Foundations of pragmatics and lexical semantics 9783111405933, 9783111042480

Citation preview

Foundations of Pragmatics and Lexical Semantics

Groningen-Amsterdam Studies in Semantics (GRASS) T h i s series of b o o k s on the semantics of natural language c o n t a i n s c o l l e c t i o n s of o r i g i n a l research o n selected t o p i c s as well as m o n o g r a p h s in this area. C o n t r i b u t i o n s f r o m linguists, philosophers, logicians, c o m p u t e r - s c i e n t i s t s a n d c o g n i t i v e p s y c h o l o g i s t s are b r o u g h t t o g e t h e r to p r o m o t e interdisciplinary and international research. Editors A l i c e ter Meulen Martin S t o k h o f

Editorial Board Renate Bartsch University of Amsterdam J o h a n van B e n t h e m University of Amsterdam Henk V e r k u y l University of Utrecht

Other books in this series: 1. Alice G.B. ter Meulen (ed.) Studies in Mode/theoretic Semantics 2. Jeroen Groenendijk, Théo M.V. Janssen and Martin Stokhof (eds.) Truth, Interpretation and Information 3. Fred Landman and Frank Veltman (eds.) Varieties of Formal Semantics 4. Johan van Benthem and Alice ter Meulen (eds.) Generalized Quantifiers in Natural Languages 5. Vincenzo Lo Cascio and Co Vet (eds.) Temporal Structure in Sentence and Discourse 6. Fred Landman Towards a Theory of Information

All c o m m u n i c a t i o n s to the editors can be sent to: D e p a r t m e n t of P h i l o s o p h y or D e p a r t m e n t of Linguistics, G N 40 University of A m s t e r d a m University of W a s h i n g t o n G r i m b u r g w a l 10 Seattle, W a s h i n g t o n 98195 1012 GA A m s t e r d a m U.S.A T h e Netherlands

Jeroen Groenendijk, Dick de Jongh, Martin Stokhof (eds.)

Foundations of Pragmatics and Lexical Semantics

¥

1987 FORIS PUBLICATIONS Dordrecht - Holland/Providence - U.S.A.

Published by: Foris Publications Holland P.O. Box 509 3300 AM Dordrecht, The Netherlands Sole distributor for the U.S.A. and Canada: Foris Publications USA , Inc. P.O. Box 5904 Providence RI 02903 USA CIP-DATA Foundations Foundations of Pragmatics and Lexical Semantics / Jeroen Groenendijk, Dick de Jongh, Martin Stokhof (eds.). - Dordrecht [etc.]: Foris. - (Groningen-Amsterdam Studies in Semantics; 7) ISBN 90-6765-264-4 bound ISBN 90-6765-265-2 paper SISO 805.5 U D C 801.5 Subject heading: semantics (linguistics).

ISBN 90 6765 264 4 (Bound) ISBN 90 6765 265 2 (Paper) ® 1986 Foris Publications - Dordrecht. No part of this publication 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 from the copyright owner. Printed in the Netherlands by ICG Printing, Dordrecht.

Table of Contents

Preface Renate Bartsch Context-dependent Interpretations of Lexical Items

VII

1

Alice ter Meulen Locating Events

27

Michael Moortgat Compositionality and the Syntax of Words

41

Ken-ichiro Shirai Japanese Noun-phrases and Particles wa and ga

63

Thomas Ede Zimmermann Transparent Adverbs and Scopeless Quantifiers

81

Godehard Link Prespie in Pragmatic Wonderland or: The Projection Problem for Presuppositions Revisited

