A lot of noise and no walnuts idiom

The letter was addressed to John and myself.
People like yourself should be given every assistance.
*the letter was addressed to herself.

Series editors Keith Brown, Eve V. Clark, Jim Miller, Lesley Milroy, Geoffrey K. Pullum, and Peter Roach

Meaning in Language An Introduction to Semantics and Pragmatics

Meaning in Language An Introduction to Semantics and Pragmatics

D. Alan Cruse University of Manchester

OXFORD UNIVERSITY PRESS

This page intentionally left blank

OXFORD UNIVERSITY PRESS

Great Clarendon Street, Oxford OX2 6DP Oxford University Press is a department of the University of Oxford. It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide in Oxford New York Athens Auckland Bangkok Bogota Buenos Aires Calcutta Cape Town Chennai Dar es Salaam Delhi Florence Hong Kong Istanbul Karachi Kuala Lumpur Madrid Melbourne Mexico City Mumbai Nairobi Paris Sao Paulo Singapore Taipei Tokyo Toronto Warsaw with associated companies in Berlin Ibadan

Oxford is a trade mark of Oxford University Press in the UK and in certain other countries

Published in the United States by Oxford University Press Inc., New York

© Oxford University Press 2000

The moral rights of the author have been asserted Database right Oxford University Press (maker)

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, or under terms agreed with the appropriate reprographics rights organizations. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above

You must not circulate this book in any other binding or cover and you must impose the same condition on any acquirer

British Library Cataloguing in Publication Data Data available

Library of Congress Cataloging in Publication Data (Data applied for) ISBN 0-19-870010-5

1 0 9 8 7 6 5 4 3 2

Typeset by RefineCatch Limited, Bungay, Suffolk Printed in Great Britain on acid-free paper by Bath Press Ltd., Bath, Avon

To Paute, Pierre, and Lisette

This page intentionally left blank

Contents

Part 2 Words and their Meanings 5 Introduction to lexical semantics

6 Contextual variability of word meaning

7 Word meanings and concepts

8 Paradigmatic sense relations of inclusion and identity

9 Paradigmatic relations of exclusion and opposition

10 Word fields

11 Extensions of meaning

12 Syntagmatic relations

13 Lexical decomposition

Part 3 Semantics and Grammar 14 Grammatical semantics

Parti Fundamental Notions 1 Introduction

2 Logical matters

3 Types and dimensions of meaning

4 Compositionality

Typographic conventions

Preface

ix

x

1 3

17 41 65

83 85

103 125

143 163 177 197 217 237

263 265

viii Contents

Part 4: Pragmatics 15 Reference and deixis

16 Speech acts

17 Implicatures

Conclusion

Answers to questions

References

Index

301 303 329

347

379 383 401 409

Typographic conventions

Small capitals For concepts; occasionally for lexical roots.

Small capitals in square brackets For semantic components.

Angled brackets For selectional restrictions

Bold type For technical terms when first introduced.

Italics For citation forms when not set on a different line.

Bold italics For emphasis.

Single quotation marks For quotations from other authors; 'scare quotes'.

Double quotation marks For meanings.

Question marks For semantic oddness.

Asterisks For ungrammaticality or extreme semantic abnormality.

Preface

The aim of this book is not to present a unified theory of meaning in language (I am not even sure that that would be a worthwhile project), but to survey the full range of semantic phenomena, in all their richness and variety, in such a way that the reader will feel, on completing the book, that he or she has made face-to-face contact with the undeniably messy 'real world' of meaning. At the same time, it aims to show that even the messy bits can, at least to some extent, be tamed by the application of disciplined thinking. As far as seman- tic theories are concerned, I have been unashamedly eclectic, adopting what- ever approach to a particular problem seems genuinely to shed light on it. If there is a theoretical bias, it is in favour of approaches which, like the cognitive linguistic approach, embrace the continuity and non-finiteness of meaning.

This is not intended to be a 'baptismal' text; it would probably not be suit- able for absolute beginners. The sort of readership I had in mind is second- or third-year undergraduates and beginning postgraduates who have completed at least an introductory course in linguistics, and who require an overview of meaning in language, either as preparation for a more detailed study of some particular area, or as background for other studies. I would hope it would be found useful, not only by students of linguistics, but also students of ancient and modern languages, translation, psychology, perhaps even literature.

Most of the material in the book has grown out of courses in general seman- tics, lexical semantics, and pragmatics, given to second- and third-year under- graduates and postgraduates at Manchester University over a number of years. I owe a debt to generations of students in more than one way: their undisguised puzzlement at some of my explanations of certain topics led to greater clarity and better exemplification; critical questions and comments not infrequently exposed weaknesses in the underlying arguments; and very occasionally, a genuine flash of insight emerged during a classroom discussion.

The final form of the text was significantly influenced by constructive com- ments on a draft by Jim Miller of the University of Edinburgh, an anonymous American reviewer, and John Davey of Oxford University Press, although, of course, full responsibility for remaining imperfections lies with myself.

Preface xi

The organization of the book is as follows. It is in four parts. Part I dis- cusses a range of basic notions that underlie virtually all discussions of mean- ing within linguistics; Part 2 concentrates on aspects of the meanings of words; Part 3 deals with semantic aspects of grammar; Part 4 introduces the core areas of pragmatics, and highlights the relations between meaning and context.

Within Part I, Chapter I provides a very general introduction to questions of meaning, locating the linguistic study of meaning within the wider context of the study of signs and communication in general. Chapter 2 introduces a set of fundamental conceptual tools, mostly drawn from the field of logic, which, because of their wide currency in discussions of semantic matters, constitute indispensable background knowledge for a study of meaning in language. In Chapter 3, a number of concepts are introduced for the descrip- tion of meanings and differences of meaning. A basic dichotomy (based on Lyons 1977) is introduced between descriptive and non-descriptive meaning and, under each of these headings, important types and dimensions of vari- ation are described. It is rare to encounter any extended treatment of these topics in semantics textbooks, yet a mastery of them is essential to anyone who wishes to talk in a disciplined way about meanings. Chapter 4 discusses the way(s) in which simpler meanings are combined to form more complex meanings.

In Part 2, Chapter 5 provides a general introduction to the study of word meanings, first discussing whether there are any restrictions on what sort of meanings words can bear, then distinguishing the meaning of a word from that of a sentence or discourse, and the meanings of full lexical items from the meanings of grammatical elements. In this chapter the major approaches to lexical semantics are also outlined. In Chapter 6, the focus is on the range of variation observable in a single word form in different contexts, ranging from arbitrarily juxtaposed homonymies to subtle modulations of sense. Chapter 7 introduces a conceptual approach to lexical semantics, beginning with a dis- cussion of whether and to what extent word meanings can be equated with concepts. The discussion continues with an outline of prototype theory, the currently dominant approach to natural conceptual categories, and its rele- vance for the study of word meanings. Chapters 8 and 9 deal with relations of sense between lexical items which can occupy the same syntactic position—in other words, paradigmatic sense relations, such as hyponymy, meronymy, incompatibility, synonymy, antonymy, complementarity, reversivity, and con- verseness. Chapter 10 looks at larger groupings of words—word fields— mainly structured by the sense relations examined in the previous two chap- ters. Chapter 11 describes the main types of process, such as metaphor and metonymy, which enable new meanings to be produced from old ones. In Chapter 12, meaning relations between words in the same syntactic construc- tion, that is, syntagmatic sense relations, are examined. Topics discussed include the nature of normal and abnormal collocations, reasons for a

xii Preface

tendency for certain types of words to co-occur, and the nature and con- sequences of selectional pressures of words on their partners in a string. Chapter 13 outlines the componential approach to the description of word meaning, which specifies meaning in terms of semantic primitives.

