,
which are given in the second column. On the right-hand side the fifty nouns
with the highest scores for reliance are listed. The left-hand chart indicates
that with a raw frequency in the pattern of 26,106 and an attraction score
of 18.45%, the noun fact is by far the most frequent noun in the pattern N-
that. Using the image evoked by the term attraction, this means that the
pattern N-i/zaf-clause attracts the noun fact in 18.45% of its occurrences.
The chart on the right-hand side depicts the same relation from the perspec-
tive of the nouns. It shows that the pattern -that is more important for the
nouns realisation, proviso, assumption, assertion, belief and insistence
than for the noun fact. For example, as many as 820 instances of the noun
realisation (which only occurs 1,185 times in the whole 225 million-word
corpus) are occurrences in the pattern -that. This yields a reliance score
of 69.2%. In contrast, the reliance score for fact is 38.13%, which means
that the noun fact relies to a relatively lesser degree than realisation on the
pattern -that for its use.
The image that I want to evoke by the terms attraction and reliance is
that of a symbiotic relation: relatively specific lexico-grammatical patterns
serve as hosts for nouns which are not completely successful linguistic signs
on their own because they are too unspecific (see Section 5.2). The lexico-
grammatical patterns attract certain nouns and the nouns in turn depend to a
variable extent on one or more patterns for their occurrence. The scores for
the two values are fundamentally affected by the rigidity of the corpus que-
ries and the effects of this, namely that unwanted hits are recorded to some
extent as well, while many basically wanted occurrences are missed (see the
discussion in Sections 4.3 and 4.4). Therefore their significance should not
be overestimated. As a general rule, however, one can consider scores for
compiled reliance of more than 15% to be interesting. Those over 20% can
be seen as an indication that a given noun is earmarked to be used as shell
noun, while scores over 30%, 40% or even 50% are clearly spectacular.
The rest of this section consists of six tables that are all arranged the
same way as Table 4.4. They contain the fifty top scorers for attraction and
reliance in the six patterns which make up the core of the corpus enquiry:
-that, N-to, -be-that, -be-to, N-wA and -be-wh. The scores in Table
4.9 have been compiled from the scores for the queries containing different
forms of the verb to be and the three w/z-elements where, when and why (see
Table 4.2). As can be seen, the lists in Tables 4.8 and 4.9 are shorter than
the others. This is due to the fact that no more than 31 and 21 noun-types
respectively were found to occur as shell nouns in the two patterns. All
other nouns had to be discarded as unwanted matches.
Since the material presented in the six tables will be discussed in consid-
erable detail in the later sections of this study, especially in Part II, I will
refrain from further comments here. In addition to the frequent types in-
cluded in the lists, the whole material used as the quantitative basis for this
study is available in the Appendix. All 670 lexemes that were found to oc-
cur in shell-noun uses are given in alphabetical order to allow readers to
check on the collocational versatility of single nouns as they move through
the text. However, for reasons of space only an addition of the reliance
scores for the patterns is provided, which is called compiled reliance. This
value is an indicator of the degree to which a given noun depends on the
patterns investigated here. Passages where certain shell nouns are discussed
in the text can be located by consulting the index of shell nouns at the end of
the book.
Table 4.4 Pattern: N-that

Total number of types: 350
Total number of tokens: 141,476
Noun Freq. in Attraction Noun Freq. Freq. Reliance

pattern in in cor-
pattem pus
fact 26,106 18.45% realisation 820 1,185 69.20%
evidence 5,007 3.54% proviso 111 250 44.40%
idea 4,812 3.40% assumption 1,391 3,151 44.14%
doubt 4,010 2.83% assertion 596 1,492 39.95%
belief 3,696 2.61% belief 3,696 9,344 39.55%
view 3,532 2.50% insistence 796 2,069 38.47%
hope 2,727 1.93% fact 26,106 68,472 38.13%
news 2,572 1.82% premise 274 765 35.82%
feeling 2,511 1.77% misapprehension 44 123 35.77%
impression 2,279 1.61% suggestion 2,033 5,854 34.73%
possibility 2,232 1.58% dictum 84 249 33.73%
claim 2,194 1.55% stipulation 48 145 33.10%
suggestion 2,033 1.44% misconception 91 284 32.04%
speculation 1,922 1.36% truism 47 150 31.33%
knowledge 1,794 1.27% reminder 812 2,688 30.21%
sign 1,738 1.23% notion 1,655 5,713 28.97%
notion 1,655 1.17% coincidence 627 2,196 28.55%
point 1,511 1.07% speculation 1,922 6,778 28.36%
warning 1,460 1.03% supposition 46 164 28.05%
fear 1,432 1.01% impression 2,279 8,206 27.77%
assumption 1,391 0.98% indication 1,063 3,949 26.92%
conclusion 1,282 0.91% adage 89 350 25.43%
argument 1,259 0.89% presumption 98 408 24.02%
chance 1,247 0.88% presupposition 8 34 23.53%
concern 1,224 0.87% doubt 4,010 17,322 23.15%
extent 1,217 0.86% acknowledgement 149 653 22.82%
reason 1,209 0.85% allegation 244 1,083 22.53%
proof 1,191 0.84% conclusion 1,282 6,170 20.78%
case 1,147 0.81% accusation 243 1,187 20.47%
indication 1,063 0.75% conviction 867 4,390 19.75%
announcement 1,023 0.72% suspicion 806 4,088 19.72%
thought 975 0.69% 410 2,110 19.43%
confirmation 19.35%
theory 939 0.66% presentiment 6 31 19.31%
statement 868 0.61% expectation 374 1,937 19.01%
danger 867 0.61% assurance 492 2,588 18.87%
conviction 867 0.61% axiom 30 159 18.69%
report 841 0.59% pretence 120 642 18.68%
surprise 821 0.58% proof 1,191 6,377 18.48%
realisation 820 0.58% possibility 2,232 12,075 18.19%
reminder 812 0.57% certainty 377 2,072 18.18%
effect 809 0.57% claim 2,194 12,068 17.77%
suspicion 806 0.57% boast 67 377 17.66%
insistence 796 0.56% guarantee 613 3,471 17.66%
message 755 0.53% contention 300 1,699 17.45%
risk 724 0.51% feeling 2,511 14,392 16.92%
confidence 703 0.50% delusion 77 455 16.61%
opinion 679 0.48% inkling 51 307 16.56%
principle 655 0.46% fallacy 51 308 16.53%
question 640 0.45% inference 62 375 16.38%
information 633 0.45% finding 96 586
Table 4.5 Pattern: N-to

Noun Freq. in Attraction Noun Freq. in Freq. Reliance

pattern pattern in cor-
pus
time 19,496 8.54% inability 1,843 2,400 76.79%
way 15,194 6.66% temerity 118 160 73.75%
attempt 13,543 5.94% disinclination 45 62 72.58%
chance 11,722 5.14% willingness 1,804 2,493 72.36%
decision 10,089 4.42% unwillingness 337 470 71.70%
ability 9,830 4.31% attempt 13,543 20,728 65.34%
right 8,522 3.74% refusal 2,406 3,892 61.82%
need 8,150 3.57% ability 9,830 16,957 57.97%
opportunity 7,799 3.42% urge 918 1,626 56.46%
effort 5,221 2.29% reluctance 1.158 2,131 54.34%
power 4,194 1.84% eagerness 285 535 53.27%
failure 4,179 1.83% readiness 676 1,368 49.42%
desire 4,159 1.82% opportunity 7,799 17,805 43.80%
plan 3,866 1.69% inclination 339 801 42.32%
place 3,791 1.66% propensity 189 449 42.09%
reason 3,499 1.53% desire 4.159 9,973 41.70%
bid 3,343 1.47% tendency 1,763 4,297 41.03%
campaign 3,098 1.36% temptation 859 2,193 39.17%
pressure 2,649 1.16% obligation 765 2,041 37.48%
position 2,476 1.09% incentive 969 2,690 36.02%
refusal 2,406 1.05% determination 2,103 5,922 35.51%
thing 2,348 1.03% permission 1,988 5,950 33.41%
determination 2,103 0.92% mducement 65 198 32.83%
permission 1,988 0.87% effort 5,221 15,926 32.78%
inability 1,843 0.81% keenness 43 133 32.33%
willingness 1,804 0.79% need 8,150 26,316 30.97%
tendency 1,763 0.77% wish 821 2,665 30.81%
capacity 1,678 0.74% chance 11,722 39,248 29.87%
intention 1,611 0.71% compulsion 188 644 29.19%
duty 1,508 0.66% failure 4,179 15,294 27.32%
room 1,375 0.60% pledge 543 2,162 25.12%
idea 1,271 0.56% decision 10,089 40,355 25.00%
freedom 1,230 0.54% audacity 70 281 24.91%
move 1,226 0.54% intention 1,611 6,484 24.85%
courage 1,201 0.53% conspiracy 836 3,427 24.39%
agreement 1,183 0.52% resolve 266 1,156 23.01%
reluctance 1,158 0.51% vow 132 616 21.43%
programme 1,130 0.50% bid 3,343 16,642 20.09%
proposal 1,107 0.49% courage 1,201 6,110 19.66%
deal 1,104 0.48% ploy 197 1,019 19.33%
commitment 1,072 0.47% invitation 744 4,062 18.32%
scheme 1,028 0.45% incapacity 38 217 17.51%
job 1,022 0.45% mandate 316 1,866 16.93%
authority 1,002 0.44% gall 77 458 16.81%
struggle 1,002 0.44% impulse 206 1,226 16.80%
moment 997 0.44% ruse 49 292 16.78%
incentive 969 0.42% capacity 1,678 10,012 16.76%
offer 960 0.42% leeway 47 294 15.99%
commission 929 0.41% injunction 184 1,165 15.79%
drive 926 0.41% intent 353 2,351 15.01%
Table 4.6 Pattern: -be-that

Noun Freq. in Attraction Noun Freq. in Freq.in Reliance

pattern pattern corpus
problem 2,672 8.62% upshot 105 313 33.55%
thing 1,532 4.94% snag 250 784 31.89%
truth 1,235 3.98% drawback 140 735 19.05%
fact 1,218 3.93% implication 274 1,514 18.10%
trouble 1,034 3.34% guess 266 1,620 16.42%
point 1,020 3.29% irony 395 3,085 12.80%
result 977 3.15% downside 51 512 9.96%
view 933 3.01% inference 37 375 9.87%
reason 897 2.89% corollary 19 198 9.60%
idea 790 2.55% hunch 43 451 9.53%
news 749 2.42% gripe 17 186 9.14%
difference 642 2.07% stipulation 12 145 8.28%
answer 633 2.04% worry 253 3,119 8.11%
theory 561 1.81% assumption 229 3,151 7.27%
reality 509 1.64% truth 1,235 17,421 7.09%
hope 482 1.56% complication 34 484 7.02%
fear 437 1.41% likelihood 125 1,857 6.73%
argument 430 1.39% disadvantage 101 1.556 6.49%
danger 395 1.27% regret 99 1,754 5.64%
irony 395 1.27% paradox 63 1,149 5.48%
feelmg 376 1.21% finding 32 586 5.46%
explanation 323 1.04% surmise 3 55 5.45%
message 278 0.90% boast 20 377 5.31%
implication 274 0.88% proviso 13 250 5.20%
concern 268 0.86% consolation 85 1,699 5.00%
guess 266 0.86% caveat 10 200 5.00%
worry 253 0.82% explanation 323 6.557 4.93%
snag 250 0.81% thesis 71 1,467 4.84%
conclusion 249 0.80% presumption 19 408 4.66%
assumption
229 0.74% peculiarity 5 109 4.59%
possibility
227 0.73% problem 2,672 59,600 4.48%
advantage
221 0.71% trouble 1,034 23,592 4.38%
197 0.64% retort 9 208 4.33%
impression 196 0.63% 4.31%
evidence 191 0.62% premise 33 765 4.29%
story 190 0.61% consensus 162 3,773 4.26%
position 188 0.61% objection 52 1,220 4.05%
difficulty 188 0.61% oddity 13 321 4.04%
suggestion 162 0.52% conclusion 249 6,170 3.68%
consensus 145 0.47% betting 46 1,251 3.67%
opinion 140 0.45% reality 509 13,863 3.61%
drawback 138 0.45% rationale 24 665 3.53%
tragedy 137 0.44% answer 633 17,957 3.52%
belief 137 0.44% misconception 10 284 3.51%
risk 132 0.43% difference 642 18,302 3.50%
experience 129 0.42% probability 45 1,286 3.48%
surprise 126 0.41% argument 430 12,367 3.46%
thought 125 0.40% subtext 9 260 3.45%
likelihood 122 0.39% grumble 3 87 3.33%
factor 119 0.38% certainty 69 2,072 3.31%
approach tenet 5 151
Table 4.7 Pattern: -be-to

Total number of tokens . 21,876
Noun Freq. Attraction Noun Freq. Freq.in Reliance

in pat- in pat- corpus
tern tern
aim 2,646 12.10% aim 2,646 9,324 28.38%
idea 1,141 5.22% brief 118 835 14.13%
job 1,041 4.76% objective 370 2,883 12.83%
task 946 4.32% countermeasure 1 10 10.00%
thing 758 3.46% ambition 373 4,678 7.97%
plan 629 2.88% intention 494 6,484 7.62%
purpose 586 2.68% task 946 13,439 7.04%
solution 551 2.52% priority 425 6,055 7.02%
step 515 2.35% trick 368 6,080 6.05%
intention 494 2.26% inclination 46 801 5.74%
answer 431 1.97% tactic 71 1,300 5.46%
priority 425 1.94% instinct 150 2,853 5.26%
goal 399 1.82% purpose 586 11,524 5.09%
ambition 373 1.71% alternative 334 7,046 4.74%
policy 370 1.69% solution 551 12,161 4.53%
objective 370 1.69% temptation 82 2,193 3.74%
trick 368 1.68% option 328 9,913 3.31%
alternative 334 1.53% bet 95 3,037 3.13%
option 328 1.50% object 187 6,027 3.10%
way 316 1.44% function 174 5,651 3.08%
role 312 1.43% step 515 16,809 3.06%
approach 307 1.40% ploy 31 1,019 3.04%
response 303 1.39% remedy 46 1,537 2.99%
strategy 303 1.39% ideal 50 1,762 2.84%
effect 286 1.31% imperative 12 426 2.82%
problem 286 1.31% mission 249 9,511 2.62%
challenge 254 1.16% strategy 303 11,607 2.61%
mission 249 1.14% mistake 115 4,540 2.53%
reaction 232 1.06% impulse 31 1,226 2.53%
advice 227 1.04% motive 50 2,036 2.46%
concern 211 0.96% idea 1,141 46,654 2.45%
method 200 0.91% answer 431 17,957 2.40%
duty 191 0.87% achievement 129 5,483 2.35%
object 187 0.85% plan 629 27,044 2.33%
function 174 0.80% gambit 6 261 2.30%
point 163 0.75% method 200 9,546 2.10%
move 154 0.70% goal 399 19,757 2.02%
instinct 150 0.69% wish 50 2,665 1.88%
action 144 0.66% job 1,041 57,599 1.81%
stage 139 0.64% challenge 254 14,460 1.76%
dream 130 0.59% reaction 232 13,248 1.75%
achievement 129 0.59% ruse 5 292 1.71%
brief 118 0.54% secret 85 5,009 1.70%
result 116 0.53% key 111 6,753 1.64%
mistake 115 0.53% duty 191 11,700 1.63%
hope 112 0.51% response 303 19,283 1.57%
key 111 0.51% tendency 64 4,297 1.49%
project 108 0.49% approach 307 20,913 1.47%
target 102 0.47% concern 211 17,148 1.23%
bet 95 0.43% obligation 25 241 1.22%
Table 4.8 Pattern: N-wA

Noun Freq. Attraction Noun Freq.in Freq.in Reliance

in pat- pattern corpus
tern
time 8,917 30.24% reason 4,614 34,941 13.21%

reason 4,614 15.64% spot 527 10,344 5.09%
point 2,691 9.12% occasion 451 9,576 4.71%
place 2,540 8.61% era 309 6,955 4.44%
area 1,551 5.26% point 2,691 68,563 3.92%
moment 1,434 4.86% situation 1,005 28,292 3.55%
situation 1,005 3.41% moment 1,434 44,496 3.22%
stage 968 3.28% setback 63 1,956 3.22%
age 823 2.79% area 1,551 52,205 2.97%
idea 752 2.55% stage 968 35,031 2.76%
period 729 2.47% time 8,917 342,869 2.60%
spot 527 1.79% place 2,540 110,008 2.31%
position 516 1.75% period 729 31,934 2.28%
case 478 1.62% age 823 44,106 1.87%
occasion 451 1.53% site 249 14,800 1.68%
era 309 1.05% blow 122 7,486 1.63%
site 249 0.84% idea 752 46,654 1.61%
region 236 0.80% conundrum 5 345 1.45%
question 178 0.60% region 236 17,332 1.36%
cnance 175 0.59% position 516 38,431 1.34%
blow 122 0.41% tragedy 67 5,974 1.12%
tragedy 67 0.23% mystery 45 5,949 0.76%
setback 63 0.15% case 478 70,636 0.68%
mystery 45 0.12% query 3 540 0.56%
explanation 36 0.12% explanation 36 6,557 0.55%
understanding 35 0.06% chance 175 39,248 0.45%
cause 14 0.05% enigma 3 677 0.44%
conundrum 5 0.02% question 178 42,406 0.42%
query 3 0.01% understanding 35 10,239 0.34%
enigma 3 0.01% cause 14 15,021 0.09%
speculation 2 0.01% speculation 2 6,778 0.03%
Table 4.9 Pattern: N-e-W/

Noun Freq. in Attraction Noun Freq. in Freq. in Reliance

pattern pattern corpus
question 959 56.02% question 959 42,406 2.26%
problem 277 16.18% puzzle 27 1,349 2.00%
thing 178 10.40% conundrum 5 345 1.45%
issue 77 4.50% unknown 4 302 1.32%
mystery 46 2.69% mystery 46 5,949 0.77%
puzzle 27 1.58% dilemma 19 2,862 0.66%
dilemma 19 1.11% problem 277 59,600 0.46%
difficulty 14 0.82% snag 3 784 0.38%
idea 13 0.76% consideration 12 4,324 0.28%
consideration 12 0.70% issue 77 32,881 0.23%
challenge 12 0.70% thing 178 80,013 0.22%
factor 11 0.64% difficulty 14 8,215 0.17%
concern 6 0.35% enigma 1 677 0.15%
conundrum 5 0.29% factor 11 10,306 0.11%
clanger 4 0.23% challenge 12 14,460 0.08%
unknown 4 0.23% achievement 3 5,483 0.05%
snag 3 0.18% breakthrough 1 3,079 0.03%
achievement 3 0.18% idea 13 46,654 0.03%
enigma 1 0.06% concern 6 17,148 0.03%
breakthrough 1 0.06% danger 4 14,398 0.03%
Chapter 5
Semantic prerequisites
Why can some nouns be used in expressions like the that I have no
money ... and the is that I have no money, while others cannot (see ex-
amples 1.1 and 1.2 in Chapter 1)? What are the semantic properties of those
nouns that allow speakers to use them as shell nouns? Previous research
suggests that the properties of abstractness (Ivanic 1991: 98-101) and un-
specificity (Winter 1992, Francis 1994: 83) play a key role. Ivanic and
Francis more or less raise these two properties to the status of defining cri-
teria. In contrast, I would like to claim that these semantic properties of
shell nouns should be regarded as no more than prerequisites which enable
the nouns to serve higher communicative and cognitive functions.
5.1 Abstractness
The impression that an expression or even a whole text is 'abstract' can

arise from various sources. In this section I will discuss the two main fac-
tors that contribute to the impression that shell nouns are abstract. First, the
idea will be examined that what these nouns denote and can refer to is ab-
stract in the sense that it cannot be seen or touched. As will be seen, a
slightly more sophisticated approach to extensional abstractness based on
Lyons' distinction between first, second and third-order entities is more
illuminating than this simple view. Second, it will be shown how the use of
shell nouns can lead to the effect that a given text has an 'abstract style'.
5.1.1 Extensional abstractness and classes of abstract entities
The most common way of conceiving of the abstractness of words is in

terms of the nature of their denotata. From such an extensional perspective,
abstract nouns are those nouns whose denotata are not part of the concrete
physical world and cannot be seen or touched. Strictly speaking, what is
abstract is not the nouns themselves, of course, but what they denote.
64 Semantic prerequisites
It is quite obvious that many shell nouns, for example fact, feeling, idea,
intention, hope and view, refer to the world of intangible notions and ideas.
For other nouns the case is more debatable. For example, the abstract na-
ture of the nouns area, region and site, which are used in the pattern ~N-wh
to refer to places and locations, is somewhat doubtful, since locations are
part of the concrete physical world. Even more questionable is the abstract-
ness of shell nouns denoting activities, such as campaign, mistake, refusal
and reaction. The noun campaign, for example, typically denotes a large
range of activities which undoubtedly take place in the physical world, ei-
ther in the field of politics, business and the media (e.g. taking part in dis-
cussions, giving speeches and interviews, producing papers, pamphlets and
advertising material), or in the even more concrete and real world of battles
and warfare.
This shows that such a simple extensional conception of abstractness is
too crude for our purposes and a more refined view of abstractness is re-
quired. One conception of abstractness that promises deeper insights has
been put forward by Lyons (1977: 442-445; 1979: 93-95, 1989: 168-
170).35 In place of the traditional two-fold dichotomy between concrete and
abstract Lyons introduces a tripartite distinction between first, second and
third-order entities.36 The guiding principle is the wish to apply the onto-
logical assumptions underlying a so-called 'naive' realism. Leaving aside
philosophical technicalities (see Lyons 1977: 109-114, as well as e.g.
Bennett 1988, Parsons 1990, Asher 1993, Peterson 1997), this means that
he suggests a common-sense classification, untainted by sophisticated
scholarly considerations, of what there is in the world.
First-order entities are persons, animals, other organisms and physical
objects which are located in space and have fairly constant perceptual prop-
erties. Lyons suggests that these characteristics make them good targets for
reference. In our context it is important to add that, conversely, these char-
acteristics turn them into very bad candidates for fulfilling the function of
temporary concept-formation (see Section 2.2). Second-order entities are
events, processes and situations "which are located in time and which, in
English, are said to occur or take place, rather than to exist" (Lyons 1977:
443). Finally, third-order entities are 'abstract' entities such as concepts,
propositions, or more generally, ideas outside place and time. As far as the
way they are encoded by language is concerned, Lyons points out that
reference to both second-order entities and third-order entities is made

most commonly, both in English and in other languages, by means of
Abstractness 65
phrases formed by the process of nominalizations. But there is a fairly

clear difference in English between the set of nominalizations that is ap-
propriate for the one purpose and the set of nominalizations that is appro-
priate for the other (cf. Vendler, 1968). (Lyons 1977: 445)
The reference to Vendler's work (which was mentioned in Chapters 1 and 2

above) can be taken as an indication that the term nominalization must here
be understood in the wider sense used by philosophers and linguists in the
generative paradigm (e.g. Chomsky 1970). This includes not only the word-
formation process of the transposition of a verb into the class of nouns, but
also that-c\a.uses, infinitives, gerunds and other constructions which can
function in lieu of nouns or noun phrases syntactically (Vendler 1967: 122-
140). While I have primarily been interested in the abstractness of shell
nouns so far, we will see shortly that this conception of nominalization
allows me to widen the scope to include the question of the abstractness of
shell contents, too.
The major improvement of Lyons' tripartite distinction lies in the cate-
gory of second-order entities. This class captures the intuition that events,
states, processes and activities are more abstract than persons and things in
the sense that they have no stable existence in the dimension of time, but
less abstract than ideas and propositions because they take place in the
physical world. Second-order entities are "observable [...] and have a tem-
poral duration", while third-order entities "can be asserted or denied, re-
membered or forgotten" (Lyons 1977: 445). The same difference is opera-
tionalized in an explicitly perception-related, cognitive framework by Horie,
who distinguishes between "directly/physically perceived events" on the one
hand, and "indirectly/mentally perceived events" (1991: 234) on the other.
Both second-order and third-order entities can be further subcategorized.
In doing so, it is important to emphasize that the typology and characteriza-
tion of the subtypes that will be put forward are specifically tailored for this
study. Although it is based on the work of linguists (e.g. Quirk et al. 1985:
202, Halliday 1994: 250) and philosophers (Vendler 1968, Bennet 1988,
Parsons 1990, Asher 1993, Peterson 1997), the proposed typology is not
true to any single one of them. A survey of the subcategorization of second-
order and third-order entities used in this study is given in Figure 5.1.
I use the term event as a superordinate term for the different types of
second-order entities in this study. In the philosophical literature, the term
eventualities coined by Bach (1981) is often found in this function (see
Parsons 1990 and Asher 1993), but the term is not commonly used in lin-
guistics. With the help of the two dimensions [DYNAMIC] and

[AGENTIVE] selected from the typology of situation types suggested by
Quirk et al. (1985: 202), I define activities as [+DYNAMIC] and
[+AGENTIVE] events, processes as [+DYNAMIC] and [-AGENTIVE] events
and states as [-DYNAMIC] and [-AGENTIVE] events. A subclass of activi-
ties, which is not included in Figure 5.1 but will prove to be relevant in
Chapters 9, 10 and 11, are accomplishments. In addition to the features
[+DYNAMIC] and [+AGENTIVE], accomplishments are characterized by the
feature [CONCLUSIVE] (see Quirk et al. 1985: 206-207, and Section 9.3.3
for more details).
Events Abstract relations

(second-order entities) (third-order entities)
activities processes states facts ideas utterances
[+DYNAMIC] [+DYNAMIC] [-DYNAMIC] states of objects linguistic ex-

[+AGENTIVE] [-AGENTIVE] [-AGENTIVE] affairs of thought pressions of
ideas
Figure 5.1 Typology of second-order and third-order entities
The subcategorization of third-order entities is less straightforward. To start

with, it is much more difficult to come up with a superordinate term to
match the term event in the domain of second-order entities. The best candi-
date, proposition, not only suffers from a proliferation of uses in language
philosophy (Searle 1969), semantics (Fillmore 1968, Dirven and Radden
1977), text linguistics (van Dijk and Kintsch 1983) and pragmatics (Falkner
1997), but is also unsuitable for third-order entities expressed by single
nouns. Asher uses the term purely abstract objects, which I find irritating
because of its length and the clash between the adjective abstract and the
reifying noun object. The only feasible alternative that I can think of is ab-
stract relation; hence I shall use this term as a counterpart for event, even
though it is perhaps still not wholly satisfactory.
On the basis of the works cited above, I subcategorize abstract relations
for the purposes of this study into facts, ideas and utterances. Facts are
abstract relations that are conceived of as 'being the case'. They are not
invested with any epistemic or truth-conditional claims but closely corre-
Abstractness 67
spond to what are called states of affairs in everyday English. An important

subclass of facts, especially for modal shell nouns, are possible facts or
facts in possible worlds. Ideas are objects of thought. Ideas correspond
most closely to what are often called propositions, but they also include
objects of thought expressed by single nouns such as idea, thought or as-
sumption, which do not fall under the notion of proposition because they are
not "saturated" (Asher 1993: 15). Finally, utterances are linguistic expres-
sions of ideas. From a philosophical point of view, the subcategory of utter-
ances is clearly the odd man out. Yet both from a linguistic point of view in
general (see e.g. Halliday 1994: 263-264, who refers to them as locutions),
and especially considering the use of shell nouns as reflected in my data, the
category is of essential importance.
How can the ontological status - or better conceptual status, since it es-
sentially depends on the mind rather than 'objective' reality - of shell nouns
and shell contents be determined? To begin with, one must realize that there
is a crucial difference between the two, even though both shell nouns and
shell contents can represent either events or abstract relations. Shell nouns
(and shell-noun phrases) are conceptually incomplete. Taken in isolation,
they cannot evoke full-fledged, saturated thoughts in Frege's (1976; 1918)
sense. Shell contents, on the other hand, which are linguistically realized by
clauses or even larger linguistic units, are saturated in this sense; they rep-
resent full propositions.
Paradoxically, it is nevertheless easier to decide on the conceptual status
of shell nouns than on that of shell contents by means of simple semantic
and conceptual analysis. Test frames like occurred at 10 o 'clock,
which accommodate only events but not facts (Vendler 1967) also help.
Nouns like event, act, action or move can thus be identified as representing
events, and fact, reason and idea as representing abstract relations. Turning
to shell contents, which are expressed by clauses, their propositions alone
can be indeterminate (Vendler 1968: 26-31, Menzel 1975: 45), as long as
they are realized by simple sentences. Consider the example She was wear-
ing a green hat. Even if what the philosopher Salmon (1986: 6) calls the
"propositional-term-forming sentential operator" that is added (yielding that
she was wearing a green hat), the conceptual status of this clause is unde-
cided. It can be turned into a fact by inserting the noun fact (the fact that
she was wearing a green hat), although its propositional content represents
a physically observable state. Conversely, when the propositional content of
a clause is extensionally abstract, as for example in Inflation has risen
again, it cannot be turned into an event by linguistic means. *The event that
inflation has risen again or *the process that inflation has risen again are
ungrammatical. This means that the shell noun can determine the ontologi-
cal/conceptual status of a proposition when it is indeterminate with regard
to extensional abstractness. When the proposition expressed as shell content
is abstract because of the meaning of its constituents, the shell noun must
also represent an abstract relation.
A look at some examples from the corpus will show how an application
of the typology presented in Figure 5.1 can lead to a more differentiated
view of what lies behind the abstractness of shell-content complexes. Start-
ing with shell contents, the infinitive clause in (5.1) basically represents a
concrete event. The mental noun aim, however, affects the conceptual status
of the shell content such that it turns it into an idea. In (5.2), on the other
hand, both the shell noun irony and the shell content represent abstract rela-
tions anyway.
(5.1) Their aim is to meet President Saddam in Baghdad, (PAPERS)

(5.2) The irony of the anti-hunting councillors' action is that this motion will
not have any significant effect on hunting in Hampshire, (MAGS)
The material collected and analysed for this study clearly suggests that shell
contents representing abstract relations by far outweigh those representing
events.
Mention should be made in passing that even first-order entities can
function as shell contents. This is possible when mental shell nouns such as
idea, notion and concept occur with o/-prepositional phrases, as in example
(5.3). However this case is comparatively rare.
(5.3) a sad person's idea of a gay person (PAPERS)
Looking next more closely at some of the shell nouns themselves, one
first realizes that the intuitively best examples like fact, case, reason, idea,
chance, possibility or aim all represent abstract relations. Nouns of this
type make up the bulk and prototypical core of the class of shell nouns (see
Section 6.1). Nouns such as campaign, reaction, trick or mistake stand for
observable physical events and are thus second-order entities. They can
function as shell nouns as well, but they represent less typical cases because
they are less frequent, collocationally less versatile than shell nouns denot-
ing abstract relations, and are often not directly linked to their shell contents
by the relation of experiential identity (see Sections 3.1.2 and 6.1). For ex-
Abstractness 69
ample, although campaign is a fairly frequent shell noun, at least in the

texts from media sources, it cannot be used with that-clauses and does not
occur in the pattern N-e-cl. And uses in the pattern N-cl also involve a
strong modal component in addition to the link of experiential identity, as in
the campaign to bring down the coalition government (PAPERS).
Two further groups of shell nouns deserve particular attention with re-
gard to the categorization as events and abstract relations. The members of
the first group are interesting because they have the systematic capacity of
denoting either events or abstract relations, indeed even both events and
abstract relations in many cases. Although other items belong to this group,
too, this property can best be illustrated with deverbal suffixations of
speech act verbs like proposal, report or declaration. Such nouns can be
used to refer to specific speech acts, in which case they represent events.
Alternatively they can denote the propositional contents of speech acts,
which must then be looked upon as abstract relations, or more precisely,
utterances.37 Relatively clear examples of the two readings, which are in
fact not easy to find, are given in (5.4) and (5.5) respectively:
(5.4) In a speech to Fianna Fail women in February 1964 he [...] made the
radical proposal that women deputies should play an active role in vet-
ting legislation in which they had a special interest, (BOOKS)
(5.5) QJ The Spanish minister then acknowledged he could not accept the
principle of the tax until the details of the proposal were made much
clearer, (PAPERS)
That the noun phrases headed by proposal in the two examples represent
different conceptual entities can be demonstrated with the help of para-
phrases. The expression made the radical proposal in (5.4) can be para-
phrased by something like proposed or performed the act of proposing.
This shows that an activity is being referred to. Any paraphrase containg
act of proposing is unacceptable for (5.5). Instead, a paraphrase highlight-
ing the propositional content of a speech act, e.g. the details of the content
of the proposal, is possible. The systematic ambiguity of derived speech act
nouns will be discussed in more detail in Section 8.3.
The second group consists of such superordinate nouns as state, event,
action, act, activity, process and move. States, events, activities and proc-
esses are physically and directly observable. As a consequence, these nouns
must be regarded as event-denoting linguistic items, and Lyons duly men-
tions them as "second-order nouns" (1977: 446). On the other hand, the
nouns are highly abstract ways of expressing such observable events. Here,
the crucial aspect is that nouns, or, to be more precise, intensionally unspe-
cific nouns are used to express events, and this brings about stylistic rather
than extensional abstractness.
5.1.2 Stylistic abstractness and grammatical metaphor
The experiences denoted by such nouns as event, action, state and process
are not so much abstract in extensional terms; rather they seem to be stylis-
tically abstract ways of talking about physical and concrete events.
In Cognitive Grammar, this phenomenon is accounted for as a particular
type "construal" of a conceptualization (Langacker 1987a, 1991). A central
part of how a conceptualization is construed by linguistic means resides in
the speaker's choice of words, and particularly in the word-classes to which
these words belong. For example in (5.1) above, the pre-linguistic cognitive
unit of 'pursuing something that one wants to achieve' is construed as a
'thing' because the speaker has opted for the nominal expression Their aim
is to ... . If he or she had chosen the corresponding verbal expression They
aim to ... , this would be interpreted in Cognitive Grammar as the construal
of a relation, or more specifically, of a process (Langacker 1987a: 244-274;
see also Ungerer and Schmid 1996: 191-194, and Section 17.2 below). The
basic idea behind the notion of construal is that speakers can use different
linguistic expressions to conceptualize situations from different cognitive
perspectives, even when they are more or less identical from an objective
point of view. Apart from the choice of word class, the distribution of figure
and ground within the clause, the perspective taken on a situation and the
allocation of attention to various aspects of the situation play important
parts (see also Section 6.2).
While cognitive linguists are mainly interested in how conceptualizations
are reflected in the language we use, they are less concerned with the spe-
cific stylistic effects of different linguistic construals. Here Halliday's Sys-
temic-Functional Grammar is more helpful. The central concept in the con-
text of shell nouns is Halliday's notion of "grammatical metaphor" (1994:
342-343). The basic idea behind this notion is that real-world states of af-
fairs can be expressed in two ways, either with 'literal' expressions or with
corresponding metaphorical expressions. For example, Halliday regards the
clause Mary saw something -wonderful as a literal statement, and the corre-
sponding clause Mary came upon a wonderfiil sight as a grammatical
Abstractness 71
metaphor (1994: 343-344). To mark the differences between 'traditional'

lexical metaphors and grammatical metaphors Halliday speaks of 'congru-
ent' rather than 'literal' expressions. 'Congruent' presumably means 'con-
gruent with the type of experience', although he does not make this entirely
clear. The notion of grammatical metaphor is helpful in the present context
because it illuminates the relation between 'concrete' and 'abstract' ways of
expressing concrete events. This can contribute to our understanding of
shell nouns.38
Let us first look at the importance of stylistic abstractness for two dif-
ferent types of shell nouns. Examples (5.8) and (5.9) are a convenient
starting-point for such a comparison:
(5.8) ... his Buddhist mind-training exercises led him to hold the firm belief
that the mind has unrecognized powers, (BOOKS)
(5.9) A very minor flaw lay in the fact that Tom Wood, alone among his
travelling companions, had met Markham face to face, (BOOKS)
Like example (5.1) above, example (5.8) exhibits an obvious parallel be-
tween the reporting expanded predicate (hold) the firm belief that... and a
corresponding reporting verbal expression such as believe firmly that.... In
both cases the /-clauses are what Halliday calls "projections" of ideas,
i.e. clauses which do not function as direct representations of experience but
as representations of representations (Halliday 1994: 250-273, 263-264; see
also Section 2.1). The conspicuousness of this parallel indicates that stylis-
tic abstractness plays a role here. Undoubtedly the author could have opted
for a verbal way of expressing what the other person's mental state was,
and this shows that in addition to projection, grammatical metaphor is at
work.
This is not the case in example (5.9). Trying to paraphrase the nominal
expression the fact that Tom Wood [...] had met Markham face to face,
one finds that this is a much less easy and straightforward task. In order to
do this, one would have to split up the clause and start out from the pro-
jected event. The result of an attempt to apply this strategy is given in (5.9'),
but one can hardly claim that this is a particularly convincing sample of
English which renders the communicative value with any precision.
?
(5.9') Among his travelling companions Tom Wood alone had met Markham
face to face, and this was annoying.
Arguably, the most unsatisfactory aspect in the paraphrase in (5.9') is the

use of this in extended reference. Trying to insert the shell nouns event and
fact - yielding this event or this fact respectively - one can unveil the func-
tion of the noun fact in (5.9): this event is unacceptable, while this fact is at
least possible. This shows what the noun fact does in this example. It
changes the conceptual status of the piece of experience denoted by the that-
clause in (5.9) from event to fact. In Asher's formal framework, this proc-
ess is captured by a so-called "abstract argument transformation" (1993:
159). Halliday would account for this by claiming that the content is pro-
jected as "fact" (1994: 264). As a consequence, the noun fact is less easily
replaceable than the noun belief m (5.8) and the whole expression is much
more difficult to paraphrase. What this suggests is that grammatical meta-
phor plays a much less important role in example (5.9) than in (5.8), as
there seems to be no satisfactory congruent way of expressing the same
experience; the shell-noun expression seems to be the only natural way of
expressing what is to be said and can thus be seen as congruent speech.
The outcome of my first comparison, then, is that some shell nouns bring
about a conceptual transformation from event to abstract relation but do not
contribute more to stylistic abstractness than is required by the experience
they verbalize.
Conversely, there are shell nouns which contribute to stylistic abstrac-
tion but do not change the conceptual status of their contents. These need
not always occur as parts of expanded predicates, as was the case in exam-
ple (5.8) above. Example (5.10) is a case in point:
(5.10) ... Sarah said: "He deserved it. I have quite a fertile imagination and I
know what hurts him and where his soft underbelly is and that's his
material possessions. My reaction was to hit him where it hurts. You're
not with someone for 27 years without knowing that." (TODAY)
Although the underlined shell content includes 'abstract' semantic elements

(partly because of metaphoric meanings), it must still be seen as represent-
ing a concrete event. The shell noun reaction itself also describes an event
or, more precisely, an activity. So neither the shell content nor the shell
noun are abstract in the sense that they denote entities which cannot be per-
ceived directly and physically. Nevertheless one has the distinct feeling that
the noun reaction advances the 'abstractness' of the passage. This feeling
can be explained if one combines Lyons', Langacker's and Halliday's ap-
proaches: it is the use of nouns rather than verbs as predicates of clauses for
Abstractness 73
the verbalisation of second-order entities that gives rise to the perceived

abstractness of a text. This abstractness is not extensional but mainly sty-
listic in nature, since the things that are being talked about belong to the
world of concrete activities and events. The use of such nouns results in
conceptualizations of activities, properties or relations as 'things', but it
does not involve an upgrading to third-order in Lyons' tripartite division of
entities.
I summarize my considerations on extensional and stylistic abstractness,
and on the interaction between the abstractness of shell nouns and shell
contents as follows:
a) Both shell nouns and shell contents may express either events or ab-
stract relations. Shell contents alone may be indeterminate with regard
to this distinction.
b) When a shell noun representing an abstract relation is linked with a
clause also representing an abstract relation, stylistic abstractness may
or may not play a role, depending on whether the same experience can
also be expressed by verbal means or not.
c) When a shell noun representing an abstract relation is linked with a
clause representing an event, stylistic abstractness is not involved be-
cause the shell noun fulfills the function of turning the event into an
abstract relation. In this case it cannot easily be replaced by a non-
nominal paraphrase.
d) When a shell noun representing an event is linked with a shell content
also representing an event, stylistic abstractness is clearly at work,
while there are no changes with regard to extensional abstractness.
To complete this systematic survey of the interaction between shell nouns
and shell contents one may like to raise the question of what happens when
a shell noun representing a second-order entity is linked with a shell content
representing a third-order entity. As pointed out in Section 5.1.1, the answer
to this question is very simple: this combination does not occur because it is
impossible to turn ideas and propositions into physical events by linguistic
means.
5.2 Unspecificity and structure-inherent semantic gaps
Such nouns as event, action, act, process and state represent second-order
entities but are special in that they are highly unspecific means of express-
ing them. In Halliday and Hasan's framework they are general nouns
(1976: 274-275). In everyday language, semantic unspecificity, generality
or schematicity (as it is called in Cognitive Grammar) is often confused
with abstractness. This is shown for example by the entry for one sense of
the adjective abstract in the Longman Dictionary of Contemporary English
(Summers 1995; henceforth "LDOCE3"), which reads "based on general
ideas or principles rather than specific examples or real events".
In linguistics, unspecificity is commonly understood not in extensional
but in intensional terms. The meaning of a noun is unspecific if it is deter-
mined by only one or a very small number of semantic dimensions.
Looking at the nouns listed in the Tables in Section 4.5, one finds that
many of them exhibit the property of intensional unspecificity. The nouns
mentioned at the beginning of this section, for example, can roughly be
paraphrased by 'something happens' {event, process), 'someone does some-
thing' (act, action) and 'something is the case' (state). Obvious as it may
seem, it is still worth noting that the only semantic dimension evoked by the
noun time is 'temporal', the only semantic dimension of place is 'locative'
and the only semantic dimension evoked by the noun fact is 'factual'.
Analogous one-dimensional meanings can be attributed to the nouns idea
and thought (sole dimension 'mental'), and reason and cause ('causal'). It
lies in the very nature of these general nouns as "superordinate members of
major lexical sets" (Halliday and Hasan 1976: 275) that their senses are
more or less mono-dimensional.
However, even a cursory glance at the lists in Section 4.5 suggests that
by no means all shell nouns are unspecific in this sense. This impression can
be confirmed by looking up the meaning of a random example, say the noun
opportunity, in a dictionary. The entry for this noun in LDOCE3 reads "a
chance to do something or an occasion when it is easy for you to do some-
thing".39 Attempting a fairly informal and naive translation of this definition
into a semantic metalanguage (see the discussion in Section 6.2), the feature
[NON-FACTUAL] can be extracted from the notions chance and occasion,
[TEMPORAL] from the word when, [AGENTIVE] from you and [DYNAMIC]
from do. Even when carried out on such a pseudo-technical level, the de-
scription in terms of features makes one wonder whether the semantic
structure it describes can still be called unspecific.
Much more complex and specific meanings can easily be found if one
actually starts looking for them. For example, the shell-noun specific
meaning of the noun irony is described by the entry "a situation that seems
strange and unexpected or amusing, or the reason it is like this" in
Unspecificity and structure inherent semantic gaps 75
LD0CE3. A proper translation of this definition into the metalanguage

would not just result in a fairly long list of features but would require that
the hierarchies, dependencies and other relations between those features be
taken into account, for example the fact that a situation can be the reason
for a cognitive state.
These examples demonstrate that shell nouns are not semantically un-
specific in the general sense advocated for example by Winter (1992). On
the one hand, semantically specific shell nouns do exist, and on the other
hand, there are such highly unspecific nouns as creature, person or object
(in a concrete reading), which cannot function as shell nouns. Nevertheless,
the intuition that shell nouns are unspecific in a certain way remains. Even
opportunity and irony, the examples just discussed, create this impression,
but it is difficult to pinpoint what the precise nature of their unspecificity is.
What are the solutions offered by previous researchers of shell-noun-like
phenomena? Ivanic (1991) looks at the unspecificity of carrier nouns from
a predominantly syntactic perspective. Discussing the example purpose she
claims that this concept is "like a syntactic specification with slots to be
filled according to each new context in which 'purpose' plays a part" (1991:
94). From the perspective of discourse relations, Francis (1994) makes the
following suggestion about the type of unspecificity that is involved. She
argues that
the main characteristic of [...] a label is that it requires lexical realization,

or lexicalization, in its co-text: it is an inherently unspecific nominal ele-
ment whose specific meaning in the discourse needs to be precisely spelled
out (Winter 1982, 1992). (Francis 1994: 83)
As the references indicate, this suggestion goes back to the work of Winter.
In his earlier publication Winter (1982: 185-186) argues that there is a type
of clauses, called unspecific clauses, which require information given in
other clauses, so-called lexical realisations, in order to be semantically
complete. In the later publication mentioned by Francis, Winter (1992)
takes a much more detailed look at nouns with unspecific meanings. He
maintains that there is a large group of inherently unspecific nouns which
serve as a "metalanguage for the clause". These nouns "do not refer to con-
crete things in the outside world", but " 'talk about' the nature of the clause
or sentence as a message in the text itself" (Winter 1992: 133). To be
communicatively effective, they have to be made specific or, as he calls it,
"lexically unique" (1992: 153).
Building on Francis', Ivanic's and Winter's work, I want to claim that

the unspecificity of shell nouns is semantic in nature after all, but in a more
specific way than was outlined above. It is a specific type of unspecificity
which can in fact be gleaned from the term shell noun and the shell meta-
phor underlying it (see Section 2.1). For shells to be able to function as
containers it is necessary that they have a gap, a hole, or some other kind of
opening or dent which can receive the content. Likewise, if a noun is to
function as a shell noun, its semantic structure must include one or several
gaps that can be filled in by the information given in the shell content.
Consider for instance the noun reason,40 As is typical of shell nouns, the
noun reason itself provides information of a very limited nature. It does
convey the information that one thing is causing or has caused some other
thing but it gives no clue as to what these things are. By evoking a two-
place relation between cause and effect, the noun reason sets up two clearly
defined semantic gaps which need to be filled. However, when it comes to
specifying these things the noun itself misses out and must rely on the con-
text to supply the necessary information, a characteristic which is of course
again typical of all shell nouns. Thus the meaning of the noun consists of
two parts: a stable and relatively well-determined semantic structure on the
one hand, and two gaps which, depending on the contexts in which the word
is used, can be filled in by a variety of pieces of information (see also Ivanic
1991 and Section 2.2 for this combination of stability and variability)41
Generalizing from this example, it can be claimed that there are specific
semantic gaps inherent in the meanings of all shell nouns. These gaps need
to be filled in order to render the nouns communicatively effective and suc-
cessful. The gaps create the impression that shell nouns are intensionally
unspecific altogether, which, as has been shown, is not the case. Since the
nature of the gaps depends on the specific meaning of a given noun, they
defy general explanations in terms of intensional unspecificity.
Both the syntactic and the discoursal properties of shell nouns fall into
place once this semantic characteristic of shell nouns has been recognized.
For the gap is responsible for the syntactic property that they can be used in
the patterns N-cl and N-e-cl (see Chapters 1 and 3), and for the discoursal
property mentioned by Francis and Winter that shell nouns require "lexical
realisation" in the co-text. The existence of a structure-inherent semantic
gap is also a precondition for the link of experiential identity which lies at
the basis of shell-content complexes (see Section 3.1.2). Since the semantic
characteristics may sometimes be difficult to assess, the best diagnostic for
the existence of a structure-inherent gap which allows a noun to function as
shell noun is whether it can be used in the patterns N-cl and/or N-e-cl.
Therefore this criterion was operationalised as the factor determining the
inclusion of a noun in the class of possible shell nouns (see Section 4.2).
Non-shell nouns such as tree, clerk or religion cannot be used in any of the
patterns in question, as was already shown in the examples in (1.2) at the
very beginning of this study.
To a certain extent, the claim that shell nouns have a structure-inherent
semantic gap can be substantiated from a purely semantic perspective as
well. This can be shown by drawing once more on dictionary entries and
pointing out the way in which the particular type of unspecificity of shell
nouns is reflected there. This line of argument has the advantage that it is
not based on syntagmatic considerations but exclusively on the meaning of
the nouns in question.
In the definitions for opportunity and irony, which, as we have seen, are
intensionally fairly specific, the inherent semantic gaps can be detected in
the uses of general dummy elements. An opportunity is described in
LDOCE3 as a chance or an occasion when it is easy to do something, and
irony is defined as a situation which has certain properties and can be a
reason for something. Although these dummy elements are general enough
to be able to function as dummies, most of them indicate constraints as to
what kind of event or abstract relation is required to fill in the gap. The
description of opportunity suggests that the gap in the semantic structure
must be filled by an expression representing an activity, and the description
for irony suggests a gapped state of affairs and a gapped fact acting as
cause. The pronoun something and other pro-forms can also be found in the
following entries for six nouns which have been selected at random from the
lists in Section 4.5. The entries are taken from LDOCE3 and have been
adapted by omitting irrelevant material.
fact a piece of information that is known to be true

intention something that you intend to do
way a method of doing something, or a manner in which
something can happen or be done
attempt an act of trying to do something
upshot the final result of a situation
remark something that you say when you express an opinion
Admittedly this random collection of examples gives only a glimpse of the

issue at stake. Yet it may still be seen as an indication that definitions of
shell nouns tend to include clues for the existence of a structure-inherent
gap. The gaps of four nouns in this list are marked by the dummy element
something, while the definitions offact and upshot make use of the dummy
element a piece of information and the general noun situation respectively.
The relative uniformity of the strategies used in LDOCE3 must not coax
us into believing that the gaps are all of a similar type. Different shell nouns
provide gaps for different types of events and propositions. This can be
revealed by translating the dictionary entries into the pseudo-metalanguage
already used above and trying to indicate the gap by visual means, as is
shown in Figure 5.2. It should be kept in mind that a proper semantic analy-
sis and description in terms of features and their dependencies and hierar-
chies would be a major effort in its own right, without being necessarily
more illuminating than this crude type of analysis and illustration.
fact
intention
way
attempt
upshot
remark
Figure 5.2 Diagrammatic illustration of structure-inherent semantic gaps
Fact, the most unspecific of these shell nouns, includes a very general gap
which may be filled by any kind of state of affairs. The stable semantic part
of the noun is that what it refers to is conceived of by the speaker as being

true. The noun itself gives no clues as to what state of affairs it is that is
portrayed as being true by the speaker. The structure of the noun intention
includes the information that someone is resolved to perform an activity, but
it gaps the nature of the activity itself. Similarly, the noun way is related to
the manner in which an event takes place or the method with which an ac-
tivity is carried out, but also gaps the precise nature of these events. The
noun attempt also gaps activities. The noun upshot gaps an event which
functions as one of the elements, namely as effect, in a cause-effect relation,
while the cause can be realized by a large variety of processes and situa-
tions. The noun remark, finally, opens up an ambiguous gap, because the
content can either represent the act of saying something, which highlights
the illocutionary force of an utterance, or the content of an act of saying
something. The latter sense corresponds to the propositional content of an
utterance (see Section 5.1.1 and 8.3.3 for more details).
These examples must suffice for a demonstration of the semantic gaps
inherent in shell nouns. They suggest that it is typical for these nouns to gap
intensionally unspecific notions such as 'activity', 'event', 'process', 'state
of affairs', 'situation', 'cause' or 'result'. It is in this respect that inten-
sional unspecificity does play an important role in the semantic structures of
shell nouns, after all.
Finally, one may wonder where these structure-inherent gaps come from
in the first place. For nouns like intention, attempt, upshot and remark,
which are derived from verbs, it is not difficult to suggest an explanation.
These nouns behave just like the corresponding verbs intend, attempt, shoot
(up) and remark, which are "ergnzungsbedrftig" in Frege's (1994; 1891)
sense, i.e. semantically incomplete, and only become saturated when they
are used in sentences. Although an equally straightforward explanation is
impossible for synchronically unanalyzable nouns like fact and way, or idea
and problem, it is highly interesting from an etymological point of view that
even these nouns ultimately go back to verbs. Thus, the noun fact is derived
from the Latin verb facere 'to do', idea from the Greek verb 'to see'
and problem from Greek 'to throw out', a prefixation from
'to throw'. According to the OED, even way can be traced back to
an Indogermanic verb with the meaning 'to carry'. Other shell nouns with
verbal roots in Latin or Greek are concept from L. concipere 'to conceive',
message from L. mittere 'to send', notion from L. noscere 'to know\ point
from L. pungere 'to prick, pierce', reason from L. reri 'to think' and the-
ory from Gr. 'to look on, view, contemplate'. Together with the
morphological evidence, these etymologies confirm an idea which was al-

ready proposed by Porzig (1930), that abstract nouns must be seen as reifi-
cations of the contents of whole clauses (see also Section 17.2). This also
accounts for the existence of gaps in their semantic structures.
5.3 Summary of Part I
Shell nouns allow speakers to put complex pieces of information into tem-
porary conceptual shells. They are a functional linguistic class whose defi-
nition is based on the three criteria temporary concept-formation, charac-
terization and link with shell content. Speakers make use of a fairly small
number of grammatical and/or text-referential devices to trigger the co-
interpretation of shell nouns and shell contents. The relation underlying
these links is that of experienced identity. This means that speakers and
hearers have the feeling that the shell nouns and the stretches of discourse
which express the shell contents are about the same thing or state of affairs.
In cognitive terms, shell nouns and the clauses and sentences expressing
shell contents activate closely related parts of the cognitive models created
by texts.
For the relation of experiential identity to be possible a given shell noun
must contain a structure-inherent semantic gap which can be filled in by the
shell content. The grammatical reflection of the existence of such a gap is
the ability of the nouns in question to occur with postnominal (non-relative)
that-, to- or w/i-clauses (N-cl) and/or as subjects of copula clauses with
that-, to- or w/2-clauses as subject complements (N-e-cl). These syntactic
properties are used as operational criteria for shell-nounhood: only nouns
which are used in these patterns with noteworthy frequencies are looked
upon as shell nouns in this study.
In addition to the specific type of semantic unspecificity, abstractness
can play a role in shell-content complexes in either one or both of two pos-
sible ways. On the one hand, shell nouns are means of expressing events or,
more typically, abstract relations (i.e. second-order or third-order entities
respectively in Lyons' framework). On the other hand, the use of shell
nouns can contribute to the stylistic abstractness of texts because they allow
speakers to verbalise events, activities and relations not just by means of
any kind of expressions, but by noun phrases. It must be emphasized, how-
ever, that this has a concretizing effect as well (see Section 17.2).
Summary of Part I 81
The formal, semantic and pragmatic descriptions of shell nouns and

shell-content complexes given in Parts II and III below are based on
authentic linguistic data. These have been retrieved systematically from a
225-million word section of the Bank of English (COBUILD). Being defined
in functional terms, the class of shell nouns is basically open-ended. It is
therefore essential to work with as representative a sample of shell nouns as
possible. This requirement is met by taking the frequencies of occurrence of
nouns as shell nouns, in terms of both types and tokens, into account. That
the frequency criterion is used as a yardstick can be justified by the From-
Corpus-To-Cognition-Principle as discussed in Section 4.1, and with refer-
ence to the practical experience of modern lexicographers, who also rely on
frequency as a criterion for representativity (see e.g. Summers 1996 and the
Introduction to LDOCE3).
The clues for the actual retrieval of instances of shell nouns from the
corpus are not the nouns themselves but their typical lexico-grammatical
environments. This strategy has two important advantages. First, with a
tolerable degree of interference only occurrences of nouns as shell nouns are
retrieved. And second, the retrieval method also functions as a heuristic for
the identification of nouns whose capacity to function as shell nouns was
not known before the corpus analysis. This is particularly important, since
an exhaustive list of those nouns which can function as shell nouns did not
exist. Such a list does not even emerge as a result of this study, because
more than the 670 nouns included here have the potential to occur as shell
nouns. The merit of the quantitative corpus approach, however, is that those
670 nouns that do make up the material for this study are the linguistically
and cognitively important ones.
Part II
The use of shell nouns
Having constructed the necessary theoretical framework I will now proceed

to a detailed description of how shell nouns are put to use by speakers. This
second part of the study can thus be seen as a contribution to descriptive
linguistics which is badly needed, because systematic linguistic accounts of
the use of shell nouns, and of abstract nouns in general, have not been
available so far. While this part is clearly not intended to be a usage record
or even dictionary of shell nouns, at some stages it comes quite close to
such, because fairly detailed semantic, grammatical, collocational and
pragmatic information is provided. Observations about the usage and collo-
cations of individual nouns are sometimes made in their own right, without
the aim of gaining additional semantic or pragmatic insights.
Since it is of course impossible to supply comprehensive descriptions of
the use of as many as 670 nouns, not all nouns can receive equal attention.
The quantitative analysis of my data outlined in Chapter 4 is one obvious
guideline for my choice of focus. Nouns with high scores for frequency in
the patterns and with high scores for reliance are more important than those
with low scores. Another basis for the distribution of attention, the degree of
typicality for the class of shell nouns, will be discussed in the first section of
Chapter 6. The semantic framework for the classification of shell-noun
uses, on which the division of this part into Chapters 7 to 12 is based, is
presented in Section 6.2.
Chapter 6
Describing shell-noun uses
6.1 Degrees of typicality
Not all shell nouns are equally good shell nouns. At several stages in the
discussion so far it has emerged that some nouns are better examples of
shell nouns than others, for example with respect to the purity of the link of
experiential identity between shell noun and content, frequencies of occur-
rence as shell noun, reliance scores, freedom of occurrence in different
patterns, ontological/conceptual status, potential for projection of events as
abstract relations and intensional unspecificity. Although these parameters
do not always converge, they allow for an assessment of single nouns and
groups of nouns with regard to their typicality for the class of shell nouns.
The members of the class of shell nouns can thus be arranged like a proto-
type category, ranging from a core of prime examples, over good and less
good to peripheral members or borderline cases. The following typicality
levels of shell nouns can be distinguished:
Prime shell nouns

The central core of the class of shell nouns consists of nouns representing
third-order entities which are the only means of conceptualizing the par-
ticular type of experience they represent. Nouns of this type are concept,
fact, issue, principle, problem and thing. Morphologically, the uniqueness
of these nouns is reflected in the fact that they are unanalyzable nominal
concepts and not, at least not in a synchronic description of the English
language, derived from either verbs or adjectives that could also be used to
express the same types of experiences. Another distinction of prime shell
nouns is that the relation of experiential identity between shells and contents
is usually very pure. It will emerge that most prime shell nouns belong to
the semantic classes of factual nouns, with some groups of mental and lin-
guistic nouns being included as well. Examples are fact, thing and reason,
idea, notion and concept; and message, rumour and legend.
86 Describing shell-noun uses
Good shell nouns

Most good shell nouns also represent third-order entities. Unlike prime shell
nouns, however, they are not the only means of projecting what they project.
In any given use a good shell noun can easily be paraphrased by verbal or
adjectival expressions (see Section 5.1.2). So the main difference between
prime and good shell nouns is that the former are in a sense less easily re-
placeable than the latter, and therefore more indispensible for the job they
are set to do. Most good shell nouns have morphologically related verbs and
adjectives. Typically, good shell nouns belong to the classes of linguistic,
mental and modal shell nouns. Examples are order, proposal and guess;
belief, assumption and plan; likelihood, certainty and permission. Some
specific eventive shell nouns with marked modal components, for example
attempt, effort and habit, are also regarded as good examples of shell nouns
Less good and peripheral shell nouns

Uses of nouns can be less good or peripheral instances of shell nouns for a
number of reasons. Many less good shell nouns do not represent abstract
relations but events. Since event-denoting nouns cannot be combined with
shell contents representing abstract relations (see Section 5.1.2), their appli-
cability is much more restricted than that of prime and good shell nouns.
Compared to prime and good shell nouns, less good shell nouns occur much
less frequently in the patterns N-cl and N-e-cl. The relation between less
good shell nouns and their contents can be more complex than that of pure
experiential identity. Less good shell nouns belong to the classes of eventive
and circumstantial shell nouns. It is especially in the latter class that one
finds peripheral instances of shell nouns whose membership in this func-
tional class is questionable. This is the borderline area between uses of
nouns as shell nouns and uses in other functions. Examples of less good
shell nouns are move, measure and reaction, and situation, way and proce-
dure. On the very fringes of the class of shell nouns, there are temporal
nouns (time, stage) and especially locative nouns (place, area), which
strictly speaking are not extensionally abstract and do not include a struc-
ture-inherent semantic gap (see the discussion at the beginning of Section
12.1, for reasons why these nouns are treated as occurring in shell-noun
uses at all.)
Features and frames 87
6.2 Explaining the meanings of shell-noun uses: features and frames
Starting out from the typicality gradient, I will now group uses of shell
nouns into semantic classes. Such a categorization of shell nouns is neces-
sary in order to gain an overview of the data and helpful for the differentia-
tion and description of the nouns. What is more, as will transpire in Part III,
it is also relevant for the pragmatic perspective on shell nouns. The time-
honoured linguistic custom of classifying is thus not just an end in itself.
The classification I want to propose allows for multiple category mem-
bership, especially of those shell nouns that can convey a particularly large
number of different meanings, e.g. point or position. To meet this require-
ment, the groups are thought of as consisting of uses rather than meanings
of shell nouns. This pragmatic perspective seems to provide a more elegant
way of dealing with the apparent proliferation of senses of some highly
unspecific shell nouns than to claim that they have an enormously large
number of different meanings. As a consequence of this approach, the
classes and subclasses of shell-noun uses should not be equated with the
word fields or lexical fields known from structuralist semantics (see e.g.
Lipka 1980, Lutzeier 1993). To the extent that there is a correspondence
between the shell-noun uses of certain nouns and the linguistic system, it
does not hold between uses and the meanings of lexemes, but rather between
uses and "lexical units" in the sense of this term advocated by Cruse (1986:
76-77) and Lipka (1992: 130-132).
While I have tried to show that intensional unspecificity is not a neces-
sary characteristic of shell nouns (see Section 5.2), it is a fact that the ma-
jority of shell nouns do not command a large number of semantic features.
Highly specific nouns such as irony or gripe are rarer in terms of types and
especially of tokens than more unspecific ones. The property of intensional
unspecificity is both a curse and a blessing for the classification and de-
scription of shell-noun uses. On the downside, the semantic analyis of shell
nouns turns out to be quite challenging because one has to deal with words
with highly general and therefore superficially similar meanings. Take for
example the nouns aim and attempt. Aim can be used to shell ideas or more
precisely mental states that are directed towards future accomplishments.
So a rough description would involve the features [MENTAL], [FUTURE] and
[ACTION] or [ACCOMPLISHMENT]. The same components also play a role in
uses of the noun attempt, but they must be arranged in a different order in
the description of this noun. Attempt is used to refer to activities that are
carried out by people with certain aims in mind, especially to see whether
something can be achieved or not. So here the order of the features must be
[ACTON], [MENTAL], [FUTURE] and [ACCOMPLISHMENT], Because of the
shared features it is crucial that the order of features is determined carefully
to reflect the relations between the semantic components.
Such relational semantic aspects are not just important for the more or
less stable meanings of shell nouns themselves but even more so for the
links between shell nouns and their contents. As was shown in Section 3.1,
these links are an essential characteristic of the use of shell nouns. Although
relational semantic features have been proposed in lexical semantics (Lipka
1979, 1992: 108, 112) to account for the meanings of converses such as
father and son or teacher and pupil (e.g. [-PARENT] and [<PARENT]),
these features are too specific to be useful for my purpose. To account for
the cognitive models which provide the basis of these links (see Section
3.1.2), I will introduce the notion offrame as a second descriptive device.
The intensional unspecificity of shell nouns is advantageous, on the other
hand, because a fairly limited number of semantic features suffice to at least
classify all the nouns that have been found to occur as shell nouns. On the
basis of the available data, five features are used as high-level classifiers for
shell nouns. This means they are seen as the first and most important
meaning components shared by larger sets of shell-noun uses. These five
features are listed in Table 6.1 with a brief explanation of the type of expe-
rience they are intended to describe, which is partly based on the subcatego-
rization of events and abstract relations put forward in Section 5.1.1 (cf.
Figure 5.1).
Table 6.1 Semantic features functioning as classifiers
Semantic feature Type of experience being described
[FACTUAL] facts, states of affairs

[MENTAL] ideas, cognitive states and processes
[LINGUISTIC] utterances, linguistic acts and products thereof
[MODAL] possibilities, abilities, permission, obligations, etc.
[EVENTIVE] activities, processes, states
It is apparent that these classifying features are neither primitive nor precise
in the way usually aimed at by feature semanticists 42 Nor do they reflect
ideological or philosophical assumptions concerning the nature of meaning.
Features and frames 89
Instead they are conceived as elements of a metalanguage which provide

fairly rough but economical characterizations of fundamental types of expe-
riences. For example, the feature [FACTUAL] is used to describe the domi-
nant semantic element of nouns that are used to shell abstract relations be-
tween entities which can be said "to be the case". Whether the expressions
are true reflections of these facts and whether they are being presupposed as
being true by the speaker are aspects that are not relevant for the conception
of the feature.43
In a similarly down-to-earth way, the feature [MENTAL] is understood as
describing ideas, i.e. cognitive states and processes attributed to a person's
mind, and the feature [LINGUISTIC] as describing people's utterances. This
includes the locutionary and the illocutionary perspectives, and even the
concrete and tangible results of utterances as in Declaration of Independ-
ence. What is meant by the feature [MODAL] is less easy to describe. Most
of the notions intended to be included in it have something to do with possi-
bilities either in the epistemic or the concrete domain. As in grammar, the
term is used to cover a fairly wide range of types of experiences, including
such notions as ability, volition, permission, obligation, as well as possibil-
ity and necessity (cf. e.g. Quirk et al. 1985, Leech 1987, Palmer 1990). The
details of this field will become clearer in Chapter 10. The feature
[EVENTTVE] is used to refer to the three types of events distinguished in
Section 5.1.1, namely activities (i.e. events involving a change that are
caused by human agents), processes (i.e. non-agentive dynamic events), and
states (i.e. non-dynamic or stative events). One additional feature, which is
not used as a classifier but plays a role in many nouns is the feature
[EVALUATIVE]. It is used to describe evaluative and attitudinal meaning
components included in shell nouns.
As emphasized above, I use semantic features simply as economical de-
scriptive devices. Their use is not meant to reflect any theoretical commit-
ment, even though I apply the convention of marking them by insertion in
brackets and spelling in capital letters. I do not conceive of these features in
an Aristotelian way as arising from binary oppositions but rather as general
components of meaning, and this is reflected in the fact that no "+"or " - "
signs are used. As far as their theoretical status is concerned, these features
are very similar to the notion of attributes as used in prototype theory (see
e.g. Ungerer and Schmid 1996: 22-31). I only refrain from referring to the
features used here as attributes because they tend to be so abstract that it is
difficult to think of them as being associated with concepts in the same way
as for example attributes like 'has wings' or 'is small' are prototypically
associated with the cognitive category BIRD. Given this theoretical back-
ground, one should not be too critical about features like [NEUTRAL], [UN-
KNOWN] or [NOT SATISFIED WITH X] which will be introduced more or less
on an ad-hoc basis in the following chapters.
The second descriptive concept to be introduced now, the notion of
frame, has a different status. As argued in Section 3.1.2, I assume that
structured cognitive models of some kind are involved in the processing of
language in general, and of shell-content complexes in particular. I have
selected the frame notion from the wealth of available concepts and terms
for various types of cognitive models (see Section 3.1.2) because it has been
used with remarkable success in three domains which are of central rele-
vance for the analysis of shell-content complexes, namely in lexical seman-
tics (see e.g. Fillmore 1985, Fillmore and Atkins 1992), in the semantics of
grammar (Slobin 1996, Talmy 1996) and in text analysis (Minsky 1975,
Winograd 1975, Petfi 1976, Rosenberg 1979).44
The conception of frames used here is borrowed from Talmy, who re-
gards an (event-)frame "as a set of conceptual elements and relations that
[...] are evoked together or co-evoke each other" (Talmy 1996: 238). Not
only is my use of the frame notion directly based on this general cognitive
conception but my analytical and descriptive method is also derived from
the way Talmy works with it. His approach is to analyse the shared con-
ceptual structures underlying similar types of events first, and then to find
out how the components of these so-called event-frames are rendered in the
linguistic output. When speakers refer to certain components of frames by
mentioning them explicitly, Talmy interprets this in terms of attention-
allocation. For example, a speaker can use adverbials to highlight sections
of the path of an object in a motion-event. One of Talmy's examples is
given in (6.1):
(6.1) The crate that was in the aircraft's cargo bay fell out of the plane into
the ocean. (Talmy 1996: 245)
Here the speaker opens two "windows of attention", one on the initial sec-
tion of the path (out of the plane) and its final section (into the ocean), but
gaps the medial section (which could have been windowed, for example, by
expressions like through the air). Although this is more implied than ex-
plicitly discussed by Talmy, his idea of how such utterances are processed
consists of two stages: the activation of a frame and the opening of windows
of attention within the frame.45
Features andframes 91
As this example indicates, Talmy is more concerned with frames repre-

senting concrete events in the real world than with abstract relations. On the
basis of cross-linguistic evidence he claims that frames such as the motion-
event frame or the causal-event frame are cognitively fundamental and pre-
sumably universal. For the abstract relations that I am chiefly concerned
with here, it would not be reasonable to mount such strong claims. Never-
theless, the notion of frames and Talmy's cognitive approach can be trans-
ferred to the abstract domain. For example, the causal nouns reason and
result can both be seen as activating a causal frame. Which of the nouns is
used, and in which lexico-grammatical pattern, depends on where the
speaker wants to direct the hearer's attention. Reason is used to highlight
the CAUSE component of the frame and result the EFFECT component.46
As a general rule, the setup of the frames I will be dealing with consists
of a small number of components and the relation(s) holding between them.
The frame representing the evidential relation for example, which underlies
the use of such nouns as evidence and proof, is seen as consisting of two
components called SIGN and BELIEF and the relation that the SIGN causes a
certain mental state or BELIEF in a person involved. It is thus a more elabo-
rate variant of the frame representing the causal relation, which consists of
a CAUSE which is experienced as being prior to, contributing to and/or re-
sponsible for the occurrence or existence of the EFFECT. These examples
show that the setups of frames are not based on strict logical or even formal
considerations concerning causality or other relations but formulated in an
experiential spirit on the basis of the 'naive' cognitive models attributed to
language users (cf. LakofF 1987, Ungerer and Schmid 1996: XI).
The following chapters 7 to 12 will be devoted to a detailed discussion of
classes of shell-noun uses, of their sub-categories, and of some usage types
of important single shell nouns. The subcategories will be called groups and
families. While classes and groups of uses are characterized by names that
are derived from their dominant semantic features, families will be named
by prime examples to facilitate their recognition. (To distinguish family
names from lexical items, family names are capitalized and inserted between
single inverted commas).
To give a general idea of the meaning shared by the nouns subsumed as
a family, semantic features and frame characteristics will be indicated. The
nouns that make up a family of uses will be given at the top of each section,
ordered by their frequency in the four lexico-grammatical patterns on which
this study focuses. The frequencies in these patterns of all nouns as well as
their frequencies in the whole corpus can be checked in the Appendix.
Chapter 7
Factual uses
7.1 Introduction
Speakers use factual shell nouns to create conceptual shells for 'abstract'
states of affairs and facts. Any experience, with the exception of first-order
entities, can be construed as a fact by means of an appropriate shell-content
construction. In this section, I will look in detail at the factual shell-noun
uses in my material. Six groups of factual shell nouns are distinguished:
neutral, causal, evidential, comparative, partitive and attitudinal. To make it
easier to keep track of the groups and families, a summary of the hierarchi-
cal structure of the class of factual shell-noun uses is given in Figure 7.1.
Groups Families Examples Section
Neutral uses 'Thing' thing, fact, phenomenon 7.2

Causal uses 'Reason' reason, cause, ground 7.3
'Result' result, effect, upshot
'Link' link, connection
'Reward' reward, compensation
Evidential uses 'Evidence' evidence, proof, sign 7.4

Comparative uses 'Difference' difference, similarity 7.5
Partitive uses 'Aspect' aspect, feature, factor 7.6
'Part' part, basis, foundation
'Example' example, exception
Attitudinal uses 'Problem' problem, snag, trouble 7.7
'Advantage' advantage, benefit
'Irony' irony, importance, novelty
'Tragedy' tragedy, shame
'Miracle' miracle, sensation
Figure 7.1 Survey of the class of factual shell-noun uses

Introduction 93
Many, but not all, families of factual uses can best be described with the
help of specifically designed frames, e.g. causal, evidential and comparative
uses. For the other families, however, it will be useful to have a very general
factual frame at our disposal. This frame consists of two simple compo-
nents, which will be called STATE OF AFFAIRS and CHARACTERIZATION. The
former component is generally mapped onto the shell content, and the latter
is incorporated in the shell-noun phrase.
7.2 Neutral uses
'Thing'
Shared semantic features: [FACTUAL]

Frame: General factual
Nouns: Fact, point, case, thing, business, phenome-
non
The neutral shell nouns which make up the 'Thing' family are among the
intensionally most unspecific nouns in the English language. Their only
common semantic denominator is the feature [FACTUAL], This unspecificity
is most evident for the noun thing itself which at first glance seems to con-
vey hardly any information at all in uses as a shell noun. While this prop-
erty puts all the more emphasis on the concept-forming function of this
noun, it casts doubt on the characterizing function. The question is whether
a noun with almost no meaning can be seen as contributing to the descrip-
tion of a piece of experience.
On closer inspection it turns out that completely neutral shell-noun uses
of the noun thing are in fact astonishingly rare. The best candidates for such
uses can be found in the pattern th- with the referring items this and that,
one and (an)other, different and same, and first, next and last. However,
even in many of these uses, especially when thing is accompanied by this,
that and last, an undertone of negative evaluation can be detected, as in the
examples collected in (7.1):
(7.1) (a) If this thing is a fraud it is a veiy expensive ... (TODAY)

(b) well, it's going back to that thing of paying homage to people
Who ... (MAGS)
(c) The last thing it wants is more hospitals ... (ECON)
94 Factual uses
Similar connotations are evoked by the large number of uses of the noun
thing in the collocations that sort/kind of thing, which should also be men-
tioned in this context.
The even more notorious collocation the thing is is also used by speakers
to express additional covert meanings on top of the 'pure' concept-forming
effect. According to LDOCE3 (s.v. thing 27), this collocation is "used when
explaining a problem or the reason for something". This would take the
noun into the groups of attitudinal and of causal shell-noun uses. And in-
deed, this additional classification - recall that multiple classification has
been declared legitimate in Section 6.2 - is warranted by the corpus data on
the collocation the thing is. Four examples of this type have already been
given in (4.10). Since these involved either a repetition of the copula is or an
omission of the conjunction that, two further examples are provided in (7.2)
to serve as reference points for the discussion here:
(7.2) (a) you're 61 now and it's time you settled down. "The thing is that he
needs a lot of loving, (SPOKEN)
(b) I'm not a doppelganger for Robert Redford. "The thing is he just
doesn't care, (SPOKEN)
From a communicative point of view, these uses of the thing is (that) ex-
hibit a very interesting conglomerate of meanings and insinuations. As sug-
gested in the LDOCE3 entry, causal aspects play a role, but perhaps in a
more intricate way than proposed by the lexicographers. My favoured strat-
egy for paraphrasing utterances like those in (4.10) and (7.2) would be to
start out from the shell content, for example from he needs a lot of loving in
(7.2a). By shelling ideas like this one with the collocation the thing is (that),
speakers achieve three things. Firstly, they turn ideas or possible facts into
facts. Secondly, they add evaluations to these facts. Although it is often
difficult to pinpoint their precise linguistic trigger, negative evaluations
seem to predominate and therefore the description as "a problem" in
LDOCE3 is probably justified. And third, speakers imply that both the fact
itself and its evaluation are of concern to both the speaker and the hearer(s)
and that this is the reason why the fact is mentioned in the first place. The
third aspect, then, can be interpreted as an emphatic linguistic gesture in-
tended to stress the relevance of what one says and to invest it with more
significance and importance. In many instances, the emphasis is partly mo-
tivated by a contrastive or counter-expectational relation to what has been
said before by one of the discourse participants. This does not seem to be
Neutral uses 95
the case in (7.2), however. The full paraphrase for the second sentence in
example (7.2a) would then be something like (7.2a'):
(7.2a1) It is a fact that he needs a lot of loving, and this concerns both you and
me, and therefore I am telling you about it.
This paraphrase shows that the CHARACTERIZATION component of the gen-

eral factual frame does have a linguistic counterpart. A slightly more ex-
plicit way of construing a similar conceptualization is the use of the noun
point in similar patterns. Here the emphasis on the relevance of the shelled
fact to speaker and hearer(s) is supported by the focusing meaning of the
noun point.
Tuggy (1996), who provides an interesting analysis of constructions of
the type the thing is is that in the framework of Cognitive Grammar, partly
argues along similar lines. According to him, expressions like the thing is
that and the problem is that are "focus formulas" (1996: 724-726). He even
goes so far as to claim that "the only function" (my emphasis) of the ex-
pression the thing (about it) is is "to focus attention on the following
clause. It means 'Hey!', and little or nothing else" (1996: 725).
Tuggy's main point of interest is not the 'normal' collocation the thing is
that but the marginal pattern the thing is is (that) which is used by many
speakers, 47 but also regarded as "grammatically erroneous" (1996: 714) by
many. He argues that the collocation is currently in the process of entering
the grammar of American English, i.e. of being "fully sanctioned" in Cog-
nitive Grammar terminology (Langacker 1987a: 66-71, Tuggy 1996: 718-
720). Tuggy traces this process back to a number of different sources, one
of which is the very fact that it is in use and thus "sanctions itself' (1996:
726). This argument of internal sanctioning has a slight smack of circular-
ity because it comes down to saying something like 'the pattern is gram-
matical because people use it', or even 'the pattern is there because it is
there'. While I accept that internal sanctioning can explain how originally
ungrammatical expressions turn grammatical, I do not think that it can ac-
count for why an expression is used in the first place. Tuggy's other claims,
especially that the double-/.? pattern is sanctioned by the single is that,
probably supported by hesitations, repetitions, mishearings and the like, and
by syntactic blends with other constructions, are perfectly convincing.
Building on his observations, I will put forward my own hypothesis as to
the source of the double-7' construction in Section 16.1.
96 Factual uses
While the unmodified collocation the thing is + that-c\axiSQ is by no

means rare, this pattern is more often used with premodifying adjectives as
in (7.3):
(7.3) (a) The worst thing is that I was told not to fight, (PAPERS)
(b) ... the most important thing is that he's black, (BBC)
These examples have not been selected at random; their choice is meant to
illustrate the two main types of CHARACTERIZATIONS evoked by the adjec-
tives that are used in the pattern, viz. evaluative and restrictive adjectives
respectively (Quirk et al. 1985: 430-431). Other frequent adjectives of the
first type are big, fanny, good, great and awful, bad, damn and terrible;
and of the second type crucial, main, only, real and whole. The semantic
nature of these adjectives supports my paraphrase of the unpremodified use
of the thing is. One could even argue that the expressions of evaluation and
relevance are only hidden in the 'naked' collocations, while they are marked
for attention in the premodified ones because they are mentioned explicitly.
The intensifying and focusing function of the nouns thing and point are
most conspicuous in uses of the two nouns in the pattern th-be-N. Espe-
cially in spoken English, collocations with restrictive adjectives which en-
hance this function, e.g. this is the main thing, that's the crucial point or
that's a very important point, are very frequent. In such uses the shell con-
tents that are referred to by the demonstrative pronouns in extended refer-
ence can represent facts as well as utterances and ideas, so that the nouns
thing and point can be seen to bridge the boundaries between factual, lin-
guistic and mental shell-noun uses.
To round off the discussion of shell-noun uses of the noun thing, it
should be mentioned with regard to example (7.3.a) that when thing is used
in the pattern -be-to, the subjective-volitional-future orientation meaning
which is typically evoked by this grammatical construction (see Section 3.2)
exerts a strong influence on the interpretation of the noun.
The upshot of the preceding few paragraphs is that the noun thing is
probably semantically less neutral in actual usage as a shell noun than its
decontextualized schematic meaning suggests. The same will be found to be
true of many other shell nouns. I have devoted so much space to this discus-
sion of the noun thing as a shell noun in order to give a glimpse of the kind
of description that is made possible by the data collected for this study.
However, it is clear that with 670 nouns to account for, only a small selec-
tion can be covered in such detail.
Neutral uses 97
The noun fact also deserves special attention because it occurs very fre-
quently as a shell nomi and can boast a high compiled reliance score. Like
thing, this noun is used by speakers in the focusing pattern N-e-cl, i.e. in
the collocation the fact is + that-c\ause, as an emphatic gesture. With the
noun fact, however, the emphasis is not so much on the relevance of the
shell content but on the claim that what is expressed in the that-c\sasQ is
true. Such uses are therefore emphatics for epistemic necessity and will be
looked at again in the section on epistemic uses (see Section 10.2). Tuggy
mentions in passing that the collocation the fact is that can be paraphrased
by something like "thing in disconformity with what is (only) be-
lieved/apparent" (1996: 722). This beautifully concise paraphrase, which
includes a reference to the contrastive meaning in the word disconformity, is
certainly applicable to example (7.4.) below, and it works well with other
instances, too.
(7.4) While this is a marvellous development, and will provide a much-

needed boost to research into the biology of Amazonia, it will not con-
tribute significantly to the fundamental problems. The fact is that nei-
ther of these institutions have significant social or political research
arms... (NEWSCI)
In contrast to thing, the noun fact is widely used in the pattern -that. With
an absolute frequency of over 26,000 occurrences in the 225 million-word
corpus the combination fact + -clause is the most frequent single collo-
cation out of those considered in my data. If one takes into account that fact
occurs in fixed expressions such as in fact, as a matter of fact, in actual
fact and in point of fact, the resulting reliance score of 38.13% becomes
even more significant.48 Collocations in the pattern th- with this or that,
which account for a large proportion of the uses of many other shell nouns,
are fairly rare with fact. The reliance score for this pattern is only 1.4%.
The collocation fact + that-clause is used by speakers to construe states
of affairs not just as being the case, but as being real and 'true'. In this
respect, fact is more specific than thing. One way of ascertaining the factual
nature of the noun fact is to look at the postnominal tfiaZ-clauses. Unlike
linguistic and mental shell nouns, the noun fact is hardly ever accompanied
by /aa-clauses with hypothetical verb forms. Past and present tenses
clearly predominate in these clauses, whereas subjunctive, future and modal
forms are extremely rare. The reason for this distinct tendency is that one
cannot make predictions about future facts and therefore speakers tend to
98 Factual uses
use other, typically mental or modal shell nouns when they want to shell
these types of experiences.
In addition to being factual in the general sense used so far here, the
noun fact may have a claim of being factive in the Kiparskys' (1971) more
restricted sense of this term when it is used in the collocation the fact +
that-clause. According to the Kiparskys, factivity obtains when the infor-
mation expressed in the complement of a verb is presupposed as being true
by the speaker. The complements of factive predicates continue to represent
true propositions even when the predicates themselves are negated. For
example, both the utterances I regret that I took the money and I don't
regret that I took the money presuppose that the speaker took the money. If
this were not the case, the speaker would either be misusing the verb regret
or would deliberately be trying to lie.
The factivity of nouns is much more difficult to assess than that of verbs
because most tests of factivity that have been proposed require predicative
expressions as their input.49 Francis (1986: 22-26) argues that some tests
can be applied to nouns and proposes a distinction between factive, half-
factive, non-factive and counter-factive nouns. According to her classifica-
tion, all the nouns discussed here as factual shell nouns also qualify as fac-
tive nouns. The trouble with the tests, however, is that some require such an
amount of paraphrasing that the whole testing apparatus ultimately breaks
down. In addition, it is doubtful whether the application of a strict notion of
factivity as proposed by the Kiparskys promises any linguistically or cogni-
tively relevant insights on the use of shell nouns. The following discussion
will therefore be restricted to some of the presuppositions and commitments
that can be involved in the use of the noun fact. Two useful examples are
given in (7.5):
(7.5) (a) The siskin is an exceptionally pretty, small green bird, a sort of
pocket version of the greenfinch. Twenty years ago it was uncommon
in England. The fact that it is now very much a regular winter visi-
tor is due almost entirely to the efforts of amateur bird feeders.
(EPHEM)
(b) We never expected John that er they would just change it like that in
the light of the fact that they're going to change local government
across the country anyway, (SPOKEN)
The speaker of example (7.5a) seems to presuppose as being true that the
siskin is "now very much a regular winter visitor" in England. Instead of
Neutral uses 99
introducing this basically new piece of information as the rheme of the

clause (see Section 3.3), he or she seems to take this fact for granted and
uses it as the starting point for the second sentence in the quoted passage.
The effect is that the addressee will not question the truth of what is strictly
speaking no more than a claim on the speaker's part, since it is construed
factively and presupposed as being true (see Section 16.1).
The utterance in (7.5b) has been chosen because it is one of the few ex-
amples where, as opposed to what was claimed above, hypothetical aspects
play a role in the subordinate that-ciause. This is particularly interesting in
the present context because factivity has to do with what we know rather
than with what we believe or expect to happen (Vendler 1980: 277). In this
example, however, the noun fact is used to shell the intention of another
person. So strictly speaking factivity cannot be involved because the
speaker is only making a prediction as to possible future events, and there-
fore modal shell nouns such as possibility, likelihood or prospect or per-
haps linguistic shell nouns such as declaration, assertion or promise would
be better suited.50
Two complementary interpretations can be offered from a pragmatic and
cognitive perspective for the finding that speakers combine the noun fact
with experiences that are not facts but other abstract entities.
First, the speakers can be seen as being so certain about their ideas and
claims, or as intending the hearers to believe that they are so certain, that
they feel that the use of the noun fact is justified. They accept, or even in-
tend, the change of the conceptual status of the shell content within the do-
main of abstract relations from idea or possible fact to fact (see Sections
5.1.1 and 5.1.2). That this shift is the product of a conscious, strategic deci-
sion is highly unlikely in view of the fact that the production of such utter-
ances is carried out within very few seconds. Thus it is perhaps not so much
an on-line decision taken during speech production, but more a habit which
is acquired on the grounds that more significance is attributed to facts in our
culture than to ideas. This leads me to the alternative and complementary
interpretation.
Precisely because facts are rated more highly than ideas, the noun fact is
used much more often than other shell nouns, at least in the pattern N-cl,
the fact that.... It is well known that linguistic elements with high frequen-
cies of occurrence are more prone to be affected by processes like semantic
bleaching or even grammaticalization than rare words and constructions. As
a result of such processes, the construction the fact that seems to have lost a
considerable part of its 'orginal' meaning and has come to be used as the
100 Factual uses
general-purpose shelling device, similarly in meaning to thing. At least in

these -that uses, it does not matter whether the shell content is strictly
speaking a fact or not; it does not even seem to matter whether the whole
shell-content complex represents a fact. What counts is simply that the con-
struction the fact that is a very handy means of shelling events and abstract
relations altogether.
These considerations highlight the fact that it is always the speakers us-
ing shell nouns who determine the conceptual and epistemic status of what
they are talking about by introducing specific CHARACTERIZATIONS in the
form of shell nouns. The objective relation between a piece of experience
(which is construed as a temporary concept by a shell noun) and the real-
world STATE OF AFFAIRS to which it corresponds is irrelevant. As a conse-
quence, those traditional tests of factivity which are based on a comparison
between linguistic expression and real-world states of affairs become
worthless. (The test suggested by Vendler (1980: 280) to overcome this
dependency on knowledge about the real world, the w/i-criterion (see also
Francis 1986: 22-23), cannot be transferred to nouns without considerable
paraphrasing efforts).
In contrast to thing, the noun fact is neutral with respect to the attitudi-
nal dimension. No emotive connotations are hidden in expressions of the
type the fact + that-dmsz. Only the -be-that collocation the fact is that
involves the elements of emphasis and counter-expectation mentioned
above.
Perhaps rather astonishingly, the nouns question and doubt can also act
in constructions which are very similar in function to the factual uses dis-
cussed here. For this to be possible they must either be combined with the
determiner o (as in 7.6) or occur in phrases like it is beyond question that
(as in 7.7):
(7.6) There is also no question that there is widespread, perhaps out-of-

control, poaching, (BOOKS)
(7.7) I rejected this firmly: it is beyond question that they must behave profes-
sionally and use reasonable care, (PAPERS)
Undoubtedly such constructions can be paraphrased by expressions with

fact, for example (7.6) by something like it is a fact that there is also wide-
spread... poaching. However, one must not overlook that they are not par-
ticularly good examples of shell-content complexes since the that-c\auses do
not function as postmodifiers or complements but as postponed subjects in
Neutral uses 101
existential there and anticipatory //-clauses (see Section 3.1.1). (Other shell-
noun uses of the nouns question and doubt will be described in Sections
8.3.2 and 9.3.2.)
The other three nouns in the 'Thing' family, phenomenon, case and
business, are much less frequent than thing, point and fact. Although they
can be used in the patterns -that and -be-that, they are found much more
frequently in the pattern th-N, and are often used in the pattern th-be- as
well. They are semantically neutral shell nouns but each of them is related
to one of the classes to be discussed below. Phenomenon evokes associa-
tions with the domain of mental entities, case with the contingency of events
(see Section 12.3), and business involves an attitudinal component (see
Section 7.7). Examples of the three nouns are given in (7.8).
(7.8) (a) "[...]. Later, mummy herself becomes the prototype for her aspira-
tions and a little envy begins to creep in." It is this supposed envy of
mummy that is the most dangerous aspect of this new phenomenon.
(TODAY; location of shell-content doubtful)
(b) Any visitor who intends to keep his vehicle in the country for more
than twelve months [...1 will have to pay duty. In this case, even if he
leaves Zambia subsequently, he will be unable to obtain refund of the
duty paid, (MAGS)
(c) ... and shook his hand like a lodge brother. Then he shook with Cas-
sidy and finally with Doug. "Congratulations, Detective. I'm glad
this business has been cleared up ..." (BOOKS)
Two further nouns related to the 'Thing' family are matter and state of
affairs. However, although both nouns are frequently found to occur in the
patterns th- and th-be-N, they did not occur in the patterns N-cl or N-e-
cl in the corpus and are therefore not included as proper shell nouns here.
7.3 Causal uses
One of the relations in which the collocation fact + -clause is often in-
serted is the causal relation. In feet, speakers have a whole array of semi-
fixed phrases containing the fact that at their disposal, which are used to
express causal meanings (see Quirk et al. 1985: 1105-1106):
102 Factual uses
because of \
by virtue of
in (the) light of
in view of the fact (that) ...
on account of
owing to
due to y
Similar grammaticalized connectives can also be found in German {auf-

grund der Tatsache dass, im Hinblick auf die Tatsache dass), Italian (per
il fatto che) and French (par le fait que-, see Conte 1996: 8, note 8). A less
awkward way of inserting complex pieces of experience in a causal relation
is to use causal shell nouns.
Instead of addressing the logical and philosophical technicalities in-
volved in the notion of causality, it will suffice for the present purpose to
understand causality as a relation between two events or facts, in which one
entity, the CAUSE, is conceived by a human conceptualizer as being prior to,
contributing to and/or being responsible for the occurrence or existence of
another, called EFFECT. The deployment of attention in this two-sided frame
gives rise to the separation of three of the four families within the group of
causal shell nouiis: by using the nouns in the 'Reason' family, speakers can
direct the focus on the CAUSE, while the nouns in the 'Result' family are
used to highlight the EFFECT element of the causal relation. Link and con-
nection, the two nouns making up the 'Link' family, can be used to focus
attention on both components of the relation. The fourth family, 'Reward',
consist of nouns which are used to conflate the EFFECT with an attitudinal
component.
'Reason'
Shared semantic features: [FACTUAL], [CAUSAL]
Frame: Causal; attentional focus on CAUSE
Nouns: Reason, thing, cause, ground
The nouns in the 'Reason' family are used to call up the causal frame out-
lined above and to direct the hearer's attention to the CAUSE component
within its cognitive setup. Nevertheless, a surprisingly large proportion of
the typical collocations in which reason occurs include explicit references to
Causal uses 103
the EFFECT components, too. To flesh out this claim, it will be helpful to
look at how the two components of the causal frame are mapped onto lin-
guistic patterns containing the noun reason. This is summarized in Figure
7.2.
Components of the causal frame:
CAUSE reason EFFECT CAUSE
Linguistic realisation using the shell noun reason:
(a) (the) reason for-PP be + f/wf-clause

why-clause be + because-clause
be + other complement
(b) (the/one/another) expressed in the be + f/iai-clause

reason context be + other complement
(c) (no/good/every/any) infinitive clause expressed in the context

reason
(d) (it was) for reason thai-clause

this
(e) (for) this reasonQ main clause
Figure 7.2 Mapping of the causal frame onto collocations with reason
Since the CAUSE component can be activated by elements occurring either

before or after the noun reason, it is given twice in this minimalist descrip-
tion of the cognitive setup of the causal frame. Corresponding examples are
given in (7.9), where the CAUSE elements are marked by {C} and the
EFFECT elements by {E}.
(7.9) (a) ... use a number of technical terms. The reason for this IE} is that in
this chapter we {C}... (BOOKS)
(b) ... and the army collapsed {E}. The reason is that Kuwait is bitterly
divided IC> (PAPERS)
(c) But Stehli has a good reason to write her story IE}. [Dotsie dies, ...]
{ C } (BOOKS)
(d) ... and it was for this {C} reason that I originally chose to tackle {El
... (MAGS)
104 Factual uses
(e) For this {C} reason, [they are coming under increasing pressure]
{E}... (PAPERS)
As can be seen, pattern (a) provides slots for the expression of both EFFECT
and CAUSE. The EFFECT component can be realized as a prepositional
phrase introduced by for (as in 7.9a) or as a clause introduced by why. A
look at a random sample of 1,000 uses of reason suggests that for-
prepositional phrases are used approximately twice as often as wAy-clauses
for this purpose. In my data the collocation reason + wAy-clause is the most
frequent pattern in which this noun was found (4,614 occurrences, reliance
13.2%, see Appendix). The CAUSE component is expressed in one of three
types of complements following the copula be : if the EFFECT is realized as a
yr-prepositional phrase, the CAUSE is usually expressed in a that-clause; if
the EFFECT is realized as a w^y-clause, one also comes across because-
clauses in addition to -clauses; and finally other types of complements
such as noun phrases are used as well. In all three variants the clause-final
complements identifying the CAUSE are clearly prominent with respect to
prosody, grammar and information distribution. Therefore, it seems justified
to say that the noun reason highlights the CAUSE element of the causal rela-
tion. The claim is further supported by the frequent occurrence of pattern
(b), where the EFFECT component is omitted and needs to be inferred from
the context. (Note that these are the only uses of reason which are directly
identifiable by the corpus query "N+is+that".)
With a reliance score of just over 10%, pattern (c), reason + infinitive
clause, is the second most frequent collocation of reason in my data. The
EFFECT is usually a dynamic entity such as an activity or a process ex-
pressed by an infinitive clause following reason in this pattern, while the
CAUSE is explained somewhere in the following context. Reason tends to be
determined by no, every or any or premodified by an evaluative adjective,
mostly good in this pattern.
Perhaps rather astonishingly, -clauses directly following reason also
express the EFFECT component of the causal relation (see pattern (d) in
Figure 7.2). This emerges from example (7.9d) which, like most examples
of its kind, involves an extrapositional construction. As a consequence, the
that-c\ausQ must be analysed as a clause-constituent in its own right which
represents the notional subject, rather than as a postmodifier or complement
to the noun reason (see Section 3.1.1). However, this syntactic analysis
does not affect the observation that the clause provides information con-
cerning the EFFECT component. Both in this pattern and in pattern (e), the
Causal uses 105
CAUSE component is being activated indirectly by means of referring items

in the determiner group of the noun phrase containing reason. The differ-
ence between (d) and (e) is that in (e), reason is part of a sentence-initial
conjunctive adverbial and the EFFECT component is expressed in the main
clause. This is illustrated in example (7.9e).
The corpus data suggest that compared with reason, its near-synonyms
cause and ground seem to be of marginal importance both in terms of fre-
quency in shell-noun typical patterns and in terms of collocational versatil-
ity. With the lexeme ground, this is to a large extent an artefact of the
method of this study which does not take plural forms of nouns into consid-
eration (see Section 4.4). Unlike the plural form grounds, which is very
common, especially in phrases like on these grounds, on similar grounds or
on the same grounds, the singular form ground is fairly rare as a causal
shell noun. It was only found in the -that collocations on the ground that
... and on the sole ground that. (In occurrences of the collocations this
ground and that ground, the noun is used in its concrete meanings 'area'
and especially 'sports ground', which are both irrelevant here.)
The noun cause, on the other hand, predominantly occurs in the ex-
panded predicate have cause to, with locution and cognition verbs such as
complain, feel, ponder and distrust as most frequent types of verbs in the
subordinate clauses.
We have seen in the previous section that many uses of the noun thing in
the collocation the thing is (is) (that) also incorporate causal meanings.
This is particularly well illustrated in the following passage taken from
spoken speech. I have taken the liberty of replacing the codes in the tran-
script by line ends to facilitate its comprehension as a dialogue:
(7.10) ... you know what are you ever going to work again - and I think this is
this this is the sadness.
Yeah.
Why why get rid of people with experience?
I think the thing is that people have seen technology change dramati-
cally within the business ... (SPOKEN)
The causal meaning of the noun thing is so prominent in this example that
thing could actually be replaced by reason here without a substantial
change in meaning. That the example represents such a clear case of a
causal meaning is mainly due to the fact that the sentence containing the
thing is is preceded by a w/ry-question which, according to the principles of
106 Factual uses
conversation analysis, demands an answer indicating a reason as preferred

second in the adjacency pair. In terms of the frame analysis for causal shell
nouns, the wAry-question must be analysed as expressing the EFFECT compo-
nent, while the tfzcrt-clause represents the CAUSE component.
'Result'
Frame: Causal; attentional focus on EFFECT
Nouns: Result, effect, outcome, consequence, up-
shot, impact, sequel, payoff
The nouns making up the 'Result' family are used to highlight the EFFECT
component of causal relations. With regard to both frequency and colloca-
tional variability this family falls into two sets. The frequent and more ver-
satile nouns are result and effect, while upshot, consequence, sequel, out-
come, payoff and impact are much rarer and mainly encountered in patterns
of type -be-that.
On the face of it, the prime members of the family result and effect are
very much alike. There are no big differences between their overall frequen-
cies of occurrence in the corpus (34,974 for result and 32,701 for effect),
between their frequencies as collocates (2,472 vs. 2,076), and consequently
not between their reliance scores (6.41% vs. 6.18%). Nevertheless, it would
be wrong to assume that the two nouns are completely identical with regard
to how they are used. The main difference between them emerges from a
closer look at their scores for the collocations in which they occur. These
scores have been extracted from the Appendix and compiled in Figure 7.3:
Noun Pattern Pattern Pattern Pattern