101

Fred Landman Paradoxes of Elimination

127

Preface

This volume contains a selection of the papers that were presented at the Fifth Amsterdam Colloquium, which took place in August 1984. The papers collected in this book center around two themes: lexical semantics, and pragmatics. Contributions to the first theme address both more theoretical, foundational issues, as well as matters of description, and some papers do both. Bartsch's paper pursues an approach to lexical meaning that relies heavily on context to determine various 'dimensions' of interpretation of lexical items. Her paper is a classification and an investigation of these dimensions, and it contains, moreover, a description of the meaning of various classes of adjectives in Dutch. The paper by Ter Meulen gives an analysis of the way in which events are located in space and time, using the framework of Situation Semantics. She discusses such problems as the relation between an event and the individuals participating in it, matters which are highly relevant for an account of temporal and aspectual properties of verbs and the like. Moortgat's contribution is to the study of productive word formation processes within the Montague grammar framework. His goal is to maintain compositionality as a constraint on interpretation, also in morphology, while at the same time keeping lexicon and syntax as separate components in the grammar. To this purpose, he adds rules of functional composition, thus underscoring the need for and the usefulness of a more flexible approach than 'orthodox' Montague grammar allows. The paper by Shirai is concerned with the content of the Japanese particles wa and ga. His claim is that a proper description of their content and function can be arrived at only if we allow epistemological concepts, such as 'well-knownness', to enter into our semantic vocabulary. Zimmermann's paper deals with meaning postulates as a means to connect lexical and logical semantics. It contains a detailed examination of various proposals for a particular meaning postulate (which is to ensure that certain adverbs are transparent), and it derives some general methodological morals from this that extend beyond the particular case discussed. To the second theme, the foundations of pragmatic theory, two papers are devoted explicitly (though various pragmatic factors also play a role in other papers, e.g. those by Bartsch, ter Meulen and Shirai). Link's contribution deals with an 'age-old' problem in the theory of presuppositions: the projection problem, i.e. how compound expressions inherit the presupposi-

VIII

Preface

tions of their components. Side-stepping the heated debate whether this is a matter of truth-conditions or of conditions of use, Link argues that it is both: a semantic notion of presupposition combined with a notion of admittance by context and general information processing strategies will do the job. The paper by Landman, finally, deals with another central concept of pragmatics: information. Using certain informational paradoxes, such as Conway's paradox and the Hangman, Landman critically discusses various theories of information representation developed within the possible worlds framework. From this short indication of the contents of the various contributions it may be clear that they share a common theoretical and philosophical interest in the foundations and applications of lexical semantics and pragmatics, yet that they also display a wide variety of approaches and frameworks. 'Logical' semantics in the broad sense of that word is no longer tied to a particular framework or set of principles, as it was in the stage of its inception. Rather it is a many-coloured thing. We hope that the papers in this volume may help to convince the reader that this constitutes a virtue, rather than a vice. A companion to this volume, containing various papers read at the colloquium which deal with the theory of generalized quantifiers and the theory of discourse representation, appears as GRASS 8. The Amsterdam Colloquium was organized by the Department of Philosophy of the University of Amsterdam, with financial support of the Department of Mathematics of the same university and of the Royal Dutch Academy of Sciences, which is gratefully acknowledged. Finally, the editors would like to thank the authors for their cooperation and patience in bringing this volume into being. The editors

Chapter 1

Context-dependent Interpretations of Lexical Items Renate Bartsch

This paper is about the semantics of certain kinds of expressions like good, strong, satisfactory in English, and flink 'm Dutch, which I call 'thematically weakly determined expressions'. They require a specification as to which respect of qualification they apply. At the same time, expressions that provide these respects are dealt with, such as predicate limiting adverbials like healthwise, financially, with respect to his health, and term modifiers like as a teacher in, for example, John as a teacher is good, which will be shown to be equivalent with John is good as a teacher. Likewise John's health is good will come out equivalent with John does well healthwise. In the semantics, I shall employ Kaplan's distinction between 'character' and 'content'. - Further, the paper includes some considerations about how the thematic dimensions or respects, in which the dimensionally weakly determined expressions are interpreted, are provided within a text.