The focus in Chapter 14, which constitutes the whole of Part 3, is on the sorts of meanings associated with various grammatical entities. First there is a discussion of the problem of whether there are any constant meanings attached to categories such as noun, verb, and adjective, and functions such as subject and object. There then follows a survey of the sorts of meaning borne by grammatical elements of various sorts, such as number and gender in the noun phrase, tense, aspect, and modality in connection with the verb, degree in the adjective, and so on.

Part 4 covers topics which are usually considered to fall under pragmatics, in that either they involve aspects of meaning which cannot be satisfactorily treated unless context is taken into account, or they are not propositional in nature (or both). Chapter 15 is concerned with reference, that is, establishing connections between utterances and the extralinguistic world. Reference is portrayed as the assigning of values to variables, the variables being signalled by definite expressions and the values being items in the extralinguistic world. Various strategies for indicating (on the part of the speaker) and determining (on the part of the hearer) correct referents are discussed, including the use and interpretation of deictic elements, names, and descriptions. Chapter 16 provides an outline of speech act theory, mainly following Austin and Searle (1969). It discusses the acts that people perform when they are speaking—acts such as stating, requesting, warning, congratulating, commanding, and so on. The range of different types of speech act is surveyed and their nature exam- ined. Chapter 17 deals with conversational implicatures, that is, those aspects of the intended meaning of an utterance which are not encoded in its linguistic structure, but are, as it were, 'read between the lines'. Different types of con- versational implicature are described and some proposed explanations of how they arise are considered.

The concluding chapter briefly surveys the areas covered in the book, sug- gests practical applications of the study of meaning, and highlights areas which are currently poorly understood, and where further research is needed. Each chapter except Chapter I and Chapter 5 contains a set of discussion questions and/or exercises, suggested answers to which will be found at the end of the book.

Part 1

In this first part of the book, a number of fundamental, but fairly general notions are introduced, which need to be grasped before the more detailed discussions in later sections can be properly appreciated. Chapter 1 has a scene-setting function, identi- fying the place of linguistic signs and linguistic communication in the broader domains of semiotics and communication in general. Chapter 2 introduces a num- ber of vital conceptual tools drawn from the field of logic. Chapter 3 surveys the range of different sorts of meaning, and dimensions of variation in meaning. Chapter 4 discusses the notion of compositionality, one of the essential properties of lan- guage, and its limits.

Fundamental Notions

This page intentionally left blank

CHAPTER 1

Introduction

1.1 Communication

1.2 Semiotics: some basic notions

1.3 Language and other communicative channels

1.4 Characteristics of linguistic channels

5 1.5 Approaches to the study of meaning

7 1.6 The linguistic study of meaning

1.7 Branches of the study of meaning 8

Suggestions for further reading 9

10

12

15

16

This page intentionally left blank

CHAPTER 1

Introduction

1.1 Communication

Meaning makes little sense except in the context of communication: the notion of communication therefore provides as good a place as any to start an exploration of meaning. Communication can be conceived very broadly, including within its scope such matters as the transfer of information between biological generations via the genetic code, the interaction between a driver and his car, and indeed any sort of stimulus-response situation. Here we shall confine ourselves to what is surely the paradigm communicative scenario, namely, the transfer of information between human beings.

1.1.1 A simple model

Let us begin with a simple model, as shown in Fig. I.I (after Lyons 1977). In the model, the process begins with a speaker who has something to

communicate, that is, the message. Since messages in their initial form cannot be transmitted directly (at least not reliably), they must be converted into a form that can be transmitted, namely, a signal. In ordinary conversation, this involves a process of linguistic encoding, that is. translating the message into a linguistic form, and translating the linguistic form into a set of instructions to the speech organs, which, when executed, result in an acoustic signal. The initial form of this signal may be termed the transmitted signal.

6 Meaning in language

Every mode of communication has a channel, through which the signal travels: for speech, we have the auditory channel, for normal writing and sign language, the visual channel, for Braille, the tactile channel, and so on. As the signal travels from sender to receiver, it alters in various ways, through distor- tion, interference from irrelevant stimuli or loss through fading. These changes are referred to collectively as noise. As a result, the signal picked up by the receiver (the received signal) is never precisely the same as the transmitted signal. If every detail of the transmitted signal was crucial for the message being transmitted, communication would be a chancy business. However, effi- cient communicating systems like language compensate for this loss of infor- mation by building a degree of redundancy into the signal. Essentially this means that the information in a signal is given more than once, or is at least partially predictable from other parts of the signal, so that the entire message can be reconstructed even if there is significant loss. It is said that language is roughly 50 per cent redundant.

Once the signal has been received by the receiver, it has to be decoded in order to retrieve the original message. In the ideal case, the message reconstructed by the receiver would be identical to the message that the speaker started out with. Almost certainly, this rarely, if ever, happens; how- ever, we may presume that in the majority of cases it is 'close enough'. All the same, it is worth distinguishing three aspects of meaning:

(i) speaker's meaning: (ii) hearer's meaning: (iii) sign meaning:

In the case of an established signalling system like language, the meanings of the signs are not under the control of the users; the signs are the property of the speech community and have fixed meanings. Of course on any particular occasion, the signs used may be ad hoc or conventional, if ad hoc, they may be prearranged or spontaneous.

speaker's intended message hearer's inferred message this can be taken to be the sum of the properties of the signal which make it (a) more apt than other signals for conveying speaker's intended message, and (b) more apt for conveying some messages than others.

1.1.2 Language as a sign system Any natural human language is a complex sign system, 'designed' to ensure infinite expressive capacity, that is to say, there is nothing that is thinkable which cannot in principle be encoded (provided no limit is placed on the complexity of utterances). Each elementary sign is a stable symbolic associ- ation between a meaning and a form (phonetic or graphic); elementary signs may combine together in a rule-governed way to form complex signs which convey correspondingly complex meanings.

Introduction 7

1.2 Semiotics: some basic notions

1.2.1 Iconicity

Signs can generally be classified as iconic or arbitrary. Iconic signs are those whose forms mirror their meanings in some respect; signs with no natural analogical correspondences between their forms and their meanings are called arbitrary. A simple example is provided by the Arabic and Roman numerals for "three": 3 and III. The Arabic form gives no clue to its meaning; the Roman version, on the other hand, incorporates "threeness" into its shape, and is thus iconic. Iconicity is a matter of degree, and usually coexists with some degree of arbitrariness. Three horizontal lines would be just as iconic as the Roman III: the fact that in the Roman symbol the lines are vertical is arbitrary, as is the fact that its size corresponds to that of letters.

Iconicity enters language in several guises. The majority of words in a natural language are arbitrary: the form of the word dog, for instance, does not mirror its meaning in any respect. However, the so-called onomatopoeic words display a degree of iconicity, in that their sounds are suggestive (to varying degrees) of their meanings:

bang clank tinkle miaow splash cuckoo peewit curlew whoosh thud crack ring wheeze howl rumble, etc.

The predominance of arbitrariness in the vocabulary is not an accidental feature, but is a crucial 'design feature' of natural language. There is a limited stock of dimensions of formal variation in linguistic signs; if all signs were iconic, it is difficult to see how universal expressivity could be achieved.