-that -be-that th-N th-be-N
result 608 977 540 231
effect 809 117 810 54
Figure 7.3 Collocational data for result and effect compared51
Since the total scores do not differ very much, we can work with absolute
frequencies here. Figure 7.3 indicates a cross-wise pattern for the use of the
Causal uses 107
two nouns in the patterns -that and -be-that, and th- and th-be- re-
spectively. The difference in usage is statistically highly significant. Perhaps
most interestingly, the frequency scores given in Figure 7.3 suggest that
speakers tend to use the noun result rather than effect in the patterns N-e-
that and th-be-N, as in the two examples in (7.11) respectively:
(7.11) (a) When women are young, their careers and pension rights would be
damaged if they took time off for their children, so grandmothers
must help. The result is that you often know your grandchildren
better than your own children! (MAGS)
(b) Goals from Andrei Kanchelskis and Steve Bruce pushed United into
fourth place with their fourth successive win. Ferguson said: "That is
our best result in the League this year." (TODAY)
As in example (7.1 lb), the noun result is used in a number of fairly specific
contexts related to competitions in sports, to examinations, elections and
many other kinds of tests and investigations. In this sense the noun is closely
related to outcome, which is mainly used to shell the final states of such
extended events.
A further, more subtle, semantic difference between result and effect is
suggested by a look at concordances of the two nouns. My impression is
that the experiences shelled by the two nouns highlight different parts of the
final section of a causal chain that both nouns focus on. The noun result is
used to highlight the state that is reached after a concrete or abstract causal
process is completed. It tends to have a more stative, thing-like quality than
the noun effect. The latter noun seems to be used more frequently when not
just the final state is in focus, but parts of the causal process itself, too.
Much more than result, it denotes dynamic types of experiences involving
changes of states.
This is confirmed by a number of converging pieces of evidence. For a
start, the i/zaf-clauses following the frequent collocation with the result that
in the corpus tend to contain stative verbs, while most of the clauses fol-
lowing the collocation to the effect that contain dynamic verbs. Unlike re-
sult, the noun effect is frequently premodified by deverbal adjectives de-
scribing the precise nature of the change that is involved, e.g. balancing,
loosening, cleansing, drying, stimulating, calming, uplifting. And only
effect produces such compounds as snow-ball effect, greenhouse-effect,
after-effect, priming-effect or knock-on effect, which do not denote a
uniquely observed EFFECT qua second element of a causal relation but a
108 Factual uses
recurrent type of process or development. Finally, a comparison between the

collocations this result and this effect also indicates that result is used to
portray experiences as final states of a causal development, while effect
tends to be used when the type of change is also of interest. It must also be
mentioned, however, that stylistic differences seem to play a role as well.
Effect is more frequently encountered in academic and technical texts than
in texts from 'everyday' colloquial sources.
Stylistic factors are also responsible for the relatively infrequent use of
the noun consequence. Although the noun occurs occasionally in the collo-
cations the consequence + that-clause and the consequence + be + that-
clause, speakers clearly prefer result and effect except in fairly formal style.
The conjunct as a consequence, on the other hand, is widespread across
different styles.
The other five nouns in the 'Result' family, upshot, sequel, outcome,
payoff and impact, have only been found in the pattern N-be-that and spo-
radically in the patterns th- and th-be- (see the Appendix). The noun
upshot has the highest reliance score of the family. 105 of its 313 occur-
rences in the 225 million-word corpus have been identified as instances of
the pattern -be-that (the upshot + be + that-clause). In assessing the re-
sulting reliance score of about 33%, one should take into account that of-
prepositional phrases expressing the CAUSE component are frequently in-
serted as postmodifiers between the noun and the copula and that these in-
stances are all ignored by the corpus query. So the actual proportion of uses
in this structural pattern is probably considerably higher.
'Link'
Frame: Causal; attentional focus on both elements of

the frame
Nouns: Connection, link
On the basis of their 'pure' meanings the nouns connection and link would
not be thought of as means of shelling a causal relation but a general rela-
tion between two states of affairs which affect each other in an unspecified
way. And examples of the latter type can easily be found. The passage in
Causal uses 109
(7.12), for instance, is taken from a scientific text, whose author seems to
try to leave the precise nature of a particular relationship deliberately open:
(7.12) ... that these increases in transition temperature have something to do

with an increase in the strength of the cooperation between the super-
conducting electron pairs and the vibrations of atoms in the solid. The
BCS theory predicts this link, but other detailed studies [...] suggest that
these superconductors involve a mechanism for electron pairing ...
(NEWSCI)
In this example, the author's reluctance to identify the precise relation he or

she has in mind also manifests itself in the choice of the neutral expression
have something to do with.
In many other cases, however, the two nouns seem to be used to construe
a causal relation, and therefore they are mentioned here. Link and connec-
tion differ from the other families of causal nouns with respect to the allo-
cation of attention. As we have seen, nouns from the other two families are
used to highlight one of the components of the causal relation, even though
the other can be mentioned as well. Link and connection, on the other hand,
allow the speaker to highlight both elements of the frame. As a consequence,
paraphrased versions of clauses containing link and connection tend to be
comparatively long because both components have to be inserted. In the
case of (7.12), the paraphrase would require the repetition of the whole
underlined section which represents the shell-content. It would include refer-
ences to two processes that are projected as states of affairs {increases in
transition temperature and increase in the strength ...) and the reference to
the relation that exists between them (have something to do with).
'Reward'
Shared semantic features: [FACTUAL], [CAUSAL], [EVALUATIVE]
Frame: Causal; attentional focus on EFFECT

Nouns: Prize, punishment, reward, compensation
The nouns in the 'Reward' family are semantically much more specific than
the ones I have discussed so far and are used much less frequently. Speakers
use them to portray and evaluate facts (or events that are upgraded by these
110 Factual uses
uses as facts) as effects of activities or states of affairs. (7.13) is a typical

example:
(7.13) We thought that being British we would have been given a better than
even chance. Instead, our reward for our persistence is that we are
worse off by 500.000 pounds due to the activities of people who evi-
dently put their interests before those of the [...] taxpayer, (PAPERS)
Even though reward is used ironically here, the example still illustrates how
speakers can insert facts into CAUSE-EFFECT relationships and at the same
time evaluate them. The expression our rewardfor our persistence could be
paraphrased by the positive consequence of our persistence. As is typical
of all nouns in the family, the activity or other cognitive entity that functions
as CAUSE in this relations, in this case our persistence, can be encoded lin-
guistically as ayr-prepositional phrase.
Generally speaking, reward and prize are used to construe facts as posi-
tive effects of previous events or activities, punishment as negative effects,
and compensation as positive effects which are seen as redressing an un-
welcome previous state of affairs or event.
7.4 Evidential uses
'Evidence'
Shared semantic features: [FACTUAL], [CAUSAL], [MENTAL]
Frame: Evidential
Nouns: Evidence, sign, proof indication, reminder,
signal, implication, finding, demonstration,
symbol, clue, indicator, symptom, manifesta-
tion, corollary, intimation
The evidential relation which lies at the heart of all nouns in the group of
evidential shell nouns is slightly more complex than the causal one. In fact
the causal relation is a part of the frame representing the evidential relation.
The general cognitive setup of the evidential frame is as follows: an observ-
able state of affairs or fact is conceived of as a reason for a mental state,
which is, in contrast to the 'Reward' family, intellectual rather than emo-
tional in nature. I will refer to the observed fact as SIGN and to the corre-
Evidential uses 111
sponding resulting mental state as BELIEF. (Note that the frame for creditive
uses in the mental domain also includes the latter component; see Section
9.3.1). A typical example of how this frame is mapped onto linguistic ex-
pressions is given in (7.14):
(7.14) Like the Postcard posse and Lloyd Cole before him, Gary Clark is proof
that the Great British Songwriting Tradition is alive and particularly
well North Of The Border, (MAGS)
This example is particularly interesting because at first glance, the SIGN

component realized by Gary Clark does not seem to have the ontological
status of a fact, but of a first-order entity. However, a closer scrutiny sug-
gests that again an ontological (or conceptual) shift, this time from first-
order to third-order entity, is at work: it is not the person Gary Clark that
functions as SIGN here, but the fact the person Gary Clark exists and be-
haves in a certain way. As is highly typical of evidential shell nouns, the
BELIEF component is realized here as the iAaf-clause complementing the
head noun proof. This mapping of the /-clause differs markedly from all
the causal uses discussed in the last section because the that-clause uses
represent facts and not ideas.
Evidential shell nouns can also be used in the pattern th-be- when
speakers want to highlight the SIGN component:
(7.15) That would be the final proof that everybody, including us, has gone
nuts, (PAPERS)
However this pattern is much rarer than the pattern -that. In fact it is
normal for occurrences of evidential shell nouns in the pattern ~N-be-that to
be followed by postnominal tfior-clauses as well, as is also the case in ex-
ample (7.15). Quantitatively the use in -that patterns alone by far out-
weighs the use in -be-that patterns. The noun indication, for example,
was found 1,063 times in the collocation indication + that-clmse and only
71 times in the pattern this/that + be + (an) indication.
The four workhorses in the 'Evidence' family are evidence, indication,
proof and sign with frequencies in all four patterns over 1,000, flanked by
reminder, signal, implication and finding (frequencies 865, 617, 525 and
389 respectively). The last noun exhibits the highest reliance score of
66.38%, which is mainly due to its frequency in the pattern th- (this find-
ing), however. As the remaining nouns demonstration (frequency in all
patterns 168), symbol (130), clue (109), indicator (54), symptom (45),
112 Factual uses
manifestation (25) and corollary (20) are far less frequent, they will not be
discussed any further.
The four frequent nouns differ mainly with regard to the conceived tight-
ness of the causal relation between SIGN and BELIEF. The essence of this
relation boils down to the epistemic question of how certain one can be
about the match between an idea and the corresponding fact in (what we
think of as) the real world. So evidential shell nouns also have some claim
to be treated among modal uses of shell nouns in Chapter 10.
Speakers construe the tightest causal link between SIGN and BELIEF
when they use the noun proof. Its use suggests that there is no doubt that
the BELIEF evoked by a SIGN corresponds to the 'true' facts out there. One
step further down the cline of epistemic certainty is the noun evidence. Ex-
ample (7.16) is an interesting case because the speaker uses proof and evi-
dence in almost identical patterns.
(7.16) In the same year, food watchdogs, led by Action and Information on
Sugars AIS), tried to ban the Mars slogan on the basis that there was no
scientific proof that chewing chocolate was of any benefit. Jack Win-
kler, of AIS, said: "There is no scientific evidence whatsoever that Mars
makes any positive contribution to working, resting or playing."
(PAPERS)
Although the major motivation for the change of noun is presumably simply
to avoid an awkward repetition, this example provides me an opportunity to
study the difference between the two nouns. The two instances suggest that
speakers use the noun proof to portray facts as necessary reasons for
BELIEFS, and the noun evidence to portray facts as possible reasons, or
simply as reasons for BELIEFS. Although this is probably a rather intuitive
observation which is hard to substantiate, I have the feeling that the other
four other nouns reminder, indication, implication and sign are used when
an even looser link between SIGN and BELIEF is to be established.
The two instances of evidential shell nouns in (7.16) are not just inter-
esting as objects of comparison, but also because both nouns are accompa-
nied by the determiner no. This is often the case with evidential shell nouns,
especially with the frequent ones. In these negative contexts, the epistemic
qualities of the evidential shell nouns give rise to an interesting contrast
with regard to invited conversational implicatures: when there is no proof
for something, for example for the idea that eating Mars bars is good for us
(see 7.16), this means we simply cannot be sure whether it is true or not;
Evidential uses 113
however, when someone says that we have no evidence that eating Mars
bars is good for us, this is not just interpreted as a statement of evidential-
ity, but in addition, or even instead, one begins to wonder whether there is
any reason for the belief at all, and even why someone has claimed that it
might be beneficial in the first place. Similar implicatures suggest them-
selves for uses of indication and sign with negative determiners. That this
interpretation is actually intended by the speaker of the quotation in (7.16)
is supported by the two other markers of negation and non-assertivity, the
postmodifying negative intensifier whatsoever and the non-assertive deter-
miner any in the postnominal clause expressing the BELIEF. If the speaker
had used the noun proof instead of evidence, any would have sounded at
least odd, if not unacceptable in the postnominal clause, since proof seems
to entail a positive assertion (there is no scientific proof whatsoever that
Mars makes 7any positive contribution to working, resting or playing).
As in the other groups and families discussed so far, the less frequently
used nouns tend to exhibit more specific semantic characteristics than the
frequent ones. The noun reminder, for example, conveys the additional
feature that the BELIEF evoked by a SIGN is not entirely new. Implication
evokes a container-image. It suggests that the BELIEF is somehow included
in the SIGN. Instead of, or in addition to, their specific meanings, the use of
these nouns may be governed by stylistic restrictions. These undoubtedly
play a role with the noun finding. This noun is mainly used in scientific
style, or academic writing in general, as is borne out by the fact that 38.5%
of all uses of the collocation the finding that were found in the New Scien-
tist section of the corpus, which amounts to only 2.7% of the whole 225
million-word corpus. However, this restriction of usage is not purely stylis-
tic but also semantic in nature; the central semantic feature added by the
noun to the evidential relation, namely the fact that the BELIEF component is
the result of a more or less scientific study of the facts shelled as SIGN, ear-
marks this noun for this register, too.
7.5 Comparative uses
'Difference'
Shared semantic features: [FACTUAL], [COMPARATIVE]
Frame: Comparative
114 Factual uses
Nouns: Difference, alternative, distinction, contrast,

analogy, discrepancy, similarity, inconsis-
tency
The cognitive setup of comparative uses of shell nouns is derived from the
basic cognitive process of comparison. It consists of two entities and a basis
for the comparison, the so-called tertium comparationis, which is shared by
both entities. The tertium comparationis is also required for cases of per-
ceived difference, since one tends to compare things that are similar in some
ways rather than completely different entities like spaceships and car bump-
ers (unless one is trying to be funny).
Semantically the nouns in the 'Difference' family fall into two obvious
groups, expressions of similarity {analogy and similarity) and expressions
of difference (all the rest). Expressions of difference clearly outweigh ex-
pressions of similarity in my data, both in terms of types of shell nouns that
are used for that particular purpose and in terms of tokens that have been
found. Occurrences of the two nouns expressing similarity amount to no
more than 105 instances, while uses of the six nouns expressing difference
add up to 2,092 instances. More than half of these are occurrences of the
noun difference, which is clearly the most important noun of the family and
will therefore take pride of place in the following discussion. It should be
added that distinction is also used as a member of the 'Aspect' family.
From a collocational point of view the nouns in the 'Difference' family
are united by the fact that they occur most frequently in the pattern th-N, to
a certain extent in the patterns -be-that and th-be-N, but not in the pattern
N-cl. There are two exceptions: difference, which was found more often in
the N-e-tfiai-collocation the difference + be + -clause than in the pat-
tern th-N, and was also found in the collocation the difference + that-
clause; and analogy, which was found in the pattern -that as well. Like all
other factual shell nouns, the nouns in the 'Difference' family show a dis-
tinct tendency to collocate with ^/-clauses rather than infinitive clauses. In
this respect, the family has one exception, namely the noun alternative
which was most often found in the pattern -be-to. In these uses, the noun
alternative apparently does not shell a comparison between states of affairs
but between different courses of actions. This is the function of the nouns in
the 'Option' family, which belong to the class of event shell nouns (see Sec-
tion 11.3).
Used in their typical environment, i.e. in the patterns N-be-that and th-
N, the nouns in the 'Difference' family shell similarities and differences
Comparative uses 115
between states of affairs. An instance of the most frequent single colloca-

tion, difference + be + that-clause, is given in (7.17):
(7.17) Like any radio station with any size of audience, Radio One will always
be worth pursuing. Of course, the fundamental difference is that it's a
public service broadcaster, (MAGS)
Difference and the other members of the family are used in a highly elliptic
way in this type of pattern. What speakers explicitly express and therefore
highlight for attention with these uses is no more than a single aspect that
distinguishes one of the compared entities from the other. This aspect is
spelled out as shell content in the complementing that-cimse. The other
components of the comparative frame are not mentioned explicitly in the
clause or sentence which contains the comparative shell noun. As in virtu-
ally all examples of this type, a more explicit expression of the comparison
is possible by means of inserting a postmodifying prepositional phrase in-
troduced by between. For the utterance in example (7.17), such an ex-
panded version is given in (7.17'):
(7.17') Like any radio station with any size of audience, Radio One will always
be worth pursuing. Of course, the fundamental difference between Ra-
dio One and other radio stations is that it's a public service broadcaster.
This expansion reflects the whole cognitive setup underlying comparative

shell nouns: the two entities that are compared to each other, and the char-
acteristic of one (or both) of the two items which gives rise to the perceived
difference (or similarity). Speakers often omit the reference to the items that
are being compared and highlight the distinctive feature (or shared feature,
in the case of similarity nouns) instead. Even in the patterns th- and th-be-
N the anaphoric elements accompanying nouns in the 'Difference' family
point to this distinguishing (or uniting) aspect.
It is important to note here that speakers do not use 'Difference' shell
nouns without preparation, 'out of the blue' so to speak. It is far more nor-
mal that the two items and the common ground shared by them are already
established in the context before the shell noun is used. In example (7.17),
this is done in the sentence preceding the one which contains the noun dif-
ference, where not only Radio One and any radio station are mentioned,
but the way for the comparison itself is also cleared by the sentence-initial
comparative adverbial introduced by like. The information expressed by the
shell noun difference is therefore not completely new but accessible because
116 Factual uses
it was already active in preceding information units (see Sections 16.1 and
16.2). While the shell-content rendered as a complementing clause contains
the only new piece of information about the public service broadcaster, it is
not brand-new either, but related to the shell noun and therefore accessible
as well.
7.6 Partitive uses
'Aspect'
Shared semantic features: [FACTUAL], [ATTRIBUTIVE]
Frame: Part-whole
Nouns: Point, aspect, factor, feature, distinction,
characteristic, essence, attribute
The nouns in the 'Aspect' family are among those whose reliance scores
are most dramatically affected by the rigidity of the corpus queries. To a
certain extent this is also the case with 'Difference' nouns, and it is also true
of the 'Part' nouns to be discussed below. The reason for this lies in the
extremely relational nature of these nouns, or more precisely, in the linguis-
tic effects of their relational nature. Shell nouns with distinct relational
meanings are often accompanied by prepositional phrases which express
reference points in the particular relations that are involved. We have al-
ready seen this effect at work in the case of difference, which is often post-
modified by a prepositional phrase introduced by between. Since the corpus
query "NN+be+that" misses out examples with such inserted postmodifiers
(see Section 4.4), the 'real' frequency of occurrence is grossly underrepre-
sented by the reliance score; the number of instances retrieved by the sys-
tem is considerably lower than the real number of occurrences of the gram-
matical structure. One potential candidate for this family, the noun element,
is not even included in the data, as it is virtually always followed by an in-
prepositional phrase or /-prepositional phrase specifying the reference
point.
The members of the 'Aspect' family are united semantically by their
potential to evoke an attributive relation. This relation gives rise to the use
of prepositional phrases that are governed by the preposition of Essentially
the cognitive basis of this relation can perhaps best be understood as a
metaphorical extension of the PART-WHOLE relation. The underlying idea is
Partitive uses 117
that aspects, features or characteristics 'are parts of or 'belong to' abstract

entities in the same way as parts 'belong to' the objects which they make up
or to which they are attached.52
With regard to their use, the 'Aspect' family are a remarkably homoge-
neous set of nouns. They were found in the patterns -be-that, th- and th-
e-N. The only exception is essence, which occurred only in the patterns th-
be- and -be-that. For the other six nouns, the pattern f/j-N was found to
be most productive, and th-be- least productive.
The meanings of the nouns essence, distinction and factor on the one
hand differ slightly from those of aspect, characteristic, feature and attrib-
ute on the other. Essence brings in the semantic component that the high-
lighted feature is "the most important and basic quality" (LDOCE3, s.v.
essence) of the entity to which it belongs. (A closely related noun is nature,
which is not included in the data, however, due to the overwhelming inter-
ference of o/-prepositional phrases). The noun essence thus incorporates a
superlative meaning which has the logical effect that what it links up with
must be unique. This is the reason why it does not occur as a shell noun in
the /2-N-collocation this/that essence. We tend to refer anaphorically to
unique things by means of the definite article the and not by means of de-
monstratives, because there is no need to exploit the pointing potential of
demonstratives when only one candidate for reference is available.53
A similar meaning is conveyed by uses of distinction in the pattern N-
be-that; the main implication here is that a certain quality or element stands
out from comparable ones and thus has a distinctive function. Sometimes
also in the pattern -be-that, but mainly in other uses, distinction produces
uses that belong to the 'Difference' family.
Factor conveys the additional meaning that a quality or facet of a state
of affairs has the potential to act as a cause. Unlike the other nouns in the
family, it is not followed by the preposition of when the related state of af-
fairs is expressed in the same clause, but by the preposition in. An example
is given in (7.18):
(7.18) ... sheer laziness. Mr Elliott acknowledges that it is "easier to get some-
one to adapt a book than to commission a new piece. Even if you have to
change it you still have a baseline to operate from". Fear of taking fi-
nancial risks is another major factor, (PAPERS)
Obviously, this passage is about possible reasons why authors may be re-
luctant to write books. At the point where the extract from the corpus be-
gins, the speaker is talking about laziness as one possible reason and goes
118 Factual uses
on to explain some observable effects. Then another aspect, fear of taking

financial risks, is introduced and shelled as another possible reason with the
noun factor. What is so useful about the word factor, as opposed to typical
causal shell nouns such as reason, is that it lends itself much more readily
to uses where the EFFECT component of the causal relation is omitted and
therefore removed from the focus of attention. Had the speaker of example
(7.18) chosen the noun reason, it would have been more or less obligatory
to specify the EFFECT by means of a w/ry-clause or ^-clause. With the
noun factor this is not necessary, because the causal meaning is blended
with the 'aspect' meaning and therefore less prominent than with reason or
cause.
'Part'
Shared semantic features: [FACTUAL], [PARUITVE]
Frame: Part-whole
Nouns: Part, basis, foundation
The nouns part, basis and foundation are used to shell general PART-
WHOLE relations. Intuition and the linguistic literature (e.g. Langacker
1993) suggest as a linguistic consequence that they rely to a large extent on
noun phrases of the type head noun + postmodifying o/prepositional
phrase. To confirm this impression, the number of occurrences of the pat-
tern part of were counted in two random selections of 1,000 uses of the
noun. Curiously, the collocation scored exactly 621 out of 1,000 instances
in both samples. This high percentage is an indisputable reflection of the
relational nature of this noun.
Although the scores for frequency in all four patterns that were investi-
gated are fairly high for the nouns part and basis (2,844 and 1,218 respec-
tively), neither noun belongs to the typicality groups of prime or even good
shell nouns for a number of reasons. Both nouns, as well as foundation,
owe their frequency scores mainly to occurrences in the patterns th-N and
th-be-N. The score for basis is influenced by the collocation on the basis
that, an expression whose meaning of contingency cannot be overlooked.
This expression exhibits a high degree of frozenness and idiomaticity. Apart
from basis in this fixed expression, the nouns do not occur in the pattern N-
cl at all, and only occasionally in the pattern -be-that. Furthermore, many
Partitive uses 119
uses of part are doubtful with respect to the criterion of temporary concept-
formation because, as in (7.19), the word is often used to refer to locations
that are already available in the text as bounded concepts.
(7.19) If the dispatch of some US troops to Macedonia indicates the eventual

establishment of a similar authority in that part of the world it should
be welcomed, (PAPERS)
A projecting construal as third-order entity is also not at work in this exam-

ple. In spite of these reservations there are instances where part, basis or
foundation are indeed used as shell nouns. Example (7.20) is one of them:
(7.20) Although the Bundesbank's president had hoped to reassure investors,

they responded by dumping government bonds. Part of the problem,
says Thomas Mayer, an economist at Goldman Sachs, an American in-
vestment bank, is that investors are confused about the central bank's
motives, (ECON)
The noun part shells the information expressed in the that-clause as a fact
here and highlights this fact as belonging to a larger problem. The parallels
to the 'Aspect' nouns are obvious, and so this usage certainly has a claim to
be regarded as a shell noun. Such a view may perhaps tell us more about
the structure of noun phrases of this type than when the part of collocation
is just regarded as nominal quantifier, analogous to a lot of or a great deal
of
'Example'
Shared semantic features: [FACTUAL], [PARTITIVE]
Frame: Class-member
Nouns: Example, exception
At the heart of the shell-noun uses of the nouns example and exception lies
the member-class relation. This relation can be seen as a second variant or
extension of the part-whole relation (in addition to the attributive relation
underlying uses of the 'Aspect' family). In this case, the WHOLE is a class
or set of entities and the PART is a member of the class.
120 Factual uses
Speakers use the nouns example and exception to achieve two opposite
effects. Example allows them to highlight a state of affairs as an instantia-
tion of a more general class of phenomena. (7.21) is a fairly typical case, in
that it illustrates the tendency to shell a particularly typical or illustrative
instantiation as an example:
(7.21) ... chocolate eggs are a fairly recent creation, dating from the last cen-
tury. Before that, real eggs, boiled and decorated, used to be given as
presents and this is a good example of a custom pre-dating the Christian
era. (BBC)
The underlined shell content gives information about a fact which then
functions as target of the referring item this. The noun example establishes
the class-member relation and shells the content as a member or instantia-
tion, while the postmodifying (^prepositional phrase contains the informa-
tion about the class. Example widely occurs in such collocations, which are
instances of the pattern th-be-N, as well as in the pattern th-N. It was also
found in N-6e-f/wtf-collocations and in reversed copular constructions of the
type xy is a good example of.
Expressions of the type the only exception was that... have a counter-
expectational meaning. They involve the tacit assumption that an entity is or
should be a member of a class, and the explicit highlighting of the fact that
it is not. In the light of this highly specific semantic structure, it is not sur-
prising that it is not a very frequent noun altogether and among the less
frequently used shell nouns as well. It was most often found in the pattern
th-be-N.
7.7 Attitudinal factual uses
The last group of factual shell-noun uses takes us back to the starting-point
of Section 7.2, the attitudinal use of the neutral noun thing. The hallmark of
attitudinal factual shell nouns is that speakers incorporate their attitude
towards certain states of affairs (or aspects thereof) into the nominal shells
they use.
The five families that are distinguished in this group are dealt with in the
order of their typicality for the class of shell nouns and their frequency. The
first two families consist of highly typical, mostly morphologically non-
derived shell nouns, some of which are extremely frequent. The 'Problem'
Attitudinal factual uses 121
family subsumes nouns which can be used to portray facts as impediments

to progress towards a certain aim. In direct opposition, the nouns in the
'Advantage' family characterize facts as making progress easier. The third
family, called 'Irony', includes some prime examples of shell nouns, but it
mainly consists of nouns that are derived from adjectives. The nouns in this
family share the potential to shell what is remarkable or interesting about a
fact in an emotionally neutral way. The two remaining families, on the other
hand, are made up of nouns that allow speakers to evaluate facts either
positively ('Miracle') or negatively ('Tragedy'). Many of these nouns are
not very good examples of shell nouns because they tend to occur in extra-
positional constructions.
'Problem'
Shared semantic features: [FACTUAL], [ATTITUDINAL], [IMPEDING]

Nouns: Point, thing, problem, trouble, difficulty,
snag, dilemma, drawback, catch, disadvan-
tage, burden, obstacle, hurdle, handicap,
downside, complication, predicament, crux
Looking at the list of the nouns supplying instances of the 'Problem' family,
one finds that some of them owe their inclusion in the class of shell nouns to
the feet that they are used metaphorically. The metaphors at work here pro-
vide a good way to explain the semantic aspects shared by the nouns in this
family.
The main metaphor is based on the image-schema of a path (Johnson
1987: 113-117), or put in a less abstract way, on the image of the move-
ment of an object or a person. This metaphor underlies the shell-noun uses
of the nouns snag, drawback, catch, obstacle, hurdle and burden. The
metaphorical mapping from a concrete onto an abstract cognitive model54
can best be explained with the example of a journey of a person along a
path towards a destination: the destination is mapped as an aim or task, the
movement along the path of the journey as progress made in the achieve-
ment of the aim, and obstacles as events or states of affairs which impede
and delay the direct and uninterrupted fulfillment of the aim or the solution
of the task. Progress in the concrete and the abstract model can be impeded
by things in your way {obstacles and hurdles), by things that hold you back
122 Factual uses
(snags, drawbacks and catches) and by heavy things that you have to carry
along (burdens). Although the morphologically derived nouns complication
and difficulty cannot really be regarded as metaphorical concepts, their use
for the expression of problems is certainly compatible with this metaphor.55
In sum, 'Problem' nouns are seen to be united by their potential to por-
tray facts as impediments to progress. I believe that this account is essen-
tially compatible with Tuggy's gloss of the noun problem in focus formula
constructions: 'Thing which is in disconformity with something (established
as) desired' (1996: 722; see also Section 7.2 above).
The lexico-grammatical use of 'Problem' nouns is marked by a distinct
preference for the patterns -be-that, th-H and th-be-N. None of the nouns
was found to occur in noteworthy numbers in the pattern N-to, and one only
very occasionally comes across uses of 'Problem' nouns in the pattern N-
that. These are clearly outweighed by occurrences of formally coinciding
postmodifying relative clauses introduced by that, which are unwanted hits
(see Section 4.3). In the Appendix no scores are therefore given for occur-
rences of 'Problem' nouns in this pattern.
In the pattern N-e-cl 'Problem' nouns are in evidence in the corpus with
all three types of complementing clauses, i.e. that-, to- and w/2-clauses. Two
variants of uses must be distinguished for combinations with that-clauses.
The first type has a strong affinity with the uses subsumed in the 'Aspect'
family (see Section 7.6). An example containing the noun snag, which is
earmarked for this type of usage - the reliance score for the pattern snag +
be + that-ciause is 31.89% - will help to show in what way:
(7.22) ... Mr Rocard may seek to be, he cannot escape the fact that his proposed
"big bang" realignment of the non-right forces in France is first and
foremost designed to produce a vehicle for his presidential bid in two
years time. The snag is that the polls show Mr Rocard as a loser against
Mr Chirac, and that two years may not be long enough for him to build a
sufficiently powerful base from which to succeed Francois Mitterrand.
(PAPERS)
The most conspicuous parallel between this example and the uses in the
'Aspect' family is that both involve the relation of particularisation. First a
state of affairs is described and then a specific aspect is extracted for spe-
cial attention. This relation could be highlighted linguistically by adding
postmodifying prepositional phrases such as about this state of affairs or
with this situation, plan, idea etc. to the attitudinal shell nouns (e.g. the
snag with his plans is that...). These insertions spell out the implicit parti-
tive relation which is shared by the nouns in the 'Aspect' family and many
uses of the 'Problem' family. The difference between 'Aspect' and 'Prob-
lem' nouns lies in their semantic structures. Uses of the 'Aspect' family can
involve attitudinal aspects, too, but when speakers want to convey these,
they have to express them by additional premodifying elements, mainly
adjectives. With nouns like snag, downside, difficulty, complication or
handicap, the attitudinal meaning is already incorporated in their stable
semantic structure. In a way, these nouns are equivalents to such expres-
sions as the negative aspect or the unpleasant characteristic. Interestingly,
expressions of this type are fairly uncommon. The reason is presumably
that their semantic and pragmatic impact is conflated in one word in the
nouns in the 'Problem' family, which thus provide speakers with more eco-
nomical means of expressing the same type of experience.56
The 'Aspect'-like use involving the relation of particularisation is espe-
cially typical of the nouns snag, drawback, difficulty, complication, down-
side, handicap, disadvantage and catch. The members of this group exhibit
the partitive meaning which lends itself to this type of usage more than the
frequent nouns problem and trouble do. While problem and trouble can be
used in the same way as well, the collocations the problem/trouble + be +
that are more frequently found to identify negative states of affairs in their
own right. From a textual point of view, these uses are less tightly inter-
woven with the preceding linguistic context, although they must of course
also be prepared. Examples (7.23) and (7.24) provide illustrations:
(7.23) Besides, Britain is not a nation as dry as the Sahara. It does not even
have a shortage of water. Large parts of the country have more than they
will ever need. The problem is that the water companies are as loath
since privatisation as they were before it to transfer the reservoirs of sur-
plus water to where they are needed, (TODAY)
(7.24) The concentration of pollen in the atmosphere may be of considerable
interest to the hay fever sufferer, so much so that the figure is usually
published in the newspapers along with the weather reports. The trou-
ble is that it is not a forecast and only records the average level reached
the day before, but it may explain why you felt so bad. (BOOKS)
The expressions the thing is (that) ... and the point is( that)..., which I have
already examined in Section 7.2, can be used in both variants that are typi-
cal of this pattern. On the whole, point seems to be used more often in the
partitive, 'Aspect'-like, usage-variant, while thing is more often used to
shell fully-saturated states of affairs. Which of the two interpretations is
124 Factual uses
more likely depends on the textual environment and on the semantic relation
of the shell content to the preceding linguistic context.
Only the five nouns problem, trouble, snag, difficulty and dilemma oc-
cur with infinitive clauses and/or w/j-clauses as complements. Since these
collocations do not give rise to factual but to eventive uses, they will be
discussed in Section 11.4 (see the 'Trouble' family).
The uses of 'Problem' nouns in the patterns th-N and th-be-N are textu-
ally and semantically less complex than those in the pattern -be-that. In
both patterns, the noun problem itself is by far the most frequent noun in
the family, and therefore we can focus our attention on the way this noun is
used.
Uses of problem in the pattern th-N are fairly homogeneous with regard
to their meanings, their embedding in the textual structure and the lexical
material surrounding them. Starting with the last point, the most frequent
determiner with anaphoric reference is the definite article, followed by the
form this. The form that, on the other hand, is fairly rare in this pattern.
The typical textual setup is that the shell-noun phrase the/this/that problem
refers anaphorically to a state of affairs which has been described in a
clause or longer stretch of text. Speakers normally prepare the link by in-
cluding linguistic elements with negative meanings or connotations in the
shell content. This semantic match between the shell content and the shell
noun problem greatly facilitates the co-activation of the content and its
negatively attitudinal shell (see Section 16.1 for more details). In (7.25), for
example, the negatively loaded words foul-smelling, pollution and drop can
be found:
(7.25) ... foul-smelling algae, the product of exceptionally high temperatures

and high levels of sea pollution, which led to a huge drop in bookings.
Fortunately this problem does not seem to have recurred this summer.
(PAPERS)
The general semantic and textual setups of uses of the noun problem in the
pattern th-be- largely coincide with the uses in the pattern th-N. States of
affairs are first described and then shelled with negative attitudes added.
The main difference between the pattern th-N and th-be- is that in the
former the characterization of a state of affairs as being problematic is pre-
supposed, while it is highlighted by the copula construction in the latter
pattern (see Section 15.1). Short utterances containing no more than the
collocations that is/was/'s the problem are particularly frequent in spoken
language (see example 7.26). A longer variant of this pattern, illustrated in

example (7.27), includes a defining relative clause after the noun problem
which contains additional information about it.
(7.26) And secondly erm depending on the risk I might have to take advice
from somebody almost as to what questions I should be asking. - Right.
Right. Mm. - That is the problem, (SPOKEN)
(7.27) ... the West must create the view that drugs are something that is to be
completely avoided, not something that is normal, acceptable, or even a
status symbol. That is the problem we face in the developing world.
(BBC)
When premodifying adjectives precede the shell noun problem, anaphoric

this is more often used as subject than that. The most frequent adjectives
are big, central, difficult, main, major, real and serious. Indeed, other ad-
jectives are fairly rare, only explicitly evaluative ones like frustrating can
occasionally be found.
To round off the description of shell noun uses of the noun problem, at-
tention should be drawn to the highly frequent collocations the problem of
+ noun phrase, the problem of + gerund, the problem of + w/z-clause and
the problem of + infinitive clause. These collocations, which are not sys-
tematically covered in this study (for reasons explained in Section 4.2), are
illustrated in (7.28):
(7.28) (a) ... if it is really serious about the problem of global warming, (BBC)
(b) He said the problem of supplying the Bulgarian people ... (PAPERS)
(c) ... there is the sticky problem of how much Japan will pay ...
(PAPERS)
As we will see in Sections 8.3.2 and 9.2, linguistic and mental nouns such
as question, idea, concept and notion are frequently used in the same pat-
tern, consisting of an noun phrase which is headed by the shell noun and
postmodified by an o/-prepositional phrase. It will be shown in more detail
that there is a very strong link of experiental identity (see Section 3.1.2)
between the head noun and the postmodifier in these uses. Like postnominal
thai-clauses and infinitives, these constructions create appositive links be-
tween the head nouns and the o/-prepositional phrases functioning as post-
modifiers.
126 Factual uses
'Advantage'
Shared semantic features: [FACTUAL], [ATTTTUDINAL], [FACILITATING]

Nouns: Advantage, benefit, bonus, blessing, asset
The 'Advantage' nouns are semantic opposites of the 'Problem' nouns.

Speakers use them to portray facts as facilitating progress towards an aim.
Unlike the 'Problem' nouns, some of the 'Advantage' nouns are used in the
pattern -that in noteworthy numbers. This pattern is illustrated in (7.29),
while (7.30) is an example of an -be-that use:
(7.29) ... and not only their families but their lovers, and their old companions
in the bar of the local, the members of their clubs, the people they had
known at work. We had a sellers' market. And the programmes had the
great advantage that the soldiers were speaking for themselves, and no-
body could say what they said was anything but true, (BOOKS)
(7.30) ... the Games. They are an experience of part of human culture that
unites rather than divides us, a melting pot of ethnic diversity. The ad-
vantage is that there is a huge audience that can hear other things vou
may have to say, (PAPERS)
There is no marked semantic difference between the meaning of the noun

advantage itself in the two uses. Both construe a state of affairs as favour-
able and as giving rise to a positive attitude. From a textual point of view,
the two examples are also similar in that both occurrences of advantage
stand in a particularizing relation to the preceding context. Both single out
positive aspects from a previously mentioned state of affairs.
The difference between the two examples and the two patterns they ex-
emplify lies in the distribution of information and the division of the sen-
tences into given and new information (see Section 3.3). Since this leads
into pragmatic and textual functions of shell nouns which will be discussed
in detail in Chapter 16, a brief comment will suffice here. In example
(7.29), the whole shell-content complex expresses new information. The
positive attitude is only introduced in the rheme, while theme of the su-
perordinate clause, the subject the programmes, has no attitudinal implica-
tions. This is different in example (7.30). Here the shell-noun phrase the
advantage is the theme from which the speaker starts out, portraying it as if
it were given information. The rheme of the main clause is represented by
the complementing that-clause expressing the shell content. This construc-

tion suggests that the positive attitude can be taken for granted at this stage
in the text to such an extent that it can be used as given information. Attitu-
dinal shell-noun phrases in the pattern -be-that, as in (7.30), contribute
more to the coherence of a text than those in the pattern -that, e.g. (7.29).
Advantage is by far the most frequent item in this family (frequency per
million: 60.2). Bonus, asset and blessing are fairly rare nouns, with attested
frequencies of 23, 18 and 9 occurrences per million words respectively.
Although benefit is more frequent (57 occurrences per million), it is not
often used as a shell noun.
Besides their occurrences in the patterns -that and -be-that, all five
nouns are attested in the pattern th-N, and all except asset in th-be-N. The
noun blessing, which carries perhaps the strongest attitudinal meaning of
the four nouns, stands out from the family because it was almost equally
frequently found in th-N and in th-be-H patterns (14 and 13 occurrences
respectively). The other three nouns exhibit the more common distribution,
according to which occurrences in the pattern th- clearly outweigh those in
the pattern th-be-N.
'Irony'
Shared semantic features: [FACTUAL], [ATTITUDINAL], [DESCRIPTIVE]
Nouns: Change, coincidence, irony, accident, para-
dox, fault, weakness, uncertainty, beauty,
folly, anomaly, twist, importance, novelty,
curiosity, oddity, charm, travesty, inevitabil-
ity, peculiarity, absurdity, paranoia
Even a brief glance at the list of nouns indicates that the 'Irony' family is
semantically very heterogenous. It must be emphasized that the nouns in
this family are not united by a similar meaning but by their potential to
portray facts in certain ways, or perhaps better, to portray facts as being
remarkable and noteworthy because of certain qualities. An example of the
prime noun in the family, irony, will help to show this mixture of factuality
and description:
(7.31) Germany's BMW and Audi have both developed seat-belt systems that
tighten up automatically in the event of a crash. The irony is that, so
128 Factual uses
far, all this has not paid off for those car makers most associated with
promoting their vehicles' safety, (ECON)
In this example, the fact expressed by means of the that-clause is charac-

terized as an irony by the speaker, as something strange or amusing. The
reason why this fact is amusing or strange can only be extracted from the
context with the help of a number of inferences. In this example, it is a clash
between the expectations of the two German car producers and the way
things turned out in reality. Borrowing parts of the descriptive framework
used by Tuggy (1996: 722) for these focus formulas (see Section 7.2), we
could say that the collocation the irony is that roughly means 'state of af-
fairs which is in amusing disconformity with something expected'. The
expectational element in this paraphrase is required as an inferential bridge
to connect the accomplishment expressed in the preceding sentence and the
statement of the fact that all this has not paid o f f . It indicates that the ac-
complishment was carried out with a certain goal in mind which has not
been achieved.
The other nouns in the family are pragmatically identical when they are
used in the pattern -be-that, but of course semantically different. How-
ever, when used in this pattern - and it is important to underscore this - all
of them seem to involve the element of disconformity which is employed by
Tuggy (1996) to capture the intuition of noteworthiness. To account for the
specific meanings of these nouns, reasons for the disconformity can be
given. Change, for example, can be glossed as 'state of affairs which is in
disconformity with something expected because it is different from it', twist
as 'state of affairs which is in disconformity with something expected be-
cause it is different from it and somehow strange', paradox as 'state of
affairs which is in disconformity with something expected because it is in-
compatible with it', novelty as 'state of affairs which is in disconformity
with something expected because it is new' and absurdity as 'state of af-
fairs which is in disconformity with something because it does not make
sense'.
It must be emphasized that apart from irony, none of these nouns occurs
particularly frequently in the pattern -be-that. However, when they do
occur as shell nouns, it is predominantly in this focusing pattern. Those
nouns that are found in the pattern -that are almost exclusively used in
extrapositional constructions of the type it was a coincidence that or it was
no accident that, which also have a focusing effect. The nouns in the
'Irony' family are thus specialized for the focusing function, and can there-
fore on the whole not be seen as very good examples of shell nouns. Finally,
it must be mentioned that change is only at the top of the list of nouns be-
cause it is used fairly frequently in the pattern th- as an eventive noun (see
Section 11.2).
'Tragedy'
Shared semantic features: [FACTUAL], [ATTITUDINAL], [EVALUATIVE],
[NEGATIVE]
Nouns: Pity, tragedy, shame, blow, disaster, night-
mare, scandal, drama, offence, disgrace,
misfortune, flaw, debacle, catastrophe, farce,
curse, nuisance
As already mentioned at the beginning of this section, most of the 'Tragedy'

nouns also produce untypical shell noun uses, since they occur mainly or
even exclusively in extrapositional constructions. Pity, tragedy, shame,
blow and scandal are the prime examples of this tendency, and the other
nouns are only less familiar in this pattern because they are much less fre-
quent altogether. Furthermore it must be emphasized that most of these
nouns are only factual when they are combined with //^/-clauses. When
they occur in the patterns th- and th-be-N, they do not project events as
facts, as in the pattern -that and -be-that, but refer back to events and
express the speakers' negative evaluations. The contrast between factual
and eventive uses is shown in example (7.32):
(7.32) (a) Roy Cornes's tragedy is that he was born with haemophilia.
(PAPERS)
(b) L J The doors are alarmed but it all happened so fast I haven't had
the opportunity yet to check whether they went off. This is a terrible
tragedy and I really don't know what made them do it. (TODAY)
Nonetheless, there can be no doubt that (7.32a) is a good example of a fac-

tual shell noun use. In this example, it is perfectly possible to insert the
noun phrase the fact after the that. In (7.32b), however, the insertion of
event after this is better than fact, if this would be acceptable at all.
130 Factual uses
'Miracle'
Shared semantic features: [FACTUAL], [ATTITUDINAL], [EVALUATIVE],
[POSITIVE]

Nouns: Wonder, miracle, luck, consolation, revolu-
tion, comfort, sensation, attraction, break-
through, virtue, merit, marvel, solace
The same restrictions apply to the positive counterparts of 'Tragedy' nouns,

which are subsumed in the 'Miracle' family. Again, extrapositional con-
structions like it is no wonder that and it was good luck that predominate in
the domain of -that uses. Most of the nouns occur in the pattern N-e-
that, but none of them are very frequent. And again, it must be stressed
especially for such nouns as revolution or breakthrough that only uses in
the patterns -that and -be-that produce factual readings; th- and th-be-
N uses are eventive. As example (7.33b) shows, even the claim that positive
evaluations are expressed is only true for uses with //zai-clauses but not
necessarily for th-be- uses.
(7.33) (a) The miracle is that it avoids being smug or snooty, (PAPERS)
(b) If things go on this way a whole political class is liable to be decapi-
tated. Good. This is a necessary revolution, which Italy would be
well advised to see coolly through to the finish, (ECON)
I have not devoted a lot of space to the discussion of the families 'Irony',
'Tragedy' and 'Miracle' because their output of proper shell-noun uses is
fairly low. What these uses reveal better than other more mundane, but
proper, factual uses, however, is the characterizing potential that shell
nouns offer speakers. For one must not forget that not just these strongly
evaluative or descriptive nouns, but also the apparently somewhat more
objective nouns in the 'Problem' and the 'Advantage' families, and all the
other nouns occurring in factual uses have the potential for characterization,
and that this potential is unconsciously exploited by speakers. In a way,
then, the term attitudinal shell noun uses is tautological, since all shell
nouns are attitudinal by definition.
Chapter 8
Linguistic uses
8.1 Introduction
Uses of linguistic shell nouns allow speakers to portray linguistic activities

and their contents and products in a number of different ways. In the pat-
terns N-that and -be-that (which are far more frequent than N-to and N-
be-to) as well as in th- and th-be- patterns, the contents of linguistic
shell-content complexes represent utterances (in the fairly broad under-
standing of this notion proposed in Section 5.1.1). The linguistic shell nouns
themselves contain the reporting speakers' characterization of the utterances
they report.57 For example in (8.1), the speaker uses the linguistic shell-
noun phrase the tired accusation to characterize the utterance they are lib-
erals and lackeys of special interest groups.
(8.1) To counter the apparent dynamism of the Democrat team, the best the
White House has been able to conjure up is the tired accusation that
they are liberals and lackeys of special interest groups, (PAPERS)
Linguistic shell-content complexes involve linguistic information of two

kinds which must carefully be kept apart. On the one hand, there is the in-
formation given in the reporting text by means of a shell-content complex.
Texts of this type are the object of this study. On the other hand, the infor-
mation given in these texts can be seen as being derived from other utter-
ances which are reported with the help of the shell-content complex. It is
with linguistic shell-noun uses, therefore, that the metacommunicative func-
tion attributed to abstract nouns by several scholars (see Section 2.1) is
most marked. The that-clause in (8.1), for example, can be traced back to
the utterance they are liberals and lackeys of special interest groups. It is
an example of what Halliday would call an embedded projected locution
(1994: 250-273, esp. 263-264; see also Sections 2.1 and 5.1.1 above).
The idea that utterances are reported is of course most conspicuous in
examples of quotations of direct speech. This is illustrated in (8.2):
132 Linguistic uses
(8.2) In a letter last week from his cell, Ali Akbar Saidi-Siijani, a well-known
Iranian writer, wrote that he had been 'falling into oblivion and forget-
ting conscience'. The Iranian regime is claiming this sad murmur as a
confession, (ECON)
However, it must be stressed that it is also involved in cases of projection

like (8.1) above, where the reporting function is less explicitly marked than
in (8.2). Uses of this type, that is combinations of linguistic shell nouns with
postnominal /2fl-clauses or infinitives, are the typical linguistic means of
reporting utterances. They will therefore be taken as the basis of the argu-
ment in this chapter. Whether the idea of utterance reports is also helpful
for the understanding of uses in other patterns will emerge as this chapter
unfolds. What seems clear right from the start is that the identifying pattern
N-e-cl involves other aspects as well (see Section 3.3). The function of
reporting can be attributed to such uses when the copula is in the past tense,
as in example (8.3):
(8.3) The Tories risk going into a general election with child benefit frozen
for four years, a state widely seen as untenable as a manifesto position,
but with no clear policy on what would replace it. Sir Norman's argu-
ment was that it should not be replaced but maintained, (PAPERS)
It is perhaps less felicitously attributed to uses in the pattern N-e-cl with

present forms of the verb be, as illustrated in (8.4).
(8.4) But the lawyers say that defining what is a major matter will give rise to
serious problems. Anthony Scrivener QC, for Channel 4, says: "The
plain answer is that the new wording does not make any improvement
a t a l l . " (PAPERS)
In such cases, the actual act of making an utterance and the wish to mark it
for special attention as focus of the clause seem to be more important mo-
tives for using a linguistic shell noun than reporting the utterance. To recon-
cile such uses as (8.4) with the idea of reporting, they will be treated as
special cases in which two acts are performed simultaneously: an utterance
is being produced and reported at the same time. The obvious advantage of
this somewhat paradoxical combination of speech acts for speakers is that
they can characterize their own utterances while making them.
When one thinks of linguistic shell-content complexes as reports of ut-
terances, one must keep two important points in mind. First, unless the re-
Introduction 133
port is given in the form of a direct quotation, it is necessarily subject to the

perception, interpretation and consequently the attitude of the reporting
speaker. In (8.3), for example, it is highly unlikely that the original utter-
ance by Sir Norman was as short as "Child benefit should not be replaced
but maintained". Rather we can assume that he said a lot more on this sub-
ject and that the shell content in the report is not more than a summary or
synopsis of his orginal statement, which represents the journalist's inter-
pretation of the gist of what Sir Norman originally said. So the idea that
linguistic shell contents simply represent prior utterances is a gross over-
simplification; but it is a very useful one as long as one bears in mind that
this is so. We must think of these 'original' utterances as theoretical con-
structs formulated for the sake of linguistic description rather than as utter-
ances that have actually been produced by some speaker at some time,
though this is of course perfectly possible.
Second, not just the report itself, but also and perhaps to an even greater
extent the characterization incorporated in the meaning of the linguistic shell
noun, are invariably controlled by the speaker of the reporting utterance
alone. This emerges nicely from (8.2), since a quotation is characterized in
two different ways by two sources. On the one hand, it is portrayed as a sad
murmur by the author of the contribution to The Economist. In addition to
his own characterization the author suggests that the Iranian regime used, or
at least would have used, a different shell noun, namely confession, to char-
acterize the same 'original' utterance. A potential declaration from this
source could thus have read "Ali Akbar Saidi-Siijani's confession that he
has been falling into oblivion and forgetting conscience ..." What this
clearly shows is that it is completely up to the reporting speaker how he or
she wishes to characterize a given utterance. Whether the characterization is
a true reflection of the communicative intentions of the original speaker
must remain an open question. The only case where a true correspondence
can be assumed is when a speaker reports his or her own previous utterance
(and is speaking truthfully and seriously).58
In sum, speakers who report utterances by means of linguistic shell-
content complexes leave their mark both on the characterization of the ut-
terance entailed in the shell noun and on the representation of the report
given as shell content.
The considerations outlined so far suggest a general frame setup for the
description of the use of linguistic shell nouns. The frame is represented
diagrammatically in Figure 8.1. It should be noted that the SPEAKER (on the
right-hand side in the diagram) of course also addresses his or her
134 Linguistic uses
UTTERANCE to one or several hearers. Although these hearers may clearly

have an influence on the SPEAKER'S utterance, I have not indicated them in
Figure 8.1 to keep the representation as straightforward as possible.
produces and renders as

addresses to shell content
ORIGINAL HEARER
ORIGINAL ORIGINAL UTTERANCE links SPEAKER

SPEAKER UTTERANCE
characterizes
as shell noun
Figure 8.1 General frame for the description of linguistic shell-noun uses
As in Chapter 7, the components of the frame are printed in small capitals

to mark their metalinguistic status. Since I will more frequently be con-
cerned with the producers of reporting texts containing shell-content com-
plexes than with producers of 'original' utterances, I have decided to use
shorter and unmarked terms for the former components. Thus, the source of
the reporting text is called SPEAKER, the source of the hypothetical 'origi-
nal' text ORIGINAL SPEAKER and the target of the 'original' text ORIGINAL
HEARER. The passage of text containing the report is called UTTERANCE
and the hypothetical 'original' utterance ORIGINAL UTTERANCE. SPEAKERS
express UTTERANCES as shell contents and shell them by means of linguistic
shell nouns. The arrows between the components ORIGINAL UTTERANCE and
UTTERANCE are there to indicate that there is a certain degree of cognitive
correspondence between the two but, unless a direct quotation functions as
shell content, never a complete equivalence. This frame is specific enough to
supply a tool for the description of linguistic shell-noun uses in contradis-
tinction to other classes, but it is also general enough to accommodate the
three subclasses which will be distinguished now.59
It is well-known since speech act theory emerged in the 1960s that utter-
ances can be viewed from at least three (or even four) perspectives. Ac-
cording to Austin (1962: 94-120), one can look upon any given utterance as
a locutionary act consisting of the production of sounds (phonetic act), the
arrangement of words in grammatical structures (phatic act) and the ex-
pression of sense and reference (rhetic act), as an illocutionary act involv-
Introduction 135
ing the pursuit of communicative aims; and as a perlocutionary act which

reflects the speaker's intentions as to the effects of his or her utterance on
the hearer. Searle (1969: 22-33) carves out the rhetic act from Austin's
conception of the locutionary act and establishes it as a separate act in its
own right, which consists of the acts of referring and predicating. He refers
to this act as the prepositional act.
Sticking henceforth to Searle's view we find that the four aspects of
speech acts are not of equal importance for the study of shell nouns. Per-
locutionary acts do play a role in the use of linguistic shell nouns in combi-
nations with infinitive clauses but they are comparatively marginal. Locu-
tionary shell nouns are extremely rare. The only candidate in the data for
what would be called locutionary shell-noun uses is the noun grumble be-
cause it has an onomatopoeic component reflecting the way something is
said. The majority of linguistic shell nouns are used to portray UTTERANCES
either as propositional or as illocutionary acts, with the latter clearly out-
weighing the former. Examples of fairly clear cases of propositional and
illocutionary uses of linguistic shell nouns are given in (8.5) and (8.6) re-
spectively:
(8.5) Not only was it Jane who raised Stephen's morale when faced with the
depressing news that his illness would eventually leave him paralysed
and speechless, but it was she who physically helped him to write his
blockbuster, (TODAY)
(8.6) The Association will give a warning that poll tax bills in some Conser-
vative districts will exceed government guidelines unless they receive a
higher proportion of the money being made available to local authori-
ties. (BBC)
The /ijf-clauses in both of these examples report the contents of the

ORIGINAL UTTERANCES. However, the shell nouns news and warning profile
two different perspectives. News evokes the propositional perspective, fo-
cusing on the information given in, or the propositional content expressed
by, the ORIGINAL UTTERANCE. The noun warning, which is morphologically
derived from a speech-act verb, profiles the illocutionary perspective. It
allows the SPEAKER to direct the focus of attention on the communicative
intentions which he or she imputes to the ORIGINAL SPEAKER'S ORIGINAL
UTTERANCE. A rough linguistic test for the distinction between the two
classes of linguistic shell nouns is that illocutionary uses pan be inserted in
the test frame given in (8.7) while propositional ones cannot.
136 Linguistic uses
(8.7) In saying [UTTERANCE], [ORIGINAL SPEAKER] performed a [shell noun].
The test is applied to examples (8.5) and (8.6) in (8.8) and (8.9) respec-
tively:
(8.8) *In saying that [Stephen's] illness would eventually leave him paralysed
and speechless. Jane performed a news.
(8.9) In saying that poll tax bills in some Conservative districts will exceed
government guidelines unless they receive a higher proportion of the
money being made available to local authorities, the Association per-
formed a warning.
The logic behind the test sentence in (8.7) is that only nouns that evoke the
illocutionary perspective can be paraphrased by the verb perform, which
spells out the performative element that is not incorporated in purely propo-
sitional shell-noun uses.
Whereas subclasses of linguistic shell-noun uses can be distinguished
fairly successfully with the help of linguistic tests, this proves to be impos-
sible for the class itself. The reason is that there is no clear boundary to the
class of mental shell nouns. Doubtful cases are for example such nouns as
admission, assumption, claim, forecast, guess and prediction and many
other nouns which seem to have the potential to denote either linguistic or
mental activities, even both of them at the same time.
How does this systematic overlap of linguistic and mental shell nouns
come about? Presumably its main source is the cognitive interdependence of
speech and thought. Looking at the relation from the mental end first, it is
well known at least since Humboldt, Sapir and Whorf that the way we think
is influenced by the language we speak. Other linguists and philosophers
have wondered whether thought without language is feasible and imaginable
at all. And finally, and probably most importantly, more often than not we
only learn about the thoughts of other persons when they tell us about them,
i.e. when they use language as a means of communicating them. Reversing
the perspective, one finds that one of the main functions of language is of
course to express thoughts. Furthermore, it is normally assumed that speech
is the result of prior mental activity (though this may perhaps not always be
true).
As a consequence of this interdependence of speech and thought, one
often cannot decide whether a SPEAKER using a shell noun is reporting an
UTTERANCE or an IDEA (which is the frame equivalent of UTTERANCE in the
mental domain; see Section 9.1). It is impossible to decide, for example,
Introduction 137
whether the SPEAKER of example (8.10) intends the underlined passage to

be understood as something Foucault has actually written or said, or as
something he has had in his mind at some time or other.
(8.10) Implicit in Foucault's work is an attack on the centrality of labour in

emancipatory politics. His thought proceeds from the assumption that
the working class, through its place in the process of production, is not
the vanguard of social change, (BOOKS)
If Foucault has actually said or written that the working class ... is not the
vanguard of social change, it can be assumed that he also had the corre-
sponding IDEA in his mind at that time. Conversely, even if he did not,
common sense certainly suggests that somehow or other, the IDEA must
have been communicated to the world outside Foucault's mind by means of
language. Or how else could the SPEAKER of (8.10) know that this is one of
Foucault's ideas? The inescapable upshot of these considerations is that
(8.10) is ambiguous, and that the ambiguity cannot be resolved. Although
this may appear to be a somewhat fatalistic conclusion from a semantic
point of view, a pragmatic perspective suggests that the ambiguity need not
be resolved at all because it does not interfere with a satisfactory under-
standing of the passage. Clearly it is not even noticed in everyday language
use.
From a systematic perspective, i.e. in the lexicon of English, this ambi-
guity is reflected in the fact that there is a large number of verbs and nouns
which are polysemous or even vague with regard to the distinction between
speech and thought reports. The verbs conclude and reveal for example (see
Leech 1983: 204), but also admit, bet, guess, claim, maintain and profess,
and many related and other nouns, can denote both mental and linguistic
activities and processes.
Nevertheless I think that a distinction between the two classes should be
made because there are also many nouns which can only be used either as
linguistic or as mental shell nouns. The best way of distinguishing between
the two uses is based on the recognition that by their very nature as physi-
cally observable acts, speech acts are agentive and dynamic events; mental
processes that take place in our black box, on the other hand, are thought of
as being less agentive and more stative in nature. Thoughts and especially
ideas and concepts are conceived of as cognitive states rather than acts. As
is well known, this distinction is encoded in the grammar of English in the
fact that verbs denoting private states of an intellectual or emotional nature
138 Linguistic uses
do not occur easily with the progressive aspect (Quirk et al. 1985: 202-
203), while verbs denoting acts of speaking exhibit no such constraints. *He
was knowing the answer is ungrammatical, while He was telling a joke is
perfectly acceptable in a suitable context.
As far as nouns are concerned, one can derive a simple test frame as a
rule of thumb for the classification required for this study. If a doubtful
noun can be used as an object of a dynamic verb such as perform, make or
give as in (8.11a), it will be classified as a linguistic shell noun. If it is bet-
ter suited for the combination with stative verbs like have or hold as in
(8.1 lb), it is dealt with in the chapter on mental shell nouns.
(8.11) (a) [ORIGINAL SPEAKER] said that [UTTERANCE], and in saying it, s/he
performed/made/gave a [shell noun],
(b) [EXPERIENCER] thought that [IDEA], and in thinking it, s/he had/held
a [shell noun}.
These test frames can supply no more than a rough guideline, though, be-
cause exceptions can easily be found. Furthermore, there are combinations
of dynamic verbs with stative mental shell nouns, such as the collocations
draw a conclusion or form a plan.
In analogy to Chapter 7, this introductory section of Chapter 8 closes
with an overview of the groups and families of linguistic shell-noun uses
and the Sections in which they are dealt with.
Groups Type of Families Examples Section

speech act
Preposi- 'News' news, argument, message 8.2
tional uses 'Adage' adage, catchphrase
'Rumour' rumour, gossip, whisper
'Myth' myth, legend, doctrine
Illocution- Assertive 'Statement' statement, observation 8.3.1
ary uses 'Proclamation' proclamation, decree
'Argument' argument, concession
'Report' report, account, story
'Prediction' prediction, bet, forecast
'Guess' guess, claim, hint
'Assessment' assessment, verdict
'Disclosure' disclosure, revelation
'Lie' lie, distortion, pretext
Introduction 139
Rogative 'Question' question, query 8.3.2
Directive 'Order' order, request, command 8.3.3

'Invitation' invitation, plea, appeal
'Advice' advice, recommendation
'Suggestion' suggestion, proposal
Commis- 'Offer' offer, bid 8.3.4

sive 'Promise' promise, pledge, oath
'Threat' threat, warning, caveat
Expressive 'Complaint' complaint, gripe, whinge 8.3.5

'Compliment' compliment, praise
Figure 8.2 Survey of the hierarchical structure of the class of linguistic shell-
noun uses
8.2 Propositional uses
Compared to the long list of nouns which are at speakers' disposal for the
characterization of UTTERANCES as illocutions (see Section 8.3), the inven-
tory of propositional shell nouns is rather limited. One possible reason is
that illocutionary shell nouns are in a way more economical than proposi-
tional shell nouns. There is no additional phonetic effort required to add a
characterization of an UTTERANCE from an illocutionary perspective, in-
stead of simply linking the UTTERANCE to locutionary nouns like message
or news. For example, if the predicates are adapted as required, there are no
further problems in exchanging the propositional noun news in (8.5) with
the illocutionary noun warning in (8.6). The propositional contents of the
respective UTTERANCES would not be affected by this move, because they
can be linked to the illocutionary shell noun just as well as to the proposi-
tional noun.
In view of these considerations one might go a step further and begin to
wonder why people use propositional shell nouns at all, if illocutionary shell
nouns help them to kill two birds with one stone. Two explanations come to
mind here. The first, which applies to the more neutral propositional uses
collected in the 'News' family, builds on the idea of the allocation of atten-
140 Linguistic uses
tion. Speakers seem to use propositional shell nouns rather than illocution-
ary ones, when they wish to direct the hearers' attention especially to the
propositional content of an utterance. When illocutionary shell nouns are
used the propositional content is expressed as well, but the focus is shifted
towards the illocutionary force of the utterance. This difference is reflected
collocationally in the fact that only a few illocutionary shell nouns, mainly
answer, conclusion, explanation, guess and suggestion, occur frequently in
the focusing pattern N-e-cl, while for the propositional shell nouns news,
message, argument and story, this is a highly characteristic and frequent
pattern.
Second, speakers use propositional nouns when they want to shell the
contents of utterances from certain perspectives. A number of different
types of propositional shell nouns are at speakers' disposal to this end. For
example, expressions like the myth that successfiil women are slim and
good-looking (MAGS) (see example ( 8 . 2 1 ) below) are a means of shelling
the propositional contents of utterances with a simultaneous characteriza-
tion from a metalinguistic perspective. There are three families of this type.
The 'Adage' family consists of nouns which construe UTTERANCES as in-
stantiations of linguistic categories, mainly as text-types of some sort. The
nouns in the 'Rumour' family report UTTERANCES in such a way that the
knowledge of their propositional content is portrayed by the SPEAKER as
being shared by many ORIGINAL SPEAKERS. And finally the 'Myth' family
consists of nouns which straddle the boundary between the other two fami-
lies and are used either way, denoting a linguistic category and/or an ex-
pression known to many speakers of a speech community.
'News'
Shared semantic features: [LINGUISTIC]
Frame: Linguistic; focus on propositional content

Nouns: Point, report, news, argument, information,
story, message, intelligence
The first propositional shell-noun to be discussed here is the 'patron saint'

of this family, the noun news. According to the first entry for this noun
given in LDOCE3, news denotes "information about something that has
happened recently". Translated into my framework this would mean that
Prepositional uses 141
news is used when the events that are reported have taken place a short time
before the ORIGINAL UTTERANCE is produced. Straightforward as this may
sound, the relation between events, utterances and their contents which un-
derlies the use of propositional shell nouns is not that easy at all, because
conceptual shifts are at work. Consider the following example:
(8.12) "The UK faces pressure on hostages" is the main headline in the

FINANCIAL TIMES following the news that two of the captives a
London journalist and a teacher with dual British/Irish nationality
were seen as recently as last Saturday by the newly released American
hostage, Frank Reed, (BBC)
This example provides a good opportunity to return to the issues of con-

ceptual shifts and projection (see Sections 5.1.1, 5.1.2 and 7.2) because the
shell content realized by the that-clause does not seem to denote an
UTTERANCE at all, but an event, the event that two of the captives ... were
seen ... by ... Frank Reed. This would mean that the shell-content complex
is used as a report of an event after all here, and not as a report of an utter-
ance. However, this is too superficial a view of the reporting situation. To
show this, one can try to substitute the shell noun news with the noun event,
as in (8.12'), and contrast it with the original version:
(8.12') "The UK faces pressure on hostages" is the main headline in the

FINANCIAL TIMES following the event that two of the captives ...
were seen... by the newly released American hostage. Frank Reed, (BBC)
Like its counterpart (8.12), (8.12') is an acceptable English sentence. But it

does not convey the same meaning as (8.12) because it is an event report. It
conveys that it was an event that has caused the Financial Times to publish
a certain headline. Example (8.12), on the other hand, says that the headline
was motivated by an utterance with the propositional content that a certain
event has happened. It implies that the people at the Financial Times re-
ceived the knowledge of the event by linguistic means. The use of the pro-
positional noun news has the effect that a clause which could just as well
represent a report of an event (when it is looked at in isolation) is construed
as a report of an utterance.
The projected propositional contents shelled by the noun news typically
represent past events. In contrast, the noun argument is earmarked for
shelling projected abstract relations, typically ideas, as reporting UTTER-
ANCES. While the shell contents of news can easily be mistaken for event
142 Linguistic uses
descriptions, the contents linked up with argument can be regarded as ideas.

The view proposed here is that people use the noun argument to shell
UTTERANCES which serve the purpose of supporting a person's opinion, of
substantiating a person's claims and of giving reasons for a person's aims
in a conversation or debate. Once a cognitive entity is shelled as an argu-
ment, it is effectively turned into an UTTERANCE, no matter what kind of
experience it may have been originally. The noun brings about a change or
shift of the conceptual status of an entity. This is also the case when, as in
( 8 . 1 3 ) and many other examples in my data, neither an ORIGINAL SPEAKER
nor its counterpart in the mental domain, an EXPERIENCER, of a potential
ORIGINAL UTTERANCE or IDEA are identified.
(8.13) But even the most recent declaration of faith put forward to defend the
progress of science as something meaningful - the argument that if it
does not trace a pathway to God, then it does at least represent a path-
way to happiness - has, in Weber's view, now been demolished, (BOOKS)
Message and information are the two semantically most neutral nouns in
the family. As is well known, people think of language as a means of send-
ing and receiving messages, but of the mind as an information processing
device. This shows that the former noun is firmly set within the domain of
language, while the latter reaches out into the cognitive domain as well. A
juxtaposition of these two nouns, as provided in ( 8 . 1 4 ) and ( 8 . 1 5 ) , helps to
profile this difference:
(8.14) A conference on sickle cell disease is spreading the message that the
blood disorder needn't stop sufferers achieving what they want, (MAGS)
(8.15) "Friendly newspaper sources" are then said to have passed on to Mr
Burnside information that a secretary in BA's headquarters had been
feeding details to Virgin, (PAPERS)
Intelligence is used much less frequently as a shell noun than the other
nouns in the family. It represents yet another example of the tendency that
semantically specific shell nouns are rarer than unspecific ones, the specific
semantic feature in this case being that the knowledge of the propositional
content it shells has been gained by secret governmental institutions and
agencies.
Both report and story are frequently used as shell nouns. Yet it is doubt-
ful to what extent their uses can be regarded as propositional shell-noun
uses. Their questionable status even emerges from a look at their colloca-
Propositional uses 143
tional data, which differ markedly from those of news, message and argu-
ment. These three typical propositional shell nouns were found more often
in the patterns -that (the news/message/argument that...) and ~N-be-that
(ithe news/message/argument be that...) than in the pattern th- (this/that
newsImessage!argument). Report and story, on the other hand, occur less
often in the pattern -that, and only story is found in the pattern -be-that
at all. Instead they are much more frequent in the pattern th-N. The collo-
cation this/fthat report was found almost 5,000 times in the corpus, this
story over 1,300 times. This makes report the fifth most frequent shell noun
in the pattern th- (behind time, way, point and case) and story the 29th
most frequent one. The enormous frequency of the collocation this report is
undoubtedly an artefact of the composition of the corpus. It is mainly due to
its use by the BBC in cataphoric introductions of contributions which iden-
tify the author of a report. See the examples in (8.16):
(8.16) This report from our correspondent in Budapest KIRSTY LANG: The
question of how the president should be chosen ...
From Belgrade, Jim Fish filed this report: Despite bad weather and
poor organisation...
Christopher Morris sent this report from Colombo: The spokesman said
the operation had been highly successful...
In these examples, report functions not as a propositional but as a meta-

linguistic shell noun (see the 'Report' family below). Story occurs in a
similar function and with a similar meaning (just think of the cover story of
newspapers), though less frequently and in a less stereotypical way.
Unlike report, the noun story is found in the pattern -be-that. In these
patterns it undoubtedly functions as a propositional shell noun. This is il-
lustrated in (8.17) with an example taken from spoken conversation:
(8.17) [... erm but of that six point five was a national price hike and then there
was a six per cent surcharge put on er fuel sold in in the Mexico City
metropolitan area and in fact in two other cities.] Erm and the story was
that this money from this would be used to finance er from this sur-
charge this money would be used to finance erm er pollution control
measures.
Story has a relatively high reliance score for the pattern th-be-H. Although
its overall frequency in the corpus is not uncommonly high, it takes the
eleventh rank in the frequency list for the pattern this/that + be + shell-
144 Linguistic uses
noun phrase. The typical collocations, which are especially frequent in spo-
ken English, fall into two classes. On the one hand, there are such frozen
expressions as that's my story, that's another story, that's a different story
and that's a long story, which are idiomatized to such an extent that their
composite meaning can no longer be explained on the basis of the meaning
of the noun story. Collocations with evaluative adjectives, on the other
hand, such as that's a good/interesting/lovely/crazy/wonderful story, do
not seem to have lost this connection but can be interpreted as metalinguis-
tic shell nouns (see 'Report'), or as propositional shell nouns whose con-
tents report extended ORIGINAL UTTERANCES.
'Adage'
Shared semantic features: [LINGUISTIC], [METALINGUISTIC]
Frame: Linguistic; focus on propositional content and

metalinguistic categorization
Nouns: Joke, nonsense, adage, preface, subtext,
aphorism, catchphrase, definition, proverb,
allegory
With the exception of the nouns joke and nonsense, the members of the
'Adage' family are all fairly rare words. Joke can muster occurrences in N-
that and -be-that patterns, in the pattern th-N, and in a remarkably large
proportion of th-be-N uses. However, we must be careful with this noun
because not all of its uses have the function of linguistic shell nouns. When
it is used in the pattern -that, as in example (8.18a), one can usually be
fairly sure that the shell content represents an utterance. For uses of the
pattern -be-that, on the other hand, this is by no means guaranteed. While
example (8.18b) can still be looked upon as a report of the content of an
utterance, the shell content in (c) can just as well be taken for a state of
affairs as for an utterance, if the factual interpretation is not even more
convincing.
(8.18) (a) It was indicative of Mr Battle's isolation from the common men of
Wall Street that he was clearly unaware of the ancient joke [...] that
Ivan Boesky always used this classic line as a preface to his requests
for illegal insider information, (BOOKS)
Propositional uses 145
(b) Our name then was Bezroff, he was a count - though the joke used
to be that anyone who owned three sheep in Georgia had a title of
some sort, (BOOKS)
(c) The real joke is that already no one can remember all the names.
(BOOKS)
Similar evaluative factual interpretations also suggest themselves for many

uses ofjoke in the pattern th-be-N.
The noun nonsense obviously involves a strong attitudinal component. It
allows speakers to express their disapproval of the propositional content of
an UTTERANCE. The noun is sometimes used in the pattern -that, but
much more frequently found in the pattern th-N, especially in spoken Eng-
lish and in less formal style.
The other nouns in the 'Adage' family are used in the patterns -that,
th- and rarely in the pattern th-be-N. In view of the short list of nouns,
readers should be reminded that I am not claiming that there are no other
metalinguistic nouns which are used to refer to stretches of text with the aim
of classifying them from a metalinguistic point of view. The list is so short
because it was the methodological policy of this study to include only those
nouns which were found to occur in N-cl and/or N-e-cl uses. Other related
nouns like chapter, passage or excerpt, which are mainly used with ana-
phoric reference in the patterns th- and th-be-N, can be found in Francis
(1986, 1994), Sinclair (1990: 391) and Winter (1992).
'Rumour'
Frame: Linguistic; focus on unidentified ORIGINAL
SPEAKER and widespread knowledge of propo-
sitional content
Nouns: Talk, rumour, gossip, whisper
The central features of these nouns are that the information shelled by them
is portrayed as being passed from one person to another in a more or less
secret manner and may be true or untrue. 'Rumour' nouns are thus par-
ticularly subtle communicative tools. Speakers can use them to perform
what can be called covert assertive speech acts. They can distance them-
selves from the proposition they state as shell content and detach themselves
146 Linguistic uses
from the responsibility of averring its truth, since the 'Rumour' nouns help
to suggest that the source of the information is unknown and that many
other people have the same infomation, too. A wonderfully explicit example
of such a use of rumour was found in a quotation in the newspaper data,
which is rendered as (8.19):
(8.19) But Forest are bottom of the Premier League and Betty Higgins, the
leader of the Labour-controlled council, said: "We had heard a rumour
that Brian would not be in charge of Forest for much longer. It was just
something that someone said." (PAPERS)
The consequence of this semantic configuration is that the nouns are rare in
the pattern -be-that which, as we have seen in various places in this
chapter, tend to be used to explicitly identify utterances and their content,
witness such expressions as the bad news was that.... Since the communi-
cative effects of the pattern -that rest to a greater extent on presupposi-
tions and implications (see Section 15.1), they provide a more subtle way of
using these nouns. (8.20) provides a second illustration of this pattern, this
time with the rarer noun talk.
(8.20) After Lukoil, Russia's largest state-owned oil-producer, had joined the
project earlier this year, there was talk that a contract to develop the
oilfields would soon be agreed upon, (ECON)
One further related noun should be mentioned here, although it is not in-
cluded in the data proper. It is the noun word, which is used the same way
as 'Rumour' nouns in such unusual collocations as word has it that ...,
word spread that... and word went out that....
'Myth'
Frame: Linguistic; focus on shared knowledge of

propositional content
Nouns: Myth, law, formula, doctrine, teaching, leg-
end, dictum, legacy, maxim, clich, truism,
stereotype, slogan, dogma, metaphor, ax-
iom, credo, motto, orthodoxy, tenet
Prepositional uses 147
The 'Myth' nouns share with the 'Rumour' nouns the idea that many people
are familiar with the utterances shelled by them. With the 'Adage' nouns
they have in common that they characterize utterances as instances of lin-
guistic categories, though these are perhaps even less clearly defined than
the 'Adage' ones. In contrast to the 'Adage' nouns, which are unambigu-
ously linguistic categories, many 'Myth' nouns also reach into the mental
domain. In a way, they oscillate between shelling speech and thought. Ax-
iom, credo, dogma, doctrine, law, maxim, metaphor, motto, stereotype and
teaching, as well as myth itself, can be used to shell both utterances and
ideas (see Section 9.2). In (8.21) for example, the core of which was already
quoted earlier, it is impossible - and, as in other cases of ambiguity that we
have encountered before, unnecessary - to decide whether myth shells the
content of an often-used, well-known statement or a widely-shared idea:
(8.21) To someone lacking in confidence, and sensitive about their weight and
body shape - and let's face it, many women channel their insecurities
into that very area, responding to the myth that successful women are
slim and good-looking - the idea of having your cake, eating it and not
paying the price in pounds of flesh may seem appealing, (MAGS)
The use of clich, dictum, formula, slogan and truism, on the other hand, is
more limited to the linguistic domain.
The favourite habitat of the 'Myth' nouns is the pattern -that. With the
exception of dictum and legacy, which are not recorded in the pattern th-be-
N, all nouns, even very rare ones like credo, dogma, tenet and truism, were
found in both th-N and th-be- patterns. Also remarkable are the high
compiled reliance scores of 38.67% and 37.35% for truism and dictum
respectively, which are largely due to occurrences of the two nouns in the
pattern -that.
8.3 Illocutionary uses
As explained at the beginning of Section 8.2, illocutionary uses highlight the

illocutionary force of an utterance, while also allowing for its propositional
content to be expressed as shell content.
The bulk of illocutionary uses is supplied by nouns which are morpho-
logically related to speech act verbs. In principle, every single speech act
verb - and according to Austin's (1962: 150) famous estimate there are
between 1,000 and 9,999 of them in English - can be turned into a noun and
148 Linguistic uses
then act as an illocutionary shell noun. In view of this enormous potential,

the reader will perhaps understand the importance of the From-Corpus-to-
Cognition Principle postulated in Section 4.1. If it were not for this princi-
ple, we would face a virtually endless list of nouns. However, the corpus
data that I have collected provides an objective criterion for the decision as
to which nouns seem to be worthy of our attention and which can be ne-
glected because they are rare.
According to my material, 110 nouns produce illocutionary shell-noun
uses in noteworthy numbers. Of these, only the following nine are not, at
least not from a synchronic perspective, morphologically related to speech
act verbs: caveat, edict, injunction, motion, oath, pretext, prognosis, ver-
dict, version. All the other 101 nouns are related to verbs in one of a num-
ber of possible ways. From a synchronic point of view, the most common
form of morphological relation is that the illocutionary noun is a deverbal
suffixation. The most frequent suffixes are -(a)tion as in affirmation, con-
firmation, contradiction, explanation and suggestion, -ment as in acknow-
ledgement, amendment, announcement, argument, pronouncement and
statement, and -ance or -enee as in assurance, grievance, insistence and
reassurance. Less frequent suffixes are -ing as in betting, ruling and un-
dertaking and -ure in disclosure. From a diachronic perspective, it is im-
portant to stress the fact that the majority of these nouns had completed the
suffixation process long before they entered the English language from
Latin or French sources.
The nouns answer, bet, charge, claim, comment, forecast, guess, grum-
ble, hint, lie, order, protest, quibble, remark, reply and vow illustrate the
second frequent morphological relation between speech act nouns and verbs,
zero-derivation (Marchand 1969: 359-389) or conversion (Quirk et al.
1985: 1558-1567). As is well-known, this type of nominalization is more
productive than suffixation with words of native origin. It accounts for a
little more than a third of the 101 nouns that are related to speech act verbs.
(Two examples with graphemic and phonological differences between noun
and verb are the pairs prophecyN - prophesyv 60 and advice^- advisev .)
Three points should be made concerning the finding that more than 90
per cent of the illocutionary shell nouns included in my data are at least
morphologically related to, most of them in fact even nominalizations of,
speech act verbs.
The first concerns the effects of the use of nouns for the description of
speech acts. As opposed to speech act verbs, speech act nouns have the
advantage that the participants of a speech act, i.e. ORIGINAL SPEAKER and
Illocutionary uses 149
ORIGINAL HEARER in my framework, can easily be omitted. When a verbal

construction is used to report a speech act (e.g. he said that..., she told us
to ..., they accused them o/...), the syntactic constraint that a clause must
have a subject prescribes that an agent is mentioned. When the speaker opts
for a passive construction (e.g. we were told to ..., they were accused of...),
the ORIGINAL SPEAKER is backgrounded but the ORIGINAL HEARER receives
all the more attention. With nominal expressions both participants can be
backgrounded, as the nominal counterparts to the three examples in the
previous sentence the statement that..., the order that... and the accusa-
tion that ... show. Like the passive, nominal constructions do allow
SPEAKERS to name the ORIGINAL SPEAKER, if they wish to do so. With pas-
sives, this is done by means of >_y-adverbials (we were told by the manager
...), and with speech act nouns by means of possessives indicating the agent,
e.g. his statement that..., her order to ... and the attorney's charge that....
Second, one must realize that, due to the very fact that they are derived
there is an enormous potential for ambiguity inherent in the meanings of
these nominalizations (see e.g. Bauer 1983: 185-187; and Section 5.1.1).
This ambiguity operates on top of and side by side with the systematic am-
biguity between speech- and thought-reporting nouns which is caused by the
interrelation of linguistic and cognitive aspects of language. Single exam-
ples of deverbal speech act nouns can exhibit highly idiosyncratic results of
lexicalization processes which may have occurred during their history either
in English or in the donor language(s). And on a more general level, an al-
most systematic tendency for ambiguity can be observed. Most nominalized
speech act nouns allow for at least three types of readings, an 'act of V-ing'
(henceforth 'ACT') reading, a 'content of the act of V-ing' reading ('CON-
TENT') and a 'concrete product of act of V-ing' reading ('PRODUCT').
The three types of readings represent second-order entities (or activities),
third-order entities (or ideas) and first-order entities respectively. A much
rarer additional possibility is the reading 'fact that someone has V-ed'
('FACT'), which also represents a third-order entity.
Two examples of shell content relations which can serve as fairly unam-
biguous illustrations of ACT and CONTENT readings were already given
in Section 5.1.1 (see examples 5.4 and 5.5). To flesh out this distinction,
four examples of the noun declaration are collected in (8.22) which illus-
trate the four possible meanings particularly clearly. It must be added, how-
ever, that (a) is highly questionable as far as its status as shell noun is con-
cerned.
150 Linguistic uses
(8.22) (a) PRODUCT of act of declaring something

Saddam Hussein is lusting for revenge and hires the Mafia to steal
the American Declaration of Independence, which he intends to
burn on CNN on July 4. (PAPERS)
(b) ACT of declaring something
Remember, your wedding is not a theatrical performance but a dec-
laration of love and commitment, (MAGS)
(c) CONTENT of act of declaring something
The declaration recommended that the Geneva-based UN Human
Rights Centre be better funded and co-ordinated, but critics coun-
tered that specific proposals on money and timetables for action were
rejected, (PAPERS)
(d) FACT that something has been declared
Western delegates insisted that, taking into account the difficulties of
achieving consensus at a meeting of 170 countries, the final declara-
tion marked a step forward in seeking to move human rights higher
up the international agenda, (PAPERS)
Unambigous examples such as those collected in (8.22) are hard to find in

the corpus. This suggests that they are fairly rare. The rule for nominalized
illocutionary shell nouns seems to be that their potential for systematic am-
biguity is hardly reduced in actual usage. Rarely is the context so specific
that it allows for a definite disambiguation. If one sticks to shell-noun uses
of nominalizations and looks at a random rather than hand-picked sample of
examples, the picture differs considerably from the one evoked by (8.22). A
set of four randomly selected examples of the collocation suggestion that is
given in (8.23):
(8.23) (a) In his statement confirming the nine arrests, the Law and Order
Minister says this should quash any suggestion that the police are
being tardy in the investigation of right-wing violence, (BBC)
(b) The BBC Foreign Affairs correspondent says her tough line appar-
ently came in response to the envoy's suggestion that President Sad-
dam Hussein was prepared to do a deal over Kuwait, (BBC)
(c) But with total party income for 1989 at pounds 8.1m, the corporate
figures also appeared to support the suggestion that Tory funds came
substantially from undeclared anonymous sources, as well as con-
stituency parties, (PAPERS)
(d) Traditionalists will sneer at Sweeney's suggestion that mulberry can
be worn with a blue or grey suit, but it goes with most outfits, (MAGS)
The first impression given by these examples is that a CONTENT reading

cannot be ruled out for any of them. Using the predicates which occur with
suggestion as major clues, one finds that they are all compatible with this
type of reading. This impression presumably arises from the fact already
noted above that the propositional content of an UTTERANCE is expressed as
shell content, even when an illocutionary and not a propositional shell noun
is used. In example (a), an ACT reading is possible as well, and the same is
true for example (b). PRODUCT readings do not really suggest themselves
for any of the examples. A FACT reading can be taken into consideration
for example (d) - yielding the paraphrase will sneer at the fact that Mul-
berry suggested that ... - but it is less convincing than the CONTENT
reading.
There will only be the occasional instance where observations concern-
ing the frequency and likelihood of a certain type of reading can be made in
the discussions of families of illocutionary shell-noun uses below. On the
whole, lack of space prevents me from probing much deeper into the prob-
lem of the semantic variability of nominalizations. What the examples given
in (8.23) seem to suggest in any case is that ambiguities between
CONTENT and ACT readings or CONTENT and FACT readings need not
be resolved at all. A successful and satisfactory understanding is perfectly
possible without an ultimate disambiguation. The probability of dramatic
misunderstandings is extremely low.
The morpho-semantic connections of illocutionary shell nouns to speech
act verbs have a third important consequence. The description of illocution-
ary uses can build on the wealth of insights gained by language philoso-
phers and pragmalinguists on the classification and description of speech
acts and speech act verbs. Thus the overall pattern for the subclassification
of illocutionary uses of shell nouns can be taken over from other authors. In
doing so, however, one has to be careful. As is well known, classifying
speech acts is one thing, and classifying speech act verbs is another. Ever
since Searle (1976: 3-4) attacked Austin on the grounds that the founder of
speech act theory had categorized speech act verbs rather than speech acts,
most speech act theorists have taken great pains to focus on the classifica-
tion of speech acts and not of single words.61 For most scholars working in
the paradigm, this approach was more in line with their pragmalinguistic
inclinations. At the present stage of this study, I am mainly concerned with
the forms and meanings of words, i.e. with grammatical and lexicological
matters rather than communicative aims and functions, and therefore my
152 Linguistic uses
account of illocutionary shell-noun uses follows speech act verb descrip-

tions and not speech act classifications.
The three best-known attempts in this field are Ballmer and Brennen-
stuhl's Speech act classification. A study in the lexical analysis of English
speech act verbs (1981), Wierzbicka's (1987) dictionary of English speech
act verbs and Leech's chapter on speech act verbs in English (1983: 198-
228). The first is largely a thesaurus-like lexicon consisting of a complex
system of lists of verbs and their semantic characteristics. As its name sug-
gests, Wierzbicka's book is also conceived as a dictionary, though illumi-
nating comments on fine semantic differences, usage and style abound. The
descriptions in both books are far more specific than anything I could aim at
in the context of my wider field of interest. Hence they are more useful as a
background for the description of single nouns than for the general classifi-
cation. Here Leech's proposal is more helpful and therefore his approach
will serve as my basis.
The core of Leech's (1983) classification of speech act verbs consists of
five classes of illocutionary verbs: assertive, directive, commissive, expres-
sive and rogative verbs. These classes are largely derived from Searle's
classification of illocutionary acts, but further supported by semantic and
syntactic considerations. Especially the last class, the class of rogative
verbs, with such members as ask, query and inquire, owes its existence to
the semantico-syntactic criterion that they select "an indirect question as
complementizer" (Leech 1983: 206). In addition to the five classes of illo-
cutionary verbs, which will be transferred to the nominal domain in what
follows, Leech (1983: 214) has "perlocutionary verbs" such as persuade,
"locutionary verbs" such as describe ... as, "neutral verbs" such as say, and
finally "phonologically descriptive verbs" such as mutter. These other
classes will not feature prominently in this study because their relevance for
nouns is negligible.
As the survey in Figure 8.2 above shows, Leech's classes of illocution-
ary verbs make up an intermediate classificatory level between the group of
illocutionary uses and the families subsumed in it. The discussion will begin
with assertive uses (Section 8.3.1), the largest intermediate category with
nine subordinate families. Then rogative uses (Section 8.3.2; one family),
directive uses (8.3.3; four families), commissive uses (8.3.4; three families)
and finally expressive uses (8.3.5; two families) will be described.
IUocutionary uses 153
8.3.1 Assertive uses
Assertive shell nouns are used to characterize utterances as (contents of)

acts of stating states of affairs and other abstract relations or (contents of)
acts of reporting events. They are the nominal counterparts of assertive
speech act verbs and most of them, though not all, are morphologically
related to verbs.
The subcategorization of assertive uses into families is no easy task.
Looking for guidance in the related verbal domain, one is confronted with
Leech's somewhat discouraging finding that assertive verbs "lend them-
selves less easily [than other speech act verbs HJS] to systematic analysis"
(1983: 223). Nevertheless Leech comes up with the helpful, albeit tentative,
suggestion that the following factors seem to be of importance for the de-
scription of assertive verbs (1983: 224). I have taken the liberty of trans-
lating Leech's suggestions into my own framework and of adding examples
of nouns, even though he deals with verbs:
a) Predictive versus retrodictive assertion: is the event or abstract rela-

tion that is reported by the UTTERANCE characterized by the SPEAKER
as being posterior (report, account) or prior (prediction, forecast) to
the ORIGINAL UTTERANCE?
b) Public versus private assertion: is the ORIGINAL UTTERANCE charac-
terized by the SPEAKER as being made known publicly (announcement,
proclamation) or privately and obliquely?62
c) Confident versus tentative assertion: is the ORIGINAL UTTERANCE
characterized by the SPEAKER as being made with trust (confirmation,
declaration) or with doubts (guess, suggestion) in the truth of the as-
sertion?
d) Informative versus argumentative assertion: is the ORIGINAL
UTTERANCE characterized by the SPEAKER as being used mainly with
the aim of informing (description, proclamation) or of persuading the
other discourse participants, of justifying, defending or arguing for his
or her positions in a conversation (defence, protest)?
As some of the examples indicate, these dichotomies do not help in creating

mutually exclusive subcategories but must be seen as semantic dimensions
that are potentially helpful for a description. Proclamation, for example, is
not just [INFORMATIVE] on the fourth dimension but also [PUBLIC] on the
second. Similarly, hint tends towards the feature [PRIVATE] on the second
154 Linguistic uses
dimension, but this is more or less conflated with the feature [TENTATIVE]
on the third. In fact, there seems to be a general affinity between public and
confident assertions and private and tentative ones. A rigid subclassification
of assertive shell nouns on the basis of these dimensions is therefore impos-
sible. Furthermore, my data suggests that a number of other fairly specific
dimensions play a role as well: a verdictive element in such nouns as as-
sessment and judgement, the feature that the propositional content of a
statement was previously unknown, e.g. in disclosure and revelation; the
idea that a statement is untrue, e.g. in such nouns as lie or distortion. On
the other hand, there are nouns in whose meanings none of the dimensions
referred to so far can be detected. These are nouns like statement itself, or
observation and remark, which of course have some additional semantic
characteristics, but are on the whole fairly neutral.
The combination of fairly general, widely-applicable features such as
those mentioned by Leech on the one hand, and of some much more specific
ones on the other, has the effect that we are confronted with two different
types of families of assertive uses of shell nouns. The families of the first
type, 'Statement', 'Proclamation' and 'Argument', all display very few
general features and consequently consist of large numbers of nouns, many
of which also belong to other families. In contrast, the families 'Report',
'Prediction', 'Claim', 'Assessment', 'Disclosure' and 'Lie' are distin-
guished by more specific semantic constellations, the family 'Disclosure'
for example by the feature that the propositional content of the ORIGINAL
UTTERANCE is portrayed as previously unknown by the SPEAKER. As a
consequence of the semantic specificity of these families, the sets of nouns
that belong to them are fairly small.
These fundamental differences require that the expository strategy used
so far must be modified to a certain extent. In what follows I will first look
at the more unspecific families which consist of large numbers of nouns.
The discussions of uses of these families will be highly exemplary and only
a small selection of possible nouns will be named explicitly and exemplified.
To make up for the fact that most nouns are not discussed in detail, I will
provide a list of possible members, in which the features that can be attrib-
uted to several nouns are indicated. This procedure has the advantage that
polysemous meanings can be gleaned from the resulting chart (see Table
8.1). The specific families with fewer nouns will then be described in the
usual way.
'Statement'
Shared semantic features: [LINGUISTIC], [ASSERTIVE]
Frame: Linguistic; focus on illocution

Nouns: e.g. account, assertion, observation, statement
Roughly speaking, 'Statement' uses shell utterances simply as acts of stat-

ing or saying something, no less, no more. Their typical lexico-grammatical
environments are the patterns -that and th-N. Every now and then one
comes across uses of the pattern th-be-N, while -be-that uses are even
rarer. 'Statement' nouns were not found to occur with postnominal or com-
plementing infinitive clauses. As far as the scores for compiled reliance are
concerned, the noun assertion stands out from other nouns in the family
with a score of more than 48%. The collocation assertion + that-clause
alone accounts for almost 40% of the total uses of this noun. In terms of
absolute frequency, however, statement (which was found 868 times in the
pattern -that, 14 times in the pattern -be-that, 591 times in the pattern
th- and 58 times in the pattern th-be-N), surpasses assertion. Two typical
examples of uses in the most frequent patterns -that and th- are given in
(8.24) and (8.25):
(8.24) He could only point to Thomas's absence and the ageless observation
that football is a funny old game, (TODAY)
(8.25) 'Thanks. Aida, and I must say it's nice to see all of you again,' Nell
responded, and she smiled warmly at Fanny and Val, including them in
this statement, (BOOKS)
'Proclamation'
Shared semantic features [LINGUISTIC], [ASSERTIVE], [PUBLIC]

Nouns: e.g. announcement, decree, notification,
proclamation, statement
The 'Proclamation' nouns incorporate the feature [PUBLIC] in their mean-

ing. When speakers use them they portray assertions as official, often also
authoritative, speech acts. The nouns normally imply that the speech acts
156 Linguistic uses
they describe are not spontaneous, but both their contents and their extra-
linguistic setting are carefully planned before the speech act itself takes
place. Many nouns have a firm place in the registers of business, admini-
stration and law. Compare example (8.26):
(8.26) The encouraging news from the Jockey Club that records were broken
last season, both in the number of horses granted hunter certificates and
runners taking part in point-to-points, was dampened at the weekend
with the announcement that the RMC group is giving up its sponsor-
ship of the sport, (PAPERS)
The nouns typically occur in the pattern exemplified in (8.26), i.e. -that.
Uses in the pattern th- are rarer. Only the nouns accusation, announce-
ment and charge were found in the pattern N-e-cl. The same nouns pro-
duce a small number of uses in the pattern th-be-N.
'Argument'
Shared semantic features: [LINGUISTIC], [ASSERTIVE], [ARGUMENTA-
TIVE]
Nouns: e.g. affirmation, argument, concession,
counterclaim, justification, reaction
To a large extent, it is impossible to identify shared semantic features of the

uses in the 'Argument' family. The reason for this difficulty is that even
more than with other types of shell nouns, these uses are pragmatically de-
termined. Speakers' choices of any of these nouns almost completely depend
on the context in which they are embedded. The common ground of these
uses is that they portray utterances metaphorically speaking as aspects of
verbal warfare: as stages and steps, attacks and retreats, weapons and in-
struments, strategies and tactics relevant for and used in the pursuit of con-
versational aims.
Although the range of argumentative aspects and semantic shades that
can be involved is enormous, a few sets of nouns with similar meanings can
still be formed:
lllocutionary uses 157
a) "Defending" nouns, which characterize assertions as acts of defending

and safeguarding one's own position: affirmation, argument, defence,
denial, justification, qualification, reassurance.
b) "Aggressive" nouns, which characterize assertions as acts of attacking
the positions of other discourse participants: contradiction, counter-
claim, criticism, critique, objection, protest, provocation.
c) "Reactive" nouns, which characterize assertions as reactions to moves
by other discourse participants: answer, contradiction, counterclaim,
defence, denial, insistence, reaction, reply, reposte, response, retort.
d) "Appeasing" nouns, which characterize assertions as acts of accepting
the positions of other discourse participants: acceptance, acknow-
ledgement, admission, concession.
The noun position itself, which is obviously also part of the metaphorical
language of verbal warfare, only rarely occurs as a true assertive shell
noun. An example of such a use is given in (8.27).
(8.27) The position was that there was no evidence against either Pottle or
Randle to justify a prosecution, (PAPERS)
In most uses, however, a strong mental component focusing on a person's

opinions and beliefs is involved. In (8.28) below, this is expressed by the
verbal construction claiming that....
(8.28) Irl Engelhardt, chairman and chief executive of Peabody, explained his
position, claiming that the union had attempted to lay down conditions
for negotiations, and that Peabody was prepared to meet reasonable de-
mands for job security, (PAPERS)
One further set of 'Argument' nouns mainly describe specific types of rela-
tions of utterances to a discourse, viz. amendment, comment, conclusion,
contribution and explanation.
Like the nouns in the other two families of assertive uses, 'Argument'
nouns typically occur in the pattern -that and in the pattern th-N, while
they are much less frequently found in the pattern -be-that and in the pat-
tern th-be-N. One interesting exception to the rule that, just like proposi-
tional uses, assertive uses are only combined with /-clauses is a subset of
reactive nouns in the 'Argument' family. In contrast to all other assertive
nouns, the nouns answer, defence, reaction, reposte and response are also
158 Linguistic uses
used with infinitive clauses, but only in the pattern -be-to, i.e. with com-
plementing infinitive clauses following a copula, and not with postnominal
infinitive clauses. Two representative examples of such uses are given in
(8.29):
(8.29) (a) [The Hankeys were employed until the contracts were finished but
were left owing the bank Pounds 88,000.] As their overdraft had
been secured on their home in South Ferriby, the only answer was to
sell, (PAPERS)
(b) Some Senators have suggested that an appropriate response would
be to cancel plans for closer economic co-operation with the Soviet
Union. (BBC)
It would be wrong to treat these examples as instances of linguistic shell-

noun uses, although they include nouns which are often found as linguistic
shell nouns. Obviously, the shell contents do not represent utterances here
but activities, and even the use of answer and response does not result in an
upgrading of these second-order entities to third-order entities. A better
interpretation of the two examples is that the shell nouns are used in non-
linguistic senses. For response, as well as for defence, reaction and re-
poste, this is perfectly compatible with the 'basic' meanings of these nouns,
while the use of answer in reference to a non-linguistic reaction is best ex-
plained as a metaphorical transfer from the domain of language to human
behaviour in general. The uses of these nouns with infinitive clauses thus
belong to the class of eventive uses, which will be discussed in more detail
in Chapter 11 below.
A list of the nouns that were found to produce uses for the families 'State-
ment', 'Proclamation' and 'Argument' is given in the chart in Table 8.1
below. For each noun, some of the features that can be relevant in certain
uses are indicated. It must be emphasized that these feature lists are of
course not exhaustive, because many other more specific features also play
a role. They must not be mistaken for complete semantic descriptions of
lexemes but are only intended as rough indications of frequently recurring
usage types. What the chart in Table 8.1 shows, in addition to the usage
information on single nouns, is that the dimension which has the widest
distinctive application in this field is Leech's distinction between mainly
informative and mainly argumentative assertives.
Table 8.1 Nouns in the families 'Statement', 'Proclamation' and 'Argument'
Features/ inform, vs. public confi- defend- aggres-

dimensions argument. dent ing sive
Nouns
acceptance argument. appeas.
account inform.
accusation argument. public aggress.
acknowledge- argument. appeas.
ment
admission argument. appeas.
affirmation argument. confid. defend.
amendment argument. public
announcement inform. public
answer argument. react.
argument argument. defend.
assertion inform. confid.
charge argument. public aggress.
comment argument.
concession argument. appeas.
conclusion argument.
confession inform. public
confirmation argument. confid.
contradiction argument. aggress.
contribution argument.
counterclaim argument. confid. aggress.
criticism argument. aggress.
critique argument. public aggress.
declaration inform. public
decree inform. public
defence argument. public defend. react,
denial argument. public confid. defend. react.
description inform.
edict inform. public
explanation argument.
insistence argument. confid. defend.
justification argument. defend.
notice inform. public
notification inform. public
objection argument. aggress.
observation argument.
proclamation inform. public
160 Linguistic uses
Table 8.1 cont.