1. CONTEXT AND CONTEXT-DEPENDENT SYNONYMY

The notion of context I employ here, comprises the interpretation of the linguistic context as well as the situational context. The linguistic context describes, or establishes, one or several reference situations, which are possible situations of satisfaction of a sentence and only in special cases coincide with the speech situation. In the linguistic context, we find expressions that refer deictically or by means of co-reference restrictions to either, parts of the speech situation, or of the reference situations. In the interpretation of a text, we interpret expressions in, what we can call, 'dimensions' of the speech situation or of the reference situations. The dimensions are place, time, speaker, hearer, third persons or entities like groups, actions, events. These I call deictic dimensions, because deictic expressions and pronouns are interpreted with respect to them. Next to these dimensions for interpretation, the context provides a 'thematic dimension': At each moment of a text, there is at least one theme, namely what the text is about, in the sense of, to which goal of the speaker or hearer this part of the text is directed. In texts that are directed towards providing information, this theme can be grasped in a thematic question 1 : The text continues in order

2

Renaie Bartsch

to answer this question that the speaker expects the hearer to want to get answeied. The thematic question defines a thematic dimension within which the following text fills in a point, i.e. gives a specification. The thematic dimension, in this way, can be defined as the set of all possible answers to a thematic question. For the kind of expression we deal with in this paper, the thematic dimension will be a set of properties that are the possible specifications under one single respect. All the dimensions that are established for the interpretation of a part of a text define the limits within which the interpretation of expressions has to be located. The context can be represented by its deictic and thematic dimensions: C = (place, time, speaker, hearer, third entities, theme) The dimensions the context exists of are identified parts of the model, i.e. semantic entities. If not stated otherwise, the notion 'context' will be used for this part of the model, instead of for the linguistic context, also called 'co-text" in text-linguistics. In what follows, I refer to the thematic dimension of the context by the index d. The disambiguation of polysemic expressions, e.g. run, takes place mainly by compatibility or incompatibility with properties of the reference situations. The same holds for homonymous expressions. But there are other polysemic words that are more dependent on the thematic dimension. Such a subclass of polysemic words are the dimensionally weakly determined expressions, such as flink in Dutch. Depending on the thematic dimension they are used in, they have a different content. The general notion expressed by flink is something like 'strong under aspect X'. The thematic dimension fills in the X. In Dutch dictionaries, we find a whole set of expressions that are contextually synonymous with flink: dapper, volhardend, dik, zwaar gebouwd, sterk, belastbaar, van grote omvang, van goede gezondheid, etc. ('brave', 'enduring' or 'persistent', 'big', 'strongly built', 'strong', 'can carry burdens mentally, or physically', 'voluminous', 'of good health', etc.). Between flink and these expressions, there exists a context-dependent synonymy: [flink]di, i [flink] d i, 2 [flink]d3 [flink]d4

= [dapper]di, i = [volhardend]di,2 = [dik]d3 = [sterk]d4 etc.

The dimensions exemplified here are d\,\ \ degree of readiness to get into possibly adverse situations for the sake of something good; d\t2'. degree of

Context-dependent

Interpretations of Lexical Items

3

endurance in adverse situations; d3: size of volume or circumference; d4: degree of physical ability and strength. The different synonyms of flink in the different context-types represented here by their thematic dimensions, are dimensionally stronger determined than flink. Dimensionally strongly determined words have the thematic dimension included in their meanings, in a presuppositional way. This means that this part of their meaning, the internal thematic dimension, is not negated or questioned or modified by simple negation or questioning or by modal modification. There are degrees in how specific a thematic dimension is presupposed by the content of the expression: In dimensionally weakly determined ones, the specification is close to zero, and has to be given in the contextually provided thematic dimension; in dimensionally strongly determined expressions the thematic dimension is specific to mure or less a degree in the content of the expression. Thus good, strong, great, heavy, more, less, do well, be successf ul are dimensionally weaker determined than brave, courageous, humourless, liberal, broadchested. When dimensionally strongly determined expressions aie used, the thematic dimension given in the context has to co-incide with, or to be included in the internal dimension of the expression. Thus, dapper is interpretable in c/1,1, but not in d3. In [dapper]d3 we have a contradiction between the internal thematic dimension, and the one given externally by the context. Generally, to resolve this conflict, the internal dimension is taken to supersede the external one, i.e. the internally coded thematic dimension replaces the contextually given one. Then the text gets incoherent on the level of the established thematic dimensions, but the hearer might look for coherence on another level of interpretation by conversational implicature. This indirect interpretation could be that the speaker wants to change the topic because of some reason, and just this reason could be relevant. Also a metaphorical interpretation would be possible, in which case the externally provided thematic dimension would be kept: [dapper]^3 could be interpreted metaphorically as meaning 'thick', were it not that for this concept Dutch has already the word dik as dimensionally strongly, and the word flink as dimensionally weakly determined expressions. Note, that also disambiguation of homonymous expressions has to be treated with respect to thematic dimensions. For example, German unterhalten in Maria unterhält Hans can mean 'entertain' or 'support', depending on whether the topic is social behaviour at a party, or financial relationships between people. Another example is German gleich, in Der Bau dieser Häuser wird gleich ausgeführt, meaning the construction of these houses will be performed in the same style, or the construction will be performed immediately, depending on whether the thematic question refers to style, or to timing.