Some iconicity is also apparent in grammar. For instance, words which belong together tend to occur together. In The tall boy kissed the young girl we know that tall modifies boy and not girl because tall and boy come next to each other in the sentence. In some languages this relationship might be shown by grammatical agreement, which is a kind of resemblance, and therefore also iconic. Another way in which iconicity appears in the grammar is that grammatical complexity by and large mirrors semantic complexity.

1.2.2 Conventionality Many of the signs used by humans in communication are natural in the sense that they are part of genetically inherited biological make-up and do not have to be learned, although a maturational period may be necessary before they appear in an individual, and they may be moulded in various ways to fit particular cultural styles. The sort of signs which are natural in this sense will presumably include facial expressions like smiling, frowning, indications of fear and surprise, and so on, perhaps many of the postural and proxemic signs

8 Meaning in language

that constitute the so-called 'body language', certain types of gesture, vocal indications of excitement, desire, etc. (whether or not linguistic), and many more. Natural signs are likely to be the most cross-culturally interpretable.

Other signs have conventionally assigned meanings; they have to be specif- ically learned, and are likely to differ in different communities. Linguistic signs are the prototypical conventional signs. Even onomatopoeic words usually have a significant conventional element; often the iconic nature of the word can only be appreciated, as it were, with hindsight. Take the Turkish word bulbul. What does it refer to? A baby's babbling? The noise of a mountain spring? In fact, it means "nightingale". Looking back, one can make the connection. It is not only linguistic signs that are conventional. Obscene or offensive gestures, for instance, can vary quite remarkably cross culturally: I was once reprimanded for pointing the soles of my feet at the Prime Minister of Iraq (in Arab culture this is disrespectful: my disrespect was entirely inadvertent). Even in Europe, conventional gestures can differ: Greeks are famously—and slightly inaccurately—said to shake their heads to say "Yes", and nod to say "No".

1.2.3 Discreteness

Some signs can vary gradually in their form, and their meanings vary in paral- lel with the change of form, like the fisherman's indication of the size of 'the one that got away'; these are called continuous signs. Other signs have fixed shapes, and must be chosen from a limited inventory: intermediate forms are not allowed, the range of possibilities is 'chunked'; such signs are described as discrete. Linguistic signs are virtually all of the discrete variety. Again, this is not an accidental feature, but has a close connection with iconicity and arbitrariness: continuous signs are necessarily iconic; arbitrary signs are necessarily discrete.

1.3 Language and other communicative channels

The prototypical scenario for linguistic communication is two people engaged in face-to-face conversation. Of course, in such an encounter, lan- guage signals are exchanged; but normally so are many other types of signal, and these modify and/or supplement the linguistically encoded message. Let us, then, briefly look at the semiotic environment of language in a typical conversation.

The signs that accompany language can be divided into two major types— paralinguistic and non-linguistic. The defining characteristic of paralinguistic signs will be taken here to be an extreme dependence on the accompanying language. Either they cannot be produced except during speech (because they are carried on the voice), or they cannot be interpreted except in conjunction

Introduction 9

with accompanying language. Examples of the first variety are abnormal volume, tempo, pitch, and voice quality; to function as signs, there must be a departure from some (personal) baseline or norm. For instance, abnormally high volume, fast tempo, or high pitch typically signal a heightened emo- tional state. Examples of the second variety include pausing, emphatic ges- tures, and gestures which metaphorically depict, for instance, direction of motion.

The functions of paralinguistic signs can be conveniently classified under three headings:

(i) Punctuation: there are signs which have functions parallel to those of punctuation in written language, mainly to segment the stream of speech so as to facilitate processing.

(ii) Modulation: this involves the addition of an emotive or attitudinal colouring to the linguistically encoded message.

(iii) Illustration: some signs 'depict' a key element in the message, such as a direction of movement, or a shape; the depiction may be relatively literal, like the hand movements of someone describing the climbing of a spiral staircase, or metaphorical, as when vertical and parallel hands accompany the setting of limits of some kind.

Not all the signs that occur alongside language are paralinguistic in the sense defined. For instance, one may smile or frown while speaking, and this may well 'modulate' the message. But smiles and frowns (and many other signs) are perfectly interpretable and capable of being produced in the absence of any accompanying language. These are therefore to be considered as non-linguistic.

1.4 Characteristics of linguistic signs

Paralinguistic signs are typically natural, continuous, and iconic, whereas linguistic signs are for the most part arbitrary, discrete, and conventional.

1.4.1 Simple and complex signs

Linguistic signs may be simple or complex. This does not just mean that they can occur singly or in groups of various sizes: the occurrence of two or more signs together does not necessarily result in a complex sign. Take the case of someone who answers a question with the word Yes, at a higher than usual pitch, and at the same time smiling. This person has not produced a complex sign with three constituents, only three simple signs simultaneously. The mean- ings of the three signs are simply added together: there is no interaction between the signs other than additivity. Contrast this with a minimally com- plex sign such as red wine: to obtain the meaning of this sign, we do not simply

10 Meaning in language

add the meaning of red to the meaning of wine (that would give us something like "wine plus redness"). What happens is the meaning of red interacts with the meaning of wine by restricting it.

There is no theoretical upper limit to the complexity of linguistic signs. This is rendered possible by the recursive nature of syntax, that is, the existence of rules which can be applied indefinitely many times (like the one which yields This is the dog that worried the cat that killed the rat that ate the com that . . .). Such rules are an essential prerequisite for the 'universal expressivity' of language—the fact that anything thinkable is expressible, or at least can be approximated to any given degree of accuracy.

1.4.2 Signs at different linguistic levels

A linguistic sign may be no more than a phoneme (or two): this is one inter- pretation of the sl- of slimy, slovenly, slug, slag, slum, slink, slattern, slut, slob, etc. which seems to indicate something unpleasant, or the gl- of glare, glimmer, glitter, glisten, glow, gleam, etc. which all have something to do with light effects. These have no grammatical status, and no contrastive value, but the intuitions of native speakers leave no doubt that they should receive some recognition. Other signs occur at higher levels of linguistic organization, from morpheme level (e.g. the -s of dogs), through word level (e.g. denationaliza- tion), clause level (e.g. the formal difference between John is here and Is John here? which signals that one is a question and the other a statement), sentence level (e.g. We'll do it as soon as you arrive as opposed to As soon as you arrive, we'll do it), up to text level (e.g. the fact that a stretch of text constitutes a sonnet is indicated by the form of the text as a whole: this form therefore constitutes a high-level sign).

The fact that a sign manifests itself at a particular level does not entail that it is to be interpreted at that level. A few examples will illustrate this point. The item the, a word, exerts its semantic effect on a whole noun phrase the little old lady who lives in the cottage on the hill; the -ed of John kissed Mary, a bound morpheme, semantically situates the time relative to the moment of utterance of the whole event symbolized by John kiss Mary; a single word like matri- mony may mark a whole discourse as being in a certain register.

1.5 Approaches to the study of meaning

Meaning may be studied as a part of various academic disciplines. There is of course a significant degree of overlap between disciplines, but characteristic- ally all have something idiosyncratic and unique in their approach (the following remarks are merely illustrative).

Introduction 11

1.5.1 Philosophy

Linguists typically take the existence of meaning for granted and accept it as an intuitively accessible 'natural kind'. They do not ask questions like How is it possible for anything to mean something? or What sort of relation must hold between X and Yfor it to be the case that X means Y? Such questions are the province of the philosopher, particularly the philosopher of language.