Features/ inform, vs. public confi- defend- aggres- react- appeas-
dimensions argument dent ing sive ing ing
pronounce- inform. public

ment
protest argument. aggress.
provocation argument. aggress.
qualification argument. defend.
reaction argument. react.
reassurance argument. defend.
remark argument.
reply argument. react.
reposte (ri~) argument. confid. react.
response argument. react.
retort argument. confid. defend. react.
ruling inform. public
statement inform. public
testimony inform. public
version argument.
I will now turn to families of assertive uses which are semantically more
specific than the three discussed so far and consist of much smaller numbers
of nouns.
'Report'
Shared semantic features: [LINGUISTIC], [ASSERTIVE], [RETRODICTIVE]

Nouns: Report, story, explanation, account, version,
tale
The uses subsumed in the 'Report' family share the specific feature that the
propositional content of the reported utterance represents a previous event.
Only the six nouns given above occurred frequently enough in this usage to
make it into the material. An example is given in (8.30):
(8.30) The former Nazi collaborator Paul Touvier appeared in court yesterday
to disprove a report that he had fled France to avoid trial for crimes
against humanity, (PAPERS)
It seems that assertive nouns that have this explicit [RETRODICITVE] com-
ponent are fairly rare. Report is a very clear case, account, story and tale
also produce good examples of the family, while explanation and version
are not always retrodictive. Account and version share the idea that the
shelled retrodictive assertive speech act is just one out of a number of pos-
sible ways of reporting an event. In this respect, they are related to the uses
in the 'Assessment' family below. 'Report' uses typically appear in the
pattern th-N.
'Prediction'
Shared semantic features: [LINGUISTIC], [ASSERTIVE], [PREDICTIVE]
Nouns: Bet, prediction, forecast, betting, prophecy,
prognosis
The list of 'Prediction' nouns is also short. Although other nouns with
similar meanings exist of course, for example divination or prognostica-
tion, these are pompous inkhorn terms not in common usage, at least not as
shell nouns. In fact, even the nouns listed above are not particularly fre-
quent either, neither in shell-noun typical nor in other uses. According to the
corpus data their frequencies range from 13.5 occurrences per million
words for bet to 1.6 for prognosis. The highest score for compiled reliance
is reached by prediction (18.87%). Betting and prognosis were only found
in the patterns -be-that and th-N, not in N-cl and th-be-N. The four other
nouns were all found in the patterns -that and -be-that, bet also occurs
in the pattern -be-to.
While 'Report' nouns are used to shell utterances as statements about
past events, 'Prediction' uses are related to utterances about future events.
Since these are much more a matter of speculation and conjecture, a more
distinct mental component can be detected in all nouns in this family. (8.31)
is a typical example of bet, obviously taken from a racing context:
162 Linguistic uses
(8.31) The Champion Hurdle at the Festival meeting is the target for the super
tough Danoli this time round and he looks an interesting proposition at
6-1 in Corals' (Switch/Delta 0800 242232) early betting. - My bet is
that he'll go off a fraction of those odds on the day, (TODAY)
'Guess'
Shared semantic features: [LINGUISTIC], [ASSERTIVE], [TENTATIVE],
[WITHOUT PROOF]

Nouns: Claim, suggestion, hint, contention, guess,
allegation
'Guess' nouns are used when SPEAKERS want to impute to ORIGINAL

SPEAKERS that they made the shelled assertion in a tentative spirit and with-
out proper proof or even evidence for its truth. Of course this characteriza-
tion need not coincide with the ORIGINAL SPEAKER'S intentions at all. Even
a perfectly confident assertion about a state of affairs, which may have been
made by the ORIGINAL SPEAKER with all signs of strong commitment to the
truth of the assertion, can be shelled as a claim, contention or suggestion
by the SPEAKER of the utterance that contains the reporting shell-content
complex. An impression of how inconspicuous such a characterization can
be is provided in example (8.32):
(8.32) ... debate about Maastricht. The rest of Europe, they tell us, has not
matched up to us and would have avoided many of their own domestic
political embarrassments had they done so. There may once have been a
certain justice in this claim, [PAPERS]
Whoever the ORIGINAL SPEAKER, the source of the ORIGINAL UTTERANCE,

was - it rather sounds as if it was a political opponent of the SPEAKER - we
can be pretty sure that they did not intend what they said as a claim, but as
a strong statement about a state of affairs.
Attention must be drawn to the fact that the two characteristics of
'Guess' uses - imputed tentativeness and imputed lack of evidence - need
not always cooccur. Although it is plausible that speakers make assertions
in a tentative way when they do not have adequate evidence for the truth of
what they say, the two aspects can also be separated. In fact, the nouns
listed in the box above fall into two subgroups. Claim, contention and alle-
gation mainly allow SPEAKERS to introduce the insinuation that the
ORIGINAL SPEAKERS do not have the evidence required for their assertion.
With these nouns, the element of tentativeness is more implied than explic-
itly expressed. Suggestion, hint and guess, on the other hand, highlight the
feature [TENTATIVE]. Here, the feature [WITHOUT PROOF] is not more than
a tacit implication. Interestingly, the influence of the SPEAKER on the char-
acterization of a speech act seems to be much stronger in the nouns of the
first group (i.e. claim, contention, allegation), because lack of evidence is
felt to be a much more drastic imputation than tentativeness.
Like 'Prediction' uses, 'Guess' uses bring along a marked mental com-
ponent. Nouns that can be seen as close counterparts on the mental side are
such conceptual and creditive nouns as assumption, conjecture, feeling and
impression (see Section 9.3.1).
By far the two most frequent nouns in the 'Guess' family are claim (with
2,995 uses recorded in shell noun contexts) and suggestion (2,555 uses).
The latter noun also belongs to another family, the family named after it in
the group of directive uses (see below). All 'Guess' nouns can also muster
comparatively high scores for compiled reliance. Suggestion, for example,
has a score of more than 43%, which is mainly due to occurrences in the
collocation suggestion that. -that uses are also responsible for the com-
piled reliance score of 31.3% for allegation. Claim, contention and guess
also have scores for compiled reliance over 20%. Although I have made it
clear that one should not overestimate the significance of these crude statis-
tics (see Section 4.5), such high scores can be seen as an indication that
these nouns are to a large extent earmarked for usage as shell nouns in the
pattern -that. Four examples of the noun suggestion in this pattern were
already provided in (8.23) above. Anticipating what will be said below on
the semantic distinction between assertives and directives and the resulting
collocational restrictions, it is important to note here that the -clauses in
all four assertive examples in (8.23) are indicative.
'Assessment'
Shared semantic features: [LINGUISTIC], [ASSERTIVE], [VERDICTIVE]

Nouns: Judg(e)ment, assessment, verdict
164 Linguistic uses
This small group of nouns is used to shell assertive speech acts as incorpo-
rating a semantic element whose communicative force is akin to Austin's
class of verdie fives. By using them as shell nouns, speakers suggest that the
assertions they shell should perhaps not be judged with the categories 'true'
and 'false', which are applicable to typical, e.g. neutral, assertions, but with
considerations like 'correct', 'right', 'wrong' etc. (Austin 1962: 141). The
three nouns differ semantically. Verdict immediately conjures up images of
courtrooms, judges and juries, although, as example (8.33) illustrates, it is
found in other contexts as well.
(8.33) I WOULD [sic] disagree with the verdict that Mick Jaeger's brainchild.
National Music Day, was a 'flop', (PAPERS)
Judg(e)ment is of course used in legal contexts, too, but it occurs also with
moral connotations and in a more general sense, which is also conveyed by
assessment. More than verdict, the two latter nouns include strong implica-
tions as to the cognitive processes that precede acts of judging and assess-
ing.
As far as their collocational tendencies are concerned, the three nouns
are very similar. All were found most frequently in the pattern th-N, fol-
lowed by -that, N-be-that and finally th-be- uses.
'Disclosure'
Shared semantic features: [LINGUISTIC], [ASSERTIVE], [UNKNOWN CON-
TENT]

Nouns: Revelation, disclosure
The semantic structure of the two nouns in the 'Disclosure' family is highly
specific. They are used to shell utterances as assertive speech acts whose
propositional contents are characterized as being unknown to the audience
of the ORIGINAL SPEAKER at the time when he or she made the ORIGINAL
UTTERANCE.
(8.34) Then, with the revelation in 1953 that the remains were a fake, the
story became a real whodunnit, (PAPERS)
Although example (8.34) proves that the noun revelation is indeed used in
this function, it must be admitted that one comes across it more frequently
in a non-linguistic sense, which is described by the entry "informal to be
surprisingly good, enjoyable, or useful" in LDOCE3. This evaluative use
can be found particularly frequently with reference to artists and sportsper-
sons and their performances in media reports, as illustrated in the two ex-
amples in (8.35).
(8.35) (a) Nick Marmby has been a revelation this season. Bursting onto the
scene, chunky and with sore chins ... (MAGS)
(b) Colledge turned out to be a revelation. Just 23, not long out of the
Royal Scottish Academy of Music,... (PAPERS)
Perhaps as a corollary of its wider semantic applicability, revelation is

collocationally more versatile than disclosure. The former was found in all
four patterns, the latter only in the pattern -that and th-N.
'Lie'
Shared semantic features: [LINGUISTIC], [ASSERTIVE], [UNTRUE]
Nouns: Lie, pretext, distortion
Unveiling the essence of lies has been the object of two studies in the
framework of prototype semantics and cognitive linguistics. Translating
Coleman and Kay's (1981) results into my framework, they showed that
three aspects of assertions are responsible for their characterizations by
SPEAKERS as lies\ that the prepositional content of the assertion is not true,
that the ORIGINAL SPEAKER of the ORIGINAL UTTERANCE is a w a r e of the
fact that it is not true; and that the ORIGINAL SPEAKER has the intention of
deceiving his or her audience. As is well-known, only prototypical lies in-
volve all three aspects. The most important of them is the third, the intention
to deceive. In her account of the verbal category lie, Sweetser (1987) em-
phasizes that lying must be looked upon in the context of cultural models of
speaking.
At this point it must be stressed once again that it is the SPEAKER of the
UTTERANCE who characterizes a speech act as deliberately untrue and de-
166 Linguistic uses
ceiving when he or she shells it with a 'Lie' shell noun. The imputations
contained in these nouns are obviously very strong. Deceptive intentions are
not just imputed to the ORIGINAL SPEAKER by the use of lie itself, but even
more so by the use of pretext, a noun which implies in addition that the
ORIGINAL SPEAKER has made an assertion in order to hide his or her true
intentions or at least to divert the ORIGINAL HEARER'S attention from them.
In this respect, pretext is closely related to the noun pretence, which also
shells such deceptive acts, but of a non-linguistic nature. Distortion is more
neutral in this respect. It allows for the possibility that the ORIGINAL
SPEAKER made an untrue assertion inadvertently, without a deliberate inten-
tion to deceive.
Probably due to their strong offensive imputations, 'Lie' uses are not
frequent at all. While lie can be found in -that uses, these are fairly rare.
An example illustrating the colloquial collocation nail the lie that ... is
given in (8.36):
(8.36) Tories said the news "nailed the lie" that Britain was suffering alone.
(PAPERS)
Distortion was not found in the pattern -that, and only once in the pattern
-be-that, but it does occur as a shell noun more frequently in the patterns
th- and th-be-N, and so it was included in the data. Pretext occurs com-
paratively frequently in the pattern -that. Th- and th-be- uses are more
common with lie than with pretext. Especially in spoken English, such pat-
tern th-be- expressions as that's a lie or that's a bloody lie are not un-
common. It is perhaps in this pattern that the fact that it is the SPEAKER
alone who imputes the lie to the ORIGINAL SPEAKER is most noticeable,
since with illocutionary nouns, the pattern th-be-~N largely has the function
of explicating the SPEAKER'S view of what kind of illocutionary act the
ORIGINAL SPEAKER has performed.
8.3.2 Rogative uses
'Question'
Shared semantic features: [LINGUISTIC], [ROGATIVE]
Nouns: Question, query
If it was not for a dozen and a half uses of the noun query in various pat-
terns that were found in the corpus, the noun question would not only be the
donor of the name of the only family of rogative uses, but also the sole
member. All other near-synonyms of question suggested by thesauri, e.g.
inquiry, interrogation, inquest, inquisition, were not found in common
usage as shell nouns.
Like assertive nouns, the prime rogative noun question can occur with
postnominal -clauses. What is more, these collocations even appear to
be assertive uses of linguistic shell nouns as well, witness examples (8.37)
and (8.38):
(8.37) ... animals like the elephant and the black rhinoceros. There is no ques-
tion that the animals are disappearing, (BOOKS)
(8.38) I rejected this firmly: it is beyond question that they must behave profes-
sionally and use reasonable care, (PAPERS)
At first glance, (8.37) seems to be an assertive use of question shelling the

statement that the animals are disappearing. (8.38) can be understood
analogously. However, as argued in Section 7.2, it is more convincing if
such uses are analysed as factual rather than linguistic uses of question
because the shell contents represent states of affairs and not utterances.
In analogy to the idea that assertive shell nouns report statements, roga-
tive shell-noun uses can be seen as means of reporting questions. Besides
the wA-clauses in postnominal and complementing function to be discussed
below, o/-prepositional phrases are of eminent importance as postmodifiers
of the noun question. However, the status of these uses as linguistic shell
nouns is also not always beyond doubt. An idea of the variety of preposi-
tional complements that are found is given in the short examples collected in
(8.39). They range in complexity from finite (a) and non-finite w/z-clauses
(b) over zwg-clauses with (c) or without a notional subject (d) to complex
noun phrases (e) and noun phrases consisting of an abstract head noun only
(f).
(8.39) (a) ... it would only renew the question of what he had been doing in her
... (BOOKS)
(b) And there was also the question of how to cook the roast beef ...
(PAPERS)
(c) ... there can be no question of your being pregnant... (BOOKS)
(d) Then it's a question of blowing dry on a low setting ... (MAGS)
168 Linguistic uses
(e) However, the question of the relation of the state to ... (PAPERS)
(f) Once again his mind turned to the question of integrity ... (BOOKS)
The plausibility of the idea that rogative shell nouns report questions de-
creases as we go through this list of examples. While for examples (8.39a)
and (b), the questions what had he been doing in her... ? and how does one
cook the roast beef? are conceivable as ORIGINAL UTTERANCE, similar
expansions are much less convincing for the less elaborated shell contents
further down the list. Mainly due to the occurrence of the determiner no, the
o/prepositional phrase in (8.39c) is better paraphrased by the factual that-
clause that you are pregnant than by an indirect question. The whole ex-
pression in (8.39d) has a modal meaning which can roughly be paraphrased
by all one has to do is blow dry .... Finally, the two nominal prepositional
complements in (8.39e) and (f) are related to w/-questions in a way (e.g.
what is integrity? or in what way does integrity play a role? for 8.39f), but
due to their brevity and abstractness, many other interpretations are also
possible. In any case a distinct mental component is involved here. What
this brief discussion of uses of question with o/-prepositional phrases has
shown, then, is that this highly versatile and frequent shell noun is not only
used as a linguistic, but also as a factual and mental shell noun (see Sec-
tions 7.2 and 9.3.2).
The typical complementizers for genuinely rogative shell-noun uses of
question and query are of course not tfwrt-clauses, but indirect questions
introduced by w/?-words. As mentioned earlier, the capacity of a speech act
verb to collocate with this type of complement is Leech's main criterion for
his class of rogative verbs, and I have followed him here with respect to
speech act nouns. The indirect questions can be linked to the noun directly
as appositive postmodifying clauses or by means of appositive of-
constructions (see Section 3.1.1). The corpus data clearly shows that com-
binations of the preposition of with a w/z-clause as prepositional comple-
ment, i.e. expressions of the type the question of how/what/when/whether
etc., are more frequent than w/i-clauses functioning directly as appositional
postmodifiers or complements {the question how/whatwhether etc.). Genu-
ine N-w/2 examples, i.e. uses of question with postnominal w/j-clauses
without the preposition of are in fact fairly rare. Their occurrence is
largely restricted to direct quotations of questions, as in (8.40).
(8.40) The best answer to the question "How will it play in '92?" should not be
"Not very much", but "Don't be small-minded", (ECON)
Illociitionary uses 169
This observation suggests an explanation for the scarcity of directly post-

nominal wA-clauses altogether. The relation of experiental identity, which
must be evoked to guarantee the co-interpretation of the shell noun and the
shell content (see Section 3.1.2), seems to be too weak to link the noun
question to questions as contents, unless these are explicitly marked as
questions. One way of marking the identity relation is to use appositive of-
constructions (Quirk et al. 1985: 1321), and the other is to signal the fact
that the shell content actually is a question by using the subject-operator
inversion typical of direct questions.
In -be-wh uses, indirect w/i-questions can be found with subject-
operator inversion (as in 8.41a) and without (as in 8.41b):
(8.41) (a) The question is, what should governments do about it? (PAPERS)
(b) The other question is what the Soviet Union should actually ...
(PAPERS)
Apparently, the author or editor of (8.41a) felt that a comma was appropri-
ate between the copula and the complementing wA-clause. This comma can
be interpreted as a signal of a direct quotation with omitted quotation marks
(see Quirk et al. 1985: 1024, note e).
By far the most frequent pattern in which the noun question was found
as rogative shell noun is the pattern th-N. A typical example with a short
distance between an interrogative sentence and the anaphoric shell-noun
phrase that question is given in (8.42):
(8.42) But what is optimum nutrition? This whole book is about that question.
(BOOKS)
Th-N uses of question are rarely premodified by adjective phrases. This is

completely different with th-be-H uses, where especially evaluative and
cohesive adjectives are found frequently in company with the noun ques-
tion. Since this type of usage, especially the collocation that's a + Adj +
question, is particularly frequent in spoken English, it will be interesting to
round off the account of rogative uses of question with a look at the adjec-
tives that are found in this particular section of my data. The adjectives
which occurred in the 70 instances of the pattern in the S P O K E N data are
listed with their frequency in Figure 8.4.
170 Linguistic uses
Evaluative (and descriptive) adjectives:

That's a good (25) question.
difficult (5)
interesting (4)
big (3)
daft (2)
hard (2)
obvious (2)
stupid (2)
valid (2)
crunch, dumb, eternal, fair, funny, general, huge, intriguing, intrusive, major,
real, rotten, terrible, tricky, wonderful (all 1)
Cohesive adjectives:
That's the next (5) question.
other (2)
different (1)
Figure 8.4 Adjectives occurring in that's a + Adj + question in the SPOKEN

corpus data
8.3.3 Directive uses
Directive uses share the common feature that SPEAKERS characterize

UTTERANCES as speech acts which were made by ORIGINAL SPEAKERS with
the communicative intention of getting their ORIGINAL HEARERS (and some-
times other people including themselves as well) to do something (Searle
1969: 66, Leech 1983: 206). Following Leech, I have divided directive uses
of shell nouns into four families: 'Order', 'Invitation', 'Advice' and 'Sug-
gestion'. The main dimensions along which these families differ are the
intensity of the directive, the attitudes imputed to the ORIGINAL SPEAKER by
the SPEAKER, especially the motives for the directive, and the group of per-
sons who are targets of the directive.
Directive uses (as well as the commissive uses to be discussed below)
are treated as linguistic shell nouns on a par with assertive, rogative and
lllocutionary uses 171
propositional uses of linguistic shell nouns in this study. In an assertive use

like (8.43a), the shell content can be traced back to a. declarative (see
8.43b), in rogative uses to an interrogative ORIGINAL UTTERANCE (see 8.44a
and b). And in directive uses, as in (8.45a), the shell content is related to an
imperative ORIGINAL UTTERANCE (see 8.45b).
(8.43) a) ... his declaration that the programme was now finished ... (PAPERS)
b) - ORIGINAL UTTERANCE: "the programme is now finished"
(8.44) a) the question whether we are detecting global warming in general ...
(BBC)
b) ORIGINAL UTTERANCE: "are we detecting global warming?"
(8.45) a) ... a demand that the dog should not be put down ... (PAPERS)
b) - ORIGINAL UTTERANCE: "do not put down the dog"
Natural and straightforward as this analogy to assertive and rogative uses

may sound, it cannot be taken for granted, as there are fundamental differ-
ences between propositional, assertive and rogative uses on the one hand,
and directive and commissive uses on the other. The major difference can
best be explained in Halliday's Systemic-Functional Grammar. Transferred
to this framework, propositional, assertive and rogative uses of shell nouns
shell propositions, while directive and commissive uses shell proposals.
According to Halliday (1994: 68-71), propositions are related to the ex-
change of information, while proposals concern the exchange of non-
linguistic "goods-&-services" (1994: 71). Leech (1983: 114-115) has a
similar distinction between propositions and mands, and there is of course
also a parallel to Lyons' distinction between second-order and third-order
entities or to my own between events and abstract relations (see Section
5.1.1).
Leaving aside rogatives and w/2-clauses, the pragmatic difference be-
tween propositional and assertive uses on the one hand, and directive and
commissive ones on the other is reflected in their collocational behaviour.
Propositional and assertive shell nouns occur only with postnominal and
complementing that-clauscs, not with infinitive clauses. These clauses have
indicative verbs. Directive and commissive shell-noun uses can occur with
f/wrt-clauses and infinitive clauses. However, when they are combined with
that-clauses, these are invariably subjunctive or accompanied by so-called
putative should (see Quirk et al. 1985: 234-235 et passim).
The source of these collocational restrictions can be found in the mean-
ing of the various shell nouns (Leech 1983: 206). Directive and commissive
172 Linguistic uses
shell nouns refer to speech acts whose prepositional contents are related to
future activities. In terms of Wierzbicka's description of the meaning of
complements (see Section 3.2,), these prepositional contents reflect subjec-
tive states of 'wanting'. This allows for infinitive clauses to occur with them
but rules out indicative -clauses, which tend to convey objective and
factual meanings (Leech 1987: 114). If that-clauses are used in such con-
texts, the subjunctive, or more often, particularly in British English, puta-
tive should, are used as indicators that an act will or should be brought
about. Prepositional and assertive uses, on the other hand, report objective
factual linguistic information. They are clear manifestations of Wierz-
bicka's semantic primitive 'say' (1988: 163). Indicative i/iaf-clauses are
therefore possible postmodifiers and complements. Infinitive clauses, on the
other hand, are ruled out because propositions rather than (non-linguistic)
activities occur as shell contents.
'Order'
Shared semantic features: [LINGUISTIC], [DIRECTIVE]

Nouns: Demand, call, request, order, motion, in-
junction, instruction, command, directive
Order' uses shell the classic case of directive speech acts, the situation that
a speaker performs a speech act with the intention of urging the hearer to do
something. Information as to possible motives for the intention is not in-
cluded in the meaning of these nouns.
An example of an 'Order' use in which the noun demand is followed by
a that-clause is given in (8.46).
(8.46) ... the government side would not agree to the demonstrators' demand
that cameras be allowed into the meeting to film the event, (BBC)
Postnominal /za-clauses were also found with all other nouns listed above,
except call and command. Postnominal infinitive clauses occur less fre-
quently than -clauses with the noun demand, but much more often than
/-clauses with the nouns injunction, instruction and request. Directive
was not found in the pattern N-to at all. A quantitative analysis of the collo-
cations order to and order that was not carried out because of the interfer-
enee by the complex conjunctions in order to and in order that. The con-
cordances prove, however, that this noun is used with both types of postno-
minal clauses.
Instances of the pattern N-e-cl are not frequent at all in the corpus,
neither with complementing tfjai-clauses nor with to-clauses. Call, com-
mand, demand, injunction, instruction and motion were not found in the
pattern N-e-cl at all, directive only in the pattern -be-to, while order and
request are used in the patterns -be-to and -be-that. All nine nouns are
found frequently in the pattern th- and occasionally in the pattern th-be-N.
The nine nouns in the 'Order' family are of course not on a par semanti-
cally and stylistically.63 Instruction is probably the most neutral directive
shell noun. Order and command are stylistically neutral as well, but both
imply that the ORIGINAL SPEAKER is in a position of authority over the per-
sons to whom the speech act is directed (cf. Searle 1969: 66). In addition,
both are used by speakers to indicate that the ORIGINAL SPEAKER has a
strong desire that the directive is complied with. Albeit to a lesser extent,
such a characterization of the intensity of the underlying attitude is also
shared by the nouns request, call and especially demand. However, when
these nouns are used, moral aspects are portrayed as being the justification
of a directive, rather than aspects pertaining to authority or social hierar-
chies in general. The nouns request, call and demand are better suited to
shelling directives from ORIGINAL SPEAKERS who are authoritatively and
socially not necessarily inferior but certainly not superior to their addressees
(Leech 1983: 220). In this respect they bridge the gap to the 'Invitation'
family discussed below. Directive and injunction are stylistically formal
directive shell nouns and are earmarked for special types of ORIGINAL
SPEAKERS: directive is used when a directive is issued by an official source,
and injunction is typically used for a specific type of directive issued by
legal courts forbidding the addressee to do something. It can occasionally be
found in a more general sense as an 'Order' shell noun, too. Motion, finally,
is another highly specialized shell noun, earmarked for proposals made at
meetings and later voted on. It is related to the 'Suggestion' family below.
'Invitation'
Nouns: Invitation, appeal, application, plea, petition
174 Linguistic uses
The 'Invitation' nouns can be regarded as the weaker branch of the 'Order'
family. They make up the continuation of request, demand and call on the
lower end of the cline of directive force. 'Invitation' nouns characterize a
speech act more as an appeal to the ORIGINAL HEARER'S willingness to do
something, than as a strict order based on authority and compulsion (Leech
1983: 217, Wierzbicka 1987: 82). This semantic facet is of course most
distinct in the noun invitation itself, which even implies that the act to be
performed is beneficial for the ORIGINAL HEARER - unless, of course, it is
used ironically or euphemistically for a strict order. With the other nouns
the weakness of the directive seems to be attributable to the fact that the
ORIGINAL SPEAKER is portrayed as having a personal rather than objectively
justifiable motive for his or her appeal. In terms of Leech's optionality or
conditionality scale (1983: 216, 219), 'Order' uses characterize a directive
as unconditional, i.e. take the ORIGINAL HEARER'S compliance for granted,
while 'Invitation' uses are conditional in that they portray the success of the
directive as being dependent on the ORIGINAL HEARER'S agreement.
The typical clausal shell-noun collocation of all five nouns is N-/o. This
pattern is illustrated in (8.47)
(8.47) Mr Gorbachov had to work hard to get his invitation to address the
Group of Seven leaders, (BBC)
Only appeal and plea occur with postnominal and complementing that-
clauses.
'Advice'

Nouns: Advice, recommendation, tip
'Advice' uses share a semantic component expressing the motive for a di-
rective. The nouns are used when the SPEAKER wishes to characterize the
ORIGINAL SPEAKER'S directive as being brought about by the belief that it is
in the interest of the ORIGINAL HEARER to carry out an activity. They are
used to provide shells conveying something like 'S tells H to do X because
s/he thinks that it is good for him/her'. See (8.48) for an illustration.
lllocutonary uses 175
(8.48) [But don't forget to inform your waitress that you don't want eggs in
your dishes. You'll never miss them in your sukiyaki.] The only other
advice is to avoid the soya-sauce container on the table, (BOOKS)
It should be noted that this interpretation of the illocutionary act of advising

is not entirely in line with Searle's claim that
Contrary to what one might suppose advice is not a species of requesting.

[...] Advising you is not trying to get you to do something in the sense that
requesting is. Advising is more like telling you what is best for you.
(Searle 1969: 67)
This passage, especially the last sentence, nicely illustrates that advising
involves a mixture of ordering and informing. The noun advice can thus
shell directive and assertive speech acts at the same time. Depending on the
context, one of the two facets can be more important than the other. This is
in line with the more general idea that many manifestations of directive
speech acts include an assertive component. Interestingly, Searle's descrip-
tion of advices emphasizes the assertive component, and so does Wierz-
bicka's (1987: 181-190), while Leech (1983: 217) attributes more impor-
tance to the directive aspect of advising.
The three nouns in this family are semantically very similar. This resem-
blance is paralleled by a large degree of collocational homogeneity, even
though the nouns are collocationally very versatile. All three are used in N-
cl, N-e-cl, th- and th-be- patterns, advice and recommendation with
both finite and non-finite postmodifiers and complements, tip only with
postnominal infinitives but not with tfiai-clauses. Advice and recommenda-
tion differ with regard to pattern N-cl uses in that the former is more fre-
quent with postnominal infinitive clauses and the latter with /W-clauses.
By far the most frequent noun is advice, followed by recommendation and
finally tip. Tip is stylistically less formal than the other two nouns.
'Suggestion'
Nouns: Suggestion, proposal, proposition
176 Linguistic uses
Conflating Leech's (1983: 217) and Halliday's (1994: 257) ideas on what
kind of speech functions suggestions serve, I see 'Suggestion' nouns as
sharing the following semantic and pragmatic setup. Their use shells direc-
tives as being motivated by the belief that it will be good for the ORIGINAL
HEARER to carry out an act X, and as being accompanied by a weak com-
mitment by the ORIGINAL SPEAKER to cooperate in carrying out X.64 'Sug-
gestion' uses can thus be seen as bridging the gap between 'Advice' uses
and the commissive 'Offer' and 'Promise' uses.
Suggestion and proposal are collocationally highly versatile. They occur
with that-clauses and infinitives in N-cl and N-e-cl patterns, and of course
in patterns th- and th-be- as well. Proposition was only found with that-
clauses in the patterns N-cl and N-e-cl, not with infinitive clauses.
As already hinted at in the discussion of 'Claim' uses (see Section
8.3.2), assertive and directive uses of the noun suggestion can be distin-
guished on the basis of the complements: assertive uses are complemented
by indicative -clauses (see the examples in (8.23)), while directive uses
are complemented by ^//-clauses with putative should (see 8.49a) or by
infinitive clauses (8.49b). Mandative subjunctives in /Aai-clauses following
suggestion were not found in the (British English) corpus:
(8.49) (a) ... response to John Major's suggestion that people should write into
Downing Street... (PAPERS)
(b) When Pearson took up Rees's suggestion to look back at what things
were really... (PAPERS)
Proposal and suggestion differ collocationally; the former is typically com-

bined with infinitive clauses, and the latter with //^/-clauses. The preference
for combining suggestion with -clauses can be due to its occurrences as
assertive noun. This does not explain the affinity of proposal and infinitive
clauses, however. This is yet another point where I would like to claim that
semantic differences give rise to collocational ones. More than suggestion,
which has a distinct assertive component, proposal is set aside for direc-
tives inducing non-linguistic actions.
8.3.4 Commissive uses
The difference between directive and commissive speech acts lies in the
question of who is to carry out an activity. The intended outcome of a di-
rective is that the hearer does something; that of a commissive is that the
speaker himself or herself is the agent of a future activity. In Searle's (1969:
56) framework, committing oneself to a future act is the propositional con-
tent condition of promises, the paragon of commissives.
'Offer'
Shared semantic features: [LINGUISTIC], [COMMISSIVE]

Nouns: Bid, offer
According to Leech (1983: 217), offers are conditional while promises are
unconditional. Translated into my framework this means that SPEAKERS use
'Offer' nouns when they feel that the ORIGINAL SPEAKER still has the option
of not doing something, depending on the ORIGINAL HEARER'S reaction to a
commissive utterance. The core of Wierzbicka's definition of the verb offer
is similar: "I say: I will cause X to happen if you say you want me to do it"
(1987: 191). In contrast, speakers use 'Promise' nouns only when they feel
that the ORIGINAL SPEAKER must do the act shelled as UTTERANCE.
With frequencies in shell-noun typical collocations of 3,458 and 1,705
respectively, bid and offer are among the more frequent shell nouns. With
regard to bid, there can be no doubt that this remarkable frequency is an
artefact of one feature of the COBUILD corpus - the preponderance of media
sources (see Section 4.2). Especially the authors of sport reports but also,
and perhaps more justified by their subject-matter, the authors of business
reports seem to have taken a fancy to the collocation bid + to-clause. The
bulk of these examples belong to the class of event uses (see Chapter 11)
rather than linguistic uses, because they shell activities rather than utter-
ances referring to future activities. Such non-linguistic uses also outweigh
linguistic ones in the case of offer, but for this noun, examples of uses with
a possible linguistic basis can also be found. One of them is given in (8.50):
(8.50) He rejected roles in such blockbusters as Die Hard, Alien 3 and Patriot
Games and turned down a million-dollar offer to play the villain in
Lethal Weapon 3 in order to concentrate on his own production, (MAGS)
Here an ORIGINAL UTTERANCE "We will let you play the villain in ..." is at
least conceivable. All in all, offer especially, but also bid, can be seen as
178 Linguistic uses
cases of nouns with a potential linguistic meaning which have reached such
an extreme position on the scale from assertiveness to directiveness that
they have virtually lost their asserting potential and are used to shell pure
(non-linguistic) proposals (in Halliday's sense). Given this extreme status, it
is almost needless to say that neither noun occurs with postnominal or com-
plementing that-clauses but only with infinitives.
'Promise'

Nouns: Promise, commitment, pledge, guarantee,
assurance, undertaking, vow, oath
With the exception of oath, the nouns in the 'Promise' family all boast
fairly high scores for compiled reliance. The score of commitment is
11.65%, and those of the other six nouns are all higher than 20%, that for
pledge being even higher than 30%.
The four highest scorers pledge, vow, promise and undertaking, on
which I will focus my attention, exhibit a homogeneous collocational range
as shell nouns. Without exception, postnominal infinitive clauses expressing
the future act promised by the ORIGINAL SPEAKER make up the most fre-
quent shell-noun environment (see example 8.51).
(8.51) The new Home Secretary has welshed on his predecessor's promise to
provide a police authority for London, (PAPERS)
These are followed by th-N and -that uses, and finally th-be-N uses. None
of the nouns was found in the pattern -be-that, and only very small num-
bers of instances of -be-to uses were found for promise and pledge (e.g.
Chief Abiola's most extravagant promise is to abolish poverty, ECON).
Like the O f f e r ' nouns, 'Promise' nouns shell speech acts as expressing
the ORIGINAL SPEAKERS' willingness to do something, and also their inten-
tion. However, as mentioned above, 'Promise' nouns are unconditional
whereas 'Offer' nouns are conditional. When SPEAKERS shell a speech act
as a promise, they take for granted that the ORIGINAL SPEAKER knew that
the ORIGINAL HEARER would want him or her to perform the act referred to
as propositional content. As a consequence, the element of contingency
included in O f f e r ' nouns, as to whether the ORIGINAL HEARER will accept

an offer or not, disappears in 'Promise' nouns.
According to Wierzbicka (1987: 205-213), the verbs promise and assure
share the aspect that they are more addressee-oriented than vow and pledge.
The same is true of the morphologically related nouns. If a speech act is
shelled as a vow or pledge, it is construed by the SPEAKER as an act by the
ORIGINAL SPEAKER of imposing an obligation on himself or herself. While
this element undoubtedly also plays a role when promise and assurance are
used, it is superimposed communicatively by the wish to assure the
ORIGINAL HEARER that an act will indeed be performed.
'Threat'

Nouns: Warning, threat, caveat
Although they have similar meanings on an intuitive level, the three nouns in
the 'Threat' family make up a fairly mixed bag. Caveat is a very rare noun.
Only 200 instances are recorded in the whole 225 million word corpus. It is
a stylistically formal noun used to denote speech acts which are made by the
ORIGINAL SPEAKER with the intention of drawing the ORIGINAL HEARERS'
attention to something they should keep in mind or take heed of.
Warning and threat, on the other hand, are frequent shell nouns. Warn-
ing is related to the 'Advice' nouns but differs in that it imputes the belief to
the ORIGINAL SPEAKER that it would be detrimental for the ORIGINAL
HEARER(S) to perform some non-linguistic act. It has a strong assertive
component as well (see also Leech 1983: 208), especially when it is used to
shell references to possible dangers, as in example (8.52):
(8.52) The Association will give a warning that poll tax bills in some Conser-
vative districts will exceed government guidelines unless they receive a
higher proportion of the money being made available to local authori-
ties. (BBC)
In addition to the directive and the assertive component, a commissive as-

pect plays a role, though an untypical one in that the ORIGINAL SPEAKER is
180 Linguistic uses
portrayed as predicting that some future event will happen rather than
committing himself to causing this event.
The commissive element is more pronounced in the noun threat itself,
and this explains the classification of the 'Threat' family as giving rise to
commissive uses. In combinations with postnominal infinitive clauses,
which by far make up the majority of its uses, threat includes a reference to
a commitment by the ORIGINAL SPEAKER to do something, often, but not
necessarily (see Wierzbicka 1987: 178-179), under the condition that the
ORIGINAL HEARER does not do something. That the condition is not a neces-
sary part of the meaning of threat emerges from example (8.53), where no
contingencies are mentioned:
(8.53) Of the two chief whaling nations, Norway is adamant in its threat to
hunt the migratory minke whale ... in the northern Atlantic, (MAGS)
The main aspect here is simply "S will do X" and this clearly boils down to
a commissive speech act. Even Leech (1983: 217), who on the one hand
emphasizes the directive force of threats by giving "insistence that h do X"
as the attitude imputed to the ORIGINAL SPEAKER by the SPEAKER, also
classifies the verb threaten as commissive.
In the rarer cases where threat is used with a postnominal that-clause, as
in (8.54) below, it does not function as a linguistic but as a modal shell
noun:
(8.54) One army under Burgoyne was to march south from Canada whilst
Howe's troops were to march north up the Hudson valley. Germain as-
sumed that the threat that the two might meet near Albany would force
the Continental Army to give battle to prevent the British from isolating
New England, (BOOKS)
It is impossible to come up with an ORIGINAL UTTERANCE for this example,

because the shell content does not represent an utterance, but a possible
event. Threat is more or less identical with danger in this function (see
Section 10.2). As we have seen above (see example 8.50), warning does
occur as linguistic shell noun with postnominal ^-clauses, presumably
because its meaning includes a greater proportion of assertiveness than that
of threat.
8.3.5 Expressive uses
Leech (1983: 210) has made it quite clear that the borderline between as-
sertive and expressive speech act verbs is very thin or, to cast it into a dif-
ferent image, that many verbs straddle the boundary between asserting
propositions and expressing feelings. Having discussed examples of the
verbs complain and reproach, he concludes that
complain, in fact, is one of a group of verbs which appear to be a mixture

of the assertive and expressive categories: others are boast, rejoice, grum-
ble and lament. When these verbs preceded [sic] a that-cisass, it seems
best to analyse them as having the sense of 'to assert in a VERBing man-
ner', where 'VERB' is the corresponding expressive sense of the verb.
(Leech 1983: 210)
The recognition that what appear to be reports of expressive speech acts are
in fact often reports of assertives is also relevant for nominal speech act
reports. An example of the noun complaint is given in (8.55) to substantiate
this claim:
(8.55) Visa's complaint that the report failed to deal adequately with certain
questions was no basis for invalidating its findings, (PAPERS)
In line with Leech's suggestion, the shell-content complex in (8.53) can be

analysed as a nominal equivalent of the verbal expression Visa asserted in a
complaining manner that....
Verbs like complain or boast do occur as genuine expressives when they
are combined with gerunds, as in Leech's example She reproached him for
eating too much. In such cases, the subordinate clause is presupposed as
being true rather than asserted, as can be shown by the fact that its truth is
preserved even when the main clause is negated: She did not reproach him
for eating too much. Unfortunately it is impossible to apply this test to
potentially expressive shell nouns. The data suggests, though, that the nouns
that I discuss as expressive uses in this section are primarily used to shell
speech acts as assertives, while the expressive component is simply added
as an extra piece of descriptive information.
Although there are not many nouns with seemingly expressive uses, it
makes sense to divide them into two families. The 'Complaint' family con-
sists of nouns which are more or less synonymous with the noun complaint.
182 Linguistic uses
The second family, called 'Compliment', is a collection of assertive nouns

with a variety of added expressive components.
'Complaint'
Shared semantic features: [LINGUISTIC], [ASSERTIVE], [EXPRESSIVE],
[NOT SATISFIED WITH X]

Nouns: Complaint, grievance, gripe, quibble, whinge,
grumble
These six nouns share the component that dissatisfaction or even anger are
portrayed as reasons for the ORIGINAL SPEAKER'S assertive speech act. As
example (8.55) above has already indicated, the shell contents of such uses
represent statements about states of affairs or events that have caused the
dissatisfaction.
Only complaint itself occurs with postnominal that-clauses. The five
other nouns, which admittedly are all fairly rare, were only found in the
pattern -be-that and some of them also in the pattern th-N. According to
Wierzbicka (1991: 180-183), whinge is particularly common in Australian
English.
'Compliment'
Shared semantic features: [LINGUISTIC], [ASSERTIVE], [EXPRESSIVE]

Nouns: Excuse, boast, compliment, praise, lament
It must be emphasized that unlike in other families, the nouns collected in

this family do not share a common meaning in addition to the feature
[EXPRESSIVE], Transferring Leech's claims as to the attitudes implicated by
the corresponding verbs to the nominal domain, the five nouns in the 'Com-
pliment' family can be characterized as follows. SPEAKERS use the shells:
boast to impute to the ORIGINAL SPEAKER exultation,

excuse to impute to the ORIGINAL SPEAKER regret,
lament to impute to the ORIGINAL SPEAKER sorrow,

compliment, praise to impute to the ORIGINAL SPEAKER admiration
over an event or state of affairs.
More finely-grained and less abstract semantic descriptions of the related

verbs can be found in Wierzbicka (1987).
Chapter 9
Mental uses
9.1 Introduction
Linguistic and mental shell-noun uses have a lot in common. The most fun-
damental parallel lies in their predominant function: linguistic shell-nouns
can be used to report UTTERANCES, and mental shell nouns can be used to
report IDEAS. However, since speakers only have direct access to their own
ideas, their reports on the thoughts of other people also depend on language.
As pointed out in Section 8.1, we normally only learn about other people's
thoughts and ideas when they tell us about them. The difference between the
two kinds of reports is that ORIGINAL IDEAS, the counterparts of ORIGINAL
UTTERANCES in the mental domain, are not conceived of as something that
is communicated, but as mental entities locked up in the mind or floating
around in conceptual space. As a consequence, there is no counterpart to
ORIGINAL HEARERS in the mental domain. The frame proposed for the de-
scription of mental shell-noun uses in Figure 9.1 is therefore almost, but not
completely, analogous to that of linguistic uses.
(= PSYCHOLOGICAL STATE)
expenences
EXPERIENCER ORIGINAL
IDEA links < SPEAKER
IDEA
Figure 9.1 General frame for the description of mental shell-noun uses
Introduction 185
(Note that I will refrain from using small capitals for the frame component
SPEAKER from now on because, as with factual uses, there is no danger of
confusion between SPEAKERS and ORIGINAL SPEAKERS in the mental do-
main.)
The correspondence between the linguistic and the mental domain can be
taken several steps further. To begin with, the distinction between proposi-
tional and illocutionary uses can be transferred to the mental domain. Men-
tal shell nouns which, in analogy to propositional linguistic shell nouns,
shell the propositional contents of IDEAS will be called conceptual shell
nouns. These are discussed in Section 9.2. In analogy to illocutionary uses
in the linguistic domain, there are mental shell nouns that allow speakers to
characterize the attitudes that they impute to EXPERIENCERS. Mental shell
nouns of this type characterize PSYCHOLOGICAL STATES and processes like
believing, wishing or doubting. I will use the term psychological-state uses
as a cover term for this second type of mental shell-noun uses, which are
discussed in Section 9.3.
Leech has demonstrated for illocutionary and psychological verbs that
the analogy between the linguistic and the mental domains goes even fur-
ther. His basic recognition is that
'speech-reporting' and 'thought-reporting' functions are logically parallel

in that they deal with the same kind of oratio obliqua.
(Leech 1983: 212)
For example, we tend to think that speakers who make statements are also
in the mental state of believing what they say. Similarly, we take it for
granted that speakers who tell the addressee to do something are in the
mental state of wishing that the addressee do it. These assumptions con-
cerning psychological states are so plausible that they are captured as "sin-
cerity conditions" in speech act theory (Searle 1969: 60-62). Transferring
Leech's (1983: 211) ideas about linguistic and psychological predicates to
the domain of shell nouns, we arrive at the correspondences laid out in Fig-
ure 9.2. As the figure shows, directive and commissive uses together corre-
spond to the psychological-state group of volitional uses. The reason for
this is that both directives and commissives are related to states of wanting
something, i.e. volitional states. They only differ with regard to the question
of whether ORIGINAL HEARERS or ORIGINAL SPEAKERS are envisaged as
agents of a future activity.
186 Mental uses
Linguistic domain: Mental domain:

Illocutionary uses Psychological-state uses
Group Examples Group Examples
Assertive statement, declaration Creditive belief, assumption
Rogative question, query Dubitative doubt, question
Directive order, command

Volitional wish, aim, plan
Commissive promise, offer
Expressive complaint, grievance Emotive65 fear, regret, surprise
Figure 9.2 The correspondence between illocutionary and psychological-state

uses
Even Halliday's (1994: 68-71) distinction between propositions and pro-

posals can be transferred to the mental domain (see also Section 8.3). Halli-
day himself (1994: 263-264) draws the parallel between embedded pro-
jected locutions and embedded projected ideas. In either domain, the pro-
jection of both propositions and proposals is possible. Applying Halliday's
framework to Leech's classification as rendered in Figure 9.2, one can say
that creditive and dubitative nouns shell propositions, and volitional nouns
shell proposals. In view of the discussion in Section 8.3, it should not come
as a surprise that the distinction between proposition-related and proposal-
related mental shell noun uses is again reflected in the type of clauses with
which the nouns in question typically combine: creditive and dubitative
nouns tend to be postmodified and complemented by /zo-clauses, and voli-
tional nouns by infinitive clauses.
Compared to linguistic means of reporting speech acts, those for report-
ing acts of thinking have not attracted a lot of attention from linguists. Phi-
losophers, on the other hand, have of course for a long time been interested
in the nature of ideas, notions and states of believing and knowing.66 Even
belief reports have been dealt with from that perspective, recently for exam-
ple in a book by Crimmins (1992). As befits a philosopher, Crimmins does
not present an analysis of the linguistic means of performing belief reports
but is mainly concerned with their truth-conditions and with ways of for-
malizing belief reports. However, Crimmins also includes pragmatic consid-
erations like appropriateness (1992: 1), as long as they are of philosophical
relevance. While it is impossible and unnecessary to review Crimmins'
Introduction 187
study here, it is reassuring that his philosophical perspective lends support

to two aspects of my own approach. Firstly, the ternary relation proposed in
the frame for mental shell-noun uses in Figure 9.1, consisting of
EXPERIENCED PSYCHOLOGICAL STATE a n d IDEA, receives s u p p o r t f r o m his
work. For Crimmins emphasizes that philosophical accounts of beliefs and
belief reports must not just be based on a binary relation between a believer
and a proposition, but have to include representations of different "ways of
believing" (1992: 3). Secondly, Crimmins' ideas confirm the claim, which is
represented in the frame by means of the two arrows, that the IDEA given as
a report by the speaker must not, or better never will, completely coincide
with the ORIGINAL IDEA experienced by the EXPERIENCER. As I have argued
at the outset of this chapter, this coincidence is even more doubtful than in
the linguistic domain because thoughts are private.
I have divided the class of mental shell-noun uses into five groups on the
basis of the analogy to the linguistic domain. The result is the categorization
summarized in Figure 9.3.
Groups (Type of psy- Families Examples Section

chological state)
Conceptual 'Idea' idea, theory, notion 9.2

uses 'Mystery' mystery, conundrum
Psycho- Creditive uses 'Belief belief, knowledge, feeling 9.3.1
logical 'View' view, opinion, thinking
state uses 'Realisation' realisation, discovery
'Illusion' illusion, fantasy, delusion
Dubitative uses 'Doubt' doubt, question 9.3.2
Volitional uses 'Aim' aim, goal, ambition 9.3.3
'Plan' plan, decision, policy
'Desire' desire, hope, wish, dream
'Reluctance' reluctance, unwillingness
'Purpose' purpose, function
'Solution' solution, key
'Motivation' motivation, incentive
'Determination' determination, courage
'Agreement' agreement, deal, contract
Emotive uses 'Surprise' surprise, regret, delight 9.3.4
'Fear' fear, worry, anxiety
Figure 9.3 Survey of groups and families of mental shell-noun uses

188 Mental uses
9.2 Conceptual uses
Together with most groups of factual uses (see Chapter 7) and propositional
linguistic uses (Section 8.2), conceptual uses form the prototypical core of
the whole class of shell nouns (see also Section 6.1 and the summary in
Chapter 13). Conceptual uses are shell nouns par excellence in almost all
respects: they are 'original' nouns, i.e. not morphologically derived from or
related to verbs or adjectives; they are the exclusive means of doing what
they are set to do, namely shelling events and abstract relations as IDEAS;
and they excel at projecting, since they turn physically observable events as
well as abstract relations into mentally perceivable DEAS. Their only deficit
is that not all of them evoke absolutely pure relations of experiential identity
to their shell contents (see Section 3.1.2).
Given this exemplary status (and their somewhat tautological name -
after all, all shell nouns are "conceptual" in a way), it is no suprise that
conceptual uses lend themselves particularly well to a description in terms
of the container-metaphor underlying the idea of shell nouns. Conceptual
shell nouns epitomize the image of conceptual shells which contain mental
entities. Examples like the thought that that man [Charles] will one day
be King (PAPERS), the old theory that you can tell a woman's true age by
looking at her hands (TODAY) and the idea of the socialist state (SPOKEN),
comply perfectly with this metaphor: the nouns theory, thought and idea
create conceptual shells, and what is inside these shells is readily expressed
in the adjacent phrases or clauses. The result of the connection between
shell noun and shell content is a temporary concept. While this is true of all
shell nouns by definition, the distinction of conceptual uses is that the con-
cepts they create are not just concepts in the metalinguistic sense in which
all shell nouns create concepts, but also from the perspective of the object-
language. With more or less marked additions, conceptual shell nouns do
not only create concepts, but they also mean concept.
In terms of the frame proposed in Figure 9.1, what conceptual uses do is
highlight the propositional content of IDEAS at the cost of backgrounding the
fact that ideas tend to be conceived and experienced by EXPERIENCERS. It
must be emphasized that this effect can also be obtained with many nouns
that have the potential to occur in psychological-state uses, especially with
non-derived nouns like goal, plan or programme, but with many others as
well, e.g. belief, view or aim (see Section 9.3). In contrast to conceptual
nouns, however, these nouns include references to the PSYCHOLOGICAL
Conceptual uses 189
STATES imputed to the EXPERIENCERS by the speaker, and therefore they

are discussed in Section 9.3.
'Idea'
Shared semantic features: [MENTAL], [CONCEPTUAL]
Frame: Mental; focus on prepositional content of IDEA
Nouns: Point, idea, position, issue, theory, notion,

thought, principle, rule, subject, image, myth,
law, theme, concept, secret, scenario, wisdom,
hypothesis, thesis, logic, topic, doctrine,
teaching, maxim, stereotype, rationale, meta-
phor, axiom, dogma, credo, motto, precept
The core of this family consists of nouns whose inherent meaning does not
go far beyond that of 'concept' or 'mental entity', e.g. point, position, idea,
issue, notion and concept. It is on these and the other more frequent ones
that I will focus my attention here.
The nouns point and position are a good starting-point because they are
familiar as suppliers of other uses (see Sections 7.2, 7.6, 7.7, 8.2 and
8.3.1). Conceptual uses of these nouns occur typically in the collocations
the point/position that and this point/position. Together with the highly
adaptable noun question, point and position mark the transition zone be-
tween prepositional linguistic and conceptual uses. For in many of their
uses with postnominal -clauses, it is literally impossible to decide
whether their shell contents represent UTTERANCES or IDEAS. Seven lines
from the concordance for the query "point+that/CONJ" taken from the
PAPERS section are given in (9.1) to demonstrate this:
(9.1)
(a) < so, I repeat my point that the Agency would be acting illegally
(b) < <t> You make the point that there are heavy penalties for >
(c) < underlined the point that this symphony marked the full >
(d) < to stress the point that for youngsters, poetry does not >
(e) < the important point that many of these men die from other >
(f) < Except for the point that he may be acting unconstitutionally,
(g) cmisses the entire point that cannibalism is and was practised to>
190 Mental uses
Although it would naturally help to have more context, even the short con-
cordance lines give an idea of the ambiguity at work here. The verbs repeat,
make, underline and stress in the first four lines suggest a linguistic read-
ing, though (c) and (d) by no means rule out a mental reading. The three
examples at the bottom of the list rather belong to the mental domain be-
cause the impersonal constructions in which they occur suggest a reading as
an apparently autonomous mental entity.
Point and position thus seem to feel equally comfortable in the linguistic
and the mental field, and this is why they were mentioned in the 'News' and
the 'Argument' families as well (see Sections 8.2 and 8.3.1). The same is
true of axiom, credo, doctrine, dogma, maxim, metaphor, motto, myth,
stereotype and teaching, which are all listed both here and in the 'Myth'
family in Section 8.2. The other nouns in the 'Idea' family that are similarly
ambiguous, essentially idea, issue, principle and rule, have been listed here
only, because they exhibit a preference for the mental domain.
I have already described the use and meaning of the noun idea in great
detail (Schmid 1993: 165-219, Schmid 1997), from both a synchronic and a
diachronic perspective. In Schmid (1993), I tried to give a unified account
of the meaning of the lexeme idea. This aim differed radically from the
present approach, which is guided by the assumption that nouns like idea
provide conceptual shells that can be adapted to different kinds of contex-
tual uses. Despite these theoretical differences, it is not necessary to say as
much about the noun idea here as its versatility and frequency would merit;
it would seem more reasonable to leave more room for nouns which have
received less attention.
Idea is one of the collocationally most versatile shell nouns. It can be
found in all patterns under examination here, including the patterns N-w/
and -be-wh, and also occurs frequently with appositive /-prepositional
phrases as well as aowi-prepositional phrases. The collocational versatility
is paralleled by an enormous semantic variability, so that the use as con-
ceptual shell noun is only one of a number of possible uses. The fact that we
will see the noun again in the discussions of the families 'Belief, 'View',
'Aim', 'Plan' and 'Purpose' testifies to its remarkable adaptability. It must
be noted, however, that all of these uses belong to the mental domain. The
most frequent collocations that give rise to conceptual uses are idea + that-
clause, idea + be + that-c\ause, this/that/such an idea, and especially idea
+ o/-prepositional phrase.
The same four types of lexico-grammatical patterns also dominate the
use of the other conceptual shell nouns. In the remainder of this subsection
Conceptual uses 191
on the 'Idea' family, I will look at these patterns in turn, starting with of-
prepositional phrases.
Besides idea, the nouns theory, principle and doctrine, and especially
notion and concept, are often found with postmodifying o/-prepositional
phrases. Thought occurs in the pattern as well, but with a much lower rela-
tive frequency. In this pattern, the appositive relation between the head
nouns and their postmodifiers, and with it the relation of experiential iden-
tity (see Section 3.1), is open to question.
(9.2) Salt made the crucial distinction between the Victorian philanthropic
notion of animal welfare and the concept of animal rights, (EPHEM)
Thus, while the two examples in (9.2) seem to suggest the interpretations
that the Victorian philanthropic notion the author is talking about is animal
welfare and that the concept is animal rights, such equations are not en-
tirely correct, since the head nouns are not conceptually identical with the
information expressed as postmodifiers. Rather, the nouns refer to mental
representations of this. The recognition that the nouns are conceptually
related to a mental entity or domain, rather than identical with it, is clearly
essential for the description of uses of the nouns principle and theory. This
is shown by example (9.3), where the concepts represented by the nouns
theory and chaos cannot be equated at all:
(9.3) These include Non-linear systems (there are courses on the weather,
modelling of biological systems and the mathematical theory of chaos)
and theoretical physics, (EPHEM)
That there is no complete relation of experiential identity between the head

nouns and their o/-prepositional phrases also emerges from the observation
that the nouns idea, notion and concept are also found with postmodifying
o/-prepositional phrases in the pattern N-e-cl. This is illustrated in (9.4):
(9.4) Their idea of heaven is to find a molecule in the body that is causing
some sort of trouble, and to work out what it is doing wrong, (ECON)
As is highly typical of these uses, the head noun is determined by a posses-

sive pronoun indicating the EXPERIENCER of the IDEA, and the prepositional
complement is an abstract noun with no determiner. In these constructions,
it is beyond doubt that the head nouns are not in a relation of experiential
identity with the postmodifying prepositional phrases but with the compie-
192 Mental uses
menting clauses. Speakers use such expressions to talk about highly subjec-
tive emotional attitudes. These uses are therefore related to the creditive
uses collected in the 'View' family below.
A last interesting formal variant of /-prepositional phrases occurs only
with the nouns idea and thought: these can be combined with gerunds as
prepositional complements of the preposition of.
(9.5) (a) Nobody likes the idea of aborting elephants but, on reflection, surely
it is preferable to slaughtering adults, (NEWSCI)
(b) I had been accepted on a degree course at college and the thought of
losing all that was too much, (MAGS)
As the illustrations in (9.5) indicate, these uses shell very powerful mental
images of hypothetical events.
If one looks at the family as a whole, it is clearly the pattern ~N-that that
provides their favourite environment. Apart from concept, the nouns men-
tioned so far all occur more frequently with ?/ja/-clauses than with of-
prepositional phrases. Notion, for example, can muster a reliance score of
almost 30% for the pattern -that. In the mental domain, these that-c\mses
are not related to Wierzbicka's semantic primitive of 'saying' but to
'knowing'. This is illustrated in example (9.6), which is certainly para-
phrased better by (9.6') than by (9.6"):
(9.6) One of the few consolations of being at the back of the fleet was the
comforting thought that other yachts acted as pathfinders, (MAGS)
(9.6') ... the comforting knowledge that other yachts acted as pathfinders.
(9.6") ... the comforting news that other yachts acted as pathfinders.
Even when the shell content represents a physically observable event, it is

not just turned into an IDEA by its combination with a mental shell noun, but
also into fact-related IDEAS by the use of //^/-clauses. As a consequence,
the paraphrase in (9.6"') is also fairly true to (9.6), although it omits the
characterizational element comforting (cf. Section 15.2).
(9.6"') ... the fact that we knew that other yachts acted as pathfinders.
Depending on the form of the verb in the that-c\axisQ, modal aspects, but
invariably epistemic and not deontic ones, can play a role as well, as is il-
lustrated in (9.7).
Conceptual uses 193
(9.7) While economic theory lends no general support to the idea that re-
moving distortions from one market must lead to an improvement in
general welfare, much free-market politics assumes that it does so.
(NEWSCI)
The nouns in the 'Idea' family are also frequently found in the pattern f/z-N.
This is not surprising in view of the fact that the main function of concep-
tual uses is to put cognitive content of any size and complexity into nicely
manipulable mental boxes. And since conceptual uses abstract from the
dynamic aspects of the mental frame and create well-bounded stative con-
cepts, they provide ideal material for anaphoric references. By far the most
frequent collocations are this idea, this issue and this point - the last of
which, one must add, is most frequently part of the collocation at this point,
which has a locative or temporal meaning akin to at this place or at this
stage, and therefore is a much worse example of a shell-noun use (see Sec-
tion 12.3) than this idea and this issue.
The pattern th-be- also attracts the nouns in the 'Idea' family, although
it must be said that most nouns are fairly rare in this pattern. The most pro-
ductive nouns are point, idea, issue, subject and theme. While the near-
demonstrative this predominates as determiner in uses of the pattern th-H,
that is far more frequent than this as pronominal subject in the pattern th-
be-N. This preponderance is presumably due to the greater potential of that
for expressing subjectivity, which plays an important role in the mainly
characterizing pattern th-be- (see Sections 3.3, 15.1 and 16.2). The most
frequent collocations of this type are listed in (9.8):
(9.8)
Pron.+Verb form Determiner Premodifier Shell noun
that's the 0 point

that's a(n) crucial, good, fair, important, point
interesting, valid
that's the key, main, only, whole point
that's an/the 0 idea
that's a(n) excellent, good, great, inter- idea
esting, lovely, terrible, terrific,
super
that's the best, whole idea
that/this would be a good, great idea
that's another, different issue
194 Mental uses
As is typical of this pattern (see Section 15.1), evaluative adjectives prevail

in these collocations, but restrictive and cohesive ones are found as well.
Attention must be drawn to the fact that many uses of point in the pattern
th-be- are ambiguous with regard to whether their shell contents represent
facts, utterances or ideas.
Finally, it must be mentioned that most nouns in the 'Idea' family are
also productive in the pattern N-be-that. However, these collocations, e.g
the idea is that or theory is that do not produce 'Idea' uses but 'Plan' uses.
I will return to them in Section 9.3.3.
'Mystery'
Shared semantic features: [MENTAL], [CONCEPTUAL], [UNKNOWN]
Frame: Mental; focus on IDEA
Nouns: Question, mystery, puzzle, conundrum, enig-

ma, unknown
For once I shall start out from a true rarity here:
(9.9) (a) How long will they stay? That is the great unknown, (PAPERS)
(b) Another unknown is what effect international pressure will have on
the latest Armenian offensive, (PAPERS)
In these two examples, the noun unknown, which can muster no more than
about one occurrence per one million words, is used to shell mental entities
representing unknown information. As is typical of unknown information,
the shell contents are expressed linguistically by explicit questions as in
(9.9a) or by indirect questions as in (9.9b), i.e. by embedded w/j-clauses.
In contrast to rogative uses of the linguistic shell noun question (see
Section 8.3.2), 'Mystery' nouns do not report acts of asking for unknown
information but shell the unknown information itself. Theoretically clear as
this contrast may be, it must be added that question is among the most sali-
ent manifestations of the fact that the domains of language and mind are
ultimately inseparable (see Section 8.1).
The other nouns in this family, except of course question, which is found
in other uses as well, occur not much more frequently as shell nouns than
unknown. The nouns share the collocational preferences illustrated in (9.9),
that is they predominantly rely on the patterns th-be- and -be-that. Post-
Conceptual uses 195
nominal w/z-clauses occur with the nouns question, mystery, conundrum

and enigma. But as with the rogative uses of the noun question, speakers
prefer to signal the identity relation between the mental entity and the un-
known information linguistically with an appositive of. This is illustrated in
example (9.10):
(9.10) ... and the even more controversial chat between Prince Charles and Mrs
Camilla Parker Bowles. The two set about probing the origin of the tapes
hoping that they would unravel the intriguing mystery of who had
bugged the royal couple and why, (PAPERS)
9.3 Psychological-state uses

9.3.1 Creditive uses
Psychological-state uses allow speakers to include a characterization of the

psychological state which they want to attribute to EXPERIENCER in their
shells for IDEAS. The hallmark of creditive uses is that the subjects of these
PSYCHOLOGICAL STATES are abstract relations (rather than events). The
particular type of IDEAS that represent the propositional content of these
PSYCHOLOGICAL STATES or emerge as their results will be called BELIEFS.
(Recall that this frame component was already introduced in the section on
evidential uses, Section 7.4)
Creditive uses are divided into four families: 'Belief, 'View', 'Realisa-
tion' and 'Illusion'. As usual, the names of these families are suggestive of
their semantic core. The nouns in the 'Belief family shell psychological
states of knowing and believing, those in the 'View' family shell what peo-
ple think about certain issues. The 'Realisation' nouns are used to empha-
size that the BELIEFS expressed as shell contents are the results of previous
mental processes. And the 'Illusion' nouns are used to portray BELIEFS as
incorrect.
'Belief
Shared semantic features [MENTAL], [CREDITIVE]
Frame: Mental; focus on psychological state

Nouns: Idea, belief, hope, feeling, impression,
speculation, knowledge, assumption, confi-
196 Mental uses
dence, conviction, suspicion, understanding,

prospect, expectation, instinct, premise,
estimate, presumption, calculation, hunch,
supposition, intuition, inkling, superstition,
conjecture, presupposition, surmise
That idea is the first item in this list is rather misleading. Its position is only
due to its high frequency as member of several families of mental uses. Idea
is not a typical 'Belief noun, because it is only found in this function as
part of the collocation have an/some/any/no idea. These fixed expressions
are used as expanded predicates, similar in meaning to the simple verb
know. Like this verb, they can be complemented with that- and wA-clauses,
but also with w/z-clauses linked to the noun with the preposition of. As ex-
plained in Section 3.1.1, uses of shell nouns that are parts of expanded
predicates are peripheral examples of shell nouns because the nominal ef-
fects are largely lost in these constructions. Other creditive nouns that also
combine with the function verb have to form expanded predicates are feel-
ing, impression, suspicion, instinct, faith, sentiment, hunch and inkling.
While the latter three less frequent nouns are more or less earmarked for
this type of usage, the former five are also often found in more typical shell-
noun uses.
Another noun that does not really deserve its place in the upper part of
this list is speculation. This noun is yet another manifestation of the pre-
ponderance of media sources in the COBUILD corpus (see Section 4.2).
While its frequency in this corpus is roughly 30 occurrences per million
words, it occurs only 10 times in the one-million word LOB corpus (Jo-
hansson and Hofland 1989). The comparatively high frequency of specula-
tion in the COBUILD corpus is due to the popularity of this noun (especially
in the collocation there is speculation that) in media texts as a marker of
tentativity. It also functions as a hidden reference to the fact that the source
of a piece of information is not known or not to be named. A typical exam-
ple is given in (9.11):
(9.11) One of the UN officials, Mr Giandomenico Picco, is now in the Syrian

capital, Damascus. There is speculation that he may meet representa-
tives there of Islamic Jihad as well as the visiting Iranian Interior Min-
ister. Mr Abdallah Nomi (BBC)
Psychological-state uses 197
Examples of this type are also untypical shell-noun uses because of the
existential /fore-construction (see Section 3.1.1) and the fact that the noun
speculation has no determiner.
Even a cursory glance at the list of the nouns in this family indicates that
the nouns in the 'Belief family are not a very homogeneous set. One would
intuitively claim, for example, that a mental state described as knowledge
differs considerably from beliefs and impressions. The major part of this
subsection will be devoted to suggesting a number of dimensions that help
to differentiate these nouns and to thus elucidate their meanings. To a
greater or lesser extent, the parameters outlined in what follows are also
valid for the 'View' family.
To begin with, the nouns can be distinguished along a dimension that can
be called imputed strength of commitment (cf. Lyons 1977: 794).67 The
examples in (9.12) differ from each other with respect to this dimension:
(9.12) (a) At any one of Lord McAlpine's five clubs [...] members can speak up
for absent friends like Mr Botnar or Mr Asil Nadir in the comfort-
able knowledge that their views have been spoken out loud by a
former treasurer of the Tory party on the Breakfast With Frost pro-
gramme. (PAPERS)
(b) The current recession has finally punctured the belief that house
prices can rise forever, (PAPERS)
(c) The impression that events are fast slipping from their control is
confirmed by an order from the Ministry of Defence authorising the
withdrawal of all military units from the streets of Moscow, (BBC)
By using the noun knowledge and modifying it with the adjective comfort-
able, the speaker of (9.12a) imputes to the EXPERIENCERS the strongest
commitment possible that their BELIEFS are true representations of past
events. While the noun belief itself, as used in (9.12b), still portrays a fairly
strong commitment, the commitment imputed by impression in (9.12c) is
much weaker. The use of impression suggests that one cannot be quite so
sure whether the shelled BELIEF is correct or not. The noun knowledge
clearly embodies the end-point on the strong side of the scale, since knowing
implies believing and having an impression, but not vice versa (Lyons 1977:
794). Other nouns tending towards the stronger end of the scale are confi-
dence, conviction and presumption. Feeling, intuition and suspicion are on
the weaker side, while assumption, belief, premise and understanding are
fairly neutral.
198 Mental uses
Related dimensions are the extent to which a BELIEF is justified and the
nature of the justification for the BELIEF. How well or badly justified a
BELIEF is tends to correlate with the imputed strength of commitment.
Strong BELIEFS tend to be thought of as being well justified, weak BELIEFS
as being badly justified. With regard to the nature of the justification of a
BELIEF, the fundamental distinction offered by the data is that between ra-
tional and emotional bases. Nouns like knowledge, understanding, suppo-
sition, calculation, premise and estimate suggest a rational basis; feeling,
intuition, instinct, inkling, hunch and especially superstition, on the other
hand, evoke an emotional basis. Metaphorically speaking, the former group
portrays BELIEFS as coming from the brain, and the latter as coming from
the heart, or the stomach. It is no coincidence that the typical collocations
are gut feeling and gut instinct, rather than 'gut knowledge or 7gut supposi-
tion. In view of the rationalist bias of Western societies, it is also not as-
tonishing that there is a tendency for strong BELIEFS to correlate with a ra-
tional basis, and for weaker BELIEFS to correlate with an emotional basis.
As in the class of assertive linguistic nouns (see the family 'Prediction'
in Section 8.3.1), a small group of creditive nouns can be singled out which
are predominantly predictive. This semantic component is reflected gram-
matically in the frequent occurrence of future tenses in the complementing
clauses, as in (9.13):
(9.13) My own expectation is that the legitimate government of Kuwait will

be re-established in the next six months and that the present government
of Iraq will probably fall as a consequence, (PAPERS)
The predictive element is undoubtedly most marked with the nouns hope,
prospect and expectation. Hope is also comparatively productive in the
pattern -be-that. Unlike the other nouns in this subset, this noun is even
found in the pattern -be-to. Since these uses are similar to, but not identi-
cal with, the uses subsumed in the 'Aim' family, I will postpone a more
detailed discussion of the noun hope until Section 9.3.3. -clauses with
future tenses, especially complementing ones following a copula as in
(9.13), were also found with the following 'Belief nouns: assumption, be-
lief, calculation, estimate, presumption, prospect, speculation, suspicion
and understanding.
A further parameter for the differentiation of creditive uses is whether
the speakers explicitly identify the EXPERIENCERS or not. If this is the case,
the next question is whether speaker and EXPERIENCER coincide or not, i.e.
whether speakers shell their own or other people's BELIEFS. The main clues
on the linguistic surface are possessive determiners - either personal pro-
nouns or proper noun genitives - because these are used to refer to
EXPERIENCERS. In the framework applied here, the first person singular
possessive pronoun my 'means' that speaker and EXPERIENCER coincide,
and the first person plural pronoun our that the speaker is one of several
EXPERIENCERS who share the same BELIEF. With regard to this dimension
one must differentiate between N-cl and N-e-cl uses.
In the pattern N-cl, belief, estimate and calculation are the only nouns in
the 'Belief family that produce noteworthy numbers of occurrences with
possessive determiners. Example (9.14) provides an illustration:
(9.14) I do hope the Conservatives' Cheltenham candidate John Taylor is

wrong in his belief that fellow Tories celebrated his General Election
defeat because he was black, (TODAY)
These three nouns were found with third-person pronouns, as in this exam-
ple, and also, but less frequently, with first-person pronouns as determiners.
On the whole, however, speakers show a distinct tendency to use 'Belief
nouns in the pattern N-cl not with possessive determiners but with definite
articles. In these cases, the EXPERIENCER must be inferred from the context.
This task can be easy to solve, as in (9.15), where the role of EXPERIENCER
coincides with the clause subject the British government. In (9.16), the
identification of EXPERIENCER is more difficult, since it is not clear whether
the writer of this passage imputes the habitual BELIEF to the Irish poet
mentioned in the previous clause, or conceives of it as an intersubjectively
shared mental entity, similarly to what is shelled by conceptual shell noun
uses (see Section 8.2).
(9.15) Mr Waldegrave's action shows the British government is anxious to

waive the usual immigration procedures on the assumption that none of
those classified as other nationals will seek to stay in Britain, (BBC)
(9.16) It deals implicitly with the mixed feelings that his love for Maud Gonne
had given rise to, and explicitly with his conviction that an Irish poet
must address an Irish subject matter. Latent in that conviction is the be-
lief that the artist's subject matter must come from the world he finds
around, (BOOKS)
Possessive determiners are much more common in N-e-cl than in N-cl

uses. They are particularly frequent in the pattern N-ie-cl with the follow-
200 Mental uses
ing nouns: assumption, belief, estimate, expectation, feeling, hope, hunch,

impression, instinct, suspicion and understanding. All these nouns are
found with first-person and third-person pronouns and proper noun geni-
tives as determiners. However, only feeling, hunch, impression, instinct,
intuition and understanding exhibit a marked tendency to be combined with
the first-person form my. For example, of 35 instances of the collocation
feeling is that ... in the SPOKEN section of the corpus, no less than 22 are
instances of my feeling is that. There are four additional instances of my
personal feeling is that..., and three instances of my own feeling is that....
Of the twelve occurrences of impression is that in the same corpus section,
as many as nine are determined by my.
How does this concentration of first-person pronouns with half a dozen
nouns in this particular pattern come about? Several explanations can be
suggested here.
One of them starts out from the recognition that the function of N-e-cl
uses is identifying the contents of a shell noun (see Section 3.3). In analogy
to linguistic uses in the pattern N-e-cl (see Section 8.1), N-e-cl uses of
creditive mental nouns can thus be seen as both reports and statements of
IDEAS. When such statements are accompanied by first-person pronouns, a
likely motive for their use is that the speakers want to emphasize that these
BELIEFS are their own personal and subjective BELIEFS. This explains uses
like my own personal feeling is that... mentioned above, which explicitly
underscore both the co-referentiality between EXPERIENCER and speaker and
the subjectivity of the BELIEF.
However, this does not yet explain why such a small number of nouns is
preferred for that purpose. Looking at these nouns once again - feeling,
hunch, impression, instinct, intuition and understanding - w e realize that
all but understanding are nouns which suggest an emotional basis for a
BELIEF. In fact, they make up the core of this group. This recognition ties in
nicely with the subjectivity hypothesis put forward in the last paragraph.
The data suggest, then, that speakers use the particular nouns in the par-
ticular construction to underscore that what they claim to be true is not
based on objective knowledge about a fact, but only represents their own
subjective BELIEF, which may be weak and badly grounded. This adds ten-
tativeness as a second possible motive for these uses, in addition to subjec-
tivity. Speakers can use such hedging gambits to indicate that they are not
so sure about their claims and to reduce their commitment to the truth of
what they say. Since the deliberate expression of subjectivity is a gesture of
self-confidence rather than tentativeness, the somewhat paradoxical situa-
tion arises that one type of expression can be used for two opposed pur-
poses. Interestingly, the same paradoxical situation has been found with
regard to the expression I think (Holmes 1990: 199, Aijmer 1997: 21-23),
which is pragmatically very similar to expressions like my feeling is and the
like.
Example (9.17) is a somewhat confusing extract from a conversation. At
one point, the speaker seems to change his or her mind and switches to an-
other construction before actually uttering the shell content. This can be
seen as a sign of uncertainty which may also give rise to the wish to state a
belief in a fairly tentative tone. The example gives a good idea of how
speakers can use this collocation to detach themselves from the responsibil-
ity for the truth of a BELIEF:
(9.17)
SPEAKER 1: ... and I don't think anyone does.
SPEAKER 2 : No.
SPEAKER 1: But on the other hand I think over the long haul the atmosphere
tends to want to restore equilibrium to itself.
SPEAKER 2: Yeah.
SPEAKER 1: And so in fact some of the [inaudible]. And I think my feeling is
that I mean I I'm not god I don't know what the true answer is
[inaudible]
SPEAKER 2: Yeah.
SPEAKER 1: the C O two
SPEAKER 2: Yeah.
SPEAKER 1: but in fact there won't be such a thing because the CO two
SPEAKER 2 : Mhm.
SPEAKER 1: continues to go up (SPOKEN)
Uses of this type establish yet another particularly close point of contact
between the linguistic and the mental domain, being very similar to uses of
the 'Guess' family (see Section 8.3.1). With the noun guess itself, uses in
the pattern N-e-cl where SPEAKER and EXPERIENCER coincide are also
very frequent. The fact that guess is often used with a modal form of the
verb be, as is illustrated in (9.18), provides support for the interpretation
that these uses are motivated by caution and tentativeness:
(9.18)
SPEAKER 1: ... whether people are going to erm renew their interests and
ethical funds are going to
202 Mental uses
SPEAKER 2 : Yes. Yes.

SPEAKER 1: to My my guess would be that er people are going to renew their
interests and er we will see erm a serious pressure from that par-
ticular source... (SPOKEN)
There is a yet a third explanation of such uses, though, which builds on

subjectivity and tentativeness. The frequency of the first person pronoun
with tentative nouns can be interpreted as a downtoning device in the service
of negative politeness (Brown and Levinson 1987: 144-172). Especially if a
speaker has to make a potentially face-threatening speech act, for example a
statement of disagreement with a previous claim of a discourse partner, it
may be more polite to portray even a strong belief as a feeling. Example
(9.19) can be explained along these lines.
(9.19) ... said: "Radiologists and dentists did not seem to carry any increased
risk. Those who worked with nuclear materials, scientists and nuclear
technicians, who could have suffered internal contamination, were at
risk. My own feeling is that alpha emitters, like plutonium, are going to
prove an important pathway through fathers to their unborn children."
(PAPERS)
In sum, I have proposed that subjectivity, tentativeness and politeness can

be motives why speakers tend to use weakly grounded, emotionally based
creditive nouns to shell their own BELIEFS. It may sometimes be difficult to
decide whether politeness plays a role or not. The other two factors corre-
late with stress distribution in spoken English. If the personal pronoun is
stressed in expressions like my impression is that, the expression of deliber-
ate subjectivity is foregrounded; if the shell noun itself receives more
prominence, then tentativity seems to be a more important motive for the
uses of the construction. I will come back to these hypotheses in the next
subsection on 'View' uses and in the functional part in Section 16.1.
Having ventured quite far into the functional domain, I would like to end
the description of this family with a brief survey of the quantitative data.
With frequencies as collocates between 4,255 and 2,737, the nouns belief,
feeling, hope and impression are among the fifty most frequently used
nouns in my data. With 1,855 occurrences in the patterns under examina-
tion, assumption is still among the first one hundred. Although most nouns
in the family were also found in th- and th-be-N, their numbers in these
patterns are comparatively low. Some of the nouns can muster remarkably
high scores for reliance in the pattern N-cl, and consequently for compiled
reliance. The highest scores are compiled in the chart in Figure 9.4:
Noun Total fre- Frequency as Compiled Frequency in Reliance on

quency in collocate reliance the pattern the pattern
225m corpus -that -
assumption 3,151 1,855 58.87% 1,391 44.14%

belief 9,344 4,255 45.54% 3,696 39.55%
supposition 164 71 43.29% 46 28.05%
presupposition 34 13 38.24% 8 23.53%
presumption 408 138 33.82% 98 24.02%
impression 8,206 2,737 33.35% 2,279 27.77%
feeling 14,392 3,795 26.37% 2,511 17.45%
suspicion 4,088 953 23.31% 806 19.72%
knowledge 17,463 2,398 13.73% 1,794 10.27%
Figure 9.4 Selective collocational data on some 'Belief nouns
'View'
Shared semantic features: [MENTAL], [CREDmVE], [ATTTTUDINAL]

Nouns: Idea, view, notion, line, opinion, conviction,
experience, attitude, perception, philosophy,
instinct, faith, awareness, thinking, logic,
stand, perspective, stance, sentiment, obses-
sion, prejudice, viewpoint, conception, ra-
tionale, persuasion, ethos, standpoint, pre-
conception, presentiment
Speakers use 'View' nouns to shell psychological attitudes towards issues

of almost any nature. These nouns do not shell what an EXPERIENCER be-
lieves to be true at a certain time in a given situation, but what an
EXPERIENCER thinks about an issue. Therefore I will use the term ATTITUDE
to refer to the specific variant of shell contents that are encapsulated by
these nouns.
I have already drawn attention to uses of the type their idea of heaven is
to ... (see example 9.4). Here idea and notion occur with the meaning typi-
204 Mental uses
cal of the 'View' family in combinations with possesssive determiners, of-

prepositional phrases with abstract nouns as prepositional complements and
complementing infinitives or âi-clauses. Like other 'View' nouns, these
expressions are used by speakers to portray more or less longstanding atti-
tudes. Another way of expressing the meaning that I attribute to the 'View'
family is to use the noun idea in the plural. Expressions like the one exem-
plified in (9.20) reflect the tendency to imagine the ATTITUDES described by
the 'View' nouns as sets of beliefs rather than single beliefs, see for exam-
ple the dictionary definitions "ideas or beliefs about something" (LDOCE3,
s.v. opinion, my emphasis) for opinion or "opinions and feelings" (ibid, s.v.
attitude, my emphasis) for attitude.
(9.20) Mill credited his wife Harriet Taylor Mill with the development of his
ideas concerning women, (MAGS)
Given this general characterization of the communicative impact of 'View'

nouns, common sense suggests that these nouns are also likely to have the
potential for highlighting the subjectivity of a BELIEF. Since the nouns refer
to ATTITUDES that may be the result of recurrent attempts to come to terms
with an issue or problem, it is only natural that such views or opinions are
more closely attached to the persons who develop and hold them. The dis-
cussion of the 'Belief family has indicated that the pattern N-e-cl is the
best testing-ground for this question because it can be seen as a special
subjectivity-highlighting device. Two highly typical examples of 'View'
nouns in this pattern are given in (9.21) as illustrations:
(9.21) (a) The Inland Revenue have a perfect right to say, "Look, this is cos-
metic," but in this particular case, since John Birt Productions Ltd is
a genuine vehicle, my own view would be that it has worked.
(PAPERS)
(b) After nearly 60 years on this earth, 46 of them spent in the pleasur-
able pursuit of fish, my experience is that people who follow field
sports, particularly angling and particularly young anglers, are far
less antisocial than those who never go near the countryside or its
pursuits, (PAPERS)
Retracing the argumentative steps followed in the subsection on 'Belief

uses, we can start by comparing the overall numbers of tokens of the pattern
N-that and -be-that. If the attitudinal 'View' nouns indeed have more
potential for the expression of subjective attitudes, they should produce
relatively more instances in the pattern N-e-cl than the 'Belief nouns,
because only a limited number of nouns in this family had this quality. The
'View' nouns were recorded 5,912 times in the pattern N-cl and 1,586 times
in the pattern N-e-cl, which corresponds to a ratio of four to one.68 In con-
trast, the ratio of N-cl to N-e-cl uses of 'Belief nouns is more than ten to
one (the scores are 18,449 N-cl vs. 1,768 N-e-cl uses). Although one must
keep in mind that these scores may also be influenced by many other fac-
tors, the difference is undoubtedly striking and may be regarded at least as a
first indication of the subjective nature of 'View' nouns.
The next measure is the proportion of nouns which were found with pos-
sessive determiners in the pattern N-e-cl. While only a small number of
nouns produce such instances in the 'Belief family, these uses are clearly
the rule rather than the exception with 'View' nouns. Not only the more
frequent nouns view, opinion, attitude and experience are found with pos-
sessive determiners but also rarer ones like perception, thinking, stance and
viewpoint69
Next, I want to take a closer look at some of the more frequent nouns
and their combinations with first-person determiners. In the SPOKEN section
of the data, which I have also used as a measuring-rod for 'Belief nouns,
the collocation view is that was found 50 times. 21 of these instances are
determined by my. The even stronger subjective collocations my own view is
that, my personal view is that and my own personal view is that were found
four, four and two times respectively. Thus, more than three fifths of the
uses of view is that in this corpus section show collocational signs of sub-
jectivity. Experience was recorded 14 times in the pattern N-e-cl with the
verb form is. Eight of these occurrences are instances of the collocation my
experience is that and one of my personal experience is that. Opinion has
13 instances in the pattern in question: two of these are instances of the
collocation my opinion is that, one of my own opinion is that, five of my
personal opinion is that and another two of my own personal opinion is
that.
In sum, the collocational findings for these nouns clearly support the
idea that they have a more subjective and attitudinal meaning than the 'Be-
lief nouns. These findings can receive further support from the lexical side,
if we look at the less frequent nouns as well: the context-independent
meanings of such nouns as thinking, stand, perspective, stance, prejudice,
viewpoint, persuasion, ethos and standpoint, and even of the highly rare
nouns preconception and presentiment, contain references to the idea that
the ATTITUDES they denote in given uses are of a personal and individual
206 Mental uses
nature. The other two aspects discussed with regard to 'Belief uses in this
pattern, tentativeness and politeness, are clearly less important than subjec-
tivity here. This is also supported by the intonational evidence that in such
collocations as my view/opinion/experience is that it is typically the posses-
sive determiner that receives more prominence than the noun.
I have focused my attention on N-e-cl uses of 'View' nouns, since these
are the typical and the most interesting ones from a pragmatic point of view.
To conclude the descriptive account, it should be mentioned that especially
view itself but also many other 'View' nouns occur in the patterns N-cl and
th-N, some of them even in the pattern th-be-N. Unlike most nouns in the
'Belief family, 'View' nouns are also frequently found in the collocation
such a N, for example such a view, such an opinion and such an attitude.
'Realisation'
Shared semantic features [MENTAL], [CREDITIVE], [RESULTATTVE]

Nouns: Realisation, discovery, recognition, analysis,
interpretation, lesson, reading, insight, con-
sideration, reasoning, reflection, inference,
equation, diagnosis, significance, recollec-
tion, anticipation, appreciation, projection,
generalization, reckoning, deduction
The nouns in the 'Realisation' family are used to portray BELIEFS as results
of acts of thinking. This means that they widen the scope of attention from
BELIEFS to ways of arriving at BELIEFS. To a certain extent, one can there-
fore claim that their meanings incorporate the features [+DYNAMIC] and
even [+AGENTIVE], although of course they denote mental rather than exter-
nal activity. The shell nouns themselves typically describe different kinds of
mental processes like realising, discovering, recognizing or reasoning, which
can lead up to certain BELIEFS. As with the other creditive shell nouns, the
shell content expresses the BELIEF. See example (9.22) for a first illustra-
tion:
(9.22) And in Baghdad, meanwhile, behind Saddam Hussein's belligerent

rhetoric may lie the realisation that the time for manoeuvre and brink-
manship is now running out, (PAPERS)
The nouns lesson, analysis, diagnosis and significance are exceptions to

the dynamic tendency of 'Realisation' uses. These nouns, which are not
straightforward derivations from dynamic cognition verbs, constitute a tran-
sition zone between 'Realisation' and 'Idea' uses. On the one hand, they
shell BELIEFS as results of mental activities. This is shown by example
(9.23), a typical use of the noun lesson in the pattern N-be-that:
(9.23) "... What we have witnessed in the past five years is massive retrench-
ment and refortification by the British in the occupied area," he ex-
plained. "The lesson is that when they rearm we improvise." (PAPERS)
The expression the lesson is that in this example roughly means 'what we
have discovered is that' or 'what our considerations of the past events have
shown us is that'. These paraphrases show that like discovery and consid-
eration, lesson has a resultative component. On the other hand, this compo-
nent is not so conspicuous on the linguistic surface as with the de-verbal
nouns, but hidden in the semantic structure of the nouns. As a result, lesson,
as well as the nouns analysis, diagnosis and significance, often evoke the
image of mental entities rather than mental processes, and it is in this re-
spect that these nouns resemble conceptual shell nouns.
The semantic difference between more dynamic and more stative 'Reali-
sation' nouns is reflected in their collocational data. Speakers clearly favour
-that constructions for the dynamic nouns realisation and recognition. In
the pattern -be-that, these nouns are not used at all, and in the patterns th-
N and th-be-N, they are also comparatively rare. The more entity-like nouns
lesson, analysis, diagnosis and significance can muster relatively large
numbers of occurrences in the pattern -be-that (as illustrated in 9.23) and
in the pattern th-N.
'Illusion'
Shared semantic features: [MENTAL], [CREDHIVE], [FALSE]
Nouns: Illusion, fantasy, misconception, fiction, delu-
sion, fallacy, misapprehension, deception,
misjudgment, miscalculation, misperception
208 Mental uses
The semantic feature shared by the nouns in this family is [FALSE], Speak-
ers use not only nouns like misconception, delusion or fallacy, but also the
potentially more neutral nouns illusion and fantasy in order to indicate that
the BELIEFS expressed as shell contents do not correspond to reality. In
(9.24), for example, a use of illusion in the pattern -that, the speaker
clearly insinuates or even implies that the BELIEF imputed to Mr Battle is
false. Similar implications can be attributed to the use of illusion in the
pattern th-be- quoted in (9.25).
(9.24) Mr Battle even insisted that Honoria and I sleep in different bedrooms,
whether from the illusion that we hadn't vet slept together, or to main-
tain the facade of Victorianism, I couldn't tell, (BOOKS)
(9.25) But that was a scientific illusion built on evolutionary terminology.
(BOOKS)
'Illusion' uses are fairly rare. They occur in the patterns -that, th- and
th-be-N, but almost never in the pattern N-e-cl. The only noun that pro-
duces noteworthy numbers of instances in the latter pattern is misconcep-
tion in the collocations the common/popular misconception is that.
9.3.2 Dubitative uses
'Doubt'
Shared semantic features: [MENTAL], [DUBITATIVE]

Nouns: Doubt, question
Like rogative uses, their counterparts in the linguistic domain, dubitative

uses cannot boast a long list of nouns. What is more, it is doubtful whether
genuine shell-noun instances of this type can be said to exist at all, as the
two potential members of the family are almost exclusively used in semi-
fixed phrases. These collocations are compiled in (9.26):
(9.26) (a) There is/can be no doubt/question that...