4

Renate Bartsch

2. DEPENDENCE ON THEMATIC DIMENSIONS A N D KINDS OF OBJECTS

In the interpretation of a thematically weakly determined predicate, we have at least one object referred to and a thematic dimension under which the object is characterized by the predication. The interpretation of the predicate depends on both of them. Consider the objects denoted by jongen, 'boy', borrel 'drink', boek 'book', and redevoering 'speech'. With respect to the dimensions, we have a rough distinction between character, physical appearance, and physical strength. For different kinds of objects, these dimensions are filled somewhat differently, for example, flink with respect to the pair character and people means 'brave', 'enduring', 'persistent', while flink with respect to the pair character and drink means 'high in alcohol content'. 1. een flinke jongen

w.r.t. w.r.t. w.r.t. w.r.t. 2. een flinke borrel w.r.t. w.r.t. w.r.t. 3. een flink boek w.r.t. 4. een flinke redevoering w.r.t. w.r.t. w.r.t. w.r.t.

character: volume: posture: physical ability: character: volume: character: volume: character/content: length: voice volume: self-presentation:

courageous, brave big strongly built strong strong liquor big glass of liquor daring content big, thick courageous long loud boasting

3. CONTEXTUALLY INTRODUCED A N D EXTERNALLY CODED DIMENSIONS

The most direct ways of expressing the thematic dimension are questions and predicate limiting adverbials. a.

b.

How does John do financially? He does well. How is John's business? He does well. Does John have problems because of the budget cuts? No, he does well. John does well financially. John's paper is stylistically bad, but methodically good. He is mentally strong, but physically weak. As far as presentation goes, his paper is good, but with regard to results, it is poor.

Context-dependent Interpretations of Lexical Items

5

Like the internally coded thematic dimension in dimensionally strongly determined expressions, a contextually explicit thematic dimension is presupposed information, and not asserted information. A. In unspecified negation or questioning, or in modal modification, the predicate limiting adverbial is not negated - this in opposition to, for example, manner adverbials: 1.

John does well financially. - No. ( = He does not do well in this respect.)

This shows that the dimension expressed in the adverbial is retained in simple negation. This is just the opposite with manner adverbs: 2.

John runs fast. - No. ( = He runs, but not fast.)

B. Predicate limiting adverbials do not allow for a paraphrase in which they are predicated - this again in opposition to manner adverbials: 1. 2.

John's paper is scientifically good. ^ T h e goodness (quality) of John's paper is scientific. John's paper is excellently composed. = The composition of John's paper is excellent.

C. Also the inference properties are opposite for both kinds of adverbials: 1. 2.

John's paper is stylistically good. * John's paper is good. John's paper is excellently written. -» John's paper is written.