1.5.2 Psychology

Meaning is a major concern of the psychology of language and psycho- linguistics. (I shall not attempt to distinguish these.) A distinctive feature here is the experimental study of how meanings are represented in the mind, and what mechanisms are involved in encoding and decoding messages. An example of a fact that could only emerge within a psycholinguistic framework is that in the lexical decision task, where experimental subjects observe strings of letters flashed on a screen and must indicate by pressing the appropriate button whether the string represents a word or not, responses are faster to words with concrete meanings than to words with abstract meanings, even when extraneous factors like length and frequency are rigorously controlled. This observation presumably provides a clue to the role of meaning in word recognition (to the best of my knowledge it is still a puzzle).

1.5.3 Neurology

Psychologists take a 'macro' view of mental states and processes. Neurologists, on the other hand, want to know how these states and processes are imple- mented at the neuronal level. A psychologist might be broadly compared with a computer programmer, and a neurologist to the designer of computer chips. Meaning, like everything else in mental life (at least if one is a physicalist) must boil down ultimately to connections between neurons.

1.5.4 Semiotics

Semioticians view language as one sign system amongst many, and seek out those features which render it so successful. They are also likely to give emphasis to marginal aspects of linguistic signification. The recent strong interest in iconicity in language represents a significant overlap between the linguistic and semiotic approaches to meaning.

1.5.5 Linguistics

It is not easy to encapsulate the linguistic approach to meaning in language succinctly. There are perhaps three key aspects. The first is that native speakers' semantic intuitions are centre-stage, in all their subtlety and nuances: they constitute the main source of primary data. The second is the

12 Meaning in language

importance of relating meaning to the manifold surface forms of language. The third is the respect paid not just to language, but to languages.

1.6 The linguistic study of meaning in language

1.6.1 What is linguistic meaning?

Here we attempt to say what is to count as meaning in language. Following an impulse towards generosity rather than austerity, we shall as a first step say that all meaning is potentially reflected in fitness for communicative intent. It will be assumed that a way of tapping into this is in terms of contextual normality: every difference of meaning between two expressions will show up as a difference of normality in some context. Thus, we know that illness and disease do not mean the same, because during his illness is normal, but during his disease is not; almost and nearly do not have precisely the same meaning because very nearly is normal but very almost is not; pass away and kick the bucket have different meanings because It is with great sadness that we report that our Beloved Ruler kicked the bucket two minutes after midnight is odd, but It is with great sadness that we report that our Beloved Ruler passed away two minutes after midnight is normal. We take normalityloddness and relative oddnesslnormality to be primitive intuitions.

It will be noticed that the move in the above characterization was from meaning to contextual abnormality. Unfortunately, the move cannot without further ado be made from abnormality to meaning, because there are other factors besides meaning which affect normality.

Let us assume that we are dealing with spoken language and that the utter- ance is correctly pronounced. The two sources of abnormality that we wish to eliminate if possible are grammatical deviance and 'meaning' that is non- linguistic in origin. Let us make the simplifying assumption that if a pin- pointed deviance is grammatical in nature, it will not prove possible to reduce it by contextual manipulation, for instance by interpreting it as metaphor, science fiction, or fairy-tale. Thus, They goes is irredeemably deviant in any context, whereas / shall go there yesterday might just make sense in a setting where time travel (or at least temporal scrambling) is possible. That leaves non- linguistic meaning to be taken care of. Consider the possibility that a certain type of delivery, not amounting to mispronunciation, may be a sign that the speaker is under the influence of some pharmacological substance. Let us make the further assumption that some speaker is deliberately trying to con- vey this information. This might well be odd in, for instance, the context of a sermon. Is this linguistic meaning? (In the case described, it is certainly a kind of meaning, and language is used to convey it.) Presumably it is not linguistic meaning, but how do we exclude it? One way is to stipulate that linguistic meaning must either be conventionally associated with the linguistic forms used, or be inferable from the latter in conjunction with contextual knowledge.

Introduction 13

One indication that the above example is not of this type would be its insensitivitv to the actual words used.

1.6.2 What are we trying to achieve?

I.6.2.I Specifying/describing meanings A very important task is to discover a way of specifying or describing mean- ings, whether of isolated words or sentences, or of utterances in context. The position taken in this book is that in general, meanings are not finitely describ- able, so this task boils down to finding the best way to approximate meanings as closely as is necessary for current purposes (lexicographers have long had to confront this problem for words).

I.6.2.2 How meaning varies with context The meanings of all linguistic expressions vary with the context in which they occur. For instance, the shade of colour indicated by a redhead and red wine are markedly different; the periods of time denoted by month in (I) and (2) are quite likely to be different:

(I) He's here for a month. (could be four weeks; not dependent on time of utterance)

(2) He's here for the month. (will depend on time of utterance, but could be 31 days)

Some variations, like the sex of the doctor in Our doctor has just married a policeman and Our doctor has just married an actress can be predicted by general principles; other variants are less, or not at all predictable. Semanticians seek a revealing account of contextual variation.

I.6.2.3 Kinds of meaning There are different sorts of meaning, each with different properties. For instance, whatever the difference in meaning between (3) and (4), it does not affect the truth or falsity of the statement:

(3) Old Joshua Hobblethwaite popped his clogs last week. (4) Old Joshua Hobblethwaite passed away last week.

I.6.2.4 What happens when meanings combine? Another vital aspect of semantics is how simple(r) meanings combine to form more complex meanings. To some extent this is a function of grammatical structure: for instance, the way red and hat combine in a red hat is not the same as the way turn and red combine in to turn red. But differences occur even within the same grammatical construction: the mode of combination of red and hat in a red hat is different from that of long and eyelash in long eyelashes (compare long eyelashes and a long river).

14 Meaning in language

I.6.2.5 Systematicity and structure; possibility of formalization All semanticians are to some extent looking for regularities and system in the way meanings behave, as this leads to maximally economical descriptions. The most dedicated to this aspect of semantics are those who attempt to model the semantic behaviour of natural language expressions by means of a strict logical or quasi-mathematical formalism. This route will not be followed in this book.

I.6.2.6 New meanings from old A striking feature of linguistic expressions is their semantic flexibility: beyond their normal contextual variability, they can be bent to semantic ends far removed from their conventional value, witness She swallowed it hook, line and sinker or You'll find her in the telephone book. The study of such extensions of meaning is an important task for semantics.

I.6.2.7 Role(s) of context It is usually assumed that linguistic expressions can be assigned some sort of context-independent semantic value, although there is much disagreement regarding exactly what this is. There is also general agreement that context is of vital importance in arriving at the meaning of an utterance. The role of context ranges from disambiguating ambiguous expressions as in We just got to the bank in time, through identification of referents (who is he, where is there, in time for what, in He didn't get there in time), to working out 'between the lines' messages like B's ignorance of the whereabouts of the corkscrew in:

(5) A: Where's the corkscrew? B: It's either in the top drawer in the kitchen, or it's fallen behind the

piano.

1.6.3 The approach adopted in this book

We are not yet in a position to rule out any approaches which yield insights, even if some such approaches appear at first sight incompatible. This book therefore takes an ecumenical position on many issues. In so far as there is a theoretical bias, it is towards the cognitive semantic position. This means, in particular, that the meaning of a linguistic expression is taken to arise from the fact that the latter gives access to a particular conceptual content. This may be of indeterminate extent: no distinction is made between linguistic meaning and encyclopaedic knowledge.