(b) There is little doubt that.. .
(c) X has no doubt that...
(d) It is beyond doubt/question that...
While especially the collocations in (a), (b) and (c) are extremely frequent,
none of them can be regarded as good examples of shell nouns. Collocations
(a) and (b) are cases of existential i/jere-constructions, (c) is an expanded
predicate with the function verb have, and (d) is an extrapositional con-
struction. In all these cases, the connection between the -clauses and the
nouns is not particularly close. And even if one treats these collocations as
shell noun-like uses, most of them would have to be regarded as factual
shell-noun uses (see also Section 7.2), because the whole noun phrase which
includes the negator no must be seen as functioning as shell. Thus, colloca-
tions (a), (c) and (d) are equivalent to statements of facts like 'it is a fact
that' or 'it is true that'. Only collocation (b) can claim to produce mental
shell-noun uses, but these are creditive rather than dubitative. This is shown
by the fact that it can be paraphrased by have the strong belief that'.
9.3.3 Volitional uses
Leaving the tfiai-clause milieu, in which creditive and dubitative uses are
normally found, we are now moving into that of infinitive clauses. As sug-
gested by the complementation of verbs (see Section 3.2) and the distribu-
tion of /Aoi-clauses and infinitive clauses in the linguistic domain (see
Chapter 8, esp. Section 8.3.3), the main reason why speakers use infinitive
clauses as postmodifiers and complements lies in the meaning of the nouns.
Volitional shell nouns can be used to shell PSYCHOLOGICAL STATES of
'wanting' that future events, especially future activities, take place. In Hal-
liday's framework, the nouns have to do with proposals rather than propo-
sitions. The specific variant of IDEAS that I am concerned with here will be
called AIMS. AIMS are thus the volitional counterparts of BELIEFS in the
creditive domain.
There are as many as nine families of volitional uses. The two semanti-
cally most general ones are 'Aim' and 'Desire', which consist of nouns
which are used to shell intended accomplishments and activities respec-
tively. 'Plan' nouns shell intended ways of arriving at AIMS. The 'Reluc-
tance' nouns are semantically opposite to 'Desire' nouns. The remaining
five families, 'Purpose', 'Solution', 'Motivation', 'Determination' and
'Agreement', are semantically more specific.
210 Mental uses
'Aim'
Shared semantic features: [MENTAL], [VOLITIONAL], [CONCLUSIVE]
Nouns: Point, idea, hope, aim, goal, ambition, inter-
est, objective, vision, object, target, ideal
'Aim' nouns shell things that EXPERIENCERS want to achieve by means of

certain activities. These types of AIMS resemble the "accomplishments" of
descriptive grammar (see Quirk et al. 1985: 206-207, and Section 5.1.1) in
that they are related to the intended endpoints of activities. They have a
conclusive element. As a consequence, 'Aim' nouns tend to highlight IDEAS
rather than PSYCHOLOGICAL STATES, because it is mainly this endpoint that
is shelled by 'Aim' uses.
The typical lexico-grammatical environment of 'Aim' nouns is the pat-
tern -be-to. An example of this type is provided in (9.27):
(9.27) Walker's next goal is to get into the Welsh squad for the Five Nations
Championship in the New Year... (TODAY)
Almost all members of this family can muster more occurrrences in this
pattern than in the pattern N-cl, which tends to be more frequent in many
other families. With no less than 2,646 recorded instances, the noun aim is
by far the most frequent noun in the pattern + be + infinitive clause (see
Table 4.7). An explanation for this marked collocational tendency will be
given below (see the 'Desire' family below).
Ideal and vision have emotive connotations which relate them to the uses
subsumed in the 'Desire' family and to emotive uses. One can often detect
yet another semantic element in these uses, which allows the speaker to
suggest that EXPERIENCER will probably not achieve what he or she is aim-
ing for. (9.28) is a good example of such a use:
(9.28) But Wallyworld's grandest vision is to make Alaska the linch-pin of a

new global rail system, (PAPERS)
As already hinted at in Section 9.3, hope is also used as an 'Aim' noun in

the pattern -be-to. We have seen, however, that it is much more common
withtfjcrt-clauses,especially with postnominal ones as in (9.29):
(9.29) All three Baltic leaders have turned up for today's meeting; not to par-
ticipate, they say, but to observe, in the hope that they'll also be able to
hold joint talks with President Gorbachev ... (BBC)
Interestingly, the EXPERIENCER and the subject of the that-clause are co-
referential here, and this is also the case in many other examples. This
raises the question why the speaker did not use the more economical infini-
tive construction instead of the that-clause, i.e. in the hope to be able to
hold joint talks. I would like to suggest two explanations. The first is that
the noun phrase the hope is not an autonomous clause constituent here but
occurs as a prepositional complement in a prepositional phrase that func-
tions as adverbial. This seems to be less suitable an anchor for an infinitive
clause than an autonomous noun phrase functioning as subject or object. A
second, semantic, explanation is that hope is not used as a volitional noun
here, but as a creditive noun, which creates an emotive shell for the BELIEF
that they'll also be able to .... This means that depending on the type of
clause that follows, hope can function as creditive or volitional noun.
The noun idea, which is also used with both infinitive and that-chuses
in the pattern N-e-cl, seems to cover the transition zone between volitional
and creditive uses, too. When it is complemented by an infinitive clause,
idea is virtually synonymous to aim. This is substantiated by example
( 9 . 3 0 ) , in which the two nouns are interchangeable:
(9.30) It's now been over a month since Soviet president Mikhail Gorbachov
cut off all oil and most natural gas supplies to Lithuania. The Kremlin's
idea is to bring the country to its knees for having declared independ-
ence from the Soviet Union, (BBC)
Like aim, idea is used here to shell an accomplishment envisaged by an

EXPERIENCER. In the pattern -be-that, as illustrated in example ( 9 . 3 1 ) ,
however, idea seems to be just as close to creditive uses of hope and ex-
pectation as to volitional ones of aim.
(9.31) The project involves a team headed by David Broome, with Geoff Bill-
ington and myself as his partners. The idea is that the three of us will
help young riders coming into the sport, (MAGS)
It is doubtful whether the noun idea actually shells an accomplishment in

this example and other instances of this collocation. I do not think that it
does. It would have been perfectly possible to express such a purely voli-
212 Mental uses
tional meaning by using an infinitive clause instead of the that-clause: the

idea is to help young riders coming into the sport. As the example stands,
the interpretation that the speaker shells a predictive BELIEF concerning a
future state of affairs is more convincing. It is in this way that the noun idea
creates a bridge between the volitional and the creditive domain.
'Plan'
Shared semantic features: [MENTAL], [VOLITIONAL], [MANNER]
Frame: Mental; focus on IDEA
Nouns: Decision, idea, plan, policy, programme, line,
scheme, project, strategy, drive, principle,
rule, attitude, brief, secret, philosophy, art,
tactic, routine, rationale, motto
Briefly, and for once without recourse to the shell-framework, here is the
difference between 'Aims' and 'Plans': 'Aims' are things we want to
achieve and 'Plans' are ways and methods we devise to pursue and achieve
them. It is, therefore, only in an indirect way that the members of the 'Plan'
family produce volitional uses. They do not shell PSYCHOLOGICAL STATES
of volition, but IDEAS that are formed during states of volition or as their
results.
A number of the nouns in the list above, namely line, principle, rule, at-
titude, secret, philosophy, rationale and motto, are not new to us. These
nouns occur also in conceptual and/or creditive uses when they are com-
bined with /-clauses. All of these nouns occur with infinitive clauses as
well, but less frequently than with -clauses and only in the pattern N-e-
cl. (9.32) gives an example:
(9.32) ... unusual meals, overeating and travel stress - they are all situations
that can create a medley of problems for your digestive system and bring
it to a virtual halt, while your internal clocks are spinning in confusion.
A basic rule is to eat lightly: small portions of food can generally be
better tolerated, (BOOKS)
In this pattern, rule and the other nouns shell an IDEA that is related to the
manner of an activity in the pursuit of an aim.
The same basic setup also applies to the other more frequent nouns and
their uses in the pattern N-cl. Especially uses of plan itself can easily be
confounded with uses of aim. However, the two types should be kept apart,
not only because, unlike 'Aim' nouns, 'Plan' nouns are less frequent in the
pattern N-e-cl than in N-cl, but also because the two sets of nouns have
different semantic structures. Example (9.33) can serve to show in what
way the nouns differ:
(9.33) In Canada a plan to slaughter the largest buffalo herd in the world has
triggered off an unprecendented revolt within a government branch
charged with the protection of that country's national park, (BBC)
At first glance, it seems as if the noun phrase a plan could simply be re-
placed by the aim without a noticeable change in meaning. However, it is
not true that such a move would not affect the meaning of the sentence.
First, the change of article (which is necessary because 7an aim to slaughter
the largest... sounds decidedly odd) brings about a difference with regard
to the reference of the noun phrase. And second, (9.33) does not really mean
that slaughtering these animals is the intended accomplishment. This activ-
ity is rather portrayed as a means of pursuing some aim which is not even
mentioned in the passage that is quoted. It would be perfectly natural for the
sentence following the one quoted in (9.33) to identify the aims behind the
plans by means of the collocation the aim is to. This shows that the 'Plan'
nouns do not shell AIMS but ways and methods of pursuing AIMS.
The same essentially also applies to decision, the most frequent member
of this family, which is the seventh most frequent noun in the whole data.
This noun incorporates the additional feature that the EXPERIENCE!*, has
spent some time deliberating or, in the case of several EXPERIENCERS, even
debating, a future course of action in pursuit of an AIM.
The fact that the element of deliberation is included in the meaning of
decision also implies that mental states of a creditive nature perhaps play a
more important role here than for other members of the family. Thus, what
Wierzbicka (1988) maintains with respect to the verb decide, namely that it
refers in its semantic structure to "both wanting and knowing" (1988: 138),
is also true of the noun decision. For Wierzbicka, the epistemic elements of
knowing and thinking make up the difference between verbs of pure voli-
tion, which do not have them, and verbs of desire, which do. Wierzbicka
also suggests that uses of decide that can be reduced to decide to by in-
serting (decide to) think (that). This would mean that uses with that-
complements highlight the epistemic meaning element. My data on the col-
location decision that does not completely support this view, because many
214 Mental uses
examples seem to be motivated either by the fact that the subjects in the
main and the subordinate clause are not co-referential or by the wish to
sound formal. (9.34a) is an example which clearly supports Wierzbicka's
thesis. In (9.34b), the subjects do not coincide. Finally, in (9.34c), I see
stylistic motives at work. Although it may play a role that make a decision
is more emphatic than the simple verb decide, and that the nominal expres-
sion stresses the fact that the decision is the end-product of a mental proc-
ess, the shell-noun expression can still be paraphrased by the simple verbal
form decided:
(9.34) (a) ... the industrial tribunal's decision that the dismissals were unfair
... (PAPERS)
(b) ... it will change its previous decision that the NBA was not against
the public... (PAPERS)
(c) ... Reid said. "Howard made the decision that he wanted to go back
to Everton." (PAPERS)
Of course, 'Aim' and 'Plan' nouns are not completely different from each
other. In addition to considerable parts of their semantic structure, the
nouns share the tendency to create the impression of being autonomous
conceptual entities which seem to be detached from an EXPERIENCER. As
with nouns like goal or objective in the 'Aim' family, this perception is
partly due to the fact that many members of the family are non-derived
nouns: policy, programme, scheme, project, strategy and many others. For
plan itself, the diachronic data indicate that the noun is not derived from the
verb but vice versa (OED, s.v. plan sb. and vb.). The marked tendency for
reification is also evident in the fact that most of these nouns can be used
for concrete entities that are the results of previous mental activities, that is
for the written texts and documents that contain plans, programmes,
schemes and projects. Especially in the texts from media sources that pre-
dominate in the COBUILD corpus, the noun plan itself, for example, occurs
most frequently in collocations like five-year plan, (UN) peace plan, 500-
days-plan, the Government's plan, the American plan, which are all am-
biguous between a concrete and an abstract reading.
An interesting semi-fixed expression involving the noun idea, and to a
much lesser extent also plan, adds positive evaluations to the volitional-
purposive-manner complex typical of 'Plan' uses. This is the extraposi-
tional construction it is a good idea to ... illustrated in (9.35):
(9.35) Test the rig to check all parts will easily pass over the leader knot. It's a
good idea to finish the leader knot off with a coating of Hawser, (MAGS)
'Desire'
Shared semantic features: [MENTAL], [VOLITIONAL], [ACTIVITY], [EMO-
TIVE]

Nouns: Desire, intention, concern, willingness, wish,
dream, intent, inclination, longing, yearning
The majority of the 'Desire' nouns, i.e. desire, concern, wish, dream,
longing and yearning, incorporate emotive features. These nouns straddle
the boundary between volitional and emotive uses because they allow
speakers to do two things at the same time, shell PSYCHOLOGICAL STATES
as AIMS and characterize them as emotional states.
A second aspect which helps to distinguish these uses from those sub-
sumed in the 'Aim' family is that the AIMS they shell do not, at least not
necessarily, represent accomplishments. They can also represent ongoing,
non-conclusive activities. A typical example is given in (9.36):
(9.36) One knows he is in continual pain and that engages sympathy at once,
but thereafter, it is his self-torturing desire to speak truthfully which is
so admirable, (BOOKS)
The activity denoted by the predicate speak truthfully does not involve an
endpoint, and is therefore not an accomplishment (in the technical sense
introduced above), but a manner of carrying out an activity that has no in-
tended termination.
'Desire' nouns differ markedly from 'Aim' nouns with regard to their
collocational behaviour, but not with regard to the type of complement, of
course. While all of them except willingness are found in the pattern N-e-
to as well, it is the pattern N-io that exerts a much stronger attraction on
them. That the use of this image is no exaggeration here can be gleaned
from the reliance scores for the pattern + infinitive clause collected in
Figure 9.5.
216 Mental uses
Noun Frequency in the Frequency in the Reliance

pattern N-to total corpus
willingness 1,804 2,493 72.36%

desire 4,159 9,973 41.70%
wish 821 2,665 30.80%
intention 1,611 6,484 24.84%
Figure 9.5 Selective collocational data on willingness, desire, wish and intention
To recapitulate the significance of these figures, those in the first line mean
that almost three-quarters of all uses of the noun willingness in the 225
million-word corpus are directly followed by infinitive clauses. This objec-
tive quantitative data raises the question why 'Desire' nouns seem to be
earmarked for postnominal infinitives, while 'Aim' nouns are much more
often used in copula constructions with complementing infinitive clauses. It
is likely that the reason again lies in the interplay between the meaning of
the nouns and the major functions of the patterns. As we have seen, the
function of mental shell nouns in the pattern -be-to is to express and iden-
tify IDEAS and PSYCHOLOGICAL STATES. This function is greatly facilitated
when the shell content has the more entity-like status of endpoints of activi-
ties, i.e. accomplishments. Shell nouns that highlight the ongoing dynamic
aspects of PSYCHOLOGICAL STATES of volition, however, lend themselves
less well to this function, since these aspects are not as easily bounded and
identified as mental entities are. This explains why aim, goal, objective and
the like are more frequent in the pattern -be-to, and why expressions like
*his willingness is/was to speak truthfully are not just unusual but un-
grammatical.
Having emphasized the difference between infinitive clauses in the pat-
terns N-/0 and N-e-to, I believe the remarkably close relation between the
noun willingness and postnominal infinitive clauses as such is also worthy
of comment. In the light of the data given in Figure 9.5, the image of a sym-
biosis between the noun and the grammatical construction (as suggested by
the terms reliance and attraction in Section 4.5) does not seem to be out of
place. One possibility of accounting for the closeness of the relation is that
the meanings of the noun itself and the construction match in a sort of
dovetailing manner: apparently, the noun provides only a gap for activities
that are to be carried out; the grammatical construction, on the other hand,
casts cognitive content into precisely that meaning - the meaning of future
events that are described from a volitional point of view. The fact that will-
ingness and infinitive clauses dovetail so nicely can be seen as a corrobora-
tion of the hypothesis that infinitive clauses convey meanings related to
subjective states of wanting to do something (see Section 3.2).
From a syntactic point of view, the data given in Figure 9.5 for -willing-
ness and the other nouns supports the idea that the appositive infinitive
clauses following these nouns must be regarded as complements (see Sec-
tion 3.1.1). Looking once more at example (9.36), in which the infinitive
clause is both semantically and formally completely dependent on the noun,
from this perspective, I think that the use of the notion of complement as
proposed by Huddleston (1984), Herbst (1988) and Biber et al. (1999) is
certainly justified here.
'Reluctance'
Shared semantic features: [MENTAL], [VOLITIONAL], [ACTIVITY], [NEGA-
TIVE]
Frame: Mental; focus on MENTAL STATE or PROCESS

Nouns: Refusal, reluctance, unwillingness, disincli-
nation
This family consists of a small number of antonyms of 'Desire' nouns. They

shell MENTAL STATES of what can be called non-volition. This is illustrated
in (9.37).
(9.37) ... and commercial associations developed that did not act in New Zea-
land's interest when circumstances changed. For example, much of the
meat processing industry was controlled by British capital, and its reluc-
tance to maintain investment and to diversify markets and products as
rapidly as the dairy industry cannot be divorced from this relationship.
(BOOKS; the comma after capital has been added for the sake of clarity,
HJS)
However, speakers can do more than just report that somebody does not
want to do something when they use these nouns. They also often imply that
it would be or have been better for the EXPERIENCE!*, to carry out an activ-
ity, in (9.37) for example the activities of maintaining investment and of
diversifying markets. This semantic component is most marked in the noun
218 Mental uses
refusal. The interesting point about the other three nouns is that it is not
only possible, but even the rule, that the EXPERIENCER to whom the reluc-
tance is imputed by the speaker may not be aware of these expectations. In
using these three nouns, then, speakers covertly express their own point of
view on what an EXPERIENCER should think or have thought, and should do
or have done.
The most frequent noun in this set, refusal, is a special case in a further
respect. It straddles the boundary between shell nouns that describe voli-
tional states and shell nouns that refer to events and acts. In fact, it even has
a claim to be mentioned in the linguistic domain as well. The three domains
merge in the semantic structure of this noun because it can be used to refer
to mental states of not wanting to do something, to acts of showing that and
acts of saying that one does not want to do something. In example (9.38), it
is the volitional element that is most prominent.
(9.38) His approach to life and politics was no more popular than his yellow
socks, and his refusal to kowtow to the prevailing political currents did
not make him a popular figure with the Tory apparatchiks of the period.
(PAPERS)
The four 'Reluctance' nouns also boast very high reliance scores for the
pattern + infinitive clause: disinclination 72.5%, unwillingness 71.7%,
refusal 61.81% and reluctance 54.34%. The only other pattern in which
they were found was the pattern th-N, but in terms of frequency these uses
are marginal compared to those in the pattern N-fo. These nouns are another
outstanding manifestation of the semantic dovetail between volitional shell
nouns and infinitive clauses. The noun refusal is particularly significant
here, uniting mental, linguistic and eventive aspects related to volition.
'Purpose'
Shared semantic features: [MENTAL], [VOLITIONAL], [DETACHED]

Nouns: Idea, purpose, function
This is yet another family in which idea tops the list of nouns because of its
general frequency of occurrence. I will focus on the other two members
here, as the use of idea is very similar to the way they are used.
Together with the partitive uses discussed in Section 7.6, these two
nouns belong to those whose representation in the dataset is most dramati-
cally reduced by the rigidity of the corpus queries (see Section 4.4). Like the
nouns part, feature and characteristic, purpose and function have a marked
relational meaning which is reflected on the linguistic surface in the fact that
the majority of their uses as shell nouns are followed by inserted of-
prepositional phrases expressing the entities to which a purpose or function
is attributed. Examples (9.39) and (9.40) are two cases in point:
(9.39) The purpose of this leaflet, the enclosed Forecast Sheets and Year-End
Checklist is to provide you with the relevant business planning tools to
help you secure the many benefits that on-going planning can provide.
(EPHEM)
(9.40) The most important function of the first aider is to ensure that the
casualty's air passages remain open, (PAPERS)
Semantically, the two nouns purpose and function are also related to the
'Aim' family, but different enough to warrant the establishment of a sepa-
rate family. To a certain extent, the two 'Purpose' nouns also shell AIMS,
but what they shell are not envisaged accomplishments of the
EXPERIENCERS themselves. Instead, they shell intended effects that are at-
tributed to concrete entities, events or abstract relations by EXPERIENCERS.
The function of a tool, for example, is what its inventor has intended it to be
good for. The purpose of an activity is the outcome intended by the agent
(EXPERIENCER). In (9.40) above, the function attributed to an abstract rela-
tion, i.e. to the role of acting as first aider (and not the concrete human be-
ing), is an expected type of behaviour. These examples show that an enor-
mous variety of conceptual entities can serve as carriers of the intended
effect (typically expressed by o/-phrases) and as AIMS (expressed by infini-
tive clauses).
'Solution'
Shared semantic features: [MENTAL], [VOLITIONAL], [RESULTATTVE]

Nouns: Solution, key, remedy, cure
220 Mental uses
Another small set of nouns, those of the 'Solution' family, are counterparts
to the 'Realisation' nouns in the creditive domain. Both families consist of
nouns that are used to shell PSYCHOLOGICAL STATES as results of acts of
thinking. The difference is that 'Realisation' nouns shell creditive states,
while the four 'Solution' nouns shell volitional states - at least in the ma-
jority of their uses. In fact, the two more frequent nouns solution and key
are also found with complementing ^-clauses. In these collocations (see
9.41), they shell creditive states as results of mental activities.
(9.41) In their report on the future of public libraries yesterday Comedia re-
searchers highlight the widespread nature of this failure of imagination.
Their solution is that a new organisation is urgently needed, to speak
for and promote the public library network at the national level.
(PAPERS)
(9.42) Once again the Government has decided to do something about under-
age sex. This time, rather than brandishing condoms in the classroom or
publishing an explicit sex guide for teenagers, it is suggesting the solu-
tion is to turn teachers into classroom spies, (TODAY)
(9.42) exemplifies the more frequent combination with infinitive clauses.

The juxtaposition of the two types profiles the difference between them. The
-clause in (9.41) represents a claim or statement of BELIEF, while the
infinitive clause in (9.42) represents an intended activity or even accom-
plishment. (9.41) is about 'knowing' and (9.42) about 'doing'.
The four 'Solution' nouns share as part of their semantic structures the
implication that EXPERIENCERS had to grapple with a problem before ar-
riving at a solution, finding the key or coming up with a remedy or cure for
it. As (9.42) has shown, real-word activities (described in the shell content)
tend to be the result of this problem-solving type of mental activity. So the
nouns straddle the boundary between mental and eventive uses. While solu-
tion and key are more on the mental side, remedy and cure tend towards the
eventive end of the scale.
'Motivation'
Shared semantic features: [MENTAL], [VOLITIONAL], [CAUSAL]
Nouns: Incentive, urge, temptation, impulse, com-
pulsion, motivation, inspiration, motive,

inducement, preoccupation, vocation
Speakers use the 'Motivation' nouns to shell mental reasons for states of
volition. The nouns allow speakers to raise and answer the question of why
an EXPERIENCER wants to achieve something. In (9.43), for example, the
speaker portrays the AIM to make headlines rather than find solutions as
the reason why the EXPERIENCER (Mr Baker) did something (which is not
mentioned in the quoted passage but could have been expressed by inserting
ayr-prepositional phrase after the shell-noun phrase).
(9.43) He said Mr Baker's prime motivation was "to make headlines rather
than find solutions" ... (TODAY)
With almost all members of the family, the shell contents that are expressed
by infinitive clauses represent the activities and especially accomplishments
pursued by the EXPERIENCERS. This applies both to complementing clauses
(as in 9.43) and to postnominal clauses, which are more frequent than the
former. The exceptions are the two fairly rare nouns preoccupation and
vocation, which only occur as shell nouns in copula clauses with comple-
menting infinitives.
As far as their stable inherent meaning is concerned, the nouns in the
'Motivation' family can be subdivided as follows:
1. Nouns that locate the reason for a volitional state in the world
outside the EXPERIENCER
l.a ... in the behaviour of other people: incentive, inducement
1.b ... as being just out there: temptation, inspiration
2. Nuns that locate the reason for a volitional state in the psychol-
o g y o f t h e EXPERIENCER
2.a ... as a temporary psychological state: urge, impulse

2.b ... as a longstanding psychological disposition: instinct, compul-
sion, preoccupation, vocation
3. Nouns that are neutral with regard to the imputed location of the
reason: motivation, motive
222 Mental uses
Of course, more specific features are also at work, for example in the noun
temptation the element that EXPERIENCERS feel that they should better not
pursue the things they want to achieve. Nevertheless the dimensions outlined
above provide a good grid for the differentiation of the nouns. It should be
added that instinct also supplies the 'Belief and the 'View' families in the
creditive domain. This multiple occurrence reflects the ambiguity that one
can have an instinct about whether something is true or not and about what
to think about an issue, but also an instinct about what one is going to do.
The noun motivation occurs in the pattern N-to in the function of the 'De-
termination' family, to which I will turn now.
'Determination'
Shared semantic features: [MENTAL], [VOLITIONAL], [CONDITIONAL]
Nouns: Determination, confidence, courage, strength,
readiness, energy, resolution, heart, nerve,
anxiety, eagerness, resolve, flexibility, resis-
tance, passion, motivation, temerity, wit,
cheek, decency, foresight, gall, impetus,
audacity, stamina, grace, zeal, keenness,
willpower, courtesy, effrontery, nous, gump-
tion
The nouns in this family allow speakers to portray PSYCHOLOGICAL STATES

and dispositions as being helpful, or even necessary, for the achievement of
an AIM. An example of the most frequent noun determination is given in
(9.44):
(9.44) His determination to oust Dr Mahathir and stop him winning his third
consecutive term in office is matched only by the Prime Minister's mis-
sion to annihilate the opposition once and for all. (BBC)
As the list of nouns retrieved from the corpus shows, a large variety of
qualities are used in this function. Since the nouns refer to traits of
EXPERIENCERS, it is not surprising that many of them are derived from ad-
jectives that describe such qualities, e.g. confident, strong, ready, anxious,
eager, flexible, decent, audacious and keen. Other nouns, e.g. determina-
tion, rsolution, resolve and resistance are derived from verbs that are
mainly used as participles in adjectival function (be determined, resolved,
resistant). Some of the nouns allow the speakers to impute interesting im-
plications as to the impact of the quality on themselves: temerity, cheek,
gall, and effrontery implicate that the speakers do not approve of the psy-
chological states that they impute to EXPERIENCERS. A neat illustration of
such an implication can be found in (9.45):
(9.45) James Patrick even had the gall to ask about the existence of a will, bold
as brass, as if he had the right to know, (BOOKS)
While the positive connotations of some nouns, e.g. determination, cour-

age, strength, energy or flexibility, are mainly due to the fact that the
qualities they describe are accepted and appreciated in Western cultures,
there are also 'Determination' nouns which have a positive evaluation in-
herent in their basic denotative meanings. These are the nouns wit, decency,
grace, courtesy, nous and gumption. A good example of this type of use
which even includes two of the nouns in apposition can be found in (9.46):
(9.46) Heseltine had the grace, or the political nous, to admit that he was
shocked by the reaction of Middle England, (BOOKS)
The favoured lexico-grammatical environment of the 'Determination' nouns

is clearly the pattern N-to. This tendency is reflected in a number of re-
markably high reliance scores for this pattern: temerity can boast a score of
73.75%, readiness 49.41%, determination 35.51%, keenness 32.33% and
audacity 24.91%.
In spite of their interesting implications, their frequency (both in terms of
types and tokens) and the impressive reliance scores of some nouns, the
members of the 'Determination' family do not belong to the prime or even
good shell nouns. For one thing, the occurrence of these nouns as shell
nouns is largely restricted to combinations with the function verb have. This
is also reflected in the examples that I have quoted, only one of which (9.44)
does not manifest this pattern. As a consequence, the 'Determination' fam-
ily is also deficient with regard to the relation of experiential identity that
they set up. Rather than by linguistic elements expressing identity, the rela-
tion between shell noun and shell content would have to explicated by inser-
tions like which is required to be able to or which is helpful to be able to.
See for example (9.46'):
224 Mental uses
(9.46') Mr Heseltine had the grace which is required to be able to admit that...
The possibility of such paraphrases indicates that the link between shell
nouns and shell contents involves much more than pure experiential identity.
'Agreement'
Shared semantic features: [MENTAL], [VOLITIONAL] or [CREDmVE],
[INTERSUBJECTIVELY SHARED]

Nouns: Agreement, deal, contract, consensus, ar-
rangement, appointment, compromise, pact,
bargain
Up to now, it was the rule that speakers imputed one IDEA or

PSYCHOLOGICAL STATE to one single EXPERIENCER or to one single group
of EXPERIENCERS; the relation was always one-to-one. The 'Agreement'
nouns deviate from this principle, attributing one IDEA or PSYCHOLOGICAL
STATE to several EXPERIENCERS, which are not thought of as being one
group. Instead, the implication that the EXPERIENCERS were previously in
different PSYCHOLOGICAL STATES with regard to an issue or activity is a
crucial aspect of the semantic structure of these nouns. The essence of what
speakers can shell with these nouns is contained in a nutshell in the original
Latin meaning of the noun consensus or, even better, of the more basic verb
consentire, which consists of the prefix con- 'together' and the base sentire
'to feel, think, judge'.
In principle, the four nouns agreement, deal, arrangement and com-
promise have the potential to produce both volitional and creditive shell
noun uses. Previously disagreeing EXPERIENCERS can decide to agree on
what they want to do or how they are going to do something on the one
hand, or on what they believe on the other. (9.47) is an illustration of a vo-
litional use of the noun agreement, and (9.48) an illustration of a creditive
one.
(9.47) It is a great help if you can couple a membership with a COVENANT,

which is an agreement to pay the Association an amount annually for at
least 4 years, (EPHEM)
(9.48) And that really explains the Countiyside Commission's cautious ap-
proach to change and our agreement that the right step forward is the
New Forest Committee and making that as effective as possible, (EPHEM)
The shell content in (9.47) represents an intersubjectively shared AIM, and

that in (9.48) and intersubjectively shared BELIEF. In the terms proposed by
Dirven (1991: 29) for the description of the verb agree, (9.47) corresponds
to 'comply to act', and (9.48) to 'concur in thinking'.
It will not have gone unnoticed that in (9.47), the example of a volitional
use, the shell content was expressed by an infinitive clause. This is of
course characteristic of these uses. As in other families, Ao-clauses can be
required either because the EXPERIENCER and the subject of the subordinate
clause are not coreferential, or, as in (9.48), because a creditive rather than
a volitional state is shelled.
Although the four nouns mentioned above can occur with either type of
clause, they exhibit marked preferences: agreement and deal typically and
frequently occur with infinitives, while arrangement and compromise,
which, one should add, are much less frequent, are more often found with
/Aai-clauses. The other five members of the family also have a bias for one
of the two patterns and the semantic functions that they typically serve;
contract, appointment, pact and bargain shell volitional states, while con-
sensus is earmarked for creditive states:
(9.49) Mr Qin has signed a contract to rent the stretch of pavement from Pe-
king municipality at about 70p per square metre a month, (PAPERS)
(9.50) He wouldn't perform at Axl Rose's thirtieth-birthday bash (Rose is a big
Nirvana fan) and turned down a spot on this summer's Metallica-Guns
N" [sic!] Roses tour. Still and all, "the general consensus is that Nir-
vana should quit." says Bjelland. (MAGS)
While pact and appointment are only found in the pattern N-cl, compro-
mise is only, and arrangement predominantly, found in the pattern N-e-cl.
Especially the nouns agreement, deal, contract, arrangement and pact are
also very frequent in the pattern th-N. As with such nouns as plan, pro-
gramme and project in the 'Plan' family, it can be argued that the wide-
spread use in anaphoric references has to do with the fact that the referents
of these nouns tend to be conceptualized as autonomous IDEAS or even con-
crete entities.
226 Mental uses
9.3.4 Emotive uses
Emotive shell nouns denote yet another type of psychological state, namely
states of emotion. They can be divided into two families, 'Surprise' and
'Fear', which can be related to creditive and volitional uses respectively.
Like creditive states, the emotional states shelled by uses in the 'Surprise'
family are linked to IDEAS representing facts, especially past events. And
like volitional states, the emotional states shelled by uses in the 'Fear' fam-
ily have to do with future events. The members of the two families share the
resultative feature that the emotional states they denote are portrayed as
results of IDEAS.
'Surprise'
Shared semantic features: [MENTAL], [EMOTIVE], [FACT-RELATED]

Nouns: Surprise, relief, pleasure, shock, regret, pain,
disappointment, anger, delight, satisfaction,
pride, guilt, joy, frustration, resentment, hap-
piness, sadness, terror, rage, astonishment,
irritation, grief, bitterness, amazement, fury,
sorrow, gratitude, annoyance, thrill, disgust,
indignation, grudge
Boasting a long list of nouns, 'Surprise' uses nevertheless are of relatively

marginal importance for the study of shell nouns. For one thing, the major-
ity of nouns are only common in the pattern N-cl - only concern, disap-
pointment, frustration, grudge, regret, relief, satisfaction, shock and thrill
were also found in the pattern N-e-cl - and in these uses, it is often more
than doubtful whether a given that-clause actually functions as an apposi-
tive postmodifier or not (see also Francis 1993: 151). Example (9.51) is a
case in point.
(9.55) President Bush has already been in contact with Mr Major and Ameri-
cans are expressing relief that a quick transition has been achieved with
no dramatic change on Gulf policy expected, (BBC)
Even the mundane observation that relief has no determiner should make us
wary, because according to Quirk et al. (1985: 1261), noun phrases with
appositive postmodifiers tend to be determined by the definite article. Closer
examination suggests that what we are confronted with here is an expanded
predicate express relief with an expressive illocutionary meaning equivalent
to 'say with relief; this predicate is complemented by a that-ciause.
Similarly unconvincing examples are represented by other superficially
coinciding constructions (see Sections 3.1.1 and 4.3) caused by inserted
adverbials in extrapositional constructions (e.g. It was with relief that we
climbed aboard, BOOKS) or other constructions (e.g. He noted with regret
that the squad was still predominantly..., PAPERS).
All the same, 'Surprise' nouns do occur as mental shell nouns when they
occur in the pattern N-e-cl, and that is why they are mentioned here. They
are used to construe PSYCHOLOGICAL STATES of emotion as results of the
knowledge of facts. Francis (1993: 151) looks at the same relation from the
complementary perspective, claiming that "what the -clause supplies is
the cause of the feeling". This constellation links these uses to the causal
uses in the factual domain (see Section 7.3), which are used to shell facts as
causes. The nouns themselves, however, denote mental states. The facts that
the EXPERIENCERS are aware of are expressed as -clauses in the pattern
-be-that. An example is given in (9.52):
(9.52) Giles smiled to himself. His one regret was that Clare had not been on
hand to witness his newly-honed technique, (BOOKS)
In this example, it is the fact that the EXPERIENCER Giles knows that Clare
had not been on hand ... that functions as the cause of the emotional state
of regretting imputed to him by the speaker. The example indicates that the
uses subsumed in the 'Surprise' family are not only related to facts, but
even factive in the technical sense of this term (see Section 7.2) because
they presuppose the knowledge of the facts. In contrast to the factivity of
the neutral nouns discussed in Section 7.2, the factivity of these nouns can
frequently be proved with the negation test: the truth of the proposition ex-
pressed in the that-clause in (9.56), Clare had not been at hand ... is not
affected when the main clause is negated, yielding His one regret was not
that Clare had not been on hand to witness ....
The members of the 'Surprise' family can be divided semantically into
four major categories as suggested in Figure 9.6:
228 Mental uses
Resulting
emotional state: Positive emotion Negative emotion
Cause:
Any kind of fact delight, gratitude, hap- anger, annoyance, bitter-
piness, joy, pleasure, ness, disappointment, dis-
relief, satisfaction, gust, frustration, fury, grief
thrill grudge, indignation, irrita-
tion, pain, rage, resent-
ment, sadness, shock, sor-
row, terror
A fact for which the pride, satisfaction concern, guilt, regret

EXPERiENCER feels re-
sponsible
Figure 9.6 The semantic differentiation of 'Surprise' nouns
A fifth subcategory consists of the nouns amazement, astonishment and

surprise which are neutral with respect to the polarity positive-negative.
Surprise, for example, collocates with pleasant just as well as with nasty.
These three nouns share the feature that the fact is portrayed as being unex-
pected for the EXPERIENCER by the speaker. A similar element, conflated
with a resulting negative emotion, can be postulated for disappointment,
frustration and shock.
'Fear'
Shared semantic features: [MENTAL], [EMOTIVE], [EVENT-RELATED]

Nouns: Fear, concern, worry, anxiety, apprehension,
dread, premonition, reservation, disquiet
The states of emotion shelled by 'Fear' nouns are negative rather than posi-
tive, and they are caused by IDEAS representing possible fixture events. The
nine nouns differ principally with regard to how strong the emotion imputed
to the EXPERIENCER is. Dread, fear and anxiety cover the strong end of the
scale, reservation and disquiet the weak end, while concern, worry and
apprehension hold a middle position. Premonition is neutral with regard to

the strength of the emotion, but suggests that the EXPERIENCE!*, does not
fully realize why he or she feels bad about something. Like disquiet, it is a
fairly rare noun.
The 'Fear' nouns occur mainly in two variants. If the EXPERIENCER is
also the protagonist of the hypothetical event that gives rise to his or her
negative emotions, this is expressed by means of noun phrases that are
postmodified by o/-prepositional phrases with /g-forms as prepositional
complements. See example (9.53) for an illustration of this type.
(9.53) The fear of contracting AIDS was very real, (EPHEM)
Obviously, these constructions exhibit a similarity to infinitive clauses in

combination with volitional uses: the EXPERIENCER also has to coincide
with the implicit subject of the participle, and the whole constructions also
express future events affecting the EXPERIENCER. They differ, however, in
that the EXPERIENCERS tend to be patients or play other non-agentive roles,
because the events that give rise to their negative feelings are beyond their
control.
If the EXPERIENCER is affected by the event in such an indirect way that
he or she cannot function as implicit subject, speakers use -clauses to
render the future events.
(9.54) This, incidentally, is a true story, not a parable, but I do not want to
divulge where the village is for fear that it will be over-run, (MAGS)
All nouns except apprehension, dread and disquiet were also found in the
pattern -be-that, where they produce good examples of shell nouns. See
(9.55) for an illustration:
(9.55) If we have any more coming along then that would be very worrying
indeed. Our concern would be that the services would be overloaded.
(TODAY)
Especially the four frequent nouns fear, concern, worry and anxiety are
also widely found in the pattern th-N. Finally, the extremely rare but nev-
ertheless authentic and acceptable collocation of the noun concern in copula
constructions with w/?-clauses as complements should be mentioned here.
An illustration of this rare collocation is given in (9.56):
230 Mental uses
(9.56) Investor's concern is just how many retailers will be around to tackle
this intensely competitive market place, (PAPERS)
Chapter 10
Modal uses
10.1 Introduction
Given that a large variety of issues have to be addressed in this study, it will
be understood that I cannot even begin to discuss the question of how the
notion of modality can be defined.70 Instead I will assume that this whole
chapter will be taken for an extensional definition of what I have in mind
when I speak of modal shell nouns.
It will suffice as a starting-point to draw attention to the crucial feature
that distinguishes modal uses from those that I have dealt with up to now. If
one neglects cases of conceptual shifts, the manifestations of shell nouns
that I have been looking at so far involved straightforward matches between
shell nouns and shell contents: factual shell nouns shell facts, linguistic shell
nouns shell utterances and mental shell nouns shell ideas. One cannot say,
however, that modal shell nouns shell 'modalities', because the shell con-
tents of modal uses also represent either facts or events. Speakers thus do
not use modal shell nouns to shell 'modalities', but to express their subjec-
tive judgments as to the degree to which it is possible, probable or certain
that the facts they shell are true or the events they shell will take place.
The nouns that can be found in modal shell-noun uses are of course re-
lated semantically to modal verbs. The clearest sign of this relationship is
that virtually all descriptions of modal verbs abound with metalinguistic
uses of nouns like possibility, necessity, permission, obligation and ability.
Despite this obvious connection - and despite the interesting fact that de-
scriptions of modal verbs are thus made dependent on nouns whose mean-
ings are taken for granted rather than explicitly defined - I have not found a
single linguistic study devoted to nouns that express modality. Only Halli-
day (1994: 267-268) mentions "nouns of modality" and "nouns of modula-
tion" briefly, but does not look at them in any detail.
From a morphological point of view, modal shell nouns are not related to
modal verbs but to adjectives expressing modal meanings. The most notable
exception to this rule is the noun need\ less frequent counterexamples are
requirement, permission, pressure and assignment. Much more typical,
however, are pairs like possible - possibility, necessary - necessity, true -
232 Modal uses
truth, certain - certainty, real - reality, able - ability, free - freedom or

responsible - responsibility.
Both types of relations, the semantic one to modal verbs and the mor-
phological one to adjectives expressing modal notions, can be exploited for
the description of modal shell-noun uses. Previous studies of modal verbs
will help to categorize modal shell-noun uses, while the adjectives will be
useful when it comes to paraphrasing and explicating the meanings of mo-
dal shell nouns.
In one respect, modal nouns are easier to handle than modal verbs: their
meanings are less ambiguous and indeterminate than those of modal verbs
(see e.g. Coates 1983: 14-18, Palmer 1990: 20-22). The verb must, for
example, can have a deontic reading, as in (10.1), and an epistemic reading,
as in (10.2) (both examples are taken from Sweetser 1990: 49). If the utter-
ances in (10.1) and (10.2) are paraphrased with shell-noun expressions, as
in (10.1') and (10.21), the difference between the two readings emerges
automatically as it were:
(10.1) Deontic: John must be home by ten; Mother won't let him stay out any
later.
(10.2) Epistemic: John must be home already; I see his coat.
(10.1') Deontic: John has the obligation to be home by ten.
(10.2') Epistemic: There is a good chance that John is at home already.
Although a small number of nouns - among them possibility and chance -

are found both in epistemic and in deontic uses, such ambiguities cause
fewer difficulties here than in the verbal domain.
On the downside, the class of modal shell nouns is even harder to delimit
than that of modal verbs. Unlike at least the core of the class of modal
verbs, modal shell nouns cannot be distinguished from other nouns on the
basis of formal criteria. Resorting to semantic criteria, one finds that opin-
ions as to what should be included in the domain of modality differ to an
extent which is perhaps only outstripped by the diversity of typologies of
modality that have been proposed. In Figure 10.1, a selective survey of the
types of meanings that are included by a few of the many scholars who have
worked on English modal verbs is given. The types are ordered according to
the way they are treated by these linguists:
Introduction 233
Lyons (1977: Necessity, possibility, factivity, obligation, permission, pro-

787-849) hibition and exemption
Coates (1983) Obligation, necessity, ability, possibility, epistemic possibil-
ity, volition, prediction
Quirk et al. Permission and possibility/ability, obligation and necessity,
(1985: 219-236) volition and prediction
Palmer (1990) Epistemic possibility and necessity, deontic possibility and
necessity, dynamic possibility and necessity, conditionals
Sweetser (1990: Root modality, epistemic modality, speech-act modality
49-75)
Halliday (1994: MODALITY: probability, usuality
89-92) MODULATION: obligation, inclination
Mindt (1995: 45- Possibility/high probability, certainty/prediction, ability, hy-
55) pothetical event or result, habit, inference/deduction, obliga-
tion, advisability/desirability, volition/intention, necessity,
politeness/downtoning, consent, state in the past, permission,
courage, regulation/prescription, disrespect/insolence
Figure 10.1 A selective survey of types of modal meanings
Adding to this already confusing picture the modes of modalities proposed

by philosophers and logicians - for example von Wright's (1951: 1-2) divi-
sion into alethic, epistemic, deontic and existential modality, or Rescher's
(1968: 24-26) framework, which includes things like temporal, boulamaic
(i.e. related to fears), evaluative, causal and conditional modality in addi-
tion to alethic, epistemic and deontic - one gets the impression that virtu-
ally nothing that I have been dealing with in this study is not modal, or at
least cannot be regarded as modal. For example, large sections of Chapter 7
involve factivity (see Lyons) or /actuality (a term used by Palmer 1986: -
8); directive and commissive uses as discussed in Chapter 8 undoubtedly
have to do with deontic modality; predictive assertive uses (see Section
8.3.1) and of course all creditive uses described in the previous chapter (see
Section 9.3.1) have a marked epistemic component; and most families of
volitional uses, for example 'Desire', 'Reluctance' and 'Aim', have to do
with notions like volition and inclination, which are not only included in the
extraordinarily wide conception of modality held by Mindt (1995), but also
in less prolific proposals (see Coates 1983, Quirk et al. 1985, Halliday
1994).
234 Modal uses
The reason why I have not treated all these types of nouns as modal shell
nouns is that I have focused on what I regard as their semantic core. Of
course it is true that directive nouns like order and command and commis-
sive nouns like promise and oath are used to place people under obligations;
first and foremost, however, all these four nouns denote speech acts, and
therefore they were treated as illocutionary shell nouns. Similarly, knowl-
edge, belief and expectation were regarded as nouns denoting psychological
states, rather than nouns expressing degrees of epistemic certainty. And by
the same token, volition as expressed by such nouns as desire, willingness
or intention was looked upon as a mental rather than a modal notion. It is of
course no coincidence that there is such a marked similarity between the
uses of these linguistic and mental shell nouns and the notion of modality:
both the nouns discussed in other chapters and modality as such have to do
with subjectivity, i.e. with reflections of speakers' "opinions and attitudes"
(Lyons 1977: 452; see also Palmer 1986: 16-17, 102-103).
I have compiled Figure 10.1 mainly in order to give an idea of the vari-
ety of accounts of English modal verbs that is available. However, the fig-
ure contains two further pieces of information that are useful here. First, if
we leave radical proposals aside and neglect terminological predilections,
there seems to be some sort of consensus that epistemic and deontic modal-
ity make up the core of the domain of modality, at least from a linguistic
point of view. On the reasonable assumption that modality as expressed by
verbs is somehow related to modality as described by nouns, these two
types should certainly also be represented here.
Second, Figure 10.1 shows that there are already many different opin-
ions on the more marginal types of modality. Nevertheless, the data require
that I add yet another one. Since none of the linguists mentioned here have
addressed the issue of modal nouns, it is not surprising that none of the
typologies suggested by them completely fits the data. The framework that
lends itself most readily to the description of my collection of nouns is the
one suggested by Palmer (1990). In addition to deontic and epistemic mo-
dality, Palmer's system includes a third "type" of modality called dynamic
modality,71 which is very useful for the description of such examples as a
unique opportunity to cut interest rates (PAPERS) and Exmoor is a won-
derful place to view stag (PAPERS). As for "degrees" of modality (Palmer
1990: 36-37), my data suggests the three degrees of possibility, probability
and necessity.12 Yet even this amendment and the 3 3 matrix of families
resulting from it do not suffice to categorize modal shell-noun uses in a
satisfactory way. One further category must be added which, however, is
Introduction 235
also implicit in Palmer's work (1990: 83-84), where "neutral" uses of the
modal verb can are distinguished from those expressing ability. According
to Palmer (1990: 83) neutral can simply indicates that an event is possible,
while ability is subject-oriented. It is precisely on this dimension that nouns
like opportunity and occasion differ from ability and capacity, and there-
fore we end up with four instead of three families of dynamic uses.
The result of this eclectic transfer from the rich findings of other lin-
guists on modal verbs to modal shell nouns is the survey of groups and
families given in Figure 10.2.
Epistemic uses 'Possibility' possibility, chance, danger 10.2

'Probability' probability, likelihood
'Certainty' certainty, truth, reality
Deontic uses 'Permission' permission, right, freedom 10.3
'Job' job, duty, role, task, mission
'Need' need, pressure, necessity
Dynamic uses 'Ability' ability, capacity, potential 10.4
'Opportunity' opportunity, chance, occasion
'Tendency' tendency, trend, propensity
'Destiny' destiny, fate
Figure 10.2 Survey of groups and families of modal shell-noun uses
As far as the frame for the description of modal shell-noun uses is con-
cerned, a generally applicable frame can only be suggested at the cost of
considerable abstraction. Such a general frame consists of two components,
a MODALITY INDICATOR and a PROPOSniON. Depending on the kind of
modality, the PROPOSITION component takes different forms, and optional
components can be added. More specific elaborations of this frame to be
suggested for the description of epistemic, deontic and dynamic uses will be
given at the beginnings of the three following sections.
10.2 Epistemic uses
Epistemic modality concerns judgments about the possibility, probability or

certainty that something is or is not the case. Even this short description,
236 Modal uses
which is borrowed almost verbatim from Palmer (1990: 50), suggests that
the PROPOSITION component of the schematic modal shell-noun frame is
elaborated as a fact or possible fact here. The result is the more specific
frame that is represented in Figure 10.3. The figure also contains an exam-
ple which shows how the frame is mapped onto a clause.
Epistemic modality frame: MODALITY INDICATOR - (POSSIBLE) FACT
Example:
MODALITY FACT
INDICATOR
But it is the possibility that some of Iraq 's vast 10 billion Dollars
nuclear weapons programme has escaped
which provokes the most unease (PAPERS,)
Figure 10.3 Frame for epistemic shell-noun uses
It is typical of all modal shell-noun uses that the MODALITY INDICATOR is

contained in the modal shell noun. The fact functioning as PROPOSITION
(and shell content) is expressed by means of a that-clause. This is the rule
for epistemic uses in the patterns N-cl and N-e-cl.
The three families of epistemic uses differ with regard to what I (with
Palmer) call degrees of modality. The three degrees can easily be gleaned
from the names of the families, 'Possibility', 'Probability' and 'Certainty'.
'Possibility'
Shared semantic features: [MODAL], [EPISTEMIC], [POSSIBLE]
Frame: Epistemic modality
Nouns: Chance, possibility, danger, option, risk,
uncertainty
The noun chance is not at the head of this list because it is the most fre-
quent 'Possibility' noun, but because it is extremely common in dynamic
uses with infinitive clauses (see the 'Opportunity' family). A second noun
that is more often found in other uses, namely in the 'Option' and the 'Way'
families in the fields of eventive and circumstantial uses (see Sections 11.3
Epistemic uses Til
and 12.3), is option. The other four nouns are all earmarked for epistemic
uses, and therefore it is on them that I will focus my attention.
The noun possibility itself provides a neutral conceptual shell for possi-
ble facts. One typical example has already been provided in Figure 10.3. As
a basis for the following discussion, a second example with a postnominal
that-clause, and one with a complementing that-clause, are given in (10.3)
and (10.4):
(10.3) The Granada move has also focused attention on the possibility that
other companies will become targets for stake-building, (PAPERS)
(10.4) He says one possibility is that FG Sagittae has recently ejected a cloud
of dust that is absorbing the star's light, (NEWSCI)
Like epistemic uses of the modal verb may (Palmer 1990: 51), both exam-
ples can be paraphrased with expressions containing the adjective possible.
But they call for different constructions, which are suggested in (10.3') and
(10.4') respectively.
(10.3') It is possible that other companies will become targets for stake-
building. The Granada move has focused attention on this.
(10.4') What is possible is that FG Sagittae has recently ejected a cloud of dust
that is absorbing the star's light.
Uses of possibility in the pattern -that are paraphrased best by extraposi-

tional clauses, as in (10.3'). If the semantic and pragmatic impact of the
paraphrase is to match that of the original, it is necessary to add the infor-
mation contained in the rest of the clause by means of an additional clause.
To match the focusing function of the pattern -be-that (see Section 16.1),
these uses are best paraphrased with W/J-clefts, as is suggested in (10.4'). In
these paraphrases, the grammatical construction can be left intact.
Initially, these paraphrases may not seem to be extraordinarily revealing;
however, the principle underlying the paraphrases has a wide field of appli-
cation. It can be applied to the semantically more complex nouns in the
'Possibility' family and even to all types of epistemic uses. The additional
semantic components included in the nouns with richer meanings can be
added as further comments, while the basic structures of the paraphrases
are the same. This is shown for two examples of the noun danger :
238 Modal uses
(10.5) There is a danger that the youthful zest is being sapped out of her and
her tennis as it was for Tracy Austin and Andrea Jaeger before her.
(PAPERS)
(10.6) Behind the scenes, solid work is being carried out in cutting costs, capi-
tal investment and realising synergy benefits; in the fullness of time, this
will bear fruit. The danger is that the damage done to the stock's image
in the meantime will be so debilitating that even when the upturn comes,
the shares will be unable to recover their rating, (ECON)
(10.5') It is possible that the youthful zest is being sapped out of her and her
tennis as it was for Tracy Austin and Andrea Jaeger before her, but this
would not be good.
(10.6') What is possible is that the damage done to the stock's image in the
meantime will be so debilitating that even when the upturn comes, the
shares will be unable to recover their rating, but this would not be good.
The paraphrases reveal that the noun danger is used to shell future events
as possible facts and to characterize them as unwelcome. In a way, then,
danger is the modal counterpart to problem, the latter is a shell for nega-
tively evaluated facts, and the former for negatively evaluated possible
facts.
It seems that in many uses of danger, this noun can be replaced by risk
without a noticeable change in meaning. But this is not always possible. In
(10.5), the use of risk would imply that there is a person who is aware of
the danger and somehow responsible for potential bad results. In example
(10.6), risk would sound decidedly odd. The reason is that what is described
as a danger in (10.6) is not a result of the activities described in the sen-
tence before. If risk were used, however, such unwarranted inferences
would certainly be drawn. What these two examples show is that risk im-
plies that an agent deliberately accepts a possible negative outcome of his or
her activities. This emerges nicely from the short passage in (10.7), where
the agent and his plans are first explicitly stated, and then the possible out-
come is shelled as a risk.
(10.7) Mr Major is going for gold; but the risk is that if the economy goes
from bad to worse, he could be disqualified altogether, (PAPERS)
Only possibility itself produces large numbers of good shell-noun uses in

the pattern -that. Danger and risk occur in the pattern almost exclusively
Epistemic uses 239
in existential i/iere-constructions, as exemplified in (10.5). Uncertainty does

not occur in the pattern -that at all.
Possibility contrasts with danger and risk with regard to how uses in the
pattern -be-that are embedded in the textual structure. On the lexico-
grammatical level, this is reflected in their typical determiners and premodi-
fiers. In the pattern -be-that, possibility occurs almost never with definite
articles, but usually with determiners and premodifiers which establish
comparative reference (Halliday and Hasan 1976: 76-87), such as one,
another, the second, the extreme and the most alarming possibility. This
indicates that collocations of the type possibility be that-clause are normally
used to characterize possible facts as one of several possibilities, others of
which are mentioned in the preceding or following context by whatever
lexico-grammatical means. This meaning is expressed more explicitly in the
semantic structure of the noun option (as e.g. in One likely option is that
he will play a key role in ..., PAPERS). Uses of risk and danger in the pat-
tern ~N-be-that, on the other hand, are almost always determined by the defi-
nite article. The frequent use of this determiner is partly motivated by the
cataphoric reference to the shell content (Quirk et al. 1985: 268), and partly
by the inferences related to the preceding context. The resultative compo-
nent inherent in the noun risk, which was illustrated in example (10.7),
gives rise to one type of such inferences. And in example (10.6), the possi-
bility characterized as a danger is also not unrelated to the content of the
preceding sentence, even though there is no resultative relation here, but a
more complex, concessive one. The preceding sentence expresses a set of
activities, and the sentence that contains the noun danger a possibility that
can become reality in spite of these activities.
All members of the 'Possibility' family also occur in the patterns th-
and th-be-N, in the pattern th-N fairly frequently in fact. Two typical ex-
amples are given in (10.8) and (10.9):
(10.8) Or perhaps you saw yourself as a poor drab little thing, working all the
hours the good Lord gives, with no reward, and no money. Consider
that latter possibility carefully, (NEWSCI)
(10.9) But will these measures stop scandals like Pindown. Lady Howe chaired
the enquiry. [Recording starts] There is no ultimate sanction this will
not happen again but what we're suggesting we hope will really reduce
that risk... (SPOKEN)
240 Modal uses
(10.8) is a very straightforward illustration of how the noun possibility can

be used in anaphoric noun phrases to create possible facts. No matter what
the proposition you saw yourself as a poor drab little thing represents
when it is viewed in isolation - perhaps a fact or an idea - it is turned into a
possibility by the use of this noun. It must be noted that these uses are more
similar in meaning to those in the pattern -be-that than to -that, in that
they normally refer to one of a number of possibilities rather than the possi-
bility as such. They involve a comparative element. In (10.9), the anaphoric
relation is much more indirect and an account of it would have to include a
large number of inferential steps. Nevertheless it is clear to the hearer that
the antecedent of the anaphoric noun phrase that risk is something like '(the
risk) that scandals like Pindown will happen again'.
'Probability'
Shared semantic features: [MODAL], [EPISTEMIC], [PROBABLE]

Nouns: Chance, likelihood, probability
The noun chance is used to express not only epistemic possibility but also
epistemic probability. As suggested by Palmer's choice of the term degree,
the distinction between the two uses is not clear-cut. Nevertheless, individ-
ual examples tend to lend themselves more readily to paraphrases with one
of the two adjectives possible or probable. For (10.10) below, for example,
I find the paraphrase it is always possible that better than it is always
probable that.
(10.10) Oral sex carries some risk because there is always a chance that the
virus could pass from the man's semen into the other person's body.
(ephem)
(10.11) Perhaps both tasks will be worthless because there is a good chance that
at least one of his rivals will beat him, (PAPERS)
For (10.11), on the other hand, the paraphrases it is likely that or even it is
very likely that sound perfectly convincing to me. These different impres-
sions are largely due to two factors: the difference between the tentativeness
contained in the adverb always in (10.10) and the confidence expressed by
the premodifier good in (10.11), and the difference between the hypothetical
Epistemic uses 241
verb form could used in the that-clause in (10.10) and the confident predic-
tive form will used in (10.11).
Apart from the degree of modality, the most striking difference between
'Possibility' and 'Probability' uses is that the latter are extremely rare in the
anaphoric patterns th-N and th-be-N. This finding can be explained when
one recalls that anaphoric uses of the noun possibility were found to include
the comparative component which is also typical of uses of this noun in the
pattern -be-that. Unlike possibility, likelihood and probability cannot be
used in such context: expressions like *one probability was that... simply
do not make sense because only one possible state of affairs can be more
likely to be the case than all others. Since anaphoric uses also seem to de-
pend on such comparative contexts, they are extremely rare with the two
'Probability' nouns.
Uses of likelihood and probability in the pattern -be-that with the
definite article as determiner are not infrequent, though they tend to occur in
formal rather than colloquial style. A typical example is given in (10.12):
(10.12) ... as having been suggested first by Geoffrey, that can hardly be sup-
posed possible in the case of the English name of Ludgate Hill. The
probability is that as a temple on a hill near Severn associated him with
that river in the west, so a still more ambitious temple on a hill con-
nected him with the Thames in the east, (BOOKS)
A more colloquial way of conveying the same meaning, which is principally

used in spoken language, is the fixed expression chances are that .... Be-
cause of the plural form of the noun, this expression is not contained in the
data proper here.
Finally, likelihood and probability also occur in the pattern -that of
course. In these patterns, the that-chusQS typically include the modal verb
will with a conditional meaning (Palmer 1990: 138-140), as exemplified in
(10.13):
(10.13) Budgerigars are particularly at risk as they frequently climb around the
mesh using their beaks, increasing the likelihood that they will consume
rust particles, (PAPERS)
Example (10.13) involves an implicit relation of contingency. It describes

what is likely to happen under certain circumstances. The modal will is
therefore not used with future reference here, but parallel to its use in the
242 Modal uses
apodosis of conditionals, for example as in If budgies climb around the

mesh using their beaks, they will consume rust particles.
'Certainty'
Shared semantic features: [MODAL], [EPISTEMIC], [NECESSARY]

Nouns: Fact, truth, reality, certainty
With the third degree of epistemic modality, epistemic necessity, we do not

only come very close to, but actually re-enter, the territory of factual uses
already surveyed in Chapter 7. However, since f actuality was not defined in
strict terms, but simply as referring to states of affairs that can 'be the
case', more than factuality is at stake here.
We have seen in Section 7.2 that the collocation the fact + tfiaf-clause is
probably more often used as some sort of general-purpose shelling device
than as a genuine marker of epistemic necessity. I will therefore ignore it in
the discussion of 'Certainty' uses in the pattern N-cl. The other three nouns,
however, do occur in this pattern with the function of characterizing a
PROPOSITION as being epistemically necessary. The rarer noun certainty
occurs relatively more often in this function than the frequent nouns truth
and reality.
(10.14) His sole allegiance is to freedom of artistic expression - to the certainty

that "poets must always stand against the powerful or truth will die".
(PAPERS)
Example (10.14) provides a good opportunity to take up a highly relevant

issue that I have not addressed so far in this chapter: the issue of how
speakers leave their mark on the conceptual content they shell (see also
Section 15.1). Example (10.14) is a particularly good case in point because
the shell content is marked as a quotation but shelled from a modal per-
spective. In analogy to Chapter 8, where it was claimed that it depends on
the shell noun whether an abstract relation is shelled as a fact or an utter-
ance, it can be argued here that it is the noun that determines the status of
the clause rather than the other way round. Thus what the speaker essen-
tially does is portray an utterance as a certainty, i.e. as a fact. This is yet
another reminder of the fact that it is always the speaker of an utterance
Epistemic uses 243
who determines the status of a conceptual unit, even if, as in (10.14), he or

she borrows the linguistic representation of this conceptual unit from some-
one else who gave it a different status. It is also a reminder that the same is
true of all the other examples we have looked at in this chapter. As will be
shown in more detail in Section 15.1, the pattern N-cl is a particularly sub-
tle device for such manipulations of the conceptual status of discourse enti-
ties, since it suggests to the hearer that the characterization contained in the
nouns can be taken for granted.
In uses of 'Certainty' nouns in the pattern -be-that, the status of epis-
temic necessity attributed to the shell content is not suggested so subtly to
the hearer, but to a certain extent deliberately highlighted. Fact, truth and
reality are all very frequent as such markers, or even emphatics, of epis-
temic necessity. See the examples given in (10.15):
(10.15) (a)... on the calculation that with these injections of western capital they
could have converted their economies into much erm er effective er
units that could actually compete in the world [coughs] markets. Now
the plain fact is that they couldn't even with these injections of of
capital they weren't really competitive in world markets, (SPOKEN)
(b) For some years it has been called the British Architectural Library,
giving the impression it is in some way part of the British Library.
The reality is that the British Library gets free copyright copies of
every British architectural book, while the RIBA has to purchase
them, (PAPERS)
(c) At one time, East Germany was thought to have a strong industrial
base and a well trained workforce, but the truth is that many facto-
ries are so antiquated they are not worth saving and they are staffed
by workers whom the West Germans judge to be lazy and lacking
any sort of drive, (BBC)
Even though these three examples have been chosen randomly, they have in
common that their emphatic element is motivated by the speakers' wish to
contrast true representations of reality with only seemingly true ones. The
contrastive relation is most conspicuous in (10.15c), where it is supported
by the adversative conjunction but, but it can easily be inferred from the
other examples as well. Tuggy's (1996: 722) account of such uses in terms
of "disconformity with what is (only) believed/apparent", which I have al-
ready quoted in Section 7.2, is certainly appropriate here. The large set of
corpus data on which I can rely here leaves no doubt that the collocations
244 Modal uses
the fact/reality/truth is + that-ciause are predominantly used in such con-

trastive contexts.
10.3 Deontic uses
According to Lyons (1977: 823-824), deontic modality differs from epis-

temic modality principally in three respects:
a) Deontic modality is not concerned with the truth of propositions but

with "the necessity or possibility of acts performed by morally respon-
sible agents"
b) Deontic modality has "an intrinsic connection" to "futurity".
c) Deontic modality "typically proceeds, or derives, from some source or
cause".
These three aspects have to be taken into consideration for the construction
of a frame for deontic shell-noun uses. The result is represented in Figure
10.4.
Deontic frame: MODALITY INDICATOR - FUTURE ACTIVITY - AGENT - AUTHORITY
Example:
AUTHORITY AGENT MODALITY FUTURE

INDICATOR ACTIVITY
He recalled the telephone the prince permission to marry Diana,
call from asking for (TODAY)
Figure 10.4 Frame for deontic shell-noun uses
The first and second of Lyons' claims are reflected in the components
(FUTURE) ACTIVITY and AGENT, and the third in the component AU-
THORITY. As far as the mapping of the frame components onto language is
concerned, the only one-to-one-mappings are between MODALITY
INDICATOR and shell nouns, and between ACTIVITIES and infinitive clauses.
Depending on the clause structure, the components AUTHORITY and AGENT
can be encoded in a number of ways.
Deontic uses 245
Looking at Palmer's (1990) descriptions of modal verbs in deontic uses,

one might be inclined to add another characteristic of deontic modality:
d) "Deontic modality is essentially performative" (Palmer 1990: 69)
Palmer goes on to explain that "by using a deontic modal, a speaker may
actually give permission (MAY, CAN), lay an obligation (MUST) or make a
promise or threat (SHALL)" (1990: 69). Two points must be made here.
First, even deontic modal verbs are not always performative. As Sweetser
argues (1990: 65), an utterance like "You must be home by ten" can be a
child's report (e.g. to a brother or sister) of an obligation that was originally
imposed by the parents. Thus even deontic modals can either create or re-
port modality. Deontic uses of nouns normally have a reporting function.
Second, with regard to Palmer's categories of making "a promise or
threat", I have explained above (see Section 10.1) that I do not consider
nouns like promise and threat to be modal shell nouns but linguistic shell
nouns, because speakers use them to report speech acts. It is important to
stress both the similarity of and the difference between illocutionary and
deontic shell nouns here, since deontic shell nouns of course have to do with
directive and commissive speech acts like commanding and promising (see
Palmer 1986: 97). Like illocutionary nouns, they report on aspects related
to such speech acts. In contrast to illocutionary shell nouns, however, deon-
tic shell nouns do not report the propositional content of directive or com-
missive speech acts, but the permission, needs and obligations resulting
from them. The reason that these two complementary perspectives on
speech acts are possible at all lies in the Janus-like nature of performatives
as simultaneous acts of saying something and of bringing something about
in the extra-linguistic world: illocutionary shell nouns focus on the speech
act, and deontic nouns on the extra-linguistic act.
'Permission'
Shared semantic features: [MODAL], [DEONTIC], [POSSIBLE]
Frame: Deontic modality

Nouns: Right, permission, freedom, licence, option,
privilege, permit, franchise, leeway
246 Modal uses
On the one hand, permission has to do with the modal degree of possibility,
and on the other, with the modality of obligations. The first relation emerges
from the fact that 'give someone permission to do something' is equivalent
to 'make it possible for someone to do something'; and the second from the
correspondence between 'having the permission to do something' and 'not
having the obligation not to do something' (Lyons 1977: 836-841, Sweetser
1990: 52). Applying the first type of explanation to example (10.16), one
could say that the Attorney-General (i.e. the AUTHORITY) is in a position to
make it possible for the person referred to as He (i.e. the AGENT) to publish
all his correspondence (i.e. to perform a FUTURE ACTIVITY).
(10.16) He was not embarrassed by the leak of this letter, but the Attorney-
General told him permission to publish all his correspondence "might
give him some difficulty." (PAPERS)
The second analysis would be that the Attorney-General is in a position to

remove any obligations not to publish all his correspondence. In terms of
the cognitive analysis of deontic or root modality proposed by Talmy (1988)
and Sweetser (1990), the Attorney-General is in a position to take away a
barrier.
The members of the 'Permission' family are very homogeneous with re-
spect to their lexico-grammatical behaviour. Their favourite environment is
the pattern N-to. Apart from right and freedom, which also occur fre-
quently in non-shell noun uses, they are comparatively rare in the pattern th-
N, and very rare in the pattern th-be-N.
Perhaps the most astonishing collocational finding is that none of these
nouns occurs in the pattern N-e-cl. Why is that so? Why are expressions
like *the permission is to publish the correspondences ungrammatical?
Two explanations can be put forward. First, such identifying uses are im-
possible because co-reference between the implicit agent of the noun, i.e. the
AUTHORITY, a n d t h e implicit a g e n t o f t h e ACTIVITY, i.e. t h e AGENT, is e x -
cluded by the semantic setup. The second possible explanation can be con-
structed on the basis of the second type of analysis mentioned above: poten-
tial uses of 'Permission' nouns in the pattern N-6e-cl would have the func-
tion of explicitly identifying and highlighting excluded alternatives or absent
barriers - clearly somewhat paradoxical notions which are therefore not
grammaticalized.
Especially permission itself, but also some of the other nouns, have a
distinct tendency to collocate with a small number of general verbs. What is
Deontic uses 247
particularly interesting about these verbs is that they fall into two groups.
Have, gain, get, win, ask for and request portray permission from the
AGENT'S point of view; give, grant and refuse, on the other hand, evoke the
perspective of the AUTHORITY. The mapping of frame components onto
clause elements depends on the verbs that are used. In the example given in
Figure 10.4, the AGENT the prince functions as subject of the clause con-
taining the noun permission and the verb ask. In (10.17) below, where the
verb refuse is used, the subject is taken by the noun phrase eight Lords,
Dukes and Earls, which represents the AUTHORITY component in the frame.
(10.17) No expense has been spared to give the production authenticity - al-
though eight Lords, Dukes and Earls refused the production team per-
mission to film in their stately homes, (PAPERS)
The AGENT is added as a direct object here, which, incidentally, of course

also functions as implicit subject of the subordinate infinitive clause.
Permission is the only noun in the family that does not evoke semantic
features in addition to expressing deontic possibility. Right has the implica-
tions that permission has a legal or moral basis and is valid over a long
period of time. Freedom is similar to right but lacks the legal and moral
implications, while putting the focus on the absence of restrictions. Licence
is used when the AUTHORITY giving permission is an official institution or
the like and when the ACTIVITIES are things which such institutions can
allow or forbid. Privilege implies that the AGENT is the only person, or one
of a small group, who has permission to do something. Permit is typically
used to refer to the written documents that contain and guarantee permis-
sion. In this use, it does of course not function as a shell noun. Franchise is
a special type of permission where the AGENT is given permission to sell the
goods or products of the AUTHORITY. And finally, leeway is largely syn-
onymous with freedom from a denotative point of view, but it conjures up
the image of a ship having the permission to stray from the prescribed di-
rection.
'Job'
Shared semantic features: [MODAL], [DEONTIC], [PROBABLE]
Nouns: Job, duty, role, task, business, mission, re-
248 Modal uses
sponsibility, commission, challenge, man-

date, agenda, assignment
It is even more difficult than in the epistemic domain, in fact quite impossi-
ble, to pinpoint the degree of obligation covered by deontic probability. The
'Job' nouns, which cover this area along the continuum of degrees of mo-
dality, shell ACTIVITIES as being more than allowed but less than urgently
required. From the AGENT'S point of view, this means that 'Job' nouns refer
to ACTIVITIES that they are supposed, meant, or even more or less required,
but not absolutely obliged or forced to do. Applying Sweetser's framework
(1990: 54) here, 'Job' nouns are used to portray resistible forces, and
'Need' nouns irresistible forces. Needless to say, the boundary to the 'Need'
family, which subsumes nouns expressing deontic necessity, is not clear-cut
at all.
The more frequent nouns job, duty and task occur mainly in extraposi-
tional constructions or in the expanded predicates have the job/duty/task to.
Role occurs only in the fairly fixed collocations have a role to play or have
an important role to play, and business produces only unwanted matches
when it is followed by infinitive clauses. Mission, however, as well as re-
sponsibility, commission, mandate and assignment occur in uses of the
type illustrated in (10.18). As the example indicates, mission is semantically
fairly specific, specialized as it is for jobs typically involving travelling, that
are carried out by soldiers, diplomats, politicians or other persons with offi-
cial functions.
(10.18) In another aspect of the American military involvement in the Gulf,

several papers carry a photograph depicting the vapour trail of the space
shuttle, Atlantis, as it blasts off on a mission to put a military spy satel-
lite into orbit, (BBC)
In the pattern N-e-cl, the situation is different. Apart from commission and
assignment, all nouns produce examples of this type, though the outputs of
business, mandate and agenda are very small indeed. By fer the two most
frequent nouns in this pattern are job and task. The typical setup of these
uses is illustrated in (10.19):
(10.19) Trackless surgeiy still has its limitations. It can be used only in ablative
procedures, in which the surgeon's job is to destroy a body's rotten fab-
ri. (ECON)
Deontic uses 249
What is so typical about this example is that the noun job is determined by
a possessive element, here a genitive noun phrase. All these possessive de-
terminers, no matter whether they are genitive noun phrases or possessive
pronouns, refer to the AGENTS who are supposed or required to carry out an
ACTIVITY. This element of the frame is therefore highlighted.
The AUTHORITY component, on the other hand, is completely back-
grounded in (10.19) and in many other uses of this type. The result of this
attention distribution in the frame is that the identity of the AUTHORITY
laying the obligation remains indeterminate. The speaker conceptualizes the
situation in such a way that we feel that there must be an obligation either
internal to the AGENT or from an external imposer. In actual fact, however,
it is the speaker alone who is responsible for the fact that an ACTIVITY is
characterized as a task. The task is not objectively given at all. The stylisti-
cally abstract way of doing this by means of a nominal expression (see Sec-
tion 5.1.2) has the side-effect that the AUTHORITY can be left open - an
effect that can only be achieved with a passive construction if the obligation
is described by means of modal verbs.
'Need'
Shared semantic features: [MODAL], [DEONTIC], [NECESSARY]

Nouns Need, pressure, obligation, requirement,
necessity, imperative
Typically, 'Need' nouns are even more abstract in this sense than 'Job'
nouns, because not even the AGENTS are identified by means of possessive
determiners or other linguistic elements. In (10.20), three examples of the
most frequent noun need are collected. These examples sound highly famil-
iar and this indicates their representativity for this type of usage.
(10.20) (a) An adviser to the Lithuanian government stresses the need to "give a
polishing" to Lithuanian foreign-ministry officials, from the top
down. (ECON)
(b) President Bush has told a meeting of world leaders in New York of
the pressing need to alleviate the plight of needy children, (BBC)
(c) They spoke of the need to give the organisation new powers of inter-
vention to deter aggression, (PAPERS)
250 Modal uses
These uses share the typical feature that both the AUTHORITY who is im-
posing an obligation that is portrayed as being irresistible and the AGENT
who is to carry out the ACTIVITY are not specified. The effect of back-
grounding the AUTHORITY is that the obligation is construed as an objective
necessity, dictated by circumstances rather than by a human imposer. A
similar effect is achieved with regard to the AGENT. Sometimes the inter-
pretation that the subject of the main clause is also the implicit AGENT of
the ACTIVITY expressed by the infinitive clause is possible. This applies to
example (10.20c). In many cases, however, and (10.20a) and (b) are good
examples, the use of 'Need' nouns seems to be a deliberate strategy de-
signed to leave open who is to take action in response to some need.
The nouns need, obligation and requirement occur frequently with the
determiner no, typically in existentiaWAere clauses, as in (10.21):
(10.21) There will be no need to change money to visit EC countries, (PAPERS)
In contrast to negated uses of the verb must, which do not negate the mo-
dality but the proposition (see e.g. Palmer 1990: 75-76), the negation before
'Need' shell nouns does negate the modality. When it is said that there is no
need to do something, this simply means that it is not necessary.
The examples of 'Need' uses given so far have already shown that these
nouns occur very naturally in the pattern N-cl. (Imperative is an exception.)
With as many as 8,150 instances, the collocation need + infinitive clause is
among the most frequent collocations in my material. As with other frequent
and close connections, for example fact + iAaf-clause or willingness + in-
finitive clause, one can think of this collocation as being based on a close
semantic dovetail between the meaning of the noun and the meaning ex-
pressed by the grammatical construction. It is important to note here, how-
ever, that this element of obligation is not mentioned by Wierzbicka in her
description of the grammatical meanings of infinitive clauses, because she
does not include modal verbs. This is the first occasion, then, where the
nominal data suggests a revision of Wierzbicka's account by adding the
notion of obligation to her semantic description of infinitive clauses.
Need and pressure are also frequent in the pattern th-N. In the patterns
N-e-cl and th-be-N, on the other hand, none of the nouns are found very
often. Uses of need, obligation and requirement are restricted to compara-
tive contexts involving cohesive determiners and/or premodifiers like one,
another, the main or the highest.
Dynamic uses 251
10.4 Dynamic uses
The group of dynamic uses is on the whole less clearly defined than those of
epistemic and deontic uses. This is partly due to the fact that the notion of
dynamic modality is more controversial than the other two types. Not many
linguists acknowledge the existence of dynamic uses of modal verbs, but
treat such uses as instances of deontic modality. In addition, it certainly also
plays a role that the modal nouns subsumed under this type here are less
homogeneous than epistemic and deontic uses.
As mentioned in Section 10.1, the concept of dynamic modality is nev-
ertheless useful for the description of my data. As I conceive of the term,
dynamic modality is at work in combinations of nouns and infinitive clauses
that are used to indicate what can possibly, probably or necessarily happen
under certain circumstances.73 The modal notion of ability can then be
seen as describing a special type of such circumstances, viz. circumstances
that are portrayed as residing in the AGENT involved in the event. It is only
in this family of modal uses that Palmer's notion of "subject-orientation"
plays a role. (It should also be noted in passing that Bybee's notion of
"agent-oriented" modality (1985: 166-169, see also Bybee and Fleischman
1995: Section I), which appears to be helpful here at first glance, is too
comprehensive for the present purpose.)
A frame that is general enough to accommodate the various types of uses
subsumed as 'dynamic modal' is proposed in Figure 10.5.
Dynamic modality frame: MODALITY INDICATOR - EVENT
Example:
MODALITY EVENT (here: ACTIVITY)
INDICATOR
It's really an opportunity to concentrate on something different from

work and home, (MAGS)
Figure 10.5 Frame for dynamic modal uses
As can be seen, the frame consists of no more than two components, the
MODALITY INDICATOR, which is as before mapped onto the shell noun, and
an EVENT, which is typically expressed by means of an infinitive clause
directly attached to the noun.
252 Modal uses
'Ability'
Shared semantic features: [MODAL], [DYNAMIC], [POSSIBLE], [SUB-
JECT-ORIENTED]
Frame: Dynamic modality

Nouns: Ability, power, failure, capacity, inability,
authority, potential, talent, skill, capability,
clout, incapacity
The modal notion of ability is thus understood as a special type of dynamic

possibility, in which the possibility of an EVENT is portrayed as being de-
pendent on qualities attributed to an agent. Since agents are involved in
most of these uses, the EVENT component of the frame is typically repre-
sented by an activity rather than a non-agentive process. In (10.22), for
example, the agent Smith is portrayed by the speaker as having qualities
which make it possible for him to rationalise difficult circumstances ....
(10.22) Smith's own ability to rationalise difficult circumstances and bring

sense to problems was formed in the East End of Glasgow ... (PAPERS)
Given this frame setup, it is not surprising that the most frequent nouns
ability, power, failure, capacity and inability are typically found with pos-
sessive elements referring to the agents who are credited by the speakers
with certain abilities. Another pattern which is typical of these nouns (ex-
cept failure) is the combination with the verb have, in which the subject of
the clause represents the agent. This is illustrated in (10.23):
(10.23) But at least Paul Merson and Neil Webb - both craftsmen in a different
style to Gascoigne - showed they have the ability to fill the Gazza void
during England's 1-0 win over the Spanish Olympic XI on Wednesday.
(TODAY)
Such uses are not particularly exciting examples of shell nouns, because
they can be directly translated into clausal versions, here for example into
showed they are able to fill.... However, one should not be rash in propos-
ing a complete correspondence between the nominal and verbal or adjectival
version here because there is a slight semantic difference. According to
Palmer (1990: 90), the modal verb be able to is sometimes used instead of
can when the actuality of an ability, as opposed to its potentiality, is to be
Dynamic uses 253
emphasized. Even though a specific temporal reference is given in example

(10.23) by means of the clause-final temporal adverbials, I would not inter-
pret this example as describing the actual ability of the two players in this
particular situation, but their general potential to fill the Gazza void. The
nouns in the 'Ability' family highlight the aspect of potentiality much more
than do the modal verbs be able to and can. Quite predictably, this tendency
is most marked with the noun potential, as is illustrated in (10.24):
(10.24) Cambridge University is said to have applied for a patent to investigate a

Fujisawa Pharmaceutical drug for its potential to prevent the spread of
fflV. (TODAY)
This example also shows that not all 'Ability' nouns are typically accompa-
nied by references to human agents; here the ability to do something is at-
tributed to an inanimate target, a drug.74 The noun potential is clearly ear-
marked for such uses.
The idea of a general and long-term ability to carry out certain activities
is also inherent in the stable meanings of the nouns skill and talent. These
two nouns are typically used in collocation with the verb have where the
subject refers to the agent of the activity that functions as EVENT. It should
be noted that these two verbs, as well as power and to an even greater ex-
tent authority, are often found in uses that have nothing to do with the no-
tion of shell nouns.
Quite the opposite is true for the antonym of ability, the noun inability,
which musters the highest score for compiled reliance (78.13%) in the
whole material. This extremely high score is almost exclusively due to oc-
currences in the pattern N-/o, where the noun boasts a reliance score of
76.79%. Uses of this kind are typically determined by possessive elements
or by the indefinite article. Especially when the agent is inanimate, the ref-
erence to it is usually not expressed in the determiner, but as a postmodify-
ing o/-prepositional phrase, as in (10.25):
(10.25) Most problems with our eyesight are caused by an inability of the lens
to focus properly. (EPHEM)
As outlined in Section 4.4, such uses are not identified and counted by the
automatic corpus query and therefore the 'true' score for reliance is even
higher than the one given by the system. In the particular case of inability,
the concordance suggests a reliance score of virtually 100% for the pattern
N-to, when such structurally correct uses are also counted. This means that
254 Modal uses
one can more or less predict that whenever the noun inability occurs, it will
sooner or later be followed by an infinitive clause expressing what a person
(or object) is not able to do. As with the nouns in the 'Desire' and the 'Re-
luctance' families in Section 9.3.3, it certainly makes sense here to regard
the infinitive clauses as complements.
While incapacity differs from inability mainly in that it is stylistically
more formal, the third noun with the feature [POSSIBLE], failure, stands in
a semantic contrast to it. Like inability, it is most commonly found with
possessive determiners referring to the agent. However, while inability is
mainly used with reference to long-term dispositions, or rather deficits, fail-
ure is also frequently used for short-term inability to carry out single activi-
ties. The difference is illustrated in example (10.26), where the shell content
represents an individual, immanent aim which the agent is portrayed as
being unable to achieve.
(10.26) His remarks are being interpreted by many here as preparation for his
expected failure to unseat Mr Patterson. Jamaica's first black leader.
(PAPERS)
The utterance in this example is not about a general and potential inability
to perform activitites of a certain type, but about the inability to carry out
one particular act successfully.
'Opportunity'
Shared semantic features: [MODAL], [DYNAMIC], [POSSIBLE], [NEUTRAL]
Nouns: Time, way, chance, place, opportunity, area,
stage, moment, position, situation, possibility,
occasion, room, approach, method, step,
region, space, scope, facility
The diverse members of of the 'Opportunity' family are united by the fact
that in this usage, they convey a meaning similar to 'situation or circum-
stances where it is possible to do something'. Thus, example (10.27) can be
paraphrased by something like 'they find themselves in a situation where it
is possible (for them) to come and talk to me'.
Dynamic uses 255
(10.27) ... you know after six weeks of me jabbering on for a couple of hours
they know me very well. - Mhm. - and so they've got that opportunity
to come and talk to me. (SPOKEN)
It should be noted that like (10.23), this example could also be paraphrased
using the modal verb can: 'and so they can come and talk to me'. What this
shows is on the one hand that can is indeed ambiguous between a subject-
oriented and a neutral dynamic reading, and on the other hand, that again
the modal nouns help to show which reading is more likely in a given use.
A closer look at the list of nouns indicates that the 'Opportunity' family
branches into the following five sub-families:
Neutral chance, opportunity, possibility, occasion

General circumstantial position, situation
Temporal time, stage, moment, step
Locative place, area, room, region, space, scope
Manner way, approach, method
The core of the 'Opportunity' family is made up of the nouns listed next to
'neutral'. As in example (10.26), these nouns have the inherent potential to
shell EVENTS as possibilities. We have seen in Section 10.2 that chance and
possibility also occur in epistemic uses. The former is far more frequently
found with infinitive clauses in dynamic function, and the latter in colloca-
tion with //jaf-clauses with an epistemic meaning. Occasion also occurs
with -infinitives, but is more frequent in the pattern th-N.
The other nouns in the family do not have this inherent element of possi-
bility. Looked at in isolation, these nouns are simply circumstantial, tempo-
ral, locative or manner concepts, and it is in this function that we will see
them again in Section 12.3. Yet some of these nouns, especially time, way
and place, are extremely frequently followed by infinitive clauses - time no
less than 19,496 times in the corpus of 225 million words on which this
study is based. Attention must be drawn to the fact, however, that the in-
finitives in many of these uses can be replaced by relative clauses. This
means that it is often doubtful if there is a relation of experiential identity
between head noun and clauses. These examples of circumstantial nouns are
therefore on the very fringe of the functional category of shell nouns.
What is nevertheless striking about these uses is that the interaction be-
tween the nouns and the infinitive clauses somehow results in the same mo-
256 Modal uses
dal meaning that is typical of and inherent in chance, opportunity and the
other neutral nouns.
(10.28) (a) Paradoxically, my being both horizontal and ill and therefore in no
position to displease her brought out the best in my mother, ...
[BOOKS]
(b) The best time to encourage your older child to start caring for a new
baby is before the birth, (MAGS)
(c) The place to make the right contacts - an activity long known as
"networking" - will be the Net. (NEWSCI)
(d) Senior Republicans admit that the economy is undermining their
base of voters. Some believe the only way to win is to press home
personal attacks on Mr Clinton... (PAPERS)
As claimed above, all these example are concerned with the notion of possi-
bility. They differ with regard to the facets of the circumstances that are
highlighted for attention by the speakers. In (10.28a), it is the general situa-
tion that gives, or in this case does not give, rise to a possibility. Inciden-
tally, such uses ofposition with be and no as determiner would even qualify
for inclusion in the 'Ability' family because they have to do with subject-
oriented possibility, i.e. ability. Such uses are ambiguous, just like the mo-
dal verb can. In (10.28b), it is the temporal aspect that is foregrounded, in
(c) the locative aspect, and in (d) the manner aspect. These three examples
can be paraphrased as 'the time when it is best possible to ...', 'the place
where it is possible to ...' and 'the only way in which it is possible to'.
The fact that neither time, place or way, nor their semantically more
specific relatives like stage, area or method have modal aspects inherent in
their meanings raises the question why these uses are conceptualized as
expressing dynamic possibility in the first place. After all, the MODALITY
INDICATOR which is normally included in the meaning of the shell noun is
missing. One possible explanation is that the MODALITY INDICATOR is ex-
pressed, or hidden, in the infinitive clause. This idea is of course not entirely
new, though it is not uncontroversial.75 It can for example be found in Quirk
et al. (1985), but it is interesting to note how carefully these authors chose
their words:
Postmodifying /o-infinitive clauses can either have a modal or a nonmodal

sense, but the modal interpretation seems to be normal.
(Quirk et al. 1985: 1269).
Dynamic uses 257
My data clearly supports the idea that postnominal infinitive clauses have a
modal meaning more often than not. Some of the paraphrases that Quirk et
al. propose for their examples, however, for example the thing we should
do for the thing to do, suggest that not just dynamic, but also deontic read-
ings of these uses, which can be glossed by should, are possible.
A second explanation for the emergence of the modal meanings of these
uses is that they arise from the interaction between circumstantial nouns and
f-infinitives. This claim is particularly convincing from the point of view
of Construction Grammar (see Fillmore, Kay and O'Connor (1988), Gold-
berg (1995)), since the proponents of this theory argue that grammatical
constructions have meanings like lexical items and that lexico-grammatical
patterns can acquire meaning in addition to the semantic impact of their
components.
To give an idea of the typical collocations of the nouns time, place and
way with -infinitives that give rise to modal meanings, I have compiled a
selective survey in Figure 10.6. Whereas the uses of way show a clear ten-
dency towards expressing dynamic possibility, those of time and place are
often ambiguous between dynamic and deontic readings.
Collocation Example
(not) have time to so I didn't have time to feel afraid, (MAGS)
there is/was no time to There is no time to stand on the cliffs at Dover (PAPERS)
a good/excellent etc. time to this is an excellent time to knock London firms off their
... (PAPERS)
it V it was time to it's time to think the unthinkable (SPOKEN)

that's/this is the time to This is the time to zet together with your partner (MAGS)
a place to Three meals a day and a place to sleep (BOOKS)
a good/wonderful/ideal etc. it's a wonderful place to work (MAGS)
place to
the place to here is the place to remember (MAGS)
the only/best/ideal etc. way authority was the only way to curb infantile aggression
to (PAPERS)
the (only/best) way (to-

clause) is/was to the only way to check is to see if they ... (SPOKEN)
find a way to
Hurd will try to find a way to break the deadlock (BBC)
Figure 10.6 Selected collocations of time, place and way with infinitive clauses
258 Modal uses
It should be remembered that there are of course other collocations where

these three nouns followed by infinitive clauses do not represent shell noun
uses, e.g. it takes time to ... or in time to .... Furthermore, attention should
be drawn to the fact that the second collocation listed for way in Figure 10.6
does not just manifest the pattern N-to but also the pattern -be-to, as the
postnominal infinitive clause is followed by the copula and another infinitive
clause which identifies the way in which it is possible to do something.
'Tendency'
Shared semantic features: [MODAL], [DYNAMIC], [PROBABLE]

Nouns: Tendency, trend, propensity, disposition
The nouns in the 'Tendency' family do not have straightforward counter-

parts among the 'classic' modal verbs. In studies that do not only address
the issue of modal verbs, but modality in general, this type of modality is
sometimes referred to as "disposition" (see e.g. von Wright 1951: 28). I
have partly included this type of modality in my discussion of 'Ability',
when I talked about long-term abilities of persons. As their place in the
framework of this description of modal shell nouns suggests, I look upon
'Tendency' nouns as expressing dynamic modality with the degree of prob-
ability. On a very general level, their semantic impact can be glossed as
'likely to happen' or 'likely to do'. Example (10.29) is a case in point:
(10.29) Multipurpose dye has a tendency to wash out or become 'non-

colourfast' in wool, (MAGS)
The mapping from frame onto clause in (10.29) is typical of the whole fam-
ily. The MODALITY INDICATOR is expressed in the shell noun, the EVENT in
of the infinitive clause, and the agent or thing involved in the EVENT is ex-
pressed as well, here as subject of the verb have. It is also typical that the
source of the probability of an EVENT, i.e. the circumstances under which it
is likely to happen, is imputed to the protagonist of the EVENT, no matter
whether it is inanimate, as in (10.29), or animate as in (10.30) below:
(10.30) In discussion with his counsellor, David came to recognise his tendency
to run away from problems and disappointments, (EPHEM)
Dynamic uses 259
More often than not, the animate or inanimate agent involved in the EVENT
is expressed explicitly, either as subject of a have-clause (10.29) or by
means of a possessive determiner (10.30). However, the agent can also be
backgrounded, for example in existential iAere-constructions (see 10.31), or
when the or this/that are used as determiners.
(10.31) So what's the link with low sugar? Recently there's been a tendency to
reduce preservatives and sugar, which also acts as a preservative, in
foods. (MAGS)
In the field of dynamic possibility, I have distinguished between subject-

oriented dynamic possibility, i.e. 'Ability', and neutral dynamic possibility,
'Opportunity'. As the examples of tendency indicate, the two variants are
united under one roof in the field of dynamic probability.
Tendency is by far the most frequent noun of 'Tendency' in my data.
Like trend, it is found in the pattern N-to, in N-e-to and even -be-that,
and in th- and th-be-N. Trend is also fairly frequent, and in fact more
frequent than tendency as a noun altogether, but it is used more often in the
anaphoric patterns than in N-cl or N-e-cl. Propensity and disposition are
much rarer nouns. Both include subject-orientation in their stable meaning.
As can be gleaned from the difference in the reliance score for the pattern
N-to, 42.1% for propensity vs. 3.9% for disposition, propensity is much
more often used as a shell noun than disposition is.
'Destiny'
Shared semantic features: [MODAL], [DYNAMIC], [NECESSARY]
Nouns: Destiny, fate
With the small 'Destiny' family, we finally leave the domain of linguistic
elements that have traditionally been looked upon as having the potential for
expressing modality. While nouns like need, necessity and obligation are
the nominal semantic equivalents to verbs that are used to express dynamic
modality, e.g. must, have (got) to, need or will, I have discussed these
nouns under deontic modality because I believe that their meanings imply an
AUTHORITY and an AGENT, even if, as we have seen in Section 10.3, these
are often not expressed on the linguistic surface. Looking for nouns which
260 Modal uses
can be used to say that it was, is or will be necessary that an EVENT takes
place, we find that the two nouns destiny and fate can be used to do pre-
cisely that. Rather than giving one or two representative examples, I will
here support my claim by giving all six concordance lines retrieved for the
collocations fate is to and destiny is to from the PAPERS and the MAGS sec-
tions of the corpus:
(10.32)
< children, your destiny is to live together under the earth," we>
< John Smith's destiny is to lead the movement which will stop
< I believe my destiny is to sing these songs and I'm sure I'll
< # her true destiny is to be a boy. The submerged them of
< heart. The Novel's fate is to disintegrate in the course of a game
< that their unhappy fate is to be pawns in other people's games.
What all these speakers are ultimately saying is that it is necessarily the
case that certain EVENTS (expressed in the infinitive clauses) will take place
- and this is, as we have seen, what dynamic necessity is all about, at least
according to the definition of dynamic modality given at the beginning of
this section. It is typical of these uses of fate and destiny that they are de-
termined by possessive elements expressing the agents or patients that are
portrayed as being necessarily - and usually unwillingly - involved in these
EVENTS. As with all modal shell nouns, except the uses of time, place, way
etc. discussed in the Opportunity' family, the MODALITY INDICATOR is
included in the meanings of the shell nouns destiny and fate.
It must be admitted, though, that dynamic necessity, as I conceive of it
here, is a fairly marginal modal shell-noun phenomenon. The two nouns do
not occur in the pattern N-cl at all, a pattern which is typical of dynamic
shell noun uses otherwise. In fact, they cannot even muster large numbers of
uses in the pattern N-e-cl, as is indicated by the fact that even the large
subsections from the newspapers and the magazines include only six in-
stances. Both nouns do occur in the patterns th- and th-be-N, but also not
too often.
All the same, 'Destiny' nouns do express dynamic necessity and there-
fore deserve to be mentioned here. What is more, they give me yet another
opportunity to stress the fact that modalities are of course not objectively
given, but construed by the speaker. In the examples collected in (10.32), it
is always the speaker who predicts that something will happen or states that
it always happens in a certain way. No matter how well the speakers suc-
ceed in portraying events as being objectively predestined, the necessity
always has a subjective basis.
Chapter 11
Eventive uses
11.1 Introduction
In Section 5.1.1, it was laid down that the term event is used in this study as
a superordinate term for mental entities representing physically observable
events which have a temporal duration. In the Transformational Generative
framework, from which I have borrowed the term eventive, events are dis-
tinguished from facts by means of test frames like Vendler's occurred
at ten o 'clock (see e.g. Menzel 1975: 122). Although I will also apply this
test frame, at least in principle, because it helps to distinguish events from
facts, it must be borne in mind that in generative grammar, the term event is
not only defined in contradistinction to fact, but also to action, act and pro-
cess, roughly as a term for activities in which agents are backgrounded. The
reason for this conception lies in the focus on the possible syntactic realisa-
tions of these types of entities, mainly fActf-clauses, infinitives, gerunds and
nominalizations (in the narrow sense).