If the relevant aspects are only stylistic quality, for example in a training course for writing compositions, then John's paper is stylistically good is equivalent with John's paper is good. The implication only holds if the thematic dimension indicated by the predicate limiting adverbial comprises the thematic dimension of the context, i.e. comprises all aspects from the context, relevant for evaluation. When we ask a friend whether he does well, we mean, without further specification, whether he does well in all respects normally considered relevant: healthwise, economically, concerning his position, concerning his family. He does well normally means 'well in all relevant respects'. 2 Other examples of predicate limiting adverbials are the dimensions for grading length, width, weight, temperature, etc. They can be explicitly coded, or internally coded in the meaning of the relative adjective. The dimensionally weakly determined predicates are the relations more, less, or the predicates much, little, high, low:

6

Penate

Bartsch

[more in length] = [longer] [more in weight] = [heavier] [high in intelligence] = [intelligent] [low in intelligence] = [unintelligent] [much in weight] = [heavy] [little in weight] = [light] [high in temperature] = [hot] [low in temperature] = [cold] In the left handside expressions, the dimension is externally coded, in the right hand side it is internally coded, and there we have dimensionally strongly determined adjectives.

4. CHARACTERS AND CONTENTS; MEANINGS AND SENSES

Although predicate limiting adverbials, at first sight, syntactically look as if they were of type < , with c for 'context' and 5 for 'world-time'-pairs. The predicative adjective good or the predicative expression do well would be of that type. A predicate limiting adverbial, then, is of type < >. Let G be the interpretation function that assigns characters to basic expressions. Let d be the index for thematic dimensions in the semantic metalanguage. With the expression a j of the object-language basically denoting the dimension d\, the predicate limiting adverbial a, is two types higher. We then have the following interpretation in the model (£, with g as variable assignment, w, world, i,time: ftti Adv lim good Pred Char]®' d> W> * = gOOdG(^i) (w, t) = G (good) (di) (w, t) Dimensionally weakly determined expressions are unstable characters, dimensionally strongly determined expressions are stable characters. In the case of a stable character, the character determines the same content for each context: For all di, dr- [a] e - * (di) = M Q * (dj). Note that, if d\ is internally coded within the meaning of a, we have:

Context-dependent

Interpretations

of Lexical

1

Items

For all dr. [a]®- « (d\) = [a]®- s (dx). For basic expressions, stability of the character amounts to: For all dit d\: G(a) (d,) = G(a) (d}). For basic expressions with a stable character, we thus can use the interpretation function F that assigns to basic expressions a content, or 'sense' in the terminology of Montague's 'Universal Grammar': For all di: G(a) (d\) = F(a). The interpretation function G assigns a character to an expression. In Montague's 'Universal Grammar', it would be a 'meaning', i.e. a function from pairs of contexts and world-time pairs to denotations, in distinction to a 'sense', i.e. a function from world-time pairs to denotations. In what follows, I shall sometimes use the term 'meaning' for Kaplan's 'character', and 'sense' for Kaplan's 'content'. In our examples, the senses or contents are mostly concepts, i.e. senses of predicate expressions. In Kaplan (1979), demonstratives have an unstable character, but this kind of character applied to a context gives a stable content, i.e. a stable function from world-time pairs to denotations. [a]®, c. 8 is a stable content if and only if: c For all w, w', t, t': -s (w, t) =

8 (w',

t').

Logically, proper names have a stable character and a stable content. Proper names in natural language are really context-dependent and therefore have unstable characters, but stable contents. Non-homonymic, nonpolysemic, and dimensionally strongly determined expressions that are not demonstratives, are of stable character and unstable content. Homonymic, polysemic, and dimensionally weakly determined expressions are of unstable character and unstable content. Even words like dog, cat, house do not have a fully stable character or meaning. The expression dog in some contexts is interpreted as real dog, and in others as toy dog. Thus, the sentence Some dogs can bark is an analytic truth about real dogs, but provides interesting information about the world of toys. Here, another context index comes into play, namely the discourse domain as part of the context.

8

Renate Bartsch

5. FORMAL TREATMENT OF RESTRICTIONS OF PREDICATE INTERPRETATIONS

First I shall give an example of the interpretation of a sentence with a dimensionally weakly determined predicate, with respect to a thematic dimension d and a world-time pair /. The predicate is the Dutch adjective flink, which means approximately 'strong under aspect X. The noun jongen has a stable character. 3 Een jongen is flink 'A boy is strong under aspect X' [Een jongen is flink]®- d- '• = [een jongen]®- d- s ("flink®- d- •> •?) = 1, if and only if: G(flink) (d) e { P) There is g'= g: G(jongen) (d) (/) (g'(x)) = 1 & P(i) (»'(*)) = 1) = There is g'r g: F(jongen) (/') (g '(x)) = 1 & G(flink) (d) (/) (g'(*)) = 1. Suppose, d = d\,\ from section 1 of this paper. Then we have that the above sentence is true, if and only if: There is g' = g: BOY (/') (g '(x)) = 1 & COURAGEOUS (/') (g '(x)) = 1. EXAMPLE:

If we want to make the interpretation of flink also dependent on the property expressed by the description of the entity to which it is applied, we have to take the character of the predicative expression to be a function from pairs of dimensions and properties to contents: [flink]®' W- G(jongen) (d)) e (PI ... ) For d = rfi,i we have then: G(flink) (di,u BOY) = COURAGEOUS e (P I ...) If, instead of BOY we would have DRINK, then the property expressed by flink under d\ would be STRONG IN ALCOHOL CONTENT. I now shall give a formal treatment of the specifications expressed by predicate limiting adverbials, and with that I shall introduce the notion of thematic intensionality. Analogously with modal operators and time operators, both also expressed by sentence adverbials, there are thematic operators, expressed by predicate limiting adverbials, for example: in every respect, in some respect, financially, as far as his health is concerned, qua content, etc. When these operators occur in a sentence, they either further specify the thematic dimension, like in John is courageous in every respect, John is courageous in some respects, John has a fine character in all respects, or they dominate the thematic dimension given in the context. In both cases, they create thematic intensionality of the expression within their scope; this

Context-dependent

Interpretations of Lexical

Items

9

means that its interpretation does not depend on the thematic dimension of the context.

1. SPECIFICATION

flink dr /v^

d.

d2

(character)

d.„

/

di.i

rv

(with respect to risk taking)

di,2 (with respect to endurance)

d.,.,. (with respect to dogs)

The meaning of courageous, for example, has internally encoded the dimension di,\. So, courageous in every respect is of the level of specification where i ranges over all respects relevant. The properties in the -range are mostly lexically expressed, the ones in the d.,.-range are to some extent lexically expressed, and the ones in the d.,.,.-range are hardly ever lexically expressed. The reason for this seems to be, that the lexical items in the less specific dimensions serve also in the more specific ones, where the more specific content is expressed by a predicate limiting adverbial, or specified in the context. As the above hierarchy indicates, thematic dimensions can stand to each other in the relationship of specification (i.e. hyponomy) and contrariness. If d\ is further specified by in every respect, then the interpretation takes place with respect to all d\,,; that means, for all specifications of d\. Thus, interpretation with respect to d\ ( = character), or with respect to di ( = physical appearance) is vague to some extent. 4 The concept, or sense of flink at di is in fact a non-exclusive disjunction of the concepts G(flink) ( d i j ) for different j. Giflink) (di), then, is: THICK or HEAVY BONED or TALL or BROAD, which are the properties at c/2,1, di,i, d2,i, and dz,4, respectively. The sign for thematic intensionality is '*', which works analogously with ' in intensional logic. We have: \

d

= For some d,: G(flink) (di).

This is a disjunction o f properties, expressible by strong

under

some

respects. Generally, specification runs as follows: For d the specifications are di, for dj the specifications are dj,,, and generally dx,, for dx as the contextually provided thematic dimension. Let (P be a variable over characters (i.e. meanings) which are functions f r o m thematic dimensions to properties. Then: [in every respect]

= \(?VdXy, (P (dx,;)

[in some respects]

= X(P3dx, / (P (dx,,)

[in character]

= \(P(P (di)

[in risk taking]

= \(P(P

[in physical appearance] = XCPCP (efe) EXAMPLES (1)

Flink tegenover kleine meisjes maar niet tegenover jongens ( ' C o u r a g e o u s with respect to little girls, but not with respect to boys')

Here, we have contrary respects that specify the dimension di, 1 by d\, 1, g and d i , i , b - The complex dimension can be expressed by Boolean operations: (X CP (