Since this book is not intended to propound a body of theory, but to acquaint non-specialists with the range of semantic phenomena in language, there is a bias towards descriptive coverage at the expense of theoretical rigour.

Introduction 15

1.7 Branches of the study of meaning in language

The following are the main broadly distinguishable areas of interest in the study of meaning. They do not by any means form watertight compartments: there are many points of overlap.

1.7.1 Lexical semantics Lexical semantics studies the meanings of words; the focus here is on 'content' words like tiger, daffodil, inconsiderate, and woo, rather than 'form'/ 'grammatical' words like the, of, than, and so on. To a non-specialist, the notion of meaning probably has a stronger link with the idea of the word than with any other linguistic unit: words are, after all, what are listed in dictionar- ies, and the main function of a dictionary is to tell us what the listed words mean. For this reason, lexical semantics perhaps provides the easiest access route into the mysteries of semantics in general, and this is one reason why it has been given a prominent place in this book, and why it comes early.

1.7.2 Grammatical semantics

Grammatical semantics studies aspects of meaning which have direct rele- vance to syntax. This has many manifestations, which can only be briefly illustrated here. One problem is the meaning of syntactic categories (problem- atic, because not everyone believes they can be assigned meanings). Consider, for instance, the differences in the meaning of yellow in the following:

(6) She wore a yellow hat. (adjective)

(7) They painted the room a glowing yellow. (noun)

(8) The leaves yellow rapidly once the frosts arrive. (verb)

Another aspect of grammatical semantics is the meaning of grammatical morphemes like the -ed of walked, the -er of longer, the re- and the -al of retrial, and so on.

Clearly this overlaps with lexical semantics, partly because some grammat- ical elements are words (like the, and of), but more particularly because some aspects of the meanings of full lexical items determine to some degree their grammatical behaviour (for instance, the fact that / am studying that question is grammatical, but not I am knowing the answer to that question).

1.7.3 Logical semantics Logical semantics studies the relations between natural language and formal logical systems such as the propositional and predicate calculi. Such studies usually aim at modelling natural language as closely as possible using a tightly controlled, maximally austere logical formalism. It is arguable that sometimes such studies shed more light on the formalism used than on the language being

16 Meaning in language

modelled; none the less, valuable insights have come from this approach. To date, most such studies have concentrated on the propositional/sentential level of meaning, and have rarely attempted to delve into the meanings of words.

1.7.4 Linguistic pragmatics

For present purposes, pragmatics can be taken to be concerned with aspects of information (in the widest sense) conveyed through language which (a) are not encoded by generally accepted convention in the linguistic forms used, but which (b) none the less arise naturally out of and depend on the meanings conventionally encoded in the linguistic forms used, taken in conjunction with the context in which the forms are used. This rather cumbersome formulation is intended to allow into pragmatics things like the identity of the individual referred to by John in / saw John today, and the assumption that the room in question had several lights in John entered the room; all the lights were on, at the same time excluding, for instance, the possibility that the person saying / saw John today had a private ad hoc arrangement with the hearer that when- ever he said John, he should be taken to mean "Mary" (since it does not arise naturally out of the normal meaning of John), and excluding also the possibil- ity of someone's inferring from a speaker's slurred speech that they were drunk (since this does not depend on the conventional meanings of the words uttered). Pragmatics is usually contrasted with semantics, which therefore deals with conventionalized meaning; obviously, the three divisions discussed above belong to semantics.

Suggestions for further reading

Much fuller accounts of the semiotic environment of spoken language can be found in Argyle (1972), Beattie (1983), Ellis and Beattie (1986) and Clark

CHAPTER 2

Logical matters

2.1 Introduction

2.2 Arguments and predicates

2.3 Sense, denotation, and reference: intension and extension

2.4 Sentence, statement, utterance, and proposition

2.5 Logical properties of sentences

19

19

21

22

28

2.6 Logical classes

2.7 Logical relations

2.8 Quantification

2.9 Use and mention

Discussion questions and exercises

Suggestions for further reading

32

34

35

36

37

39

This page intentionally left blank

CHAPTER 2

Logical matters

2.1 Introduction

This chapter introduces a set of fundamental conceptual tools, mostly drawn from the field of logic, which, because of their wide currency in discussions of semantic matters, constitute indispensable background knowledge for a study of meaning in language. The level of treatment here is fairly elementary; some of the notions introduced will be further refined in subsequent chapters.

2.2 Arguments and predicates

A closely linked pair of concepts which are absolutely fundamental to both logic and semantics are argument and predicate. No attempt will be made here to explore the philosophical background and underpinning of these notions: the basic notions are fairly accessible and they will be treated in an elementary fashion. Put simply, an argument designates some entity or group of entities, whereas a predicate attributes some property to the entity denoted by the argument, or a relation between the entities denoted by the arguments, if there is more than one. Thus, in John is tall, we can identify John as the argument, and is tall as the predicate. In John likes Mary, both John and Mary are arguments, and likes is the predicate which attributes a particular relationship between the entities denoted by the arguments; in John gave Mary a rose, there are three arguments, John, Mary and the rose, with gave as the predicate. The combination of an argument and a predicate forms a proposition: notice that a proposition may have only one predicate, but may have more than one argument. It is not clear whether there is any theoretical upper limit to the number of arguments a predicate may take, but the most one is likely to encounter in linguistic semantic discussions is four, exemplified by Mary paid John £500 for the car:

Arguments: Mary, John, £500, the car Predicate: paid (for)

20 Meaning in language

An argument may have a more or less complex internal structure. For instance, a whole proposition, itself possessing argument(s) and predicate, may constitute an argument, as in John was surprised that the man was tall.

Arguments: (i) John (ii) that the man was tall Predicate: was surprised Argument (ii) Argument: the man

Predicate: was tall

There are various ways of incorporating propositions as constituents of complex arguments, and there is no limit to the resulting degree of complexity. An account of this is beyond the scope of this book.

Predicates are commonly described as one-place, two-place, three-place, etc. according to the number of arguments they take, so that, for instance, is poor is a one-place predicate, and teach is a three-place predicate (John taught Mary French). But what does it mean to say that teach is a three-place predicate? How do we determine how many places a predicate has?

This is, in fact, a very difficult question, but we can get some handle on it by looking at a few verbs. Let us start by looking at teach. One aspect of the problem is immediately obvious when we look at examples such as the following:

(1) John taught Mary French. (2) John taught French for two years. (3) A: What does John do?

B: I think he teaches at Lowhampton High. (4) Anybody who teaches teenagers should get double salary.

Do we say that teach is three-place in (1), two-place in (2) and (4), and one- place in (3)? While there is some justification for such an analysis, there is intuitively a sense in which the (overt) argument structure of (1) is basic, and irreducible. Logically, for an act of teaching to take place, there must be some- one who does the teaching, someone who undergoes the teaching (whether or not they actually learn anything!), and some item of knowledge or skill which it is hoped will be acquired by the latter. Without at least one each of these requirements, the notion of teaching is not logically coherent. What, then, are we to make of (2)-(4) above? It seems that we assume that the missing argu- ments could in principle be supplied, but the speaker has not supplied them, presumably because they are not currently relevant, or perhaps in some cases because they are extremely obvious. In (2), we do not have a vision of John discoursing in solitude on the French language; nor do we imagine that John in (3) does something like sneezing, which needs neither audience nor topic. By the same sort of criteria, read is a basically two-place verb, in spite of the existence of sentences such as John is reading and John is learning to read. We cannot make sense of read unless we have someone to do the reading and some coded signal to decode. As a final example, consider buy. This requires four

Logical matters 21

arguments, as in John sold the car to Mary for £500. If John receives no money, he is simply giving the car away; if there is no car, Mary is just giving John some money; if there is no one to receive the money and concede ownership of the car, then Mary is throwing her money away and taking possession of the car. And so on. In this way, it is usually possible to determine a basic logically minimum number of arguments for a predicate.