General eventive uses 'Event' event, change, action 11.2
Specific eventive uses 'Attempt' attempt, effort, move 11.3

'Tradition' tradition, habit, custom
'Option' option, priority, alternative
Attitudinal eventive 'Trouble' trouble, problem, difficulty 11.4

uses 'Success' success, achievement, triumph
'Mistake' mistake, crime, fault
Figure 11.1 Survey of the class of eventive shell-noun uses
The division of the class of eventive uses is represented in Figure 11.1.

Three groups can be distinguished. The first and the second group differ
chiefly with regard to the criterion of intensional specificity. The character-
262 Eventive uses
istic feature of the third group is, in analogy to attitudinal factual uses (see
Section 7.7), the predominance of attitudinal components.
Like the other classes of shell-noun uses, eventive uses can also be de-
scribed with the help of a frame. In analogy to the general frame proposed
for the description of a number of families of factual uses, this frame is very
simple. It consists of an EVENT (expressed as shell content) which can take
the more specific variants of ACTIVITY, PROCESS or STATE, and a
CHARACTERIZATION of the EVENT, which is contained in the shell noun
itself. This frame is represented in Figure 11.2.
Eventive frame: EVENT - CHARACTERIZATION
Example:
CHARACTERIZATION EVENT (ACTIVITY)
Sarah's next move was to block what she thought were attempts to draft
her daughter into munitions work, (BOOKS)
Figure 11.2 Frame for eventive shell-noun uses
11.2 General eventive uses
'Event'
Shared semantic features: [EVENTIVE], [NEUTRAL]
Frame: General eventive

Nouns: Position, situation, change, event, act, action
The list of nouns that occur in general eventive uses is short, suprisingly
short, in fact. The reader would perhaps have expected such nouns as proc-
ess, activity or state to be included here. The fact is, however, that these
nouns do not qualify for the status of shell nouns according to the syntactic
criteria laid down in Section 4.2: process, activity and state do not occur in
the patterns N-cl or N-e-cl.
If we ignore position and situation for the time being, even the other
nouns given as suppliers of general eventive uses are not prime examples of
shell nouns. Two of them do not give rise to eventive uses when they occur
in the patterns N-cl or N-e-cl, and the other two are rare and collocation-
General eventive uses 263
ally restricted: event occurs only in the pattern -that, but the collocation in
the event that is conditional and not eventive, and change occurs only in the
pattern N-be-that, but these uses are attitudinal factive ones of the 'Irony'
type (see Section 7.7); and while speakers do use act and action in the pat-
tern N-be-to, this is only possible when they insert restrictive premodifiers
like only or first, as in the examples in (11.1):
(11.1) (a) Gillies's first act was to fire Ed Haves as lawyer for the foundation.
(MAGS)
(b) The first action was to place the vessel under cover and remove the
deckhouse, (MAGS)
At first glance, it may be a surprise that these intensionally unspecific

nouns, which would intuitively be regarded as prime candidates for eventive
shell nouns, do not feature more prominently here. However, looking at
them more thoroughly it turns out that these nouns do not have the struc-
ture-inherent semantic gap that is required from a noun if it is to function as
a shell noun (see Section 5.2). For these nouns do not gap states, processes
and activities, but they refer to them, and this of course rules out their use
in the patterns N-cl and N-e-cl. The only variant of these constructions
that is possible is the identifying one, provided that, as in the examples in
(11.1), a restrictive element like first, only or a superlative form is added as
nucleus of the topic.
The patterns in which the nouns listed above and similar nouns like pro-
cess, development, activity and state really begin to thrive are the anaphoric
patterns th- and th-be-N. This is only natural, given that they have the
potential for referring to different types of events. Four examples, one each
with this and that in the patterns th-N and th-be-N, are given in (11.2):
(11.2) (a) ... effortless grace with which a blob of seemingly featureless frog-
spawn transforms itself into a tadpole - complete with skin, muscle,
gut and all the intricate wiring of the nervous system. Out of the
simple came forth complexity. If this process leaves you cold then
here is your cure: I prescribe a course of Wolpert. (NEWSCI)
(b) If a class is conscious of what advantages it has, it will necessarily
generate notions that tend to create and preserve that situation.
(BOOKS)
(c) ".... For political reasons," said Black Dominique, "Clarence has
decided that our man should be flown to Morocco in a diplomatic
bag. We'll give him to King Hussein". This was a new develop-
264 Eventive uses
ment: hitherto we'd planned to pop up unexpectedly in Spain, where

we'd confront the imposter, Juan Carlos, (PAPERS)
(d) "But first the worm must be killed. I'm sure the chef doesn't know
how. It must be killed!" Danio repeated. "Then perhaps you should
kill it". - "I cannot," Danio said. "You know I cannot." - "No, that
would be an ignoble act," Hanuman said, (BOOKS; COBUILD codes
indicating speaker turns omitted)
The examples give a good idea of how these nouns are used as means of
referring back to descriptions of events. In the pattern th-N, especially when
no premodifiers are added as in (11.2a and b), such references come very
close to uses of this and that in extended reference (Halliday and Hasan
1976: 52) because the CHARACTERIZATION components that the nouns can
muster are highly general in nature. The only semantic contributions of the
nouns process and situation are [EVENTIVE], [DYNAMIC], [NON-AGENT-
IVE], and [EVENTIVE], [NON-DYNAMIC], [NON-AGENTIVE] respectively. In
Halliday and Hasan's framework, these nouns manifest the transition zone
between reference and lexical cohesion. In the pattern th-be-N, the nouns
also function as semantically unspecific shells, but here they are typically
accompanied by descriptive or evaluative premodifiers which carry the in-
formational and prosodie focus of the clause.
Although I have just drawn on the features used for the definition of sec-
ond-order entities in Section 5.1.1, one must be careful not to mix up the
metalinguistic, technical terms 'event', 'activity', 'process' and 'state' with
the corresponding non-technical English nouns, whose description is at
stake here. In everyday English the noun event, for example, is not used as a
superordinate term for activities, processes and states, but to refer to con-
glomerates of activities and processes, in such a way that the individual
agents that may be involved are backgrounded and the activities and proc-
esses are viewed holistically. Its meaning is thus closer to the conception of
events in generative grammar (Menzel 1975: 121-139).
Vendler (in a lecture referred to by Menzel 1975: 50) and Menzel (ibid.)
make interesting claims as to the semantic difference between the nouns act,
action, activity and process. According to Vendler, speakers express disap-
proval when they use act. Menzel argues that when the noun act is used
with postmodifying gerunds linked to the head noun with of, as for example
in the act of stealing the bike, "an act is an action which the speaker con-
siders to be important" (1975: 52, 179-195). In the anaphoric uses dis-
cussed here, these differences between act and action cannot be detected. As
General eventive uses 265
fer as activity is concerned, Menzel suggests "that activities are generic

actions" (1975: 182). In a similarly general way, Menzel (1975: 52-57) sees
the element of duration as the main feature of processes and the noun proc-
ess., and adds that processes evoke the perspective of the patient.
Though most of Menzel's claims seem at least compatible with what the
corpus data indicate, one must not forget that they are mainly made with the
aim of constructing rules that generate only grammatical sentences where
matrix verbs and complements match. My own findings are more in line
with the short usage descriptions given by Chalker (1996: 94).76 Speakers
seem to use act to describe single and complete agentive events, and action
to emphasize that the agent has a purpose in performing an activity. This
means that act evokes a more holistic perspective than action, which is used
to invite the hearer to imagine the course of an action and the processes that
are involved. Probably influenced by such typical plural collocations as
leisure activities and terrorist activities, the noun activity portrays actions
as particularly lively and energetic, or 'busy' as Chalker (1996: 95) calls
them. Process is not only used to refer to processes (in the technical sense)
as in (11.2a), but also, similar to event, to conglomerates and series of ac-
tions.
I have excluded the two nouns position and situation from my consid-
erations so far because they are special cases in two respects. First, when
they occur with infinitives, these nouns normally give rise to dynamic modal
uses. When they occur with that-clauses, they produce either neutral factual
or, in the case of position, assertive linguistic uses. Second, when these two
nouns actually produce general eventive uses, these represent states rather
than activities or processes. But in order to achieve this, speakers must use
postnominal w/z-clauses rather than infinitives. Two examples are given in
(11.3) and (11.4):
(11.3) I'm in the fortunate position where it's become the norm for me to do
something different each time, (MAGS)
(11.4) We now have a situation where thrombolytic therapy is very widely
accepted and very widely practised, (BBC)
Although in principle nothing can be said against the claim that such exam-
ples shell STATES in a very general way, I hesitate to make this claim with-
out reservations because, as we will see in the next chapter, w/?-clauses are
typically used as complementizers of circumstantial shell nouns. I will
therefore leave the question as to whether these are proper examples of gen-
266 Eventive uses
eral eventive uses open until the discussion of general circumstantial uses in
Section 12.2.
11.3 Specific eventive uses
As the names of the groups indicate, specific eventive uses differ from gen-
eral ones with respect to their semantic specificity. Unlike general eventive
uses, they include fairly distinct characterization components. There are
three families of specific eventive uses.
'Attempt'
Shared semantic features: [EVENTIVE], [PURPOSIVE]

Nouns: Attempt, effort, campaign, move, struggle,
test, conspiracy, plot, initiative, trick, race,
fight, measure, quest, rush, ploy, gamble,
crusade, enterprise, endeavour, venture, ruse,
manoeuvre, precaution, countermeasure
Speakers use the 'Attempt' nouns to attribute the CHARACTERIZATION of

goal-orientation, i.e. a feature [PURPOSIVE], to EVENTS, or more typically,
to ACTIVITIES. Even this short and abstract description indicates that we are
re-entering the familiar terrain of volitional uses like 'Desire', 'Aim' and
'Plan'. Like these families, 'Attempt' nouns have to do with goals and in-
tended accomplishments, but 'Attempt' uses shell the ACTIVITIES under-
taken in the pursuit of aims, while 'Desire', 'Aim' and 'Plan' uses shell
PSYCHOLOGICAL STATES of volition. It is a natural consequence of this
relation that the shell contents of 'Attempt', 'Desire', 'Aim' and 'Plan'
nouns all typically represent the envisaged accomplishment (which are ex-
pressed by means of infinitive clauses). Furthermore, the experiencers of
volitional states tend to coincide with the agents of such activities. Both sets
of nouns also share the distinct implication that the outcome of the
ACTTVITY, that is the question whether the accomplishment will be achieved
or not, is uncertain. All these similarities notwithstanding, 'Attempt' nouns
differ from mental nouns in that they themselves denote events.
With no less than 13,776 attested occurrences, the collocation attempt +
io-infinitive holds the fourth rank in the frequency list of combinations of
Specific eventive uses 267
nouns with clauses, only outstripped by fact that (26,106 occurrences), time
to (19,496) and way to (15,194). An example is given in (11.5):
(11.5) The British foreign secretary, Mr Hurd, has failed in an attempt to get
the Israeli government to change its mind about refusing to accept a UN
mission to investigate last week's killings in Jerusalem, (BBC)
In terms of reliance scores, however, attempt surpasses the other three

nouns, with a compiled reliance score of 66.46% and a reliance score of
65.33% for the pattern N-to. The latter score does not come as a surprise of
course, given that the 'Desire' nouns, for example, and their negative coun-
terparts, the 'Reluctance' family, show a similarly strong tendency to collo-
cate with infinitive clauses. If one also takes into account that attempting to
do something roughly means 'starting to do something in order to see if one
is able to achieve it', one realizes that there is another close conceptual tie
with the 'Ability' uses, which, needless to say, are also typically comple-
mented by infinitive clauses. This means that we are accumulating colloca-
tional evidence for the claim that infinitive clauses have to do with a con-
ceptual complex revolving around notions like wishes, aims, accomplish-
ments, obligations, abilities and attempts.
The collocation attempt + infinitive is used in a large variety of con-
structions. The major recurrent environments are: X is/was an attempt to, X
made no attempt to and in an attempt to. The same types of collocations
supply the favourite habitats of the noun effort, which is semantically also
very similar. Attempt suggests more than effort does that the ACTIVITY may
be unsuccessful. Effort, on the other hand, indicates a stronger intensity of
the shelled ACTIVITY. Even stronger characterizations are included in the
collocations struggle + infinitive and fight + infinitive. Race and rush imply
that several agents compete in the attempt to achieve something. Campaign
is mainly used in political and business contexts and carries the implication
that the ACTIVITIES extend over a longer period of time. Like campaign,
move is a favourite of the press. Trick and ruse, as well as conspiracy, plot
and ploy are used to characterize insidious and deceptive ACTTVRNES carried
out underhand with the aim of harming someone else, while gaining an ad-
vantage for oneself; the latter trio has the additional feature that several
agents cooperate in such attempts.
The most frequent nouns attempt, effort and campaign do not occur in
-be-to uses, but many of the other nouns do, namely move, struggle, test,
trick, measure, quest, ploy, ruse, venture, manouevre and countermeasure.
268 Eventive uses
With the exception of trick, which produces uses of the 'Way' family in this
pattern (see Section 12.3), the output of these nouns in the pattern -be-to
is fairly low. And like act and action, most of them require restrictive or
comparative elements like another, best, next, latest or simpler.
The noun attempt itself is surprisingly rare in anaphoric uses, and so is
effort. Campaign, move, test and measure on the other hand are widely
found in the patterns th-H and th-be-N. Illustrations are given in (11.6):
(11.6) (a) It was in the aftermath of this crash that he converted to Judaism.
This move inspired his own joke in reply to a fellow-golfer who
asked him what his handicap (golfing standard of playing) was. He
remarked: "I'm a one-eyed Jewish negro. What more do you want."
(BBC)
(b) But his sponsors, Wheeler, who hadn't seen much action from their
main attraction, insisted he ride the final of the Grundig World Cup
on the high altitude course at Vail. Colorado. According to Kluge,
this was a disastrous move, (MAGS)
'Tradition'
Shared semantic features: [EVENTIVE], [HABITUAL], [TEMPORAL], [MAN-
NER]

Nouns: Tradition, habit, convention, custom, ritual
This small set of rather specific nouns is used to indicate that an EVENT or
ACTTVITY has been carried out in the same way for a long time. They in-
volve the features of continuity and habituality. While all five nouns can
muster small numbers of occurrences in the patterns N-cl or N-e-cl, these
certainly do not make up their usual environments. When speakers want to
link these nouns not just anaphorically but structurally to their shell con-
tents, they prefer to express the shell contents in postmodifying of-
prepositional phrases, as illustrated in (11.7):
(11.7) The school has a well-established tradition of teaching and research in

these areas, (EPHEM)
Specific eventive uses 269
Uses of the nouns in anaphoric patterns are also found, although it should
be added that none of them is used particularly frequently as a shell noun.
An illustration of a th- use of the noun habit is given in (11.8):
(11.8) Every new report on the effects of smoking adds to the weight of scien-
tific knowledge and common sense against the perils of this dangerous
habit, (PAPERS)
A semantic differentiation of this group of nouns is given in a usage note

attached to the entry on habit in LDOCE3; I will therefore say no more
about them here.
'Option'
Shared semantic features: [EVENTIVE], [COMPARATIVE]

Nouns: Option, priority, alternative, choice, prefer-
ence, speciality
Another not too large family of nouns, the Option' family, can be seen as
the eventive counterparts to comparative factual nouns (see Section 7.5). In
fact, two of the nouns, option and alternative, can occur with both that-
clauses and infinitives in the pattern N-e-cl. Combined with that-c\a.uses,
option produces epistemic 'Possibility' uses, and alternative comparative
factual uses. In the pattern -be-to, which is the favourite environment of
'Option' uses (see 11.9), both nouns give rise to eventive uses.
(11.9) The herbalist's priority is to encourage the body to proceed with the
change as smoothly as possible, rather than relying on temporary, and
often delaying, alleviation of hormone replacement therapy, (BOOKS)
In such -be-to uses, 'Option' nouns shell an ACTIVITY as one of several

alternatives. While option, alternative and choice are neutral as to which
course of ACTIVITY is preferred, priority and preference include the
CHARACTERIZATION that the ACTIVITY identified in the infinitive clause is
favoured. Speciality includes the additional meaning that the agent performs
the shelled ACTIVITY particularly proficiently, frequently and willingly.
Priority and alternative can be found with a variety of premodifiers and
270 Eventive uses
determiners. Choice is typically used with focusing elements, and prefer-

ence and speciality with possessive determiners.
11.4 Attitudinal eventive uses
As signalled by their name, attitudinal eventive uses are the eventive coun-
terparts to attitudinal factual uses. To start with, uses of problem, trouble,
difficulty, dilemma and snag in the pattern -be-to must be mentioned here
(see Section 7.7). These nouns make up the 'Trouble' family. The other two
families of attitudinal eventive uses, 'Success' and 'Mistake', are distin-
guished on the basis of the simple polarity between expressions of positive
and of negative attitudes.
Two further sets of nouns clearly belong here, although they will not be
explicitly listed, since they are only eventive in uses in the anaphoric, but
not in the clausal patterns. In the latter they produce factual uses. The
nouns in question are the attitudinally positive event descriptions break-
through, miracle, revolution and sensation, and their negative parallels
blow, catastrophe, debacle, disaster, drama, farce, nightmare, scandal
and tragedy. The noun tragedy, for example, is factual in an utterance like
Roy Cornes's tragedy is that he was born with haemophilia (see example
7.32a), but when it is used in anaphoric construction, as in example (7.32b)
repeated here as (11.10), it refers back to the description of an event:
(11.10) L J The doors are alarmed but it all happened so fast I haven't had the
opportunity yet to check whether they went off. This is a terrible trag-
edy and I really don't know what made them do it. (TODAY)
Even though the meaning of these nouns is predestined for event descrip-
tions, the combination with that-clauses turns them into shells for states of
affairs.
'Trouble'
Shared semantic features: [EVENTIVE], [ATTITUDINAL], [MANNER],
[DEONTIC]
Nouns: Problem, trouble, difficulty, dilemma, snag
Attitudinal eventive uses 271
None of the five members of this family is new to us; all five of them can
also give rise to factual attitudinal uses when they are combined with that-
clauses. Their uses with infinitives, which are at stake here, are fundamen-
tally different from the factual ones. What collocations such as the problem
is to ... evoke besides the attitudinal meaning is a semantic complex involv-
ing the manner of activities and deontic modality. Example (11.11) and the
paraphrases proposed in (11.11') and (11.11"), especially their highlighted
passages, show how the components of this structure interact:
(11.11) Several years ago I was involved in a project aimed at measuring the
reflectivity of metallised glass. The problem was to safeguard the many
civil radar sites round Britain from encroachment by property develop-
ment. Increasing numbers of buildings were being designed then with
metallised glass, (NEWSCI)
(11.11') ... we had to safeguard the many civil radar sites round Britain from
encroachment by property development, but we did not know how to do
it. ...
(11.11")... measuring the reflectivity of metallised glass. The problem was how
to safeguard the many civil radar sites.
The speaker of (11.11) uses the pattern to portray the activity expressed in
the infinitive clause as an objectively given necessity or need, or an obliga-
tion which has been issued by an unnamed authority. This deontic meaning
must be due to the grammatical meaning of the infinitive clause because it is
not inherent in the structure of the noun problem (see also Section 10.3) As
suggested by the paraphrase in (11.11'), the collocation the problem was to
expresses that how this task can be solved was not known at the reported
time. That lacking knowledge as to the manner of solving the task indeed
plays a role can be shown by inserting the interrogative pronoun how in the
utterance (11.11). This test results in the paraphrase given in (11.11"),
which contains a subordinate interrogative w/z-clause. As argued by Quirk
et al. (1985: 1051), subordinative interrogative clauses resemble wh-
questions semantically "in that they leave a gap of unknown information".
The fact that the insertion of the pronoun how is perfectly unproblematic
suggests that there is at least one interpretation of (11.11) which brings a
manner element into the focus as unknown information. We will see later
that the source of the feature [MANNER], which is not marked on the lin-
guistic surface in examples like (11.11) either in the nouns or in the shell
contents, is again probably the infinitive construction (see Section 12.3).
272 Eventive uses
Depending on how a given instance is interpreted, then, uses of 'Trou-

ble' nouns in the pattern N-e-io can have eventive, modal or circumstantial
meanings. Circumstantial readings are particularly likely when 'Trouble'
nouns occur in the pattern -be-wh. Problem occurs as many as 277 times
in this pattern in the corpus, dilemma 19 times, difficulty 14 times and snag
3 times. An illustration is given in (11.12).
(11.12) The bicycle police logged 40 arrests in the first three months of opera-
tion. Now there is so much competition to use the bikes that a rota has
been drawn up. One problem is what to do with the bicycle after an ar-
rest. (PAPERS)
'Success'
Shared semantic features: [EVENTIVE], [ATTITUDINAL], [POSITIVE]
Nouns: Success, achievement, improvement, triumph,
coup, innovation, refinement
Looking at the first example of'Success' nouns given in (11.13) below, one
may get the impression that there is not much difference to the factual uses
because the infinitive could easily be replaced by a tfiaf-clause (which is
given in (11.13')).
(11.13) Sher's achievement was to find the right rhetorical tone wheedling,
mean, enraged, self-pitying for all the different, high-pitched tableaux.
(PAPERS)
(11.13') Sher's achievement was that she found the right rhetorical tone whee-
dling, mean, enraged, self-pitying for all the different, high-pitched tab-
leaux.
It could of course be argued that (11.13') has a more factual ring to it than
(11.13), and that, as a result, (11.13') must be regarded as a factual use of
the noun achievement. However, it is not necessary to make such an unpro-
vable claim, because the corpus data suggests that people hardly ever use
the combination illustrated in (11.13') - except when the predicate in the
complement clause is the copula be. This stative verb sounds decidedly odd
in an infinitive clause which is equated with achievement, because this noun
Attitudinal eventive uses 273
has a structure-inherent gap for an accomplishment. As a result speakers

tend to opt for that-clauses under these circumstances. These few examples
suggest either a factual or a facetious reading.
In the pattern -be-to, where they are found most widely, attitudinal
eventive nouns tend to be used to express CHARACTERIZATIONS of past
events. As a result, simple past forms of the copula, as in (11.13), are more
frequent than present tense forms. However, present tenses with generic
time reference are possible as well, as is shown by example (11.14):
(11.14) The school's triumph is to get the children to sit still, read, recite sto-
ries ... (PAPERS)
'Success' nouns are also found in the patterns th- and th-be-N.
'Mistake'
Shared semantic features: [EVENTIVE], [ATTITUDINAL], [NEGATIVE]

Nouns: Mistake, crime, fault, error, sin, offence,
folly
The negative counterparts of 'Success' nouns are also normally found in the
pattern -be-to, th- and th-be-N. Like achievement, mistake occurs very
rarely in the pattern -be-that when the verb in the subordinate clause is the
copula. Mistake is the only member of this family which is also found in the
pattern N-to. Almost invariably, these uses are extrapositional constructions
of the type illustrated in (11.15):
(11.15) It's a mistake to think a model can be an actress, because the two pro-
fessions don't have much in common, (MAGS)
The noun folly is used in this type of construction, too, but without a deter-
miner (e.g. it is folly to believe that..., PAPERS).
Another parallel to the 'Success' family is that 'Mistake' uses also gen-
erally have past tense copulas, speakers tending to express their attitudes to
past as opposed to other events. This is illustrated for a -be-to use and a
th-be- use in (11.16a) and (b) respectively:
274 Eventive uses
(11.16) (a) He had done absolutely nothing wrong. His biggest crime was to sit
near the fire exit, (MAGS)
(b) I signed the deal just before he went and that was my biggest mis-
take. (TODAY)
Chapter 12
Circumstantial uses
12.1 Introduction
This chapter deals with nouns referring to situations, times, locations, man-
ners of doing things and conditions for doing things. Typical examples are
situation and context, time and moment, place and region, way and proce-
dure, and condition and proviso. A good case can be made for excluding
locative and temporal nouns from the class of shell nouns. For one thing,
especially the locative ones are not abstract in the extensional sense of this
term (see Section 5.1.1), since they refer to places in the real world which,
after all, are ontologically concrete. It is also doubtful whether these loca-
tive and temporal nouns incorporate the kind of structure-inherent semantic
gap that was claimed to be a semantic prerequisite for shell nouns (see Sec-
tion 5.2). And finally, there are reservations on the syntactic side, because,
as will be shown in the course of this chapter, the W/J-clauses following
nouns like time and place are by no means straightforward examples of
appositive postmodifiers or complements, but are often better analyzed as
relative clauses or adverbial clauses. Biber et al. (1999: 626-630), for ex-
ample, treat cases like time w/jen-clause or place where-clause as "relative
clauses with adverbial gaps". (Note that these "gaps" are the ones filled by
the wA-elements in the relative clauses, not the gaps in the semantic struc-
tures of shell nouns.)
Chiefly for three reasons, I have decided to include these nouns as pe-
ripheral cases of shell nouns nevertheless. The first is the evident concep-
tual, if not syntactic, analogy between clear cases of shell nouns like fact +
that-c\ause and aim + to-clause and the combinations time + w/zew-clause
and place + where-cimse. Like the clear cases, the latter two pairs evoke
the link of experiential identity between what the nouns and the clauses
stand for. In fact the last two pairs are particularly good examples of the
semantic match between shell nouns and shell contents, since the nouns and
the complementizers share the semantic dimensions 'locative' and 'tempo-
ral'. Second, even though the locative and temporal nouns do not include a
semantic gap comparable to the one in the structure of good shell nouns,
they are intensionally so unspecific that they are also in need of additional
276 Circumstantial uses
conceptual content to be informative. Expressions like come to the place or

meet me at the time are incomplete from a communicative point of view,
because the two nouns are too unspecific to be able to function as deictics
on their own. And third, nouns like place and time or area and stage have
the effect of conceptual encapsulation. As opposed to the world of objects
and organisms, which is inherently fragmented and structured into individu-
als, the temporal dimension and three-dimensional space are both unstruc-
tured continua. Locative and temporal expressions help the human concep-
tualizer to partition these continua by singling out specific portions. In this
respect, time and place are more similar to abstract nouns like fact or aim,
which partition portions of the amorphous abstract domain (see Section
17.1), than to concrete general nouns like person, creature or object. Al-
though they are semantically just as unspecific as time and place, the latter
three nouns do not contribute to conceptual partitioning, since they are used
to refer to individuals. Therefore concrete general nouns are excluded from
the class of shell nouns, whereas the general locative and temporal nouns
are included, even though they are of course situated on the fringe of the
class.
As their name suggests, circumstantial shell nouns can be used to create
shells that contain circumstantial information about events. They do not
characterize aspects of the conceptual core of an event, which consists of
the participants and their activities, but give information about additional,
non-focal aspects like setting, time or manner. Circumstantial shell nouns
thus shell information that is typically expressed by adverbials in simple
declarative sentences. The conception underlying this class of shell nouns
can perhaps be characterized better within verb-centered grammatical theo-
ries like dependency grammar, valency grammar, case grammar or Talmy's
(1991, 1996) event-frame approach. In Tesnire's well-known terminology
(1959: ch. 48), circumstantial shell nouns do not shell the actants and what
they do, but the circonstants.
Circumstantial uses share with modal uses that there is no ontological
correspondence between shell nouns and shell contents. Just as modal shell
nouns do not shell 'modalities' but events and facts, circumstantial shell
nouns do not shell 'circumstances' but events. And again rather like modal
uses, circumstantial uses reflect the speakers' interests in these events. As
the following description of groups and families of uses will show, speakers
can either highlight the totality of the circumstances under which an event
takes place or single out location, time, manner or condition for special
attention. These possibilities are summarized in Figure 12.1:
Introduction 277

General circum- 'Situation' situation, position, context 12.2
stantial uses
Specific circum- 'Place' place, area, region, spot 12.3

stantial uses 'Time' time, stage, moment, period
'Manner way, approach, method
'Condition' condition, case, proviso
Figure 12.1 Survey of the class of circumstantial shell-noun uses
The analogy with modal uses also suggests the general frame for the de-
scription of circumstantial uses. Since only events, but not facts, can occur
as shell contents, the frame is even simpler than the one for modal uses. It
consists of the components EVENT (expressed as shell content, and mani-
fested in one of the more specific types ACTIVITY, PROCESS or STATE) and a
CIRCUMSTANTIALITY INDICATOR, which is incorporated in the shell noun.
Circumstantial frame: EVENT - CIRCUMSTANTIALITY INDICATOR
Example:
CIRCUMSTANTIALITY EVENT (STATE)
INDICATOR
We would the rather interest- that we could prove anything we please
be in ing situation to postulate, (NEWSCI)
Figure 12.2 Frame for circumstantial shell-noun uses
The only family which cannot be described with the help of this general
circumstantial frame is 'Condition', and so a special frame will be intro-
duced for this family.
12.2 General circumstantial uses
'Situation'
Shared semantic features: [CIRCUMSTANTIAL]
Frame: General circumstantial

Nouns: Position, situation, context
I have already pointed out in Section 11.2 that it is difficult to keep general
eventive and general circumstantial uses of the nouns position and situation
apart. For convenience, I reproduce one of the examples given above,
(11.4), as (12.1) here:
(12.1) We now have a situation where thrombolytic therapy is very widely

accepted and very widely practised, (BBC)
The prepositional content of the where-clause in this example describes a

state - the state that thrombolytic therapy is very widely accepted and very
widely practised. If the determiner of the shell-noun phrase were a definite
article and the clause describing this state were introduced by that, one
would not hesitate to regard the example as an eventive use shelling a state.
There is even a case for an interpretation as a factual use under these cir-
cumstances. The trouble is, however, that proper shell-noun uses of situa-
tion with appositive, rather than relative, that-clauses are extremely rare,
especially in comparison with postnominal where-cluses.77 Because of the
strong interference of relative clauses, I have not given a score for situation
+ that-cleaise in the Appendix. The actual number of instances is below 50.
In contrast, the collocation situation + where-clause can muster more than
a thousand occurrences in the corpus. I prefer to look upon such uses of
situation as highlighting circumstantial rather eventive information because
the w/z-word where evokes a locative perspective on the scene. This sug-
gests that the speaker is more interested in the circumstances than in the
events themselves.78
The main circumstantial parameters for the descriptions of events are
place and time and therefore not only where- but also when-clauses are
possible complementizers for situation and position. It is interesting to note
that both situation and position are far more frequently followed by where-
clauses than by w/ze-clauses. This preponderance of where may be due to
the historical, morphological and conceptual connections of the two nouns
with the locative domain, and/or to the cognitive primacy of concrete space
over abstract time.
Context does not occur with w/j-clauses as complements, but in small
numbers in the pattern -be-that. This finding ties in with the observations
on where and when in the previous paragraph, since the meaning of context
is more abstract than those of situation and position, derived as it is from
the linguistic domain. The combination with that-c\auses can thus again be
considered semantically and cognitively motivated.
General circumstantial uses 279
All three nouns listed in the family header are highly productive in ana-
phoric references, especially in the pattern th-N. These collocations can also
give rise to general eventive uses. If one insists on a resolution of the ambi-
guity - the gains of which are questionable, though - the clause constituent
in which the anaphoric noun phrase occurs provides a possible linguistic
criterion. Uses in adverbials can then count as circumstantial uses (see ex-
ample 12.2) and uses in subject or object function (see 12.3) as eventive
uses.
(12.2) The music business is dead? I don't agree, it's only resting. So what can
the Press do in this situation? (MAGS)
(12.3) As more and more people want to compete at county level, shows are
restricting the number of horses each rider can enter. And this situation
is making life veiy difficult for riders trying to bring horses on. (MAGS)
Finally, attention should be drawn to the lexeme circumstance which is also

very common as a general circumstantial shell noun, especially in the plural.
Since plural forms were excluded from the quantititave data for this study,
no score for the frequent collocations in/under these circumstances (see
example 12.4) is given in the Appendix.
(12.4) If I had discovered, for instance, that the rubber exhaust pipe had just
parted from the manifold, it would have been difficult to prove fraud, as
that could merely be the result of poor workmanship, although even in
these circumstances, it would be unlikely that an insurance company
would pay up for the loss as the owner, quite evidently, made no attempt
to mitigate the loss, (MAGS)
Unlike situation and position, circumstances is not ambiguous between a

circumstantial and an eventive reading.
12.3 Specific circumstantial uses
'Place'
Shared semantic features: [CIRCUMSTANTIAL], [LOCATIVE]
Nouns: Point, place, area, position, region, spot, site
When they use the nouns of the 'Place' family, speakers highlight the loca-
tions of EVENTS as CIRCUMSTANTIALITY INDICATOR. It must be stressed
that I am only concerned with the uses of these nouns with postnominal wh-
clauses and in anaphoric patterns here. Uses in the pattern N-fo give rise to
dynamic modal readings (see the discussion in Section 10.4).
Not surprisingly, the members of the 'Place' family combine most natu-
rally with postnominal where-clauses. Illustrations of the three most fre-
quent nouns point, place and area are given in (12.5):
(12.5) (a) The Polynesians developed such a vessel centuries ago, there was
even one built in Britain in the reign of Charles Two, but only re-
cently have we progressed to the point where they have become reli-
able enough for efficient ocean crossings - the catamaran, (PAPERS)
(b) One of the most potent symbols of the Cold War the Checkpoint
Charlie border crossing at the Berlin Wall ~ has been dismantled.
For twenty-nine years, it stood at the heart of Berlin, at the place
where Soviet and American tanks confronted each other after the
Berlin Wall went up in 1961. (BBC)
(c) ... we have to remember that people in rural areas often are earning
low incomes erm but they still have a right to live in the area where
they work, (SPOKEN)
Only point and position, which can have both locative and temporal read-
ings, are found with veAen-clauses as well, but much less often than with
w/jere-clauses. In terms of absolute numbers of occurrences, point is the
most frequent 'Place' noun in the pattern N-w/j with wAere-clauses, closely
followed by place and then area. If one takes the overall frequency of the
three nouns into account, the picture changes slightly, place having a lower
reliance score than area for the pattern + where-clause.
Looking at the three examples in (12.5), one may get the impression that
they do not qualify as shell noun uses at all because the postmodifiers are
relative and not appositive clauses.79 This is the analysis of these patterns
put forward by Quirk et al. (1985: 1252-1257), for example. It is of course
true that the word where could be replaced by the relative pronouns at
which or in which in all three examples without a noticeable change in
meaning. I am not sure, however, if it would be right to regard this as une-
quivocal evidence against appositive readings. Such a strict interpretation
would step out from the assumption that there is a clear dividing line be-
tween appositional postmodifiers and other types of postmodifiers such as
Specific circumstantial uses 281
relative clauses. That this is not the case has already been shown in Section
3.1.1, and it is also mentioned by Quirk et al. in a note (1985: 1262),
though only with respect to nonrestrictive appositions and relative clauses. I
believe that the uses under discussion here must be seen as manifestations of
the feet that there is no clear-cut categorical distinction between appositive
and relative clauses, certainly not in the circumstantial domain, and perhaps
not even with //jai-clauses in the mental and linguistic ones. It will be re-
called that the main criterion for apposition is identity of reference and that
it is difficult to check this criterion if one of the groups in question does not
refer because it is a clause. If we resort to the weaker restriction of 'experi-
ential identity' then, nothing contradicts interpretations of the examples in
(12.5) as shell nouns. For example, the expressions the place and where
Soviet and American tanks confronted each other after the Berlin Wall
went up in 1961 clearly refer to the same location and therefore meet this
criterion. What this suggests is that, depending on the image that one pre-
fers, these uses are either indeterminate between appositive and relative
readings or hold a middle position on the cline between clear appositional
and clear relative clauses.
The second linguistic environment favoured by 'Place' nouns is naturally
the pattern th-N. Point and position are semantically extremely versatile in
this pattern, with readings ranging from the factual through the linguistic
even to the mental domain. The proportion of clearly locative readings of
the collocations this/that point/position is fairly low. Unquestionable loca-
tive readings are however produced in large numbers by place in the pattern
th-N, and even more frequently by area. The latter is also used in trans-
ferred meanings in abstract domains, but much less often than point and
position. Both place and area are determined far more frequently by this
than by that. The ratio of this to that uses without intervening premodifiers
is roughly three to one in my data.
To give the reader an idea of the premodifiers that are typically used
with 'Place' nouns in the pattern th-N, I have extracted the relevant data for
place, area and region from the PAPERS corpus section. The adjectives are
listed in Figure 12.3. Looking at the data for this first, it is interesting that
area is more frequent than place. While place is almost exclusively com-
bined with evaluative adjectives, area occurs also with descriptive ones like
huge and with restrictive adjectives (crucial, entire, important, principal,
vital, whole). Region occurs mainly with adjectives indicating the geo-
graphical location or the character of a place. Somewhat unexpectedly, the
adjectives occurring with that are on the whole more descriptive than
evaluative. Here, place can muster more occurrences than area.
this Adj this Adj this Adj that Adj that Adj that Adj
place area region place area region
awful crucial (2) autonomous extravagant geographical same
beautiful entire charming elusive notorious
cosy grey intractable grandiose particular
dark huge (2) mountain particular unclassifiable
empty important (2) northwestern same
fabled lovely poor small (2)
foul middle remote soulless
little particular scenic
splendid principal Serbian
strange safe subterranean
terrible tumbledown temperate
unhappy unfortunate textile
unique vital
whole (3)
Figure 12.3 Collocational data on this/that+Adj+place/area/region from PAPERS
Th-N uses of 'Place' nouns are not always shell nouns, as the combination
of deictic determiners and locative nouns can of course be used in deictic
rather than anaphoric function. The collocation this place, for example, is
often used in deictic function to refer to a house or flat. Therefore one
should not overestimate the significance of this highly productive pattern for
this study. The same is true of the uses of 'Place' nouns in the pattern th-
be-N, which tend to involve evaluating references to locations in the extra-
linguistic world.
'Time'
Shared semantic features: [CIRCUMSTANTIAL], [TEMPORAL]
Nouns: Time, stage, moment, period, age, era
The 'Time' nouns are the temporal analogue to the 'Place' nouns. Their
description can therefore be shorter because a number of general points that
are valid for both domains have already been made.
Like 'Place' nouns, 'Time' nouns are used in the patterns N-cl, th- and
th-be-N, and, again like the latter, they are not very reliable shell nouns in
th- and th-be-N. 'Time' nouns are of course not combined with postnomi-
nal where-clauses but with wAew-clauses. See the illustrations in (12.6):
(12.6) (a) 'So Young' comes from a time when me and my close friends were
caught in a bit of a desperate drugged-up situation and there was one
particular night when one person nearly died, (MAGS)
(b) It will be a real wrench to leave the club after eight years but I have
got to a stage when I must now look at what is best for my future.
(TODAY)
(c) They helped raise money for the Clinton campaign, lent the Clintons
their California home, and introduced them to the Beverly Hills hair-
dresser Christophe. That was the moment when the glamour turned
tawdry, with the controversy over the president's Dollars 200 haircut
while Air Force One blocked the runways at Los Angeles airport.
(PAPERS)
The examples show that 'Time' nouns also resemble 'Place' nouns with
respect to the unclear status of the postnominal clauses. It is again possible
in all three examples to replace when by at which or in which.
Example (12.6c) is a manifestation of both patterns N-cl and th-be-N, a
combination which is not rare and also found with 'Place' nouns. In such
uses, an EVENT is taken up by an anaphoric pronoun this or that which
functions as subject. This pronoun is then linked by means of the copula to
a use of a 'Time' or 'Place' noun in the pattern N-w/i. People exploit the
shell contents of these constructions to express a second perspective on the
EVENT. As illustrated in (12.6c), the first EVENT description is typically
attitudinally neutral, while the second contains an evaluative appraisal of
what happenend at the given time.
Example (12.6a) raises a further issue about 'Time' and 'Place' nouns
which I have not addressed so far. The reader may have noticed that the
final part of this example contains a second combination of a temporal noun
and a following when-clause: one particular night when one person nearly
died. For two reasons, I do not regard this expression as an instance of a
shell noun. From a semantic and conceptual point of view, the concept night
does not qualify as shell noun because it is conceptually bounded. Even
though the boundaries of nights may be fuzzy in the same way as the
boundaries of hills and mountains are (see Quine 1960: 126) - the question
being at precisely what time nights begin and end - the word night repre-
sents a saturated notion with no structure-inherent semantic gaps. This is
different with all the nouns which occur in specific circumstantial uses.
The sequence of night + veAen-clause in (12.6a) also differs from the su-
perficially similar sequence time + when-clause in the same example from a
syntactic point of view. My impression is that the connection between night
and the following wAe-clause is less tight than that between time and the
other wAe-clause, and that the former wAe-clause comes much closer to
being an autonomous adverbial clause than the latter. This is in line with
Quirk et al. (1985: 1254), who regard the word when in their example "the
day when she arrived" (emphasis original) as a "wA-adverb".
With a frequency of occurrence of almost 9,000, the collocation time
when is among the most frequent ones in the whole corpus. Of course the
frequency of occurrences in the pattern /A-N is much higher but, as already
mentioned, most of them involve direct exophoric rather than anaphoric
references. As pointed out in Section 10.4, the modal collocation time +
infinitive is also highly productive. As a result, time emerges from this
study as the most frequently used shell noun altogether. This finding by no
means reflects the importance of this noun as a shell noun, since it is chiefly
brought about by the fact that time is the most frequent noun in the whole
corpus altogether and by the fact that all exophoric uses are also included in
the quantitative data.
'Way'
Shared semantic features: [CIRCUMSTANTIAL], [MANNER]

Nouns: Way, approach, method, technique, practice,
trick, procedure, recipe, norm, knack, gambit
In their discussion of relative pronouns as adverbials, from which I have

already quoted a small passage above, Quirk et al. (1985: 1254) remark
"that there is no relative how parallel to when, where and why to express
manner with an antecedent noun". Since their grammar is not just based on
the intuition of highly competent native speakers but also on the corpus data
collected in the Survey of English Usage, it is not surprising that my corpus
data support this observation - almost completely, that is. For I have found
half a dozen instances in the SPOKEN data where speakers use the combina-
tion way how in a manner analogous to place where, time when, and also
reason why, which, incidentally, I have not mentioned here because I regard
it as producing factual uses (see Section 7.3). Two of these examples are
given in (12.7):
(12.7) (a) You're hurting me 'cos of the way how you've hurt. That's what
you're doing, (SPOKEN)
(b) I can I know a way how to get up and get onto the roof and ...
(SPOKEN)
Of course the significance of such examples should not be exaggerated, and

I would not begin to argue that they constitute a falsification of Quirk et
al.'s claims to be reckoned with. On the other hand, I do not want to discard
them as "performance errors" either. Especially (12.7a) leaves the impres-
sion that the word how was not used inadvertently by the speaker, but delib-
erately in order to highlight that it is the manner of an activity that he or she
is talking about. In (12.7b), it may play a role that the use of way is more
like means in that it seems to convey an instrumental meaning which could
be paraphrased by a way with the help of which you can ....
If the collocation the way how scarcely (if at all) exists, then how is the
combination of manner shell nouns with EVENTS realized? The answer is
simple but consequential - simple, because the link is of course realized by
infinitive clauses, and consequential, because this would mean that the same
kind of semantic dovetail that we find between place and w/jere-clauses,
time and w/jerc-clauses, and reason and w/ry-clauses holds between way and
fo-infinitives. This leads to the inevitable conclusion that infinitive clauses
have an inherent meaning of manner, in addition to the complex of wishes,
aims, achievements, obligations, abilities and attempts attributed to them so
far (see Section 11.3). Additional support for this claim comes from nouns
which have no inherent manner meaning and are therefore an even better
testing-ground. As we have seen, the members of the 'Trouble' family
problem, trouble, difficulty and dilemma, which can by no means be said to
include the dimension 'manner' in their stable semantic structures, turn
ambiguous between an eventive reading and a manner reading as soon as
they are combined with infinitive clauses (see Section 11.4).
The fact that 'Way' nouns are combined with infinitive clauses has one
further consequence for their description here: their uses in the pattern N-to
are systematically ambiguous between manner readings and dynamic modal
readings. Since I have already addressed this aspect in Section 11.4, I will
not further dwell on it here, but focus my attention on uses of 'Way' nouns
in the pattern -be-to and th-N, which are more typical of 'Way' uses any-
way.
The major characteristics of uses in the pattern -be-to will be outlined
with reference to the following four typical examples:
(12.8) We are winning against drugs but the only foolproof way is to bring in
blood tests, (PAPERS)
(12.9) BILLINGS: But you must have tried other methods in the past, what
were they? NEWEY: Well the usual method is to apply synthetic lac-
quer. transparent lacquer, over the surface of the object, (BBC)
(12.10) Eric's approach is to shoot it in the shade on a bright sunny day, at f/11
and 1/30 sec on Fuji RDP 100. (MAGS)
(12.11) To get sturdy seedlings, the trick is to keep the compost slightly on the
dry side, never too wet, but make sure that it doesn't dry out fully.
(MAGS)
Example (12.8) is representative of uses of the noun way in the pattern N-

be-to because it involves the focusing adverb only. This is very typical of
such uses. The other typical collocations of way in this pattern, one way is
to, another way is to and adjectives in the comparative or superlative, are
also mainly restrictive and comparative in nature. Comparative determiners
and premodifiers are also found with many other nouns in this set, but
rarely with trick, practice and recipe.
Example (12.9) illustrates habituality, a second aspect commonly found
in these uses. The main lexical means of introducing this aspect in the pre-
modifier are the adjectives standard, usual and traditional. Besides the
noun method, mainly approach, technique, practice and procedure are
found in such uses.
In (12.10), habituality combines with a reference to the agent who is
portrayed as performing a certain ACTIVITY habitually in a certain way.
Such combinations of 'Way' nouns with possessive determiners are com-
mon with approach, recipe and gambit, though it must be added that the
latter two nouns are perhaps too rare in my data to warrant such a claim.
Finally, (12.11) illustrates the typical usage of the noun trick in the pat-
tern -be-to. What is so typical about it is that it has no premodifier and is
determined by the definite article. Less typical, but all the more illuminat-
ing, is the fact that the aim of the shelled ACTIVITY is expressed explicitly in
the clause-initial adverbial infinitive clause. While it is quite normal that the
aim is hidden somewhere in the context, this is a good reminder that most of
the uses that are at stake here at least imply that the ACTIVITY is carried out
by the agent with a certain aim or achievement in mind. The impact of the
CIRCUMSTANTIALITY INDICATOR contained in the 'Way' shell nouns is that
the ACTTVITY is portrayed as an instrument in pursuit of the aim. With this
recognition the question as to the grammatical meaning of infinitive clauses
has come full circle, because we have found the intrinsic link between man-
ner and the wish-aim-obligation-ability complex.
In uses of the types illustrated in (12.8) to (12.11), the aim or intended
achievement of an instrumental ACTTVITY can also be expressed by means
of a postnominal infinitive clause - which would then, not surprisingly un-
der the circumstances, give rise to dynamic modal uses. See the illustration
in (12.12), where the postnominal infinitive represents the intended
achievement and the complementing infinitive the ACTIVITY portrayed as
instrumental.
(12.12) Some believe the only way to win is to press home personal attacks on
Mr Clinton, bringing up the draft-dodging issue ... (PAPERS)
Alternatively, the aim or achievement can be expressed by means of post-

modifying /-prepositional phrases with gerunds as prepositional comple-
ments. An illustration of this usage is given in (12.13).
(12.13) He argues that the swiftest and most effective way of checking the
population explosion is to encourage women to take paid jobs, (ECON)
It is certainly unfortunate that this example is missed as an instance of way

by the automatic corpus query and wrongly attributed to explosion, because
the pattern way + of + gerund is fairly productive. In a random sample of
1,000 uses of the noun way, no less than 70 instances of this pattern were
found.
Like locative and temporal circumstantial shell nouns, 'Way' nouns are
also very frequently used in the pattern th-N. In contrast to the two other
families, however, exophoric references are fairly rare. The only situation
where they seem to be feasible are utterances which are accompanied by
pictures or gestures indicating the way in which an ACTTVITY is, or is to be,
performed. And even in such situations, the only noun likely to occur is
way. By far the majority of uses of 'Way' nouns in the pattern ih- are
anaphoric. See example (12.14):
(12.14) ... so the ribbon can be used another ten times. The potential is there to
save a million pounds a week, and as over 95% of the ribbons used in
the UK are imported, it would be rather nice to cut the balance of trade
deficit this way. (BBC)
More interesting than such purely anphoric uses, which merely shell an
ACTIVITY as a manner or instrument, are uses with premodifiers which al-
low speakers to express additional descriptive or attitudinal information. To
give an impression of the kinds of premodifiers that are typically used, I
have compiled the adjectives and nouns that occur in the PAPERS section of
the corpus between the determiner this and the nouns approach, practice
and way (see Figure 12.4). The most striking feature of this list is the enor-
mous diversity of the adjectives that were found. Only new occurs more
than once with one noun and with more than one noun. This adjective is also
the most frequent premodifier of method, procedure and technique, which I
have not included in Figure 12.4 because the number of relevant examples
was too low for them to be illuminating. There is a wide range of adjectives
which can hardly be regarded as giving rise to something like collocations,
which are by definition recurrent co-occurrences.
approach practice way

acid innovative bizarre arbitrary voluntary
amateur inquisitorial dubious bottle wider
authoritarian irrational common constructive
cautious latter heinous delightful
fastidious layered horrible gradualist
fundamental liberal outrageous hierarchical
gradual linear sales moribund
haphazard market scandalous new (2)
humane monolithic sinister original
moribund straight painless
new (2) sympathetic particular
original voluntary straight
particular wider sympathetic
Figure 12.4 Collocational data on the pattern this+Adj+approach/practice/way
Nevertheless, the adjectives suggest conclusions concerning the meaning of

the nouns. The most eye-catching tendency is certainly the accumulation of
negatively evaluative and distancing adjectives occurring with practice.

Applying ideas put forward by Louw (1993) and Stubbs (1995: 24-29), one
could say that practice has a decidedly negative "semantic prosody".
Speakers seem to choose this noun when they do not approve of the way in
which ACTIVITIES are habitually carried out. Approach and way, on the
other hand, occur with surprisingly few evaluative premodifiers. It is true
that words like amateur, fastidious, irrational, sympathetic, delightful and
perverse contain evaluative elements, but such premodifiers are clearly
outweighed by predominantly descriptive ones. This suggests that the nouns
themselves are attitudinally neutral.
'Condition'
Shared semantic features: [CIRCUMSTANTIAL], [CONDITIONAL]
Frame: Conditional
Nouns: Case, condition, event, provision, proviso,
criterion, stipulation, restriction, limitation,
constraint, precondition
The last family to be described here has a more complex cognitive setup
than the previous ones because it involves the notion of contingency. The
nouns in this family are used to shell the dependence of one event on an-
other. The clearest examples of conditional uses are occurrences in the pat-
tern N-e-cl, two of which are illustrated in (12.15) and (12.16).
(12.15) A further precondition was that the Single Act which was regrettably
unbalanced in 1985 by the absence of a common policy towards the out-
side world, should be renegotiated to include such a provision, (PAPERS)
(12.16) The only proviso is that the novice must weigh less than 13 stone.
(PAPERS)
The examples will help me to relate these nominal expressions of condition-

ality to the more familiar ones in terms of if-then clauses, and by doing so,
to produce a frame for their description. As is well-known, conditional
clauses involve two propositions that are related in a way such that the truth
of one proposition, which is called apodosis and expressed in the main
clause, is dependent on the truth of the other, the protasis, which is ex-
pressed in the //Clause. Since I have refrained from dealing with my mate-
rial in terms of the truths of propositions, I have described this relation

above as the dependence of one event on another. Nevertheless, it will be
helpful to call the dependent event APODOSIS and the event on which some-
thing depends, PROTASIS.
In the -be-that constructions illustrated in (12.15) and (12.16), the
PROTASES correspond to the shell contents and are encoded by that-clauses.
This emerges from the fact that the /za-clauses can be replaced by if-
clauses, if the tense in the //-clause and the main clause are adapted, as is
shown in (12.15') and (12.16'):
(12.15') This will only happen if the Single Act, which was regrettably unbal-
anced in 1985 by the absence of a common policy towards the outside
world, is renegotiated to include such a provision.
(12.16') This will only happen if the novice weighs less than 13 stone.
The shell nouns themselves do not, contrary to what one may perhaps ex-
pect, refer to the APODOSES, but they create the conditional relation by their
equation with the PROTASIS. They roughly contain what is paraphrased as
this will only happen / / i n (12.15') and (12.16'). It is only by virtue of this
meaning of the nouns that the that-clauses acquire their conditional mean-
ing. In both (12.15) and (12.16) references to the APODOSES are not in-
cluded, but it is clear that they are mentioned in the preceding context.
-be-that uses of 'Condition' nouns are not particularly frequent. They
only occur with the nouns which have the conditional meaning as part of
their semantic structure, but not with the general nouns case and event.
When 'Condition' nouns are used in the patterns N-cl or th-N, the
APODOSIS is usually mentioned explicitly. It is in these two patterns that the
general nouns case and event occur as conditional nouns. The normal con-
text for conditional uses of these two nouns is a prepositional phrase intro-
duced by in which functions as adverbial. In the pattern N-cl, illustrated in
(12.17), the APODOSIS is expressed as the main clause, and the PROTASIS as
the -clause which follows the noun:
(12.17) An independent judge will pick an entry at random. In the event that
the contents do not comply with the rules, further envelope(s) will be
drawn until the winner has been found, (TODAY)
As example (12.17) and its straightforward paraphrase in (12.17') below

show, such uses can simply be seen as more formal variants of conditional
//clauses. From this point of view, the collocation in the event that can be
looked upon as a grammaticalized complex conjunction.
(12.17') An independent judge will pick an entry at random. If the contents do

not comply with the rules, further envelope(s) will be drawn until the
winner has been found, (TODAY)
Uses in the pattern th-N, however, involve more than just stylistic abstract-
ness, because they contribute considerably to the condensation or integra-
tion of information (see Section 17.3). Example (12.18) demonstrates how
this works:
(12.18) Any visitor who intends to keep his vehicle in the country for more than
twelve months, unless he take up residence, will have to pay duty. In
this case, even if he leaves Zambia subsequently, he will be unable to
obtain refund of the duty paid, (MAGS)
In this example, the shell-noun phrase this case refers back to information
given in the preceding sentence and shells it as PROTASIS. Note that this is
only possible because it functions as complement to the preposition in. Just
like in (12.17), the APODOSIS is then expressed in the main clause of the
sentence. In contrast to (12.17), however, it would only be possible to for-
mulate the same conditional relation between the PROTASIS and the
APODOSIS without the shell noun case, if parts of the PROTASIS were re-
peated.
Chapter 13
Summary of Part II
As a conclusion to the descriptive part of this study, I will provide two sur-
veys of the category of shell nouns. The first is given in Table 13.1. As can
be seen, all classes, groups and families discussed in Part II are listed to-
gether with a small set of frequent and typical examples. In addition, infor-
mation on lexico-grammatical patterns and on types of postnominal or com-
plementing clauses is provided. For each family, the patterns which attract
the members of the families are listed in the order of their importance for the
family. If the members of a family do not occur in a certain pattern, the
corresponding reference is missing; if they are hardly attracted by one or
several patterns, the abbreviations are put in brackets. In addition to that-
clauses and to-clauses (i.e. infinitives) and w/2-clauses, postmodifying of-
prepositional phrases are given in the last column when nouns occur with
them. For all families whose members can be combined with either that-
clauses or infinitives, the less frequent complement type is put in brackets.
Although any attempt at summarizing such large amounts of detailed in-
formation involves the danger of oversimplification, some general tenden-
cies can be pointed out. With regard to the lexico-grammatical patterns, it
can be observed that factual nouns rely more strongly on the pattern N-e-cl
than linguistic and mental nouns do. The reason, as will be suggested in
Section 16.1, is that factual shell nouns are particularly useful for the fo-
cusing function fulfilled by this pattern. Especially linguistic but also men-
tal shell nouns are more often used in the patterns N-cl and th-N. This can
be explained by the potential of these nouns for reporting and referring to
(the contents of) utterances and ideas (see Section 16.2). For eventive uses
and circumstantial shell nouns in particular, the pattern th-N is by far the
most important pattern. Apart from the members of the 'Attempt' family,
eventive uses are hardly found in the pattern N-cl. Circumstantial nouns are
very rare in the pattern N-e-cl, and their uses in the pattern N-cl are often
doubtful with regard to the syntactic link between the nouns and the clauses
following them.
As far as the relation between shell nouns and the types of clauses with
which they combine is concerned, my findings from the domain of nouns
give strong support to what Givn has claimed for the domain of verbs,
Summary of Part II 293
namely that there is a systematic semantic "isomorphism" (1990: 515) be-

tween the types of complements and the items governing them. More spe-
cifically, the evidence collected on the complementation of nouns supports
and complements the claims made by Wierzbicka (1988), Givn (1990),
Frajzyngier and Jasperson (1991) and others for verbal complementation
(see Section 3.2). Let us consider each of the various types of postnominal
and complementing clauses in turn. Infinitives have been found as post-
modifiers or complements of the following types of nouns:
- directive and commissive nouns in the linguistic domain;

- volitional nouns in the mental domain;
- nouns expressing permission, obligation, ability and dynamic modality
related to the possibility, likelihood and necessity of events;
eventive nouns, especially those producing 'Attempt' uses;
- circumstantial nouns related to the manner of events.
Manifestations of Givn's notion of 'manipulation' and Wierzbicka's se-

mantic prime 'want' can easily be found in the first three items in this list.
Wierzbicka's claim that the infinitive is used to express future orientation is
also confirmed. In addition, the findings also suggest a close association
between the infinitive and deontic modality (especially obligation), dynamic
modality (especially ability), the notion of attempting and the manner in
which events take place. These affinities have more or less been overlooked
in studies of verbal complementation; presumably, the reasons are that some
of the deontic and dynamic modal verbs, for example can and must, are
linked to lexical verbs without complementizers, and that with verbs like be
able to, be allowed to and try to, the connection with -infinitives seems so
natural that it is simply taken for granted.
TTratf-clauses were found in combination with:
- factual nouns;
prepositional, assertive and expressive nouns in the linguistic domain;
- conceptual and creditive nouns in the mental domain;
- epistemic nouns in the domain of modality.
This list confirms the affinity between -clauses and the domains of
'knowing' (which includes 'believing') and 'saying'.
294 Summary of Part II
mi
e
f
<3 tf S Q Q Q 53
s I
that
-S rS
* rS
- 'S
2 %%%
tu

<u
2I "3
I S?
tu tV
u
Ci G <1 -o
tu
& <3 -
<u<5 -Cl
-c

J5
tu
? 2 V

2 2 22 2

o - <1 -
1
ei
: s;
2 -o 2

2 5,2 S 'S
*5*
* * tu
-
Oh
tu
2iA 2" "2" 2
2" "
" tu
TS
s <)i <<<5 s

-C 5 - - tu
-s
s s
2 V tU tV
u
- 3 - 2
tu
" 'S 'S 2 2 S -s -s 222 s; -e 2V 2V 2 2?
s b
S'o tu
o . n s,
^ S
tu tgu

a. o
a
o
>
0>
-s;
0. a,5s
I)
^
IR g - s
! vc, 1s tu a
tu
-g o" .15
n g u g3 . tu 5 'S"
t "K S <5? Si a,
tu U ^ bo
R H 5
g- " u W - -S -S
1 S -; ^
S
' :
v. Js. C S ?tu fe
.tu
"
O
tu -

a,
s"
I

g ~
S

I
)
0
1
ig
u o .S u
P Pi J Pi S
g 8 cS
s a et)
I
a a ce g .s
-S O
3 I ti 1
o a ce
2 U o Oh
3
Mw
3 O
IS o
3
<N
.
S
3 O a 5 3 5 5 5 5 S 'S g
S ' S ' S S ' S ' S ' S ' S ' S 1 5> i? 5 s *
cuV<u? ??? Cu? t % %

<u 9 % %<
otu
?
"9 e <5 -9 "9 -o * .
*S S "9 -V
S SI IS'S -s:
. 't "oi l "o V .9 tu
tu 9^ ti) (UtU
OH s: S S
e tu 1-o 1o
S S

s* s* s* 3
13 " " -o<u 3S 3 'S3 <1 O 'S
S 'S 'S y 'S 'St ? 3 2 ?" " S3"'S3
'Vvv3'v'v3v3 3 3'S 'S3 ? S 'S S 'S
3 3
r !
So ^y OR S
5 S
l<n
u S" -S,sc ^ ,
Ci
w O a - * I
U .2SS8 jaM g
!
O
ta I
C/5 I
STcus,
3 (X
"o
23 9> u3 3enbO I J73ce S?
oI o 5
E5
fr
I I .S cu
e 'Sd s
3
S8 Vi3 < j o .SQ i
o Pi
Sa w
IS 83
I
2 9 6 Summary of P a r t I I
o Q
j
S ' S S
o
S S S 5

S ' S ' S " S 1 P S P S S S S S S S S S ' S S
ta ta* ta* ta* ta ta ta ta* ta ta ta ta* tv*
S ?
tu
- O
?
cu
-Cl
cu
- o
? ?? V
ta
- O
tu
" O
X
<2
ta
4 .
5
V tu
5
V

55 - S
i i
5? 3
 - o - o cu .
i
y
* 3 o - c - o O ^
3? s? I I
I
S s
?t u s s S 2 V S S
i 2 % 2 2 2 i2 2i 2 2 2 2 2 2
1-7 a - O
S -S
r . S > S
CL, s: o
-S> ' S
"i
i
H
a

s?
o
^
-
t S
o
W $ s tu
g 5 o S
S ' ^
o
s
tu S o)
Q
- o > 2:
i G
u
e
T g
H
9 " s " u
.
Io
x> f

2
Q_ .. O t/3 I
1 : " s O " u
S
Q < Q Pi O
U
oa
-J5
>
U
t a
sa

CO <
co
3
3
o 3
ctj
I g
o

S
i
1
Ih O
>
I Q

-

a,
S

.P O jS S 5 S S S 5 1
-s: -s:
I t s
o
V V V V
tu tu
-O <1 -
e 3

?
-
3 3 3
tu tu
<1 <1
S
icl.
3 3 <u
s <u S 'S tu "9 - 3
s <v s O

i 3 t S s: s .
SV, I O SV,
& -S S i
^ - it sf
8 tf ^ I" IP o
I ft, O
G R '' 5 s tu tu
ft. 2
1 scu -s a S
8 ? ,
8 S -i: 1 s 4, -Ci I
Q
S a<n
C
~ S
H, .2
I
M l * g ^ I U
3 g
5 cd c biB
^S-SS IW g
o
g-
g o 3
o I o
t s'3
Ss
o
'S
8.
on

3
i
o

,
g
co
m
(U I a
G >
S w
3 M
U ^
g
298 Summary of Partii
On the most general level possible, the data collected here suggests a
clear division between event-related nouns, which combine with infinitives,
and nouns that are related to facts and propositions, which take that-clauses
as postmodifiers and complements. If one accepts that the meanings of the
nouns match up with the meanings of the clause types, this general division
clearly confirms Frajzyngier and Jasperson's claim that infinitives belong to
the de re domain and thai-clauses to the domain of de dicto. It also supports
Halliday's fundamental distinction between 'proposals' (which are related
to "goods-&-services") and propositions (which are related to the exchange
of information; Halliday 1994: 70-71). The only types of nouns which sys-
tematically straddle the boundary between event-orientation and fact-
orientation are directive and commissive nouns in the linguistic domain and
volitional nouns in the mental domain. The very meaning and nature of
these nouns makes it possible for them to take either type of complement,
depending on whether they are linked up with future events or possible fu-
ture facts.
An alternative way of accounting fo the interaction between shell nouns
and types of complements is to claim that shell-content complexes are Con-
structions in the sense of Construction Grammar (see Fillmore, Kay and
O'Connor 1988, Goldberg 1995). This account is particularly useful for
cases where this interaction results in meanings that cannot be derived from
the parts making up a pattern (see Section 10.4).
In Table 13.2, an alternative perspective on the material is proposed.
The matrix representation chosen for this table enables the semantic rela-
tions and contiguities between classes of shell nouns to be indicated to some
extent. As can be seen, the four classes of factual, linguistic, mental and
eventive uses provide the main division of the field. This reflects the fact
that shell nouns can shell experiences of four types, viz. facts (including
states-of-affairs), utterances, ideas and events. Circumstantial uses are of
marginal importance in contrast to the four other classes and are used to
shell aspects of events. As was pointed out at the beginning of Section 10.1,
modal nouns also shell events or, in the case of epistemic modality, facts. In
a spirit not unlike that proposed by Ransom (1986; see Section 3.2), Table
13.2 suggests that the dimensions of attitude and modality cut across the
four major classes. All four classes include groups of mainly attitudinal
nouns, and the relation between the types of modality and the other classes
emerge as well. Not just in the factual and the eventive, but also in the lin-
guistic and the mental domain, there is a clear connection between epistemic
modality and fact-relatedness, and deontic (and dynamic) modality and
Summary of Partii 299
event-relatedness. In the classes of linguistic and mental shell nouns, it can

be seen that illocutionary and psychological-state uses can also be differen-
tiated internally on the basis of the distinction between event-relatedness and
fact-relatedness.
A second feature of Table 13.2 is that it depicts the typicality gradient in
the category of shell nouns (see also Section 6.1). The dark hatching indi-
cates prime instances of shell nouns. This area is made up of most factual
nouns and of the classes of propositional and conceptual nouns, all of which
are also fact-related. (Only the 'Tragedy' and the 'Miracle' families are less
good examples on syntactic grounds.) These classes consist of synchronic-
ally unanalyzable nouns with the potential for projecting any kind of com-
plex experience into the world of abstract relations. Good shell nouns are
similar in their functional potential, but they are less indispensable as they
are derived from verbs or adjectives. Finally, on the periphery of the cate-
gory, there are general eventive and circumstantial nouns, whose use in the
function of shell nouns is fairly restricted.
300 Summary of Partii
Part III
Functions of shell nouns
Chapter 14
Introduction to Part III
I have devoted considerable space to the questions of what nouns are used
as shell nouns and in what ways. This was necessary because abstract
nouns in general, and the nouns that typically lend themselves to shell-noun
uses in particular, have so far been neglected in linguistic literature. Pre-
sumably due to their very abstractness and concomitant elusiveness, a sys-
tematic descriptive account of such nouns has been lacking. In the third and
final part of this study, I will turn to the question of why speakers use these
nouns. In answering this question, I will generally focus on prime and good
examples of shell nouns rather than peripheral ones.
I have long pondered on whether it would be more appropriate to talk of
effects of shell nouns in this part, instead of functions. This would have
shifted the focus of this part slightly, from the purposes that shell nouns are
used for to the communicative impact they have. The difference between
functions and effects is largely paralleled by the contrast between a speaker-
oriented and a hearer-oriented perspective on speech events. Linguistic ele-
ments and constructions can be said to have functions in the sense that they
are exploited by speakers in order to achieve certain effects in the hearers.
This parallel shows that functions and effects are ultimately two comple-
mentary perspectives on the single central question of Why?. It should not
give rise to confusion, if I thus resort occasionally to descriptions in terms
of effects rather than functions when this perspective is more illuminating.
Often, one can only attribute a function to a thing or phenomenon when one
has observed its effects.
What this short discussion of the notion of function also shows is that,
as I conceive of them here, functions ultimately reside in the language user
and not in linguistic elements themselves. Thus when it is claimed, for ex-
ample, that "shell nouns have the function of concept-formation" this is just
shorthand for "shell nouns lend themselves to being exploited by speakers in
the service of concept-formation".
This part falls into three chapters. In Chapter 15,1 will suggest a num-
ber of semantic (and syntactic) benefits that may induce speakers to use
shell nouns. To a large extent, this chapter gives more substance to the cri-
terial function of characterization (see Section 2.2). Chapter 16 fleshes out
304 Introduction to Partili
the criterial function of linking, by providing a more detailed analysis of the

textual functions of shell-noun uses. In addition, pragmatic and rhetorical
aspects of shell-noun uses will be discussed in that chapter. In Chapter 17,
finally, I will look at cognitive functions of shell nouns, which have mainly
to do with the third criterial function of temporary concept-formation. The
idea behind the notion of cognitive function is that the use of shell nouns
can be motivated by properties of the cognitive system and its processes. I
regard this discussion of cognitive functions as the pivotal part of this
study, because many descriptive aspects, as well as the other pragmatic
properties of shell nouns, can ultimately be traced back to the potential of
shell nouns to turn complex cognitive content into temporary concepts.
First, though, it is necessary to introduce a small set of terms which are
helpful for the functional description of utterances. The notions to be dis-
cussed are given, accessible and new information, and the pairs theme -
rheme, topic - comment and topic - focus. My aim here is not to give an
exhaustive survey of the research on these topics but simply to clarify my
use of these terms.
Given, accessible and new information

As will be explained in greater detail in Section 16.2, the well-known oppo-
sition between given and new information (cf. e.g. Halliday 1967b, Chafe
1976, Clark and Haviland 1977) must be refined so as to yield a three-fold
distinction for the purposes of this study. Building on the important work of
Prince (1981),80 Chafe (1987, 1994) and others, three degrees of familiarity
of information can be distinguished. The criteria for the distinction are the
knowledge which is assumed to be shared by the discourse participants, the
activation states of their cognitive systems and the "activation cost" (Chafe
1994: 73), i.e. the cognitive effort needed by the discourse participants,
especially the hearer, to activate a concept or proposition ,81
Evoked/active/given information: Evoked concepts are components of
cognitive models that are assumed to be active in both the speaker's and the
hearer's minds. These have traditionally been referred to as given informa-
tion. For the obvious reason that they are already active, it takes the hearer
the least cognitive effort to activate evoked concepts. Therefore,
evoked/active concepts are often either gapped, i.e. substituted by zero, or
expressed by means of pronouns with little intonational prominence.
NewAnactive information . On the other end of the scale,82 there are new
concepts, which have to be activated by hearers from scratch as it were. As
Introduction to Partili 305
a consequence, they require the largest degree of cognitive cost. New con-
cepts are thought of as being previously inactive components of cognitive
models which have to be retrieved from long-term memory (in which case
they are called unused entities by Prince) or are introduced to the cognitive
system for the first time (i.e. brand-new entities). New concepts are typi-
cally expressed by indefinite noun phrases providing detailed descriptions.
Accessible/semi-active information. The third activation state of con-
cepts holds an intermediate position between evoked/active/given concepts
on the one hand, and new/inactive ones on the other. Accessible concepts
are thought of as being semi-active. They may have been fully active earlier
in the discourse or may be evoked as parts of an already active cognitive
model. The cognitive cost for the activation of accessible information is
higher than for given information but lower than for new information. As
far as the linguistic realisation of accessible concepts is concerned, a range
from stressed pronouns to noun phrases determined by definite articles or
demonstratives is available.
Theme - rheme, topic - comment, topic - focus

In unmarked sentences, there is a close correlation between given informa-
tion and the so-called theme of the sentence, and between new information
and the rheme. Whereas given and new information are distinguished with
regard to the dimensions of assumed familiarity and state of activation,
theme and rheme are looked upon in terms of 'aboutness' (see e.g. Lam-
brecht 1994: 118-127). According to Halliday, whose understanding goes
back to the Prague School linguists (see e.g. Firbas 1964),
the Theme is the element which serves as the point of departure of the
message; it is that with which the clause is concerned. [...] The Theme is
the starting-point for the message; it is the ground from which the clause
is taking off. (Halliday 1994: 37)
Although Halliday emphasizes that the theme is not defined in terms of

position in a clause, he makes it clear that as a general rule it is realized and
can therefore be identified as "that element which comes in the first position
in the clause" (1994: 38). The rheme is the remainder of the clause, where
the theme is developed. Unmarked themes, i.e. clause-initial elements that
represent given information and make up the starting-point of the message,
can be regarded as containing presupposed information.83
306 Introduction to Part III
A closely related distinction is that between topic and comment. This di-
chotomy goes back to Hockett (1958: 201-203) and is mainly used by
American linguists as a close notional equivalent to the contrast between
theme and rheme (e.g. Bolinger 1977: 32-38, Givn 1983, Chafe 1987: 22,
Gundel 1988).84 Another terminological variant is the pair topic and focus
(Sgall et al. 1973, Chafe 1980b, Sgall et al. 1986: 175-265, Dik 1989: 263-
287, Lambrecht 1994). Sometimes the pairs are also mixed, as in Quirk et
al. (1985: 1360-1377), who contrast theme with focus. I will use the term
focus in this third part because it evokes links to the cognitive ability of
attention distribution, and this is what I am ultimately concerned with. Since
I think that theme should be accompanied by its traditional partner rheme, I
will not follow Quirk et al. but contrast focus and topic.
Although various tests for the identification of topics have been put for-
ward,85 there are no generally applicable syntactic, semantic or pragmatic
procedures for the distinction between topic and focus. Since there is no
room in this study to discuss these problems and the relevant literature, it is
all the more important to provide a clear definition of my use of the two
terms. I will use the term focus for that part of a clause or sentence which is
marked for the highest degree of attention. This is in line with Dik's con-
ception o f f o c u s , who states that
focality attaches to those pieces of information which are the most impor-
tant or salient with respect to the modifications which S wishes to effect in
PA [i.e. roughly the hearer, HJS], and with respect to the further develop-
ment of the discourse. (Dik 1989: 266)
As is well-known - cf. for example Quirk et al.'s (1985: 1357) "principle of

end-focus" - this prosodically and informationally most prominent part
tends to be found towards the end of clauses in unmarked circumstances. In
must be added, however, that such descriptions of focus and prominence
are of course not entirely correct, because they suggest that these notions
have an objective existence. More realistic from a cognitive point of view is
Grosz's statement (1981: 84) that "if an entity is in focus, it is the object of
someone's focusing; it cannot be impersonally in focus". Accordingly,
Grosz defines focusing as "the active process, engaged in by the partici-
pants in a dialogue, of concentrating attention on, or highlighting, a subset
of their shared reality" (ibid.). One should keep this in mind, even though it
is stylistically convenient to use the notion focus as if the referent had an
objective existence.
Introduction to Part III 307
Dik makes it quite clear that topicality and focality do not exclude each
other. According to him, "certain topical elements may at the same time be
focal to the communication" (1989: 266). Although Halliday's notion of
focus is more closely tied to the phonological structure of information units
than Dik's, it is important to note with him that each "information unit has
either one primary point of information focus or one primary followed by a
secondary" (Halliday 1967b: 203). It must be added that there are con-
structions where the secondary focus precedes the primary and is therefore
sometimes regarded as an anticipatory focus (Quirk et al. 1985: 1388).
The understanding offocus advocated here in terms of salience and po-
tential for attraction of attention can also be related to Langacker's concep-
tion of prominence (see Ungerer and Schmid 1996: 156-204 for a survey).
Two aspects of this approach will help to illuminate focus constructions as
discussed in Section 16.1. First, supporting Halliday's view, Langacker
distinguishes between primary and secondary figures in relations, or tra-
jectors and landmarks, which carry primary and secondary prominence
respectively. Second, Langacker treats prominence as a recursive notion.
Not just sentences and clauses but also phrases and even words are looked
upon as involving gradients of prominence. From this perspective, there is
no problem in claiming different degrees of attentional prominence for dif-
ferent parts of matrix clauses and again for embedded clauses, a phenome-
non which is typical of shell-content complexes in the patterns N-cl and N-
be-cl (see Section 16.1).
The term topic is not defined in terms of aboutness, as is done by Dik,
nor in terms of prominence, but in terms of position, similarly to Halliday's
conception of theme. My notion of the term can be defined most straight-
forwardly with reference to Chomsky's (1965: 221) well-known conception,
according to which the term is used to refer to the left-most noun phrase
dominated by S in the surface structure. This definition includes no impli-
cations as to the attentional status of topics, and therefore there is no reason
why the topic of a clause should not also carry the secondary or even pri-
mary focus.
Chapter 15
Semantic functions
From a semantic perspective, the major function of shell nouns is the char-
acterization of the complex information which is given in the shell content
(see Section 2.2). This function can be realized linguistically by two parts
of shell-noun phrases, by the shell head nouns and by premodifiers. In the
descriptive part of this study, I have largely focused my attention on the
head nouns of shell-noun phrases and mentioned premodifying adjectives
only in connection with some of the most frequent nouns like question,
point and idea. In the first section of this chapter, I will take a more sys-
tematic look at the way the nouns alone allow speakers to portray the shell
contents according to their needs. In the second section, I will show how
speakers exploit the potential provided by the premodifiers of noun phrases
for different types of characterizations.
15.1 The characterizing potential inherent in shell nouns
Shell nouns have a unique potential for the characterization of complex

information (see Sections 2.2 and 5.2), which derives from a specific com-
bination of informational flexibility and conceptual stability. On the one
hand, the gaps inherent in the semantic structures of these nouns make it
possible that the nouns are interpreted together with other stretches of dis-
course. The stable denotation, on the other hand, can be exploited by speak-
ers for the characterization of the information.
More often than not, the characterization entailed in a given noun goes
entirely unnoticed, since one tends to overlook that the choice of shell nouns
is completely up to the speaker of an utterance. In the words of Francis,
who makes this claim for her class of anaphoric nouns, shell nouns "pur-
port to represent" the information given as shell content, but "in effect [...]
what they often do is to create that position" (1986: 37, emphasis original).
As has been shown in Chapter 8, the conceptualization activated by a par-
ticular speech act noun, say by accusation, affirmation, promise or excuse
for example, may in fact not even coincide with the communicative inten-
tions of the original speaker of an utterance. Linguistic, and to an even
The characterizing potential inherent in shell nouns 309
greater extent, mental shell nouns like hope, belief or plan are always re-
flections of what the speaker imputes to the ORIGINAL SPEAKER'S or
EXPERIENCER'S communicative intentions and thoughts. It is invariably the
speaker of a reporting utterance who defines the semantic details in relation
to which an utterance or idea is spelled out.
The four lexico-grammatical patterns which have been examined in this
study differ with regard to the extent to which they allow speakers to char-
acterize the information expressed as shell content. In what follows, I will
look at the four patterns in turn, starting from the most overtly marked and
least subtle cases in the pattern th-be-N.
Th-be-jV
The pattern with the most conspicuous characterizing potential is th-be-H.
In fact, as was already noted in Section 3.3, the main motivation of expres-
sions like that is a marvellous achievement (SPOKEN) is the characteriza-
tion of a piece of information. It is therefore not surprising that among the
more frequent shell nouns, especially attitudinal and highly descriptive
nouns like problem, trouble, pity, tragedy and story are strongly attracted
by this pattern. Of course, less specific high-frequency nouns like thing,
case, point, question and the superordinate circumstantial shell nouns are
often found in this pattern as well, but these tend to be accompanied by
evaluative adjectives in this pattern (see Section 15.2). The less frequent
nouns that can muster the highest reliance scores for this pattern - crux,
quibble, fallacy, misperception, mistake, generalization, pity, delusion,
misjudgment, ploy, misconception and compliment - also have highly de-
scriptive or attitudinal meanings. Remarkably, as many as five of these
twelve nouns at the top of the reliance list for the pattern th-be- belong to
the 'Illusion' family {fallacy, misperception, delusion, misjudgment and
misconception).
Th -N
In the pattern th-N, the characterization included in shell nouns is less
prominent. In the majority of examples, there is a close match between shell
nouns and shell contents. As mentioned above, the characterizing function
of the nouns is therefore hardly noticed. It is all the more interesting when
one comes across examples in which the speaker's influence is more salient.
One such example was (8.32) in Chapter 8. Another one of this type is
given in (15.1):
310 Semantic functions
(15.1) Others have concluded that political changes have reduced the utility of
applied force of all varieties, naval ones included. Since, as we shall see,
navalists pride themselves on the political usefulness of navies in peace-
time, this charge is a serious one. (BOOKS)
The antecedent and shell content of the shell-noun phrase this charge is a
that-c\2MSQ which functions as complement to the verb concluded. So the
shell noun which would match the shell content best from a purely semantic
point of view would be conclusion. Other fairly neutral mental shell nouns
like belief or idea would certainly also be compatible with the shell content.
Instead, however, the fairly strong illocutionary shell noun charge is used
here. Of course one can only speculate as to the speaker's reasons for this
choice. Perhaps the speaker felt that the noun charge allowed for a more
dramatic introduction of the danger faced by the navy, and prepared the
ground for a more thorough 'defence' against the 'charge'.
Shell nouns from all classes are found in this pattern, with no systematic
tendencies as to which types are used.
N-be-cl
The characterizing potential of uses of shell nouns in the pattern N-e-cl is
almost as large as that of th-be- uses, but it is far less conspicuous. Its
most interesting manifestations are found in the classes of factual, modal
and mental shell nouns. From the first class, not just explicitly attitudinal
nouns like problem, trouble, snag, advantage and irony (see Section 7.7),
but also basically neutral ones like thing and point, and causal, evidential
and comparative nouns (Sections 7.2 to 7.5) must be mentioned here. Typi-
cal examples of strong characterizations in the modal and mental domains
are the emphatics of epistemic necessity fact, truth and reality (with that-
clauses); the deontic nouns job and task (with infinitives); creditive nouns
like view, hope, feeling and assumption (with ^/-clauses); volitional
nouns like aim, goal, objective, idea, plan, intention, purpose and solution
(with infinitives); and attitudinal mental nouns like fear, concern and worry
(with /ja-clauses).
All these nouns, even fairly neutral ones, are used by speakers to attach
attitudinal labels to the events, states of affairs and ideas expressed in the
complementing clauses. A few representative examples of such instances in
context will make this clearer. To start with, example (15.2) is an illustra-
tion of how speakers exploit the pattern to avoid a strong commitment to
their own personal beliefs (see Section 9.3.1):
(15.2) ... gay experience of the seventies which can be seen as a modern demo-
cratic and large scale version of the turn of the century Capri experience.
But rather than a feeling of liberation and emancipation, the overall
impression is that, sexually, everybody reifies everybody else, (MAGS)
In this particular case, the speaker achieves this by shelling the piece of
information he or she wants to convey as an overall impression. This char-
acterization as a creditive state which is not attributed to a specific
EXPERIENCER implies something like 'everybody believes this'.
The speaker of example (15.3) pursues completely different aims. His or
her wish is to underscore the truth of what is strictly speaking also just a
statement of a personal belief (see Section 10.2).
(15.3) In addition, the USA's line on Kashmiri self-determination is more

cautious, and Washington has become increasingly flexible in listening
to India's position. The truth is that India now needs friends in the west
- particularly the USA - with considerably more financial muscle than
the struggling Soviet Union, (BBC)
This is achieved by shelling the claim with a very strong epistemic modal
noun, the noun truth itself.
Example (15.4) involves a deontic modal noun, task, with a comple-
menting infinitive clause (see also the discussion of example (10.19) in Sec-
tion 10.3).
(15.4) The basic assumption of these therapies is that dis-ease [sic!] may lie
within the structure of a relationship or relationships rather than within
an individual. The therapist's main task is to retain neutrality in a
situation where blame or illness is often focused on one member of the
marriage or family, and to attempt to understand the dynamics of the
system and then explain them to the participants, (BOOKS)
The relation between this shell noun and the shell content is so natural that
one tends to forget that it is again the speaker who portrays the activity
expressed by the infinitive clause not just as a task, but even as therapist's
main task.
The last example takes us back to the claim made above that the char-
acterizations in the pattern N-e-cl are less conspicuous than in the pattern
th-be-N. Undoubtedly, the nouns impression and especially truth in (15.2)
and (15.3) are fairly strong indicators of this function. The reason why I
think that even in such examples, the attitudinal element is introduced in a

more subtle way than in the pattern th-be- lies in the distribution of shell
nouns and shell contents in the clause. In the pattern th-be-N, the anaphoric
reference to the shell content functions as topic in the clause, and the char-
acterization incorporated in the shell noun represents the focus. This means
that the characterization is what the clause leads up to, it is presented as
new information and deployed as the most prominent part of the clause, so
that hearers will pay a high degree of attention to it.
In the pattern N-e-cl, the situation is reversed. It is the shell content that
carries the focus of the clause, as even a cursory look at examples (15.2) to
(15.4) shows, in which the underlined sections are invariably found at the
end of the clauses. The shell nouns, on the other hand, and with them the
characterizational component of the whole relation, are presented to the
hearer as topic. It is portrayed as given or at least accessible information, as
a point of shared knowledge on which the message can be built. The pattern
can thus be seen as a conversational gambit or a rhetorical strategy which
allows speakers to introduce their attitudes towards events and state of af-
fairs in a highly subtle way.
In short, the crucial difference between th-be-N and N-e-cl uses with
regard to characterization is that the former highlight the linguistic elements
that realize this function in the prominent end-position, while the latter pres-
ent it as being presupposed and suggest that it can be taken for granted.
This does not mean, however, that the characterization is unimportant in the
pattern N-e-cl and not taken in by the hearer. In contrast, one could imag-
ine that the characterizational component acts like a prime in a psychologi-
cal priming-task, activating information that influences and facilitates the
semantic integration of the shell content (see Section 17.3).
N-cl
Presuppositions can also be at work in uses in the pattern N-cl, even though
the clause positions of the noun phrases that contain both shell nouns and
shell contents are not predetermined as in the pattern N-e-cl. Therefore
general observations concerning the distribution of information are not pos-
sible. The main relevance of N-cl uses for what is at stake here lies in the
possibility to use shell nouns in this pattern as means of determining the
conceptual status of clauses (see Sections 5.1.1 and 5.1.2).
The best manifestations of this function are found with -clauses. As
has been shown in Section 5.1.1, the propositional contents of -clauses
can be indeterminate with regard to whether they represent events or facts,

as long as they are looked at in isolation. It is only when a higher predicate
or, in the present context, an abstract head noun is introduced that the con-
ceptual status is fixed.
There have already been many occasions in the descriptive part of this
study to draw attention to this phenomenon. Fairly disparate nouns such as
fact, proof, news, argument, thought, possibility, certainty, job and destiny
have provided such occasions.86 The time has now come, however, to pro-
vide a more coherent and systematic account. Let us first look at an exam-
ple taken from the mental domain:
(15.5) In Discovery this week we'll be looking at the idea that information is a
force not only in human affairs but in the physical universe, (BBC)
In (15.5), the shell content is characterized as a mental entity by the con-

ceptual shell noun idea. It is important to note, however, that the preposi-
tional content of the -clause itself contains no indications as to the kind
of abstract relation (see Section 5.1.2 for my use of this term) that it repre-
sents. Looked at in isolation, the clause information is a force not only in
human affairs but in the physical universe is a proposition; a more specific
statement about its conceptual status cannot be made. The neutrality and
flexibility of the clause can be demonstrated by replacing the mental shell
noun idea by nouns from other shell noun classes. The results are given in
(15.6):
(15.6) (a) In Discovery this week we'll be looking at the fact that information
is a force not only in human affairs but in the physical universe.
(BBC)
(b) In Discovery this week we'll be looking at the news that information
is a force not only in human affairs but in the physical universe.
(BBC)
(c) In Discovery this week we'll be looking at the possibility that infor-
mation is a force not only in human affairs but in the physical uni-
verse. (BBC)
(d) *In Discovery this week we'll be looking at the event that informa-
tion is a force not only in human affairs but in the physical universe.
(BBC)
(e) *In Discovery this week we'll be looking at the situation that infor-
mation is a force not only in human affairs but in the physical uni-
verse. (BBC)
The substitution test reveals that shell nouns representing abstract relations,
namely facts (a), utterances (b) and possible facts (c) can be combined with
the shell content of example (15.5) just as well as the mental noun idea.
From the classes of factual, and especially mental and linguistic nouns, a
large number of more specific shell nouns could be substituted for idea:
evidence, finding; belief, assumption, realisation, conjecture, claim, ob-
servation, statement. This shows that the speakers of such utterances are
fairly free to design the semantic complex resulting from the interaction
between shell noun and shell content precisely according to their needs and
wishes.
That there are limits to this freedom emerges from the inacceptability of
(15.6d) and (e), where an eventive and a circumstantial shell noun respec-
tively are substituted for idea. The reason why these shell nouns are impos-
sible here is that the propositional content of the shell content in (15.5) is
clearly abstract. As pointed out in Section 5.1.2, it is impossible to turn
abstract states of affairs into concrete events by means of shell nouns.
The reverse process, that is the shelling of that-clauses which can repre-
sent events by virtue of their propositional content, is not only possible but
in ample evidence (see the references in note 86).
(15.7) Vanity Fair led the field, gleefully detailing everything from Tartt's
friendship with professional enfant terrible and American Psycho
author, Brett Easton Ellis, to the news that the film rights had been sold
for the kind of money you could buy a small island with, (MAGS)
In example (15.7), the shell content bears all signs of physically concrete
events. Even though the main participants of the event are backgrounded by
means of the passive constructions, the verb sell makes a clear reference to
a commercial transaction and the past perfective had been sold marks it as
having actually taken place. All the same, the head noun news has the effect
of portraying the event as the propositional content of an utterance. With
the help of the head noun fact, the same that-dause could have also been
characterized as a fact.
How have linguists accounted for this phenomenon so far? As mentioned
in Section 3.2, those working in the early generative paradigm have tried to
explain this effect (with predicates) by postulating the existence of such
head nouns as fact, proposition and event in the deep structure (Menzel
1975). Halliday (1994: 250-273) refers to this process as projection. It
must be noted that examples (15.5) and (15.6a-c) are also manifestations of
projection in Halliday's sense. Undoubtedly, however, the effect is less

marked in these examples, as both the shell nouns and the shell contents
represent abstract relations. "Abstract argument transformations" (as such
examples as (15.7) are called by Asher 1993: 159) from potentially concrete
to abstract entities are more spectacular in a way.
There is yet a further modern linguistic theory which can be adapted to
provide an illuminating and profitable account of the phenomenon that shell
nouns in the pattern N-cl may be used to determine the ontological or con-
ceptual status of what is said in the shell content. This is Fauconniers' cog-
nitive-linguistic theory of mental spaces (1994 and 1997). In plain words,
the essence of mental space theory is that linguistic expressions provide
guidelines for the construction of mental spaces, which can be understood
as some kind of cognitive models. Technically speaking, mental spaces are
theoretical constructs represented as structured sets of elements and the
relations holding between them (Fauconnier 1994: 16). It is not necessary to
be familiar with the details of mental space theory to be able to appreciate
how it can be applied to shell-noun uses in the pattern N-cl. The crucial part
of mental space theory for my purposes is the notion of "space builders"
(Fauconnier 1985: 16-18). As the name suggests, space builders are lin-
guistic expressions that "may establish a new space or refer back to one
already introduced in the discourse" (Fauconnier 1994: 17). According to
Fauconnier,
space-builders may be prepositional phrases (In Len's picture, In John's

mind, in 1929, at the factory, from her point of view), adverbs (really,
probably, possible, theoretically), connectives (if A then , either
or ), underlying subject-verb combinations (Max believes , Mary
hopes , Gertrude claims ). Space builders come with linguistic
clauses, which typically [...] predicate relations holding between space
elements. (Fauconnier 1994: 17)
In sentence (15.8), for example, which is taken from Fauconnier, both the
matrix clause Max believes and the adverbial in Len's picture build mental
spaces.
(15.8) Max believes that in Len's picture, the flowers are yellow.
By virtue of the grammatical connection of these space builders to the

clause following them, the relation holding between the space elements rep-
resented by flowers and yellow is interpreted as being set within these men-
tal spaces, of which one is embedded in the other. Although Fauconnier

does not mention van Dijk, one can well argue that van Dijk's idea of lexi-
cal items like think as "world-creating predicates" (1977: 101; 1981: 153,
272) are precursors of the notion of space builders.
While the list given in the quotation above does not include nouns which
govern clauses, I can see no reason why shell nouns, especially factual,
linguistic, mental and modal ones, should not be regarded as space-builders,
too. With respect to space-building, these nouns function exactly the same
way as what Fauconnier calls "underlying subject-verb combinations".
Nominalizations like Max's belief that ..., Mary's hope that ... and
Gertrude's claim that... build mental spaces of 'beliefs' just as their verbal
counterparts do. And in the same way, non-derived factual shell nouns in
the pattern N-cl build mental spaces of 'facts', linguistic shell nouns mental
spaces of 'utterances' and modal shell nouns mental spaces related to 'pos-
sibilities', 'needs' and the like.
What is gained by thinking of shell nouns in the pattern N-cl as builders
of mental spaces? I think there are two points to be made in favour of this
approach. First, the metaphor underlying the notion of mental spaces is
appealing. It provides a reasonable and illuminating cognitive explanation
of why abstract shell nouns (as opposed to eventive and circumstantial
ones) can affect the conceptual status of clauses so dramatically. And sec-
ond, the notion of space-builders puts shell nouns into a wider perspective.
It unites them functionally not only with potential verbal relatives, but also
with adverbials, conditionals and other connectives which, as we have seen
at several points in Part II, are indeed close semantic and pragmatic coun-
terparts. Circumstantial, especially conditional uses, modal uses and causal
uses are relevant here.
Finally attention should be drawn to the rhetorical aspects of the less
conspicuous types of characterizations in the patterns N-cl, N-e-cl and th-
N. What do speakers gain by hiding their attitudes in parts of clauses that
have little prominence, and by determining or changing the conceptual
status of propositions? To answer this question, one must first realize that
conceptual entities like facts, possibilities, ideas, beliefs, assumptions,
statements, contentions and claims differ with regard to their status and the
significance that is attributed to them in a given culture. In rationalist West-
ern societies, facts, truths, verified knowledge and strong statements are
rated much more highly than ideas, beliefs, feelings or weak claims. As a
consequence, speakers try - especially when they are engaged in arguments
and controversial debates - to portray their own ideas and positions as
facts, truths, certainties and safe knowledge, and the ideas of their oppo-
nents as hypotheses, claims, theories and impressions. As this section has
shown, shell nouns are a very handy means of pursuing such aims because
they allow speakers to take up their own ideas and statements and those of
the other discourse participants and to characterize them the way they wish.
Especially when such characterizations are of the more hidden, inconspicu-
ous type, shell nouns thus provide speakers with a powerful rhetorical strat-
egy for manipulating the development of dialogues. As will be shown in
more detail in Section 16.1, the possibility provided by the pattern N-cl to
present information as given is also useful for that purpose.
15.2 Characterization expressed by premodifiers
So far, I have looked at the characterizing potential of shell nouns them-

selves, while backgrounding the fact that they are often accompanied by
additional characterizing expressions which function as premodifiers in the
noun phrase. Since the principal formal realisations of the functional pre-
modifier slot are adjective phrases, it is on these that this section will focus.
The characterizing potential of the premodifiers of shell-noun phrases
can be observed best if one tries to transform uses of deverbal shell nouns
with premodifiers into verbal expressions. In (15.9), two examples of asser-
tive illocutionary uses of the nouns announcement and observation are
given as input for such paraphrases:
(15.9) So when the bland officiai announcement was made on April 13th that
Chinese-British talks on Hong Kong's political future were to begin in
Beiing on April 22nd. Mr Patten could well afford the magnanimous
observation that this was "a victory for common sense", (ECON)
It is no easy task to render all aspects of the meanings conveyed by these

two shell-noun phrases with the help of verbal constructions. Evidently, the
only acceptable paraphrases for the first instance, which are given in
(15.10) and (15.11), also involve noun phrases as parts of adverbials. A
substitution of the premodifier by a non-nominal adverbial as given in
(15.12) is stylistically awkward.
(15.10) When it was announced in a bland official statement that...

(15.11) When it was announced in a bland official manner that...
(15.12) When it was announced blandly and officially that...
The problems with these paraphrases suggest that the shell-noun phrase
allows the speaker to include an additional characterization of the speech
act which cannot be included in a verbal construction equally well. This
impression is supported by the second shell-noun phrase in example (15.9),
where a paraphrase like Mr Patten observed magnanimously that... would
be acceptable but perhaps not be regarded as a particularly idiomatic sam-
ple of English. What is more, the additional information expressed by the
verb afford is not even reflected in this paraphrase.87
It seems, then, that compared to the fairly limited possibilities provided
by the grammar of English (and many other languages) for the modification
of verbal meanings, the potential inherent in the structure of noun phrases is
enormous. It may well be argued that this potential is one of the reasons
why expanded predicates like take a quick look or have a good cry are in-
creasing in importance (Algeo 1993: 214), and that it has an influence on
speakers' uses of shell nouns.
From the semantic and functional perspectives, the adjectives that I have
found to occur in shell-noun phrases fall into five groups:
a) descriptive adjectives, e.g. bungled/clandestine/unprecedented/unsuc-

cessful attempt to ... (PAPERS)
b) evaluative adjectives, e.g. that's a horrendous/terrible/tremendous
problem (SPOKEN)
c) classifying adjectives, e.g. the Fascist/scientific/medical/anthroposo-
phical thinking is that... (PAPERS)
d) restrictive adjectives, e.g. the main/only/real/whole problem is that...
(SPOKEN)
e) cohesive adjectives, e.g. that's the next/the other/a different question
(SPOKEN)
The first and second of these groups are derived inductively from the adjec-
tives found in my material. In principle, the difference between the two
groups is clear - descriptive adjectives express qualities of things, evalua-
tive adjectives express the speaker's attitude towards things. Nevertheless,
there are many instances of adjectives where it is by no means easy to de-
cide whether the speaker is using them for a descriptive or an evaluative
purpose. Even adjectives denoting size like big, enormous or huge are cases
in point. The adjectives of the third type, classifying adjectives, are men-
tioned by Sinclair (1990: 66-67) as means of identifying the class to which
something belongs. The name of the fourth group, restrictive adjectives, is
Characterization expressed by premodiflers 319
taken from Quirk et al. (1985: 430-431), who state that "restrictive adjec-
tives restrict the reference of the noun exclusively, particularly, or chiefly".
Given the contribution of these adjectives to the determination of reference,
it is not surprising that some of them are regarded as postdeterminers by
Sinclair (1990: 70). Restrictive adjectives tend to have focusing, intensify-
ing or emphasizing meanings. Finally, the notion of cohesive adjectives is
based on Halliday and Hasan's work (1976: 76-87), where it is shown that
adjectives like same, other, different and next are means of creating the type
of cohesion called comparative reference.
In this section, I will only be concerned with descriptive, evaluative and
classifying adjectives, because the other two groups do not predominantly
contribute to the characterization of information but serve other functions,
primarily focusing and linking. I will therefore return to these adjectives in
Chapter 16. Because of the functional similarity of descriptive and evalua-
tive adjectives mentioned above I will only distinguish between them when it
is necessary or illuminating to do so.
Th-be-
As can be expected, uses in the pattern th-be- provide an ideal environ-
ment for descriptive and, in particular, evaluative adjectives. Whether this
potential is exploited, and to what extent, largely depends on the style of a
stretch of discourse. To demonstrate this, I have extracted all instances of
shell nouns from the concordance for the query "that+would+be+l,lJJ+
NN", which retrieves expressions of the type that would be a terrible crime
and that would be the final proof for.... The result is given in (15.13).
(15.13)
SPOKEN
a nice acknowledgement the best place one exciting thing
the wrong attitude a good place the best way
a big benefit a big problem (2) the better way
a major compromise the primary purpose the boring way
a terrible crime a sufficient reason the cheapest way
a big difference our best site an appropriate way
a good excuse a neat solution (2) a good way
the best idea the ultimate success a fair way
an excellent idea the major thing a better way
a good idea (6) the ideal thing (2) a nice way
my natural inclination a good thing
a useful option a legitimate thing
BBC
a significant achievement a welcome development the cruellest thing

a considerable blow our wildest dream (2) the last time
a serious blow a trickier point a great triumph
a major breakthrough another major setback his final triumph
BOOKS
an ignoble act a rare irony a thankless task

a complete admission a different matter (2) the best thing (2)
a likely assignment a big mistake a good thing (2)
a political decision the great move the last time
the normal development a good place a good time (2)
an excellent idea a perfect place a good way
a good idea (3) the only reason the best way
a great idea a neat solution a stupid way
a major improvement a constant struggle
MAGS
a real incentive a great shame the best way (2)

the next job a sad story
a complete lie a terrible thing
PAPERS
our fondest dream the best outcome no great surprise

a sensible change a happy outcome the best thing
a real change a good place the only thing
the logical conclusion a real problem a foolish thing
a foolish decision the final proof a visible truth
a reasonable fear an important factor the wrong way
a different matter the immediate prospect
a great mistake the lesser reason
ECON
a remarkable achievement a severe blow the surest sign

a demanding agenda a treacherous business the prudent way
a reckless assumption a big change
a major blow a surprising judgment
Characterization expressed by premodifiers 321
TODAY
no mean achievement a correct decision the only reason

a great achievement a cardinal error an evolutionary reason
a real achievement a significant event a crazy situation
a welcome achievement a good excuse a different story
an honest answer a tough job an ideal time
a dangerous area a true luxury a real tragedy
the big attraction the right place the ultimate triumph
of invaluable benefit a routine procedure a sensible way
a further blow really definitive proof
NEWSCI
a reasonable assumption a real blow
From a qualitative point of view, the most remarkable aspect of the list in
(15.13) is that with the exception of modal nouns, all classes of shell nouns
are represented here. The nouns in the list fall into three groups. The first
consists of the semantically general and highly frequent nouns idea, thing
and way. These are particularly frequent in the material from the sources
SPOKEN and BOOKS. It can therefore be assumed that they are typically used
in face-to-face communication. Especially when the head nouns are highly
unspecific, one can well argue that the informational and intonatory focus in
such expressions is not on the nouns but on the adjectives. Nevertheless, I
would certainly not go as far as Thompson (1989: 253) does, who claims
that when they modify such "empty" head nouns, the adjectives must be
seen as having a predicative function.
The second group is constituted by attitudinal shell nouns which include
evaluative components as parts of their own meaning. Examples are
achievement, benefit, blow, breakthrough, mistake, problem, setback,
shame, success, tragedy and triumph. Thirdly, there are semantically spe-
cific but attitudinally neutral nouns like attitude, compromise, incentive,
job, proof, reason and sign. These are on the whole less strongly attracted
by the pattern th-be- than the nouns in the first two groups.
Although one should not put too much emphasis on quantitative obser-
vations here, it is worthwhile relating the absolute frequencies of occurrence
of these patterns to the size of the different subcorpora. These scores are
listed in Table 15.1. They clearly suggest that the characterizing function of
adjectives is exploited most frequently in spoken English.
Table 15.1 Relative frequency of shell nouns in the pattern 'that would be Det
Adj N'
Subcorpus Absolute score Size of subcorpus in Relative frequency per

million words million words
SPOKEN 42 20.18 2.08

TODAY 26 26.61 0.98
ECON 10 12.13 0.82
BOOKS 32 42.13 0.76
BBC 13 18.42 0.70
PAPERS 22 64.66 0.34
NEWSCI 2 6.09 0.32
MAGS 8 30.14 0.26
The source with the second highest score is the tabloid Today, followed by
ECON, BOOKS, BBC and the quality newspapers. Although the differences
between the scores are not so drastic here and low scores, especially those
under ten, are of course not very reliable, it is not surprising that TODAY
takes such a high rank in the list. After all, it is typical of tabloids that they
copy the style of spoken language and allow themselves considerably more
evaluative statements than serious newspapers.
In contrast to descriptive and evaluative adjectives, classifying ones are
very rare in this pattern. In the sample in (15.13), not a single adjective of
this type is included.
Th-N
Mental, eventive and circumstantial shell nouns occur most frequently with
premodifiers in the pattern th-N. Factual and linguistic ones are found in
this pattern as well, but modal ones are very rare. Since modal shell nouns
do not encapsulate states of affairs or events themselves but only portray
them from certain perspectives, they are bad carriers for evaluations.
As far as the adjectives are concerned, there is an astonishing range of
attitudinally strong and 'weighty' adjectives, i.e. long and rare rather than
short and frequent ones. Since the variety of the adjectives defies further
generalizations, I will illustrate the characterizing potential with a few ex-
amples extracted from the data from the PAPERS section on the pattern this
+ Adjective + Noun. These are provided in (15.14):
Characterization expressed by premodi fier s 323
(15.14)
Nouns Mainly descriptive adjectives Mainly evaluative adjectives

Factual
evidence clear, exculpatory, fresh damning, dramatic, over-
whelming
problem ridiculous, tremendous
Linguistic
argument old feeble, tendentious
question extraordinary, simple, tricky fascinating, important, painful,
perplexing
Mental
attitude casual, dismissive, equivocal, bizarre
exclusive
decision historic, sudden, tactical appalling, catastrophic, scan-
dalous,
feeling old, persistent awful, horrible, horrid
idea elusive, innovative, novel, barmy, big, brilliant, danger-
unusual ous, foolish, intriguing,
spiffing, wonderful
plan iconoclastic, massive, simple crazy, dangerous
Eventive
act irrevocable, unwitting, unwar- absurd, blasphemous, heinous,
ranted prosaic
attempt lyrical, unusual bold, daft, dotty
crime horrific, loathsome, sinister,
terrible, tragic
development fragmented, inevitable, new,
overdue, sudden, unexpected
event prestigious extraordinary, great, happy,
historic, strange, stunning,
valuable
Circumstantial
approach authoritarian, fastidious, original, sympathetic
gradual, haphazard, humane,
innovative, inquisitory, liberal,
Machiavellian, moribund, new,
practical
place awful, beautiful, cosy, fabled,

unhappy, unique, splendid,
strange
practice bizarre, heinous, horrible,
outrageous, scandalous, sinis-
ter
situation anarchic, complex, vulnerable appalling, awful, encouraging,
excruciating, horrific, scan-
dalous, sordid, uncomfortable,
unsatisfactory
The list in (15.14) testifies to the fact that an enormous range of descriptive
and evaluative adjectives occur in this pattern. Especially with the general
nouns question, idea, event, place and situation, but also with the attitudi-
nal nouns problem and crime, the evaluative function seems to outweigh the
descriptive. The general nouns are also the best proof for the idea that the
nouns themselves are mere conceptual shells introduced as anchors for the
anaphoric link created by the determiners and the characterization contained
in the adjectives. The list also reflects once more the negative semantic
prosody of practice and suggests the same for decision, feeling and situa-
tion, because negative adjectives clearly prevail in this uses.
In the pattern ih-, one finds also classifying adjectives. These are much
rarer than descriptive and evaluative ones, however, and mainly occur with
a fairly limited set of nouns. As an illustration, the most important nouns
found in the PAPERS section are given in ( 1 5 . 1 5 ) :
(15.15)
Nouns Classifying adjectives
action civil, legal
approach collectivist, Fabian
attitude moral, public, xenophobic
belief absolutist
catastrophe economic, gastronomic, social
conundrum theological
course Keynesian
idea Victorian
problem constitutional, dyslexic, economic
process democratic, educational, therapeutic
question legal
Characterization expressed by premodiflers 325
The adjectives listed in (15.15) do not describe properties of the states of

affairs or events denoted by the nouns, but they relate them to certain
spheres of experience. A legal question, for example, is not the opposite of
an illegal question but a question that is related to law. While many classi-
fications of this type may not seem spectacular, there is some potential for
powerful characterizations. Thus, inconspicuous classifications like Fabian
approach or Victorian idea can have very strong implications and serve as
subtle conversational gambits because the adjectives stand for complex
stereotypes and even ideologies and are rich in connotations.
N-be-cl
In the pattern N-e-cl, characterizing premodifiers are much less prominent
than in the patterns th-be- and th-N. Only a small number of general de-
scriptive and evaluative adjectives such as basic, good, great, important,
interesting and new are found in noteworthy numbers. These are most fre-
quently combined with neutral, partitive and attitudinal factual nouns, espe-
cially thing, point, factor, aspect and problem. Rarely does one come
across uses with modal nouns, for example possibility, and eventive nouns
like move and measure. Notable exceptions are the three nouns that can
express epistemic necessity in this pattern, truth, fact and reality. Descrip-
tive and evaluative adjectives are also rare in combination with linguistic
and mental shell nouns (but see the collocations given in example (16.5)).
All this suggests that the premodifiers in this pattern are not predomi-
nantly used for the purpose of characterization. This ties in with my claim
(see Section 15.1) that the characterizational element is not marked for at-
tention in N-e-cl uses, but presented in such a way that it is being taken for
granted.
N-cl
In spoken English, shell-noun uses in the pattern N-cl occur only rarely with
premodifying adjectives. A possible reason may be that the informational
focus of these uses is clearly on the much more weighty postnominal clauses
so that only little processing capacity is available for expressing additional
information by means of a premodifier. In carefully planned written lan-
guage, this is different. In contrast to the patterns th-be- and th-N, how-
ever, by far the majority of the premodifiers in the pattern N-cl have de-
scriptive rather than evaluative function. Besides a small number of factual
nouns like aspect, evidence and fact and modal nouns like chance and dan-
ger, mental and linguistic nouns clearly prevail in this pattern. Several types
of descriptive premodifiers can be distinguished here on the basis of dimen-
sions which I have already used for the description of these classes of
nouns.
One very common type of characterization which applies to both mental
and linguistic shell nouns concerns the strength of commitment to an asser-
tion or belief (see Section 9.3.3). Examples of typical collocations of this
sort are given in (15.16); they manifest different degrees of strength of
commitment:
(15.16) clear allegation

safe, strong, shaky assumption
clear, firm, growing, strong belief
tentative, vague conclusion
absolute, growing, unassailable, unshakeable conviction
confident expectation
strong feeling
growing realisation
A second frequent type are adjectives which refer to the degree to which an
idea is accepted and shared. See the examples in (15.17):
(15.17) general, popular, widespread belief

general, unanimous conclusion
broad, general consensus
shared conviction
general, widespread expectation
general, widespread feeling
general, widespread, popular notion
widespread perception
Speakers can also express their judgments as to the truth of a belief by

means of adjectives:
(15.18) false, unjustified assumption

erroneous, mistaken belief
absurd conclusion
false, misleading, unjustified impression
Finally, assertions can be portrayed as explicit or implicit and public or

private:
Characterization expressed by premodifiers 327
(15.19) public, tacit acceptance

implied accusation
public, tacit admission
implicit assumption
implied expectation
implicit promise
Shell nouns that are followed by infinitives are less frequently premodified
than those that are followed by that-clauses. On the other hand, nouns with
infinitive clauses combine with a greater variety of adjectives, occasionally
even evaluative ones. Examples of adjectives that were found in combina-
tion with the frequent nouns ability, attempt and plan are given in (15.20)
as an illustration:
(15.20) adept, amazing, diminishing, exhilarating, increased, innate, mental,

newfound, proven, uncanncy, unnerving ability
blatant, brave, bungled, clear, energetic, fresh, good, idealistic, ill-fated,
unsuccessful, valiant, zealous attempt
ambitious, clear, controversial, cynical, existing, ongoing, original plan
As the combinations listed in (15.16) to (15.20) show, many of these adjec-

tives spell out or specify dimensions that are parts of the semantic struc-
tures of the head nouns.
I want to sum up the major points of this chapter. In Section 15.1,1 have
looked at how speakers take advantage of the characterizing potential inher-
ent in the context-independent meanings of shell nouns. The degree to which
this potential is exploited, and the way in which this is done, depend on the
pattern in which a given noun is used. The pattern th-be-~N provides speak-
ers with a direct and overt way of expressing their attitudes about events
and abstract relations. In the other three patterns, the characterization in-
cluded in shell nouns is less marked and therefore often does not attract a
high degree of attention, especially when there is a close match between the
cognitive content of shell noun and shell content. On the other hand, speak-
ers can take advantage of the inconspicuousness of these characterizations
in order to introduce their attitudes in a less obtrusive but all the more sub-
tle and cunning way. Especially the pattern N-cl allows them to determine
or change the conceptual status of the events and abstract relations to which
the shell nouns are linked. This can be seen as a rhetorical strategy which
has the aim of manipulating the establishment of states of affairs, ideas, and
assertions in the universe of discourse as being true, false, hypothetical,
justified and so on. At the same time, the nouns in these patterns can be
used to construct information in such a way that it is taken for granted by
the discourse participants (see also Section 16.1 below).
The premise of Section 15.2 was that noun phrases have a greater
structural potential for additional specifications and characterizations of
experience than verb phrases do. I have tried to show that this potential is
exploited for different purposes in the four patterns. In the pattern th-be-N,
premodifiers are mainly used in the service of evaluations. This ties in with
the overt form of characterization that is characteristic of this pattern. Uses
of the pattern th-N exhibit a balance between descriptive and evaluative
premodifiers. Classifying adjectives are found in this pattern as well. In the
pattern N-e-cl, characterizing premodifiers are rare in comparison with
restrictive ones (which will be discussed in Section 16.1). Finally, in the
pattern N-cl, descriptive adjectives prevail. These tend to be related to di-
mensions that are part of the semantic structure of the head nouns.
Chapter 16
Pragmatic, rhetorical and textual functions
In this chapter, the scope of attention will be extended beyond the bounda-
ries of clauses and sentences. The first section will look at how speakers can
use shell-noun patterns to distribute and construct information in such a
way as to achieve specific pragmatic and rhetorical aims. In this section, the
pattern N-e-cl takes pride of place. The patterns th- and th-be- are not
so interesting from the pragmatic and rhetorical perspective, but more from
a textual one. In section 16.2, I will look more closely at the anaphoric
function of shell-noun expressions in these two patterns, focusing on the
question of how the links between shell nouns and shell contents are estab-
lished. Then I will turn to the signposting function of shell nouns (Section
16.3). Unlike Winter (1977: 3) and Francis (1986: 2), I will not use this
image for intra-sentential links and short-distance anaphora, but reserve it
for cases of shell nouns which signal long-distance relations across the
boundaries of sentences and especially paragraphs.
16.1 Focusing and topicalizing
While the bulk of this section will be devoted to pragmatic and rhetorical
functions of the pattern N-e-cl, a few remarks on the patterns N-cl, th-N
and th-be- should be made from this perspective as well. These will also
help lay the foundation for the discussion of the pattern N-Z>e-cl.
To start with the pattern th-be-N, uses of this type are prime examples
of an unmarked distribution of topic and focus.
(16.1) For a while there I was thinking, vou know. I'm gonna write pop songs,
dammit. And that was a big mistake, (MAGS)
In example (16.1), the attitudinal shell-noun phrase at the end of the clause
undoubtedly makes up the focus of the clause. Whether it is the adjective
big or the noun mistake that would carry the main stress in a spoken version
of the example is not important here. As is normal for unmarked clauses,
the end of the clause is not just the focus but it also contains the new infor-
330 Pragmatic, rhetorical and textual functions
mation. The leftmost noun phrase of the clause, on the other hand, i.e. the
topic that, expresses given information. This is particularly clear here be-
cause the anaphoric pronoun that serves no other function but to refer to
previously mentioned information. It must be added, however, that as far as
stress and attention allocation are concerned, the pronoun is definitely more
prominent than the two linking devices and and is and can therefore be seen
as carrying a weak secondary peak of prominence. Since the clause-internal
structure of the pattern th-be- is predetermined by definition, this distri-
bution of topic and focus applies to all instances of the pattern.
Uses of the pattern th-N are syntactically much more variable; general
statements concerning topic and focus are therefore impossible for this do-
main. Yet since the shell-noun phrases in this pattern by definition include
signals for anaphoric reference, one can assume that they represent given or
accessible rather than new information (see Section 16.2 for more details).
This recognition in turn suggests that these shell-noun phrases must be
found more frequently towards the beginning of clauses, and thus as topics,
than towards the end. This conclusion is clearly supported by the corpus
data, in which by far the majority of shell-noun phrases in the pattern th-N
occur in clause-initial position or after one or two clause-initial adverbials.
N-cl uses are also syntactically versatile. Shell-noun phrases of this type
can function as (parts of) all types of clause constituents - even as parts of
(expanded) predicates. Here, it is more illuminating to reverse the perspec-
tive and draw conclusions concerning the information and attention status
from the clause position. When shell-noun phrases in the pattern N-cl occur
as topics, one might expect that the shell nouns or the shell contents or both
elements represent given or at least accessible information. Looking at the
data, it turns out that cases where the information is actually given are very
rare. Much more common is the phenomenon that the information is por-
trayed by the speaker as being accessible or even given, even though explicit
clues as to where the information has been introduced are missing. Example
(16.2) is a case in point:
(16.2) Villaverde, southern Madrid, finds the big Socialist hope has a lot to
prove
By JOHN HOOPER
FOR Spain's ruling Socialists, "Supeijudge" Baltasar Garzn is an al-
most priceless asset. His decision to stand for them at the general elec-
tion on June 6 seems to guarantee something that has been called into
question too much of late - their honesty, (PAPERS)
Focusing and topicalizing 331
As in many similar examples, both the shell noun and the shell content basi-
cally represent new information here. (The only cohesive tie to the previous
sentence is the possessive determiner his) This should be uncontroversial
here, since the headline is included in this quotation and this shows that the
quoted passage is the beginning of the article. It is of course possible that
the passage is the beginning of an opinion column, in which the writer is
taking for granted that the reader is already familiar with the corresponding
news item. More likely, however, this is a rhetorical gambit by the writer.
Although such strategies may easily cause frustration with readers that are
not so well informed, they are by no means rare. What the speaker essen-
tially does is present the act and the content of the judge's decision as in-
formation that is known to himself and his readers, in short as information
which can be presupposed (see Allerton 1978: 166). It is in patterns of this
type that the shifts of the conceptual status of the shell content discussed in
Section 15.1 are particularly subtle and effective.
Much more frequent than such instances, however, are uses of shell-
noun phrases in the pattern N-cl in focus position. This is illustrated in ex-
ample (16.3), which also includes an illustration of the way in which ana-
phoric uses in the pattern th- refer back to given information:
(16.3) For example, suppose we wanted to decide whether or not company size
determines management remuneration, and formulated the hypothesis
that average management remuneration is larger in bigger firms. This
hypothesis can be either true or false and will be either accepted or re-
jected. (BOOKS)
In my view, the whole first shell-noun phrase in this example must be

treated as being in the focus of attention. This does not exclude a further
analysis of the clause with regard to the internal distribution of attention.
While the prosodie nucleus is presumably on the stressed syllable of the
word bigger, it seems reasonable from a syntactic (and semantic) point of
view to regard the whole complex noun phrase as focus
I will now turn to the pattern N-6e-cl, which is the most interesting one
from the point of view of information dynamics. Although the syntactic
structure of this pattern is also predetermined by definition, two major types
of topic-focus distributions must be distinguished, depending on the shell
nouns and their determiners and premodifiers.
The hallmark of the first type is that the shell-noun phrase at the begin-
ning of the clause is redundant from a prepositional point of view. This is
the case with the neutral factual shell nouns thing and point (see Section
7.2) and with the noun question in rogative linguistic function (Section
8.3.2). While examples have already been provided in the descriptive sec-
tions on these nouns, it will be helpful to have additional ones here as refer-
ence-points. These are given in (16.4):
(16.4) (a) Why [...] why get rid of people with experience?
I think the thing is that people have seen technology change dra-
matically within the business, (SPOKEN)
(b)... there was more equal equal income distribution but you know let's
not worry about that. The point is that everybody is better off [pause]
with this reallocation of resources, (SPOKEN)
(c) The old one is withering and the new one is flourishing, that's as it
should be. The question is when will the new one take over and pull
u p t h e o l d o n e , (PAPERS)
That the shell-noun phrases are redundant informationwise in these exam-

ples can easily be shown by omitting them together with the copulas and the
complementizer that. The results of this elimination test, which are given in
(16.4'), are not only perfectly acceptable but render the complete proposi-
tional content of the original versions:
(16.4') (a) people have seen technology change dramatically within the business
(b) everybody is better off [pause] with this reallocation of resources
(c) when will the new one take over and pull up the old one
If, then, the shell nouns are not required from a propositional point of view,
why are they used in the first place? As already explained in Section 7.2,
the basic motivation for utterances of this type is that they allow speakers to
put more emphasis on the shell content. With regard to such examples of the
nouns thing, point and question, Tuggy's claim (1996: 725) that their
function is "to focus attention on the following clause" and that they mean
"'Hey!' and little or nothing else" is correct.
As far as the distribution of topic and focus is concerned, these uses can
easily give rise to confusion, because they basically comply with the ex-
pected situation, at least with regard to the focus. After all, it is normal that
the focus is found at the end of the clause. The topic, on the other hand, is
suspicious, precisely because of its redundancy. It is ultimately this fact that
suggests that what we are confronted with here is not a 'normal' or 'canoni-
cal', but a "focus construction" (see e.g. Dik 1980: 215-229) involving a
marked focus. Inconspicuous as the construction may seem, it involves an

effect sometimes referred to as focalization (Sornicola 1994: 4638), which
marks the focus for extraordinary attention.
Further support for this claim can be provided by paraphrasing the ex-
amples in (16.4) with wh-cleft sentences, which are among the prime exam-
ples of focusing constructions (see e.g. Dik 1980: 210-229, Jarvella and
Nelson 1982: 74-75). Appropriate paraphrases are given in (16.4"):
(16.4") (a) What is important is that people have seen technology change dra-
matically within the business
(b) What is important is that everybody is better off [pause] with this
reallocation of resources.
(c) What is important is when will the new one take over and pull up the
old one.
The paraphrases with the adjective important as anticipatory focus (Quirk

et al. 1985: 1388) seems to capture both the propositional content and the
communicative impact of the examples.
The focusing function of these uses of thing, point and question is fre-
quently supported by premodifying restrictive adjectives. Thing collocates
with important, main, only and whole, point with basic, crucial, main and
important; and question with big, crucial, main and only. In uses of this
type, especially with the restrictive premodifiers first, only and superlative
adjectives, even the general eventive nouns act and action can occur in the
pattern N-e-cl with past tense copulas.
No matter whether uses of thing, point and question are supported by
restrictive adjectives or not, the wish to put an extra focus on a piece of
information can in principle be traced back to four sources. First, speakers
may want to highlight a piece of information for attention simply because
they feel it is particularly important. The ubiquity of the pattern in the idio-
lects of some speakers of English suggests that this emphatic focus may in
fact turn into a habit in spoken English. Second, the focus construction can
be motivated by the fact that what the speaker says is in contrast to what
has been said before or what is believed by the hearer(s). This is a special
variant of the emphatic focus, which can be called contrastive focus (Han-
nay 1983: 217). The element of contrastiveness is captured in Tuggy's
(1996) paraphrases by the notion of disconformity (see Section 7.2). Third,
the hidden evaluative connotations which, as discussed in Section 7.2, can
often be detected in uses of the thing/point is that can play a role. This may
have to do with hesitation, the fourth and final motivation for uses of the
pattern. Speakers may simply use the pattern in order to gain time for the
formation of their ideas and the translation of these into language. As has
been shown in (16.4) and (16.41), the thing is that and the point is that
leave a wide range of ways of continuing and completing the sentence, and
are therefore versatile enough to function as hesitators.
Thing, point and question make up a separate group of focus construc-
tions because only these three nouns are semantically so general that they
are truly redundant from a propositional point of view. The uses of this first
type are therefore only of a focusing character. This is different with the
second type of uses of the pattern N-e-cl, in which a much greater range of
nouns can be found. These nouns are semantically specific enough to add a
characterization to the shell-content complex. With regard to the topic-focus
relation, this has the consequence that in contrast to the first type, the shell-
noun phrases that function as topics also attract a certain degree of atten-
tion. Although the phenomenon is much less strongly manifested than in
typical cases like fronting, left-dislocation or cleft-sentences, this can still be
understood as a weak form of topicalization because the topic clearly re-
ceives secondary prominence. The uses of the nouns that will be discussed
now can thus be seen as giving rise to a combination of focusing and topi-
calizing: a special primary focus is deployed on the sentence-final clause
(i.e. the shell content) and a smaller anticipatory peak of prominence on the
topic (i.e. sentence-initial shell-noun phrase).
In fact, this combination of focusing and topicalizing can be achieved
with the nouns thing, point and question as well, but only when they are
accompanied by descriptive or evaluative premodifiers. This is rare, espe-
cially with the noun point, which, perhaps because its meaning is inherently
focusing, only occurs with restrictive adjectives. Yet collocations like the
interesting/amusing thing is that... and the unanswered, difficult, awkward
question is why... are in common use, and in these the adjectives do carry a
certain degree of prominence.
Much more frequent, however, are uses of more specific shell nouns in
such focusing-cum-topicalizing constructions. These nouns can be further
divided into subgroups according to the type of information that carries the
anticipatory prominence peak (see Table 16.1 below for an overview).
To begin with, there are attitudinally neutral nouns. These are either re-
lational factual nouns which insert states of affairs into certain type of rela-
tions, or fairly general linguistic or mental nouns. All these nouns are in-
variably determined by articles rather than possessive pronouns. Some, but
by no means all of them, collocate with descriptive or more often evaluative

adjectives, and most of them with restrictive ones. An exemplary list of such
uses is given in (16.5):
(16.5) Factual uses:

the big, main, only difference is that...
the clear, horrible, main implication is that...
the basic, central, main, major, real, simple reason is that...
the result is that...
Linguistic uses:
the answer is that/to ...
a possible, likely, obvious, sensible explanation is that...
the explanation is that...
the goodfad news is that...
the (final, immutable) message is that...
the (official) story is that...
Mental uses:
the common, general, implicit, underlying assumption is that
the aim, ambition, goal, intention, plan, purpose, solution is to ...
the idea is that/to ...
Topical, i.e. secondary, prominence must be attributed to those elements of

these clause-initial noun phrases which have the greatest characterizing
potential. When the noun is not premodified, it is its context-independent
meaning that is marked for secondary attention. When there is a character-
izing premodifier, as is normally the case with news for example, the pre-
modifier carries the secondary intonational and informational focus.
A second subgroup consists of inherently attitudinal nouns. With the ex-
ception of problem, trouble, difficulty, dilemma and snag, which can be
complemented by infinitive clauses (see Section 11.4), all uses in this sub-
group are factual or mental ones which are combined with i/aaZ-clauses (see
Sections 7.7 and 9.3.4). Among the factual ones are problem, trouble,
irony, snag, advantage, drawback and tragedy, and among the emotive
(mental) nouns concern, fear, hope and worry (see also the scores for at-
traction in Table 4.6 in Section 4.5). Far less frequent are the nouns down-
side, paradox, boast, consolation, peculiarity or oddity, but the reliance
scores listed for them in Table 4.6 indicate that they are also strongly at-
tracted by this pattern. All these nouns allow speakers to topicalize the at-
titudinal meanings they convey. Given that they incorporate attitudinal se-
mantic elements, it is not surprising that they are hardly ever used with
evaluative premodifiers in the pattern N-e-cl. The only premodifiers that
do occur are restrictive ones in intensifying function, as in the big/key/basic
problem, the big/main/major advantage, the big/only snag or the real/
ultimate irony.
The third subgroup comprises the three shell nouns which function as
markers of epistemic necessity, fact, truth and reality (see Section 10.2).
Although speakers may sometimes begin sentences with the fact/truth/re-
ality is that... simply in order to stress the well-foundedness and correct-
ness of their beliefs or statements, a more frequent motive is that their as-
sertions differ from what has been stated before. The use of these nouns can
thus more often be interpreted as a contrastive focus on the epistemic status
expressed by the topical shell nouns than as an emphatic focus. While the
three nouns occur widely without premodifiers, all three of them can also
muster fairly typical adjectival collocates. These are given in (16.6):
(16.6) the plain, sad, simple fact is that...

the crude, harsh, stark reality is that...
the blunt, brutal, hard, harsh, sad, simple truth is that...
With the fourth subgroup of focusing-cum-topicalizing nouns we return to

the complex of tentativeness, subjectivity and politeness discussed in Sec-
tion 9.3.1. These aspects are relevant for the collocations collected in
(16.7):
(16.7) 'BelieF family

my (personal, own) feeling is that...
my hunch is that...
my, our impression is that...
my instinct is that...
my intuition is that...
'View' family
my (personal, own) experience is that...
my, our (own, personal) opinion is that...
my, our view is that...
The hallmark of these uses is that they involve creditive nouns in the pattern
N-e-cl with first-person possessive determiners. As already pointed out in
Section 9.3.2, it depends on the nouns (and, of course, on the context) where
the secondary prominence peaks are located in these uses. These expres-
sions can function as explicit markers of either tentativeness or subjectivity.
Sometimes they can be interpreted as general politeness markers. The
purely creditive nouns in the first group tend more towards the expression
of tentativeness. Those in the second set, which denote what people think
about something rather than what they believe to be true, are more likely
candidates for strong and deliberate expressions of subjectivity. When more
intonatory prominence is on the possessive determiners, as would intuitively
seem probable with the second set, the subjectivity interpretation is more
convincing, and when the nouns themselves are stressed, the secondary
prominence peak is on the element of tentativeness denoted by the nouns.
To summarize this section, a survey of the five types of focusing uses of
the pattern N-e-cl is given in Table 16.1.
Table 16.1 Types of focusing uses in the pattern N-e-cl
Types of nouns Pragmatic Type of Topicalized se-

function(s) focus mantic aspects
Thing, point, question focusing, emphatic or 0
(hesitating) contrastive
Relational factual nouns (e.g. focusing and emphatic context-indepen-
difference, result, reason), topicalizing dent meanings of
general mental (e.g. aim, the nouns;
idea, plan) and linguistic (evaluations ex-
nouns (e.g answer, explana- pressed in premo-
tion, news) difiers)
Attitudinal eventive and focusing and emphatic speakers' attitudes
factual nouns (problem, topicalizing (or contras-
snag, trouble) and emotive tive)
nouns (concern, fear, worry)
Nouns expressing epistemic focusing and contrastive speakers' charac-
necessity (fact, reality, truth) topicalizing (or em- terization of epis-
phatic) temic status
Creditive nouns with first focusing and emphatic tentativeness or
person possessives as deter- topicalizing (or contras- subjectivity
miners (e.g. feeling, impres- tive)
sion, view)
Finally, I would like to add a coda concerning the double-' construction, as

in the thing is is that, which is discussed by Tuggy (1996; see also Section
7.2). Tuggy (1996: 722) observes that the acceptability of the double is-
construction correlates with the schematicity, i.e. the semantic unspecificity,
of the head nouns. The more specific a head noun is, the less acceptable is
the double-/.? construction (e.g. 7the upshot is is that...). He does not give
an explanation for this phenomenon, but if we take a number of his other
points into account and add the findings summarized in Table 16.1, a fairly
consistent picture emerges. Tuggy observes, first, that the focusing effect of
focus constructions is particularly strong when the information added by the
initial noun phrase is semantically "empty" (1996: 739); second, that hesi-
tation may be one of a number of factors that sanction the double- con-
struction, though he does not regard it as an important one (1996: 729); and
third, that the first is is often stressed when the noun thing is used in the
pattern (1996: 714).
Taken together, all this suggests that the double-w construction has
emerged from the canonical single-zs fActf-construction.89 With unspecific
and uninformative nouns like thing, point and question, i.e. the group of
only topicalizing uses, speakers tend to pronounce the copula with more
prominence, because the head nouns are semantically even more 'empty'
than the copula. In the thing is that I have no money, the copula is more
prominent than in the irony is that ... or the upshot is that ..., where the
nouns carry secondary prominence. In the thing is that..., it is even typical
to pronounce the copula with a falling tone. No matter whether this is ulti-
mately motivated by hesitation or not, it has the effect that the copula is no
longer experienced as a copula, since as simple linking devices, copulas
normally carry very little prominence. Having rested on the copula for an
extraordinarily long time and having given it an uncommonly large degree
of intonatory prominence and a very untypical intonation contour, speakers
seem to forget that what they just said was a copula and get the impression
that the copula is still missing. As a consequence, they continue the clause
with a copula, which then winds up being the second. The result is the dou-
ble-zs construction. I would argue that the double-is construction developed
along these lines. Once established, it may of course be transferred to other
similar patterns, that is 'sanction' related patterns in Tuggy's terminology,
as well.
Linking 339
16.2 Linking
In contrast to the previous section, this one will chiefly be concerned with
the patterns th- and th-be-N. As Section 3.1.1 has shown, the linking
function is realized in the patterns N-cl and N-e-cl by means of the gram-
matical structure of clauses. No more needs to be said about this issue here.
Instead I will focus on the anaphoric links established by the patterns th-N
and th-be-N. Cataphoric links across sentence-boundaries, which can also
be created by shell-noun phrases, are discussed in Section 16.3.
Since it has already emerged in Section 3.1.2 that only cognitive theories
of anaphora can account for the links of experiential identity, I will rely
exclusively on theories of this type for a theoretical background of this sec-
tion. These models are grounded in cognitive notions like familiarity, given-
ness and accessability, which were first brought to bear on the issue of
anaphora by Chafe (1976), Du Bois (1980), Clancy (1980), Prince (1981)
and Givn (1983, 1985).90
I will focus on two questions in this section. Firstly, how do the ana-
phoric links created by shell noun uses compare to other forms of anaphoric
reference across sentence boundaries? And secondly, how is the link be-
tween shell-noun phrases and shell contents established? The major reason
why this second question must be asked is that the antecedents of shell-noun
phrases (i.e. the shell contents) are by definition not realized by noun
phrases but by clauses or longer stretches of text. They are cases of "ex-
tended reference" (Halliday and Hasan 1976: 52). This has the effect that
the antecedents are more difficult to identify than with anaphoric references
between two noun phrases, and therefore the question of how the link is
established is more pressing.
To start with, then, let me situate anaphoric uses of shell nouns in
Prince's pioneering taxonomy (the gist of which was already introduced in
Chapter 14), and in two more recent approaches, the "Givenness Hierarchy"
proposed by Gundel, Hedberg and Zacharski (1993, see also Gundel 1996)
and the "Accessibility Marking Scale" put forward by Ariel in various pub-
lications (see e.g. 1990, 1996).
All three theories share the basic assumption that the forms of anaphoric
expressions correlate with the assumed cognitive status of the antecedents.
The higher the activation status of the antecedent is, the less explicit is the
referring form, right down to 'zero', i.e. to ellipses, for antecedents that are
in the focus of attention. The differences between the three models concern
theoretical conceptions of the cognitive statuses of concepts and the details

of the range of forms and cognitive statuses that are proposed.
Prince's taxonomy is based on the idea of "assumed familiarity" (1981:
233), and Gundel et al.'s hierarchy on the equally simple but hard-to-define
notion of "givenness" (1993: 275-285). Both notions reflect assumptions
attributed to a cooperative speaker concerning the addressee's knowledge
and attention states. Although the same is essentially true of Ariel's Acces-
sibility Theory, there is one further twist. Here, linguistic expressions are
regarded explicitly as "markers" of accessibility. It is assumed that speakers
select their referring expressions in such a way as to mark them for a cer-
tain degree of accessibility and that this is interpreted accordingly by ad-
dressees. Accessability markers differ with regard to how 'heavy' and ex-
plicit they are (i.e. their phonological size or 'attenuation'), how semanti-
cally specific they are (i.e. their informativity), and with regard to their
potential to select antecedents (i.e. their rigidity; see Ariel 1996: 21). Light,
unspecific and non-rigid elements like ellipses ('zero') or it mark concepts
as being highly accessible. Heavy, specific and rigid expressions like the
man in the corner mark concepts as being unaccessible (or containing new
information). Factors that influence the accessibility of antecedents are the
distance between the anaphora and the antecedents, the number of compet-
ing referents in the context, the degree of salience of the antecedents, and the
degree of unity and coherence between the contexts in which the anaphora
and the antecedents occur, i.e. whether they occur within one clause/sen-
tence/paragraph or not (Ariel 1990: 22-30).
The range and order of expressions that are taken into consideration in
the three models are juxtaposed in Figure 16.1. The primary aim of this
figure is to provide the reader with the background necessary for the fol-
lowing discussion of shell nouns.
Where can uses of shell nouns in the patterns th- and th-be- be lo-
cated on these scales? To begin with the pattern th-be-N, the pronouns
functioning as subjects in this pattern must be regarded as referring to 'acti-
vated' information in the Givenness Hierarchy. According to Gundel et al.
(1993: 278), this means that the antecedent, i.e. the shell content, is repre-
sented in short-term memory, but not at the current centre of attention. If it
were, the pronoun it would be chosen instead of this or that. In Prince's
model, uses of this type would be regarded as textually evoked. Statistical
data provided by Ariel (1996: 23) shows that demonstrative pronouns alone,
as opposed to demonstrative determiners accompanying nouns, are pre-
dominantly used when the antecedent is in the clause preceding the anaphor,
Linking 341
or at least within the same paragraph. This finding is confirmed by my cor-

pus data, in which the shell content is frequently found either in the same
sentence as the demonstrative pronoun and the shell noun, or in the one
preceding them. Since demonstrative pronouns alone are neither particularly
informative nor rigid designators, the distance between the anaphora and the
shell content must be fairly small and the contextual coherence fairly strong.
Assumed Familiarity Scale (Prince 1981: 237)
New Inferrable Evoked
Brand-New Unused (Noncontaining) Containing (Textually) Situationally

Inferrable inferfable Evoked Evoked
Unanchored Anchored
Givenness Hierarchy (Gundel, Hedberg and Zacharski 1993: 275)

in uniquely type
focus activated > familiar identifiable > referential > identifiable
that
it this that the indefinite this
this
Accessibility Marking Scale (Ariel 1990: 73, 1996: 21)

zero < reflexives < agreement markers < clitiziced pronouns < unstressed pronouns <
stressed pronouns < stressed pronouns + gesture < proximal demonstrative (+ noun
phrase) < distal demonstrative (+ noun phrase) < proximal demonstrative (+ noun
phrase) + modifier < distal demonstrative (+ noun phrase) + modifier < first name < last
name < short definite description < long definite description < full name < full name +
modifier
Figure 16.1 Three cognitive models of anaphora
For the pattern //i-N, it is interesting to compare uses of the proximal de-
monstrative determiners this and these with those of the distant ones that
and those. In the Givenness Hierarchy, expressions of the former type (e.g.
this statement) are rated as referring to 'activated' information, while those
of the latter type (that conclusion) are rated as referring to 'familiar' infor-
mation - a weaker degree of givenness, which does not necessarily mean
that a concept is represented in current short-term memory (Gundel et al.
1993: 278). Ariel also judges the proximal demonstratives in pronoun or
determiner function, i.e. this and these, as marking a higher accessibility
than the distal demonstratives that and those. The reason for the different
assessments is that the antecedents of this and these require that the antece-
dent is not only activated, but "speaker-activated", which means that it must
have been either introduced by the speaker or at least included in the
"speaker's context space" (Gundel et al. 1993: 279). Especially when it is
used as a determiner, that on the other hand can create references to things
outside the speaker's context space, for example to remarks of another
speaker. As is illustrated in example (16.8), which is taken from Gundel et
al. (1993: 288), this is at least doubtful in such contexts:
(16.8) A: I think that my novels are better than his.

B: I agree with that (statement)/??this (statement).
Translated into the frameworks introduced for Chapter 8 and 9, this means
that linguistic and mental shell nouns are normally determined by that in the
pattern th-N, if SPEAKER and ORIGINAL SPEAKER or EXPERIENCER do not
coincide.
In short, the cognitive anaphora models suggest that speakers can only
make use of the pattern th-be- when they assume that the shell contents
are still 'activated' in their hearers' minds. The same is true of uses of the
pattern th- with the proximal demonstrative determiners this and these,
while the distal demonstratives that and those mark the shell content as
being no more than familiar. It should be emphasized, however, that the
emotive proximity of the speaker to the shell content is the crucial factor
here, and not 'cognitive' aspects (in a narrow sense of the term) related to
knowledge and assumed familiarity.
I now turn to the second question, the question of how the selection of
the antecedents (or shell contents) is controlled. In the framework of the
anaphora models, the question is why the speaker can be so sure that the
shell content is activated or at least familiar to the reader. After all, as
propositions or even more complex pieces of information, they are not par-
ticularly salient, since they are not conceptually bounded (see Section 17.1).
Linking 343
In the pattern th-be-N, the lack of salience of the shell content is com-
pensated for by other factors. As we have seen, speakers only use this pat-
tern when the distance between anaphora and antecedent is so small and the
contextual coherence so strong that there are practically no competing ante-
cedents. The most typical case is that the anaphoric demonstrative pronouns
are sentence-initial and take up a clause-final focus from the preceding
clause.
In the pattern th-N, one encounters many examples where there is not
just a greater distance between shell-noun phrase and shell content, but also
a discourse topic boundary between the two. As the next section will show,
the shell-noun phrases are in fact often used to signal discourse topic
changes. What is more, it is typical of such examples that the antecedents
cannot always be pinpointed exactly but somehow reside in the gist of the
preceding paragraph. It is precisely in order to make sure that readers will
be able to identify the antecedents all the same that speakers use anaphoric
shell-noun phrases rather than single demonstrative pronouns. Shell-noun
phrases like this idea, this remark, this move or this issue are both more
informative and more rigid as designators than demonstrative pronouns are.
They allow speakers to help readers to interpret their anaphoric references.
How, precisely, does this work? The shell-content framework that I have
put forward in this study suggests that what ultimately governs the selection
of the antecedent is the semantic match between the gap inherent in the se-
mantic structure of the shell nouns and the meaning of the shell content (see
Section 5.2). Linguistic shell nouns, for example, have a structure-inherent
gap which can only be filled by utterances, mental shell nouns have a gap
specifically designed for ideas, and factual shell nouns a gap for states of
affairs. Depending on the further details of their context-independent mean-
ings, shell nouns can of course contain much more specific gaps, which also
require more specific types of shell contents. As a general rule, the identifi-
ability of the antecedent correlates with the specificity of the shell noun.
This is illustrated in the four examples in (16.9):
(16.9) (a) ... that these increases in transition temperature have something to do
with an increase in the strength of the cooperation between the su-
perconducting electron pairs and the vibrations of atoms in the solid.
The BCS theory predicts this link, but... (NEWSCI)
(b) ... skeletons of men, women and children have been found smashed
up, cut and burned. Since 1902 archaeologists have usually attributed
such finds to cannibalism, though few have explained their reasons

for this explanation adequately, (NEWSCI)
(c) Yet another theory suggests that the goal of the welfare state, within
a society in which economic competition under capitalism dominates,
must be to effect gradual reform. This Fabian approach argues for
... (BOOKS)
(d) The perfect compromise for them was an Art Deco style that com-
bines both elements - classic white china and stylist wood and
chrome fittings with graphic patterns and dramatic colour schemes.
So Ben and Sue decided to use this idea as their starting point.
(MAGS)
In (16.9a), there is a very tight match between the fairly specific meaning of
the noun link and the meaning of the shell content, especially the predicate
have something to do with and the relation that it sets up. Although a non-
nominal anaphoric reference by means of this would presumably also have
been successful in this example because of the short distance and the strong
coherence, the shell noun makes the reference much more precise. An al-
most equally neat dovetail holds in (16.9b) between the shell-noun phrase
this explanation and the information in the preceding clause. In this exam-
ple, the demonstrative this alone would not work, however, because its ref-
erence is not rigid enough. If this were used alone, readers would probably
activate a fact rather than an utterance as antecedent, namely the fact that
such finds are manifestations of cannibalism. The noun approach in (16.9c)
is less specific and therefore less informative than link and explanation.
Consequently, the reference is less rigid. Although I have marked the pas-
sage from the goal to gradual reform, which is also an example of a shell-
content complex, as shell content, it is not entirely clear whether the antece-
dent is only the shell-noun phrase the goal or the shell content to effect
gradual reform or the whole shell-content complex. While this indetermi-
nacy could be traced back to sloppy and imprecise language from a dog-
matic point of view, the theory of shell nouns suggests that there is no am-
biguity of reference, since the whole shell-content complex results in the
activation of one idea anyway. Note that the shell noun is obligatory in this
example, not because it determines the reference, though, but because the
main verb argue requires it. In (16.9d), the highly unspecific shell noun
idea is used. Again I have marked the whole preceding sentence as shell
content and antecedent, but again the reference is highly unrestricted.
Whether the shell noun could be omitted here is doubtful. Although idea
Linking 345
does not bring in much more semantic determinacy than the feature
[MENTAL], it results at least in the impression that the reference is precise.
The cognitive reasons for this will be discussed in Sections 17.2 and 17.3.
To a certain extent, the semantic match between shell noun and shell
content also supports the link in uses of the pattern th-be-N, even though
shell nouns do not accompany the anaphoric demonstratives as immediate
neighbours but are only equated with them by means of the copula. Via this
link, shell nouns, particularly specific ones, can also contribute to the selec-
tion of the antecedent. Example (16.10) is a case in point:
(16.10) ... five years ago Julie's legs were as thin as my arms but when she went
to the boarding school she really improved. With all the physiotherapy
and swimming she built herself up and has been managing to walk using
a frame. That is a marvellous achievement, (MAGS)
In this example, there is a very good match between the shell noun
achievement and the shell content. The semantic structure of achievement
includes a gap for conclusive activities, i.e. accomplishments (see Section
5.1.1). This gap dovetails particularly well with the first clause in the shell
content, she built herself up, because the meaning, the tense and the aspect
of the verb add up to the representation of a completed conclusive activity.
That the shell noun also matches with the second clause is mainly due to the
meaning of the main verb manage, the tense and aspect fit the gap of
achievement less well.
I have claimed that the semantic match between the gaps in shell noun
meanings and the shell contents is the reason why shell contents can be
marked as activated antecedents, even though they have little cognitive sali-
ence (as opposed to persons and objects for example). This also holds true
for anaphoric uses of shell nouns which are not determined by demonstra-
tives but by definite articles. According to Gundel et al.'s Givenness Hier-
archy, uses of this type reflect that the antecedent is not in focus, not acti-
vated and not familiar but still 'uniquely identifiable'. This is typically the
case after discourse topic boundaries, for example at the beginnings of new
paragraphs in written texts asin(16.11):
(16.11) SEARS, the troubled retailer, will today announce it has ended its 334
million outsourcing deal with Andersen Consulting and will employ 650
staff who were transferred to the consultants as part of the agreement.
The announcement will be made as part of ... (The Times, 15 January
1998, p. 23).
The shell content is uniquely identifiable here because of the semantic match
and the morphological relation between announce and announcement,
which is a case of lexical cohesion according to Halliday and Hasan (1976).
If speakers cannot assume that the antecedent is activated, familiar or
uniquely identifiable, they can resort to one of two strategies. They can
either add a reformulation of the shell content in a non-restrictive appositive
clause to the anaphoric reference, to make sure that the right information is
reactivated. An example of this type, taken from the London-Lund Corpus
of spoken conversation, is given by Meyer (1992: 87):
(16.12) A: I've read some of - not particularly known as a Shakesperean

critic, Eliot, his more general essays on Shakespeare.
B: Do you know his remarks on Hamlet?
A: Yes, I have read them, sir.
B: What are they? Will you give me the gist of his approach?
A: He, he believes that, that Shakespeare attempts in Hamlet some-
thing which he, he didn't understand himself even, and to that, to
that there, that's why it raises so many problems of interpretation
[16 tone units later]

B: Now let's go back to Hamlet then. Do you agree with Eliot's view,
that this is an imperfect play?
As can be seen, the anaphoric reference is not established by a demonstra-

tive in this example but by means of the name Eliot. Even though this con-
tributes immensely to the specificity and rigidity of the reference, speaker
obviously assumed that, after a number of intervening information and tone
units, the antecedent would not be active any more, and therefore added the
non-restrictive appositive clause. Competition between possible antecedents
also plays an important role here, because the intervening information units
are presumably also about Eliot and his views.
This brings me to the second strategy, which is not new to us at all.
Speakers can also use shell nouns in the pattern N-cl with definite articles
or, similarly to (16.12), with possessive determiners to take up information
that has been given before. The only difference to the first strategy is that
the appositive clauses are restrictive in the pattern N-cl, which means that
the information contained in them is regarded as being necessary. This is
illustrated in (16.13) with an example taken from the issue of 13 February
1998 of The Independent:
Linking 347
(16.13) Mohamed al Fayed remained at odds with Princess Diana's representa-

tives yesterday after claiming that the crash in which she and his son.
Dodi, died was the result of a conspiracy. In an unprecedented state-
ment, her office attacked speculations about the Princess's death and
said it was upsetting her sons, princes William and Harry.
Crash Investigators in Paris also refused to be drawn on Mr Fayed's
claim, saying they would wait until the inquiry was over. "Mr Fayed has
his own opinions but we are not prepared to comment on them. The in-
vestigation is already before a French judge and only when he concludes
his investigations will we have an opinion," said one of the investigating
team.
The conspiracy theorists' view that the death of the Princess was no
accident remained, at first, confined to sections of the Arab media ...
The shell-noun phrase in the third paragraph both refers back to the under-
lined section in the first paragraph and reformulates it. That such a re-
minder of the information is necessary is mainly due to the long distance
between the two elements. As for the assumed cognitive status of this in-
formation, I would argue that the restrictive appositive clause in (16.13)
marks a lower degree of accessibility than the non-restrictive one in (16.12).
Non-restrictive clauses can be traced back to such assumptions by the
speaker as think that you still know what I am referring to, but just to
make sure, and for your convenience, I repeat the information'. When writ-
ers use restrictive clauses, as in (16.13), on the other hand, they seem to
have the attitude don't assume that you still know what I am referring to,
even though I have said it before, therefore I repeat it so that you need not
look for the information in the preceding text'. That a whole paragraph
intervenes between the second shell-noun phrase and the shell content is
certainly an important factor here. As will be shown in more detail in the
next section, uses of this type can be seen as combining anaphoric and cata-
phoric reference.
Pattern N-e-cl uses can also involve cross-sentential anaphoric links,
even though both parts of the shell-content complex occur in the same
clause. The means of creating such links are the cohesive adjectives already
mentioned in Section 15.2 above. The most important and frequent items
are next, (an)other, different, same and the ordinal numbers first, second
etc. On the basis of Halliday and Hasan's notion of "comparative reference"
(1976: 76), these items can be understood as suggesting that the anaphoric
noun phrases and their antecedents are similar in certain respects and there-
fore belong to one class of things. This can be observed in example (16.14):
(16.14) ... to deteriorating memory and mental performance. We get aluminium

from cooking utensils and food packaging, including aluminium foil.
People with indigestion risk an extra dose, from many types of antacids.
Another hypothesis is that nerve cells are simply not getting enough
blood supply, (BOOKS)
The shell-content complex in this example needs no explanation. What I am

concerned with here is how the comparative item another affects the pre-
ceding context. It has the effect that the preceding passage is declared to be
a hypothesis as well.
Such backward characterizations by means of comparative cohesive ad-
jectives do not come out of the blue, but are carefully prepared by speakers
(or more typically writers). Since such sequences can of course not be iden-
tified automatically, I again use an example that is not taken from COBUILD.
The source is The Economist, 17 January 1998, p. 34-35:
(16.15) ... the book is still dangerous to the government for two reasons. First, it
may encourage people to believe that an embittered Mr Brown is bidding
his time until he can exact revenge by ousting Mr Blair. In fact, while
Mr Brown still harbours ambitions to become prime minister [...], his
relationship with the prime minister is nevertheless a lot closer than
many other chancellors have enjoyed. "You cannot be with Gordon for
an hour without him phoning Tony or Tony phoning him," says a cabi-
net colleague.
The second danger is that the book brings into open all the bitchy
Westminster gossip about...
The first sentence sets up the text structure of the following two paragraphs
by introducing two reasons why a certain book is dangerous. The first rea-
son is then introduced by the numerative text signal first, and the second
reason by the shell-noun phrase the second danger. This noun phrase has
the welcome double effect that it reminds readers that the previous para-
graph was about a danger and informs them that the complementing clause
is about a second danger.
In addition to illustrating the specific anaphoric function of cohesive ad-
jectives in the pattern N-e-cl, the last three examples make up a smooth
transition to what will be at issue in the next section, namely the function of
signalling textual units and especially their boundaries.
Signposting 349
16.3 Signposting
The image of signposting will not be used for intra-sentential links and
short-distance anaphora, but only for cases of shell nouns which signal
textual units and the boundaries between them or, more generally, the
structures of texts.91 As suggested by Winter's work on the topic, signposts
are always accompanied by their realisations, that is by the pieces of in-
formation that they point to and require in order to be informative.
The examples used in this section differs from the others in this study in
three ways. First, I will not use material from the COBUILD corpus, because
manifestations of the signposting function are more easily spotted by the
traditional method of reading texts or simply browsing through them. Since
the best sources are well-edited texts, I have chosen print media, one issue
each of The Times, The Independent and The Economist, and a linguistics
textbook, Mey (1993), as sources. The newspapers, from which I have al-
ready quoted in Section 16.2, will be referred to by the abbreviations TIMES
(The Times, 15 January 1998) and INDY (The Independent, 13 February
1998). Second, the examples will naturally be much longer than those used
in the micro-analytical descriptive part. And third, the examples will include
uses of shell nouns other than in the four patterns which have been de-
scribed in detail.
In addition to signposts or signals and their realisations, two notions
will play a central role in this section. The first is what Brown and Yule
(1983: 73) refer to as "the very attractive pretheoretical notion" of dis-
course topic, which is defined as "what is being talked/written about"
(ibid.). Unlike the notion of (sentential) topic, with which I have worked in
the previous sections, discourse topic is thus understood in terms of 'about-
ness' here. It is not necessary for the purposes of this study to elevate the
notion from its lowly status of being 'pretheoretical', because I will only
need it to refer to what a piece of discourse is about.92
The second notion, that of paragraph, will be treated in a similarly
down-to-earth way, even though, or precisely because, it has so far also
defied all attempts to cast it into a rigid technical definition.93 Despite Lon-
gacre's (1979: 116) and Hinds' (1977: 83) warnings that especially in
newspapers, paragraph boundaries are more dictated by eye appeal and
appearance than by discourse-semantic considerations, I will use the term
paragraph to refer to orthographic paragraphs unless indicated otherwise.
The crucial relation between the two notions of discourse topic and
paragraph is that the beginnings of paragraphs are understood as either
representing topic boundaries as such, or as containing linguistic elements

that signal topic boundaries, so-called "topic boundary markers" (Brown
and Yule 1983: 94). Related terms are "topic change markers" (van Dijk
and Kintsch 1983: 204) and "frame shifters" (Polanyi and Scha 1983: 266).
What I am proposing is that not only conjunctions like but, however and so
(van Dijk 1977: 139-140) function as topic boundary markers and topic
changers or shifters, but also shell nouns. Passing remarks to the same ef-
fect can be found in Brown and Yule (1983: 100). More generally, shell
nouns can be credited with an organizational function in texts.94 The present
section will give a survey of the two main types of signposts that are real-
ized by means of shell-noun phrases: anaphoric signposts which serve as
topic change markers, and cataphoric signposts which are used by writers to
guide readers through their texts.
Anaphoric shell-noun phrases with demonstratives and definite articles
as determiners are very frequently found at the beginnings of paragraphs
(Francis 1994: 87). Examples of this type have already been provided in
(16.13), (16.14) and (16.15) in the last section. Another example is given in
(16.16) for the discussion here:
(16.16) SCORES of children with hopeful, gap-toothed smiles gaze out from the
pages of Be My Parent, a bi-monthly newspaper published by the British
Agencies for Adoption and Fostering.
It is one of the main market places for advertising children for whom
local authorities are seeking long-term carers. It is a highly competitive
business; the supply of children far exceeds the pool of peope prepared
to provide them with a home and a family life.
The problem was highlighted this week when a newspaper in Oxford
published... (INDY, p. 10)
Even this short extract testifies to the claim made above that orthographic
paragraphs are indeed not determined by text-semantic considerations in
newspapers. Clearly, the first and the second orthographic paragraph make
up one text-semantic unit. The third orthographic paragraph must be seen
as representing a new text-semantic paragraph, because the author zooms in
from the issue of children seeking homes in general terms to one particular
case, which was highlighted by an item in a newspaper. This topic shift is
both facilitated and marked as such by the shell-noun phrase the problem.
On the one hand, this noun phrase creates a concept-forming and evaluating
link to the two previous orthographic paragraphs, especially to the passage
marked as shell content by underlining in (16.16). On the other hand, the
Signposting 351
shell-noun phrase also makes up the starting-point of the information given

in the next paragraph. It is no coincidence that it functions as sentential
topic in the first sentence. Apart from linking, concept-formation and char-
acterization, the three basic functions of shell nouns, this shell-noun phrase
does two things, then. It preserves (discourse) topic continuity, by relating
the new paragraph to the previous ones, and it marks the topic change, by
serving as a starting-point for new information (cf. Francis 1986: 66).
In contrast to explicit forms of text deixis (see Kurzon 1985: 192-198),
shell-noun phrases of this type are not only extremely frequent in textbooks
and other types of expository writing, but also in newspaper reports. Al-
though a fairly large variety of shell nouns occur in this function, the noun
move is undoubtedly the favourite of the press. Apart from this and other
eventive shell nouns like measure, process and change, linguistic shell
nouns like report, announcement, statement, advice, warning, hint and
plea and mental nouns such as agreement, arrangement, initiative, plan
and decision prevail. Slightly more rarely one comes across factual nouns
(problem, snag, reason, difference), modal nouns (danger, opportunity)
and circumstantial nouns (approach, way). Fairly frequent, on the other
hand, are specific deverbal eventive nouns which were not regarded as shell
nouns proper in this study because they cannot be used in the patterns N-cl
or N-Ae-cl. Examples of this type are arrival, departure, rise, split or
surge. Although the functions of these nouns are very similar to those of
shell nouns, they are not included in this study because they are semanti-
cally too specific to be able to meet the syntactic requirements.
In principle, shell-noun phrases as boundary markers and topic shifters
can be found at "nodal points", as Conte calls them (1996: 5), in all sections
of newspaper articles. There are two places, however, where newspaper
writers deploy them particularly often in this function: at the beginning of
the second paragraph of an article, and at the beginning of the last para-
graph. An example of the first type is given in (16.17):
(16.17) Our ceasefire stands, say IRA

By Kim Sengupta in Belfast and Colin Brown in London
Northern Ireland ministers were last night fighting to ease the crisis over
the Ulster Peace talks with a clear hint that the Loyalist Ulster Demo-
cratic Party could be allowed back into the negotiations at the end of this
month.
The move was seen at Westminster as a signal that Sinn Fein could be
suspended for a few weeks ... (INDY, p. 2)
The development and structure of the text is highly stereotypical in such

examples. The articles start with a fairly neutral and matter-of-fact descrip-
tion of one piece of hard news. Readers are simply informed here of what
has happened. In terms of van Dijk's "thematic structure" (1988: 30-48) of
news articles, the first paragraph in (16.17) and in many similar cases rep-
resents the Main Event (1988: 53). This information is then taken up by an
eventive or factual shell-noun phrase, or, in the frequent case of reported
statements, by a linguistic shell-noun phrase, which serve as sentential top-
ics and starting-points for background information or generalizations. As in
(16.17), the foci of such clauses typically represent what van Djik calls
Verbal reactions or Comments (1988: 54-57) by other parties that are in-
volved in or affected by events, or first evaluations by the writers them-
selves. Interestingly, example (16.16) above manifested the complementary
pattern, which is also possible but less frequent. The author of this example
started out from a general observation concerning children in need of places
in homes and then turned to one particular case. Again, however, this shift
was marked by a shell-noun phrase at the beginning of the third ortho-
graphic (and second text-semantic) paragraph (the problem).
The second type of topic boundary is in a way a mirror image of the first
at the ends of newspaper articles. It is illustrated in (16.18):
(16.18) Infant classes to reduce in size

By Judith Judd
Education Editor
MORE THAN 100.000 infants will be taught in smaller classes from
September as the first step towards fulfilling the Government's pledge to
cut class sizes.
[10 orthographic paragraphs intervene]
Both the authorities and the Campaign for State Education, a parents'
pressure group, sav that ministers should work towards a norm of 25.
rather than 30 in a class to give schools enough elbow room.
The initiative may prove controversial with parents who want their
children to attend popular schools. Legislation will restrict parents' right
of appeal if it threatens to breach class size limits. Mr Blunkett said that
authorities would bring forward proposals to resolve the problem, [end
of article; INDY, p. 6)
Signposting 353
(16.18) exemplifies how shell-noun phrases are used in newspapers to take

up the Main Event, which is usually also the discourse topic of an article
(here an initiative of the Government aiming to reduce class sizes), in order
to add final Comments, outlooks on possible Consequences (van Dijk 1988:
54), evaluations or other types of background information. This need not
always to be done at the beginning of the last paragraph but can also occur
at the beginning of the last but one or last but two.
What (16.16), (16.17) and (16.18) suggest, then, is that paragraph-
initial shell-noun phrases facilitate the signalling of what van Dijk calls
"news schemata" (1988: 49-59). This impression can easily be confirmed
even by a cursory glance into an issue of any of the more serious British
papers. In tabloids, the tendency is on the whole less marked, mainly for
stylistic reasons.
The signposting strategy is also very common in expository writing, es-
pecially in textbooks, whose authors tend to make an effort to structure their
texts as clearly as possible. As an example, a fairly long passage taken from
Mey (1993: 55-56) is rendered here as (16.19). Mey is trying to provide a
soft introduction to Grice's Cooperative Principle here.
(16.19) Let us consider some examples. Suppose I utter a sentence such as
Many of the delegates opposed the motion.

On a normal reading, such a sentence would convey the impression
that although many delegates voted for the motion, a number of them
were against, and voted accordingly. In fact, this is the interpretation
that I, as the sender, want the receiver to accept. Normally the sentence
would thus not be taken to mean that all of the delegates voted against
the motion, even though, strictly speaking ('many' does not say how
many), such a reading could be consistent with the normal reading - es-
pecially if I complete my utterance by adding something like
In fact, all of them did.
The question is why anybody would say 'many' rather than 'all', if in
actual fact there were no others? If I could have used the stronger ex-
pression ('all'), why didn't I?
What we're confronted with here is a pragmatic principle: speakers
try to be understood correctly and avoid giving false impressions. If what
I say is logically correct, and true according to some abstract semantic
'rule', but still confuses or misleads my hearer, then my utterance will
not have its proper effect: I will be misunderstood.
There seems to be a general understanding that people, when thev

rive out information, prefer to do so in the spirit of parsimoniousness. In
Gazdar's terminology (1979: 56-8), we can talk about a scale of expres-
sions, from stronger to weaker; an example is the following scale
(adapted from Levinson 1983: 132):
all, most, many, some, few, none,

where the strongest 'scalar' expression occurs to the left, with strength
decreasing as one moves right (though the ordering of 'some' before
'few' can be questioned from this point of view).
The rule seems to be that we. by using a weaker expression, exclude
the stronger ones: that is. the use of 'many' implies that 'all' cannot be
used, at least in a 'canonical' (that is. non-expanded) context.
The principle that is invoked here is a particular case of the general
principle of cooperation. [...]
Not all delegates opposed the motion.
This sentence is both more 'rigorous' and easier to verify than the
original one; besides, as we have seen, it is implied in it (by conversa-
tional implicative, as the term goes); so why don't I just sav this and
avoid all misunderstandings ?[ 1
The answer 2 to that question lies in the way language users go
about their communicative business. Communication is not a matter of
logic or truth, but of cooperation;... ?
The first point worthy of attention is the frequency of shell-content com-

plexes in this passage, irrespective of the functions that individual instances
fulfill. It must be added, however, that the number of shell-noun phrases is
by no means uncommonly large here, at least not if compared to the rest of
the book in question. The main observation in the present context, however,
is that many of these shell-noun phrases occur at the beginnings of para-
graphs. As in the newspaper articles, these shell-noun phrases function both
as connectives and as new starting-points. One difference to the newspaper
examples is that in most of the paragraph-initial instances in (16.19), the
signposting function is less explicitly marked because the nouns are accom-
panied by clausal shell contents as well. When the shell-noun phrases are
highlighted (as in my reproduction of Mey's passage), their function of
signposting readers through the content of the text is particularly conspicu-
ous.
Signposting 355
The fourth shell-noun phrase in bold-face types in this example, a prag-

matic principle, takes me to the second major category of signposts, exclu-
sively cataphoric ones. As signalled by the colon after this noun phrase, its
function is to tell the reader that specific information is to follow. This in-
formation is expressed as shell content in the next clause. In addition to
existential i/zere-clauses (see 16.19), combinations of demonstratives and
the copula (this is the problem or the problem is this) are often used to
introduce such cataphoric references. An interesting illustration from The
Times, in which the latter strategy is combined with a that-clause in a fairly
extraordinary manner, can be found in (16.20):
(16.20) It is, after all, 310 years since William of Orange came to the throne,
and while we remain a Protestant nation, the anti-Catholic virus should
by now have worked itself out of the system.
The charge is this: that Britain's most ancient order of chivarly. whose
members are appointed by the reigning monarch, and which recognises
distinguished service to the nation, is a Protestant bastion, barred to
Catholics, (TIMES, p. 18)
Since this passage has been printed in a fairly renowned newspaper, one
should not question the quality of the writing - especially in view of the fact
that this passage appeared on the same page as not only the lead article but
also a contribution by Tony Blair. One can assume, then, that the author
had a specific reason for using both a cataphoric reference and the struc-
tural signal that to link shell noun and shell content with each other. My
feeling is that this double bond gives even more prominence to the shell
content than the simple focusing N-e-cl construction (see Section 16.2),
because the colon after the pronoun this is read as an extra pause. I am
aware that this does not explain why the writer chose to use the that-link as
well. Conceivable motives are that he felt that the tie would be closer or that
the conceptual boundaries of the information would be stronger if it was
expressed in a that-c\ause rather than a simple clause (see Section 17.1 on
conceptual partitioning).
Strictly speaking, simple cataphoric references such as those discussed
so far are more interesting from the point of view of linking than from the
signposting perspective. Cataphoric signposting comes into its own, how-
ever, when the structures of longer passages are anticipated by cataphoric
plural shell-noun uses. In the following example of this type, which is taken
from the opinion column of The Independent, I have included references to
the text structure in braces at the ends of the units in question. Capital let-
ters in braces refer to signposts, numbers in braces indicate realisations of
signposts.
(16.21) Rupert Murdoch features in three challenges {A} to Blair

DONALD MACINTYRE
THE GENTLE ART OF COMPROMISE
Rupert Murdoch - even since the spectacular defection of the Sun to its
cause - still haunts the collective psyche of the Labour Party. Three un-
resolved issues {B}, currently exercising the government's best minds,
all have features {C} in common, of which Mr Murdoch's looming
presence, real or perceived, is only one {1}. The newspaper industry has
a direct interest in the outcome the outcome [sic!] of each {2}. The pub-
lic is less interested in them than the newspapers would like them to be
{3}. And finally, Tony Blair has directly focussed his lawyer's mind on
all of them {4}.
The first concerns the trade unions and how to put into practice the
manifesto... {5}
[35 intervening lines].
It will probably take weeks rather than days for the discussion to reach a
conclusion.
The second is the issue of privacy and the press. {6}
[27 intervening lines].
But it will still look to the Government's critics like tampering with the
Bill in order to satisfy the owners of tabloid newspapers.
The last, of course, is predatory pricing. {7}
[39 intervening lines]
... he will probably not finally decide whether to do so until he has fully
examined the Government's case.
There are possible compromises {D} of varying effectiveness in all
three cases {E}. On privacy, if the Government does produce the ex-
emption that the PCC [...] is seeking, it will surely (at long last) have to
come up with a system of compensation for individuals who have been
wronged by newspapers invading their privacy. {8} On predatory pric-
ing, [5 intervening lines]. {9} On unions [10 intervening lines]. Like it
or not, of all three issues {F}, it is union recognition that in the long
run will matter most in the Parliamentary Labour Party. {10}
Signposting 357
As the references show, this comment is not just rigidly structured, but the
reader is also carefully guided through the text by signposts. A synopsis of
the text structure is given in Figure 16.2. The synopsis reveals how shell-
noun phrases are employed as cataphoric and anaphoric signposts in this
text. What is particularly interesting about the three cataphoric signposts,
{A}, {B} and {D}, is that their cataphoric function is not marked on the
linguistic surface. In all three cases, it is simply stated that there are three
challenges, three unresolved issues and possible compromises, but it is left
to the readers to actively interpret the shell-noun phrases as referring for-
ward in the text. In the framework of de Beaugrande and Dressier (1981),
then, these noun phrases contribute to the coherence95 of the text, but they
do not create cohesion (because of the lack of surface marking). Neverthe-
less, expressions of this type, and especially existentiaWere clauses of the
type there are three aspects/issues/etc., are found very frequently as cata-
phoric signposts.
{A} Signpost (cataphoric): three challenges

{B} Signpost (cataphoric): three unresolved issues (same targets as {A})
{} Signpost (cataphoric): features in common
{1} First realisation to {C}: Mr Murdoch's looming presence .
{2} Second realisation to {C}: The newspaper industry...
{3} Third realisation to {C}: The public...
{4} Fourth realisation to {C}: And finally, ...
{5} First realisation to {A} and {B}: The first...
{6} Second realisation to {A} and {B}: The second...
{7} Third realisation to {A} and {B}: The last...
Signpost (cataphoric): possible compromises
{E} Signpost (anaphoric): all three cases (points back to {5}, {6} and
{7}, as well as {A} and {B})
{8} First realisation to {D}: On privacy,...
{9} Second realisation to {D}: On predatory pricing,...
{10} Third realisation to {D}: On unions, ...
{F} Signpost (anaphoric): all three issues (points back to {5}, {6} and {7},
as well as {A}, {B} and {E})
Figure 16.2 Analysis of signposts and realisations in example (16.21)

The realisations to which the signposts point are not signalled by shell-noun
phrases here but by other connective devices. Of the first set of realisations
constituted by {1}, {2}, {3} and {4}, only {4} is marked explicitly by the
conjunct finally, but this suffices to show readers that this is the last item in
the list of features in common (if they have not realized it anyway). The
second set of realisations, {5}, {6} and {7}, are marked by elliptical noun
phrases with ordinal numerals in comparative cohesive function. And in the
third set, {8}, {9} and {10}, the text structure is signalled by clause-initial
adverbials expressing 'respect' (Quirk et al. 1985: 706-707).
Shell-noun phrases are not only used as signposts, though, but also as
means of introducing and signalling their realisations. One example of such
an uptake was included in (16.15) above, where the writer first deployed a
cataphoric signpost (for two reasons), marked the first realisation with the
simple enumerative form first, and the second with the shell-noun phrase the
second danger. Here is a second example, taken from the textbook by Mey
(1993):
(16.22) As we will see later, there are basically two ways of going about study-
ing conversation and other basic linguistic interaction: in one, we can
just study what's going on, trying to describe it as exactly as possible,
and figuring out what the options are for the participants to join in at
any given point, and what their choices are of expressing themselves to
their own and others' satisfaction. This line of approach is followed by
the so-called conversation analysts (who will be discussed in greater de-
tail in chapters 10-12).
Another, more theoretical approach tries to go 'behind conversa-
tion', as it were, [8 intervening lines].
This latter approach, needless to say, comes closer to what I have
defined as pragmatics.
(Mey 1993: 48-49).
As in (16.15), there is a cataphoric shell-noun phrase in this example, sign-

posting two ways of going about.... And again as in (16.15), the first reali-
sation is marked by a numeral, here a cardinal numeral in an elliptical noun
phrase. Especially when the first realisation follows directly after the sign-
post, shell-noun phrases are not often used to signal the realisation. The
second realisation, however, is marked by the shell-noun phrase another,
more theoretical approach, and again, this seems to be a fairly common
strategy which serves the purpose of ensuring that the realisation is identi-
fied correctly. In this particular example, both realisations are taken up
Signposting 359
again by anaphoric shell-noun phrases (this line of approach and this latter
approach). To my mind, the frequent repetition of the shell noun approach
is a clear sign of the author's willingness to sacrifice stylistic elegance for
clarity and conspicuousness - which is certainly a commendable move given
the intended readership of a textbook.
In sum, Chapter 16 has made it quite clear that considerations which go
way beyond single clauses and sentences may affect the use of shell nouns.
In Section 16.1, I have shown that the wish to deploy an emphatic or con-
trastive focus of attention is a frequent motive for the use of a shell noun in
the pattern N-e-cl. Some variants of this type are particularly interesting
because they combine a primary focus on the shell content with a secondary
prominence peak on the shell noun itself or its premodifier. With certain
types of shell nouns, mainly creditive mental ones like feeling, impression
or intuition, this secondary focus on the topicalized shell-noun phrase can
be exploited for rhetorical purposes like expressing tentativeness or subjec-
tivity. In Section 16.2,1 have situated anaphoric uses of shell nouns within
the range of anaphoric expressions. Depending on the specificity of the
nouns and other factors such as the distance between shell noun and shell
content, shell nouns tend to be used to refer to activated or familiar infor-
mation. A second question addressed in Section 16.2 concerned the relation
between anaphoric shell-noun phrases and their antecedents. It was argued
that the semantic/cognitive match between the gaps inherent in the semantic
structures of shell nouns and the meaning of the shell contents control the
identification of antecedents. Section 16.3 has been devoted to the sign-
posting function of shell-noun phrases. Two types were discussed. It was
shown how paragraph-initial shell-noun phrases function as (discourse)
topic boundary markers, which simultaneously refer back to the previous
paragraph and serve as starting-pont for the new. Finally, shell-noun
phrases which function as cataphoric signposts signalling the structure of
the text following them were discussed.
Chapter 17
Cognitive functions
I have already said at the outset of Part III that I regard the cognitive func-
tions of shell nouns as their ultimate raison d'tre. To appreciate the role of
cognitive functions, it is helpful if one abstracts from individual uses of
shell nouns and from the reasons why speakers use them in given contexts,
and begins to ask why shell-content complexes have become part and parcel
of the linguistic system (of English and at least the few European languages
I am somewhat familiar with). It is the aim of this chapter to bring these
functions into the limelight.
This chapter is divided into three sections, one on conceptual partitioning
(17.1), one on reifying and hypostatizing (17.2), and one on integrating
(17.3). It must be emphasized, however, that the three (or four) aspects are
tightly interwoven and have only been separated from each other for ex-
pository reasons.
17.1 Conceptual partitioning
The term conceptual partitioning originates from the work of cognitive

linguists. It is used by Dinsmore (1991), Talmy (1991) and Fauconnier
(1997), but with different things in mind. In Fauconnier's theory of mental
spaces, a brief reference to which was already given in Section 15.1, the
function of partitioning is attributed to space builders. Fauconnier argues
that space builders like Susan believes or Max hopes partition off informa-
tion, "by relativizing it to different domains" (1997: 38). The idea is that by
opening up mental spaces, space builders separate the information relevant
in these spaces from other pieces of information. Dinsmore (1991) con-
ceives of his so-called "partitioned representations" in a very similar way.
While I have argued that shell nouns in the pattern N-cl can also be re-
garded as space builders, this is still not the understanding of conceptual
partitioning that I am concerned with here.
My notion of conceptual partitioning is borrowed from Talmy (1991).
Talmy (1991: 483) argues that clauses contribute to the conceptual parti-
tioning of events to the effect that what would otherwise be a continuum, e.g
Conceptual partitioning 361
in space, time or other dimensions, is experienced as one bounded concep-

tual entity. According to him, there are different types of such entities,
though the only example he names is the category of events.
I claim that shell-content complexes partition conceptual entities of two
different kinds, which differ primarily with regard to their size and their
saturatedness. Both types divide our experience and consciousness, which
must be imagined as a continuous and amorphous flow of information
rather than a set of bounded and structured propositions, into chunks of
different granularity. The clausal shell contents in the patterns N-cl and N-
be-cl (realized by that-clauses, wA-clauses or infinitives) partition cognitive
entities representing events or abstract relations, i.e. entities of a proposi-
tion-like nature (see Section 5.1.1); they partition fully saturated cognitive
units which have a predicative power (and truth-values).
The shell nouns themselves, on the other hand, since they are nominal
elements and function as heads of noun phrases, partition concepts, i.e.
unsaturated cognitive entities. (A concept alone cannot be used to predicate
something of something else.) Obviously, this is the essence of the concept-
forming function of shell nouns which was mentioned as early as in Section
2.2. Experience is conceptually encapsulated, as the effect is very aptly
referred to by Francis (1986: 36-38, 1994: 85) and Conte (1996: 2-4).
If the description of concepts as 'unsaturated' cognitive entities, which
ultimately goes back to Frege's (1994; 1891) ideas about mathematical
functions, evokes the image of a deficient entity, this is highly misleading.
As I will show, concept-formation is perhaps the single most important
effect of the use of shell nouns altogether.
Taking a random example from the descriptive part to flesh out these
claims, I reproduce (7.5a) as (17.1) here.
(17.1) The siskin is an exceptionally pretty, small green bird, a sort of pocket
version of the greenfinch. Twenty years ago it was uncommon in Eng-
land. The fact that it is now very much a regular winter visitor is due
almost entirely to the efforts of amateur bird feeders, (EPHEM)
The that-clause in (17.1) represents a saturated cognitive entity. This kind

of one-to-one mapping between clauses and complex, proposition-like cog-
nitive entities representing events and abstract relations comes so naturally
that it is often taken for granted, even by linguists. Among those who draw
attention to the relation are Danes (1976: 30), van Dijk and Kintsch (1983:
362 Cognitive functions
37-38), Foley and van Valin (1985), Pawley (1987: 333), Givn (1990:
515, fii.1) and Talmy (1991: 482-483).
Being a noun, the shell noun fact itself does not partition a complex and
saturated abstract relation but a simple concept. Although the possible shift
in conceptual status from assertion to fact may play a role (see Section
15.1), one may claim that conceptual partitioning as concept is in fact the
main motive why the speaker uses the noun fact. The speaker uses the noun
to cast the shell content, to which it is linked by means of the appositive
relation, into one single concept. This is useful for him, as he wants to insert
the information into a causal relation, in which the fact that it is now very
much a regular winter visitor acts as EFFECT (within the frame description
used in Section 7.3) and the efforts of amateur birdfeeders as CAUSE.
From a grammatical point of view, it would of course have been possible
to omit the concept-partitioner the fact. This is shown by the acceptability
of (17.1):
(17.1') The siskin is an exceptionally pretty, small green bird, a sort of pocket
land. That it is now very much a regular winter visitor is due almost en-
tirely to the efforts of amateur bird feeders.
In longer sentences, however, which contain more information and express

more complex relations, an omission of the fact or similar shell nouns can
result in an overload of the capacity of the short-term buffer. According to
van Dijk and Kintsch (1983: 349), the capacity of this buffer is limited to
three atomic propositions. It must be emphasized, however, that the atomic
propositions in their framework do not necessarily correspond to clauses but
are defined as consisting of a predicate and its arguments. In this definition,
noun phrases like a regular winter visitor are also propositions. If a very
complex proposition consisting of many more than three atomic proposi-
tions is inserted into an abstract relation, it is possible and even likely that
the hearer is no longer able to keep track of the incoming information, un-
less the information is partitioned into smaller conceptual chunks. Example
(17.2) is a case in point, though by no means an extreme one:
(17.2) The fact that the head of one of Nigeria's northern-based regiments,
based in Kaduna. has already pledged support for President Bbangida
reinforces the view that the coup is ethnically-based, (BBC)
Conceptual partitioning 363
Since the that-clause here contains quite a lot of information, hearers will
only be able to process it easily if it is put into the conceptual shell provided
by the noun. It is to this more simple, integrated conceptual entity (see Sec-
tion 17.3) that both speakers and hearers can then cling.
If large information units are to be partitioned as simple concepts, there
must be an indication that the large piece of information is to be identified
with the small one, and that the small one can therefore stand for the large
one. The key to this relation is the notion of experiential identity. As out-
lined in Section 3.1, speakers have various linguistic constructions at their
disposal to signal this link, most notably the copula construction, apposi-
tional constructions and anaphoric or cataphoric references. If it were not
for the link of experiential identity, it would be impossible to explain how
proposition-like entities can be cast into simple concepts. The effect of con-
ceptual partitioning is particularly dramatic in the case of the patterns th-
and th-be-N, where even previously unbounded pieces of information ex-
pressed in whole paragraphs or even longer passages come to be partitioned
as concepts that are expressed by simple noun phrases.
The partitioning of complex pieces of information as concepts is of
course only possible if the specificity of information that can be expressed
by clauses and larger pieces of text is drastically reduced. This is achieved
by integrating the information in the shell noun. This important cognitive
effect of the use of shell nouns will be described in more detail in Section
17.3. Before that, however, I will turn to a pair of effects which are closely
related to the formation of concepts, the reiflcation and hypostatization of
cognitive content caused by the use of nouns.
17.2 Reifying and hypostatizing
Not only nouns, but also verbs, adjectives, adverbs and even the members
of closed word classes can be said to create concepts of some kind. What is
the distinction of nouns, then, in this respect? Why is it that shell nouns,
rather than shell verbs or shell adjectives, seem to be so extraordinarily
useful? Since the question of what distinguishes the word class of nouns has
also been addressed by cognitive linguists, I will start out by trying to an-
swer it from this perspective. It is to a brief survey of Langacker's (1987a,
1987b, 1991) view of word classes, therefore, that I will turn now.
For Langacker, "a noun designates a thing" (1987a: 189; emphasis
original). He goes on to propose that
a thing is properly characterized as a region in some domain, i.e. every

nominal predication designates a region. Count nouns represent a special
but prototypical case, in which the designated region is specifically con-
strued as being bounded in a primary domain.
(Langacker 1987a: 189; emphasis original)
Of course, this definition is in need of further clarification. Basically, Lan-

gacker accounts for the distinction between the major classes of nouns,
adjectives, verbs and adverbs in terms of the basic cognitive ability of per-
ception. More precisely, two perceptual sub-abilities play a crucial role.
The first is the ability to single out, or profile, parts of the perceptual input
as a prominent figure against the ground (1987a: 183). It is this process of
profiling that is referred to by the verb designate in the quotation above
(and not the everyday sense of the term). Accordingly, a word is a noun if it
profiles a region, and a count noun if it profiles a bounded region. A region
is defined as "a set of interconnected entities" (Langacker 1987a: 198), and
a region is bounded "when there is some limit to the set of interconnected
entities it comprises". What all this comes down to, then, is that one can
think of count nouns as words whose denotata are experienced as stable and
coherent conceptual gestalts. Gestalts are understood in the present study as
holistically perceived and conceptualized, and therefore integrated and uni-
tary, cognitive structures.96
At this point, the second ability comes into play, the ability to scan the
environment, i.e. to compare perceptual inputs acquired at different, nor-
mally subsequent, points in time (Langacker 1987a: 102-103). If we watch
an event, for example the arrival of a train, we compare our perceptions
with regard to the relation between the involved objects, here the train and
the envisaged point of arrival, and note how they change from one point in
time to another. This type of perception, which can be described by a clause
like the train is arriving, is called "sequential scanning" by Langacker
(1987a: 248-249). Sequential scanning is represented diagrammatically in
Figure 17.1.
The same scene can also be described by the nominalization the train's
arrival. Such nominal expressions are interpreted by Langacker as reflect-
ing a different type of scanning called "summary scanning". The perspec-
tive on the identical situation that is evoked by the nominal expression is
represented in Figure 17.2. Here, the different temporal stages are summa-
rized in one picture. The bold line around the stages of the event indicates
that they are profiled "collectively as a thing" (Langacker 1987a: 248) by
the nominal expression. The event is thus construed as one single gestalt,
Reifing and hypostatizing 365
whose essence differs from its constituent stages. In contrast, the verbal
expression represented in Figure 17.1 profiles the changing relation.
Figure 17.2 Summary scanning: arrival (based on Langacker 1987a: 144)
In sum, then, the difference between nouns and verbs in Cognitive Grammar
concerns two dimensions: nouns profile individuals, events and abstract
relations as 'things' and suggest a summary scanning of scenes, while verbs
profile relations and suggest sequential scanning. This explains to some
extent why nouns are experienced as denoting single unitary entities which
have a stable existence ('gestalts'), and verbs as denoting processes and
relations.
Returning to the area of shell nouns, there are two suitable testing-
grounds for this hypothesis. The first is supplied by deverbal linguistic and
mental shell nouns like order, promise, belief, assumption and desire, since
examples of these nouns can be juxtaposed with the corresponding verbal
expressions. Examples (17.3) and (17.4) are repeated from Chapter 8,
where they appeared as examples (8.51) and (8.53). I have chosen these two
examples because they can be converted into verbal expressions particularly
well, as is shown in (17.3') and (17.4'):
(17.3) Of the two chief whaling nations, Norway is adamant in its threat to
hunt the migratory minke whale ... in the northern Atlantic, (MAGS)
(17.3') Of the two chief whaling nations, Norway adamantly threatens to hunt
the migratory minke whale ... in the northern Atlantic, (MAGS)
(17.4) Visa's complaint that the report failed to deal adequately with certain
questions was no basis for invalidating its findings, (PAPERS)
(17.4') Visa complained that the report failed to deal adequately with certain
questions, but this was no basis for invalidating its findings.
When the speech activity of threatening is conceptualized by the nominal

expression its threat, it seems to have an autonomous existence, to be stable
for a period in time - a property which is clearly more characteristic of
things than of activities. It seems to be there, as it were, rather than take
place. The same is true of the speech act described by the noun complaint in
(17.4). In contrast to the verbal expression Visa complained that, the nomi-
nalization Visa's complaint that in (17.4) acquires a thing-like gestalt,
which appears before our mental eye as one coherent and stable region
rather than as a sequential and changing process. In short, both nominal
expressions evoke the ideas of 'things' rather than processes or activities,
even though strictly speaking they denote events. This cognitive effect of the
use of nouns, which results in a conceptualization of any type of cognitive
representation as 'thing', can be called reification.
Not just deverbal nominalizations, which are not among the prime ex-
amples of shell nouns, but also non-derived nouns like fact, idea or problem
exhibit this effect. 97 To test this claim, examples like (17.1), in which the
shell noun fact seems to be redundant from a grammatical and a pragmatic
point of view, though not from a cognitive one, are particularly illuminating.
I repeat the example and its paraphrase here for convenience.
Reifying and hypostatizing 367
(17.1) The siskin is an exceptionally pretty, small green bird, a sort of pocket
land. The fact that it is now very much a regular winter visitor is due
almost entirely to the efforts of amateur bird feeders, (EPHEM)
(17.1') The siskin is an exceptionally pretty, small green bird, a sort of pocket
land. That it is now very much a regular winter visitor is due almost en-
tirely to the efforts of amateur bird feeders.
Can Cognitive Grammar also explain why the speaker in example (17.1)
uses the noun phrase the fact despite its redundancy? I think it cannot.
Here is an outline of what Langacker has to say about such examples:
According to him (1991: 432), both the combination of the fact + that-
clause in (17.1) and the -clause alone in ( 17.1') are analyzed as elabo-
rations of the main-clause primary figure, i.e. in plain terms, as manifesta-
tions of the grammatical constituent subject. They only differ in that the
former is a 'normal' nominal and the latter a nominalization of a relation.
Langacker explains such cases as (17.1) as "a correspondence between the
profiles of two full nominis" (1991: 432), here the fact and that it is now a
regular visitor. Not surprisingly, he calls the correspondence "a kind of
apposition" and argues that "two nominis that designate the same con-
ceived entity but describe it in different ways combine to form a higher-
order, doubly-grounded nominal with the same profile" (ibid.). In the same
context, he calls the clause "a reified proposition" and notes that it repre-
sents a process with nominal construal which can also function alone as
participant of the main clause (this is the case in 17.1'). Langacker treats the
notion of reification mainly as a process involved in nominalizations (1991:
34-35). In this respect, he does not seem to have broken away from the gen-
erative tradition in which he was brought up. The reason why a speaker
might prefer such a doubly-grounded expression to a simple one is not ad-
dressed by Langacker.
My view is that a more literal understanding of the notion of reification,
in the sense of 'making a thing out of something', is more helpful to account
for the role of shell nouns than the technical generative conception of the
term. I believe that speakers use two nominis, a noun phrase and a that-
clause, side by side to activate one single idea, because nouns have a higher
potential for reification than nominal -clauses. Much more powerfully
than clauses, nouns create the illusion that what they stand for is similar to
a 'thing' with respect to stability in time and conceptual unity. Although, as
has been shown, this is the basic insight behind Langacker's idea of nouns
as designating 'things', he makes the mistake of extending the category of
nouns too far (1987a: 242).
Long before the advent of cognitive linguistics and Cognitive Grammar,
philosophers and linguists have used the term hypostatization to attribute
the combined effect of concept-formation and reification to the use of
words.98 The Swiss linguist Ernst Leisi explains the notion as follows:
Die Mythologie, der scholastische Realismus und die platonische Ideen-

lehre sind die grten Beispiele fr die Neigung der Sprachgemeinschaf-
ten, jede Erscheinung irgendwelcher Art, sofern sie durch ein Wort be-
zeichnet werden kann, zu vergegenstndlichen (allenfalls zu personifizie-
ren) und mit einer selbstndigen, von anderen Erscheinungen abgelsten
Existenz zu begaben, sie also zur akzidenzlosen Substanz zu erheben. Die-
se Erhebung zur Substanz nennen wir, dem Sprachgebrauch der Philoso-
phie folgend, H y p o s t a s i e r u n g. (Leisi 1975: 26)
(Mythology, scholastic realism and the Platonic theory of ideas are the
most notable examples of the tendency of language communities to reify
(possibly personify) every phenomenon of any kind, provided that it can be
denoted by a word, and to endow it with a self-contained existence, re-
moved from other phenomena, thus elevating it to a non-arbitrary sub-
stance. Following common usage in philosophy, we call this elevation to
substance hypostatization.)
Although Leisi shows that the members of all open word classes give rise to
the hypostatization of experience, it is nouns that lend themselves most
readily to the process. According to Leisi, nouns hypostatize intended
meanings as things, adjectives as properties, and verbs as actions. The par-
allels to Langackers definition of word-classes are obvious. In contrast to
Langacker, however, whose main criterion of reification is syntactic func-
tion, Leisi claims that the potential of nouns to hypostatize pieces of cogni-
tive content as 'things' is a property inherent in nouns on the level of the
linguistic system and thus independent of their function in the clause. Leisi
does not raise the question of where this quality comes from, but it can be
speculated that its ultimate cognitive source lies in the fact that prototypical
nouns such as book, rattle or rope do in fact represent (categories of) con-
crete things.
What all this comes down to, then, is that speakers use shell nouns be-
cause the cognitive entities represented by nouns are more like concrete
'things' than those represented by clauses. They are better to grasp, carry
Reifying and hypostatizing 369
along and manipulate, and easier to understand, remember and insert into
abstract relations. As a result, they are better conceptual reference points
than clauses (van Hoek 1997: 52-60), more accessible than clauses (Ariel
1996: 22), and much more easily singled out for individual conscious
awareness than clauses (Langacker 1991: 97). From these properties follow
other linguistic and cognitive advantages of the use of nouns for complex
cognitive entities: nouns can function as anchors for evaluations and other
types of characterizations (see Sections 15.1 and 15.2), as topics or start-
ing-points in clauses (16.1), as carriers of anaphora (16.2) and as signposts
(16.3). This is why I have claimed that the cognitive functions are the ulti-
mate reason for the existence of shell nouns. It is important to emphasize,
however, that in contrast to 'normal' full-content open-class nouns, the
concepts partitioned by shell nouns, as well as the concomitant reification
effects, are highly ephemeral. Both effects are on-line phenomena restricted
to the actual use of a given shell noun in a given context.
This list of cognitive benefits of shell nouns leaves me with an obvious
paradox to account for. If, as claimed here, shell nouns indeed reify com-
plex pieces of information, how is it possible that some of the most fre-
quently used shell nouns, for example fact, idea or news, can be used to do
precisely the opposite, namely to transform cognitive units representing
events in the concrete world into abstract relations (see Section 15.1)? The
answer is that two different kinds of concreteness are involved here. The
concreteness of events is experienced as real, ontological concreteness;
events take place in what we believe to be the real world. The concreteness
conjured up by shell nouns like fact and idea, on the other hand, is only
apparent, even though it is cognitively all the more effective; it is conceptual
rather than ontological in nature. The use of shell nouns can thus be seen as
a manifestation of the concretizing function of metaphor (Schmid 1999).
Not only clearly metaphorical expressions like she dropped the idea or they
canvassed an idea exploit this function, but also, on a much more general
and pervasive level, nouns that are used for descriptions of events and states
of affairs. As Halliday puts it, "the nominal expression in the grammar
construes iconically an objectified entity in the real world" (1994b: 142). It
is in this sense that his notion of grammatical metaphor (cf. Section 5.1.2)
does have considerable justification after all.
17.3 Integrating
The potential of shell nouns for conceptual partitioning and reification cru-
cially depends on the link of experiential identity. This link ensures that
simple concepts are co-interpreted and co-activated together with cogni-
tively complex events and abstract relations. The result of this connection -
and this is the third crucial cognitive effect of shell nouns - is a cognitively
less complex conceptual unit, in which the information represented by the
shell noun and the shell content are integrated in such a way that the noun
can stand for the whole cognitive complex. This semantic integration
(Bransford and Franks 1972: 213-220, Conte 1996: 5) in turn has the
highly beneficial consequence that the amount of attention that has to be
devoted to the larger piece of information is diminished. This results in a
relief of the short-term memory (Chafe 1994: 119) or the short-term work-
ing buffer (van Dijk and Kintsch 1983: 349), which means that the dis-
course participants are able to take in new information. Just like a personal
pronoun like he or she is "a cognitively easier form" (Bolinger 1977: 4)
than a full noun phrase, a shell-noun phrase is a cognitively more economi-
cal linguistic unit than a clause.
Let me retrace these steps with the help of an example. Although in-
stances of all patterns would basically be suitable as illustrations, the ef-
fects of integration can be demonstrated particularly well with a plural use
of a shell noun in the pattern th-N, because the shell contents of such uses
tend to be uncommonly complex. Example (17.5) is a case in point:
(17.5) The company said yesterday that it would sell or close its 12 remaining
abattoirs, was cutting chicken production from over three million birds a
week to two million, and had abandoned property trading. These meas-
ures resulted in an extraordinary charge of pounds 92 million, which
wiped out the year'S profits, (PAPERS)
In the first sentence that I have quoted, readers are presented with three
chunks of information about activities of a company, all of which are ex-
pressed by clauses. When readers have processed this sentence, one can
assume that they have formed a mental representation of its content, or bet-
ter its 'gist', in their minds. The second sentence then provides information
about the consequences of these activities. Given the size of the shell con-
tent, it is of course impossible to simply repeat the three clauses and then
add the new information, since the short-term buffer would be full before
Integrating 371
the reader arrives at the new information. As a result, the cognitive system
would presumably sooner or later stall. In order to make it possible for
readers to be able to focus their attention on the new information, while at
the same time keeping the other information accessible but not in focus, the
writer introduces the noun phrase these measures as topic. By means of this
move, the three chunks of information given in the preceding sentence can
act as starting-point without burdening the short-term buffer.
The way I attribute it to shell nouns, the cognitive process of integration
is not only in accordance with de Beaugrande's "general constraints on
process models" (1982), but also compatible with several specific psycho-
logical models of text comprehension." With regard to de Beaugrande's
constraints, shell-content complexes seem to play an important role in com-
bining the retrospective representation of prior text, the perception of cur-
rent text, and the predictive representation of subsequent text in the proces-
sor's memory (1982: 183).
The integrating function of shell nouns finds its place perhaps most
readily in the phase of semantic integration in the "wholistic [sic] idea
model", which is described and supported with experimental evidence by
Bransford and Franks (1972). Although the two authors do not mention
abstract nouns explicitly, they provide evidence that the abstraction of ho-
listic ideas seems to be important for the recognition and recall of text pas-
sages. Undoubtedly, nouns whose main function is to create and name such
conceptual gestalts will facilitate this process. In view of the signposting
function of paragraph-initial shell nouns especially (see Section 16.3), it is
also interesting to mention Bransford and Franks' (1972: 237-241) claim
that the formation of holistic semantic structures is particularly important
for the memory of paragraphs.
In the well-known framework of Kintsch and van Dijk (1978), the inte-
grative function of shell nouns can be seen as an instruction by the writer
for his or her readers to take over information from their last processing
cycle into the next one (1978: 368). Being conceptually partitioned, the
propositions encapsulated by shell nouns will acquire prominent positions in
the coherence graphs, and are therefore favourite candidates for a transfer
to the next cycle in the memory buffer. This claim is also compatible with
more recent studies on the role of a memory buffer in text-processing, no
matter whether they include the idea of cyclical processing (see Miller and
Kintsch 1980, Fletcher 1981) or not (Haberlandt and Graesser 1990).
In Britton's proposal (1994), which builds on van Dijk's and Kintsch's
ideas but is specifically designed for the understanding of expository text,
the integrating process is modelled by an operator called UNITIZE. This

command has two functions which come particularly close to my notion of
integration. It ensures that the information which must be held in active
memory is unitized and thereby reduced so that there is capacity to read on.
And it is used as a chunking device,
to summarize a section of the text, to construct a macrostructural state-

ment of it, to reduce an extended argument to a brief statement of its con-
clusion, to extract the point of a section of a text, or to extract the point of
the whole text. (Britton 1994: 650)
In spite of the superficial terminological correspondence, my notion of inte-

gration differs from the use of the same term in the so-called Construction-
Integration Model of text comprehension (Kintsch 1988, 1994; see also
several contributions to Weaver et al. 1995), which has also developed out
of the early Kintsch and van Dijk (1978) and van Djik and Kintsch (1983)
models. In the Construction-Integration Model, integration is conceptual-
ized in a connectionist spirit as a recursive process of spreading activation
in a network. The result of the process is the stabilization of the cognitive
system, particularly at the ends of processing units like sentences (Kintsch
1988: 168-169, 1994: 732). My own notion of the integrating function of
shell nouns could be brought to bear here as one of a number of factors
which can contribute to speeding up this stabilization process, the result of
which is a mental representation of a chunk of text. The question of whether
shell nouns can actually be credited with such a role could presumably be
investigated experimentally by comparing the comprehension of texts with
and without shell nouns.
The obvious claim that the integrating function of shell nouns contrib-
utes to the coherence of texts is also compatible with models of text com-
prehension which do not accept the idea of a memory buffer. In Le Ny's
(1991: 206-209) model, for example, which builds on the idea of a constant
and gradual activation and de-activation of representations, shell nouns can
be seen as impeding the decay of a proposition in memory and as facilitating
its re-activation. And in Garnham's (1987, 1989) approach, which is based
on the notion of mental model, shell nouns can be incorporated as an im-
portant type of link between the parts of abstract mental models, which
supports the formation of integrated discourse representations.
It is important to emphasize that the details of information which are
given in the shell contents, for example in the three clauses in (17.5), are not
Integrating 373
just dumped, i.e. deleted from short-term memory. There is a good reason
why the process at work here is not called elimination or repudiation but
integration of information. What happens is that the details of the informa-
tion are condensed or conflated in a different conceptual gestalt rather than
eliminated altogether. In order to explain this integrative process, I would
like to resort once again to the semantic and cognitive match between the
meaning expressed by shell contents and the gaps in the semantic structures
of shell nouns. Due to this match, the resulting cognitive structure is an
integrative blend100 between the information given by the shell noun and that
given by the shell content. In example (17.5), it is the match between the
description of three actions taken by the company and the gap for activities
included in the noun measure which controls the integration of information.
As a result, the information expressed by the shell content is still fairly ac-
cessible and can be retrieved again at a relatively low activation cost.
The general cognitive benefits of the integrating function of shell nouns
should have become evident by now. How important this function is in ac-
tual discourse depends crucially on the nature of a text or piece of dis-
course, especially its subject-matter and the type of topic continuity that
controls the coherence of the text (Givn 1983). Discourse about concrete
events, as perhaps epitomized by fairy tales and other 'simple' narrative
texts with temporal sequencing and participant-related topic continuity
(Longacre 1983: 2-6), can do without shell nouns in integrating function,
because there is no need to reduce the complexity of information chunks.
Hearers and readers can make progress perfectly well from one utterance or
sentence to the next without having to take along complex pieces of infor-
mation. Temporal sequencing and participant-continuity enable them to
proceed in a step-by-step fashion, with each step consisting of first focusing
on a new piece of information and then removing it from the centre of atten-
tion.
It is clearly in abstract expository discourse that the integrating function
of shell nouns comes into its own. According to Longacre (1983: 3-6), ex-
pository writing is characterized by the absence of a temporal succession of
events and by the absence of agent-orientation. Instead, coherence is upheld
by the continuity of abstract discourse entities and topics (i.e. more by gen-
eral thematic continuity than by action and participant continuity in
Givn's (1983) terminology). Since abstract relations are cognitively far
more complex than the representations of participants and temporal stages,
the short-term buffer would constantly be at or beyond its capacity limita-
tions if both the given, i.e. already active, information and the new focal
information were to be processed in full detail. The integration of abstract

information (and the resulting reduction of its complexity and its load on the
active buffer) is thus a necessary precondition for the successful processing
of abstract expository writing. The same is also true of spoken discourse
about abstract topics. If it were not for the integrating potential of shell
nouns, people engaged in a scientific discussion would not be able to talk
about reasons and results of, differences and similarities between, or proofs,
implications or examples of abstract relations. Too much information would
have to be processed at one time.
An example from an expository text will serve to substantiate these
claims. The example is taken from an article by Whorf entitled "Language,
mind, and reality" (Whorf 1956: 247-248; the emphases by small capitals
are original):
(17.6) ... And my task is to explain an idea to all those who, if Western culture
survives the present welter of barbarism [i.e. World War Two, HJS],
may be pushed by events to leadership in reorganizing the whole human
future.
This idea is one too drastic to be penned up in a catch phrase. I would
rather leave it unnamed. It is the view that a noumenal worlda world
of hvperspace. of higher dimensionsawaits discovery by all the sci-
ences. which it will unite and unify, awaits discovery under its first as-
pect of a realm of PATTERNED RELATIONS, inconceivably manifold and
vet bearing a recognizable affinity to the rich and systematic organiza-
tion of LANGUAGE, including au fond mathematics and music, which are
ultimately of the same kindred as language. The idea is older than Plato
and at the time as new as our most revolutionary thinkers. It is implied
in Whitehead's world of prehensive aspects, and in relativity physics
with its four-dimensional continuum and its Riemann-Christoffel tensor
that sums up the PROPERTIES OF THE WORLD at any point-moment; while
one of the most thought-provoking of all modern presentations, and I
think the most original, is the Tertium Organum of Ouspensky. All that I
have to say on the subject that may be new is of the PREMONITION IN
LANGUAGE of the unknown, vaster worldthat world of which the
physical is but a surface or skin, and yet which we ARE IN and BELONG
TO. For the approach to reality through mathematics, which modern
knowledge is beginning to make, is merely the approach through one
special case of this relation to language.
The view implies that...
Integrating 375
Whorf is introducing the central topic of his article in this passage. Since
the topic is highly abstract, he goes about this introduction in a very cir-
cumspect way. The first hint he deploys is the indefinite shell-noun phrase
an idea, which functions as a cataphoric signpost, even though it is not
marked as such. This shell, which, as must be emphasized, is still empty at
this point, is taken up at the beginning of the second paragraph as topic of
the sentence and discourse topic of the paragraph. It is not until the second
sentence of this paragraph, however, that Whorf actually begins to fill the
shell, which he reiterates as the view, with content. As it happens, the con-
tent is of an enormous cognitive size and complexity. The only way to add
more information about this massive chunk of information is of course to
fell back on the shell noun, which now contains the information in inte-
grated form. This is precisely what Whorf does, of course, when he enlarges
on the history and current manifestations of the idea, informs the reader of
what he is going to contribute to the subject and, in the final paragraph
quoted only partially here, talks about implications of the view. It is only by
virtue of the hypostatizing and integrating potential of the shell nouns that
Whorf is also able to refer to the abstract relation by means of an anaphoric
it in the middle of the second paragraph. Such a reference would have been
impossible without a nominal shell.
The cognitive benefits of the shell nouns in this example are beyond
controversy. Speakers and writers can however also exploit the cognitive
functions of shell nouns for less laudable purposes. People can speak in
more or less empty shells, for example, if they do not know the specifics
that would be needed to fill them or do not want to disclose them to their
hearers. Many good examples of such (ab-)uses of shell nouns can be found
in interviews with politicians. Especially when they are forced by journalists
to comment on events or developments that they themselves have only just
been made aware of, politicians use shell nouns to avoid committing them-
selves to specific information or promising specific courses of future action.
An example of this type is given in (17.7):
(17.7) She said: "The information this [research The Times] paper provides
today does mean we need to take stock quite urgently of the implica-
tions of this new evidence as far as the policy is concerned."
The Times, 30 August 1996 (emphasis added)
Presented as it is in isolation, the communicative impact of this passage will

probably remain more or less opaque. As is indicated by the inverted com-
mas, the bulk of it represents a quotation, which is taken from an interview

with a politician. The context of this specific passage is that new scientific
findings suggest that the so-called mad cow disease (BSE) would be dying
out within five years, no matter whether cattle are killed in large numbers or
not. This information helps to interpret what the contents of the shell-noun
phrases the information and this new evidence are, since both nouns encap-
sulate precisely this piece of knowledge. More interestingly, however, the
contents of the other two shell nouns implications and policy are explained
nowhere, neither within the limits of the quoted passage nor in the article
from which it is taken. I would argue that this omission is by no means ac-
cidental but due to a deliberate strategy of the speaker of the quotation, a
junior agricultural minister, who is eager to avoid any public commitment to
future acts and measures, or perhaps does not know what exactly she is
talking about. In such an endeavour or calamity, 'empty' shells nouns come
in very handy indeed.
Let me finally summarize the insights gained in Chapter 17. To a large
extent, the set of cognitive functions and processes that I have discussed can
be captured by the notion of gestalt-formation. Shell nouns exist, and
speakers use them, because they can cast any types of experience that can-
not be described by a simple concrete word, no matter how cognitively
complex they are, into context-dependent conceptual gestalts. This explains
the following functional properties of the representations evoked by shell-
content complexes. First, as demanded by the gestalt principle of closure
(Koflka 1935: 151), the gestalts created by shell nouns represent bounded
conceptual regions. Second, in line with the gestalt principle of continuation
(Koka 1935: 153), these gestalts are integrated and unitary conceptual
structures. Third, as claimed by the Gestalt Psychologists (see e.g. Boring
1950: 588-590), the concepts created by shell nouns do not emerge by a
simple addition of the information given by the nouns and the information
given by the shell contents, but represent holistic conceptual structures in
their own right; this is the result of an integration process. And finally, the
conceptual gestalts created by shell nouns have the potential for reification
and hypostatization. They create the impression that they represent 'things'
rather than events or abstract relations, and that these things have a stable
and autonomous existence.
Chapter 18
Conclusion and outlook
With the discussion of cognitive functions in the previous section our tour
from authentic corpus data to the cognitive system draws to a close. This
study has provided ample evidence that a certain type of abstract nouns
allows speakers to create context-dependent conceptual shells. The relevant
nouns derive this potential from the very fact that they are nouns, which
entails that they have the potential for reifying and hypostatizing chunks of
experience as integrated conceptual gestalts. The crucial semantic prerequi-
site for the capacity of nouns to function as shell nouns is the existence of a
gap in their semantic structure. This gap is not only a precondition for shell-
nounhood, however, but it also controls the co-activation of shell nouns and
shell contents, especially when the two are not linked by grammatical con-
structions but by means of anaphora. The match between the structure-
inherent gap of a given shell noun and its shell content also facilitates the
process of integration. Although experimental evidence - over and above
the detailed linguistic analysis - has not been provided, it is claimed that
this process has crucial cognitive benefits, namely a reduction of the amount
of information that must be held in an active state by the discourse partici-
pants and a resulting relief of the load on the short-term memory buffer.
These cognitive functions of the use of shell nouns can be claimed to be
at work to a large extent in all their uses. The same is true, again to a high
degree, of the function of characterization. Some kinds of uses, especially
those of the noun fact in expressions of the type the fact that, may perhaps
be almost exclusively motivated by the cognitive need to create a concept.
On the whole, however, when they use shell nouns speakers can in fact
hardly avoid characterizing the information given in the shell content. Ironi-
cally, the characterizing potential of shell nouns is nevertheless hardly no-
ticed in many instances, and this is again due to the dovetail-like match
between the gap in the shell noun and the shell content. However, the char-
acterizing function can also be exploited by speakers in a highly deliberate
way, chiefly to manipulate the conceptual status of complex cognitive enti-
ties in the universe of discourse and to express their personal attitudes to-
wards and evaluations of these pieces of information.
378 Conclusion and outlook
In addition to cognitive and semantic functions, shell nouns can be used

in the service of a number of pragmatic, rhetorical and textual functions.
These are on the whole less pervasive than the cognitive and semantic ones
and depend crucially on the particular kind of usage of a given shell noun.
In constructions of the type the thing is that or the irony is that, shell nouns
are primarily used to focus attention on the clause following the copula.
When nouns other than the highly unspecific thing, point and question are
used in this pattern, this focusing effect is accompanied by a topicalizing
effect, which puts a secondary, anticipatory prominence peak on the major
characterizational element of the clause-initial shell-noun phrase. In ana-
phoric uses of the types this problem and that's the big question, shell
nouns make an important contribution to the identification of the conceptu-
ally unbounded antecedents to which they are related. The identification is
controlled by the semantic match between shell nouns and shell contents.
The semantic support for the anaphora is particularly helpful when there is
a great distance between the two groups of linguistic elements. Needless to
say, it contributes to the coherence of a text as well. The linking function
can also be exploited in order to signal structural boundaries between text-
semantic units and organizational structures of texts and their parts.
This, then, has basically been the answer to the questions of why shell
nouns are used by speakers and why they exist at all.
As far as the possible linguistic realisations of these functions are con-
cerned, I had to limit my attention to a selection of nouns and to the main
lexico-grammatical patterns in which they occur. The basic criterion was
the frequency of lexemes in uses that qualified as linguistic manifestations
of the three functions of characterization, concept-formation and linking on
lexico-grammatical grounds. This criterion was operationalized with the
help of a large corpus, which - granting a considerable skewage towards the
variety of English used by the media - can be seen to be representative of
the use of English and of the system of the English language. The 670 nouns
which met the criterion were differentiated with regard to their typicality for
the class of shell nouns. The basis for this assessment was a combination of
the quantitative results with collocational, morphological, semantic and
pragmatic considerations. The resulting typicality structure is summarized
in Table 13.2.
For the quantitative study of linguistic phenomena, extremely large
numbers of instances are needed. It is only possible to comply with this
requirement if examples can be collected and processed automatically.
Chiefly as a result of this methodological limitation, the description of the
Conclusion and outlook 379
use of the 670 nouns focused on four lexico-grammatical patterns. Of these,

two constructions received special attention (those in which shell contents
are realized by clauses that are linked to the shell nouns by structural
means, i.e. the patterns N-cl and N-e-cl), because their strong ties to their
shell contents allowed for a better assessment of the meanings of shell
nouns. In this analysis, the relation between the lexical meanings of the shell
nouns and the grammatical meanings that can be attributed to that-clauses
and infinitive clauses play a central role. On the whole, the extensive mate-
rial on the complementation of nouns looked at in this study has confirmed
and augmented the findings on ^//-clauses and infinitive clauses from
studies of verb complementation. On a very general level of description,
that-clauses are used for the description of fact-related shell contents, and
infinitives for event-related ones. In addition to the semantic complex of
volition, manipulation and future orientation that was previously attributed
to infinitive clauses for example by Wierzbicka (1988) and Givn (1990),
deontic and dynamic modality, ability in particular, and manner as well
were found to be related to infinitive clauses.
Even though the study has come full circle with the discussion of func-
tions of shell nouns, there are of course a number of loose ends which must
be pursued further in the future. To start with, the limitations arising from
the corpus method with regard to the nouns that are taken into consideration
and the patterns in which they occur should be overcome. Compounds and
even fixed expressions that can function as shell nouns, and nouns that are
predominantly used as shell nouns in the plural, would be obvious starting-
points for further research. As far as the lexico-grammatical usage of shell
nouns is concerned, other types of postmodifications, mainly prepositional
phrases, must be looked at more closely, and those patterns that I did inves-
tigate must be looked at less mechanically by looking at grammatical
structures rather than just sequences of words (as the corpus queries do).
A second large field open for further research is the use of shell nouns in
relation to text-types and to the distinction between spoken and written lan-
guage. Although the composition of the COBUILD corpus allowed me to
arrive at some fairly general observations concerning this matter, there is
still a lot to be done. My general expectation would be that the frequency of
shell nouns correlates with the abstractness of the subject-matter of a text,
with the style-level and with the overall communicative function. Shell
nouns are more frequent in texts on abstract topics that are written in a
neutral or formal style and serve an expository or argumentative function.
Colloquial spoken English, I expect, will turn out to be characterized by a
380 Conclusion and outlook
high frequency of a small number of typical and short shell nouns like thing,
fact, reason, point, problem, idea, feeling, aim, plan or question, and the
virtual absence of 'heavy' nouns like declaration, pronouncement, af-
firmation, precondition, preconception or misapprehension.
While the distribution of shell nouns across text-types would certainly be
an interesting research topic, it would be even more rewarding to examine
the relation between functions of shell nouns and text-types and medium-
related variation. Even my bird's-eye view of this issue has for example
indicated that the focusing function, especially with the nouns thing, point
and question, is more often exploited in spoken than in written language.
The same is true of strongly attitudinal and evaluative realisations of the
characterizing function (mainly of the pattern th-be-N). Shell nouns with a
signposting function, on the other hand, are of course typical of expository
writing and, in specific forms, particularly of newspaper texts. The inte-
grating function is especially important for all forms of highly abstract dis-
course, both in the written and the spoken medium.
Finally, an area of study should be mentioned which I have had to ex-
clude almost completely mainly for reasons of internal coherence, namely
the history of shell nouns. This field provides two sets of interesting ques-
tions. On the one hand, a systematic investigation of the etymologies of shell
nouns, in particular of those that are unanalyzable from a synchronic point
of view of course, could possibly support my claims concerning their basic
cognitive functions. As far as I have looked at it, the etymological evidence
suggests that with the exception of very few nouns like cause, which go
back to logical relations, shell nouns tend to be derived from verbs. This
supports Porzig's (1930) idea that all abstract nouns are ultimately reifica-
tions of the contents of whole sentences. On the other hand, some work must
also go into the question of when these abstract nouns came to be used as
shell nouns. This can be very revealing. Diachronic material on the noun
idea that I have looked at (Schmid 1997: 91-105), for example, indicates
that when this noun entered the English language in the 15th century, it was
exclusively used for the hypostatization of mental entities. The first linguis-
tic means of expressing the shell contents in this process were postmodify-
ing o/-prepositional phrases. According to the quotations in the OED, from
which the material in Schmid (1997) was taken, postnominal and comple-
menting tfjcrt-clauses and wA-clauses did not occur as shell contents of the
noun idea before the 18th century, and infinitives not before the 19th cen-
tury.
Appendix
In this appendix, information is provided on the frequencies of 670 nouns in

eight lexico-grammatical patterns. The data is taken from the 225 million-
word British section of the COBUILD corpus (see the description in Section
4.2). The columns give the following pieces of information:
1. Noun List of the 670 nouns which qualified as shell-nouns ac-

cording to the criteria laid down in Sections 2.2 and 4.2
2. -to Frequency of occurrence in the pattern "nomi + infinitive
clause"
3. -be-to Frequency of occurrence in the pattern "noun + is|was|has
been|will be|would be|would have been + infinitive clause"
4. -that Frequency of occurrence in the pattern "noun + that-
clause"
5. -be-that Frequency of occurrence in the pattern "noun + is|was|has
been| will be|would be|would have been + /Aa/-clause"
6. -wh Frequency of occurrence in the pattern "noun + Wi-clause"
Frequency of occurrence in the pattern "noun + is|was|has
7. N-be-wh been|will be|would be|would have been + w/z-clause"
Frequency of occurrence in the pattern "this|that + (adj) +
8. 77n's/that-N noun"
Frequency of occurrence in the pattern "this|that + is|was +
9. Thislthat-be- (adj) + noun"
10. Freq. as Frequency of the noun as collocate in shell noun-typical
collocate patterns: sum of columns 2 to 9
11. Freq. in corpus Frequency of the noun in the 225 million-word corpus
12. Comp, reliance Compiled reliance: score in column 10 divided by the score
in column 11 (see Section 4.5)
It should be noted that especially for rarer nouns, minor differences between
the frequency scores given in the Appendix and the frequencies of the nouns
in individual patterns as retrieved directly from the corpus are possible. The
382 Appendix
reason is that the scores given here are taken from the one thousand most
frequent nouns in the collocation program for the different subsections of
the corpus. It is, therefore, possible that noun-pattern combinations with
very low frequencies are not included.
Appendix 383
, sP
O** 0sÔ
o^ â
^ ôON Ô ^.C s^ N-O NN
ON O?NO S
N O
O sONN
ONO
N 1 ?ONN?ONN
? NOPNNP ON
^r- S o\ m M O rs
I
OVI'/ o oo oo ^r m 00 m O 00
)" ON ON T)-
l
ow-i M i m r i ^ O I
o a <So T
<Nf r ' - o<oS i' rr no NO ro -<

'
N
mO fl-l
1
r -< 1-i
<N ' 1
u m
i l N r- m m NO NO in oON NO 100<NO NO o o m N NO
, 00 NO
u-i -
m m <N <
O
m 00 00 m 00 m m <N o1"vi
NO <mN<N 00 <
m 11S
-
g no" co r i N
O rr fS O
& 11 rf r w -
T fl-T sr -* co ON" <N o"
<N <N '- CS r <N
ta o
~
< r-- O
TfrHr-TrmOO\Or-CTt 1
o ^r(Nr-1

< f-
t m ^ H vo h >/> vo m <N r- o- <N en< m
> VI no
m
fe
|
s
oo oo
) m o\ oo
oo ^r r-
^ t^
<N
ON M(S
- o\ r-
0\ r-
h -
"I
S * r^ m mt rs i M t; 2 ^ ^ ^ o\ m ^f
g *

s i
oi TJ- CN
Ms
< <s oo
m oo
( o\ tj- m ON 0\ 00
S
11o\<NTt
TT <N oo TT [ 0\ O
<N <N
o q (S
o.
<N N
ooO Ti- r-
es - NO
<N <N -q-
5 a
r- r- m
oo On
00K r- ' m
-si-
On"

I
i. <uu s Sf s >
O Sc 2
I f '2 >" I
I irQ."5~3SQ
ti
5 g S-S O
Sf 5 5 s 6 &c CM-- I
* t
sU iOi tj Sc -S -S -Si
S
O - .2 J?
-O -Cl 8o a8
<3
<3s
384 Appendix
vOx s ox sON x vO sO s so s ?x S
B< V ox o o o o o x c x cr- CX o x o x o o x sp s
0 s o v OS OS o x
r- os co >n 00 O <N tj- co VO 00 <S - t^ so
I < sq in 00 O (S SO co cts O SO m s ^ m 00 1 Os -rf
co co iri wS ri so co as S CN 1 oo' I CTs 00 s " 00 S so
I CN 1- Csl SO

O I ri o I T T Os SO 00 m >ri TT <N t< Os 00 00

SO w-l O - >n CTS Os 00 o SO m O T-H O SO t< r - SO Os O m <M s n 00
m so "r m >n t r- <N O - <s m <N 00^ I" >n
& g oo" oo" ,1 /- r " oo" t-" <N o" <N ( f" oo" so" so" <* I m " s o
fr I
< <s t1 <N <N
fe O
CTsCTsrNmmcicSCTsNsoCTsciNvioosO'
u-i s o c i t^ t ^ vi . oo o m v i oo cts s o o o os 00 u i } xr\
l Tt Ifl co oo m
o so" <N
fe
|
s
so 00 o o r - os < s 00 . - I ^H <N
"- O O <N <S <N CTs
m
a i
~ <N <N H N M r f i f l r - N O O u l r t i n T f ^ l O t f l ' <N T

a CTS o co o ^ - ^ - s o ^ v i o o c o c o o
1 t ^ (S
s i - ^ co"
es
o oo m t ^ f* v i o oo > ^ o cts
w <-H CO CO m cN co <N
s VO <N

3 00c0O00CTsr-<t^c0O so SO <N oo
> es co o oo <S ^ r r u-l <N Os O os as ir)
s O .-i 1-1 CN m co T f
z,
I r- u-l " y
o CN
a
h o\ TT f m
r - <N o co os
<N CO <N co w-i >n
c c
o 1
s
I i S I 2 a
I ^ g -I
a g I & & -g ? I- S ~
2 ~ ts ' S 2
s a, Ci, , CL Ci, .9 I s 1 1
o s. - o
I
2
0 81 "5

Appendix 385
"*"\0
0s*yO
\ ^x-
p 0\osx o^x vcpx s0s osÔcx ONN
o o N ON ox CX CX S
P SN OO
N sO
ON

soOx ovxOoxN=
oNx? ox ox >
0
m no on - m viNO <Nm 00 ON Or- NO00 TT es r^TT Ti o Tjn- ON
OO O < Sf r- ON O00 NO On Tf m m 00
r^ r-i NO<s< N >O n <N< vi -< NO(N <Noi m O n
<N C S <N
>nS oi oi O n m
' i
ONmrnooviONu-irôiONTtomv-iTi
h^nNinoOr-oOi^fnOiOrjO O
on <NN
NO 00Or-
m N
O' ON>
m/"l mOm
00 N M3oW-l^"r
<NM
n
* m >yi o on <N m rs" r-" in m" no" o~n<Non
fr
te s
N t M O h O t ^ O O r . _S in oo h t i - i ^ t n o o M ^ h
^-OoltîNmmTl.imviTS-
On I m (N <<N <N m
N
NO
O fS O
O
(mmmt^ on pitîniini/iM'i
< on mm ^ NO NO
3
l i
a S o>on r
m r-i
fNC^mrf^-ÔNinviONNOVl " 1
M -o\(S00 VI mo mo o/o vio m
<No- om rs m
e i ~ oo mooooo'nmoio'
/>
mt ^
<
<N
N
1
*
oooo m
Tj- no M ooHmh v>^ o vo o<N om
i> S
S
S C
N O on nono i-i
>n om r-<N r-
CS m o>nn no ^-ioo NO
1
or- <s oo
3 oo m 00 On
TT O
1
R g
0 <u s S .
t-t
0 a f 6C0 I So S s
3 1 S I ? ? 3
O s i s A oS3 S
o O Qi Qj
S so
-o -il: -sCl s
386 Appendix
x - x x vO yO s? N? VPx
o. o o o
x x
c x cr- o c x ox ox c x - N e v
ox ox o ox o
x x
m <N I CO o l m
o
00 <S m O) VO 00 om 0s
I
c
o
o^
vn
00 r - <N 00 >/ \ o s T T VO <N 00 o 00 r - m o w-i vq
00 m' o< c - i r n o ^ ^
'
Cvi fi vo oi ( VO vi VO oi wS -!
t( Ol (S m 04 fo T<
t
VO O l VO I 00 < 00 vo 00 VO 00 m m VO VO
!
00 O l O VO
CO m m m < - VO ON m m
a\ o i VO m f m a\ o r - o i
rr,
r n o . MD Ol < < VO <N o ^ O,
3 r o " o " r
fT r i r V
1
( S ro" o f >n o T ^ r i o T vo" vo" r-" ~ o
& r- m CS m Ol
fe O
VO 00 Os \ m VO ( S m <tc q u-l VO " > o VO - oo oo o

00 00 1 m
1 1
<N - - r - 00 VO <N u-i
q
o r-- Ol m es
O^ < 00 r - VO v o en
. - oo* rn* oT t
tL.
o o i o t t o o ^ t ^ t o Oi(S -S- < oi f > -q-

oo m o oo i n < ( S f -
m <
S
g
m o c~> ol ' OO OO h O - - ^ t m
O l O l --h Ol M t ' (S
00
1 r- r-

i o m ( vo o r- m 00
>n ^H vo ~ ~ "- vo <s oi
a

Ol r-
Ol
t<
/ t -
r - o i vo
VO
VO
<s
oi
vo
1 >n m VO
oi
VO 00 ov o i \ oo o O M
S m r- o t~> O l oo VO rr o <N
vo rr <N r - Ol ON O
!
< J
S c
v & M
^ >U j (ij S - C
5 ' S - t .
a
s
^SS * ft.
S 8 s J S . S 3 s ^ s
,
o
y
S"
? ?
_
?

a
?
o
J so roo s<Jo
Appendix 387
eu cx V sO
s oN
N V? X cx cx
o v C
^x csox
ox CX cx ON o <N cx lys 0s ox
00 o m in ~io O TT on 1 (* m m (N 00 VO m O fo
f O ci m - m I ' <S CO VO 0 <n <N t-; 0 C\ ro <N f
f- <s -; m <N t 1
* l-H in ri co Ti- en fi O v (N ON VO 00 0 vo ci ZU en o< vi o< oi CO
co r m <N <N <N 11
2
rl1 00 m 00 1I m vo si- Ov r- m ON 00 <N
11 00 > - m r- <s 00 <N o
1I - 00
r- 0 0 rs ON n ON 1 1VI1
11 \ VO I <s
vOft TT 1 r-; co ro vo
00" > t> r-" " t* r-i r r-" m" m" r ro vo" vo"

m '(Nm (S 1- 0\ (S in t (S t IH to m ON (
i s
s
m <N \ oOTj-oomOHHij-rnmTft^ -Io<So I

i n N t <S
o o t^H
13 <N <N ri Tf S oo >n
s?
J .2
mv-iinrscoONNt-^rNi'^^-^^(^<
5 ^ r- Tt. ^ l O O c rN O-foooot^v-imon
<-> <N ci >n
oo
rs
^ m m m oo oo >o
o <n
g i
o
-f
y. *
<N m oo ON vo
tJ vo <N <S > m
) <N VO vo O <N 00
<N <s
s Tf (s r~ <n m o1 (S > ^ V Cl o
<N f> oo o\ vo o o
<N m
i
<N
 -S? Cl
J
a 3- "3- 'S- ^1 S,
.8 S
<c> -S
^ ^ ^R s s
s r* r* R S R R
so c or*
r oe os oi
s s s o o o oo
t j u o u oy ou
o o u
388 Appendix
N so x sOn sO so
ONsoOnN ? vO
O ONvQ
N sO ON v0Os0X ON sOON vtfN
OON ON ON O N VO sO
N x
. V c-
vi O CS NO c Oo <N O O
N cr-? O
in
NN
O
N 00 00 1m (
N O
11q 00N ON <N c~N r-c
I 00 - 1 11ON I1 a\ o 11(N rs m r-> m m o 1 1 fo ON <N 00 m m
NO oi <N ci o ^ 00 <N o r^ <s
, 1 ) S

^ ^ (N 1 1 ^
rn
h m r- u-l o00 O -H O On On Ôn >n C

o00 o<nC<") o_ on
00N n t 1 on 00
U-1 O On O N r-
i/I <s ON NO O Nl oo
NO CN ri ON NO 00
OO NO
& g m O11 ON ^ 00 m
<N 11<s 1<N m m <3 11 ' m <N N On
on"p-" m"r" P-" no"<N oo" o" r co" <s" <N NO"
m
2 &
to
^ ot j
m i p- O <N o .-c W
On-l f-
m 00
VN NO Al r- m 00 r- m > n1 o NO m
00 <N / 11N
OO
N <N <N <S o 00 o00 ) 11m m <s m (1100 ON TJ-
<s
8 s
fe
<N 'S" - On m 00 ON 00 ? NO u-i m NO m 00 <N
s 00

no O >n NO NO N
mO 00 ON P- o to 00 ) r-
VO <N <S 1O m
<N <N m in oOn 00
m no fN)
m
S s NO m NO
g
5r *
1 on m \ r^ m m Tt vi Tt in in h
e on o\
m
S OC O h on t
no u-i oo
NO no no
1 q NO O
NO
11 00 m
O <N O
<N

I CSa o
C y <
s .S & au"ss

S 'S I i
so too a os o s a,
o
aO aO sS 'C s C
a C o -o "
u u <j u U 'C U <J
uCu a U ^S ^S 3
Appendix 389
N? SO
N vp sp x \?S ^ s? sPS -OSsoOS6s s?
CL ox es e OS So 's?
es S ^
es S ox OS s?
<n <r\cN
s os ?
S -s OS O
m O O St s?
os s
os sO
os
00 T o 00 m <S O Os 00 m Os <n 1

r- <N CS SO m Os 00 O 00 Os yr\(S o Vl ci T Os Os rs) O o
fS r-
Csj
Os so en r uS ci S 00 r t fs| oo' fi r^ SO TT 00 oo csi rs Os
<S <s o m t
2
VI <N so - SO
1 >Tio t>
Os r- m Os m rs <N Ul Os es m N 00 SO oo
- <O SO rr,
SO O o r- (N 1
rsl so (N Cl
1< o) o"is S
00 u-l <- O CO
m es 00, ci m r-. 00 -; so. m
S o" "T,
X" rl oo" rn so" 1 so" Os" v o oo" " <N , TT Os" (S so"
2 fr TT <N I
1 r-
t- 00
O Os r- Os 00 OS - o Wf ro - r- 00 O O
rr SO so <S o 00 so f) 00 <n SO s ^ O 1 oo Os1 so so
m <N < m (S, " <N CN <N so
<s" ' < V <N fi
|
fe S
n 00 Os m 00 <S (S so 00 ci m m 1 O r-< Os m VI CS|

13 SO 1< <N <s en m
lS32 ?J
e
a s t- n m <
fij <N
. I TJ- S m I 1
,
I
'y. *
oo m N M
s j oc m

m m t^ q TT O o oo o so
s 00 M "3-rs)r~tÔsOsfSso 00 <N CS t
m es - < TT
z,
1 ( m
so
SO
2 a
Os so Os m
~ <s in O
"
390 Appendix
\0
qN sO V
^C .O S V
-O O*. sCO^x sO V?x s sOx sOx sPx SPx ^
cO
x vP
ox sO sOx spx
V V o oN ?x o ox O o o c oN V? vO o
ox Tf
o^ cx o o
o\ oo h o " m O 00 o n O
(- O(Tv>n
>n on >n CO in - <s <n ~ CS 1-1
m o o w N N v i v o o o es t^ co ^ TT CS ON CO CO Tt co n in eo
vo - co <n C-~ ON l/S 00 VO <s r CS CS vo
CS es >n CO
VO < 00
- TT TT 00 t"- r - V
M S N O i i l O O ^ O O i n O VO 'S" CS h ri If
& 3 co -<3- oo} O (<1
CS ' O O W C l O O O O f O O
VI ^ Ov c^ 00 t >/> r- ov in m co
fr es" f-" oC ' 1- 1- < -1
co >
(S
ta S
in ^ * oc m t
es - co OmOONNiflKHO'nOOiOOiN^ V3 I" m 00
CS .-i .-1 es c oon es o ^ o
1" "->" i-* r- es" in
fe s
r- 00.'CSneSVOOtio<N
es ^ lo ^ T' T< u-l <S
<N <N 00 es O 00 V"! 00 O (S CS CO t oo o

ON io
CS t- es co

"I
1
VO I O O
M ^ t
-s
co CS VO vo ov
m CS <N VO O
? O oo
t"
o VO
> co oo es
<s
<n >n >n vo oo n r- oo oo -

<s ON CS O 00 CS 00 VO vo
Ov in (S CS VO
s; jj
Iiltllliilsiiitlailiili
Appendix 391
Q. V S
y4O
(V\0
> o^\0 O
\0^ \0O*\0O^
\P dv
e^ s
V s
V se\ e\ ox ox ox v=x sO
x vo s= so
ox csO
x V?
6V x
vo r oo 00 in 00 O O o\ r- 00 00 S T
<om cm 00 t ( 1 CO VO 00 Ov o
00 <N ov - m CN o co 00 VO 1 00 vq <S 00 CO - r^ co
5i 3 ov CO CN Ov oi S V
O1 r-' vo
<N
oi
' '
rn ov S co CN CN co
<N <S ' 1
h
m Ov -<CO o co ov Ov >/"i 3- r- V O m <N VO Ov - 00 O
vi ov T
ov o o co >n VO VO co "
u-l 00 VO O VO o I m Ov oI-H C
co >/~> N
sr 5 vo" co" m < 1
r-; 00K vo r- m </1 co O VOK CN vo rr
2 & <N s co" ci r- r* vo" o" <N oo" o" >/-f o" w-T vo" P-" -" r-"
<N co - m VO I < <S
h o
- VO cn CO co I1c V
o I O vn VO U-l m er Ov co co CN
co 00 <Nm m
<N P-; V- >o 00
wO w-i >n
m ts CO V O co 00 co 00 Ov' <r
- co
vo CN f <N >n P- 1
v co
,-T u-T oo" <N
<N
ta S
<N <N CN
I < co rr vo >n m m - <N fN < vn N vo O
mN <
v CN r- OV Ov 00
S t- VO I1 r- fI co<N vo
m es r< CS co CN
S
S >8
55 VO 00 00 Tt <r < N Ov U-l r-

O <s m Ov VO Ov 00 O
m t^ v
v
rr\ 00 o oo >rv co
mN vo
o CN CN C
00
<N
00 VO <n VO<
00 UN-! >n
CN <N 2 vo
Ov en C
CN
N
13

u VO
o\ vo
m
1 oCO
1 s
r fl
00 CS co
11 <N
VO co co VO
s <N r
s
VI VI r- 00 O TT <N - VO 00 OV
3 00 ov o CO m O1 Ov /-> VO CN C
coN
CN I <N
1
WT-
, vo" t
1 <N
1< co
CN CN vo
CN
a ov
vo -
so
* 1
8 S
e
tv CV1 <oIsO>C >- C
s S 5 3 y
S? ? m 2 S s l ^ 11?
o & . 3 S e e
55 S ^ S S S S S S S <5, <S. 'S, <5, S. S, <S. <S, <S, <5, S, .
392 Appendix
f N vC O sPs sO x x
o ^s V S ^ ox s? s vP sO
On ox ox ff tfv ? ox V?n QV
Vf oo
>
OP
i/-> ooP- m0 fio 00 o m o\ r ov <S 00 Tf >/"l m CTv<S O N O m Ov O
m 1 V
<N O m <s vo r M m oo o t m <N fS ~ ~ I O f<"> r- <S vo r- vo
c-i r^ V O S 0000 vi vi ro VO < t> vo S (S VO <N H Tf o1 1^
vo r ci ri mf n vi VO VO < vo VI
M \ 00 h rmn om ^
o fvo
s iloooo
ntvi
3 VOn \ 00 V)
" r-^ t rn ^ oc_ Tt < m
<s
" r- m 00 m
tt vo^
""
/ < ' < vo* ^ ro w- r~ < vrT
2 & m
1
to o
otN>'00-)V>ot~~oo<Nt^voovfn>/-i
o ^ o v t ^ f ' i O h ^ c i O r H in m t ' 0 > v o m o o < N ^ - o r f v o o o o o
^ oc m <N ^ <N t h r- m es \ I<s^ ,m . ^ (S <N o\ 1/1
MS
fe
Tf m ^ r \
S
vo m mf . \ O
I-v >n
00
S as
.a
<1
venoin
m ^vi- ^cl- m
o\oho oTj
r - ^ - o^ Oî mfîio
ih in Tf h <N ci (N m
.>
."S*
e i
5 1
I <S
m 00

-a
vo vo o\ r-
m
o
o\
a\
2 s m
r- o <s trt oo
S VO r- Tj- m r- r- e> r-
Z, <N
Appendix 393
N
so

x
> /-1 TJ- < c Wcxc"
x
oO oVPoxSPox so
N N
ox osO cx ox se
x v?
oN \0
ox sO QS \PV-O -O
QS sO Q\ so NP
in
so fi Cl < <N
OS TT< ; h - " VO o' Os
<Nu-l O SO O O O -> 00 O O
t - 00 OsCI SO Ci 00 /
fi <N <S
< ri<m m ri ri -< ^ fi 00 fi S fi t fi
<S
~ <s <N ri fS <N fi <S <<
N
N
1
tr-lOOr-iNOTmi
- o oo o oo wi Os m jI - <N T < <s <N Os
t 00 < N so
<r\ >o wi m ts f- <n so >n 00 r- '/v SO
- (S O Os "T, <N o<N v <N
& g vo fS
S & fi" so" <N SOr- o" fi"
I o"
1 00 SO
1
<N
ta o
r~-(NOvCT*NcNoooooNOvmoô i o>nootôm
IN o ^ fS iviTf*-<csf-'3't v i o \ o o u i t N t - - c N o c > v i t
- /-> on
fN (N M

a j
f~ os
S oo w-i Sw-iOvCTvw-ioooi-l^rôo
N i f l r n i o \o
i M<-ic i n<nt f ' N
< t^ m
<N oo m
S 13

?i f
*> V. M m io
vo
r) 00
M tci co vi
Cl oOsM1" 0\
S <N <N
S
S m in M
oo oo
rr (N o <Ih(Nm ooO O Os
Tt " N O
'S VO h oo m r) ( <S
"
<u
Tt ON
r
ON
SO SD
<N O
<N
ss
s S _I
s
S
Seii^ 'S. . -s -a I ~
o . I .
O . 1
& & S, 5 I J J ^ t
: S -1.1 -I .S .g -I .g .
. ."
.
394 Appendix
O N
xP sS O= y O y=
N O \S 0 s
O v P nsO v O y \p \0 yO \0 xÔ s O vp yO \ sO SN0 s p s p yp yp
Q0SX
m
CL V jS
tfN ONS o'*.
N ^ NS ON ON Q^ S QS Q^ V QS QS QS S Q
eS Q
o^ Q
^ Q
S'
o S o' Q S'
o
m
m On O
IO es ON o T j - m O O O O O O O O N ' i '
,1 0 0
m
00 00 00 00 Tl On ON NO Cl Tl o <N 00 <N NO
t m TT s Tt t I*
^
la
o o T< m On Cl Tl - On NO ON NO m - O 1- 00 r -
m
o O n o ON r<
1 TJ- ON rr r^ ^ TT NO NO NO NO r- Tl 00 Tl 00 oo Tl
< N 0 0 m ci 00 <D 00 < N t n r
tj- m m m
9
5 ri ri Cl"
o
r- r i <N ri <N <s On" r i no" r i m
2 & ToT t^ Tl
u
to o
"
Tl O T i c s o r ^ - t . o o o <N s r- 00 o Tl s - On - I1 f o ^r
-e < < < -
m
t- <S Tl 00 r C l o o O Tl Tl <N
oo o Cl ' <N 00 1 < N <s m
(S
si
fe
m r <N Tt t- m o ro
s Ti <N <S
m
.13 ^
<N CO t- OO O 4 vi ^ T t^ r-c N O o <N r- o
<N
- J-l
T l .I > r- m Tt <s ro r- fN
e *
NO OI
M
<1
m
r oo TI ic NO
TI
-s s

m m
o r- t (N ON ON r- no oo
NO ^ I NO I1 ON m m oo
o NO f-
NO NO t 00 Tj" N O
<s
<N < M n
CO 0 0 On ON t
ti- c i NO m NO o oo 00 ON )
I 00 ON m no

-, -s 'S 0 R 5 ~

S
i ~ -a U s
-s sy s s Sf e s S? t c
fi) k
s S ^
S 6"
5
S S S feos ^ s h c
5 ~ ~ -Sf a S s S -SP - a s; s;
6J O) &J
o G G ^ ^ ^ ^ S <S> < - -5 s ? ! B .
.s s s s s s s s s K s s f s i ? e s C .R .S .R .R R .S 'C
.R .R
Appendix 395
N SO NO s o s p SO xOX NO v o N^ NP SP SO \P sP P \P \P \ P N?
OS ON OS ON OS OS s? ON o ? N ? ON
sON ? NP^ ON ON ON ON ON ON N? ON ON N Q\
ON ON N
O s o oo m m 00 VI f O ON OO m ON CO ON CS o oo nj- on m r s en
<s <? v i | 1 < S V i ON o o r - i n m a vi o oo ; <n o o o
SO O O r S 0 0 SO v i <s en Tt Vi e n ,; so (n < S SO tT ^ so o ci
( -> 11 co f -I <-> OI
o\ r) in ' - ^ - < ' ' 3 - <

c - > s o o o O N O O O N V > O N O O o m > '/ " )' r >^ o <CN
N <fNN O~ i - Tt ri O .' <\
&
o- o- <- 3 - o v oo^ < n NO -
k r- rt_ t^ < \ < > \
s
T f " tri r i t-" so" c T r i so" f-" e s " m " > s o " <N * no" " c o
S & m vi m in
' O O o v i T i - t ^ o o v N O T j - ' 0 0 T j - O N 0 0 0 0 r - - ' 3 - 0 0 0 0 . ' O N t ^ t ^ N o

v i m o v i t ^ m ^ m ' - - . o o \ r-1 o\ o t t^ o
oo s o t^ f o r- ^- u-i m r ^ m ^ j - t j - e s <N
<s" r*" <s"
S
to
O Ifl t t^ " VI r<"> t en i- r- so s o <s

13 ii m <s
<S <S
o n 1I r o I oo no .-i
a s
ci o ^ V \o m c-> vi o o fi m 00 Tj-
a * ON on o vi i- t m oo m
-i vi tn ^ <s
S *
y r-
2 t
ON SO o i o (N SO
m <N r-
f s
S
Ti- O m O 00 CM V I TT ON ( S O
-ra- 00 V) Os ON ) in
<S - r) <S
s
Z,
"S
C
o
- S
g
o . .
396 Appendix
a sO sp ^p vP sO SO
SO sO NO N? sO \ON f ON ON N ON O Np O
NN
ON
0s NN
ON QN N ON n ' AN o^- ON N? 0X
00 NO O O 00 m (N m Tf o 00 t1 ON CO o 00 co - (S 00 1
on (N 00 <N W) F" o ON NO en 00 r^ en m co o CN| NO 1' ' 00 co On
* r r <S 00 CO u-i NO On r-i ON (S TT >ri 00 en m H co o co On (S wi On co <N 00
I1
t < co '
^
CN NO NO On co r - ON (N 00 NO 'Vi m NO TT NO o o t On
NO O NO NO I-H <N o (1 NO es <N 00 m 00 rlo NO
On ON O
NO NO
co
no o r- co r r 00 > O rn ""!> es <N On t "V in f NO^ ' ' o ON '
" 3 00 r-" <N On" ro On" r f irf of <s co"
2 & ti- <N
fe g
On On m <N o o O NO On - O vo m NO NO co I- co ) . CS NO
on - <N *n CS 00 >/-> co On m NO NO | NO ON ON On On - CN O
' co "I NO eo | 00 v-l (S 1 t'

fe s
^ 3
en <s *- O Tt 00
13 no ~ r- rt
^ <N t - TT
O eNcot^w-it'cNOeScomoo
tj- l Tj- O
~ ~ CSOOeNt^t^NOCNNOOO
vo Tt h - o
no
i
ri O 0 i n f S 0 \ f S i N O ^ co rr
>I CO <' ' t'
? s <N
m CN in m o
es >r> w-i ^ rr ON co
t- --
o O ON O t
<N O TT U~l U-l
<s <N "-i
3 a
00 o >n <N vo
4 oo TT oo / (N
00 - CS
C
2
I I ?
&> S ta
!
g 3 y
"3 fe, <5 M I H s -S ? S O ta
S - S S >4 u s -s, a , ^ a g .0 > O
s * .8 o - .
o I I 1 1 U U e c te
S E S

Appendix 397
CTv
ri < lVO Oo\ T
<00
<N t"- vt oi m
tn^no oi ôoor o^ iôr vn itvN
om m vs-T
Vi <N O
tiT r CS f1
00 <N oo
>3- <S
er" s "ri O o vo"
ro
Tf o\ >n m
m C3 1 o O
"- vo" -! <s" <s cT
- f
" o"
& <N -- o)
fe o
O'-'iN'^rm'^-mo ocrvvoii-ioov-ivfiv
j H (S H <S m - v i f i tr s- o t ^f>
f t r s 1o o
f
,2 f i <N
"o
V"l
I
CN <N Tt - "
/> fi
fS VO
S
fc O N i f l1O O t m O U i o>/-i r-
es
( - (S fS f>

-" CN
13
O
TT
vo - o

<s ) (S
V
TfO ^1V)

m
<S T < -
(S
T
-- o u-i

- oo

tI fl <s
VO O <-i r-
VO>nvo r-
m oVO
r-I 1/i
1rsfl
z,
398 Appendix
000s N?o x N SO0X

es cx ox 0s=s- c O
x
^ so^
O ON ON N
s ON O
ON s?N x N? ON O
NN
?ONN
? N0ON
. on (S 00 r^ r- ON ON 0On On 00 om m On gN 00 On r-1
1 = mm r-
rq
o m on m 00
o
so O n so p <N - 00 r-; O " o O

- (S <s (S m uS 1 I, T ri ^ NO 00 00 ON On
no NO ro 00
1, mT T~H m rn
^ 1
>n m O sf 00 >/-> on o\ <s m O N ON SO oo

O
m o00 N fO no r- * 00 o 1 <s
o Om
OO ' NNO
OO N s
om o<3 (S
<N O f- q 11<N m 00
" 3 no"t-" on"on" no"<N oo" no" r-" n >n fO no"
2 & I,
, o m I -"
o" rq ri
to o
NO r< on
r">
, t 00 00 00 V
o 00 m
I so
1trl V
, II f-
I-H o00 m o NO r^ o NO NO - rNi 00
m mN om
CD <
m
es m <N t00 r 00 t- m 00 Ol
'
O <S 00 NO NO
NO
oo" t " V
fT ri oo" m"
ri
r |
fe S
On ON <S <N 00 oI-H ON1 so m ON

3 m m (
<N
o
NO
<N
i
fn f On fi m m O m ro NO r-*00 m 00 .00 m
a a? <N NO , on <N CS SOrtm < 00 u-i <N t SOI, (N N
OO
n NO o fri
*> u-l m - nok m
g i ri ri ri
On o
1 <N t
>n
ri
U-1 on
.
rt r- m m <N W gr rq 00
*> s
i< r- so >n NO <N o

on m o NO NO On On O
S t-
NO - 00
tN VO N mO
00 <s ON ON
mN ,
< rr <
NN
O
2 a
O m
onx o >n oo
NO <
1 00 00
00 ND NO
ON On NO NO
NO N O Tj"
r- >1 On -; 00
r-" fn" en
,
c
eo R S o
c

s
o
.5 fe
a.
S; "
! J "C 'e
s^ o ft, " U
3
2
S iQJ Q
-

'c os . -S. U R e
i J gQ) - '- -s:
fc er ,
er So o S S
, ft,
8 - .
ft, ft. ft, ft, ft, ft. ft,
O
ft, ft, ft, ft, ft,ft,"ft, "ft.
Appendix 399
s N? sP yO yO sP sp sO -yO V? v P V? N? V?
. cx ox o ?x o x o x o x ox o x o^ o x o x Q\ Q\ q \ Q\ Q\ Q\ c x o x v ox ox ox ox ox ox
r- VO O so Ov CS r ^ VO o Ose s - O Tf i n es e s o o >/-i en e s
I m f 00 <r T <S <S en <S Ov e s o O esm ^ r - r O SOSOe s en O 00
VO t" (S Os r-' en r <S VO r VO TT iri o o en e s en w-i Ov Os en
en 11 ( S r' ( S I i1 T-H T-H en
Os en
q e s VO en >/n Os
es o Oso >n ^ r - VO ( es 00 so
-SD so en
O i< en r - SO
>n e ^ VO e s 00 VO o Os o VO o n en o t
>n t< CD <3 i n
CS o CS en VO, e s r - m 00 en m Ov I r - en 00 T
Os SO
& es" , oo"
o" oo" es" o " 00* en" , e s es" en"
SO en r- en
fe .
Os1/1 Oso 00 CS r - o r~ en en r - r-
rtVO
en Ov en OV >o VO r - 00
(S T o so en o o Os Ov C eo Tf v o r - O f en
VO r - r- m v q v 00 - es en ii
rtes" m " es" en"
3 tlOse(1
r s so
( -
C l /i VO CS r r
VO o >n es
es
es
Tt - CO e s vo cs Osen

<N 00 VO Osen Os es VO e s 00 m v o 00 en Os00

s r- t o m Os 00 o en e s Ti
en 00 TT
T
en en 00
os ' " i , en
r> es VO rn
S * * SO 1
ov
1 VO
ov o r- es Tf in
00 Ov <N
^ s o i es

vo - VO Os eS m ,_ } o 00
S
J
Ov t ' Os e s m t j - t io m Os
1 "t -< <N
en o IN
I VO - r- vo
en en o r- . VO Tf Os
3
I
Ov o v
VO
00 r- en oso T
v o .< m
z, 00 TT
vo
<N

s tu S
2 c
S ,
s
a a 8 g-
_ S .. I
- &> ' s .a
^ . s M s? I
l ' I G ts Ts ' S ftp
ftO ' S fcj ? "Si >
S S o SS
9 O O O O I S g e c c S s s
I 5
g <4. a . a,
400 Appendix
so
N s= v= ox O? s? ox N
V^
U-1
O
x x x
lonl o<N oO o ? ox ox OV tfvox Vi V ov

l
ff^
f sov
l
O Ox svn
OV s(5O
<s
v
TJ- Ov vo <
N Ov Tf Cl fS VO < 00
1100 ON O
N 00 n VO oo
I-; 00 m <N
l l <N vq - m
f CS w-i <S o es 00 O

l i . 4 oo'1 li w-i VO uS S rn (N 1ri 1i fi vS l fi
S 00
fi oi o11CS (S 00 1O tiv
T
t-
l l <N (S
2
m S o S m U-1 V00 00 o TT o\ O
O "S" vo 00
r^ 1t- 1O
>rN
> 00 VOr- O NVOVt
r- (S V Or-
u-l o
On (S n o
VOon <N vtO_<1 o TT
m o
S v VO < m 00 <N T
TT
T Vi TT <N
Ov"vo" es m"11om n"Tf" om
o"11 vo" vo"
OV" r-" on"Ov" w-T
TT
ta o
m
ri vt m on O v
O O on oVOm m t- m < <N O VO1m
1O t^
O mo " ONOv VO
1 m vo
vo V
1 <N <r <N <N <N 11VO<s r- 11*n m >n V
f- l vl VO00 n
fSO O
n
n" vo" ( 11" 1-1 1" ri
ta
s
m m On <s oo <s in
3 <N <N VO <N fi 90
fi oo rs
<c

1 CM
Ii
^
u-l O t^l^nv-ific-i^j-Tj-oooi
<N <3- </-> <s ^ M ^t
OOOv0\ n m o
m <s - î <r> ni
<N fS <NO ^ o
-O r-
<N
<N <S 00 <t 1 t-<S-H>0 vo lf> m

vo
i
VO

S 00 (S
O IMMfl <s
- Ov il -"T O <S
n Tt vo
1 Ol
m m Oi
tm 00
<N o\(S ^
1
41
5 s
V
Ors
m vo
oo f>
<N -q-
00
o s
< 00
N
T ON O
m On
1 \or~ on
oo
0\
oo r-
o
ss

!

I
S. <a g R 5S? &fs
s* S c ta -S 1-
J 'CO S
0 la s ? t
o 4J
S .c .o ,> I -o u S S"
S S i c p r r 1a, a,
I
a, a, a. a . a ^ a . a . a . a . a . a , a. a, a, a, a, a,
Appendix 401
O*N N
O - ONO et N
> TIS Or- <n (N 00 00 et >/- ri t< O TI
< <s r- NON
O f) o
< ON
C
"
et <s o or< "V O NO
1 N cs- N
OTi O/- 001 N
et o <N <N1 r-
oO NO NO O
" I et r oo" co en" , NO" co et r co NO" <s"
<r <N
h r H O O ^ O O O O S f(NoovihiiowOiiflr-'HÔN
lWriMMO^nOÔW
"-fSÔONÔrHrjCf-fNOintÔr-ovir-ri o rs " OO
<N 00
rr o
H u
~ <N
in oo et on <N >n Tf es no oo m vi
3 on
et
NO O
S
.a
fc no
t-
ON ->3-
00 -<
et
oN
NO
O O- *-< et o o ^ o o i o o o o w i m c i p i o s >/-i
no ON NO < ' fs r- <N
S
r- on
</->
-o <s on
et oo
1 r- vo
m i-i 0(S-5f
,-h m -s-
fS O-"3N
- ONf
On 0m
O 00 0 ON
<> s
3
g o m vo O oo 1 00 <N
N <N
O -
( o( <s r~ m <s
"
S S * O
<N NO
-e<>u <N 00
m et o
r- ON
ON q
fet vo ON
I
o
* "e
o s g!
c8 SJ ? 0 '5 e^ g
O
s 8 , & fc t o"
tS
I _S
"? Q < 00 I
OU /
,
O > >
'S
a N
a ae a a S
a a8 bO . .
C c a
e
s ti <e4
g
402 Appendix
x x ^
ON
N?
^-s? ox S?
>/-> en 00
x NP
x "s? x x Vf X x
00V or- cVOO
00 o oen N 000or- V xOCX Vx? x
cnxf Tt
oN ox osO
o oxces ooo t--
x
? V1
VO VO 00
es t rr O es oo 11 00 VO rr >n 11 p es
,1
VO en vq VI 00
en o o r enov vi
* en en VOO N VO 00 es
1 n il en es vi es en vi
en es es 1
Tf en VO S es
es 11 es H ci

en
o<N es TT 00 n o\
es v> r- 00
oo o r-
- oen 00 VOV OO VO NVOen O v -
00 vi O N
O
00N
en
en <s en rr en V
n es O,es 11 Ses
00
On "V O
es n m 00
es^ 00 ON oes en
> 00K
en"en"t-" en es" es" es" vf oo" vo" es" <N r-" r-" Ov"11 en es" o" n
& O
h
09
& g
H ohioaittoOmONt Ôv ii H
i nr i- n- t tÔveSeSentÔOOnvotêSt^TfOs
hhomoi^rJ^nMiriinOiO
rHtOnOON'THOOHOO vo m r-1 es en es t^ 11 TT t u
S' n" es"
ta S
00
1 oen M
n Ovenr< h s en
a (N
t^
1 Cf
S oo in M -i-<3-esesvocncncnn/vono C
ss vo t en vi es t ov es on t n o \ ^/>',3- 00S 1-1 Oco
e * es on n rs t
"
1
f -s
*
> r-t
VI en
^ oo
oo r r- o\ oo
On
<s oo en O
t>v es on f"/"! ot 00
t 00
3 m
es -5T r-c VI Tj- n <S ovo
O
enr
00 00
a VO en <en
N en vo
O oo
a
vo vo o oo O
O VO > - es
ON On eS ent-
en
j es en n eses>o
es
R 1 s> s
I s H I
g fc S os !
>
"3 S
u
U ->"3
s
I gs
5? ~ "S a.

S
o P
bott S
i lâ,o a. sa. so so% S,S
? " "S "3 I
I I I
Appendix 403
vO NO N? sO ô SO SO VO sO sO vO N? ^P SO s o N? S
^ ox o x CX cx ox cx ox ox ox ox cx ox cx ON o x c x
- t( VI ci Os U-l v o TT O 00 00 O O ( N <n 00 <N T f 00 00 <N
I 00 a\ - Cl m r- so r ^ <N On m On - ( S o On Vi ON
as

a > iri <N < S <r <N I-H vS
<N ^
<N wS 00 u-i 00 ON ^ f<S 11 ro
' ^
< >/-> <S as 00 as 00 00 ( S 00 00 O r - NO NO On O

m o
^
> t u-i
- On <N o r - 00 <N es o Os
00 00 >n O On 00 m < On 00 o o ON
so."Ts <o 1 <N 00K 1 ' 00K O , 00 On ON I'
"T,
-
as m" SD SO m " " o " 1< r SO
r i oo" c o >n ( N ( I " rr
2 & <N m
fe
^ o1
>/-> m NO >n T 00 NO I VO ON m ON m o U- ON NO <n On NO NO - On 00

t- r 00 o ( S 00 f m 00 O SO sO 1
l-H
00 m 00 <N r"> o
n
00 00 <s
<s m o m t m ^f m NO

oo" iI <N of
fe s
i l
CN o CS m t- 00 m r C1 <N I< r NO m o <n ON NO NO On On I

s f<"> v o TT m <N en U-l
^
a j>
e *
Os m On <N SO m as ' O
s
CS O
CN o - Os cOi r- so m m
<S
t <N t> O< s I< NO G o
00
<N <
00
-
c> so ON <N 1
e
i

? *
^ *
r- r- r-
<r <s m
-9 S

t o oo 00 ve 00 Ô 00 f 1
Tt O 00 t^ t I' V I Ol o o on m m
Pi m m <s m
t CS o Tt
M oo
m
2
<N as r- 00 T f
<N 3" ^ <N ^h
r- m O <N
I

_ i
<u f3 C . 1 , C
_ .s () S < 2 , a- s
S c c o
s -o
5 ^
-S ~ S e I g 1 1 1 1 ^ ^
O b ^ -a o a * Vi J s * ) .).
t o c s 1 2 2 .
"C S S g g
g a

s

< '<

^ J ^ } ? ?
404 Appendix
s? N? VO sO
ox vP N? S OS oN S? sx OS "5 SO vO X? s? S
o ox
00 s* s?
Os
cr-
oc
s? ox
VI O
ox S ox OS S S
Os Os 00 Os os o m OS
es OS
Os m Os
s?
S
Vi
S
VI os
so os
I m SO
00 o
tsO SO 00 o < 1 s SO
00 fS <N
m m <N SO f ' 00 m 00 m O
OS
-s <N rn SO <N Csj SO f i <S Os H >ri <S m H S S VI SO c-i vi Tf
m <N
s ^ '
Os (S SO <S so 00 TT Os SO 00 m Os VI O f> m Os
Os o Os r- 00 00 r-
OS SO1 <s O 00 T< TT r-
, !
V1 o
o, sO__ r-
<N O m I m so
q It mVI 00
"is
00
m
Os
Tf 00 SO Ol o so
SOk m SO^ rN
* 3 <N oo" vi" r-T rs
, so"

so" o" vi" F^r m" Os" o" so" vi" oo" t<
& <N <N f l fs| <N
ta O
r <N vi SO SO W- co <N m <r< 00 OS VI

vi <N
SO r vi m Om < r m SO so
<s u-l r- O m Os VI - m Os VI1 -
<N m Ol r- fS o <N < m Tf 00
<N so" esf
|
ta S
Os m
t < fi SO
<N vi <N SO <s m tt <S <N r- VI 00 r- o r- O
m
13 VI VI VI tt
m rt
n
ffn Os Os m O
oo SO
<N m
11 O o - Vi r-
I * Os 00
Os mI O 00 O
00 OS Os m o
' o
VI OS fN 00
S O VI fs) fn I Tj-

<N r-;" f < TT rn q r-
e * TT
5 1
\ vi < r-
1 o < so
vi os
< o
r- )- v>
< f1t m <S es trs
os <N <N ' Os ro
m vi
S IN
fS m oo os oo m
s <S <s tj- so -- vi
Os oo
S
i o Os m <N os
<N O-l O 1
vi oo Os O fS SO O
i ON OS Cl
r- so
m vi en
<N
O O O
sO f- O
t S tu
'c 9 -S a I s S)
bo $.3tu
o ) g s s a,
Sg> S <n n <n o o
I S
<0 .*5
3 S
"5 "3
a, a,
7 a,
ET <? n t l al s ' s
&>
Appendix 405
O "> TJ- 00 Os 00 o f l 00 en <r ( H f o en

<1 VO ON Os >n vo o O
<S <S TJ-
i vo > f 00 os Os o es 1 en O
vo <n vi i"- Ov
> S r-i en" U-f I
es ca m m es en O en es
' V vo" oo" es" en" en" oo" es"
<
r es" es" oo" r-"
2 O & en
fc.
" u
Tt <n i c 1 T f o e n < / - i v o e n o o o \ o > O T t ^ o o o v o o o o o v o e s t ^ e N

i o /"> en r- ^ ^_ . ^ . . - l
oo e rh^ in
^ i / - * . Q o tTh rvt^ H m ifi
r- vi u-i \ NriîCNOOrJOv o 1/-1 es
S S cs" es"
ta TS
S
es es oo es m oo oo es O en o m oo fi o h- 00 es t-
* VO r- es as <s (S os vo oo
a
s
f i o >-i r- oo oo _ .r- J
o ^ vom^rmiiooooin
."3 fc oo es ^ h M t ri CS 00 es o m m vo
n <N es f) es t m
e l
/"> f en oo en
O 00 00 es a\
en es es vo o es
oo en o f 00 vo
t os
S m O o oo * m os
VO fl >n vo es
Tt es -> oo
c
s o
2 : :
C . -s; I M rs
a,-H s _ _ k I s
- bo s
a, 5; & vS 4 J 5 So Jt. y
P.. s c s
e ~
S f c c c
t
a a a r sr 3 5 a s s s
406 Appendix
sO vO \C ^ -, c ô . o
ox CX o x 0s"

CX c x 0s V o x
oN 0 s
1 1 OsCN
O^ 0s o^ O*1 V
11
I
P-
P-
00
m
VO
O
CN O VO o n VO CO CN
CO Ov VO ON m
00
VI
00

VO
VO q
m
0 0 VO en m
t^
0 0 CN CO
r-
00
1
00
u VO vi p^
, VO co CO \ VO
, < ri 0 0 oi r-i 00 ro CN CN c o ,i
I m CN co VO
^
h
r- m
il
00
a\ CO 0 0 m 00
11 0 0 0 0
fO <N o t (S Os
m CN
OsOCN o r o VO
1 1 CN
0 0 CN VI OV i n <N r - m
CN Osc o
VO VO - (S
CD 0 0 00 K CN m m 1" < 3 11 r o VO^
er1 s " T ,
o " " CO*" v o " , CN v T CN " " " " v i " r n <s <3" o " 1_ v T
&
ta o
00 11 11
m
11 <N 1 1 1
"
vo 00 11 r -
CN m 00
os
0 0 VO CN 0 0 <s VO 0 0 VO 0 0 VO <N VO CN CO
Os
OsOs
CN
<3- o vi m CN <N VO CN ( I/-1 in f<> O m O
CN V i c n " , CN CN 1' VI r - t r1 > CN fO 11
-" I r - P-" r-T
i l >/->
fe s
<3- f 0 0 -<3- CO <3- CN V I - m VI VO 0 0 VO 11 CN m CN o
CO rl 0 0 1 ^ 1 1 1
S
os
VO
1"
CN < c<-> o
11
Ov
11 '
VO VO 0 0 Osv i <N co ON co CO 00 VO o >/-i O 0 0 ON CO 11 *r> 11

VI CN
P - ; 1 c o
1 c o

q -
co
1I VO

t^ <N 1 r <N
fN fN
CN
1 11
m 00
<N 1
Ti , '
P-"
(
00 53"
P-
>
-
1 en
"
CN 0 0
s 13
r- m vi

<N CN
CO VO CN 0 0 0 0
11 c o rt
00 00
m co
VO i n
CTv
VI
O
v> ^ <N

'l ^
3
VI gr i n
VO r -
00 \ ( <N >Si CN Os
OS \ O 0 0 VO
V
I
00 00
co

co m
'S" 00 "S- CN CN
2 s
if>
- m
<N
co
os
00 00
oC
Appendix 407
. co 00 <1- coNo O O ONV

COu-l CN 00 C 1on 0co0 r-;
oO CN NOOno0n 0OONr-
q n
00o
o a <s sr <
N 0CN
0 >n co >n
co
- < T co co oo m >/->
U >
N
COo0NO
uN-i
0C
C
NC
co
5 ONO
r- N
OON
f- O im
O1
1N
N
< co
O00 w
NO -l co
ON
O co (SN0O0Nm
NONO <N N
NO
O
O
n
oOO
oC noOon
11
<1TT
CO,- CNo"
u-T r-"
, ,o ri CNCNen
1
Onf^'cocN^ÔCNT
_.00, t-. ooinoifi^tooNMifiÔ^fi^N
CN Tf 1 m
o" - "t PI f o"vi
r'~ -
y-in"o
VO m
Ms
ta
o<ns rr <
o NO
3 <N
N <N
o
s
S V- C
ON l t-
es 1 r- ^cNr^cNôococ-iONÔNiôcN
mmO <N
t'ooo r- cn oo <s ^
<n

S * m <N
51
rt co
>} om
co ortt^ <N N
O CO
O SN
S <N

S ON O
f- nN
<N oO rCN r-
<N O NO
1
1 m
cn- r- no
co
on o>co >O
w-1 U-1o
o\
I oOo 0S0t
g . g?
s 0 feCL3s?g o
I T5 If-1 2 -2 1 I s s-
s1 5 .?; -S"'S -S
Notes
1. See the discussion in Section 5.2 on the specific type of unspecificity that is
involved in the semantic structure of shell nouns.
2. Abstracting from the complexities of speech events, the term speaker is used
as a synonym to text producer here and in this whole study. As such it refers
to the speakers of spoken utterances as well as to the writers of written texts.
Analogously, the term hearer will be used generically to refer to the recipi-
ents of spoken utterances and of written texts. The distinction between a text
producer, e.g. a journalist or editor in the case of a radio broadcast, and the
actual speaker will only be made if and when it is helpful and necessary.
3. Together with such linguists as Foley and van Valin (1985), Givn (1990:
515, fii. 1), Talmy (1991: 482-483) and others, I will assume that clauses
are the syntactic units that code mental propositions, and that propositions
in turn are mental representations of events or states of affairs. Furthermore
I assume (in line with e.g. van Dijk and Kintsch 1983: 37-41 and Givn
1990: 896) that clause/proposition/event-sized chunks of information are the
basic psychological units in the processing of coherent discourse.
4. For more information about this corpus, the so-called Bank of English at the
COBUILD project in Birmingham, from which the material for this study was
extracted, see Section 4.2.
5. Note that this piece of information consists of three complex propositions
and is therefore beyond what the human short-term memory (Miller 1956,
Chafe 1987: 22, 1994: 119), short-term working buffer (Kintsch and van
Dijk 1978: 368-370, van Dijk and Kintsch 1983: 349) or working memory
(Baddeley and Hitch 1974, Baddeley 1990) can handle.
6. Halliday and Hasan also mention that general nouns are superordinate
members of major lexical sets and that the combination of general noun plus
specific determiner is very similar to a reference item (1976: 275).
7. In her two later publications, Francis (1993: 149-150) and (1994: 90-93)
proposes very similar classifications.
8. As Ivani herself admits, the second motivation is a little misleading in that
what both Ivanic and I are concerned with are not attributing relational pro-
cesses, which have Carriers as subjects, but identifying relational processes,
and in these the role of the subject is called Identified by Halliday (1994:
122).
9. The notion of hypostatization, which refers to precisely this combination of
illusions (see e.g. Leisi (1975: 26-27), Lipka (1977, 1992: 16) and Conte
(1996: 4)) has not gained widespread currency so far. I have therefore not
410 Notes
used the term here, but will come back to it in the more detailed discussion
of cognitive functions of shell nouns in Section 17.2.
10. This investigation is based on a cognitive approach to the study of language
(see Section 2.3 and the discussion of a cognitive view of reference and
anaphora in Section 3.1.2). From this perspective, the notion of context is
defined as a cognitive representation of the interaction between concepts (see
Ungerer and Schmid 1996: 47-49). Whether the activation of such a repre-
sentation is triggered by linguistic or extra-linguistic stimuli is not a crucial
question from this perspective. I will therefore generally not distinguish
between the context (of situation) and the (linguistic) co-text, but use the
term context as a cover term. Wherever it is important to make or emphasize
the distinction, I will speak of linguistic context and situational context re-
spectively.
11. The concept of co-interpretation is used by Halliday and Hasan (1976: 314)
to explain what they see as the semantic relation underlying different types
of reference (in their framework). From a more modern point of view, the
nature of this notion can hardly be called semantic any more. As it presup-
poses an active human interpreter, it is better regarded as a pragmatic and
cognitive concept and it is precisely in such a way that it is understood in
this study. In this pragmatic-cognitive conception, the notion of co-
interpretation will prove very useful for the purposes of this study, since it is
general enough to be used as a cover term for the various grammatical and
semantic relations that exist between shell nouns and shell contents.
12. It is tempting to replace the fairly clumsy notion of lexico-grammatical
pattern by the more concise term collocation. However, having pondered
about the issue for a long time, three considerations have finally convinced
me that collocation is not a good term for what I have in mind here.
First, the patterns I am concerned with are more defined by grammatical
than by lexical co-occurrences. In Firth's original framework (see 1957;
1951: 194-203), they would therefore have to be treated as cases of colliga-
tion rather than collocation (see also Mcintosh 1961, Greenbaum 1970,
Mitchell 1971, Carter 1987: 56).
Second, as most linguistic terminology, collocation is used in many
different ways by scholars from different backgrounds (see also Herbst
1996). After Halliday's (1966) and Sinclair's (1966) attempts to exploit
collocation as an argument for the importance of the study of lexis, two con-
ceptions of collocation seem to prevail today: in corpus linguistics, it has
been used fairly loosely in the sense of a "tendency for words to occur to-
gether" (Sinclair 1991: 71). With little concern for semantic and grammati-
cal relations between the words involved, collocation is more or less turned
into a statistical measure of co-occurrence tendencies in this paradigm (see
e.g. Clear 1993). Language teachers and lexicographers, on the other hand,
Notes 411
have stressed the importance of such semantic relations and of the difference
between usual and restricted collocations (see Aisenstadt 1979, Cowie
1981: 223-230, Hausmann 1985: 118).
Third, many different nouns can occur in the four lexico-grammatical
patterns that I have introduced above, and the patterns themselves are vari-
able as well with regard to the forms of the verb be and the reference items
that can be used (this, that, the etc.). So the patterns correspond to types of
collocations rather than to collocations as such.
On all these grounds, I have decided to work with the term lexico-
grammatical pattern, or simply pattern here. I will use the term collocation,
however, when I am talking about actual 'con-locations' of specific words,
especially in the second part of this study.
13. Items creating comparative reference such as same, such and other are ne-
glected here because they are of minor importance compared to demonstra-
tive reference. Expressions such as another accusation also involve some
kind of identity, namely the presupposition that the items that are compared
belong to one superordinate class or at least share certain semantic dimen-
sions. If this is not the case, comparisons are impossible.
14. See for example Lyons (1977: 177-197) and Lipka (1992: 58-60).
15. In fact, it is generally doubtful whether anaphora refer only to those ele-
ments in the text which they can replace grammatically. As Brown and Yule
have for example shown (1983: 204-214), hearers must often activate much
more knowledge than only what is being referred to by the antecedent. More
criticism of the co-referential view of anaphora can be found in Garnham
(1987: 77-88, 149) and Rickheit and Strohner (1993: 225-226).
16. It is interesting to note especially for German-speaking linguists, that there
is at least one cognitively-oriented approach of reference that was even pro-
posed before Halliday and Hasan (1976), namely Kallmeyer et al.'s Theorie
der vermittelten Referenz 'theory of mediated reference' (1980: 213-226; 1st
ed. 1974). Kallmeyer et al. replace the idea of identity of reference for the
interpretation of anaphoric personal pronouns with the view that such pro-
nouns give instructions for a connection with the antecedent noun phrase,
which, and this is the major cognitive aspect of their theory, do not refer to
reality itself, but to a cognitive model of reality called Wirklichkeitsmodell
'model of reality' (1980: 23 et passim).
17. For philosophically-minded readers it will be obvious that such a cognitive
view is necessarily relativistic. Things and situations can be experienced in
totally different ways by different people when 'objective' truths no longer
count. See Lakoff (1987: 304-337) for a discussion of some important issues
and a defence of the relativist cognitive view.
18. See e.g. Fillmore (1975), Lyons (1977: 667-674), Conte (1981), Ehlich
(1982), Levinson (1983: 85-89), Krenn (1985: 37-70), Mey (1993: 95-99).
412 Notes
19. The term universe-of-discourse is for example used by Lyons (1977: 508,
1979: 94), the term textual world by Beaugrande and Dressier (1981: 84-90
et passim), discourse representation by Brown and Yule (1983: 206), mental
model e.g. by Johnson-Laird (1983), Garnham (1987: 88) and van Dijk
(1995: 394), situation model by van Dijk and Kintsch (1983: 336) and
Kintsch (1988, 1994), mental space by Fauconnier (1994, 1997), frame by
Minsky (1975: 212), Fillmore (e.g. 1975: 124, 1985: 223), and Talmy
(1996: 237), script by Schank and Abelson (1977: 42) and schema by Ru-
melhart and Ortony (1977) and Kintsch and van Dijk (1978: 373).
20. On apposition, and on the appositive relation between head nouns and post-
modifying clauses in particular, see Jespersen (1927: 27), Poutsma (1929:
619), Quirk et al. (1985: 1260-1263, 1271-1272, 1321), Meyer (1992: 21-
25), Francis (1993: 148-155).
21. To mention the most important milestones only, see e.g. Rosenbaum (1967),
Vendler (1968), Kiparsky and Kiparsky (1971), Karttunen (1971a, b), Men-
zel (1975), Bresnan (1979), Freed (1979), Ney (1981), Ransom (1986),
Wierzbicka (1988: 23-168), Givn (1990: 516-536), Mair (1990), Frajzyn-
gier and Jasperson (1991), Langacker (1991: 438-463).
22. See Rosenbaum (1967: 16-21), who still calls it "identity erasure transfor-
mation", as well as Kiparsky and Kiparsky (1971) and Menzel (1975: 22-
26).
23. See e.g. Biber and Finegan (1991: 204-211), Johansson (1991: 3), Leech
(1991: 12-15), Fillmore (1992: 38-39), Leech and Fligelstone (1992: 127-
138), Svartvik (1992b: 8-13), Clear (1993: 273-274), Svartvik (1996: 9-13)
Biber, Conrad and Reppen (1998: 1-12), Kennedy (1998: 5-12, 268-294).
24. See Langacker (1987a: 59-60, 1991: 159-163) and Geeraerts et al. (1994: 10
et passim) for the notion of entrenchment.
25. See Section 4.3 on intrusions of non-shell nouns into these sections of the
data.
26. The threshold varies between two and five occurrences. This is in line, for
example, with the approach of Clear (1993: 277), who regards three occur-
rences as the bottom threshold for the significance and relevance of pairs of
words.
27. I would like to thank Ramesh Krishnamurthy, Corpus Manager at
COBUILD'S Bank of English, for providing me with this data. In all examples
in this study that are taken from this corpus, the subcorpus is indicated by
the abbreviations in small capitals used in this list, e.g. SPOKEN, BBC, MAGS.
28. The preponderance of media sources in the COBUILD has of course also been
noticed by competing corpus compilers and especially corpus-oriented lexi-
cographers, see e.g. Summers (1996: 266, note 4).
29. What is at stake here is the balance between the precision and the recall of a
retrieval system (cf. Clear 1993: 275). A system, or a query statement for
Notes 413
that matter, exhibits precision to the extent that it only yields items that are
of interest to the researcher. The recall of a query statement is high when it
manages to retrieve all items which are interesting for the researcher.
30. I would like to thank Tim Lane for letting me use this program at such an
early stage of its development that even members of the COBUILD staff only
became aware of it through my results.
31. In the case of query strings, the lexical item with the lowest frequency will
be selected as the node by the search program.
32. The operational criterion for the inclusion of nouns in the material was their
occurrence among the most frequent 1,000 items either for the 225m-word
corpus or for one of the subcorpora. This means that there is no rigid quan-
titative bottom threshold but a de facto threshold between roughly 2 and 5.
Since the frequencies are not based on actual counts but on the lists of the
collocate program, minor deviations from the actual frequencies for the
nouns with very low frequencies as collocates are possible. These vagaries at
the bottom end of the frequency scale should not be problematic since it is
the top end that I am chiefly interested in (see the rationale put forward in
Section 4.1).
33. See Tuggy (1996) on the construction the thing is is that, and my further
discussion in Sections 7.1 and 16.1.
34. I am not going to rely on the much more elaborate measures of T-score and
Mutual Information (see Church and Hanks 1990), which are used at
COBUILD (Clear 1993) and also in the British National Corpus, because
these statistics are so technical that most linguists cannot see behind the
formulas. Indeed, even people who work with these scores have told me that
the actual linguistic significance of the two notions has remained somewhat
opaque to them (cf. also Stubbs 1995 for a critical discussion). A further rea-
son for coming up with statistical measures which are specifically tailored to
shell-content complexes is that the nouns always occur in a perfectly pre-
dictable slot in these patterns. In contrast, T-score and Mutual Information
are calculated for all items within a defined span, without taking their rela-
tion to the node into consideration. This procedure, which may be useful and
necessary if one is concerned with co-occurrence tendencies in general, un-
necessarily blurs things that are perfectly clear in the case of shell-content
complexes.
35. It should be noted that the three versions of Lyons' system are not com-
pletely identical. My account is based on the classic 1977 version.
36. A much more fine-grained typology of abstract entities is provided by Asher
(1993: 15-62). However, both his general approach (which is both conceptu-
alist and formalist) and his aims (to analyse reference to abstract entities in
the framework of Discourse Representation Theory; see Kamp and Reyle
414 Notes
1993) are so different from the stance taken here, that a detailed account of
his approach would presumably lead to confusion rather than new insights.
37. Unfortunately, the term utterance itself suffers from the same indetermi-
nacy. It is therefore important to remember that it was defined as a technical
term denoting a third-order entity or abstract relation.
38. It will also emerge in Section 17.2 that the notion of grammatical metaphor
may have some justification after all. The notion captures the apparently
paradoxical fact that in spite of their stylistic abstractness, shell nouns share
with lexical metaphors the function of concretizing abstract entities (see
Schmid 1999).
39. In this context, it is worth drawing attention to the fact that many dictionary
definitions of shell nouns are circular, making use of other similar shell
nouns, in this case chance and occasion.
40. Hoey (1994) provides a detailed corpus-based analysis of the noun reason as
a common signal in discourse". See Section 7.3 for further details on the use
of this noun.
41. My notion of structure-inherent semantic gaps has of course nothing to do
with the notion of "gapping" used in generative grammar for the ellipsis of
clause elements (see e.g. Radford 1988: 98 et passim).
42. For discussions of the status of semantic features and other related aspects
see Lipka (1979, 1992: 98-116), Sprengel (1980) and Schneider (1988: 40-
67).
43. The reason why the mundane conception of the semantic notion of factuality
held here is so much emphasized is that the notion of fact plays an impor-
tant role as a rigidly defined concept in other frameworks (see e.g. Vendler
1967 and 1980, Kiparsky and Kiparsky 1971 and Kartunnen 1971a).
44. An overview of the development of the notion of frame - starting out from
its use in Artificial Intelligence (e.g. Minsky 1975) and the mainly syntactic
conception in Fillmore's earlier work on frames (e.g. 1977) to a cognitive
conception, as proposed by Fillmore today (Fillmore and Atkins 1992) and
most notably by Talmy (1996) - is given in Ungerer and Schmid (1996: ch.
5).
45. As stylistic variants of the notion of windowing, I will use the expressions of
focusing one's attention, of highlighting components of frames and of pro-
filing components of frames as figure, which are based on metaphors from
the source domains of visual perception, the lighting of stages and Gestalt
psychology respectively.
46. To mark the pre-linguistic, cognitive status of the components of frames, I
will henceforth use small capitals for their names, e.g. CAUSE, EFFECT, STATE
OF AFFAIRS, CHARACTERIZATION, BELIEF, SIGN. This should not lead to con-
fusion with the references to the corpus section because there is no overlap
in their functions.
Notes 415
47. Tuggy confines his attention to American English. It is not clear whether
this is because this is his native variety or because he assumes that the pat-
tern the thing is is (that) is not common in other varieties of English. My
corpus data suggest in any case that the collocation is used in spoken British
English as well, and not even only in colloquial style.
48. In a sample of 1000 concordance lines for the noun fact, as many as 475
lines contained the fixed phrase in fact. If one diminishes the score for the
overall frequency in the corpus proportionately, the compiled reliance value
for fact rises to 71.93%.
49. See Vendler (1980) and especially Francis (1986: 19-26, 92-95) on the fac-
tivity of nouns.
50. It must be emphasized, however, that the reference to the future in the that-
clause in (7.5.b) is expressed by means of be going to + infinitive, which
evokes a strong connection to the present, either to the fulfillment of present
intentions or the results of present causes (Quirk et al. 1985: 214-215). It is
presumably the reliance on present evidence that sanctions the combination
of fact with going to. Whether a less present-oriented means of expressing
reference to the future, for example the w/7/-future, would be acceptable in a
f/jai-clause with a personal subject postmodifying the noun fact is doubtful.
51. Note that the total number of occurrences of the two nouns as given in Fig-
ure 7.3 is not identical with the one given in the text above because both
nouns have also been found in a small number of occurrences with to-
clauses which have been neglected in the figure.
52. The part-whole relation thus seems to be cognitively more basic than the
relation underlying attributive uses of shell nouns. The latter are neverthe-
less discussed before partitive shell nouns because they are better examples
of shell nouns on the basis of their overall frequency as shell nouns and their
frequency of occurrence in the pattern -be-that.
53. This claim is based on Quirk et al.'s remarks concerning the "logical" use of
the definite article (1985: 270).
54. See Ungerer and Schmid (1996: 120-130), Schmid (1997: 111-114) and
(1999) for the conception of metaphor used here.
55. The noun downside is based on a different metaphor which is less easy to
explain. Presumably, the general metaphorical schema "good is up" and
"bad is down" (Lakoff and Johnson 1980: 16), which is also at work in ex-
pressions like things are looking up or things are at an all time low, plays a
role here. According to the OED (s.v. handicap), the origin of the noun
handicap is obscure but probably related to horse-racing contexts.
56. This explanation is reminiscent of the notion of blocking used in word-
formation for the phenomenon that a possible complex lexeme is not used
because of the existence of a simple form (see e.g. Bauer 1983: 87-88).
416 Notes
57. On the notion of reporting from a linguistic perspective see Francis (1986:
77-81), Winter (1982: 197-200), Leech (1983: 207-216), Halliday (1994:
252-257) and Tadros (1994: 75-77).
58. That a complete correspondence between a prior utterance and the report
must not be taken for granted even in the case of self-reports has been shown
by work in forensic linguistics; see e.g. Coulthart (1992: 245).
59. For now, I have abstracted from the fact that ORIGINAL UTTERANCES have
contents, i.e. that they are about events, states of affairs and so on. I will
delve into these intricacies when I start looking at groups and families in
Section 8.2.
60. Note that in American English the spelling is prophesy both for the noun
and the verb.
61. Besides Searle (1976), see e.g. Bach and Harnish (1979), Allan (1986: 190-
204), Reiss (1985: 26-68). Whether the efforts of these linguists to keep
speech act verbs and speech acts apart have actually been successful is a
different matter.
62. It is doubtful whether there are actually nouns, or for that matter verbs, that
have the meaning 'making privately and obliquely known' given by
Leech. Even Leech's own examples intimate, imply and hint (1983: 223) do
by no means always have this quality.
63. See Wierzbicka (1987: 37-48) on the meaning and use of some morphologi-
cally related verbs.
64. See also Wierzbicka (1987:188) on the feature of offered cooperation in-
cluded in the veib propose.
65. Note that Leech's term for this group is "attitudinal". I have replaced his
term because the function of characterizing psychological attitudes is shared
by all mental shell nouns, and because what is at stake here are not so much
attitudes (in the sense of longstanding ideas and beliefs), but emotions.
66. See e.g. Frege (1976; 1918), Geach (1957), Armstrong (1973), Bogdan
(1982).
67. The notion of strength of commitment is borrowed from Lyons' discussion of
epistemic modality. There can of course be no doubt that this type of modal-
ity plays a role in all creditive shell nouns. For reasons explained in Section
10.1 the issue of modality is backgrounded in this chapter on mental shell-
noun uses.
68. Because of the frequency of the fixed expression the bottom line is that, the
-be-that collocations of line are not included in this score.
69. In this context, other means of identifying EXPERIENCER, some of which
again take us to the periphery of the class of shell nouns, should briefly be
mentioned. The nouns view and opinion, for example, occur frequently in
the fixed expressions someone is of the view/opinion that, which, like have
an idea and similar phrases mentioned earlier, function as expanded predi-
Notes 417
cates involving nouns. Further borderline case are take the view that, hold
the view that and express the view that.
70. See von Wright (1951), Rescher (1986), Lyons (1977: 787-849), Coates
(1983), Palmer (1986 and 1990), Sweetser (1990: 49-75) and Mindt (1995)
for a range from logical, semantic, syntactico-semantic and cognitive-
linguistic to pedagogical approaches.
71. Of course the notion of dynamic modality is not uncontroversial, mainly
because it does not seem to be on a par with epistemic and deontic modality
(see e.g. Ungerer's (1990) review of another of Palmer's books, Palmer
(1987), where Palmer also works with dynamic modality). Interestingly, dy-
namic modality is explicitly excluded from Palmer (1986:12).
72. Palmer himself is fairly vague about the third degree of modality; he does
not provide a name for it, not even in his matrix representation of types and
degrees provided on his page 37.
73. Note that this conception is more similar to that given by Palmer (1990: 37)
for "neutral modality" than for "dynamic modality". The problem with the
transfer of Palmer's system, however, is that dynamic modality is defined in
contradistinction to deontic modality by means of two criteria that are not
very helpful here. First, dynamic modality is claimed to be "subject-
oriented" (Palmer 1990: 36), in that it has to do with ability and volition.
And second, unlike uses of deontic modal verbs, dynamic uses are not used
performatively (Palmer 1990: 83). These problems would suggest that I use
the term 'neutral' instead of 'dynamic modality', but 'neutral modality' does
not encompass ability, and apart from that, it is too vague a term for a fairly
specific phenomenon.
74. Palmer (1990: 85, 136-137) introduces the notion of 'power' to account for
such uses.
75. Wierzbicka (1988: 27-59) makes similar claims. In contrast, Mair (1990:
84-92) argues that there is no modality inherent in infinitive clauses. It must
be added, however, that at this point of his study, Mair is not concerned with
postmodifying clauses but with infinitives and gerunds functioning as clause
subjects.
76. It should of course be mentioned that the COBUILD guide by Chalker (1996)
is based partly on the same material as my study.
77. Of course, one could also argue that such uses as example (12.1) must be
excluded because the w/jere-clause is a relative clause that can be para-
phrased by in which.... This issue will be discussed below in Section 12.3.
78. One may wonder in passing where this indeterminacy between eventive and
circumstantial uses comes from in the first place. It is certainly no coinci-
dence that the shell contents of these uses are all STATES rather than dy-
namic events. Given this finding, it is only a small step to the claim that de-
scriptions of real-world states are not very different from descriptions of the
418 Notes
circumstances of states. After all, what are states if not configurations of

certain circumstances? Although from a linguistic point of view, state de-
scriptions also consist of subjects representing persons, things and the like,
and of predicates expressing the states in which they are, it is clear from a
cognitive, especially a perceptual, point of view that both the agents and the
whole states are much less salient than in dynamic events.
79. A second possible interpretation of these wA-clauses, as adverbials, is dis-
cussed further down with respect to temporal uses.
80. See Brown and Yule (1983: 182-189) for a concise critical summary of
Prince's (1981) main points.
81. A largely compatible cognitive view of the contribution of old and new in-
formation to the coherence of discourse is given by Givn (1990: 893-944).
It is neglected here mainly because of its irritating dependence on the-mind-
is-a-computer metaphor. Other, equally finely-grained approaches are
Ariel's Accessability Theory (1990, 1996) and Gundel et al.'s Givenness Hi-
erarchy (Gundel, Hedberg and Zacharski 1993, Gundel 1996). The relevance
of these approaches for the study of shell nouns will be looked at in more
detail in Chapter 15.2.
82. Cf. also Allerton (1978: 147) and Krenn (1985: 27), as well as Fiibas'
(1964) notion of communicative dynamism on the idea of a scale or cline
from given to new information.
83. See Chomsky's (1971: 199-214) use of the term "presupposition" for theme,
and Lambrechts (1994: 52-53) useful discussion of types of presupposition.
84. The definition given for topic here, and my use of the notion in Sections
14.1. to 15.2, is sometimes called sentential topic (e.g. Brown and Yule
1983: 70). In Section 15.3,1 will work with a second conception of the term
topic, called discourse topic (Brown and Yule 1983: 71-83).
85. See e.g. Chomsky (1971: 200-214), Sgall et al. (1986: 207-216), Gundel
(1988: 13-47).
86. See the discussions of examples (5.10) and (5.11) in Section 5.1.2, example
(7.5) in 7.2, example (7.14) in 7.3, examples (8.12) and (8.13) in 8.2, exam-
ple (9.6) in 9.2, examples (10.8) and (10.14) in 10.2, example (10.19) in
10.3, and (10.31) in 10.4.
87. This is another systematic 'advantage' of using shell nouns. When utterance
and thought reports are not introduced by verbs but by nouns, the verbs can
be used to express additional information which must be expressed in a
separate clause when the verb serves as peg for the report.
88. That units larger than phrases can function as focus was already pointed out
in response to Chomsky by Sgall et al. (1973: 164): "The focus does not
necessarily coincide with a single phrase, but an assumption readily occurs
that, in general, the focus consists in a sequence of elements ranging from a
certain point (chosen among those determined by the range of permissible
Notes 419
focus [sic!]) along the scale to the lexical unit carrying the intonation cen-
tre".
89. It should perhaps be mentioned that many native speakers would of course
intuitively deny the existence of the double- onstruction; see Tuggy (1996:
714) for corroborative evidence.
90. The most recent cognitive anaphora theory by van Hoek (1997), proposed
within the framework of Langacker's Cognitive Grammar (1987a, 1991), is
only concerned with intra-sentential anaphora and therefore not useful for
my purposes.
91. Another image that is closely related to signposting is signalling, as used by
Hoey (1979). Note that my conception of the signposting function is broader
than that of Kurzon (1985), who is only interested in items with text deixis
like this section or the above figure.
92. For attempts to define the notion of discourse topic and to distinguish it
from sentence topic, see Keenan and SchiefFelin (1976), van Dijk (1977:
119, 131-142), Brown and Yule (1983: 71-83), Givn (1983: 9-15) and van
Dijk and Kintsch (1983: 155-156, 169).
93. See Hinds (1979: 145-150), Longacre (1979), Brown and Yule (1983: 95-
100) for approaches to definitions of the notion of paragraph.
94. See Francis (1986: 27-45; 1994: 86) and Conte (1996: 5-7) on the text-
organizing function of encapsulating nouns, as well as Koeppel (1993 : 200-
201, 220-225, 233-234, 239-240) on Textgliederungssignale 'text segment-
ing signals' in various types of German texts.
95. For more recent views of the notion of coherence, whose significance I have
taken for granted in this study, see Charolles and Ehrlich (1991), Ehrlich
and Charolles (1991), Le Ny (1991), and the contributions in Gernsbacher
and Givn (1995a), especially Gernsbacher and Givn (1995b), Givn
(1995), Sanford and Moxey (1995) and Trabasso et al. (1995). The general
trend in these publications, which I strongly support, is to define coherence
as a mental entity (Givn 1995) or, even better, a mental state. Trabasso et
al. (1995: 190), for example, define coherence as "a subjective state of mind
that results from generating or evaluating ideas in relation to one another".
96. Useful sources for the notion of gestalt from the psychological and the lin-
guistic perspective are Koflka (1935), Boring (1950: 586-613), Wertheimer
(1958; 1923), Lakoff (1977: esp. 246-247) and Ungerer and Schmid (1996:
33-37).
97. As suggested in Section 5.2, even these synchronically unanalyzable nouns
can in fact be traced back historically to verbs. This supports Porzig's (1930)
idea that the reifying potential of abstract nouns should be explained by re-
lating them to full clauses.
98. See Leisi (1975: 25-27) and his notes on page 26, as well as Lipka (1977;
1992: 16) and Conte (1996: 4).
420 Notes
99. In addition to the models mentioned explicitly, see the contributions to

Britton and Graesser (1996a), in particular the surveys of recent models by
Graesser et al. (1996) and the synopsis by Graesser and Britton (1996).
100. Those readers who are familar with Mental Space theoiy will be reminded of
Fauconnier's notion of blending at this point (see Fauconnier and Turner
1996, Fauconnier 1997: 149-186). It must be emphasized, however, that
when Fauconnier talks about blends, there are always several mental spaces
involved, while my idea of integration (a term that is used by Fauconnier,
too; see 1997: 185-186) concerns the relation between the space builder and
the internal configuration of a single mental space.
References
Aijmer, Karin
1997 I think - an English modal particle. In: Toril Swan and Olaf Jansen
Westvik, eds., Modality in Germanic languages. Historical and
comparative perspectives, Berlin - New York: Mouton de Gruyter,
1-47.
Aijmer, Karin and Bengt Altenberg, eds.
1991 English corpus linguistics. Studies in honour of Jan Svartvik, Lon-
don - New York: Longman.
Aisenstadt, Ester
1979 Collocability restrictions in dictionaries. In: Robert R.K. Hartmann,
ed., Dictionaries and their users, Papers from the 1978 BAAL
Seminar on Lexicography, Exeter Linguistic Studies vol.1, no. 4,
71-74.
Algeo, John
1993 Having a look at the expanded predicate. In: Bas Aarts and Charles
F. Meyer, eds., The verb in contemporary English: theory and de-
scription, Cambridge: Cambridge University Press, 203-217.
Allan, Keith
1986 Linguistic meaning, 2 vols., London - New York: Routledge and
Kegan Paul.
Allerton, David J.
1978 The notion of 'givenness' and its relations to presupposition and to
theme. Lingua 44, 133-168.
Ariel, Mira
1990 Accessing noun phrase antecedents, London: Routledge.
1996 Referring expressions and the +/- coreference distinction. In: Fro-
theim and Gundel (1996), 13-36.
Armstrong, David M.
1973 Belief, truth and knowledge, London: Cambridge University Press.
Asher, Nicholas
1993 Reference to abstract objects in discourse, Dordrecht etc.: Kluwer.
Austin, John L.
1962 How to do things with words, Oxford: Clarendon.
422 References
Bach, Emon
1981 On time, tense, and aspect: An essay in English metaphysics. In:
Peter Cole, ed., Radical pragmatics, New York: Academic Press,
62-81.
Bach, Kent and Robert M. Harnish
1979 Linguistic communication and speech acts, Cambridge/Ma. - Lon-
don: Press.
Baddeley, Alan D.
1990 Human memory: Theory and practice, Hove, UK: Lawrence Erl-
baum Associates.
Baddeley, Alan D. and Graham J. Hitch
1974 Working memory. In: Gordon H. Bower, ed., Recent advances in
learning and motivation, Vol. Ill, New York: Academic Press, 47-
90.
Baker, Mona, Gill Francis and Elena Tognini-Bonelli, eds.
1993 Text and technology. In honour of John Sinclair, Philadelphia -
Amsterdam: John Benjamins.
Ballmer, Thomas and Waltraud Brennenstuhl
1981 Speech act classification. A study in the lexical analysis of English
speech activity verbs, Berlin etc.: Springer-Verlag.
Bauer, Laurie
1983 English word-formation, Cambridge: Cambridge University Press.
Bazell, C.E, J.C. Catford, M.A.K. Halliday and R.H. Robins, eds.
1966 In memory ofJ.R. Firth, London: Longman.
de Beaugrande, Robert and Wolfgang Dressier
1981 Introduction to text linguistics, London: Longman.
Bennet, Jonathan
1988 Events and their names, Oxford: Oxford University Press.
Biber, Douglas and Edward Finegan
1991 On the exploitation of computerized corpora in variation studies.
In: Aijmer and Altenberg (1991), 204-253.
Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad and Edward
Finegan
1999 Longman grammar of spoken and written English, London etc.:
Longman.
Bogdan, Radu J., ed.
1982 Belief: Form, content and function, Oxford: Oxford University
Press.
Bolinger, Dwight
1968 Entailment and the meaning of structures. Glossa 2, 119-127.
References 423
1977 Pronouns and repeated nouns, Indiana University Linguistics Club:

Indiana.
Boring, Edwin G.
1950 A history of experimental psychology, 2nd ed., New York:
Appleton Century Crofts.
Bransford, John D. and Jeffrey J. Franks
1972 The abstraction of linguistic ideas: A review. Cognition 1: 211-249.
Bresnan, Joan W.
1979 Theory of complementation in English syntax, New York - London:
Garland.
Britton, Bruce K.
1994 Understanding expository text: building mental structures to induce
insights. In: Gernsbacher (1994), 641-674.
Britton, Bruce . and Arthur C. Graesser, eds.
1996 Models of understanding text, Mahwah/N.J.: Lawrence Erlbaum
Associates.
Brown, Gillian and George Yule
1983 Discourse analysis, Cambridge: Cambridge University Press.
Brown, Penelope and Stephen Levinson
1987 Politeness. Some universals in language usage, Cambridge: Cam-
bridge University Press.
Bybee, Joan
1985 Morphology. A study of the relation between meaning and form,
Amsterdam - Philadelphia: John Benjamins.
Bybee, Joan and Suzanne Fleischman, eds.
1995 Modality in grammar and discourse, Amsterdam - Philadelphia:
John Benjamins.
Carter, Ronald
1987 Vocabulary, London etc. : Allen and Unwin.
Chafe, Wallace
1976 Givenness, contrastiveness, defniteness, subjects, topics and point
of view. In: Li (1976), 25-55.
Chafe, Wallace, ed.
1980a The pear stories. Cognitive, cultural and linguistic aspects of nar-
rative production, Norwood/N.J.: Ablex.
Chafe, Wallace
1980b The deployment of consciousness in the production of a narrative.
In: Chafe (1980a), 9-50.
1987 Cognitive constraints on information flow. In: Tomlin (1987), 21-
51.
424 References
1994 Discourse, consciousness, and time. The flow and displacement of

conscious experience in speaking and writing, Chicago - London:
The University of Chicago Press.
1996 Inferring identifiability and accessibility. In: Fretheim and Gundel
(1996), 37-46.
Chalker, Sylvia
1996 Collins COBUILD English Guides 9: Linking Words, London:
HarperCollins.
Charolles, Michel and Marie-France Ehrlich
1991 Aspects of textual continuity. Linguistic approaches. In: Denhire
and Rossi (1991), 251-267.
Chomsky, Noam
1970 Remarks on nominalization. In: Roderick A. Jacobs and Peter S.
Rosenbaum, eds., Readings in transformational grammar,
Waltham/Mass. etc.: Ginn and Company, 184-221.
1971 Deep structure, surface structure, and semantic interpretation. In:
Danny D. Steinberg and Leon A. Jakobovits, eds., Semantics. An
interdisciplinary reader in philosophy, linguistics and psychology,
Cambridge: Cambridge University Press, 193-216.
Church, Kenneth W. and Patrick Hanks
1990 Word association norms, mutual information and lexicography.
Computational Linguistics 16.1, 22-29.
Clancy, Patricia M.
1980 Referential choice in English and Japanese narrative discourse. In:
Chafe (1980a), 127-202.
Clark, Herbert H. and Susan E. Haviland
1977 Comprehension and the given-new contract. In: Roy O. Freedle,
ed., Discourse production and comprehension, Norwood/N.J.:
Ablex, 1-40.
Clear, Jeremy
1993 From Firth Principles. Computational tools for the study of collo-
cation. In: Baker, Francis and Tognini-Bonelli (1993), 271-292.
Coates, Jennifer
1983 The semantics of the modal auxiliaries, London - Canberra: Croom
Helm.
Coleman, Linda and Paul Kay
1981 Prototype semantics: the English word LIE. Language 57, 26-44.
Conte, Maria-Elisabeth
1981 Textdeixis und Anapher. Kodikas/Code 3, 121-132.
References 425
1996 Anaphoric encapsulation. In: Walter de Mulder and Liliane Tas-

mowski, eds., Coherence and Anaphora. Belgian Journal of Lin-
guistics 10, 1-10.
Coulthard, Malcom
1992 Forensic discourse analysis. In: Malcolm Coulthard, ed., Advances
in spoken discourse analysis, London - New York: Routledge.
Coulthard, Malcolm, ed.
1994 Advances in written text analysis, London: Routledge.
Cowie, Anthony P.
1981 The treatment of collocations and idioms in learner dictionaries.
Applied Linguistics, 2 (3), 223-235.
Crimmins, Mark
1992 Talk about belief, Cambridge/Ma. - London: MIT Press.
Cruse, D. Allan
1986 Lexical semantics, Cambridge: Cambridge University Press.
Ciystal, David
1967 Word classes in English. Lingua 17, 24-56.
DaneS, FrantiSek
1976 Zur semantischen und thematischen Struktur des Kommunikats.
In: FrantiSek Danes and Dieter Viehweger, eds., Probleme der
Textgrammatik, Berlin, 29-40.
Denhire, Guy and Jean-Pierre Rossi, eds.
1991 Text and text processing, Amsterdam etc. : North Holland.
van Dijk, Teun
1977 Text and context, London: Longman.
1981 Studies in the pragmatics of discourse, The Hague etc. : Mouton.
1988 News as discourse, Hillsdale/N.J.: Lawrence Erlbaum Associates.
1995 On macrostructures, mental models, and other inventions: A brief
personal history of the Kintsch-van Dijk Theory. In: Weaver et al.
(1995), 383-410.
van Dijk, Teun and Walter Kintsch
1983 Strategies of discourse comprehension, New York: Academic
Press.
Dik, Simon C.
1980 Studies in Functional Grammar, London etc.: Academic Press.
1989 The theory of Functional Grammar. Part I: The structure of the
clause, Dordrecht/Holland - Providence/RI, USA: Foris.
Dinsmore, John
1991 Partitioned representations. A study in mental representation, lan-
guage understanding and linguistic structure, Dordrecht: Kluwer.
426 References
Dirven, Ren
1991 Schema and subschemata in the lexical structure of the vert agree.
Cahiers de l'Institut de Linguistique de Louvain 17, 25-42.
Dirven, Ren and Gnther Radden
1977 Semantische Syntax des Englischen, Frankfurt/Main: Athenaion.
Du Bois, John W.
1980 Beyond definiteness: The trace of identity in discourse. In: Chafe
(1980a), 203-274.
Ehlich, Konrad
1982 Anaphora and deixis: same, similar, or different? In: Robert J. Jar-
vella and Wolfgang Klein, eds., Speech, place and action. Studies
in deixis and related topics, Chichester etc.: Wiley, 315-338.
Ehrlich, Marie-France and Michel Charolles
1991 Aspects of textual continuity. Psycholinguistic approaches. In:
Denhire and Rossi (1991), 269-285.
Falkner, Wolfgang
1997 Verstehen, Miverstehen und Miverstndnisse. Untersuchungen
an einem Korpus englischer und deutscher Beispiele, Tbingen:
Niemeyer.
Fauconnier, Gilles
1994 Mental spaces, Cambridge: Cambridge University Press.
1997 Cognitive mappings in language and thought, Cambridge: Cam-
Fauconnier, Gilles and Mark Turner
1996 Blending as a central process of grammar. In: Adele E. Goldberg,
ed., Conceptual structure, discourse and language, Stanford: Cen-
ter for the Study of Language and Information, 113-130.
Fillmore, Charles J.
1968 The case for case. In: Emon Bach and Robert T. Harms, eds., Uni-
versals in lingustic theory, London etc.: Holt, Rinehart and
Winston, 1-88.
1975 Santa Cruz lectures on deixis, Bloomington/Indiana: Indiana Uni-
versity Linguistics Club.
1977 The case for case reopened. In: Peter Cole and Jerry M. Sadock,
eds., Syntax and semantics, Vol. 8: Grammatical relations, New
York etc.: Academic Press, 59-81.
1985 Frames and the semantics of understanding. Quaderni di
Semantica VI, 222-254.
1992 'Corpus linguistics' or 'computer-aided armchair linguistics'. In:
Svartik (1992a), 35-60.
References 427
Fillmore, Charles, Paul Kay and Peter O'Connor

1988 Regularity and idiomaticity in grammatical constructions. The case
of LET ALONE. Language 64, 501-538.
Fillmore, Charles J. and Beryl T. Atkins
1992 Toward a frame-based lexicon: The semantics of RISK and its
neighbors. In: Adrienne Lehrer and Eva Kittay, eds., Frames,
fields, and contrasts, Hillsdale/NJ.: Lawrence Erlbaum Associates,
75-102.
Firbas, Jan
1964 On defining theme in functional sentence analysis. Travaux lin-
guistique de Prague 1, 267-280.
Firth, J.R.
1957; 1951 Modes of meaning. In: J.R. Firth, Papers in linguistics 1934-1951,
London etc.: Oxford University Press, 190-215. [Originally pub-
lished in Essays and Studies, The English Association, 1951],
Fletcher, Charles R.
1981 Short-term memory processes in text comprehension. Journal of
verbal learning and verbal behavior 20, 546-574.
Foley, William A. and Robert D. van Valin
1985 Information packaging in the clause. In: Timothy Shopen, ed.,
Language typology and syntactic description, Vol. I Clause struc-
ture, Cambridge: Cambridge University Press.
Frajzyngier, Zygmunt and Robert Jasperson
1991 -clauses and other complements. Lingua 83, 133-153.
Francis, Gill
1986 Anaphoric nouns, Discourse Analysis Monographs 11, Birming-
ham: University of Birmingham Printing Section.
1993 A corpus-driven approach to grammar. In: Baker, Francis and
Tognini-Bonelli (1993), 137-156.
1994 Labelling discourse: an aspect of nominal-group lexical cohesion.
In: Coulthard (1994), 83-101.
Freed, Alice F.
1979 The semantics of English aspectual complementation, Dordrecht
etc.: D. Reidel.
Frege, Gottlob
1976; 1918 Der Gedanke. In Frege, Gottlob (1976), Logische Untersuchungen,
2nd ed., ed. by Gnther Patzig, Gttingen: Vandenhoek and Ru-
precht, 30-53. [Originally published in Beitrge zur Philosophie
des Deutschen Idealismus I, 2, 1918.]
1994; 1891 Funktion und Begriff. In: Frege, Gottlob (1994) Funktion, Begriff,
Bedeutung, 7th ed., ed. by Gnther Patzig, Gttingen: Vandenhoek
428 References
and Ruprecht, 18-39. [Originally presented as a paper to the Je-

naische Gesellschaft fr Medizin und Naturwissenschaft,
9.1.1891.]
Fretheim, Thorstein and Jeannette K. Gundel, eds.
1996 Reference and referent accessability, Amsterdam - Philadelphia:
John Benjamins.
Garnham, Alan
1987 Mental models as representation of discourse and text, Chichester:
Ellis Horwood.
1989 Integrating information in text comprehension: the interpretation of
anaphoric noun phrases. In: Greg N. Carlson and Michael K. Ta-
nenhaus, eds., Linguistic structure in language processing, Dor-
drecht: Kluwer, 359-399.
Geach, Peter
1957 Mental acts, London: Routledge and Kegan Paul.
Geeraerts, Dirk, Stef Grondelaers and Peter Bakema
1994 The structure of lexical variation. A descriptive framework for
cognitive lexicology, Berlin etc. : Mouton de Gruyter.
Gernsbacher, Morton ., ed.
1994 Handbook of psycholinguistics, San Diego etc.: Academic Press.
Gernsbacher, Morton A. and Talmy Givn, eds.
1995a Coherence in spontaneous text, Amsterdam - Philadelphia: John
Benjamins.
Gernsbacher, Morton A. and Talmy Givn
1995b Introduction. Coherence as a mental entity. In: Gerasbacher and
Givn (1995a), V-X.
Givn, Talmy, ed.
1979 Syntax and semantics. Volume 12: Discourse and syntax, New
York: Academic Press.
Givn, Talmy
1983 Topic continuity in discourse: An introduction. In: Givn, Talmy,
ed., Topic continuity in discourse: A quantitative cross-language
study, Amsterdam: John Benjamins, 1-41.
1985 The pragmatics of referentiality. In: Deborah Schifirin, ed., Mean-
ing, form and use in context, Washington/DC: Georgetown Univer-
sity Press.
1990 Syntax: A functional-typological introduction, Vol. 2, Amsterdam:
John Benjamins.
1995 Coherence in text vs. coherence in mind. In: Gernsbacher and Gi-
vn (1995a), 59-115.
References 429
Goldberg, Adele
1995 Constructions: A Construction Grammar approach to argument
structure constructions, Chicago/IL: University of Chicago Press.
Graesser, Arthur and Bruce K. Britton
1996 Five metaphors for text understanding. In: Britton and Graesser
(1996), 341-351.
Graesser, Arthur C., Shane S. Swamer, William B. Baggett and Marie A. Sell
1996 New models of deep comprehension. In: Britton and Graesser
(1996), 1-32.
Greenbaum, Sidney
1970 Verb-intensifier collocations in English: An experimental ap-
proach, The Hague: Mouton.
Grosz, Barbara J.
1981 Focusing and description in natural language dialogues. In:
Aravind K. Joshi et al., eds., Elements of discourse understanding,
Cambridge etc.: Cambridge University Press, 84-105.
Gundel, Jeanette .
1988 The role of topic and comment in linguistic theory, New York -
London: Garland.
1996 Relevance theory meets the Givenness Hierarchy. An account of
inferrables. In: Fretheim and Gundel (1996), 141-154.
Gundel, Jeannette ., Nancy Hedberg and Ron Zacharski
1993 Cognitive status and the form of referring expressions in discourse.
Language 69, 274-307.
Haberlandt, Karl and Arthur C. Graesser
1990 Integration and buffering of new information. In: Arthur C.
Graesser and Gordon H. Bower, eds., Inferences and text compre-
hension. (The psychology of learning and motivation 25) San Di-
ego/CA.: Academic Press, 71-87.
Halliday, Michael A.K.
1966 Lexis as a linguistic level. In: Bazell et al. (1966), 148-162.
1967a Notes on transitivity and theme in English. Part 1. Journal of lin-
guistics 3, 37-81.
1967b Notes on transitivity and theme in English. Part 2. Journal of lin-
guistics 3, 177-274.
1993 Quantitative studies and probabilities in grammar. In: Hoey (1993),
1-25.
1994 An introduction to functional grammar, 2nd ed., London: Edward
Arnold.
Halliday, Michael A.K. and Ruqaiya Hasan
1976 Cohesion in English, London - New York: Longman.
430 References
Hannay, Mike
1983 The focus function in Functional Grammar: questions of contrast
and context. In: Simon C. Dik, ed., Advances in Functional Gram-
mar, Dordrecht/Holland - Cinnaminson/USA: Foris, 208-223.
Hasan, Ruqaiya
1984 Coherence and cohesive harmony. In: James Flood, ed., Under-
standing reading comprehension: Cognition, language, and the
structure of prose, Newark/Del.: International Reading Associa-
tion, 181-219.
Hausmann, Franz Josef
1985 Kollokationen im deutschen Wrterbuch. Ein Beitrag zur Theorie
des lexikographischen Beispiels. In: Henning Bergenholtz and Joa-
chim Mugdan, eds., Lexikographie und Grammatik. Akten des Es-
sener Kolloquiums zur Grammatik im Wrterbuch, 28.-30.6.1984,
Tbingen: Niemeyer, 118-129.
Herbst, Thomas
1988 A valency model for nouns in English. Journal of Linguistics 24,
265-301.
1996 What are collocations: Sandy beaches or false teeth. English Stud-
ies4, 379-393.
Hinds, John
1979 Organizational patterns in discourse. In Givn (1979), 135-157.
Hockett, Charles
1958 A course in modern linguistics, New York: MacMillan.
van Hoek, Karen
1997 Anaphora and conceptual structure, Chicago - London: The Uni-
versity of Chicago Press.
Hoey, Michael
1979 Signalling in discourse, Discourse Analysis Monographs No. 6,
Birmingham: English Language Research, University of Birming-
ham.
1991 Patterns of lexis in text, Oxford: Oxford University Press.
1994 A common signal in discourse: How the word reason is used. In:
John McH. Sinclair, Michael Hoey and Gwyneth Fox, eds. (1994),
Techniques of description. Spoken and written discourse. A Fest-
schrift for Malcolm Coulthard, London - New York: Routledge, 67-
82.
Hoey, Michael, ed.
1993 Data, description, discourse. Papers on the English language in
honour of John McH. Sinclair, London: HarperCollins.
References 431
Holmes, Janet
1990 Hedges and boosters in women's and men's speech. Language and
communication 10, 185-205.
Hone, Kaoru
1991 Cognitive motivations for event nominalizations. In: Lise M. Do-
brin, Lynn Nichols and Rosa M. Rodriguez, eds., Papers from the
27th Regional Meeting of the Chicago Linguistic Society, Part I:
The General Session, Chicago: Chicago Linguistic Society, 233-
245.
Huddleston, Rodney
1984 Introduction to the grammar of English, Cambridge etc.: Cam-
Ivani, Roz
1991 Nouns in search of a context: A study of nouns with both open- and
closed-system characteristics. International Review of Applied Lin-
guistics in Language Teaching 2, 93-114.
Jarvella, Robert J. and T. Russell Nelson
1982 Focus in information and general knowledge in sentence under-
standing. In: Jean-Francois Le Ny and Walter Kintsch, eds., Lan-
guage and language comprehension, Amsterdam: North-Holland,
73-86.
Jespersen, Otto
1927 A Modern English grammar on historical principles. Part III:
Syntax. Second volume, Heidelberg: Carl Winters Universitts-
buchhandlung.
1940 A Modern English grammar on historical principles. Part V: Syn-
tax. Fourth volume, Copenhagen: Ejnar Munksgaard.
Johansson, Stig and Knut Hofland
1989 Frequency analysis of English vocabulary and grammar. Based on
the LOB corpus. Volume I. Tag frequencies and word frequencies,
Oxford: Clarendon Press.
Johansson, Stig
1991 Computer corpora in English language research. In: Johannson and
Stenstrm (1991), 3-6.
Johansson, Stig and Anna-Britta Stenstrm, eds.
1991 English computer corpora. Selected papers and research guide,
Berlin - New York.
Johnson, Mark
1987 The body in the mind. The bodily basis of meaning, imagination,
and reason, Chicago - London: University of Chicago Press.
432 References
Johnson-Laird, Philip N.
1983 Mental models, Cambridge: Cambridge University Press.
Kallmeyer, Werner, Wolfgang Klein, Reinhard Meyer-Herrmann, Klaus Netzer
and Hans-Jrgen Siebert
1980 Lektrekolleg zur Textlinguistik. Band 1: Einfhrung, 3rd ed., K-
nigstein/Ts.: Fischer Athenum.
Kamp, Hans and Uwe Reyle
1993 From discourse to logic. Introduction to modeltheoretic semantics
of natural language, formal logic and Discourse Representation
Theory, Dordrecht etc.: Kluwer.
Karttunen, Lauri
1971a Some observations on factivity. Papers in Linguistics 4, 55-70.
1971b Implicative verbs. Language 47, 340-358.
Keenan, Elinor Ochs and Bambi B. Schieffelin
1976 Topic as a discourse notion: A study of topic in the conversations of
children and adults. In: Li (1976), 335-384.
Kintsch, Walter
1988 The role of knowledge in discourse comprehension: a construction-
integration model. Psychological review 95, 163-182.
1994 The psychology of discourse processing. In: Gernsbacher (1994),
721-740.
Kintsch, Walter and Teun van Dijk
1978 Toward a model of text comprehension and production. Psycho-
logical Review 85, 363-394.
Kiparsky, Paul and Carol Kiparsky
1971 Fact. In: Danny D. Steinberg and Leon A. Jakobovits, eds., Seman-
tics. An interdisciplinary reader in philosophy, linguistics and psy-
chology, Cambridge: Cambridge University Press, 345-369.
Koeppel, Rolf
1993 Satzbezogene Verweisformen. Eine datenbankgesttzte Unter-
suchung zu ihrer Distribution und Funktion in mndlichen Texten,
schriftlichen Texten und schriftlichen Fachtexten, Tbingen: Gun-
ter Narr.
Koflka, Kurt
1935 Principles of Gestalt Psychology, London: Routledge and Kegan
Paul.
Krenn, Monika
1985 Probleme der Diskursanalyse im Englischen. Verweise mit this,
that it und Verwandtes, Tbingen: Gunter Narr.
References 433
Kurzon, Dennis
1985 Signposts for the reader: a corpus-based study of text deixis. Text 5,
187-200.
Labov, William
1973 The boundaries of words and their meaning. In: Charles-James N.
Bailey and Roger W. Shuy, eds., New ways of analyzing variation
in English, Washington/DC.: Georgetown University Press, 340-
373.
1978 Denotational structure. In: Donka Farkas, Wesley M. Jacobsen and
Karol W. Todiys, eds., Papers from the Parasession on the lexicon,
Chicago: Chicago Linguistics Society, 220-260.
LakofF, George
1977 Linguistic gestalts. In: Papers from the thirteenth regional meeting,
Chicago Linguistic Society, Chicago: Chicago Linguistic Society,
236-287.
1987 Women, fire and dangerous things, Chicago: University of Chicago
Press.
Lakoff, George and Mark Johnson
1980 Metaphors we live by, Chicago - London: University of Chicago
Press.
Lambrecht, Knud
1994 Information structure and sentence form. Topic, focus and the
mental representation of discourse referents, Cambridge etc.:
Cambridge University Press.
Langacker, Ronald, W.
1987a Foundations of cognitive grammar. Vol. I., Theoretical prerequi-
sites, Stanford: Stanford University Press.
1987b Nouns and verbs. Language 63, 53-94.
1991 Foundations of cognitive grammar. Vol Two., Descriptive applica-
tions, Stanford: Stanford University Press.
1993 Reference-point constructions. Cognitive Linguistics 4, 1-38.
LDOCE3, see Summers ed. (1995)
Leech, Geoffrey N.
1981 Semantics. The study of meaning, 2nd ed., Harmondsworth: Pen-
guin.
1983 Principles of pragmatics, London - New York: Longman.
1987 Meaning and the English verb, 2nd ed., Harlow: Longman.
1991 The state of the art in corpus linguistics. In: Aijmer and Altenberg
(1991), 8-29.
1993 100 million words of English. English Today 33 (9) 9-15.
434 References
Leech, Geoffrey N. and Steven Fligelstone

1992 Computers and corpus analysis. In: Christopher S. Butler, ed.,
Computers and written texts, Oxford: Blackwell, 115-140.
Leisi, Ernst
1975 Der Wortinhalt. Seine Struktur im Deutschen und Englischen, 5.
Aufl., Heidelberg: Winter.
Le Ny, Jean-Francois
1991 Coherence in semantic representations: Text comprehension and
acquisition of concepts. In: Denhire and Rossi (1991), 205-221.
Levinson, Stephen C.
1983 Pragmatics, Cambridge etc.: Cambridge University Press.
Li, Charles N., ed.
1976 Subject and topic, New York: Academic Press.
Lipka, Leonhard
1977 Lexikalisierung, Idiomatisierung und Hypostasierung als Probleme
einer synchronen Wortbildungslehre. In: Dieter Kastovsky und
Herbert E. Brekle, eds., Perspektiven der Wortbildungsforschung.
Beitrge zum Wuppertaler Wortbildungskolloquium vom 9.-10. Juli
1976, Bonn: Bouvier.
1979 Semantic components of English nouns and verbs and their justifi-
cation. Angol Filolgiai Tanulmnyok 12/Hungarian Studies in
English XII, Debrecen, 187-203.
1980 Methodology and representation in the study of lexical fields. In: D.
Kastovsky, ed. (1980) Perspektiven der lexikalischen Semantik.
Beitrge zum Wuppertaler Semantikkolloquium vom 2.-3. Dezem-
ber 1977, Bonn: Bouvier, 93-114.
1992 An outline of English lexicology, 2nd ed., Tbingen: Niemeyer.
Longacre, Robert E.
1979 The paragraph as a grammatical unit. In: Givn (1979), 115-134.
1983 The grammar of discourse, New York and London: Plenum Press.
Louw, Bill
1993 Irony in the text or insincerity in the writer. The diagnostic poten-
tial of semantic prosodies. In: Baker, Francis and Tognini-Bonelli
(1993), 157-176.
Lutzeier, Peter R., ed.
1993 Studien zur Wortfeldtheorie/Studies in lexical field theory, Tbin-
gen: Niemeyer.
Lyons, John
1977 Semantics, 2 vols., Cambridge: Cambridge University Press.
1979 Deixis and anaphora. In: Myers (1979), 88-103.
References 435
1989 Semantic ascent: A neglected aspect of syntactic typology. In: Doug

Arnold, Martin Atkinson, Jacques Durand, Claire Grover and
Louisa Sadler, eds., Essays on gramatical theory and universal
grammar, Oxford: Clarendon Press, 153-186.
Mcintosh, Angus
1961 Patterns and ranges. Language 37 (3), 325-337.
Mair, Christian
1990 Infinitival complement clauses in English. A study of syntax in dis-
course, Cambridge etc.: Cambridge University Press.
Marchand, Hans
1969 The categories and types of present-day English word-formation. A
synchronic-diachronic approach, 2nd ed., Mnchen: Beck.
Menzel, Peter
1975 Semantics and syntax in complementation, The Hague - Paris:
Mouton.
Mey, Jacob L.
1993 Pragmatics. An introduction, Oxford: Blackwell.
Meyer, Charles F.
1992 Apposition in contemporary English, Cambridge etc.: Cambridge
University Press.
Miller, George A.
1956 The magical number seven, plus minus two: some limits on our ca-
pacity for processing information. Psychological Review 63, 81-97.
Miller, James R. and Walter Kintsch
1980 Readability and recall of short prose passages: a theoretical analy-
sis. Journal of experimental psychology: human learning and
memory 6, 335-354.
Mindt, Dieter
1995 An empirical grammar of the English verb. Modal Verbs, Berlin:
Coraelsen.
Minsky, Marvin
1975 A framework for representing knowledge. In: Patrick H. Winston,
ed., The psychology of computer vision, New York: McGraw-Hill,
211-277.
Mitchell, T.F.
1971 Linguistic 'goings-on': Collocations and other matters on the syn-
tagmatic record. Archivum Linguisticum 2, 35-69.
Myers, Terry, ed.
1979 The development of conversation and discourse, Edinburgh: Edin-
burgh University Press.
436 References
Ney, James W.
1981 Semantic structures for the syntax of complements and auxiliaries
in English, The Hague etc.: Mouton.
Palmer, Frank R.
1986 Mood and modality, Cambridge: Cambridge University Press.
1987 The English verb, 2nd ed., London - New York: Longman.
1990 Modality and the English modals, 2nd ed., London - New York:
Longman.
Parsons, Terence
1990 Events in the semantics of English: A study in subatomic seman-
tics, Cambridge/MA.: MIT Press.
Pawley, Andrew
1987 Encoding events in kalam and English: Different logics for report-
ing experience. In: Tomlin (1987), 329-360.
Peterson, Philip L.
1997 Fact proposition event, Dordrecht etc.: Kluwer.
Petfi, Jnos
1976 A frame for FRAMES: A few remarks on the methodology of se-
mantically guided processing. In: Proceedings of the Second An-
nual meeting of the Berkeley Linguistic Society, Berkeley: Univer-
sity of California Institute of Human Learning, 319-329.
Polanyi, Livia and R.J.H. Scha
1983 The syntax of discourse. Text 3 (3), 261-270.
Porzig, Walter
1930 Die Leistung der Abstrakta in der Sprache. Bltter fr deutsche
Philosophie 4, 66-77.
Poutsma, Hendrik
1929 A grammar of late modem English. Part I: The Sentence. Second
Half: The composite sentence, 2nd ed., Groningen: P. Noordhoff.
Prince, Ellen F.
1981 Toward a taxonomy of given-new information. In: Peter Cole, ed.,
Radical pragmatics, London - New York: Academic Press, 223-
255.
Quine, William V.
1960 Word and object, Cambridge/ MA. : MIT Press.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik
1985 A comprehensive grammar of the English language, London - New
York: Longman.
References 437
Radford, Andrew
1988 Transformational grammar, Cambridge: Cambridge University
Press.
1997 Syntax. A minimalist introduction, Cambridge: Cambridge Univer-
sity Press.
Rable, Wolfgang
1972 Satz und Text, Tbingen: Niemeyer.
Ransom, Evelyn .
1986 Complementation: Its meaning and forms, Amsterdam: John Ben-
jamins.
Reiss, Nira
1985 Speech act taxonomy as a tool for ethnographic description: an
analysis based on videotapes of continuous behavior in two New
York households, Amsterdam - Philadelphia: John Benjamins.
Rescher, Nicholas
1968 Topics in philosophical logic, Dordrecht: Reibel.
Rickheit, Gert and Hans Strohner
1993 Grundlagen der kognitiven Sprachverarbeitung, Tbingen - Basel:
Francke.
Rosenbaum, Peter S.
1967 The grammar of English predicate complement constructions,
Cambridge/MA.: MIT Press.
Rosenberg, St.T.
1979 Frame-based text processing. In: Dieter Metzing, ed., Frame con-
ceptions and text understanding, Berlin - New York: de Gruyter,
96-119.
Rumelhart, David E. and Andrew Ortony
1977 The representation of knowledge in memory. In: Richard C. Ander-
son, Rand J. Spiro and William E. Montague, eds., Schooling and
the acquisition of knowledge, Hillsdale /N.J. - N.Y.: Lawrence Erl-
baum, 99-135.
Salmon, Nathan
1986 Frege 's puzzle, Cambridge/Ma. : MIT Press.
Sanford, Anthony J. and Linda M. Moxey
1995 Aspects of coherence in written language: A psychological per-
spective. In: Gernsbacher and Givn (1995a), 161-187.
Schnk, Roger C. and Roger P. Abelson
1977 Scripts, plans, goals and understanding. An inquiry into human
knowledge structures, Hillsdale/N.J.: Lawrence Erlbaum Associ-
ates.
438 References
Schmid, Hans-Jrg
1993 Cottage and Co., idea, start vs. begin. Die Kategorisierung als
Grundprinzip einer differenzierten Bedeutungsbeschreibung, T-
bingen: Niemeyer.
1997 The historical development and present-day use of the noun idea as
documented in the OED and other corpora. Poetica 47, 87-128.
1999 Functions of metaphor. In: Karin Pittner and Robert J. Pittner, eds.,
Beitrge zu Sprache und Sprachen 2. Vortrge der 5. Mnchner
Linguistik Tage, Mnchen 1995, Mnchen: Lincom Europa, 163-
172.
Schneider, Edgar W.
1988 Variabilitt, Polysemie und Unschrfe der Wortbedeutung, 2 Vols.,
Tbingen: Niemeyer.
Searle, John R.
1969 Speech acts, Cambridge: Cambridge University Press.
1976 A classification of illocutionary acts. Language in Society 5, 1-24.
Sgall, Petr, Eva Hajicov, Eva BeneSavo
1973 Topic, focus and Generative Semantics, Kronberg: Scriptor.
Sgall, Petr, Eva Hajicov and Jarmila Panevov
1986 The meaning of the sentence in its semantic and pragmatic aspects,
Dordrecht etc.: D. Reidel.
Shibatani, Masayoshi and Sandra A. Thompson, eds.
1996 Grammatical constructions: Their form and meaning, Oxford:
Oxford University Press.
Sinclair, John McH.
1966 Beginning the study of lexis. In: Bazell et al. (1966), 410-430.
1990 Collins COBUILD English grammar, London - Glasgow: Collins.
1991 Corpus, concordance, collocation, Oxford: Oxford University
Press.
Slobin, Dan I.
1996 Typology and rhetoric: Verbs of motion in English and Spanish. In:
Shibatani and Thompson (1996), 195-220.
Sornicola, R.
1994 Topic, focus, and word order. In: R.E Asher, ed., The encyclopedia
of language and linguistics, Vol. 9, Oxford: Pergamon, 4633-4640.
Sprengel, Konrad
1980 ber semantische Merkmale. In: D. Kastovsky, ed., Perspektiven
der lexikalischen Semantik. Beitrge zum Wuppertaler Semantik-
kolloquium vom 2.-3. Dezember 1977, Bonn: Bouvier, 145-177.
References 439
Stubbs, Michael
1995 Collocations and semantic profiles. On the cause of the trouble with
quantitative studies. Functions of Language 2, 23-55.
Summers, Della, ed.
1995 Longman dictionary of contemporary English, 3rd ed., Bungay:
Longman. (=LDOCE3)
Summers, Delia
1996 Computer lexicography: the importance of representativeness in
relation to frequency. In: Thomas and Short (1996), 260-266.
Svartvik, Jan, ed.
1992a Directions in corpus linguistics. Proceedings of Nobel Symposium
82, Stockholm, 4-8 August 1991, Berlin - New York: Mouton de
Gruyter.
Svartvik, Jan
1992b Corpus linguistics comes of age. In: Svartvik (1992a), 7-13.
1996 Corpora are becoming mainstream. In: Thomas and Short (1996),
3-13.
Sweetser, Eve E.
1987 The definition of lie : An examination of the folk models underlying
a semantic prototype. In: Dorothy Holland and Noami Quinn, eds.,
Cultural models in language and thought, Cambridge: Cambridge
University Press, 43-66.
1990 From etymology to pragmatics. Metaphorical and cultural aspects
of semantic structure, Cambridge etc.: Cambridge University Press.
Tadros, Angele
1994 Predictive categories in expository text. In: Coulthard (1994), 69-
81.
Talmy, Leonard
1988 Force dynamics in language and cognition. Cognitive Science 12,
49-100.
1991 Path to realization: A typology of event conflation. In: Proceedings
of the Seventeenth Annual Meeting of the Berkeley Linguistic Soci-
ety, Berkeley: Berkeley Linguistic Society, 480-519.
1996 The windowing of attention. In: Shibatani and Thompson (1996),
235-288.
Taylor, John R.
1995 Linguistic categorization. Prototypes in linguistic theory, 2nd ed.,
Oxford: Clarendon Press.
Tesnire, Lucien
1959 Elments de syntaxe structurale, Paris: Librairie C. Klincksieck.
440 References
Thomas, Jenny and Mick Short, eds.,

1996 Using corpora for language research. Studies in Honour of Geof-
frey Leech, London - New York: Longman.
Thompson, Sandra A.
1989 A discourse approach to the cross-linguistic category 'adjective'.
In: Roberta Corrigan, Fred Eckman and Michael Noonan, eds.,
Linguistic categorization, Philadelphia - Amsterdam: John Ben-
jamins, 245-265.
Tomlin, Russell S., ed.
1987 Coherence and grounding in discourse, Amsterdam - Philadelphia:
John Benjamins.
Trabasso, Tom, Soyoung Suh and Paula Payton
1995 Explanatory coherence in understanding and talking about events.
In: Gernsbacher and Givn (1995a), 189-214.
Tuggy, David
1996 The thing is is that people talk that way. The question is is Why?.
In: Eugene H. Casad, ed., Cognitive linguistics in the redwoods.
The expansion of a new paradigm in linguistics, Berlin - New
York: Mouton de Gruyter, 713-752.
Ungerer, Friedrich
1990 Review of Leech (1987) and Frank R. Palmer (1987) The English
Verb, 2nd ed.,Anglia 108, 155-159.
Ungerer, Friedrich and Hans-Jrg Schmid
1996 An introduction to cognitive linguistics, London etc.: Longman.
Vendler, Zeno
1967 Linguistics in philosophy, Ithaca/N.Y.: Cornell University Press.
1968 Adjectives and nominalizations, The Hague: Mouton.
1980 Telling the facts. In: John R. Searle, Ference Kiefer and Manfred
Bierwisch, eds., Speech act theory and pragmatics, Dordrecht:
Reidel, 273-290.
Weaver, Charles A. Ill, Suzanne Mannes and Charles R. Fletcher, eds.
1995 Discourse comprehension. Essays in honor of Walter Kintsch,
Hillsdale/N.J.: Lawrence Erlbaum Associates.
Wertheimer, Max
1958; 1923 Principles of perceptual organization. In: David C. Beardsley and
Michael Wertheimer, eds., Readings in perception, Princeton/N.J.:
D. van Nostrand, 115-135. [Originally published as Untersuchun-
gen zur Lehre von der Gestalt, Zwei, Psychologische Forschungen
4 (1923), 301-350.]
References 441
Whorf, Benjamin L.
1956 Language, thought, and reality. Selected writings of Benjamin Lee
Whorf, ed. by John B. Carroll, Cambridge/MA. - New York - Lon-
don: MIT Press and John Wiley and Sons.
Wierzbicka, Anna
1972 Semantic primitives, Frankfurt: Athenum.
1980 Lingua mentalis: The semantics of natural language, Sydney - New
York: Academic Press.
1985 Lexicography and conceptual analysis, Ann Arbor: Karoma.
1987 English speech act verbs. A semantic dictionary, Sydney etc.: Aca-
demic Press.
1988 The semantics of grammar, Amsterdam - Philadelphia: John Ben-
jamins.
1991 Cross-cultural pragmatics. The semantics of human interaction,
Berlin - New York: Mouton de Gruyter.
Winter, Eugene O.
1977 A clause-relational approach to English texts. Instructional Science
6, Vol. 1 special issue, 1-92.
1982 Towards a contextual grammar of English, London: Allen and
Unwin.
1992 The notion of unspecific versus specific as one way of analysing the
information of a fimd-raising letter. In: William C. Mann and San-
dra A. Thompson, eds., Discourse descriptions. Diverse analyses
of a fund-raising text, Amsterdam - Philadelphia: John Benjamins,
131-170.
von Wright, Georg H.
1951 An essay in modal logic, Amsterdam: North Holland Publishing
Company.
Index of shell nouns
Note: The bold numbers in the following index of shell nouns refer to pages
where shell nouns are discussed as members of particular families of uses.
The extent to which bold numbers occur in the entry for a shell noun is thus
an indication of the semantic versatility of the noun.
ability 4, 5, 25, 231-232, 252, 327 anger 182,226,228

absurdity 127, 128 announcement 57, 148, 153, 155-
acceptance 157, 159 156, 159, 317, 345, 346, 351
accident 127, 128 annoyance 226, 228
account 102, 138, 153, 155, 159- anomaly 127
161 answer 5, 59, 60, 132, 138, 140,
accusation 10, 57, 131, 149, 156, 148, 157-159, 354
159, 308 anticipation 206
achievement 60, 62, 261, 272-273, anxiety 187,222,228,229
309, 321, 345 aphorism 144
acknowledgement 57, 148, 157 appeal 139, 173, 174,
act 6, 67, 69, 73-74, 262 application 173
action 6, 60, 67-74, 261-265, 268 appointment 224-225
adage 57, 144 appreciation 206
admission 136, 157-159, 327 apprehension 228-229
advantage 8, 22, 27, 59, 92, 126- approach 4, 254-255, 276-277,
127, 335 284, 286-288, 323-325, 344,
advice 5, 60, 139, 148, 175, 351 351, 358
affirmation 148, 156-159, 308, area 4, 61, 64, 86, 254-256, 276-
380 277, 280-282
age 61, 282 argument 12, 57, 59, 132, 138,
agenda 248, 320 140-143, 148, 156-160, 313,
agreement 58, 224-225, 351 323
aim 4, 7-8, 15, 20, 36, 60, 68, 70, arrangement 224-225, 351
87, 187-188, 210-213, 216, 275, art 212
276, 335, 337 aspect 11, 52, 92, 116-118, 123
allegation 57, 162-163, 326 assertion 55, 57, 99, 155, 159
allegory 144 assessment 154,163-164
alternative 60, 114, 261, 269 asset 126-127
amazement 226, 228 assignment 231,248
ambition 60, 187, 210, 335 assumption 4, 17, 55, 57, 59, 67,
amendment 148, 157, 159 86, 136, 137, 163, 186, 195-
analogy 114 200, 202- 203, 314, 326-327,
analysis 11,206-207 335, 366
444 Index of shell nouns
assurance 57, 148, 178-179 cause 11, 41, 61, 74, 102-105,
astonishment 226, 228 118, 380
attempt 4, 5, 58, 77-79, 86-87, caveat 59, 148, 179
266-268, 318, 323, 327 certainty 57, 59, 86, 232, 235,
attitude 203-206, 212, 321, 323- 242, 313
324 challenge 60, 62, 248
attraction 130 chance 3, 5, 23, 25, 45, 57, 58,
attribute 116-117 61, 68, 74, 232, 236, 240, 254-
audacity 58,222-223 256, 325, 414
authority 252-253 change 128-129, 137, 262-263,
awareness 203 351
axiom 11, 57, 146-147, 189-190 characteristic 116-117, 123, 219
charge 148-149, 156, 159, 310
charm 127
bargain 224-225 cheek 222-223
basis 118-119 choice 269-270
beauty 127 claim 10, 57, 136-137, 162-163,
belief 4, 23, 55, 57, 59, 71, 72, 314, 316
86, 186, 188, 195-203, 234, clout 252
309, 310, 314, 316, 324, 366 clue 110-111
benefit 126-127 coincidence 57, 127-128
bet 60, 137, 148, 161-162 confort 130
betting 59, 148, 161-162 command 172-173, 186, 234
bid 58, 177 comment 10, 148, 157, 159
bitterness 226, 228 commission 49, 58, 248
blessing 126-127 commitment 58,180
blow 61, 270, 320-321 compensation 109-110
boast 57, 59, 181-182, 335 complaint 181-182, 186, 366
bonus 126-127 complication 59, 121-123
breakthrough 62, 130, 270, 320, compliment 182-183, 309
321 compromise 224-225, 321
brief 60, 212 compulsion 58, 221
burden 121 concept 11, 68, 79, 85, 125, 188-
business 93, 101, 248 192
conception 203
concern 25, 57, 59-60, 62, 215,
calculation 196-199 226-229, 335, 337
call 172-174 concession 156-159
campaign 58, 64, 68-69, 266-268 conclusion 10, 57, 59, 138, 140,
capability 252 157, 159, 310, 342
capacity 58, 235, 252 condition 11, 275-276, 289
case 3, 5, 6, 11, 57, 61, 68, 93, confession 133,159
101, 143, 289-291 confidence 57,197, 222, 240
catastrophe 129, 270, 324 confirmation 57, 148, 153, 159
catch 121, 123 conjecture 163, 196, 314
catchphrase 144 connection 102, 108-109
consensus 59, 224-225 deduction 206

consequence 106, 108-110 defence 153,157-159
consideration 62, 206-207 definition 144
consolation 59, 130, 335 delight 187,226,228
conspiracy 58, 266-267 delusion 57, 207-208, 309
constraint 289 demand 49, 172-174
contention 57, 162-163 demonstration 110-111
context 11,278-279 denial 157, 159
contract 224-225 description 9, 153, 159
contradiction 148, 157, 159 desire 8, 58, 215-216, 234, 366
contrast 114 destiny 259-260, 313
contribution 157, 159 determination 58, 222-223
conundrum 61-62, 194-195, 324 diagnosis 206-207
convention 268 dictum 57, 146-147
conviction 57, 196-197, 199, 203, difference 4, 59, 114-116, 337,
326 351, 354, 365
corollary 59, 110, 112 difficulty 59, 62, 121-124, 261,
counterclaim 156-157, 159 270-272, 285, 335
countermeasure 60, 266, 268 dilemma 62, 121, 124, 270-272,
coup 272 285, 335
courage 58, 222-223, 233 directive 172-173
courtesy 222-223 disadvantage 59, 121, 123
credo 146-147, 189-190 disappointment 226, 228
crime 273-274, 319, 323, 324 disaster 129, 270
criterion 289 disclosure 148, 154, 164-165
criticism 157, 159 discovery 206-207
critique 10, 157, 159 discrepancy 114
crusade 266 disgrace 129
crux 121, 309 disgust 226, 228
cure 219-220 disinclination 58, 217-218
curiosity 127 disposition 258-259
curse 129 disquiet 228-229
custom 268 distinction 114, 116-117
distortion 154, 165-166
doctrine 146-147, 189-191
danger 57, 59, 62, 180, 236-239, dogma 146-147,189-190
348, 351, 358 doubt 5, 11, 57, 100, 186, 208
deal 58, 119, 224-225 downside 59, 121, 123, 335
debacle 129,270 drama 129, 270
decency 222-223 drawback 59, 121, 123, 335
deception 207 dread 228-229
decision 4, 5, 58, 212-214, 323, dream 60, 215
351 drive 58,212
declaration 10, 69, 99, 149, 150, duty 25, 58, 60, 248
153, 159, 186, 380
decree 155, 159
eagerness 58, 222 feature 52, 116-117, 219

edict 148, 159 feeling 25, 57, 59, 64, 163,195-
effect 57, 60,106-108 203, 323, 337, 359, 380
effort 5, 8, 58, 86, 266-268 fiction 207
effrontery 222-223 fight 266-267
endeavour 8, 266 finding 57, 59, 110-113, 314
energy 222-223 flaw 129
enigma 61-62, 194-195 flexibility 222-223
enterprise 266 folly 127,273
equation 27, 206 forecast 136, 148, 153, 161
era 61,282 foresight 222
error 273 formula 146-147
essence 116-117 foundation 118-119
estimate 196-200 franchise 245
ethos 203, 205 freedom 58, 232, 245-247
event 6, 33, 65-74, 77, 129, 141- frustration 226, 228
142, 262-265, 289-291, 314, function 218-219
321-324 Jury 226, 228
evidence 57, 59, 110-113, 314,
321-324
example 119-120 gall 58,222-223
exception 119-120 gambit 60, 284, 286
excuse 182, 308 gamble 266
expectation 57, 196, 198, 211, 234 generalization 206, 309
experience 59, 203-206 goal 60, 188, 210, 214-216, 344
explanation 59, 61, 140, 148, 157- gossip 145
161, 337, 344, 348 grace 222-224, 263
gratitude 226, 228
grief 226, 228
facility 254 grievance 148, 182, 186
fact 3-6, 8-11, 15, 22, 30, 33, 53, gripe 59, 87, 182
55-57, 59, 64, 67-68, 71-74, 77- ground 102, 105
79, 85, 93, 97-102, 129, 151, grudge 226, 228
192, 242-244, 250, 261, 267, grumble 59, 135, 148, 181-182
275-276, 313, 314, 325, 336, guarantee 57,178
337, 361-362, 366-369, 377, guess 59, 86,136-137, 140, 148,
380, 415 153, 162-163, 201, 202
factor 59, 62, 116-118 guilt 226, 228
failure 58, 252, 254 gumption 222-223
faith 196,203
fallacy 57, 207-208, 309
fantasy 207-208 habit 86, 268-269
farce 129,270 handicap 121, 123, 415
fate 259-260 happiness 226, 228
fault 127, 261, 273 heart 222
fear 57, 59, 118, 228-229, 335 hint 154, 162-163, 351, 375, 416
hope 5, 57, 59, 60, 64, 195-202, intelligence 140

210-211, 309, 316, 335 intent 58, 215
hunch 59, 196, 198, 200 intention 25, 58, 60, 64, 77-79,
hurdle 121 215-216, 234,
hypothesis 189, 348 interest 131,210
interpretation 206
intimation 110
idea 3-6, 10-11, 21, 23-26, 41, intuition 196-200, 359
45, 52, 57-68, 74, 79, 85, 123, invitation 58, 174
125, 189-196, 203, 210-215, irony 59, 68, 74-77, 127-128, 335-
218, 310, 313-314, 323-325, 338
335, 337, 343, 344, 366, 369, irritation 226, 228
374-375, 380, 416 issue 11, 62, 85, 189-193
ideal 60, 210,
illusion 207-208
image 189 job 25, 58, 60, 248-249, 313, 321
impact 106-108 joke 144-145
imperative 60,249-250 joy 226, 228
impetus 222 judgement 10,154
implication 10, 59, 110-113 justification 156-159
importance 127
impression 57, 59, 163, 195-197,
200-203, 311, 337, 359 keenness 58, 222-223
improvement 272 key 60, 193, 219-220
impulse 58, 60, 220-221 knack 284
inability 58,252-254 knowledge 57, 195-198, 203, 234
incapacity 58, 252-254
incentive 58, 221, 321
inclination 58, 60, 215, 233 lament 181-183
inconsistency 114 law 146-147, 189, 325
indication 57,110-113 leeway 58, 245, 247
indicator 110-111 legacy 146-147
indignation 226, 228 legend 85, 146
inducement 58, 220-221 lesson 206-207
inevitability 127 licence 245
inference 11, 57, 59, 206 lie 148, 154, 165-166
information 5,140-142 likelihood 4, 23, 32, 59, 86, 99,
initiative 266, 351 240-241
injunction 58, 148, 172-173 limitation 289
inkling 57, 195-198 line 203, 212, 416
innovation 272 link 102, 108-109, 344
insight 206, 368 logic 189, 203
insistence 55, 57, 148, 157, 159 longing 215
inspiration 220-221 luck 130
instinct 60, 195-200, 203, 221-222
instruction 172-173
mandate 58,248 notion 4, 57, 68, 79, 85, 125, 189-

manifestation 110-112 191, 203
manoeuvre 47, 266 nous 222-223
marvel 130 novelty 127-128
maxim 146-147, 189-190 nuisance 129
measure 4, 86, 266, 268, 351, 373
merit 130
message 4, 57, 59, 79, 85, 139- oath 148, 178, 234
143, 305, 312, 335 object 10, 60, 66, 75, 210-211
metaphor 146-147,188-190 objection 59, 157, 159
method 60, 254-256, 284-288 objective 60, 210, 214, 216
miracle 130,270 obligation 4, 8, 58, 60, 231, 249-
misapprehension 54, 57, 207, 380 251
miscalculation 207 observation 154-155, 159, 314,
misconception 57, 59, 207-208, 317
309 obsession 203
misfortune 129 obstacle 121
misjudgment 207, 309 occasion 61,235,250,254-255,
misperception 207, 309 414
mission 60, 248, 267 oddity 59, 127, 335
mistake 60, 64, 68, 273-274, 309, offence 129,273
329 offer 58, 177, 186
moment 49, 58, 61, 254-255, 275, opinion 5, 57, 59, 203-206, 336,
282-283 416
motion 148, 172-173 opportunity 5, 25, 58, 74-77, 234,
motivation 220-222 254-256, 351
motive 60, 220-221 option 60, 236-239, 245, 269-270
motto 146-147, 189-190, 212 order 4, 49, 51, 86, 148-149, 172-
move 4, 10, 58, 60, 67, 69, 86, 174, 186, 234, 366, 371, 373,
266-268, 343, 351 378
mystery 61-62, 194-195 orthodoxy 146
myth 140, 146-147, 189-190 outcome 106-108
necessity 231, 234, 249, 259 pact 224-225

need 4, .8, 25, 58, 231, 249-251, pain 226, 228
259 paradox 10, 59, 127-128, 335
nerve 222 paranoia 127
news 3, 4, 6, 57, 59, 135, 136, part 52, 63, 118-119, 219
139-146, 313-314, 335, 337, passion 222
369 payoff 106, 108
nightmare 129,270 peculiarity 59, 127, 335
nonsense 10, 144-145 perception 203, 205
norm 284 period 61, 282
notice 159 permission 4, 58, 86, 231, 233,
notification 155,159 245-247
permit 245 preoccupation 220-221

perspective 11, 135, 203, 205 presentiment 57, 203, 205
persuasion 203, 205 pressure 58, 231, 249-251
petition 173 presumption 57, 59, 196-198, 203
phenomenon 93, 101 presupposition 57, 196, 203
philosophy 203, 212 pretence 57, 166
pity 129, 309 pretext 148, 165-166
place 4, 10, 14, 25-26, 30, 34, 53, pride 226, 228
58,61,74, 86, 193,234,254- principle 57, 85, 189-191, 212,
260, 275-276, 280-282, 285, 376
324 priority 60,269
plan 4, 58, 60, 86, 123, 138, 188, privilege 245
212-214, 225, 309, 323, 327, prize 110
335, 337, 351, 380 probability 59, 234, 240-241
plea 173-174 problem 3-8, 11, 15, 26, 45, 49,
pleasure 226, 228 59-62, 79, 85, 95, 121-125, 238,
pledge 58, 178-179 270-272, 285, 309, 318, 321-
plot 266-67 324, 335-337, 350-352, 355,
ploy 58, 60, 266-268, 309 366, 378, 380
point 3-6, 11-12, 36-37, 57-61, procedure 86, 275, 284, 286, 288
79, 87, 93-97, 101, 116, 121, proclamation 153, 155, 159
123, 140, 143, 189-190, 193- prognosis 148, 161
194, 210, 280-281, 308, 332- programme 58, 188, 212, 214, 225
338, 378, 380 project 49, 60, 212, 214, 225
policy 49, 52, 60, 212, 214, 376 projection 206
position 3, 58-61, 157, 173, 189- promise 99, 178-179, 186, 234,
190, 254-256, 262, 265, 278- 245, 308, 366
281, 331 pronouncement 148, 160, 380
possibility 57, 59, 68, 231-232, proof 57, 91, 110-113, 313, 319,
234, 236-241, 254-255, 313, 321
317, 325 propensity 58, 258-259
potential 252-253 prophecy 161
power 58, 252-253 proposal 4, 58, 69, 86, 175-176
practice 284, 286-289, 324 proposition 5, 33, 175, 314
praise 182-183 prospect 99, 196, 198, 320
precaution 266 protest 148, 153, 157, 160
precept 189 proverb 144
preconception 203, 205, 380 provision 289-290
precondition 289, 380 proviso 55, 57, 59, 275, 277, 289
predicament 121 provocation 157, 160
prediction 136, 153, 161 punishment 109-110
preface 144 purpose 60, 75, 200, 218-219
preference 269-270 puzzle 62, 194
prejudice 203, 205
premise 54, 57, 59, 195-198
premonition 228, 374
qualification 157, 160 reposte 157-160

query 61, 152, 166-168, 186 request 172-174
quest 266, 268 requirement 231,249-251
question 4, 10-11, 26, 45, 57, 61- resentment 226, 228
62, 100, 125, 166-170, 186, reservation 228
189, 194, 198, 208, 308, 318, resistance 222-223
323-325, 332-334, 337-338 378, resolution 222-223
380 resolve 58,222-223
quibble 148, 182, 309 response 60, 157-160
responsibility 232, 248
restriction 289
race 266-261 result 59-60, 91, 106-108, 337,
rage 226, 228 370
rationale 59, 189, 203, 212 retort 59, 157, 160
reaction 4, 60, 64, 68, 72, 86, revelation 154, 164-165
156-158, 160 revolution 130, 270
readiness 58,222-223 reward 109-110
reading 206 right 58, 245-247
realisation 55, 57, 206-207, 314 risk 57, 59, 236-240
reality 59, 232, 242-244, 325, ritual 268
336-337 role 60, 248
reason 3-4, 11, 57-59, 61, 67-68, room 58, 254-255
74-77, 79, 85, 91, 102-105, 118, routine 212
285, 321, 337, 351, 380 rule 189-190,212
reasoning 206 ruling 148, 160
reassurance 148, 157, 160 rumour 4, 85, 145-146
recipe 284, 286 ruse 47, 58, 60, 266-268
reckoning 206 rush 266-267
recognition 206-207
recollection 206
recommendation 174-175 sadness 226, 228
refinement 272 satisfaction 226, 228
reflection 206 scandal 129, 270
refiisal 58, 64, 217-218 scenario 189
region 61, 64, 254-255, 275, 277, scheme 7, 8, 49, 58, 212, 214
280-282 scope 65, 206, 254-255
regret 25, 59, 186, 226-228 secret 60, 189, 212
relief 25, 226-228 sensation 130, 270
reluctance 58, 217-218 sentiment 196, 203
remark 5, 77-79, 148, 154, 160, sequel 106, 108
343 shame 129, 321
remedy 60,219-220 shock 226, 228
reminder 4, 57, 110-113 sign 57, 110-113, 321
reply 148, 157, 160 signal 110-111, 329
report 3-4, 6, 10, 57, 69, 140-143, significance 206-207
153, 161, 351 similarity 114
sin 273 tactic 60, 212

site 61, 64, 280 tale 160-161
situation 3-5, 29, 61, 86, 254-255, talent 252-253
262-265, 275, 278-279, 324 talk 145-146
skill 252-253 target 60, 210
slogan 146-147 task 25, 60, 248, 311
snag 59, 62, 121-124, 271-272, teaching 146-147, 189-190
310, 335-337, 351 technique 284, 286, 288
solace 130 temerity 58,222-223
solution 60, 219-220, 320 temptation 58, 60, 220-222
sorrow 226, 228 tendency 58, 60, 258-259
space 254-255 tenet 59, 146-147
speciality 269-270 terror 226, 228
speculation 57, 61, 195-198 test 266, 268
spot 61,280 testimony 160
stage 60-61, 86, 193, 254-256, theme 189, 193
276-277, 282 theory 57, 59, 79, 188, 189, 191,
stamina 222 194
stance 203, 205 thesis 59, 189
stand 203, 205 thing 3-8, 10-11, 52, 53, 58-60,
standpoint 203, 205 62, 85, 93-97, 100-102, 105,
statement 5, 12, 57, 148-149, 154, 120-123, 240, 257, 321, 332-
155, 160, 186, 314, 317, 342, 334, 338, 378, 380, 413, 415
351 thinking 203, 205, 318
step 60, 254-255 thought 57, 59, 67, 74, 188-192
stereotype 146-147, 189-190 threat 179, 180, 245, 366
stipulation 57, 59, 289 thrill 226, 228
story 59, 140-144, 160-161, 309 time 4, 14, 30, 45, 49, 53, 58, 61,
strategy 60, 212, 214 74, 86, 143, 254-260, 267, 275-
strength 222-223 276, 278, 284-285
struggle 58, 266-268 tip 174-175
subject 189, 193, 375 topic 189
subtext 59, 144 tradition 4, 268
success 272, 321 tragedy 59, 61, 129, 309, 321, 335
suggestion 10, 57, 59, 140, 148, travesty 127
150-153, 162-163, 175-176 trend 258-259
superstition 196, 198 trick 4, 60, 68, 266-268, 284-287
supposition 57, 196, 198, 203 triumph 272-273, 321
surmise 59, 196 trouble 8, 59, 121-124, 270-271,
surprise 57, 59, 186, 226, 228 285, 309, 335, 337
suspicion 57, 196-200, 203 truism 57, 146-147
symbol 110-111, 125 truth 4, 8, 232, 242-244, 311,
symptom 110-111 325, 336-337
twist 127-128
uncertainty 127, 236 warning 4, 57, 135, 139, 179-180

understanding 61,196-200 way 4, 14, 45, 58, 60, 77-79, 86,
undertaking 148, 178 143, 254-260, 267, 275, 284-
unknown 62, 194 289, 321, 351
unwillingness 58, 217-218 weakness 127
upshot 4, 59, 77-79, 106-108, 338 whinge 182
urge 49, 58, 220-221 whisper 145
willingness 25, 58, 215-217, 234,
250
venture 266, 268 willpower 222
verdict 148, 163-164 wisdom 189
version 160-161 wish 8, 186, 215-216
view 11, 57, 59, 64, 188, 203-206, wit 222-223
337, 375, 416-417 wonder 130
viewpoint 1 1 , 2 0 3 , 2 0 5
virtue 130 worry 59, 228-229, 335, 337
vision 210
vocation 220-221 yearning 215
vow 58, 148, 178-179
zeal 222
Index of subjects
aboutness, 305-307, 349 anaphoric nouns, 10-11, 240, 279,

abstract entities, 16-17, 63, 99, 308, 347
117, 315, 413-414 APODOSIS, 290-291
abstract noun, 3-6, 10-11, 24, 63, apposition, 4, 30, 223, 281, 367,
80, 83, 131, 191, 204, 276, 303, 412
371, 377, 380, 419 assumed familiarity, 305, 340, 342
abstract relation, 66-69, 72-73, 77, attention,
80, 85-89, 91, 99-100 focus of, 118, 135, 306-307,
abstractness, 9, 63-74, 80, 168, 331, 339, 359
291, 303, 379, 414 and information status 330
extensional, 63-68, 70-74, 275 windowing of 90-91
accessible information, 116,304- ATTITUDE, 203-205
305, 312, 330, 340, 369, 371, attraction, 53-55, 130, 215-216
373 attribute, 29, 89
activation, 28-29, 90, 304-305, AUTHORITY, 244-250 259
339, 344, 372-373, 410
activation cost, 304, 373
ACTIVITY, 246, 249-252, 262-269, BELIEF, 91, 111-113, 195-212,
286-288, 311, 345 220, 225
adjectives boulamaic modality, 233
classifying, 318-319, 324, 328 buffer, short-term memory 362,
cohesive, 169-170, 194, 318- 370-373, 377, 409
319, 347-348
descriptive, 170, 264, 281-282,
288-289, 318-319, 322-326, carrier noun, 12-13, 75
334-335 CAUSE, 91, 102-106, 110, 362
evaluative, 96, 104, 125, 144, chunk, of information, 6, 12, 14,
169-170, 194, 264, 281-283, 361, 370-377, 409
289, 309, 318-319, 322-328, cleft sentence, 37, 237, 333-334
COBUILD, 5, 7-8, 21, 42-47, 409,
334-336
412-413, 417, 420
restrictive, 96, 194, 263, 268, cognition-utterance verbs, 35
281, 286, 318 Cognitive Grammar, 20, 27, 70,
74, 95, 365, 367-368, 419
AGENT, 244, 246-251, 259 cognitive model, 20, 28-29, 80,
AIM, 209-215,219,221-225 88-91, 121, 304-305, 315, 341,
alethic modality, 233 411
anaphora, 18, 27, 29, 31, 329, coherence, 127, 340-344, 357,
cognitive models of, 339-343, 371-373, 378-380, 418-419
349, 369, 378, 410-411,419 cohesion, 10, 319, 357
454 Index of subjects
cohesive link, 18 discourse topic, 343, 345, 349,

co-interpretation, 9, 21, 29, 80, 353, 375, 418-419
169, 410 EFFECT, 91, 102-110, 118, 362
collocation, usage of the term, emphatic focus, 333, 336
410-411 entrenchment, 39, 412, 420
collocational versatility, 68 epistemic modality, 33, 97, 112,
comparative reference, 239,319, 192, 232-244, 248, 251, 255,
347, 411 336-337, 416-417
compiled reliance, 53, 55-56 EVENT, 251-255, 258-262, 266,
complementation, 268, 277, 280, 283, 285
of shell nouns, 4,21-24,209, existential-rftm?, 24-25, 250, 357
293, 379 expanded predicate, 25, 71, 72,
of verbs, 31-35 105, 196, 209, 227, 248, 318,
complementizer, 152, 332 417
concept-formation, 14, 17, 19, 37, EXPERIENCER, 142, 185-200, 203,
41, 64, 80, 119, 303-304, 351, 210-229, 342, 416
361, 368, 378 extended reference, 11, 18, 23, 28,
conceptual partitioning, 276, 355, 72, 96, 264, 339
360-363, 370 extraposition, 11,24
conceptual status of entities, 67-68,
72, 85, 99, 142, 243, 312-316,
327, 331, 362, 377 FACT, 236
concordance, 47, 189, 190, 254, factivity, 233
260, 319, 415 of nouns, 98-100,227,263
Construction Grammar, 257, 298 of verbs, 32
container noun, 5 , 1 1 factuality, 127, 233, 242, 414
contrastive focus, 333, 336, 359 feature, semantic, 66, 74-78, 87
conversation, 8, 36, 42, 43, 106, figure, and ground in Cognitive
142, 143, 153, 201, 346, 358 Grammar, 70, 307, 364, 367,
conversion, 148 414
corpus query, 44, 46-52, 104, 108, focalization, 333
116, 254, 287 focus, 306-307
corpus tag, 44, 50 of attention, 118,135,331,
339, 359
focus construction, 307, 332-334,
de dicto, 35, 298 338
de re, 35, 298 focus formula, 95, 122, 128
deictic function, 17, 276, 282 focusing
deixis, 29, 351 function, 135, 237, 306, 319,
deontic modality, 192, 232-235, 329-338, 378, 380, 414
244-251, 257-259, 270-271, function of the pattern N-e-cl,
293, 298, 310, 311, 379, 417 37, 97, 140, 355
discourse representation, 29, 372, formal coincidence, 48-51
412 frame, as a descriptive device, 29,
88-95, 414
Index of subjects 455
frame shifter, 350 of shell nouns, 14-15,18-21,

French, 11,102,148,347 304, 319, 338-348, 355, 378
fronting, 334 of the pattern th-N, 36, 343-348
full-content noun, compared to linking verbs, 22-23, 27, 338
shell nouns, 15-18 locution, 105, 131
FUTURE ACTIVITY, 185,244,246 low-content noun, 6
gap, structure-inherent semantic, manipulation

73, 76-80, 86, 263, 271-275, of discourse participants, 8, 317
343, 345, 373, 377, 414 of the conceptual status of
general noun, 6, 10-11, 74, 78, entities, 72-75, 243, 327, 377
276, 290, 324, 409 manipulative verb, 35
German, 11, 102, 128, 419 mental model, 29, 372, 412
gestalt, 364, 366, 373, 376, 419 mental space, 29, 315-316, 360,
gestalt principles, 376 412
given information, 7, 37, 127,
metaphor, 9, 13, 71, 76, 121, 146-
304, 305, 330, 331
147, 188-190, 316, 369, 415,
Givenness Hierarchy, 339-341,
418
345, 418
modality, 231-260, 271, 293, 298,
grammatical metaphor, 70-72, 369,
379, 416-417
414
MODALITY INDICATOR, 235-236,
244, 251-252, 256, 258, 260
IDEA, 136-138, 142, 187-194, 212,
224
new information, 10, 37, 126,
identifying function of the pattern
304-305, 312, 330-331, 340,
N-e-cl, 27, 31, 132, 200, 246,
351, 370, 418
263, 409
identity of reference, 27-28,30, nominalization, 65, 148, 364-367
281, 411
illocution, 155-182
o/-prepositional phrase, 31,68,
imputed strength of commitment,
108, 204, 219, 229, 253, 268,
197-198
287, 292, 380
inferrable information, 18, 341
with partitive uses, 116-120
insertions, 48-51
with rogative uses, 167-168
Italian, 102
with conceptual uses, 190-192
ontological status, 67,111
ORIGINAL HEARER, 134, 149, 166,
label, 12-13, 17, 75, 310
Latin, 79, 148, 224 170, 174, 176-180, 184-185
left-dislocation, 334 ORIGINAL IDEA, 184, 187
lexical cohesion, 18, 264, 346 ORIGINAL SPEAKER, 133-148, 162-
lexical unit, 87 166, 170-185, 201, 308, 342
linking function,
456 Index of subjects
ORIGINAL UTTERANCE, 133-135, reporting, 71, 131-136, 141, 153,

141-142, 153-154, 162-180, 161-162, 167, 186, 245, 292,
184, 416 309, 416
ouvertures, 8 restrictive clauses, 23, 346, 333-
336, 347
rheme, 37, 99, 126, 304-306
paragraphs, shell nouns at the sanctioning, 95, 338
beginnings of 349-353 saturation of discourse entities, 67,
PART, 117-118, 120 79, 284, 361, 362
partitioning, conceptual, 276, 355, scanning, 364-365
360-363, 370 schema, 29,412,415
path, 90, 121 script, 29, 412
picture, corpus program 47 second-order entities, 41,64-70,
postmodifier, 48-49, 52, 104, 125, 73, 80, 149, 158, 171, 264
226 semantic match between shell nouns
postmodifying clause, as opposed and contents, 124, 275, 343-
to complement 23 346, 378
postnominal clause, discussion of semantic prosody, 289, 324
the term 21-26 shell-content complex, definition
premodifier, 7, 46, 240, 286, 287, of, 8
288, 317, 325, 335, 359 shell-noun phrase, definition o f ,
presupposition, 57, 195,203,411, 7-9
418 SIGN, 91, 110-113
profile, in Cognitive Grammar signalling, lexical 10, 13, 348,
135, 142, 364-367 353, 358-359, 419
projection, 71, 85, 132, 141, 186, signpost, 10, 349-350, 355-359,
314 369, 375
proposition, 65-68 sincerity condition, 185
PROPOSITION, 235-236, 242 situation model, 29, 412
PROTASIS, 289-291 space builder, 315,360
PSYCHOLOGICAL STATE, 185-189, SPEAKER, 133-138, 140, 149, 153-
195, 203-228, 234, 266 154, 162-166, 170-174, 177-182
speech act, 10, 50, 69, 132-138,
145-186, 202, 234, 245, 308,
re-activation, 7, 28, 372 318, 366, 416
realisation, of signposts 75-76, speech act verb, 10, 69, 147-148,
349, 356-358 151-153, 168, 181, 416
realism, 16, 64, 368 STATE, 262,265,277,417
reference to fact, 23 structural semantics, 20
referential linkage, 30 style, 43, 63, 108, 113, 145, 152,
register, 113 241, 252, 319, 322, 344, 379,
reification, 17, 66, 34, 214, 360, 415
363-370, 376-377, 419 subjectivity, 35, 193, 200-206,
relative clause, 3, 23, 26, 50, 122, 234, 336-337, 359
125, 255, 275, 278, 281, 417 substitution view of anaphora, 27
reliance, definition of 53-56 suffixation, 148
Index of subjects 457
Systemic-Functional Grammar, 12, universe-of-discourse, 29, 412

20, 27, 70, 171 unspecific noun, 10, 70, 75, 93,
263
unspecificity of shell nouns, 6, 9-
tentativeness, 162, 200-202, 206, 10, 63, 74-80, 85-88, 93, 338,
240, 336-337, 359 409, 420
textdeixis, 351,419 UTTERANCE, 134-139, 141, 145,
text reference, 18, 23, 28 151, 153, 165, 170, 177, 184,
textual world, 29,412 189
thematic structure, 352 utterance launchers, 8
theme, 37, 126, 189, 193, 304-
307, 418
third-order entity, 41,63-66,73, verbal complementation, 31-36,
80, 85-86, 111, 119, 149, 158, 293
171, 414
topic, 304-307, 312, 329-334,
343, 349-352, 359, 371-375, wA-clause, 3, 22, 23, 125, 168,
380, 418-419 169, 271, 381
topic boundary marker, 350, 359 word class, 17, 70, 363-368
topic change marker, 350 world-creating predicate, 316
topicalizing, 9, 329, 334-338, 378
Transformational Grammar, 5-6,
24, 32-33, 35, 261, 412 zero-derivation, 148
typicality, degrees of 14, 83-87,
118, 120, 298-299, 378
Topics in English Linguistics
Edited by Bernd Kortmann and
Elizabeth Closs Traugott
Mouton de Gruyter Berlin New York
1 Niels Davidsen-Nielsen, Tense and Mood in English. A Comparison with Dan-

ish. 1990.
2 Historical English Syntax. Edited by Dieter Kastovsky. 1991.
3 English Computer Corpora. Selected Papers and Research Guide. Edited by
Stig Johansson and Anna-Brita Stenstrm. 1991.
4 Donka Minkova, The History of Final Vowels in English. The Sound of Muting.
1991.
5 Lia Korrel, Duration in English. A Basic Choice, Illustrated in Comparison
with Dutch. 1991.
6 Andreas H. Jucker, Social Stylistics. Syntactic Variation in British Newspa-
pers. 1992.
7 Ken-ichi Takami, Preposition Stranding. From Syntactic to Functional Analy-
ses. 1992.
8 Bas Aarts, Small Clauses in English. The Nonverbal Types. 1992.
9 New Directions in English Language Corpora. Methodology, Results, Soft-
ware Developments. Edited by Gerhard Leitner. 1992.
10 History of Englishes. New Methods and Interpretations in Historical Linguis-
tics. Edited by Matti Rissanen, Ossi Ihalainen, Terttu Nevalainen and Irma
Taavitsainen. 1992.
11 Early English in the Computer Age. Explorations through the Helsinki Corpus.
Edited by Matti Rissanen, Merja Kyt and Minna Palander-Collin. 1993.
12 Towards a Standard English: 1600-1800. Edited by Dieter Stein and Ingrid
Tieken-Boon van Ostade. 1993.
13 Studies in Early Modern English. Edited by Dieter Kastovsky. 1994.
14 Ronald Geluykens, The Pragmatics of Discourse Anaphora in English. Evi-
dence from Conversational Repair. 1994.
15 Traute Ewers, The Origin of American Black English. e-Forms in the HOO-
DOO Texts. 1996.
16 Ilse Depraetere, The Tense System in English Relative Clauses. A Corpus-
Based Analysis. 1996.
17 Michiko Ogura, Verbs in Medieval English. Differences in Verb Choice in
Verse and Prose. 1996.
18 Spanish Loanwords in the English Language. A Tendency towards Hegemony
Reversal. Edited by Flix Rodriguez Gonzles. 1996.
19 Laurel J. Brinton, Pragmatic Markers in English. Grammaticalization and
Discourse Functions. 1996.
20 Christiane Dalton-Puffer, The French Influence on Middle English Morphol-
ogy. A Corpus-Based Study on Derivation. 1996.
21 Johan Elsness, The Perfect and the Preterite in Contemporary and Earlier
English. 1997.
22 Carl Bache and Niels Davidsen-Nielsen, Mastering English. An Advanced
Grammar for Non-native and Native Speakers. 1997.
23 English in Transition. Corpus-based Studies in Linguistic Variation and Genre
Styles. Edited by Matti Rissanen, Merja Kyt and Kirsi Heikkonen. 1997.
24 Grammaticalization at Work. Studies of Long-term Developments in English.
Edited by Matti Rissanen, Merja Kyt and Kirsi Heikkonen. 1997.
25 Axel Hbler, The Expressivity of Grammar. Grammatical Devices Expressing
Emotion across Time. 1998.
26 Negation in the History of English. Edited by Ingrid Tieken-Boon van Ostade,
Gunnel Tottie and Wim van der Wurff. 1998.
27 Martina Hcker, Adverbial Clauses in Scots: A Semantic-Syntactic Study.
1998.
28 Ingo Plag, Morphological Productivity. Structural Constraints in English Der-
ivation. 1999.
29 Gustav Muthmann, Reverse English Dictionary. Based on Phonological and
Morphological Principles. 1999.
30 Metaphor and Metonymy at the Crossroads. A Cognitive Perspective. Edited
by Antonio Barcelona. 2000.
31 Generative Theory and Corpus Studies. A Dialogue from 10ICEHL. Edited
by Ricardo Bermdez-Otero, David Denison, Richard M. Hogg and C. B. Mc-
Cully. 2000.
32 Manfred G. Krug, Emerging English Modals. A Corpus-Based Study of Gram-
maticalization. 2000.
33 Cause - Condition - Concession - Contrast. Cognitive and Discourse Per-
spectives. Edited by Elizabeth Couper-Kuhlen and Bernd Kortmann. 2000.
34 Hans-Jrg Schmid, English Abstract Nouns as Conceptual Shells. From Cor-
pus to Cognition. 2000.

English Abstract Nouns As Conceptual Shells, From Corpus To Cognition

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

English Abstract Nouns As Conceptual Shells, From Corpus To Cognition

Încărcat de

Drepturi de autor:

Formate disponibile

English Abstract Nouns as Conceptual Shells

G Berlin New York 2000

Printed on acid-free paper which falls within the guidelines of the

Als Habilitationsschrift auf Empfehlung der Philosophischen Fakultt fr

Die Deutsche Bibliothek CIP-Einheitsaufnahme

Copyright 2000 by Walter de Gruyter G m b H & Co. K G , D-10785 Berlin

This book would not be what it is without the support of a number of

Part I Foundations: Theory, terminology and methodology 1

Part II The use of shell nouns 83

6. Describing shell-noun uses 85

6.2 Explaining the meanings of shell-noun uses: features

12. Circumstantial uses 275

Partili Functions of shell nouns 301

14. Introduction to Part III 303

Foundations: Theory, terminology and

(1.1) (a) Determiner + (Premodifier) + Noun + postnominal that-c\ms&, wh-

(1.2) (a) *The boy that I had no money ...

Table 1.1 Examples of shell nouns

Shell nouns make up an open-ended functionally-defined class of abstract

(1.3) The Government's aim is to make GP's more financially accountable,

a) Theoretical and methodological perspectives: How can shell nouns and

2.1 The term shell noun

Ivanic (1991), who speaks of carrier nouns, exploits a related image.

2.2 Defining shell nouns and shell-content complexes in functional

This functional definition of shell nouns has a number of fundamental

strate this, it will be helpful to compare shell nouns to full-content nouns on

(2.1) (a) Full-content nouns: teacher, cat, journey

terminers with nouns specify the reference of particular expressions. Neither

Figure 2.1 The converging balance of shell nouns

as will be shown in greater detail in Chapter 5, certain types of nouns lend

2.3 A brief note on the theoretical stance

This study is not consistently set within a single theoretical framework. As

This chapter looks more closely at a number of syntactic, semantic and

3.1 Triggering co-interpretation

3.1.1 Lexico-grammatical patterns of shell-noun uses

Pattern Abbreviation Example of the general pattern

Shell noun + postnominal N-cl (3.1) Mr Bush said Iraq's leaders

Figure 3.1 Lexico-grammatical patterns favoured by speakers for the use of

(3.5') To stop and think about it is a good idea.

(3.6') The speculation is that he might move into politics.

These examples are not particularly good examples of shell-content com-

Similar affinities to simple verbs can be observed with occurrences of

(3.7) Britain is a great place to live and work in ... (MAGS)

Traditionally the infinitive clauses in such examples as (3.7) have been

(3.7') a great place in which/where it is possible to live and work

Cases of this type tend to be analysed as variants of relative clauses as well.

seem to share a syntactic or semantic basis which could be used to explain

3.1.2 From identity of reference to experiential identity

reality, but only whether it is represented in the cognitive models of situa-

3.2 The semantic contributions of different types of complements:

(3.10) Joan is likely to get married.

is symptomatic, for example, that in MenzePs (1975) study of the semantics

THAT complements can be shown to be derived from either SAY clauses

The infinitive complement, on the other hand

is associated with a personal, subjective, first-person mode: want',

3.3 Basic functions of shell-noun typical patterns

wish to characterize a piece of information in a certain way, in (3.15), for

4.1 The From-Corpus-to-Cognition Principle

Finally, for non-native students of a language a corpus acts both as

Frequency in text instantiates entrenchment

The upshot of this principle, which provides the methodological foundation

a) an objectively controlled set of data, which consists of

4.2 Data retrieval