But our problems are not over. Acts of reading, teaching, and selling (and sneezing) take place at particular times and particular places. We may say that unlocated, timeless acts of reading, etc., are logically impossible. It is also necessarily the case that the ambient temperature has such and such a value. Whereas it is relatively easy to establish the minimum number of arguments for a predicate, how do we establish a maximum? Or perhaps there is no such thing? After all, if we learn that John sneezed, we assume that the event happened at a particular place and time, and that even if the speaker did not specify these items, they could in principle be specified, along with countless other things. What is the difference in status, therefore, between the time and place of John's sneezing, and the subject that John teaches? One approach is to say that although a verb like sneeze in a sense conjures up a rather complex picture of a person in a setting acting in a certain way, it highlights only certain aspects of that scene, but without obliterating or denying the rest. These highlighted aspects are what distinguish the act denoted from other possible acts. This means that if we wish to ascertain whether a John-sneeze- event has occurred, we need only observe happenings pertaining closely to John; we do not need to check time, place, or temperature. In this way we can check that the minima we previously established for the number of arguments a predicate takes are generally also maxima.

2.3 Sense, denotation, and reference: intension and extension

Language is used to communicate about things, happenings, and states of affairs in the world, and one way of approaching the study of meaning is to attempt to correlate expressions in language with aspects of the world. This is known as the extensional approach to meaning.

The thing or things in the world referred to by a particular expression is its referent(s): in saying The cat's hungry, I am (normally) referring to a particular cat, and that cat is the referent of the expression the cat. The whole utterance attributes a particular state to the cat in question. We can also consider the whole class of potential referents of the word cat, namely, the class of cats. This, too, is sometimes called the reference of the word cat. But this is clearly different from the designation of particular individuals as in the case of The cat's hungry, so, to avoid confusion, we shall follow Lyons and say that the class of cats constitutes the denotation of the word cat. So, in the case of

22 Meaning in language

The cat's hungry, the word cat denotes the class of cats, but the cat refers to a particular cat.

The alternative to an extensional approach to meaning is an intensional approach. Take the word cat. Why do we use it to refer to cats, rather than, say, to platypuses or aardvarks or spiny anteaters? One answer is that the word is associated with some kind of mental representation of the type of thing that it can be used to refer to, and aardvarks do not fit the description associated with the word cat. This representation constitutes what is called the sense of the word (or at least part of it). We shall assume in this book that the main function of linguistic expressions is to mobilize concepts, that concepts are the main constituents of sense, and that sense (and hence concepts) constrains (even if it does not completely determine) reference. (It should be noted that some authors, for instance Lyons, understand sense in a different way. For them, sense is a matter of the relations between a word and other words in a language. So, for instance, the sense of cat would be constituted by its relations with other words such as dog (a cat is necessarily not a dog), animal (a cat is an animal), miaow (The cat miaowed is normal but ?The dog miaowed is not).)

2.4 Sentence, statement, utterance and proposition

A number of distinctions need now to be made which at first sight might seem to be academic nit-picking of the worst sort. However, they are absolutely vital for clarity in semantics. For the sake of simplicity of exposition, the distinctions will first of all be explained in connection with declarative sen- tences only; how the various notions apply to non-declarative sentences will be dealt with later.

2.4.1 Sentence meaning; truth conditions; propositional content A sentence is a grammatical unit, that is, it is a string of words of a particular type, whose well-formedness conditions are specified in the grammar of the language. Thus, The cat sat on the mat and John put his hat on the table are sentences of English; John put on the table and Is the of mother boy swim are not. We shall not spend too much time on discussing what distinguishes a sentence from other grammatical units, such as a phrase, or a word, except to say that a sentence must contain at least one independent clause (i.e., one that does not need another clause, to be grammatical), and a clause must contain at least one predication (i.e., an argument-predicate structure). The sentence is thus the smallest linguistic unit that can be used in an act of 'telling'. Compare the following for normality/oddness:

(5) A: Tell me something nice. B: Chocolate.

Logical matters 23

A: What do you mean? B: Well, chocolate is nice.

(6) A: Tell me something nice. B: Love is a many-splendoured thing. A: Ah! How true!

The oddness of B's reply in (5) consists in the fact that it neither overtly forms a sentence, nor permits a plausible sentence to be reconstructed from the context, as in, for instance, (7):

(7) A: He asked me what I wanted. B: What did you tell him? A: Icecream.

Here, the full act of telling is reconstructible as I want ice cream, which involves a sentence. Henceforward we shall use our intuitions as to what con- stitutes a sentence.

Notice that we said above that the sentence was 'used in the act of telling': the sentence itself, on its own, does not in fact tell us anything. What does the sentence The cat sat on the mat tell us? Is it true? There is no way of knowing, or rather, the question does not make sense: sentences of themselves do not necessarily have truth values (some, arguably, have: A molecule of water con- tains two hydrogen atoms and one oxygen atom). Yet presumably no one would wish to say that The cat sat on the mat was meaningless. Let us introduce the notion sentence meaning to designate the semantic properties a sentence pos- sesses merely by virtue of being a well-formed sentence (of English), before any question of context or use arises. A sentence possesses this meaning exclusively by virtue of the words it contains, and their grammatical arrange- ment. (I assume, here, that words have at least some context-independent conventionally assigned semantic properties.) We may assume that the gram- mar of a language is associated with principles of composition, that is, rules which tell us how to put together the meanings of the constituents of a con- struction to get the global meaning of the construction. Thus, in The big cat sat on the small mat we know, for instance, that smallness is attributed to the mat and not to the cat, and that a superior vertical position is attributed to the cat; we know, furthermore, what sort of animal is involved, and that only one of them would be intended as a referent in any actual use of the sentence. And so on. In a fairly obvious sense, the meaning of a sentence will constrain the uses to which it can be put, at least without the setting up of additional ad hoc conventions. (For example, // / say How are things?, you will know it's my husband on the phone: that is not what How are things?, by general convention, means.)

Although a sentence, outside of particular uses, does not have a truth value, it does have truth conditions, that is, conditions which must hold for the sentence to be used to make a true statement. Thus, before we can truthfully

24 Meaning in language

say, on some occasion, The cat is on the mat, there must be some relevant feline occupying a specific position relative to an appropriate item of floor-covering. Those aspects of the meaning of a sentence which determine whether a state- ment the sentence is being used to make, in a particular situation, is true or false, are collectively known as the propositional content of the sentence. Two sentences with identical propositional content will yield statements with the same truth values on all occasions of use, as for instance: John caressed Mary and Mary was caressed by John. By the same token, if two sentences have different propositional content, there will necessarily exist some conceivable situation in which they will yield statements with opposite truth values.

Propositional content does not by any means make up the whole of sen- tence meaning, as conceived here. For instance, the interrogative meaning of a question is not included, nor the imperative meaning of a command (more generally, illocutionary meaning falls outside propositional content). The force of such words as yet, still and already (sometimes called conventional implicature) is excluded; the following two sentences, for instance, have the same propositional content, since their truth conditions are the same, but one would not want to say that they were identical in meaning (nor are they appropriate in the same circumstances):

(8) John has not arrived. (9) John has not yet arrived.

Other aspects of meaning which do not fall under propositional content are expressive meaning:

(10) It's very cold in here. (11) It's bloody cold in here.

and features of register such as the formal/colloquial distinction:

(12) My old man kicked the bucket yesterday. (13) My father passed away yesterday.

To qualify as part of sentence meaning, a semantic property simply has to be a stable conventional property of some linguistic expression. We shall not even demand that it be manifested on every occasion of use of the expression in question: it is sufficient that it be potentially present, and not attributable to context. We shall be particularly impressed by evidence that expressions with closely similar propositional content do not possess the property in question. An example of this is the potential expressive nature of the word baby compared with infant (see Chapter 3 for a more extensive discussion of non-propositional meaning).

2.4.2 Statement meaning

One does not make a statement simply by producing a sentence of declarative

Logical matters 25

form. Someone in a language class, for instance, practising the tenses of Eng- lish, who says:

The cat sat on the mat. The cat sits on the mat. The cat will sit on the mat.

is not making a series of statements. The first requirement for a statement to have been made is that a proposition must have been expressed; the second requirement is that an appropriate commitment be made to the truth of the proposition. Let us dwell on this for a moment, beginning with the notion of a proposition.

2.4.2.1 Propositions A simple proposition attributes some property to an entity, or a relation between two or more entities. It is either true or false (even if it is not practic- ally or even physically possible to ascertain which): truth or falsehood is a sign that at least one proposition has been expressed. A proposition is not a specif- ically linguistic entity (although we shall not dwell too long on the problem of what sort of entity it is, and in what Platonic or other realm it subsists). The same proposition may be expressed by an indefinitely large number of sentences:

John saw Mary. John saw his sister. Mary was seen by Peter's uncle. etc.

These can all express the same proposition, provided, of course, that Mary is John's sister, and that John is Peter's uncle, and so on.

Why, then, is The cat sat on the mat not a proposition? As it stands, it is neither true nor false. It becomes true or false when it is asserted of some specific cat and some specific mat. Until definite referring expressions in a sentence have been assigned referents, it does not express a specific prop- osition. A sentence like The cat sat on the mat can be used to express an indefinitely large number of different propositions (i.e., with reference to dif- ferent cats and different mats) on different occasions of use.

2.4.2.2 Epistemic commitment A proposition may be 'entertained', without any stance being adopted towards its truth or falsehood, as, for instance, in a logic class, where proposi- tions are entertained, and their interrelationships studied, such as

(14) All aardvarks are purple. Tinkerbelle is an aardvark. Hence, Tinkerbelle is purple.

26 Meaning in language

However, a proposition on its own cannot actually communicate anything: it is not an item of knowledge. To communicate, it must be energized with some kind of illocutionary force. For instance, the proposition expressed by The earth is a polyhedron is of no interest until someone claims it is true (or false): a true or false proposition is an item of knowledge. The illocutionary force in such a case would be that of assertion. A (minimum) statement is therefore a proposition uttered, as Lyons puts it, 'with epistemic commitment'. There are other possibilities for illocutionary force, to be studied in due course.

2.4.2.3 Incomplete propositions Consider the following exchange:

(15) A: What's the time? B: Half-past four.

Is B telling A anything? Obviously he is. Then what proposition is he express- ing? Again the answer is obvious: "The time is 4.30". But this is not what he actually says. In such cases, it is up to the hearer to reconstruct the full form of the proposition on the basis of contextual clues, but there is no doubt that it is the full form of the proposition that the speaker intends to convey, and this should form part of statement meaning.

We are now in a position to spell out what statement meaning consists of. First of all, statement meaning incorporates in its entirety all aspects of sen- tence meaning which belong to the sentence used in making the statement. Secondly, statement meaning includes a specific proposition or propositions, the identification of which requires that (a) referents be assigned to any def- inite referring expressions in the sentence, and (b) incomplete propositions be completed. It is assumed that the proposition(s) is expressed with the appropriate epistemic commitment.

2.4.3 Utterance meaning

Even a fully elaborated statement meaning may not adequately represent everything the speaker intended to convey by uttering the sentence to which it applies. To take a very simple example, consider the following:

(16) A: Have you cleared the table and washed the dishes? B: I've cleared the table.

In normal circumstances, it would clearly be part of B's intended message that s/he had not washed the dishes. Yet this cannot be obtained by elaborating or completing the proposition expressed. This is even clearer in the following case where it is obviously the speaker's intention to convey the proposition that A is too late for supper:

(17) A: Am I in time for supper? B: I've cleared the table.

Logical matters 27

These extra (i.e. covert) propositions are expected to be inferred by the hearer on the basis of contextual information, but they go well beyond the mere filling out of missing bits in what is actually said. We shall give the name utterance meaning to the totality of what the speaker intends to convey by making an utterance, within certain necessary limits.

Two people might, for instance, have an arrangement such that How was your day? means "How was your day?", but How was the day today? means "My husband is going to his club this evening—we'll have plenty of time for fun". We would wish to exclude this sort of meaning from utterance meaning. I think the key point is that such a use requires a special ad hoc stipulation on the part of speaker and hearer: the hearer's understanding of the clue does not arise entirely from his general knowledge of the rules and conventions governing the use of the forms in question.

Once again, utterance meaning subsumes statement meaning, but only in the sense that the latter must be traversed in order to arrive at the former, i.e. is a necessary step in derivation: utterance meaning does not necessarily incorporate statement meaning as a proper part (although, of course, it may do).

2.4.4 Non-declaratives

So far, we have looked only at declarative sentences. However, the notions elaborated above apply equally to non-declarative sentences. Take the case of a question. The general notion of sentence meaning is probably unproblematic here. The equivalent to statement meaning will be question meaning. Recall that the propositional content of a statement determined whether it was true or false in a given situation. Now questions do not have truth values, but they do have propositional content. Although a question does not have a truth value, a given question-statement pair has what might be called an answer value, that is, the statement is or is not an answer to the question, and if it is, it is either true or false. This answer value is determined equally by the propositional content of the question and the statement:

(18) A: What day is it today? (said on Tuesday) B: It's Monday today, (false answer) A: What day is it today? (said on Monday) B: It's Monday today. (true answer) A: Who do you love best? B: It's Monday today. (no answer)

Just as a declarative sentence has truth conditions, we can postulate that a question-and-answer pair, considered as sentences, has answer conditions, that is, the conditions which must hold for the statement to constitute a true answer to the question.

The same trick can be played with imperatives, except that a linguistic

28 Meaning in language

response is not always, perhaps not usually, what an imperative sets out to elicit. Generally what is required is an action (in the broadest sense, which covers such cases as Don't move!). In the case of a fully contextualized com- mand, where a full command meaning is operative, a given action either counts as compliance or does not, that is, it has a compliance value, equivalent (in the relevant respects) to the truth value of a statement. What its compliance value is, is determined by the propositional content of the command. An imperative sentence has a set of compliance conditions, but no compliance value (how do you obey Put it there, without knowing what it is, or where there is?).

Let us now try to summarize and generalize. Declarative sentences, in gen- eral, do not, in themselves, make statements, interrogative sentences do not ask questions, and imperative sentences do not issue commands. These only result when the sentences are fully contextualized. Each of these, however, has propositional content, which governs what illocutions it can be used to per- form when properly contextualized, that is, which statements can be made, questions asked, and commands issued, etc.

2.5 Logical properties of sentences

2.5.1 Logical relations between sentences

A number of logical relations between sentences make a not infrequent appearance in semantic discussions, especially in lexical semantics. They are related to, but not identical to, relations that logicians recognize between pro- positions. It is, however, necessary to bear in mind their logical basis, if confu- sion is to be avoided. Five relations will be recognized here: implication/ entailment, equivalence, contrariety, contradiction, and independence.

2.5.1.1 Entailment This and the following relations strictly speaking hold between propositions, and that is how we shall treat them to begin with; the connection with sen- tences will be dealt with later. Entailment is the relation which holds between the P and the corresponding Q items in the following:

P It's a dog. John killed the wasp. All dogs are purple.

Q It's an animal. The wasp died. My dog is purple.

Notice that we are assuming that these sentences express propositions; that is to say, the sentences are being used in a particular context with particular reference. We are further assuming that in the first pair it refers to the same entity in each case, and in the second pair, the same wasp and the same event are being referred to. To say that proposition P entails proposition Q means

Logical matters 29

that the truth of Q follows logically and inescapably from the truth of P, and the falsity of P follows likewise from the falsity of Q. So, in the first pair of sentences above, if it is true of some entity that it is a dog, then it follows ineluctably that it is an animal, and if it is not an animal, then there is no way it can be a dog. Similarly, in the second pair, if John killed some wasp, then we cannot avoid the conclusion that the wasp died, and if the wasp did not die, then it cannot be the case that John killed it.

Entailment, as used by linguistic semanticists, is to be distinguished from what logicians call material implication. A proposition P materially implies another proposition Q if and only if (henceforward iff) it is never the case that P is true and Q false. At first sight this seems to be essentially the same as entailment. However, there is a crucial difference: the definition of material implication makes no reference to the meanings of the propositions, merely to a relation between their truth values; entailment, on the other hand, (some- times called strict implication) is essentially a relation between meanings. To illustrate this point, consider the propositions It's a dog and All bachelors are unmarried. It can never be the case that the first is true while the second is false, so we have a case of material implication. But this is not entailment, because it is not a consequence of the meaning relations between the two propositions, but of the fact that the second proposition cannot under any circumstances (except by altering the meanings of the components) be false. On the other hand, although it is true that It's a dog materially implies It's an animal, because the former cannot be true while the latter is false, it is also the case that the former entails the latter, because the truth-value relation holds by virtue of the relation in meaning between dog and animal. Material implica- tion is essentially of no interest to linguistic semantics, although the status of All bachelors are unmarried as being always true is of interest (see 2.5.2.1 below: analytic expressions).

Strictly speaking, entailment does not hold between sentences, because sentences do not have truth values. However, one frequently encounters in semantic texts statements to the effect that such-and-such a sentence entails some other sentence. This can be taken as a kind of shorthand for something slightly more complex. Saying that sentence S1 entails sentence S2 means that in any context where S1 expresses a true proposition, S2 also necessarily expresses a true proposition, provided that corresponding definite referring expressions in the two sentences are co-referential. Obviously, in the case of It's a dog and It's an animal, the two occurrences of it must refer to the same entity for the logical relation to hold, and in the case of John killed the wasp and The wasp died, we must be talking about the same wasp, and the time references must be the same.

Two other properties of entailment must be emphasized. The first is that the relation is not determined by context: it is context independent, since it depends entirely on the meanings of the constituents of the sentences. Con- sider a case where John has in front of him a box of coloured disks, in which

30 Meaning in language

all the red disks are round in shape, and all the green disks are square. In such circumstances, the truth of John picked a square disk from the box follows inescapably from the truth of John picked a green disk from the box. But clearly this relation of truth values does not arise from relations between green and square, but from the context: it would in principle have been just as easy to have all the red disks square, and the green disks round. On the other hand, the relation between It's a dog and It's an animal is independent of any particular contexts.

The second property is that the truth of the entailed sentence must follow inescapably from the truth of the entailing sentence. It is not enough for it to be usually true, or even almost always true; it has to be unthinkable that it might be false. Consider the relation between It's a dog and (i) It's a pet and (ii) It can bark. Most dogs that most people encounter are pets, but there are such things as wild dogs, so the relationship is merely one of expectation. This is not entailment. Likewise in the case of (ii), most dogs can bark, but a dog with a defective larynx does not thereby cease to be a dog, so the relationship is not logically necessary. Only logically necessary, context-independent relation- ships count as entailment. (We shall modify this position in Chapter 3, but for the moment it stands.)

2.5.1.2 Equivalence Propositional equivalence between two sentences can be straightforwardly defined as mutual entailment. That is, in effect, equivalent to saying that the two sentences always express the same proposition (provided, of course, that corresponding definite referring expressions are co-referential). The following are examples of equivalence:

John killed the wasp. The wasp is dead. It began at 10 o'clock.

The wasp was killed by John. The wasp is not alive. It commenced at 10 o'clock.

If it is true that John killed the wasp, then it is also true that the wasp was killed by John and if it is true that the wasp was killed by John, then it is also necessarily true that John killed the wasp; a parallel two-way entailment holds between the members of the other two pairs.

2.5.1.3 Contrariety Contrary propositions may not be simultaneously true, although they may be simultaneously false. The following are examples:

John killed the wasp. John killed the wasp. This paint is red.

The wasp is alive. Mary killed the wasp. This paint is green.

It cannot be simultaneously true of some wasp both that John killed it and that it is still alive; on the other hand, if the wasp is actually dead, but it was

Logical matters 31

Bill who killed it, then both John killed the wasp and The wasp is alive are false. We can define this relation in terms of entailment, by saying that S1 and S2 are contraries iff S1 entails not-S2, but not-S2 does not entail S1 (and vice versa). Thus, This paint is red entails This paint is not green, but This paint is not green does not entail This paint is red, since it might well be, for instance, yellow.

2.5.1.4 Contradiction Contradictory propositions must have opposite truth values in every circum- stance: that is, they cannot be either both true or both false. In any particular circumstance, one member of a contradictory pair must be true and the other false. The following sentences exemplify contradictory pairs:

The wasp is dead. John is still singing. No dogs are brown.

The wasp is alive. John is no longer singing. At least some dogs are brown.

If John is still singing, then it is false that he is no longer singing; to this extent, this is like contrariety. However, there is a crucial difference: if it is false that John is still singing, then it must be the case that he is no longer singing, and if it is false that he is no longer singing, then he must be still singing. We can define contradiction in terms of entailment, by saying that S1 and S2 are contradictories iffS1 entails not-S2, and not-S2 entails S1 (and vice versa).

2.5.1.5 Independence For some pairs of propositions, the truth values vary independently of one another: they may be both true, both false, or one true and the other false:

John is retired. It is Tuesday today.

Mary is married. Christmas day falls on a Wednesday this year.

The relations described in this section have an important role in the analysis of meaning relations between words, as we shall see in later chapters.

2.5.2 Analytic, paradoxical, and synthetic sentences

2.5.2.1 Analyticity Analytic sentences are sentences which automatically express true proposi- tions in any context, by virtue of the meanings of their constituent words and their arrangement. The following sentences are therefore analytical:

Bachelors are unmarried. John's uncle is a man. This proposition is either true or false.

2.5.2.2 Paradox Paradoxical sentences automatically express false propositions: