Palgrave Studies in Pragmatics, Language and Cognition Series Editors: Richard Breheny and Uli Sauerland Series Advisors: Kent Bach, Anne Bezuidenhout, Noël Burton-Roberts, Robyn Carston, Sam Glucksberg, Francesca Happé, François Recanati, Deirdre Wilson Palgrave Studies in Pragmatics, Language and Cognition is a series of high quality research monographs and edited collections of essays focusing on the human pragmatic capacity and its interaction with natural language semantics and other faculties of mind. A central interest is the interface of pragmatics with the linguistic system(s), with the ‘theory of mind’ capacity and with other mental reasoning and general problem-solving capacities. Work of a social or cultural anthropological kind is included if firmly embedded in a cognitive framework. Given the interdisciplinarity of the focal issues, relevant research will come from linguistics, philosophy of language, theoretical and experimental pragmatics, psychology and child development. The series aims to reflect all kinds of research in the relevant fields – conceptual, analytical and experimental. Titles include: Anton Benz, Gerhard Jäger and Robert van Rooij (editors) GAME THEORY AND PRAGMATICS Reinhard Blutner and Henk Zeevat (editors) OPTIMALITY THEORY AND PRAGMATICS María J. Frápolli (editor) SAYING, MEANING AND REFERRING Essays on François Recanati’s Philosophy of Language Corinne Iten LINGUISTIC MEANING, TRUTH CONDITIONS AND RELEVANCE The Case of Concessives Mark Jary ASSERTION Ira Noveck and Dan Sperber (editors) EXPERIMENTAL PRAGMATICS Klaus Petrus (editor) MEANING AND ANALYSIS New Essays on Grice George Powell (editor) LANGUAGE, THOUGHT AND REFERENCE Uli Sauerland and Penka Stateva (editors) PRESUPPOSITION AND IMPLICATURE IN COMPOSITIONAL SEMANTICS Uli Sauerland and Kazuko Yatsushiro (editors) SEMANTICS AND PRAGMATICS From Experiment to Theory
Hans-Christian Schmitz ACCENTUATION AND INTERPRETATION Christoph Unger GENRE, RELEVANCE AND GLOBAL COHERENCE The Pragmatics of Discourse Type
Palgrave Studies in Pragmatics, Language and Cognition Series Series Standing Order ISBN 978–0–333–99010–0 Hardback 978–0–333–98584–7 Paperback (outside North America only) You can receive future titles in this series as they are published by placing a standing order. Please contact your bookseller or, in case of difficulty, write to us at the address below with your name and address, the title of the series and the ISBN quoted above. Customer Services Department, Macmillan Distribution Ltd, Houndmills, Basingstoke, Hampshire RG21 6XS, England.
Meaning and Analysis New Essays on Grice Edited by
Klaus Petrus University of Bern, Switzerland
Selection and editorial matter © Klaus Petrus 2010 Chapters © their individual authors 2010 All rights reserved. No reproduction, copy or transmission of this publication may be made without written permission. No portion of this publication may be reproduced, copied or transmitted save with written permission or in accordance with the provisions of the Copyright, Designs and Patents Act 1988, or under the terms of any licence permitting limited copying issued by the Copyright Licensing Agency, Saffron House, 6-10 Kirby Street, London EC1N 8TS. Any person who does any unauthorized act in relation to this publication may be liable to criminal prosecution and civil claims for damages. The authors have asserted their rights to be identified as the authors of this work in accordance with the Copyright, Designs and Patents Act 1988. First published 2010 by PALGRAVE MACMILLAN Palgrave Macmillan in the UK is an imprint of Macmillan Publishers Limited, registered in England, company number 785998, of Houndmills, Basingstoke, Hampshire RG21 6XS. Palgrave Macmillan in the US is a division of St Martin’s Press LLC, 175 Fifth Avenue, New York, NY 10010. Palgrave Macmillan is the global academic imprint of the above companies and has companies and representatives throughout the world. Palgrave® and Macmillan® are registered trademarks in the United States, the United Kingdom, Europe and other countries. ISBN: 978–0–230–57908–8 hardback This book is printed on paper suitable for recycling and made from fully managed and sustained forest sources. Logging, pulping and manufacturing processes are expected to conform to the environmental regulations of the country of origin. A catalogue record for this book is available from the British Library. Library of Congress Cataloging-in-Publication Data Meaning and analysis : new essays on Grice / edited by Klaus Petrus. p. cm.—(Palgrave studies in pragmatics, languages and cognition) ISBN 978–0–230–57908–8 (hardback) 1. Grice, H. P. (H. Paul) 2. Language and languages – Philosophy. I. Petrus, Klaus, 1967– P85.G735M43 2010 401—dc22 10 9 8 7 6 5 4 3 2 1 19 18 17 16 15 14 13 12 11 10 Printed and bound in Great Britain by CPI Antony Rowe, Chippenham and Eastbourne
2010012502
Contents Acknowledgments
vii
Contributors 1
viii
Introduction: Paul Grice, Philosopher of Language, But More Than That Klaus Petrus
2
Paul Grice and the Philosophy of Ordinary Language Siobhan Chapman
3
Intuition, the Paradigm Case Argument, and the Two Dogmas of Kant’otelianism Jay David Atlas
4
Grice on Presupposition Anne Bezuidenhout
5
Irregular Negations: Implicature and Idiom Theories Wayne A. Davis
1 31
47 75
6
A Gricean View on Intrusive Implicatures Mandy Simons
7
Speaker-Meaning, Conversational Implicature and Calculability Jennifer Saul
103 138
170
8
Some Aspects of Reasons and Rationality Judith Baker
184
9
Showing and Meaning: On How We Make Our Ideas Clear Mitchell Green
202
10 Illocution, Perillocution and Communication Klaus Petrus
221
11
235
Speaker-Meaning and the Logic of Communicative Acts Christian Plunze
12 The Total Content of What a Speaker Means Al Martinich
v
252
vi
Contents
13 On Three Theories of Implicature: Default Theory, Relevance Theory and Minimalism Emma Borg
268
14
288
Contextualism in the Philosophy of Language Nikola Kompa
15 WJ-40: Issues in the Investigation of Implicature Laurence R. Horn
310
Name Index
341
Subject Index
345
Acknowledgments I would like to thank the authors for their readiness to take part in this project. My special thanks go to Richard Breheny, Robyn Carston, Jill Lake, Priyanka Pathak and Jonna Truniger as well as to the members of the research group meaning.ch, especially to Sarah-Jane Conrad, David Lüthi, Jonas Pfister und Annina Schneller. My work on this anthology was supported by the Swiss National Foundation for the Promotion of Scientific Research (project-numbers 100011-105658, 101511-113985 and PP001-114812/1).
vii
Contributors Jay David Atlas, University of Massachusetts, USA Judith Baker, York University, Canada Anne Bezuidenhout, University of South Carolina, USA Emma Borg, University of Reading, UK Siobhan Chapman, University of Liverpool, UK Wayne A. Davis, Georgetown University, USA Mitchell Green, University of Virginia, USA Laurence R. Horn, Yale University, USA Al Martinich, University of Texas at Austin, USA Nikola Kompa, University of Bern, Switzerland Klaus Petrus, University of Bern, Switzerland Christian Plunze, University of Frankfurt, Germany Jennifer Saul, University of Sheffield, UK Mandy Simons, Carnegie Mellon University, USA
viii
1 Introduction: Paul Grice, Philosopher of Language, But More Than That* Klaus Petrus
I shall first proclaim it as my belief that doing philosophy ought to be fun. (Grice 1986: 61)
1. From metaphysics to the philosophy of language The British philosopher H. Paul Grice (1913–1988) is regarded as an eminent representative of Ordinary Language Philosophy and is well-known for his works in the philosophy of language. With only two papers – ‘Meaning’ (1957) and ‘Logic and conversation’ (1967) – he made it into every serious textbook dealing with the philosophy of language, linguistics, communication, or cognitive sciences. Grice, however, was not only a philosopher of language. At least he would have felt exceedingly uncomfortable, if at the department of the University of California–Berkeley he had been introduced to a visitor as ‘Mr Grice, our man in Philosophy of Language’ (Grice 1986: 64). Grice did not attach a lot of importance to the division of philosophy into separate faculties. Philosophy, as he states in his ‘Reply to Richards’, published in the unofficial festschrift Philosophical Grounds of Rationality. Intention, Categories, Ends, ‘is one subject, a single discipline. [ ... ] Or, one might even dare to say, there is only one problem in philosophy, namely all of them’ (Grice 1986: 64). This does not rule out that philosophy has a main subject which connects all putative sub-disciplines with one another. Quite the reverse: ‘It might be held that the ultimate subject of all philosophy is ourselves, or at least our rational nature’ (ibid.: 65; his emphasis). Indeed, rationality is the topic which figures as a kind of leitmotif throughout Grice’s philosophy. He was deeply convinced that human beings – or more precisely, persons – are essentially rational beings (Grice 1991: 140ff.). 1
2
Klaus Petrus
Thus Grice looked again and again at different aspects of human behaviour at the individual stages of his work, no matter whether these aspects were of a linguistic or other kind. His special interest was directed at the mental processes underlying human behaviour and making it explicable as rational behaviour at all (Truniger 2006). Grice was aware of the fact that this approach presumes a comparatively rich ontology. It was a central part of his late philosophy to show that not only psychological concepts like intention, but also concepts like final cause, essential property and especially absolute value are indispensable to any adequate theory of human rationality (Grice 1986, 1991, 2001). Hence in his paper ‘Meaning revisited’, published in 1982, Grice wrote that ‘the notion of value is absolutely crucial to the idea of rationality, or of a rational being’ (Grice 1989: 298). Also in his ‘Reply to Richards’ he stated: ‘I believe (or would like to believe) that it is a necessary feature of rational beings, either as part or as a consequence of part of, their essential nature, that they have a capacity for the attribution of value’ (Grice 1986: 72). As far as the ontological inventory is concerned, Grice committed himself frankly to a ‘new ontological Marxism’ (Grice 1991: 131). In more concrete terms, he followed the principle of metaphysical constructivism according to which concepts or entities of any kind can be postulated within a theory as long as they possess explanatory force (Grice 1986: 68ff., 89f.). Hence it was very important to Grice to concern himself with the methodological requirements of theory construction, and to account for the introduction of certain concepts or entities at the different stages of theory development. This project of theory construction, which Grice called ‘Theory-theory’ (ibid.: 87), seemed to him to be an integral part of any serious science. All the same, within those areas dealing with human rationality, Grice had a certain hierarchy of theories in mind (see Chapman 2005: 173f.). Metaphysics takes the position of some ‘first philosophy’ insofar as it is fundamentally devoted to the question of which materials (categories, substances, subjects, attributes etc.) need to be presumed by every theory (Grice 1986: 86ff., 1991: 23–91). Within metaphysics, every further branch has its own Theory-theory. One of these branches is philosophical psychology, which is concerned with the nature and function of psychological concepts, and marks out the range of psychological beings (Grice 1991: 121–61). A special case of philosophical psychology is rational psychology, which in turn deals with the essence of rational beings as well as with the question of how their behaviour is to be rationally explained (Grice 2001). Finally Grice mentions a sub-branch of rational psychology, which deals with a special form of rational behaviour, viz. linguistic behaviour: the philosophy of language (see section 8, figure 1.3).
2. Grice’s conception of the philosophy of language This hierarchical structure of disciplines is interesting in various respects. In particular, it becomes clear that the philosophy of language is not a
Introduction 3
preferred domain. Yet it is not just a discipline among others either. Rather it is part of a comprehensive theory of rationality within the bounds of which the concept of value is obviously of paramount importance. It is only in his later writings that Grice laid particular stress on this very connection (Grice 1989: 297f., 369, 1991). In his ‘Reply to Richards’, for instance, he writes in full accord with the above exposition that ‘the problems which emerge about meaning are plainly problems in psychology and metaphysics, and I hope that as we proceed it will become increasingly clear that these problems in turn are inextricably bound up with the notion of value’ (Grice 1986: 73). There are, of course, many earlier hints suggesting that Grice considered his investigation into the philosophy of language as contributions to a general account of rationality. This becomes particularly obvious with respect to his famous Cooperative Principle which he puts forward within the framework of his theory of conversation presented in ‘Logic and conversation’. There is good reason to believe that the participants’ cooperative behaviour can only be explained assuming that we are dealing with essentially rational beings (Baker 1989; Grandy 1989; Petrus 1996; Sbisa 2001). Consequently, it has even been suggested that the Cooperative Principle be replaced by a more basic principle of rationality (Kasher 1976). Yet it is perfectly possible to interpret Grice’s theory of speaker’s meaning straight out as an analysis of rational communication (Meggle 1981; Kemmerling 1979, 1986; Petrus 1999), which is, moreover, tightly connected with his considerations about folk-psychological explanations (Grice 2001; see Grandy and Warner 1986: 15ff.). I shall return to these two points in more detail below (sections 5.3 and 6.2). Grice’s idea of a hierarchically structured edifice of sciences is informative in yet another respect: like metaphysics or philosophical psychology, the philosophy of language has its own Theory-theory. Those who concern themselves with it are, in other words, not only dealing with genuinely semantic facts but also with the question of how they approach these facts, how they construct their theories, and for what reason they postulate certain concepts or entities within these theories. Grice has himself pursued this kind of Theory-theory of language (as one could call it) from the very beginning. In his 1957 paper ‘Meaning’ he tries to show that the concept of intention is the basic notion of any adequate theory of meaning (see below, section 6.2). But in the same breath, he clearly states the following: ‘I must disclaim any intention of peopling all our talking life with armies of complicating psychological occurrences’ (Grice 1989: 221). As mentioned before, Grice maintains that concepts should only be built into a, say, theory of meaning if they really possess explanatory force. The proof that it is so is part of the construction of this very theory. Concepts which do not have any explanatory value are superfluous from a theoretical point of view, and should therefore not be used. Grice followed this principle without exception. In his 1978 paper ‘Further notes on logic and conversation’ he
4
Klaus Petrus
refers to it as Modified Occam’s Razor, which he defines as follows: ‘senses are not to be multiplied beyond necessity’ (Grice 1989: 47). Grice’s remarks on the method of ordinary language philosophy (OLP) are relevant in this connection too. Without doubt, throughout his lifetime Grice stuck to the central thesis of this philosophical direction, and defended it several times against more or less mischievous distortions: conceptual analysis of ordinary expressions is definitely an indispensable and integral part of (philosophical) theory construction (Grice 1986: 57ff., 1989: 172ff.). Another thing, however, he was certain about as well: ‘to practice conceptual analysis is not necessarily to practice philosophy’ (Grice 1989: 174). Although he considered linguistic botanizing as practised by John L. Austin and others at Oxford both necessary and useful, he was nevertheless in doubt as to the universal explanatory force of this method (Grice 1986: 57). Often enough it is attributed to OLP that it – unlike ideal language philosophy – conceives of systematic theories as being impossible. Perhaps this is merely a cliché. Should it be correct though, Grice would certainly not be a typical representative of this school (Soames 2003: 216; Atlas 2005: 45). He was not content with informal, case-by-case investigations, but doubtless thought of himself as a constructor of theories possessing as much explanatory force as possible – be this a theory of value, a theory of perception, or a theory of meaning.
3. Meaning and use That Grice always considered the philosophy of language to be a methodological project as well becomes perfectly visible as soon as it comes to his attitude towards a further slogan (or cliché) of OLP: ‘Meaning is use’. According to Grice, this slogan stands for a procedure which always follows more or less the same pattern (Grice 1989: 4ff.). Let us assume that we are interested in certain concepts such as ‘good’ or ‘true’. Those who are committed to OLP will search for ordinary language sentences in which these words occur. In doing so they will find out that the normal use of these sentences implies that a certain condition C is fulfilled. One will, for instance, only say that Pearl Jam is a good grunge band if one is commending it (= condition C); or one will say that it is true that Ferdinand is a dangerous bull only if one knows that it is so, or if one has enough evidence to say so (= condition C). The fact that the standard or appropriate use of sentences like these presumes that a condition C is fulfilled is then taken as evidence that the fulfilment of C is a part of the meaning of those sentences containing the words in question. This procedure was very popular with the people in Grice’s academic surroundings. A case in point is Peter Strawson, who tried to show in his 1950 paper ‘On referring’ that Bertrand Russell’s theory of descriptions does not do justice to the way speakers ordinarily use sentences containing descriptive
Introduction 5
phrases to make statements (Strawson 1950). According to Russell, the proposition expressed by sentences of the form (1) has one of the standard truth values and can be completely characterized using quantifier-variable notation like (2) (Russell 1906): (1) The present King of France is bald. (2) ∃x(Fx & ∀y (Fy → x = y) & Gx) Against this Strawson objects that normally the expression ‘the F’ (i.e. ‘the present King of France’) is correctly used only if there is an F (= condition C). Therefore this condition is a part of the meaning of ‘the F’. If it is not fulfilled – if, as Strawson was later to put it, the presupposition that there is an F is false – ‘the F is G’ cannot be used to express a proposition which is either true or false. From this Strawson draws the further conclusion that ‘neither Aristotelian nor Russellian rules give the exact logic of any expression of ordinary language; for ordinary language has no exact logic’ (Strawson 1950: 344). In Introduction to Logical Theory (1952), Strawson applies the insight that expressions of a natural language do not have an exact semantics which could be captured by the means of classical logic to the analysis of truthconditional constants such as ‘&’, ‘⋁’, or ‘⊃’ (see also Strawson 1986). In this case too he denies that these expressions can capture the meanings of the natural language counterparts ‘and’, ‘or’, or ‘if’. Let us look at the following sentence (3): (3) Mr X is in the kitchen or in the living room. Taken as a strict truth-functional operation, (3) is true iff either one or both of the disjuncts are true, and false iff both disjuncts are false. However, in ordinary language, Strawson holds, the word ‘or’ has another meaning: people who use a sentence like ‘p or q’ do not do this because they already know that p is true, or because they already know that q is true. Rather, they utter this sentence only if they are not sure which of both is true (= condition C). If this condition is not fulfilled there is, according to Strawson, a misuse of language. In other words: It is a part of the meaning of a suchlike sentence that it is used correctly only if the speaker does not know that p is true, and does not know that q is true either. Thus it cannot be correct that the meaning of the word ‘or’ is adequately captured by the formal device ‘⋁’ (Strawson 1952: 78ff.). 3.1 Meaning versus use Grice does not hold the view that this procedure is fundamentally inappropriate or wrong; and he concedes that it sometimes leads to correct results. But at the same time he draws attention to the fact that the identification of meaning
6
Klaus Petrus
with use can sometimes be the source of fundamental philosophical errors (Grice 1989: 4ff.), since by doing so one runs the risk of overlooking central differences between concepts like saying, meaning and use, or of confusing formal or semantic with pragmatic components of language (Grice 1986: 59). It is important to note that Grice’s criticism of the slogan ‘Meaning is use’ is first of all methodologically motivated. Grice emphasizes several times that it is the task of conceptual analysis to determine the actual meaning of an expression. Hence it is decisive to clearly distinguish between the actual meaning and other, additional aspects of the use of an expression. In order to do so Grice has first of all to show that there exists something like the actual meaning at all. The 1956 paper ‘In defense of a dogma’, written with Strawson, was to serve exactly this purpose (Grice 1989: 196–212). Had Quine been right with his objections against the analytic/synthetic distinction (Quine 1953), the concept of meaning would turn out to be a myth indeed, and conceptual analysis in Gricean style would be impossible. In a next step, Grice needs to explain what the difference between the meaning and further aspects of the use of an expression is. A first approach is made in ‘The causal theory of perception’ dating back to the year 1961. Even though this paper is not dealing with issues of the philosophy of language in the narrow sense of the word, there is a section titled ‘Implication’, which contains an answer to Strawson’s analysis of the relation between truth-conditional constants and their ordinary language counterparts (Grice 1961: 131ff.; unfortunately, this section has been left out in the Studies). Grice too concedes that uttering a disjunctive statement ‘p or q’ normally (as he still used to say at that point) implies that the speaker does not know that p, and that she does not know that q either (= condition C). Nevertheless he denies that the correctness of this implication constitutes a condition for the truth or falsity of the disjunctive statement. Neither is it a part of the meaning of ‘p or q’, or something which is implied by this very sentence (or by the statement made). What is implied, in other words, is no Strawsonian presupposition (Grice 1989: 269ff.). This is so because it holds for presuppositions that β is a presupposition of α, just in the case the truth or falsity of α requires the truth of β. To state it more exactly, α loses its truth-value if β is false. Against that Grice maintains that the disjunctive statement ‘p or q’, or the proposition expressed by it, does not cease to be true or false if the implication proves to be wrong. Hence what is implied does not contribute anything to the truth conditions of the utterance. It is not part of what was said by it, but owes itself to the use the speakers make of the sentence in question – and this is actually based on a meaning which has been captured by classical logic. As far as this use beyond meaning or – more exactly – beyond what is said is concerned, Grice comments: [T]he fact that the utterance of the disjunctive sentence normally involves the implication of the speaker’s ignorance of the truth-values of the
Introduction 7
disjuncts is, I should like to say, to be explained by reference to a general principle governing the use of language. Exactly what this principle is I am uncertain, but a first shot would be the following: ‘One should not make a weaker statement rather than a stronger one unless there is a good reason for so doing’. (Grice 1961: 132; his emphasis) This passage anticipates a lot of what Grice is going to elaborate on in his William James Lectures. Part of it is making this principle or ‘pragmatic rule’ (as Strawson, with reference to Grice, calls another principle in footnote 1, p. 179 of his Introduction to Logical Theory) more precise, as well as erecting a typology of different implications. The first leads to Grice’s considerations about the status and the role of the conversational maxims, whereas the latter results in the distinction between conventional and conversational implicatures (as he will call the above-mentioned implications). Taken together, they both form the central components of Grice’s famous theory of conversation (see section 5). Against this background, the theory of conversation turns out to be a methodological instrument (and not simply a theory of rational communication; see Lüthi 2006). And thus considered it is part of some Theory-theory as well (see section 2). What matters to Grice is a criticism of the slogan ‘Meaning is use’ – at least insofar as it amounts to an identification of meaning with use, and thus to a confusion of semantic and pragmatic aspects of language (Grice 1986: 59). 3.2 Meaning as a function of use All this may suggest that Grice took the view that meaning has nothing to do with use. This, however, is not correct. On the contrary, it is thanks to him that it became obvious that the meaning of an expression is a function of what speakers do with it, or, as Grice explains in his paper ‘Meaning’ (1957) for the first time, mean by this expression on particular occasions of use. Grice is convinced that it is constitutive of the basic concept that a speaker by doing something under certain circumstances means something in doing so. All other semantic concepts – such as, for instance, word-meaning, sentence-meaning, or what is said – are derivative and should be explicated in terms of the basic concept of speaker’s meaning (see figure 1.2 on page 17). Thus considered, Grice’s challenge consists in the development of a theory of language which neither identifies meaning with use, nor completely divorces the two. The former, if one likes to put it this way, is to be achieved by the theory of conversation, the latter by the theory of meaning. Like the theory of conversation, Grice’s theory of meaning can be regarded from the viewpoint of a methodological enterprise, or as part of some Theory-theory, since in the end, Grice simultaneously pursues the project of conceptual analysis the purpose of which is to determine the actual meaning of an expression. For
8
Klaus Petrus
this purpose, however, a solid and workable theory of meaning is called for (see section 6).
4. Theory of language In his William James Lectures Grice repeatedly indicates what the purpose of a theory of language should consist in. What it comes down to is the determination of the total signification of an utterance (while ‘utterance’ is to be understood in a very general sense in the context of Grice’s writings). In his 1968 paper ‘Utterer’s meaning, sentence-meaning, and word-meaning’ he proposes the following distinctions within the total signification: [ ... ] a distinction between what a speaker has said (in a certain favored, and maybe in some degree artificial, sense of ‘said’), and what he has implicated (e.g. implied, indicated, suggested), taking into account the fact that what he has implicated may be either conventionally implicated (implicated by virtue of the meaning of some word or phrase which he has used) or nonconventionally implicated (in which case the specification of the implicature falls outside the specification of the conventional meaning of the words used). (Grice 1989: 118; first two emphases by Grice)
what U meant
what U conventionally meant
what U said
what U nonconventionally meant
what U conventionally implicated
what U conversationally implicated
what U particularized conversationally implicated Figure 1.1
Grice’s theory of language
what U nonconversationally nonconventionally implicated
what U generalized conversationally implicated
Introduction 9
As we shall see later on, Grice is especially interested in a particular kind of nonconventional implicatures which he calls conversational implicatures, and which he subdivides into particularized and generalized implicatures (Grice 1989: 40f.). What, moreover, (at least indirectly) follows from passages of ‘Further notes on logic and conversation’ is that Grice equates the specification of the total signification of an utterance x made with a specification of what the utterer U means by uttering x (see Levinson 1983: 131; Neale 1992: 520; but see also Saul 2002). Roughly, this results in the spectrum shown above (figure 1.1), which a theory of language à la Grice should be able to cope with.
5. Conversation and implicatures Grice is very concerned about stating the notions presented in figure 1.1 more precisely, and explaining how they are interconnected. Without doubt, the concept of what is said plays a central role in this respect. On the one hand Grice seems to take the view that what is said by an utterance is a variety of what is meant, or that what is said can be explicated in terms of what is meant (see figure 1.2). The concept of what is said is therefore an integral part of Grice’s theory of meaning. On the other hand it is of eminent importance to the theory of conversation and the characterization of implicatures, insofar as implicatures are usually thus characterized, that U says something and means (implicates) something over and above that. In other words, the concept of what is said constitutes the interface between Grice’s theory of conversation and his theory of meaning (see section 6.1). 5.1 Conventional implicatures What U said by an utterance can, according to Grice, be understood ‘to be closely related to the conventional meaning of the words (the sentence) he has uttered’ (Grice 1989: 25). As Gottlob Frege already wrote in his paper ‘Der Gedanke’ (Frege 1918/19), Grice too notices that the conventional meaning of an uttered sentence does from time to time both fall short and go beyond what is said (Neale 2001; Horn 2007). The first holds if the sentence contains, for instance, indexical and/or ambiguous expressions (e.g. ‘He is in the grip of a vice’). For an identification of what is said one needs to fix the referents of these expressions, and to eliminate ambiguities. The latter holds if the uttered sentence contains conventional devices which signal that U – as Grice puts it – over and above some central speech act performed a further, non-central speech act (Grice 1989: 122). A first example is already given in Grice’s paper ‘The causal theory of perception’. If U utters the sentence: (4) Sally is poor but she is honest.
10 Klaus Petrus
she strictly speaking performs two speech acts: (i) U says that Sally is poor and that she is honest; (ii) additionally, U indicates that there is a contrast between poverty and honesty (or that somebody – perhaps U herself – thinks that this is so). According to Grice, it is decisive that the conventional device ‘but’ in (4) plays a part in figuring out what U meant, or – as Grice puts it – conventionally implicated. This very expression, however, plays no part in determining what U said by (4). In other words, the same is said in (4) and (5): (5) Sally is poor and she is honest. The reason for this is that the conventional implicature generated by ‘but’ (i.e. (ii)) contributes in no way to the truth conditions of the utterance (Grice 1961: 127; but see Bach 1999). This becomes immediately obvious since the conventional implicatum can be false without what is said being false (as regards the difference between implicatures and presuppositions see section 3). Grice terms these implicatures ‘conventional’ because they result from the conventional meaning of words like ‘but’ or ‘therefore’. In order to see that by (4) it is meant that there is a contrast between (Sally’s) poverty and honesty nothing more than knowledge of the linguistic conventions which rule the use of ‘but’ is needed. 5.2 Nonconventional implicatures It is in this respect that all the other forms of implicatures, which could accordingly be termed ‘nonconventional’, differ from the conventional ones (see figure 1.1). Grasping them requires of the audience some extralinguistic considerations which provide the key for working out or calculating the implicature in question (Grice 1989: 31, 39). As regards conversational implicatures the assumption is, roughly speaking, that U cooperates or wants to make a meaningful contribution to the conversation by her utterance. This in turn requires that, for instance, the utterer U and the addressee A know which purpose their conversation serves (and know this of each other as well; but see Gu 1994). That U is cooperative means that she observes the following principle (and that each of the participants supposes that this is the case and that they know this of each other): ‘Make your conversational contribution such as required, at the stage at which it occurs, by the accepted purpose of the talk exchange in which you are engaged.’ (Grice 1989: 26) To this Cooperative Principle (CP) Grice subordinates four categories of maxims and sub-maxims. Maxims of Quantity: (1) Make your contribution as informative as is required (for the current purpose of the exchange); (2) Do not make your contribution more informative than is required. Maxims of Quality: Try to make your contribution one that is true; (1) Do not say what you believe to be false; (2) Do
Introduction 11
not say that for which you lack adequate evidence. Maxim of Relation: Be relevant. Maxims of Manner: Be perspicuous; (1) Avoid obscurity of expression; (2) Avoid ambiguity; (3) Be brief (avoid unnecessary prolixity); (4) Be orderly (ibid.: 26 f.). There has been much speculation about the interconnections of these conversational maxims (CM), their status, their order of importance, or the possibility of a reduction of these maxims to just a few or even to one single maxim (for an overview see Rolf 1994: 102ff.). Grice himself does not claim that this list is complete, or that the maxims are independent of one another (ibid.: 27). At any rate it can be noted that the maxims which belong to (CM) are predominantly the kind of maxims which govern all rational communicative practices, at least insofar as typically linguistic purposes – as, for instance, the maximally effective exchange of information – are concerned. It is exactly this very purpose of (CP) Grice has in mind, and to which he accordingly geared the maxims (ibid.: 28). Even though Grice does not provide a definition of conversational implicatures, he nevertheless lists a number of important features (Grice 1989: 30f.). That by uttering x U conversationally implicates that p roughly means: (a) that by x U neither says nor conventionally implicates that q; (b) that x can only be grasped in accordance with (CM) (or at least (CP)), if by x U means that p; and (c) that by x U means that p, while relying, among other things, on A recognizing that (b). Grice’s basic idea consists therein that typical instances of conversational implicatures result from situations in which the utterance x would be inappropriate on the basis of its conventional meaning alone. Assuming that A wants to know whether Mr X makes a good philosopher, and thereupon U replies: (6)
Mr X has excellent handwriting and is always very punctual
by saying this, U clearly violates one or even several maxims (e.g. the maxims of Quantity or the maxim of Relation). In such a situation U relies on A trying to interpret (6) in such a way that it does not violate (CP) or (CM), and moreover on A being capable of ‘working out’ the implicatum. Grice calls this procedure ‘exploitation’ of the maxims, in which it is central to him that a suchlike violation is made intentionally and overtly. This at least holds for those conversational implicatures which Grice is particularly interested in: particularized conversational implicatures (see figure 1.1). In contrast to generalized conversational implicatures it takes special characteristics of the context of the utterance in order to work them out (as in example (6)). Generalized implicatures on the other hand are relatively context-independent (Hirschberg 1985; Levinson 2000). A violation of (CM) is not detected because the implicatum normally goes with the utterance. Yet there would be a violation if the implicatum was not meant as well
12
Klaus Petrus
(Grice 1989: 37ff.). Examples for generalized conversational implicatures are not merely sentences like (7)
Mr X is meeting a woman this evening.
by which it is normally implicated that Mr X is meeting somebody other than his wife. Also sentences we have already come across (in section 3) and which contain logical constants belong under this rubric as, for instance, (3)
Mr X is in the kitchen or in the living room.
By means of (3) it is normally implicated that the speaker U does not know in which of the two rooms Mr X is. Apart from conversational implicatures, Grice mentions under the rubric of nonconventional implicatures en passant one further type: nonconventional nonconversational implicatures (Grice 1989: 28; see figure 1.1). They differ from conventional implicatures insofar as it is not the case that they are generated on the basis of the conventional meaning of the uttered sentences. On the other hand they differ from conversational implicatures because other maxims than the ones Grice presented (CM) play a part in the inference to what is meant by U, such as, for instance, maxims of politeness. Whether there exists a clear-cut distinguishing criterion between these maxims and (CM) is left open by Grice. As has already been mentioned, there are a lot of indications suggesting that Grice above all searched for maxims which – on the assumption that we are dealing with the maximally efficient exchange of information – govern all reasonable communication. 5.3 Conversation and rationality Grice provides us with a number of tests for the presence of conversational implicatures which are, however, rather controversial (Grice 1989: 39, 43f., 59, 270f.). In the face of these controversies it seems to be more important to focus on the aspect that implicatures can be ‘worked out’, i.e. that they are calculable. In ‘Logic and conversation’, Grice develops a general pattern of how A is capable of grasping a conversational implicature: U has said that p; there is no reason to suppose that he is not observing the maxims, or at least the Cooperative Principle; he could not be doing this unless he thought that q; he knows (and knows that I know that he knows) that I can see that the supposition that he thinks that q is required; he has done nothing to stop me thinking that q; he intends me to think, or is at least willing to allow me to think, that q; and so he has implicated that q. (Grice 1989: 31)
Introduction 13
It has often been criticized that this pattern is psychologically implausible (e.g. Sperber and Wilson 1995; Recanati 1995; Gibbs 1999). Doing so presumes that (CP) and (CM) are socio-psychological and thus empirically verifiable instances of conversation. This does not seem to correspond with Grice’s view. He does not want to claim that we do in fact follow these maxims, but rather that it is reasonable for us to follow them (Grice 1989: 29). The reasons are not of an extrinsic kind; i.e. they do not approach the participants ‘from the outside’ and do not urge them in the form of, for instance, social pressure to behave in a conformist way: the observance of (CP) and (CM) is not a quasi-contractual matter (ibid.). According to Grice, these maxims seem rather to be of a subjective kind. That they are observed is subject to the free will of the participants, and results from the setting of certain goals which they want to achieve: So I would be able to show that the observance of the Cooperative Principle and maxims is reasonable (rational) along the following lines: that anyone who cares about the goals that are central to conversation/communication (such as giving and receiving information, influencing and being influenced by others) must be expected to have an interest, given suitable circumstances, in participation in talk exchanges that will be profitable only on the assumption that they are conducted in general accordance with the Cooperative Principle and the maxims. (Grice 1989: 29f.) As this passage shows, Grice at the same time seems to hold that (CP) and (CM) are observed ceteris paribus – and thus claim objective validity in certain respects. But what could that mean? A possible answer can be found in Grice’s considerations about the relation between rationality and value. As already mentioned in section 1, it is an essential feature of rational beings – Grice calls them ‘persons’ – that they are capable of attributing values (Grice 1986: 72). With that goes, among other things, the capability of evaluating goals and means which serve the attainment of these goals (ibid.: 112ff.). In the light of this idea, the participants’ readiness to follow (CP) and (CM), as well as their capability of assessing linguistic behaviour in its respective context as appropriate or inappropriate, can be regarded as a manifestation of this essential feature of rational beings. Just as much as U, with a view to the goal she pursues, is concerned about taking the most effective and thus considered the most ‘valuable’ way, A – with this goal given – is concerned with working out in a calculability process the rationally most acceptable interpretation of what U meant. What U (and A as well) do in the actual case, becomes intelligible only if the participants mutually assume that they are essentially rational beings – and thus beings which ceteris paribus follow (CP) and (CM) in conversational exchange. That A and U, as Grice puts it, ‘cooperate’ could, thus considered, mean that Gricean persons mutually assign rationality (and with this the capability of attributing
14
Klaus Petrus
values) to each other (Truniger 2006: 90ff.). In any case, the (mutual) supposition that (CP) and (CM) are observed would according to this interpretation bear objective validity insofar as it constitutes a condition of the possibility of conversation. Without this supposition the linguistic behaviour of those who engage in conversation would not be explainable at all. 5.4 The two sides of the theory of conversation In the preceding paragraphs I presented Grice’s theory of conversation as a contribution to a general theory of rationality. This corresponds with his idea of a hierarchically structured edifice of sciences (see figure 1.3 on page 24). Grice’s theory of conversation is part of his theory of language, which in turn belongs to the philosophy of language. He conceives of the philosophy of language as a domain of rational psychology which for its part belongs to philosophical psychology. All these areas are finally to be located within ‘first philosophy’, i.e. metaphysics. Given that these domains all together revolve around human rationality, it is a matter of consequence that they are dealing with aspects of this subject in an increasingly specialized manner. The theory of conversation is such a sub-branch; its topic is linguistic behaviour insofar as it can be interpreted as rational behaviour. Nevertheless, it should not be forgotten that the theory of conversation has another side too. As explained in section 3.2, it is essential to Grice’s understanding of conceptual analysis to clearly distinguish between the meaning of an expression and other facets of its use. It is exactly this very purpose the theory of conversation has to serve as a methodological tool: to prevent that meaning and use are telquel identified. In his ‘Reply to Richards’ Grice notes the following: In my own case, a further impetus towards a demand for the provision of a visible theory underlying ordinary discourse came from my work on the idea of Conversational Implicature, which emphasized the radical importance of distinguishing (to speak loosely) what our words say or imply from what we in uttering them imply; a distinction seemingly denied by Wittgenstein, and all too frequently ignored by Austin. (Grice 1986: 59, his emphasis) Hence Grice’s theory of conversation has two sides: as a philosophical domain it covers a certain domain of a general theory of rationality; its topic is rational linguistic behaviour. And in methodological respects it serves as an instrument for the distinction between meaning and use.
6. Meaning and rationality Like the theory of conversation, Grice’s account of meaning has its two sides. As we shall see later on, it has made a contribution to the theory of
Introduction 15
rationality as well (see below, section 6.3). At the same time it serves an important methodological purpose. Even though Grice is against an identification of meaning with use, this does not mean that the one has nothing to do with the other. On the contrary, Grice is firmly convinced that the meaning of an expression is a function of what a speaker U does with it or what she means by it (Neale 1992: 542 as well as this chapter, section 3). In this context, Grice needs, among other things, to clarify what it actually means to mean something by an utterance. 6.1
The analysis of speaker’s meaning
A first outline of such an analysis can be found in his famous 1957 paper ‘Meaning’, in which against the background of the distinction between natural and nonnatural meaning he pursues the question of what it means that utterances have meaning at all, and, on the other hand, the question of what exactly they mean. Roughly speaking, his analysis consists of two steps. In a first step, the nonnatural meaning of a certain utterance x is equated with what a certain speaker U meant by x under certain circumstances. What she meant by x is in turn explicated by which effect U intended to produce in an audience A. In concrete terms, Grice’s analysis goes as follows (Grice 1989: 92, 219): (GA) ‘U meant something by uttering x’ is true iff, for some audience A, U uttered x intending: (1) A to produce a particular response r (2) A to think (recognize) that U intends (1) (3) A to fulfill (1) on the basis of this fulfillment of (2). If (1)–(3) are present, U meant, or in brief: M-intended, something by uttering x. What exactly U meant by uttering x results, according to Grice, from a specification of the reaction r intended on the part of A (ibid.: 92, 105, 220). Hence Grice explicates both the fact that U meant something by x, and the fact of what she means by uttering x, in terms of psychological concepts. In the first case it is the concept of intention, whereas in the second case it is the concept of (as Grice calls it) the M-intended effect, which is always a propositional attitude on the part of A. With respect to ‘imperativetype’ utterances it consists in A’s intention to do ψ; with respect to assertive or ‘indicative-type’ utterances the effect consists in A believing (that U believes) that p (ibid.: 111, 123). It is important to note that these psychological concepts are the basic concepts of Grice’s theory of meaning (see also below, section 6.2): Grice conceives of meaning something by x as intending (1)–(3) or producing a reaction in the addressee A in the way explicated in the analysis. Therefore it is central to determine the status as well as the role of these concepts in
16
Klaus Petrus
respect of the explanation of linguistic acts. This is a project Grice does not realize within the framework of his theory of meaning but within his philosophical psychology or rational psychology (Grice 1991: 121–61, 2001; see Grandy and Warner 1986a: 15ff.; Plüss 2001). Intentions (1)–(3) have been the subject of extensive and lengthy debates (e.g. Strawson 1964; Searle 1969; Schiffer 1972). Disregarding the (technical) problems of the Gricean analysis, one thing stands out immediately: conventions do not occur. In fact it is the crux of this approach that for U meaning something by x it is not necessary that there exists a previously conventionally established relation between the means by which something is meant (a gesture, a drawing, or even the utterance of a sentence) and what is meant. Nevertheless Grice does of course not want to deny that there is something like conventional meaning. But what is meant, i.e. nonnatural meaning, is not always conventional meaning as well (the other way round, however, this holds). Also he does not deny that, in a certain respect, linguistic acts are always conventional acts, namely insofar as their medium (e.g. natural language) is conventional. Yet Grice thinks that such acts could, at least in principle, be performed without conventions as well. This idea, by the way, has some far-reaching consequences. Already in the 1960s it was applied to the analysis of a central type of linguistic actions, viz. to illocutionary acts conventionally interpreted by Austin (Austin 1962; see Strawson 1964; Searle 1969; Schiffer 1972; Bach and Harnish 1979). The debate between intentionalists and conventionalists goes on to this day (see Petrus 2006). Finally, Grice does not want to deny either that especially in the case of linguistic communication conventional means are often used to mean something which exceeds what is said. Conversational implicatures are the apt example for this (see section 5.2). Grice, however, does not content himself with an analysis of speaker’s meaning in terms of intentions. In ‘Meaning’ he already indicated that it ought to be the objective of a second step to explicate all other semant-ically relevant concepts on the basis of this prior analysis: in particular the concept of the conventional meaning of an utterance-type (termed ‘timeless meaning’ by Grice), and the concept of what is said by an utterance (Grice 1989: 217). As has already been emphasized in section 5, it is the concept of what is said which is important to Grice in order figure out (conventional as well as conversational) implicatures, which themselves in turn are a variety of what is meant by the speaker (see figure 1.1). Presented in a simplified way, Grice aims at proving the conceptual dependencies illustrated in figure 1.2 (Neale 1992: 543). The arrows in the scheme, i.e. ‘α→β’ are to be understood as indicating that the analysis of concept α is logically prior to the analysis of concept β. And this is the case iff the concept α plays a role in the analysis of concept β, and the concept β does not play a role in the analysis of concept α. The idea,
Introduction 17
(2) utterance-type meaning
(1) utterer’s meaning
(3) what is said
(4) what is implicated Figure 1.2 Hierarchy of concepts in Grice’s theory of meaning
roughly speaking, is to provide an analysis of (1) first of all. In a second step this analysis is used in the analysis of the meaning of an utterance-type. What is said by an utterance is then explicated in terms of (1) and (2), and after that what is implicated (4) is explicated in terms of (2) and (3). Thus conceived, Grice’s theory of meaning consists of two parts. The first part consists in an analysis of (1), which is to be achieved by (GA), i.e. the Gricean analysis of speaker’s meaning. The second part consists in showing that (3) can be analysed on the basis of (2), and that (2) in turn can be analysed on the basis of (1). Let us call this project, which Grice particularly tackles in his paper ‘Utterer’s meaning, sentence-meaning, word-meaning’, the Gricean programme (Grice 1989: 117–37; see Schiffer 1972; Bennett 1976; Loar 1981; Avramides 1989). 6.2 Meaning and rationality Grice’s theory of meaning – no matter at which stage of his programme – is based on a very special mechanism, which is sometimes referred to as the ‘Gricean mechanism’ (Bennett 1976: chapter 5). Roughly speaking, its point is that the speaker U makes certain assumptions about what could lead the addressee A to produce the reaction r intended by U. Yet it is not (only) the natural features of U’s x-ing which lead A, for instance, to believe that p. The reason for producing this reaction should, according to Grice, rather be one which U herself provides to A. From U’s perspective the reason is that A recognizes that U intends to make A believe that p. In other words, it is exactly this recognition which should be (among others) a reason – and not merely a cause (Grice 1989: 92) – for A to produce the reaction intended by U, i.e. to believe that p (see condition (3) in (GA), section 6.1). In a simplified version, the pattern of the Gricean mechanism (for assertive utterances) looks as follows (Kemmerling 1991: 322): (GM)
(i) (ii) (iii) (iv)
By x-ing U intends A to believe that p. A recognizes that (i). A believes because of (ii) that if (i) then p. Therefore: A believes that p.
18 Klaus Petrus
Hence the point of (GM) is that A’s recognition of the primary intention which U pursues in the sense of the Gricean analysis – i.e. (i) – leads to its fulfilment, i.e. to (iv). In this respect (GM) can be conceived of as an intention-fulfilling-mechanism (Kemmerling 2001: 79). But it would be wrong to assume that (ii) is always sufficient for A to produce the reaction r intended by U. It is likely that A makes further assumptions, and that these assumptions count as additional reasons to believe that p. As regards assertive utterances of U, A’s assumption that x is a suitable means for U’s purposes, that U is sincere and truthful, or reliable with respect to p, can very well be part of these reasons (Grice 1989: 294f.; see Petrus 1999: 216ff.). Either way it is assumed in this model that it is – as Grice occasionally puts it – within the rational control of A to produce the reaction intended by U. All this might suggest that Grice’s analysis requires that A really recognizes what U intended, or even that A actually produces the reaction intended by U because of this recognition. But in (GA) there is no mention of this at all. The analysis only requires that U has certain intentions. Therefore it is important to note that the above statements about A are strictly speaking assumptions which U makes about A. That U relies on the Gricean mechanism is to say that U presumes that it is within the rational control of A to produce the reaction intended by her – and not that this reaction is actually within the control of A (see Petrus 1999: 227ff.). That U makes suchlike assumptions has its reason and reveals the deeperlying rationality, which is inherent to the Gricean theory of meaning. If U did not rely on A taking her recognition as a reason to produce reaction r, or if U did not presume that A on her part assumes that U’s x-ing is a suitable means, that U is sincere, truthful and reliable, U would hardly have a chance of reaching her goal, and of producing the desired reaction in A (Grice 1989: 219). In other words, if U did not rely on the Gricean mechanism, and did not make the relevant assumptions about A, she herself would have no good reason to do what she is up to. It is only on the assumption that U relies on the Gricean mechanism that her behaviour can be explained as rational behaviour at all. 6.3 The two sides of the theory of meaning These brief remarks show that in Grice’s theory of meaning, too, rationality plays a central part: meaning something by an utterance is an attempt to influence someone in a rational way. To this it could be objected – similarly as regards Grice’s theory of conversation (see section 5.3) – that Grice’s approach seems to be highly implausible from a psychological point of view. But this objection is liable to miss the point. What matters to Grice is to work out mechanisms and concepts which are indispensable to making explicable the behaviour of beings which, in the Gricean sense, mean something by a sign. And a crucial part of this explanation is provided by the Gricean mechanism.
Introduction 19
As in the case of the theory of conversation, it should be taken into consideration that Grice’s theory of meaning does not merely make a contribution to a theory of rational, linguistic behaviour. Rather – or more precisely, at the same time – it serves a methodological purpose. Although meaning and use are not to be identified (see section 3.2), Grice holds that these two components are not completely isolated from each other. According to him, the meaning of an utterance is rather a function of its use. What expressions mean depends on what speakers do with them or, à la Grice, what they mean by them. The answers to the questions of what exactly this means, and in which sense it holds for basic semantic concepts like the concept of meaning of an utterance-type or what is said as well, are provided by Grice’s theory of meaning. Analogous to the theory of conversation, the theory of meaning too has always its two sides: as a philosophical domain it covers a certain domain of a general theory of rationality; its topic is rational linguistic behaviour. And in methodological respects it serves as an instrument to show that meaning is a function of use.
7. Meaning, saying, implicating In section 6.1, I distinguished two parts of Grice’s theory of meaning which, of course, are closely related: the one part – the Gricean analysis – deals with an explication of speaker’s meaning in terms of intentions. The second part – the Gricean programme – is concerned with making use of this analysis when it comes to the explication of the meaning of an utterance-type and what is said. While some people think that the first part can be successfully realized along the lines of Grice’s proposal, there is much scepticism about the second part. And indeed, the Gricean programme proves to be extremely difficult in detail, as Grice’s own remarks show only too clearly (Grice 1989: 117ff.). 7.1
What is said
A special problem is posed by the concept of what is said. As can be seen in figure 1.2, Grice maintains that this concept can at least partly be explicated by the concept of the meaning of an utterance-type X. All the same, the transition from ‘When uttered by U, X meant ‘p’ to ‘By uttering X, U said that p’ would be too rash. In order to say something by X, X has to possess a certain feature: doing X needs to be a linguistic act (Grice 1989: 87). The hand-signalling of a biker may well mean ‘p’, but he does nevertheless not say that p (at least not – as Grice never omits to add – in the favoured sense of ‘say’). Grice does justice to this fact by expressly demanding that X ‘is an occurrence of an utterances type S (sentence)’. At least according to some scholars (see Bach 2001a: 15), what is said à la Grice must then correspond to ‘the elements [of the sentence], their order, and their syntactic character’ (Grice 1989: 87). For the sake of simplicity, let us say that according to these
20
Klaus Petrus
conditions what is said is syntactically correlated with the sentence uttered by U (see section 7.2.1). Off-hand it may well seem as if what is said coincides with a special sort of meaning of an utterance-type, namely with sentence-meaning. This, however, is not correct either. Even if we disregard for a moment the problems which are posed by indexical and ambiguous expressions (see section 5.1), there are still utterances which do in fact have sentence-meaning, but by means of which nothing is said in the Gricean sense. Examples for this are provided by instances of non-literal speech such as irony (Grice 1989: 34). Assuming somebody utters sentence (8) ironically: (8)
Fred is an honest man.
Of course it is correct to say that (8) means ‘Fred is an honest man’. But it would be wrong to say that by (8) U said that Fred is an honest man. The reason for this is that by uttering (8) U did not mean that Fred is an honest man: U has not the intention to make A believe that Fred is an honest man. If at all, by (8) U has made as if to say that Fred is an honest man (ibid.: 34). Thus what is said (in the strict sense) is always a variety of what is meant (see figures 1.1 and 1.2). According to Grice what is said is therefore located in a sphere where sentence-meaning and speaker-meaning overlap, and which he terms ‘conventional meaning’ (ibid.: 119). But this too is not yet correct. This time, the reason is constituted by the conventional implicatures. Assuming U utters the sentence: (4) Sally is poor but she is honest. Grice would not want to deny that both what U means by (4), and what this sentence means is that Sally is poor but honest. Nevertheless he disputes that U said so (ibid.: 88). As explained in section 5.1., Grice claims that the meaning evoked by ‘but’ does not contribute to the truth conditions of (4): the (conventional) implicatum should always be capable of being false without what is said by an utterance becoming false – which, by the way, shows that what is said in the Gricean sense must always be propositional. Grice rather thinks that by (4) U performs two speech acts – a central one and a non-central one – and that only the propositional content of the central speech act is to be regarded as part of what is said. For this reason he stipulates that what is said is strictly speaking a form of what is centrally meant (ibid.: 88). 7.2
Problems with what is said
In recent years, an intense debate about the concept of what is said has flared up, taking up Grice’s considerations more or less directly (for an
Introduction 21
overview: see Pfister 2007). Since the concept of what is said has traditionally been assigned to the domain of semantics, and the concept of implicature, on the other hand, to pragmatics, this debate has at the same time been revolving around the semantics/pragmatics distinction (see contributions in Bianchi 2003; Szabo 2005; Preyer and Peter 2007). In the meantime a number of competing positions have evolved – literalism, minimalism, indexicalism, contextualism, or syncretism – and an end to the debate is not within sight. Indeed, this debate poses a challenge to Grice as well. As has become clear from the preceding section, what is said à la Grice possesses at least the following three features: what is said (i) is syntactically correlated with the sentence uttered; (ii) is a proposition; and (iii) is (centrally) meant by U. The problem with what is said now results from the observation that many (perhaps even most) sentences of a natural language – even after the determination of the referents of indexical expressions, and after disambiguation – do not express a content which possesses features (i)–(iii). Accordingly, nothing is said by all these sentences, at least in the Gricean sense. Let us consider the following examples: (9) Maria departs. (10) Gioacchino is too old. Although these uttered sentences are syntactically complete, (9) and (10) are nevertheless semantically incomplete. They do not express a proposition, but at best a propositional radical (Bach 1994; but see Cappelen and Lepore 2005). In other words, nothing is said by these sentences, since feature (ii) is not present. In order to express a proposition, they need to be completed, that is to say it must be specified from where Maria departs, and what Gioacchino is too old for (see Bach 1994). This completion needs to be done in accordance with what U meant by these sentences, and thus depends on the context of the utterance. The propositions expressed by (9) and (10) could, for instance, be spelled out as follows: (9*) that Maria departs from Palermo (10*) that Gioacchino is too old to play in the football team After having been completed like this, what is expressed by the examples possesses indeed features (ii) and (iii). Nevertheless it is questionable to what extent the so-specified content still does justice to requirement (i) – at least if syntactic correlation, as suggested by Grice, requires that every element of what is said correspond to some element of the uttered sentence. If one held this opinion, all features (i)–(iii) would be fulfilled, and by sentence (10), for example, it would be said in the Gricean sense that Gioacchino is too old to play in the football team. Otherwise nothing would be said by it.
22
Klaus Petrus
Another conflict between the features of what is said results from sentences of the following type: (11) Maria and Gioacchino are engaged. (12) Maria has nothing to put on. Once again we are dealing with syntactically complete sentences. Moreover they are – unlike (9) and (10) – semantically complete: they express a proposition, and thus fulfil requirement (ii). Yet it is – normal circumstances provided – highly unlikely that these sentences express what U meant by them. It is to be expected that U meant something more specific than he explicitly expressed. But then they do not possess the required feature (iii). Accordingly, the proposition expressed by these sentences needs to be expanded in accordance with what U meant (Bach 2001a; Carston 1988; 2002: 21ff.; Recanati 2004: 23ff.). In the case of (11) and (12), such an expansion could yield the following propositions meant by U: (11*) that Maria and Gioacchino are engaged to each other (12*) that Maria has nothing appropriate for the party to put on If the examples are expanded like this, what is expressed by them possesses in turn features (ii) and (iii). Nevertheless, the question whether the thus specified content still does justice to requirement (i) arises once more. If it were answered in the affirmative, all the features (i)–(iii) would be given, and by (12) it would be said that Maria has nothing appropriate for the party to put on. Otherwise nothing would be said. 7.3 Three conceptions of what is said The examples above show that in the end we are dealing with the question as to which of the above mentioned features priority should be given – and at the same time with the question of how narrowly or broadly the concept of what is said should be interpreted. Starting out from Grice’s considerations, we can distinguish at least three conceptions of what is said (Harth 2003). According to the minimal conception of what is said, only the content resulting after reference fixation of indexical expressions and disambiguation ought to be regarded as what is said. As a consequence of this conception, what is said needs always to fulfil the requirement of syntactical correlation, viz. feature (i). What is said by (10) is therefore that Gioacchino is too old; what is said by (11) is correspondingly that Maria and Gioacchino are engaged. According to the propositional conception of what is said, what is said must always be truth-evaluable, that is to say, needs to possess feature (ii). Accordingly, the content of an utterance needs (where necessary) to be completed in such a way that a complete proposition is expressed by the sentence
Introduction 23
in question. Everything else – and thus any possible expansions – is not counted as being part of what is said. What is said by (11) is correspondingly that Gioacchino is too old to play in the football team, whereas in (12) it is ‘merely’ said that Maria and Gioacchino are engaged. According to the maximal conception of what is said, what is said ought not only to be truth-evaluable, but also and always to state the proposition meant by U. Accordingly, the content of an uttered sentence needs (where necessary) not only to be completed, but also to be expanded. What is said by (11) is therefore that Gioacchino is too old to play in the football team; and what is said by (12) is that Maria and Gioacchino are engaged to each other. There are a lot of reasons speaking for and against these various conceptions of what is said. Which of them is finally chosen may well depend on the theoretical purpose that the concept of what is said is to serve. As has been emphasized several times, Grice himself makes high demands on this concept: what is said constitutes the interface between the two facets of his theory of language, viz. the theory of meaning and the theory of conversation. On the one hand, the concept of what is said is inherent to the concept of speaker’s meaning; on the other hand it plays a central part on all the occasions on which aspects of the use of an utterance are to be determined which do not belong to the meaning in the actual sense (i.e. implicatures). Since the delimitation of what is implicated and what is said also depends on the question of which components within the total meaning of an utterance contribute to its truth conditions, it may well be supposed that Grice would favour the propositional conception of what is said – though together with the, to his mind crucial, remark that the completion in question (where necessary) ought to be carried out in accordance with what U meant by her utterance (since what is said is always a form of what is meant). While doing so Grice would of course (have to) admit that he has overlooked the problem of semantic underdeterminacy which results from sentences like (9) and (10). Expansions, on the other hand, would in Grice’s view presumably be regarded as belonging to the domain of implicatures, although for different reasons. A lot speaks for regarding cases like (12) as a particular form of non-literality. Although all the components of the uttered sentence are to be taken literally, the utterance as a whole needs to be interpreted in a nonliteral sense (Bach 2001b). On this assumption U would, strictly speaking, have said nothing by (12) – as in the case of irony – but merely made as if to say that Maria has nothing to put on (see section 7.2). Whether this is the route Grice would have taken in view of all the problems, is, as indicated, quite controversial. But on the other hand nobody disputes that the concept of what is said is assigned a key function in Grice’s conception. In anticipation of the ensuing debates, Stephen Neale made the following remark towards the end of his 1992 essay on Grice’s philosophy of language: ‘The more one reflects on his work, the more one feels that
24 Klaus Petrus
the notion of what is said is for Grice a fundamentally important notion in philosophy’ (Neale 1992: 555).
8. Philosophy of language and the theory of rationality In the course of the preceding pages I tried to go into some aspects of Grice’s philosophy of language in more detail and to integrate them into a broader picture. This picture is the picture of human rationality, to draw which was Grice’s primary concern (see figure 1.3). The general frame of his theory of rationality is constituted by metaphysics. Within this province exists a multitude of domains which look at the subject in increasingly more specific ways. Among them rank philosophical psychology, rational psychology, and – last but not least – the philosophy of language: its preferred, specialized field of research is defined by linguistic behaviour insofar as it can be interpreted as rational behaviour. Grice always claimed to develop theories with the most general explanatory force possible. Thus regarded, it is crucial to his philosophical work to always and constantly reflect upon the principles of theory construction. Among others there is the question of which concepts are to be introduced
Metaphysics Philosophical Psychology Rational Psychology Philosophy of Language Theory of Language Theory of Langu
Theory of Meaning
Theory of Conversation
Utterer’s Meaning → Saying →
Implicating
Methodology
Topic
Methodology
Meaning as a function of use
Theory of rational linguistic behaviour
Distinction between meaning and use
Figure 1.3 The philosophy of language within a general theory of human rationality
Introduction 25
at which stage of theory construction, and why this is done. This question is posed in all scientific branches, and consequently also in the philosophy of language. Answering this question is within the scope of Theory-theory. In this respect Grice seems to be firmly convinced that this methodological enterprise is not something which is exterior to the theory of language, or a mere addition from outside. Rather (as the name suggests) Theory-theory is a component of this theory of an aspect of human rationality. This, in the end, might be the source of the fascination of the Gricean approach: Grice’s theory of language (i.e. the theory of meaning and the theory of conversation) is not only concerned with rational linguistic behaviour, but it is also and at the same time to be regarded as a contribution to the construction of a theory about rational linguistic behaviour.
9. The chapters In the anthology at hand, most aspects mentioned in this introduction are deepened and sometimes controversially discussed. In her contribution, Siobhan Chapman sheds light on the complex relationship between Grice and ordinary language philosophy (OLP). In particular, she suggests a movement in Grice’s work initially towards the doctrines of OLP and then gradually away from them as his philosophy became increasingly formal. Jay Atlas too starts out from Grice’s emphasis on the ‘ordinariness’ of the intuitions about meaningfulness, synonymy and analyticity. Among other things, he examines critically Grice and Strawson’s ‘argument from intuition’ in support of the analytic/synthetic distinction against Quine’s attack on it, as well as their reductio argument against Quine’s ‘elimination’ of the synonymy relation. Anne Bezuidenhout devotes her contribution to the debate between Russell and Strawson about the status of implications of existence and uniqueness associated with the use of definite descriptions. She tries to show that Grice’s claim that they are conversational implicatures rather than Strawsonian presuppositions can be accommodated within a presuppositionalist framework. Wayne Davis has a close look at different types of negations that are not used in accordance with standard rules of logic, and raises, among others, the question of whether such irregular negations are pragmatically or semantically ambiguous. In his opinion, an implicature theory works well only for evaluative implicature denials, while other irregular negations are semantically ambiguous in an unusual way. In her contribution, Mandy Simons argues for a reconceptualization of conversational implicatures, according to which they ought to be conceived of as inferences generated by observations about what a speaker has (merely) expressed. This includes observations about sentence parts whose content is
26 Klaus Petrus
not part of what is asserted. In order to show how Gricean maxims can be applied to non-asserted content, Simons offers some reformulations of the maxims, especially of the first part of the maxim of Quantity. According to Jennifer Saul there is a real conflict in Grice’s theory between, as she puts it, ‘Speaker Meaning Exhaustiveness’ (i.e. the claim that Grice’s notion of speaker-meaning divides exhaustively into what is said and what is implicated) and Grice’s Calculability Criterion. Saul comes to the conclusion that abandoning Speaker Meaning Exhaustiveness is the preferable alternative, which moreover leaves us with an interesting theory of conversational implicatures as a normative notion. The central topic of the contribution of Judith Baker is the role of rationality, which was repeatedly emphasized in the course of this introduction. However, Baker is primarily interested in the relation of practical reasons to rationality and focuses on one particular claim and the arguments that support it, i.e. that rationality is displayed in a wider variety of actions than those done for reasons. Mitchell Green, for his part, deals with meaning and communication and tries to revitalize the discussion of illocutionary force by relating it to social norms. In particular, he argues that force is an aspect of speaker-meaning and that according to new research in the evolutionary biology of communication, speaker- and natural meaning are more complexly interrelated than current consensus allows. Klaus Petrus too concerns himself with the relation between illocution and Gricean-style communication. Unlike many intentionalists, he holds that Grice’s theory of meaning is not suitable for the analysis of illocutionary acts, because they are not communicative acts. It is, however, appropriate for one kind of speech acts which are easily confused with illocutionary acts and which are communicative acts indeed: perillocutionary acts. Communicative acts are likewise the topic of the contribution of Christian Plunze. He goes into the claim that the recognition of a communicative intention implies communication which allows different interpretations. According to one interpretation, it is appropriate to speak of successful communication if the communicative intention is recognized. According to another interpretation, the fact that the addressee does not recognize this intention does not show that the speaker does not communicate at all. Plunze offers an account of communicative acts which enables him to explain why the latter interpretation holds. In his chapter, Al Martinich presents a modified version of Grice’s view of the total content of what a speaker means. According to Martinich, what is said is as much a component of pragmatics as is what is implicated. This modified Gricean theory requires a defence against criticisms of the concept of what is said advanced by some minimalists, as well as against criticisms of the concept of what is said and conversationally implicated advanced by some contextualists.
Introduction 27
Emma Borg takes such a minimalist point of view. Her starting point is the fact that in recent literature the precise account Grice offered of implicature recovery has been questioned and alternative approaches have emerged from different semantic programmes. Borg tries to show that the most popular accounts – the default inference view and the relevance-theoretic approach – face significant problems and that an account emerging from semantic minimalism is best placed to accommodate Grice’s distinction between what a sentence means and what utterances of it implicate. Nikola Kompa goes into semantic minimalism as well and contrasts it with different versions of contextualism. Like Borg, Kompa assumes that a defining feature of semantic minimalism is the so-called ‘Basic Set Assumption’, i.e. the assumption that there is only a small set of obviously context-sensitive expressions (mainly indexicals, demonstratives, ambiguous expressions). By looking at various forms of context sensitivity, Kompa tries to show the basic set assumption to be mistaken. Finally Laurence Horn presents once more an overview of the lively debate about the status of conversational implicatures and their relation to what is said. The background is provided by criticism of the customary semantic/ pragmatics distinction particularly voiced by contextualists. Horn makes the case for a relatively orthodox Gricean notion of what is said and tries to show that a minimally modified Gricean model provides a satisfying approach to the fundamental distinction between what is said and what is meant.
Note * An anonymous referee noticed that my introduction suggests, in places, that Grice’s work is set against a ‘grand scheme’ which he systemizes in the way I have expounded. This is countered with the claim that Grice himself was ‘the most evasive about definitive interpretations of the nature of his enterprise’, as the referee remarked. I agree with the latter. However, I believe that Grice did indeed perceive this work on philosophy of language, in particular, within a wider framework. My introduction is an attempt to expatiate on this framework based on the remarks made by Grice himself in his ‘Reply to Richards’, for instance. This work was supported by the Swiss National Foundation for the Promotion of Scientific Research (project-numbers 100011–105658 and PP001–114812/1). Special thanks to Jonna Truniger, Sarah-Jane Conrad and Jonas Pfister. I dedicate this piece of work to Andreas Graeser.
References Austin, J. L. (1962) How to Do Things With Words (Oxford: Clarendon Press). Atlas, J. D. (2005) Logic, Meaning, and Conversation: Semantic Underdeterminacy, Implicature, and Their Interface (Oxford: Oxford University Press). Avramides, A. (1989) Meaning and Mind: An Examination of a Gricean Account of Language (Cambridge, MA: MIT Press).
28
Klaus Petrus
Bach, K. (1994) Semantic slack: what is said and more. In S. Tsohatzidis (ed.) Foundations of Speech Act Theory: Philosophical and Linguistic Perspectives (London: Routledge), pp. 267–91. Bach, K. (1999) The myth of conventional implicatures. Linguistics and Philosophy, 22: 367–421. Bach, K. (2001a). You don’t say. Synthese, 128: 15–44. Bach, K. (2001b) Speaking loosely: sentence nonliterally. Midwest Studies in Philosophy, 25: 249–63. Bach, K. and Harnish, R. M. (1979) Linguistic Communication and Speech Acts (Cambridge, MA: MIT Press). Baker, J. (1989) The metaphysical construction of value. Journal of Philosophy, 10: 505–13. Bennett, J. (1976) Linguistic Behaviour (Cambridge: Cambridge University Press). Bianchi, C. (ed.) (2003) The Semantics/Pragmatics Distinction (Stanford: CSLI Publications). Borg, E. (2004) Minimal Semantics (Oxford: Oxford University Press). Cappelen, H. and Lepore, E. (2005) Insensitive Semantics: A Defense of Semantic Minimalism and Speech Act Pluralism (Oxford: Blackwell). Chapman, S. (2005) Paul Grice, Philosopher and Linguist (Basingstoke and New York: Palgrave Macmillan). Carston, R. (1988) Implicature, explicature, and truth-theoretic semantics. In R. Kempson (ed.) Mental Representations (Cambridge: Cambridge University Press), pp. 155–81. Carston, R. (2002) Thoughts and Utterances: The Pragmatics of Explicit Communication (Oxford: Blackwell). Cosenza, G. (ed.) (2001) Paul Grice’s Heritage (Turnhout: Brepols). Frege, G. (1918/19) Der Gedanke. Eine logische Untersuchung. Beiträge zur Philosophie des deutschen Idealismus 2: 58–77. Reprinted in English translation under the title ‘Thoughts’ in Frege, G. (1977) Logical Investigations (Oxford: Blackwell), pp. 1–30. Gibbs, R. W. (1999) Interpreting what speakers say and implicate. Brain and Language, 68: 466–85. Grandy, R. (1989) On Grice on language. Journal of Philosophy, 10: 514–25. Grandy, R. and Warner, R. (1986a) Paul Grice: a view of his work. In R. Grandy and R.Warner (eds.) (1986b), pp. 1–44. Grandy, R. and Warner, R. (eds.) (1986b) Philosophical Grounds of Rationality: Intentions, Categories, Ends (Oxford: Oxford University Press). Grice, H. P. (1961) The causal theory of perception. Proceedings of the Aristotelian Society 35 (suppl.): 121–52. Grice, H. P. (1986) Reply to Richards. In R. Grandy and R. Warner (eds.) (1986b), pp. 45–106. Grice, H. P. (1989) Studies in the Way of Words (Cambridge, MA: Harvard University Press). Grice, H. P. (1991) The Conception of Value (Oxford: Oxford University Press). Grice, H. P. (2001) Aspects of Reason (Oxford: Oxford University Press). Gu, Y. (1994) Pragmatics and rhetoric: a collaborative approach to conversation. In H. Parret (ed.) (1994) Pretending to Communicate (Berlin: de Gruyter), pp. 173–95. Harth, M. (2003) Concepts of what is said. In G. Meggle (ed.) (2003) Saying, Meaning, Implicating (Leipzig: Leipziger Universitätsverlag), pp. 72–83. Hirschberg, J. (1985) A Theory of Scalar Implicature. PhD thesis, University of Pennsylvania. Published 1991 in the series Outstanding Dissertations in Linguistics (New York: Garland).
Introduction 29 Horn, L. (2007) Toward a Fregean pragmatics: Voraussetzung, Nebengedanke, Andeutung. In I. Kecskes et al. (eds.) (2007) Explorations in Pragmatics (Berlin: Mouton de Gruyter), pp. 39–69. Kasher, A. (1976) Conversational maxims and rationality. In A. Kasher (ed.) (1976) Language in Focus: Foundations, Methods and Systems (Dordrecht: D Reidel Publishing), pp. 197–216. Kemmerling, A. (1979) Was Grice mit “meinen” meint. In G. Grewendorf (ed.) Sprechakttheorie und Semantik (Frankfurt a. M.: Suhrkamp), pp. 67–118. Kemmerling, A. (1986) Utterer’s meaning revisited. In R. Grandy and R. Warner (eds.) (1986b), pp. 131–55. Kemmerling, A. (1991) Implikatur. In A. Stechow and D. Wunderlich (eds.) Semantik/ Semantics (Berlin: de Gruyter) 319–33. Kemmerling, A. (2001) Gricy actions. In G. Cosenza (ed.) (2001), pp. 73–99. Levinson, S. C. (1983) Pragmatics (Cambridge: Cambridge University Press). Levinson, S. C. (2000) Presumptive Meanings: The Theory of Generalized Conversational Implicature (Cambridge, MA: MIT Press). Lewis, D. (1969) Convention: A Philosophical Study (Cambridge, MA: Harvard University Press). Loar, B. (1981) Meaning and Mind (Cambridge: Cambridge University Press). Lüthi, D. (2006) How implicatures make Grice an unordinary ordinary language philosopher. Pragmatics, 16: 247–74. Meggle, G. (1981) Grundbegriffe der Kommunikation (Berlin, New York: de Gruyter). Neale, S. (1992) Paul Grice and the philosophy of language. Linguistics and Philosophy, 15: 509–59. Neale, S. (2001) Implicature and colouring. In G. Cosenza (ed.) (2001), pp. 139–84. Petrus, K. (1996) Kommunikation, Rationalität, Adäquatheit. Logos, 3: 293–303. Petrus, K. (1999) Kommunikation als Kooperation. In M. Flügel, T. Gfeller and C. Walser (eds.) Werte & Fakten (Bern, Stuttgart: Paul Haupt Verlag), pp. 213–34. Petrus, K. (2006) Illokutionäre Akte: Alte Begriffe, ein neuer Weg. Linguistische Berichte, 206: 117–46. Pfister, J. (2007) The Metaphysics and the Epistemology of Meaning (Frankfurt: Ontos Verlag). Plüss, H. (2001) Bedeutungstheorie und philosophische Psychologie bei Paul Grice (Bern: Paul Haupt Verlag). Preyer, G. and Peter, G. (eds.) (2007) Context-Sensitivity and Semantic Minimalism: New Essays on Semantics and Pragmatics (Oxford: Oxford University Press). Quine, W. V. O. (1953) Two dogmas of empiricism. In W.V.O. Quine (1953) From a Logical Point of View (Cambridge, MA: Harvard University Press), pp. 20–46. Recanati, F. (1986) On defining communicative intentions. Mind and Language, 3: 213–42. Recanati, F. (1995) The alleged priority of literal interpretation. Cognitive Science, 19: 207–32. Recanati, F. (2004) Literal Meaning (Cambridge: Cambridge University Press). Rolf. E. (1994) Sagen und Meinen. Paul Grices Theorie der Konversations-Implikaturen (Opladen: Westdeutscher Verlag). Russell, B. (1906) On denoting. Mind, 14: 479–99. Saul, J. (2002) Speaker’s meaning, what is said, and what is implicated. Noûs, 36: 228–48. Sbisa, M. (2001) Intentions from the other side. In G. Cosenza (ed.) (2001), pp. 185–206. Schiffer, S. (1972) Meaning (Oxford: Oxford University Press). Searle, J. R. (1969) Speech Acts (Cambridge: Cambridge University Press).
30 Klaus Petrus Soames, S. (2003) Philosophical Analysis in the Twentieth Century: Volume 2: The Age of Meaning (Princeton: Princeton University Press). Sperber, D. and Wilson, D. (1995) Relevance. 2nd edn. (Oxford: Blackwell) [1st edn, 1986]. Stanley, J. (2000) Context and logical form. Linguistics and Philosophy, 23: 391–434. Strawson, P. F. (1950) On referring. Mind, 59: 320–44. Strawson, P. F. (1952) Introduction to Logical Theory (London: Methuen). Strawson, P. F. (1964) Intention and convention in speech acts. Philosophical Review, 73: 439–60. Szabo, Z. G. (ed.) (2005) Semantics versus Pragmatics (Oxford: Oxford University Press). Truniger, J. (2006) Nachdenken über Rationalität als philosophisches Fundamentalprojekt. Unplublished MA thesis, University of Zurich. Warner, R. (2001) Introduction: Grice on reasons and rationality. In H. P. Grice (2001), pp. vii–xxxviii.
2 Paul Grice and the Philosophy of Ordinary Language Siobhan Chapman
The story of Paul Grice’s relationship to what has become generally known as ‘ordinary language philosophy’ (OLP) is long and involved. It is one of initial enthusiasm, growing uneasiness and distance, but continued loyalty. Along with John L. Austin and Peter F. Strawson, Grice is one of the philosophers now most readily associated with the Oxford-based movement. Indeed his theory of conversation is widely regarded, among linguists at least, as one of the most successful products of OLP. In his development of this theory Grice certainly drew on a recognizably ‘ordinary language’ approach. And he continued to return to this in his later work, some of it conducted decades after the decline of OLP and thousands of miles away from Oxford. But there is another side to the story. Grice himself had reservations about OLP and increasingly voiced concerns about its limitations. Much of his own philosophical work, including the theory of conversation, relies on assumptions, strategies and methodologies that contrast sharply with those of OLP. This chapter will consider the nature of OLP, its origins and development. It will investigate the tensions in Grice’s relationship to OLP by considering the evidence for the growth of its influence on his work and subsequently his departure from it. It will trace his academic interaction with J. L Austin, and the nature of some of the other influences and inspirations in his work. It will also suggest ways in which, despite his own reservations, Grice’s bestknown work might be celebrated as a valuable contribution to the business of OLP itself. He offered some striking and new insights into the processes of meaning in communication: insights that are crucial to the success of a philosophical method based on the close analysis of natural language. OLP had its origins in deeply-felt enthusiasms and strong antipathies. The enthusiasms were for what was perceived to be a new way of working, even a way of finally solving many of the age-old problems of philosophy. The antipathies were towards many of the assumptions and approaches with which philosophers had worked in the past. There was suspicion of anything that relied on specialist philosophical terminology. Speculative and metaphysical philosophy was mistrusted, but so too were some earlier 31
32 Siobhan Chapman
incarnations of analytic philosophy, the tradition to which OLP itself can be seen as broadly belonging. In particular, there was a distaste for the approach to philosophy exemplified by the Vienna Circle, the group of scientifically-minded philosophers, led by Moritz Schlick and including the logician Rudolf Carnap, who developed the doctrines of logical positivism. During the 1920s and 1930s, the Vienna Circle in effect proposed to relaunch empiricism as a viable philosophical doctrine. Empiricism’s earlier heyday had been during the Enlightenment, when it was championed by philosophers such as John Locke (1993 [1690]) as a touchstone for truly justified beliefs and an antidote against the need to posit innate ideas in order to explain human knowledge. However, Kant’s arguments that experience alone cannot be sufficient to account for the full range of human knowledge seemed to present problems for empiricism. Kant (1929 [1781]) pointed out that the truth of certain types of propositions is neither necessary nor known from experience. Mathematical propositions are a highly pertinent case. Kant demonstrated that ‘5 + 7 = 12’ is not necessarily true (our understanding of ‘5’ and of ‘7’ alone cannot account for our belief in its truth) nor confirmed by any observable facts in the world that are independent of the very system of mathematics to which this statement belongs. The Vienna Circle were confident that recent developments in logic were sufficient to rescue empiricism from Kant’s apparent counterexamples. Russell and Whitehead (1910) had argued that a rigorously symbolic logical notation was sufficient to express mathematical propositions; mathematics was in effect reduced to logic. Wittgenstein (1961 [1921]) built on this demonstration, arguing that since logic is a closed system its propositions are necessary truths, or tautologies. If mathematics can be reduced to logic then the statements of mathematics too are necessary truths. The logical positivists therefore justified the inclusion of mathematical propositions in the category of analytic statements that are exempt from empirical testing. We know the truth of ‘bachelors are unmarried’ because we know the language in which it forms a valid sentence and in just the same way we know the truth of ‘5 + 7 = 12’ because we know the system of mathematics to which it belongs. This left the logical positivists free to stipulate that all other sentences, that is all synthetic sentences, must either be subject to an identifiable process of verification, or else be dismissed as meaningless. They therefore felt justified in rejecting from serious philosophical discussion the traditional topics of the more metaphysical branches of philosophy, including statements of ethics, aesthetics and religion. As Schlick himself put it, ‘The empiricist does not say to the metaphysician “what you say is false,” but, “what you say asserts nothing at all!” He does not contradict him, but says “I don’t understand you.” ’ (Schlick 1932: 107) In the decades that followed the heyday of the Vienna Circle, that is during the 1930s, 40s and 50s, a variety of reactions against logical positivism were propounded by philosophers of science and of language. Although
Paul Grice and the Philosophy of Ordinary Language
33
they differed from each other in emphasis and philosophical outlook, it is possible to identify a shared core to their various objections. They all found something fundamentally unempirical in the work of the Vienna Circle, and claimed that this was a fatal flaw in a doctrine that claimed such strict empirical credentials. Most famously, Karl Popper argued that the statements that logical positivism identified as meaningful and true could never actually be demonstrated to be more than probable. The inductive accumulation of data in support of a statement could never lead to certain empirical knowledge, because it was always possible that the very next piece of relevant evidence might prove to be a counterexample. In the place of verifiablity as the criterion for a meaningful statement, Popper proposed falsifiability as the criterion for a properly scientific statement. So for Popper true knowledge was indeed empirical knowledge, but it did not gain this status because it had been verified by experience, or even confirmed by a body of data. Rather, ‘every scientific statement must remain tentative for ever’ (Popper 2002 [1935]: 280, original emphasis). This proposal has been hugely influential in the philosophy of science, and also in linguistics, especially for those linguists who believe that the study of language should properly be conducted as a natural science. Other responses to logical positivism came from philosophers who were more explicitly concerned with language and who argued that the philosophy of the Vienna Circle was based on some assumptions about language that had no empirical credentials. Perhaps the best known of these responses is that from W. V. O. Quine. Quine had been invited to attend some meetings of the Vienna Circle as a young man and acknowledged a life-long intellectual debt to Carnap, but he became increasingly uneasy with what he saw as logical positivism’s unfounded and unsupported assumptions about language. The main focus for his anxiety was the purported distinction between analytic and synthetic sentences. This anxiety found fullest expression in Quine’s writings of the 1950s, but it is hinted at in a paper he published in 1943. In this, Quine notes that the concept of analyticity depends on the concept of synonymity, since an expression can be said to be analytic only if its subject and predicate contain expressions that are synonymous. Synonymity in turn depends on a notion of meaning. ‘Just what the meaning of an expression is – what kind of object – is not yet clear; but it is clear that, given a notion of meaning, we can explain the notion of synonymity easily as the relation between expressions that have the same meaning’ (Quine 1943: 120, original emphasis). By the early 1950s, Quine was not so much worried by the current absence of a clear account of meaning, as convinced that no such account could be found; meaning was a philosopher’s construct with no empirical justification. It was clear that words exist, and that objects in the world exist, but ‘meanings themselves, as obscure intermediary entities, may well be abandoned’ (Quine 1951: 22). If Quine was right in this, the effects for the whole programme of the
34
Siobhan Chapman
logical positivists would be catastrophic. Their criterion of meaningfulness depended on a viable distinction between analytic and synthetic sentences, and this distinction in turn depended on the existence of determinate, identifiable meanings, independent of context or individual speaker. Quine’s rejection of meaning as an unempirical construction was accompanied by a conviction that language should be studied on an experiential basis. His writings include frequent references to the business of the ‘field linguist’, often described as learning about new and exotic languages by observing the behaviour of their speakers. Quine himself seems never to have engaged in such activities, however, or to have paid attention to the reports of those who had. One near contemporary of Quine who did attempt active empirical study of language was the Norwegian philosopher Arne Naess. Naess’s work was received with some suspicion by the philosophical establishment of the time, and is now little read, but it provides an interesting part of the story of the empiricist reaction against logical positivism. Like Quine, Naess attended some meetings of the Vienna Circle in the early 1930s and was initially impressed by what he heard there, especially by the work of Carnap. Also like Quine, he became worried that logical positivism was based on some assumptions about language that were not subjected to the same empirical rigour as the rest of their philosophy. Looking back on his philosophical development some sixty years later, he described how he had come to realize that ‘Carnap seemed to accept a presupposition that he had philosophically adequate knowledge of what ordinary words and sentences of his mother-tongue mean. It includes the words “true”, “probable”, “certain” ... How could the logical empiricists boast about a scientific attitude when they relied so much on intuition when speaking about the use of words?’ (Naess 1993: 17–18). Naess became convinced that philosophical use of such terms was legitimate only if based on empirical investigations into how they are actually used and understood by ordinary people. He developed various techniques for this investigation, most prominently a form of interview by questionnaire that in some ways prefigured methods developed in sociolinguistics thirty or more years later. His first monograph was a report on an empirical study of the term ‘truth’, in which he concluded that there are many more, and many more sophisticated, ordinary understandings of ‘truth’ than philosophers usually acknowledge. Naess was ambitious for his new and idiosyncratic style of philosophy, arguing that ‘There are scarcely any of the traditional philosophic problems which are not suitable for this procedure’ (Naess 1938: 162). For a more detailed account of Naess’s philosophy and method, see Chapman (2008). In charting the major mid-century reactions to logical positivism it is perhaps overly simplistic to equate the rise of OLP with the work of any single individual. Traces of the general approach it embodies can be found in the work of philosophers such as G. E. Moore, Ludwig Wittgenstein
Paul Grice and the Philosophy of Ordinary Language
35
and Gilbert Ryle. But it is perhaps most closely associated with the work of J. L. Austin, and Austin will be taken here as representative of its main doctrines. It was Austin who worked assiduously to establish a school of philosophy dedicated to the study of ordinary language among his Oxford colleagues in the years immediately after the Second World War. As a result, it was Austin’s version of OLP with which Grice came into most regular and most prolonged contact. In the later 1940s Austin established a series of weekend meetings to discuss philosophical topics with a hand-picked group of junior colleagues, among whom Grice was a prominent and enthusiastic member. Austin described these meetings as his ‘Saturday mornings’. Austin’s philosophy rested on some similar assumptions to Naess’s, particularly that philosophers would be advised to pay attention to how words and phrases are ordinarily used. Where he differed substantially was in how he proposed that this should be done: in other words, in what he saw as the appropriate method for the empirical study of language. Austin did not have direct first-hand experience of the Vienna Circle, but like many British philosophers of his generation he was introduced to the ideas of logical positivism when he read A. J. Ayer’s Language, Truth and Logic when it was first published in the mid 1930s (Ayer 1971 [1936]). There was much in this book that he admired but his lingering doubts, articulated and developed during many personal conversations with Ayer, eventually formed into a philosophical approach that was in many ways diametrically opposed to that of the Vienna Circle. Austin was troubled by what he saw as an inappropriately dismissive attitude to ordinary language. The Vienna Circle’s emphasis on logic, and on the claim that many apparent statements are in fact meaningless pseudostatements, made them wary of the potentially misleading nature of natural language. Carnap made this position explicit when he cautioned that ‘the fact that natural languages allow the formation of meaningless sequences of words without violating the rules of grammar, indicates that grammat-ical syntax is, from a logical point of view, inadequate’ (Carnap 1932: 68). For Austin, ordinary language was worthy of serious philosophical attention in its own right. It could offer a valuable tool to the philosopher because, unlike ad hoc philosophical coinages, it had developed over generations to meet the needs of people describing the world around them. In his paper ‘A plea for excuses’ Austin famously extolled the philosophical expediency of consulting ordinary language: ‘Our common stock of words embodies all the distinctions men have found worth drawing, and the connections they have found worth making, in the lifetimes of many generations: these surely are likely to be more numerous, more sound, since they have stood up to the long test of the survival of the fittest, and more subtle, at least in all ordinary and reasonable practical matters, than any that you or I are likely to think up in our arm-chairs of an afternoon – the most favoured alternative method’ (Austin 1956: 30).
36 Siobhan Chapman
Austin dedicated himself to the meticulous examination of this common stock of words. Given that this process was so central to his philosophical outlook, it is surprising that he left no explicit commentary of any length on philosophical method. He did urge his philosophical audiences on a number of occasions to ‘use the dictionary – quite a concise one will do, but the use must be thorough’ (Austin 1956: 134, original emphasis). This is evidently a process that Austin used many times himself: going through a dictionary to build up lists of related words and to contemplate nuances of use. By this means Austin hoped to elucidate particular philosophical problems. It was a method that he liked to describe as ‘linguistic botany’. His close colleague J. O. Urmson offered a fuller account of this methodology, drawing on his own recollections of working with Austin and on some unpublished notes discovered after Austin’s death. The researcher would draw up lists of examples of the usage of a philosophically salient range of vocabulary, based on the results of searching the dictionary, on personal intuition, and perhaps on examples of usage taken from published sources. Only after careful analysis of the examples collected was it legitimate to consult what philosophers had previously said about a topic, or to advance any new discussion. By that stage ‘it is an empirical question whether the accounts given are correct and adequate, for they can be checked against the data collected’ (Urmson, Quine and Hampshire 1965: 80). Much of the widespread and at times vitriolic opposition to Austin’s work focused on the validity of his chosen philosophical method. Either explicitly or implicitly, many challenged the claim that the authenticity of the accounts of usage on which OLP depended was ‘empirical’. After all, Austin and his colleagues engaged in no systematic sampling of language use, no recording of naturally occurring conversations, and no questioning of subjects outside of their own limited and homogeneous community. Writing a decade after Austin’s death but summing up the reactions of many of his contemporaries, C. W. K. Mundle complained that Austin’s accounts of language ‘describe how English is spoken by the few people who are as fastidious as himself and as sensitive to the nuances and etymology of its words’ (Mundle 1979 [1970]: 82). Austin’s defence against criticisms of this type hinged on what he saw as the ideally collaborative nature of philosophical inquiry. Whenever possible, linguistic botanizing should be undertaken as a group enterprise by philosophers working in teams, such as that provided at his Saturday mornings. This established that no example of usage was accepted without discussion and consensus, ensuring the objectivity of the final data. Grice’s attitude to logical positivism seems to have followed a similar trajectory to Austin’s. In the late 1930s he shared the general enthusiasm among young Oxford philosophers for Language, Truth and Logic, and in particular for Ayer’s suggestion that rigorous linguistic analysis could dissolve many philosophical problems of long standing. Nevertheless, as he
Paul Grice and the Philosophy of Ordinary Language
37
recalled years later, he always harboured worries about the sweeping and dogmatic claims Ayer was making (Grice 1986: 48). A survey of Grice’s early work suggests that these worries became increasingly entrenched, and demonstrates a corresponding rise in the influence on his thinking of Austin’s alternative method of analysis. What is apparently the earliest surviving manuscript of Grice’s writing, a paper on ‘Negation and privation’ from the late 1930s, shows distinctly verificationist leanings. In this, Grice addresses the problem posed by our knowledge of negative facts: that is, by how we can be confident in making negative statements such as ‘This is not red’ and ‘I am not hearing a noise’ (Grice c.1938). He argues that in such cases our knowledge is derived from inference: from our positive knowledge that the object in question is green, for instance, or from our entertainment of the proposition ‘I am hearing a noise’, together with an absence of knowledge that this proposition is true. The verificationist tendencies in this paper are apparent in Grice’s proposal to analyse natural language expressions that contain a potentially problematic term (‘not’) using logically more acceptable expressions that do not contain this term. This second set of expressions can be subject to an identifiable process of verification involving sense data (in the case of ‘This is green’) or introspection (in the case of the absence of knowledge that ‘I am hearing a noise’ is true). By the time he published his first article, ‘Personal identity’, in 1941, Grice was displaying much more interest in the evidence of ordinary language. He no longer tries to ‘translate’ natural language expressions into empirically verifiable propositions. Rather, he consults issues of how certain words are ordinarily used as an aid towards philosophical analysis, in a manner that is certainly reminiscent of Austin’s emerging methodology. In effect, Grice proposes to revive a version of the ‘memory theory’ of personal identity, an approach that dates back at least to John Locke. For Grice, sameness of identity is assured by continuity of memory. This continuity need not be unbroken; that is, it is not necessary for us to remember all events from our early life in order to claim to be the same person in later life. Rather, a continuous series of experiences and of memories of previous experiences is built up throughout a person’s lifetime. Grice’s account of identity is technical and rather formal in nature, but significantly it is premised upon a reflection on how the personal pronoun ‘I’ is used in everyday language. Grice identifies three different types of usage: those in which ‘I’ can be glossed as ‘my body’ (‘I fell down the cellar steps’), those in which ‘my body’ does not capture the full meaning of ‘I’ (‘I played cricket yesterday’), and those in which ‘my body’ seems an even less appropriate substitute (‘I am thinking about the immortality of the soul’). His point, of course, is that consulting ordinary language shows that personal identity is not always understood to rest just on bodily identity, but that in some cases it is necessary to take account of ‘something about the sort of thoughts and intentions and decisions I had’ (Grice: 1941: 74).
38
Siobhan Chapman
Seven years after the publication of this article the influence of OLP was even more clear in Grice’s philosophy, but so too were the early signs of how and why he would become disillusioned with some of the assumptions and working methods on which it was founded. 1948 is a significant year in the story of Grice’s intellectual development because it is the year in which he wrote and presented his seminal paper ‘Meaning’. This was eventually published in The Philosophical Review in 1957, but by then it had languished unchanged in a desk drawer for almost a decade, a victim of Grice’s notorious perfectionism and resulting reluctance ever to offer his work up for publication. ‘Meaning’ begins with a rather abrupt appeal to ‘Consider the following sentences’ (Grice 1957: 213), and the presentation of a series of linguistic examples. Grice goes on to distinguish between two different uses of the term ‘means’: to label these as ‘natural’ and ‘nonnatural’ meaning, and to develop an account of linguistic meaning as a type of nonnatural meaning, dependent on a complex set of intentions on the part of the speaker. In effect, he employs the analysis and comparison of uses of a particular expression, the methodology characteristic of OLP, in approaching that most central of issues in the philosophy of language, the nature of meaning. ‘Meaning’ may be in many ways a canonical example of OLP, but it also contains the germs of the two main ways in which Grice’s work was diverging, and was to continue diverging, from Austin’s. In brief, he was firstly developing an enthusiasm for general, explanatory, rather formal theories, and he was secondly becoming increasingly aware of a need to be rigorous and explicit in the distinction between sentence-meaning and speakermeaning. ‘Meaning’ may have drawn on ordinary language examples, but it was at heart an ambitious attempt to offer a theory of linguistic meaning, based on a formalized account of the psychological concept of speaker intention. Austin, on the other hand, excelled at the gathering, analysis and categorization of linguistic examples. Grice began to detect a shallowness, or an indication of incompleteness in this process. That is, Austin was strangely reluctant to go beyond this process to make any more general claims; he argued that linguistic analysis was the essential starting point to any philosophical engagement, but when it came to it he seemed to shy away from such engagement. What Austin was achieving was a series of meticulous accounts of particular areas of the vocabulary of English, with no principled way of determining what was philosophically interesting. In later life, Grice recalled that he raised these doubts with Austin, but met with characteristic intransigence (Grice 1983). In his presentation of ‘Meaning’, and his subsequent work on his intentional account, Grice voiced the necessity of making a clearer distinction than was currently being made between sentence-meaning and speakermeaning. He discussed this distinction in ‘Meaning’ in terms of ‘the standard meaning, or the meaning in general’, and ‘what a particular speaker or
Paul Grice and the Philosophy of Ordinary Language
39
writer means’, which ‘may well diverge from the standard meaning’ (Grice 1957: 216–17, original emphasis). The relationship between standard meaning and speaker-meaning was an uneasy issue in OLP. Austin never adopted the motto ‘meaning is use’ that is sometimes associated with the later work of Wittgenstein, but his insistence on investigating the meaning of a word by analysing the different circumstances in which it would or would not be applicable certainly pointed towards this position. In his development of speech act theory, his distinction between the locutionary act, determined by the conventions of the language, and the illocutionary act, determined by further conventions of usage, would seem to commit him to a distinction between meaning per se and meaning in use. But Austin himself was apparently uneasy about this commitment, and reluctant to acknowledge the value of the distinction to his philosophy. In his lectures on speech acts he claimed that ‘The total speech act in the total speech situation is the only actual phenomenon which, in the last resort, we are engaged in elucidating’ (Austin 1962: 147, original emphasis). Grice’s relationship with OLP became increasingly complex during the years following his presentation of ‘Meaning’. He was a regular and apparently enthusiastic member of the Saturday morning meetings; indeed he took over leadership of the group after Austin’s early death in 1960. He was one of the few Oxford philosophers explicitly to talk about and to defend OLP, for instance in a lecture that he delivered at Wellesley College, Massachusetts (Grice 1958). However, he was also becoming interested in and impressed by a very different style of philosophy of language. In his philosophical memoir he recalled that in the 1950s, and increasingly during the 1960s, his outlook was influenced by American philosophers such as W. V. O. Quine and Noam Chomsky. This is of course a rather disparate pairing; Quine and Chomsky differed in almost every aspect of their thinking, and were embroiled in philosophical disputes that lasted decades. But what they had in common, and what interested Grice about them, was their commitment to establishing general, explanatory accounts of language. It was this very feature that made them unlikely targets of admiration for a philosopher of ordinary language. Certainly, Syntactic Structures (Chomsky 1957) had been read at Austin’s Saturday mornings when it was first published, and Austin had declared himself impressed. But what fascinated Grice was not so much Chomsky’s bold attempt to take on the intricacies of grammar, which seemed to him to be the main source of Austin’s interest, but the fact that Chomsky’s approach did not offer ‘piecemeal reflections on language’; on the contrary, ‘one got a picture of the whole thing’ (Grice 1983). This interest in general explanatory theories, together with his preoccupation with the distinction between sentence-meaning and speaker-meaning, continued to impact on Grice’s work as he developed the ideas that he was to deliver in his William James lectures on ‘Logic and conversation’ in 1967.
40 Siobhan Chapman
In these lectures Grice introduces his central insight that the significance of any particular utterance can be analysed into the separate components of ‘what is said’ and ‘what is implicated’, and further that meaning determined by linguistic rule and meaning intended by individual speaker are mediated by a principled set of generally accepted constraints on behaviour that could be classified in terms of an overarching expectation of cooperation. His definition of a distinct level of meaning to be described as ‘what is said’ is tentative. ‘In the sense in which I am using the word say, I intend what someone has said to be closely related to the conventional meaning of the words (the sentence) he has uttered. Suppose someone to have uttered the sentence He is in the grip of a vice. Given a knowledge of the English language, but no knowledge of the circumstances of the utterance, one would know something about what the speaker had said, on the assumption that he was speaking standard English, and speaking literally’ (Grice 1975: 25). In typically hedged and almost apologetic fashion, Grice here introduces conventional meaning into his philosophy of language: the notion that the rules of the language themselves specify something that might be recognized as ‘literal meaning’, independent of considerations of context and of individual speaker. Grice acknowledged ‘what is said’ reluctantly and with reservations. He admitted towards the end of his life that the exact relationship between linguistic meaning and speaker-meaning was the philosophical question that had given him the most trouble. In effect, he always hoped that speaker intention might prove to be the deciding factor in all aspects of meaning, but found a separate ‘what is said’ philosophically expedient. The reasons for this can be found in Grice’s mixed motivations in developing his theory of conversation. He was interested not just, not even primarily, with accounting for how people communicate in conversational situations. He was at least as much concerned with offering a new solution to some age-old philosophical problems, and in particular with showing how formal logic could provide an explanatory base for many aspects of linguistic meaning. This latter motive was strikingly at odds with some of the central tenets of OLP, with its insistence on attending to the particularities of usage rather than trying to reduce language to logic. In this, OLP corresponds to the ‘informalist’ position that Grice sets out near the start of his lectures. He paraphrases the informalist as claiming that: ‘The philosophical demand for an ideal language rests on certain assumptions that should not be conceded; these are, that the primary yardstick by which to judge the adequacy of a language is its ability to serve the needs of science, that an expression cannot be guaranteed as fully intelligible unless an explication or analysis of its meaning has been provided, and that every explication or analysis must take the form of a precise definition that is the expression or assertion of a logical equivalence’ (Grice 1975: 23). Grice hoped to avoid the informalists’ reluctance to give precise logical definitions of natural language expressions, together with the corresponding
Paul Grice and the Philosophy of Ordinary Language
41
formalists’ position that natural language was simply messy and imperfect compared to logic. He would achieve this if he could show that equivalence to logic is indeed demonstrable, but only at the level of ‘what is said’. Further principles concerned with the use of language were then necessary to explain the difference between this and what is often understood when linguistic expressions occur in context. So for instance the natural language expression ‘and’ might really be equivalent to logical conjunction at the level of what is said, but takes on implications of temporal sequence and causality in an example such as ‘She fell over and grazed her knee’. These same mechanisms that Grice developed in the service of a logical account of linguistic meaning also proved fruitful in explaining a host of other processes that have caught the imaginations of philosophers, and particularly of linguists. These include metaphor, irony and insinuation. Grice’s division of what Austin called ‘the total speech situation’ between ‘what is said’ and ‘what is implicated’ depended more heavily and more explicitly on the distinction between sentence-meaning and speaker-meaning than had his intentional account in ‘Meaning’. This has not prevented him from being most often associated with the notion that meaning is dependent on context of use, or of being on what P. F. Strawson once described as the ‘communication-intention’, as opposed to the ‘formal semantics’ side of the debate about meaning (Strawson 1969: 171–2). Some more recent commentaries have emphasized the centrality of formal meaning, independent of use, to Grice’s programme. Kent Bach has commented of the distinction between what sentences mean and what speakers mean in using them that, ‘Whereas Wittgenstein adopted a decidedly anti-theoretical stance toward the whole subject, Austin developed a systematic, though largely taxonomic, theory of language use. And Paul Grice developed a conception of meaning which, though tied to use, enforced a distinction between what linguistic expressions mean and what speakers mean in using them’ (Bach 2006: 147). Brian Loar has gone further, listing Grice alongside Frege and Russell as a philosopher endorsing ‘truth theories of language’. Refreshingly, he places the logical positivists of the Vienna Circle in the opposing camp supporting ‘use theories of meaning’. Logical positivists with their interest in the logical analysis of formal languages are usually seen as divorced from concerns of use. But according to Loar’s definition, use theorists ‘hold that the meaning of a term is determined by how it is used, for example, how to verify an instance, or how to use it in inferences’ (Loar 2006: 85). On this criterion, the logical positivists fit the rôle of use theorists. The trend towards logical and formal explanations set during the development of the theory of conversation continued. When Grice took up a chair at the University of California, Berkeley in 1967, he was drawn in part by the more formalist styles of philosophy to be found in America. These contrasted favourably for him with the subjective and anecdotal style that Austin had made predominant in Oxford. Grice’s subsequent work touched
42
Siobhan Chapman
on many different philosophical topics, including rationality, philosophical psychology, ethics and even metaphysics. Much of this work shows the formalist influence. For instance, ‘Vacuous names’, published just two years after his move to Berkeley, is a response to Quine on the topic of reference that is replete with logical terminology (Grice 1969). However, his personal notes reveal that he remembered Austin’s methodology, and used it, throughout his life. Time and again he began work on a philosophical topic by jotting down examples of the ways in which a particularly salient word or group of words might ordinarily be used. His preliminary notes often include lists of example sentences and comments on appropriateness or otherwise of use. Grice’s continued use of linguistic botanizing found its way into the public presentations of his work, albeit in modified and tidied form. For instance, in his philosophy of reason, originally presented as a series of lectures in the late 1970s, he exhorts his audience to consider the philosophically salient distinction between ‘explanatory reasons’, ‘justificatory reasons’ and ‘justificatory-explanatory reasons’. He introduces the distinction by means of a barrage of examples that illustrate the three different ways in which, he claims, the word ‘reason’ is ordinarily used. For instance, ‘The reason why the bridge collapsed was that the girders were made of cellophane’ can be contrasted with ‘The fact that they were a day late was a reason for thinking that the bridge had collapsed’ and with ‘John’s reason for thinking Samantha to be a witch was that he had suddenly turned into a frog’ (Grice 2001: 37–40). Despite superficial similarities, however, what Grice was doing in work such as this was rather different from Austin’s ideal for philosophy. Firstly, Grice’s linguistic botanizing and his lists of examples seem to have been drawn entirely from personal introspection and the consultation of his own usage. There is no evidence that he drew systematically on the insights of others, or that he attempted to reproduce anything like Austin’s Saturday mornings at his new base in Berkeley. He did not try, and was perhaps not much interested in, living up to Austin’s aspirations to empiricism and objectivity in his references to ordinary language. Secondly and perhaps more significantly, Grice’s later linguistic botanizing was always the springboard to general, ambitious and often highly formal theories. Here again the work on reasons provides a telling example. The lectures that began with the presentation of some whimsical linguistic examples to support a distinction between different types of ‘reason’ continue with a detailed examination of the Kantian view that there is a single human faculty of reason, manifest in cases of both practical and non-practical reasoning. Grice’s support for this view is expressed in terms of a formal analysis of reasoning statements in which the term ‘Acc’ is introduced to represent the rationality operator, and the symbols ‘˫’ and ‘!’ represent the two different mood operators. In contrast, Austin’s musings on the ordinary use of expressions
Paul Grice and the Philosophy of Ordinary Language
43
were more often than not the endpoint as well as the starting point to his philosophical studies. For more on Grice’s continued championing and use of the methods of OLP see Chapman (2005). By the end of his life, Grice’s relationship to OLP was complex almost to the point of paradox. He was known by many, and celebrated by some, as a leading exponent of OLP during his Oxford days, and as an advocate of it during his career at Berkeley, long after it had ceased to be fashionable. He continued to celebrate the advantages of consulting intuitions about usage, both in his published works and public presentations, and also in his private notes and advice to graduate students. Yet the work for which he was already best known was in many ways at odds with, even a negation of, some basic assumptions of OLP. And much of his more recent work displayed a formality or a speculative nature that would have appalled Austin. A couple of years after Grice’s death, his former colleague P. F. Strawson commented in a review of Studies in the Way of Words on his uneasy relationship to OLP, particularly in relation to his development of the theory of conversation. Strawson remarked that the lectures on logic and conversation necessarily implied criticism of OLP. OLP seemed to require that an account of the meaning of an expression must draw on details of the conditions of its use: ‘Grice’s own view, on the contrary, is that, at least in many such cases, meaning on the one hand, and additional conditions of appropriate use on the other, are quite different things; and that the existence of the conditions in question is to be explained by the interaction of meaning, strictly and literally understood, and certain quite general principles of conversational interchange’ (Strawson 1990: 155–6). In expressing the point of Grice’s departure from OLP, Strawson also offers a succinct account of his major contribution to the study of language. Strawson sees this central aspect of Grice’s project in ‘Logic and conversation’ as sitting uneasily with his apparent defence of OLP elsewhere. But he acknowledges that Grice’s achievement should ultimately be seen as ‘not a repudiation, but a refinement of the “ordinary language” philosophers’ approach’ (Strawson 1990: 157). It offers OLP a better understanding of, and a better chance of working effectively with, its material. Strawson’s summary of Grice’s relationship to OLP points the way to an unusual evaluation of his work. Grice’s distinction between ‘what is said’ and ‘what is implicated’ has been celebrated in linguistics as setting the agenda for much of the discussion during the past forty years over the borderline between semantic and pragmatic meaning. It has also been taken seriously in more formal philosophical discussions of meaning based on logic, because of its attempt to retain the logical connectors as the basis for various aspects of natural language semantics, and to explain apparent deviations from them in terms of separate principles of language use. But Strawson hints at a positive evaluation of Grice’s insight from within the framework of OLP itself or, since OLP is now little practised in any
44
Siobhan Chapman
recognizably Austinian form, of any analytical approach that relies on close attention to how language is used. Before Grice’s sharp, if problematic, division between ‘what is said’ and ‘what is implicated’, philosophers dealing with language use were in danger of being misled by particularities of the circumstances in which expressions were most commonly used. They had no principled way of distinguishing between the semantic features of particular words and expressions, and the general assumptions that accompany any use of language; they might accommodate into their account of a particular expression an aspect of its apparent meaning that was actually a general feature of language use. Grice’s 1961 paper ‘The causal theory of perception’, written while he was developing the ideas that were to appear in his William James lectures, offers one example of this potential philosophical benefit. Grice turns his interest in the way in which language is used towards a consideration of the philosophical problem of perception. He tentatively defends a causal account, in which people receive impressions through their sense organs that they infer are caused by physical objects. Such an account has traditionally been supported with reference to the everyday use of expressions such as ‘so-and-so looks ϕ [e.g. blue] to me’. There must be something, such as the philosophers’ ‘sense data’, that corresponds to what the subject is experiencing when such an expression is used. A potential objector might point out that not just any case of perception could appropriately be described in this way. A sense of doubt or denial about the applicability of the property ϕ to the object under discussion always accompanies the use of such an expression. Hence, the expression does not truthfully describe most normal occurrences of perception, so the argument that philosophers must take account of it does not hold. Grice tentatively proposes to counter this objection by analysing the use of such an expression in the absence of doubt or denial not as literally false but as extremely misleading because of ‘a general feature or principle of the use of language’ (Grice 1961: 125). In effect, it would be misleading to add the extra information that the object looked blue to the speaker if this were unnecessary, and if a simple statement that ‘so-and-so is blue’ would do. Therefore ‘so-and-so looks ϕ to me’ stands as a logically, if not pragmatically, possible account of any experience of perception. Philosophers must indeed pay attention to this and to the support it offers to the causal theory. In this way Grice tackles a philosophical question by means of an analysis of ordinary language, but an analysis that is informed by his own insight into the potential differences between semantic meaning and meaning in use. It is an analysis that would not be possible using, for instance, Austin’s method of paying equal attention to all discernible nuances of meaning. More generally, understanding of meaning in ordinary language, and hence the prospects for philosophical progress based on its analysis, has been much enhanced by Grice’s chief intellectual legacy.
Paul Grice and the Philosophy of Ordinary Language
45
References Austin, J. L. (1956) A plea for excuses. Proceedings of the Aristotelean Society. Reprinted in J. L. Austin (1961) Philosophical Papers (Oxford: Clarendon Press), pp. 123–52. Austin, J. L. (1962) How to Do Things with Words (Oxford: Oxford University Press). Ayer, A. J. (ed.) (1959) Logical Positivism (Glencoe, IL: The Free Press). Ayer, A. J. (1971) Language, Truth and Logic (Harmondsworth: Pelican) [first published 1936]. Bach, K. (2006) Speech acts and pragmatism. In M. Devitt and R. Hanley (eds.) (2006), pp. 147–67. Carnap, R. (1932) The elimination of metaphysics through logical analysis of language. In A. J. Ayer (ed.) (1959), pp. 60–81 [originally published in German in Erkenntnis]. Chapman, S. (2005) Paul Grice, Philosopher and Linguist (Basingstoke and New York: Palgrave Macmillan). Chapman, S. (2008) Language and Empiricism, After the Vienna Circle (Basingstoke and New York: Palgrave Macmillan). Chomsky, N. (1957) Syntactic Structures (The Hague: Mouton). Devitt, M. and Hanley, R. (eds.) (2006) The Blackwell Guide to the Philosophy of Language (Oxford: Blackwell). Grice, H. P. (c. 1938) Negation and privation. In H. P. Grice papers, BANC MSS 90/135 c, The Bancroft Library, University of California, Berkeley. Grice, H. P. (1941) Personal identity. Mind, 50. Reprinted in J. Perry (ed.) (1975) Personal Identity (Berkeley: University of California Press), pp. 73–95. Grice, H. P. (1957) Meaning. The Philosophical Review, 66. Reprinted in H. P. Grice (1989), pp. 213–23. Grice, H. P. (1958) Postwar Oxford philosophy. In H. P. Grice (1989), pp. 171–80. Grice, H. P. (1961) The causal theory of perception. Proceedings of the Aristotelian Society. Reprinted in H. P. Grice (1989), pp. 224–47. Grice, H. P. (1969) Vacuous names. In D. Davidson and J. Hintikka (eds.) Words and Objections (Dordrecht: D. Reidel Publishing), pp. 118–45. Grice, H. P. (1975) Logic and conversation. In P. Cole and J. Morgan (eds.) Syntax and Semantics, vol. 3 (New York: Academic Press). Reprinted in H. P. Grice (1989), pp. 22–40. Grice, H. P. (1983) In conversation with Richard Warner and Judith Baker. Tape in H. P. Grice papers, BANC MSS 90/135 c, The Bancroft Library, University of California, Berkeley. Grice, H. P. (1986) Reply to Richards. In R. Grandy and R. Warner (eds.) Philosophical Grounds of Rationality (Oxford: Clarendon Press), pp. 45–106. Grice, H. P. (1989) Studies in the Way of Words (Cambridge, MA: Harvard University Press). Grice, H. P. (2001) Aspects of Reason (Oxford: Clarendon Press). Kant, I. (1929) Critique of Pure Reason (Norman Kemp Smith, trans.) (London: Macmillan Press [first German edition, 1781]. Loar, B. (2006) Language, thought and meaning. In M. Devitt and R. Hanley (eds.), pp. 77–90. Locke, J. (1993) An Essay Concerning Human Understanding (London: Everyman) [1st edn1690]. Mundle, C. W. K. (1979) A Critique of Linguistic Philosophy, 2nd edn (London: Glover and Blair) [1st edn 1970].
46
Siobhan Chapman
Naess, A. (1938) “Truth” as Conceived by Those who are Not Professional Philosophers (Oslo: Jacob Dybwad). Naess, A. (1993) Logical empiricism and the uniqueness of the Schlick seminar: a personal experience with consequences. In F. Stadler (ed.) Scientific Philosophy: Origins and Developments (Dordrecht: Kluwer Academic Publishers), pp. 11–25. Popper, K. (2002) The Logic of Scientific Discovery (London: Routledge) [1st German edn 1935, 1st English edn 1959]. Quine, W. V. O. (1943) Notes on existence and necessity. The Journal of Philosophy, 40: 113–27. Quine, W. V. O. (1951) Two dogmas of empiricism. In W. V. O.Quine (1953) From a Logical Point of View (Cambridge, MA: Harvard University Press), pp. 20–4. Russell, B. and Whitehead, A. (1910) Principia Mathematica (Cambridge: Cambridge University Press). Schlick, M. (1932) Positivism and realism. In A. J. Ayer (ed.) (1959), pp. 82–107 [originally published in German in Erkenntnis]. Strawson, P. F. (1969) Meaning and truth. Inaugural Lecture at the University of Oxford (Oxford: Oxford University Press). Reprinted in P. F. Strawson (1971) LogicoLinguistic Papers (London: Methuen), pp. 170–89. Strawson, P. F. (1990) Review of Paul Grice Studies in the Way of Words. Synthese, 84: 153–61. Urmson, J. O., Quine, W. V. O. and Hampshire, S. (1965) A symposium on Austin’s method, 62nd annual meeting of the American Philosophical Association, Eastern Division. In K. T. Fann (ed.) (1969) Symposium on J. L. Austin (London: Routledge), pp. 76–97. Wittgenstein, L. (1961) Tractus Logico-Philosophicus (D. F. Pears and B. F. McGuinness, trans.) (London: Routledge and Kegan Paul) [1st German edn 1921, 1st English edn 1922].
3 Intuition, the Paradigm Case Argument, and the Two Dogmas of Kant’otelianism Jay David Atlas
1. Introduction One of the most famous philosophical controversies, and one of the most significant for the history of late twentieth and early twenty-first century philosophy, was one in which H. Paul Grice and Sir Peter Strawson defended the analytic/synthetic distinction against Morton White’s and Quine’s arguments against it (in 1956, in 1950, and in 1951 respectively). I shall specifically examine Grice and Strawson’s Paradigm Case Argument in support of the distinction and their reductio argument against Quine’s ‘elimination’ of the synonymy relation; in so doing I will also be addressing Scott Soames’s use of the same form of argument against Quine’s ‘elimination’ of sentence-meaning. I shall argue that those reductio arguments fail to undermine Quine’s and White’s arguments against the analytic/synthetic distinction and, considering Grice’s latest views of the controversy, show how the arguments of Atlas (2005) provide a Gricean and Chomskyan defence of the analytic/synthetic distinction. I shall also examine critically Saul Kripke’s famous ‘argument from linguistic intuition’ in defence of essential properties. It too is a form of the Paradigm Case Argument in the Gricean fashion. I shall argue that the arguments from linguistic intuition fail to support the dogma of essential properties. In the next section I shall point out to future historians of English-speaking twentieth-century philosophy some of the salient features of the development of the ideas.1
2. Historical remarks The Quine of the late 1940s and early 1950s oscillated back and forth in his talk of ‘analytic’ and ‘a priori’, the first a semantic term, the second an epistemic one. After all, it is a self-conscious, logical empiricist tenet that sentences known a priori are just those that are analytically true. The careful 47
48
Jay David Atlas
distinctions among ‘necessary’, ‘a priori’, and ‘analytic’ that we associate with Kripke’s Naming and Necessity (1972/1980) were not introduced into analytical philosophy by Kripke, though he certainly made philosophers aware of their importance by the remarkable use he made of the contingent a priori and related notions. The basic distinctions, contrary to logical empiricist conflations, had already received serious discussion by Morton G. White and were well-known to Stanley Cavell and Burton Dreben, who were graduate student Teaching Assistants in the Harvard University course in which White lectured on this material prior to publication of his 1956 book Toward Reunion in Philosophy.2 In fact, the attack on the distinctions between analytic/synthetic and a priori/a posteriori knowledge that we associate with Quine’s 1951 ‘Two dogmas of empiricism’ was preceded by White’s 1950 ‘The analytic and the synthetic: an untenable dualism’ in Hook (1950) which was reprinted in Linsky (1952), a famous anthology in philosophy of language of the period. The version of ‘Two dogmas’ first published by Quine in The Philosophical Review contains in the ‘acknowledgment’ footnote at the bottom of the first page an appropriate acknowledgment of his Harvard colleague White’s prior published discussion. Quine also acknowledges the importance of White’s arguments to the development of Quine’s views, especially the ‘two sides of a dubious coin’ argument mentioned and footnoted in the first paragraph of Section 1 of Quine’s essay.3 In Chapter 16 of Volume 1 of Scott Soames’s Philosophical Analysis in the Twentieth Century (2003), Soames discusses Quine’s attack on the analytic/ synthetic distinction and, as he notes again in the second volume (2003: 224–5), he believes that Quine’s arguments fail. In fact he uses against Quine the form of argument that he reports Grice and Strawson to have used against Quine’s ‘Two dogmas of empiricism’ in their well-known 1956 essay ‘In defense of a dogma’. Grice and Strawson (1956: 146) had offered a reductio ad absurdum argument against Quine, writing, ‘We want only to point out that if we are to give up the notion of sentence-synonymy as senseless, we must give up the notion of sentence significance (of a sentence having meaning) as senseless too’. Soames (2003: 280) uses the same form of argument to show that Quine’s position must also be eliminativist about ‘ordinary’ notions of sense, reference, and truth of sentences or utterances – which he takes to be an absurd consequence. From Soames’s discussion of Quine’s Word and Object (1960), one would not suspect that Quine had addressed Grice and Strawson’s several forms of argument, one of which Quine believed suffered the logical defect of a metaphysical fallacy of subtraction (Quine 1960: 206–11). Quine (1960: 206–7) writes, ‘... it is argued that if we can speak of a sentence as meaningful, or as having meaning, then there must be a meaning that it has, and this meaning will be identical with or distinct from the meaning that another sentence has’. Quine comments that ‘... we could as well justify the hypostasis of sakes and unicorns on the basis of the idioms “for the sake of” and “is hunting unicorns” ’.
Paradigm Cases, Analytic-Synthetic, Essentialism
49
3. Grice and Strawson’s Paradigm Case and reductio arguments Grice and Strawson claim that our notion of a sentence being meaningful is an ‘ordinary’ notion as well as a philosophical one; the notion is one that is central to our ‘ordinary’ conceptual scheme and not to be abandoned cavalierly. Thus they take seriously the force of their reductio (quoted above). In the seventh paragraph of their essay they put the point first for philosophical use of the terms ‘analytic’ and ‘synthetic’ as follows: We can appeal, that is, to the fact that those who use the terms ‘analytic’ and ‘synthetic’ do to a very considerable extent agree in the applications they make of them. They apply the term ‘analytic’ to more or less the same cases, withhold it from more or less the same cases, and hesitate over more or less the same cases. This agreement extends not only to cases in which they been taught so to characterize, but to new cases. In short, ‘analytic’ and ‘synthetic’ have a more or less established philosophical use; and this seems to suggest that it is absurd, even senseless, to say that there is no such distinction [Paradigm Case Argument]. For, in general, if a pair of contrasting expressions are habitually and generally used in application to the same cases, where these cases do not form a closed list, this is a sufficient condition for saying that there are kinds of cases to which the expressions apply; and nothing more is needed for them to mark a distinction [Sophisticated Paradigm Case Argument].4 They then extend the argument to use of the terms in ordinary language as follows: Yet the denial that the distinction ... really exists is extremely paradoxical. ... For we frequently talk of the presence or absence of relations of synonymy between kinds of expressions – for example, conjunctions, particles of many kinds, whole sentences – where there does not appear to be any obvious substitute for the ordinary notion of synonymy ... . Is all such talk meaningless? Is all talk of correct or incorrect translation of sentences of one language into sentences of another meaningless? It is hard to believe that it is. ... If talk of sentence-synonymy is meaningless, then it seems that talk of sentences having a meaning at all must be meaningless [Reductio Version One].5 ... Grice and Strawson defend this last quoted sentence’s claim, what I call ‘Version One’ of their Reductio argument, in the next few sentences of their essay by saying that if one had a notion of a sentence-meaning, one would be able to define a notion of sentence-synonymy as, roughly, two sentences having the same set of meanings. This defence just directly begs the question
50 Jay David Atlas
against Quine’s criticisms of sentence-meanings as propositions. Yet Grice and Strawson never claim that a notion of a sentence-meaning is just an ‘ordinary’ notion with an obvious linguistic use and application. Nor do they claim that their notions of ‘having a meaning’ or ‘meaning something’ are ‘ordinary’ notions with uses and applications. They assume that their own applications of ‘meaningless’ and ‘meaningful’ constitute ‘ordinary’ uses of the terms. But the latter pair of notions does not enter into the defence of their claim, unlike ‘meaning something’ or ‘having a meaning’. They continue with what I call ‘Version Two’ of their Reductio argument: We want only to point out that if we are to give up the notion of sentence-synonymy as senseless, we must give up the notion of sentencesignificance (of a sentence having meaning) as senseless too [Reductio Version Two].6 Quine (1960: 207) replied to Grice and Strawson as follows: Another defense cites our undeniable intuitions of synonymy and analyticity [my emphasis]. These I recognize, but I have urged ... that they do not sustain a synonymy concept suited to identity of propositions, or meanings. It should be said also that the arguments ... were doubtless meant in part as defenses merely of the intuitions and not of a synonymy adequate to propositional identity [my emphasis], though the distinction never gets drawn. As defenses of the intuitions those arguments may be acknowledged as fair enough, always excepting the one involving the fallacy of subtraction. But so construed the arguments are no defense of propositions [my emphasis], however strongly motivated by a wish to defend them. Whistling in the dark is not the method of true philosophy. For Quine, intuitions are one thing; a theoretically adequate concept of synonymy, that is, a concept that provides a scientifically adequate account of the identity of propositions (not sentences, not utterances, but theoretical posits, abstract entities, called ‘propositions’ – the meanings of sentences), is another. Quine thinks that an intuitive concept does no work in articulating a theory of an abstract object, a theory of a proposition: ‘the meaning of a sentence’. Quine does not deny the existence of semantic intuitions of synonymy; he rejects their scientific adequacy in defining the identity of propositions. The disputants see the use of intuitions and ‘ordinary’ concepts differently. Quine’s last objections to Grice and Strawson apply to Soames’s reductio ad absurdum argument against Quine, which, like Grice and Strawson’s Reductio argument, imputes to Quine a radical eliminativism of our ‘ordinary’ notions (Soames 2003: 280). But Soames, notably, has a nuanced assessment of the Grice-Strawson Paradigm Case Argument. Soames (2003: 371)
Paradigm Cases, Analytic-Synthetic, Essentialism
51
writes, ‘The argument rests on the supposed fact that the sentences classified by philosophers as analytic are a more or less homogeneous collection that would be recognized by virtually anyone who had been given a rudimentary introduction to the distinction. ... Although there may well be a kernel of truth in their assumptions, there is reason to think that matters are more complicated than they realized’. The kernel of truth for Soames is the analogous argument that there is a ‘projectible’ distinction – my way of putting Soames’s (and Grice and Strawson’s) point with Nelson Goodman’s terminology7 – between sentences that are necessary and known a priori and sentences that are contingent and known a posteriori. Of course, this in itself will not entail a projectible distinction between the analytic and the synthetic, given Morton White’s and Saul Kripke’s reasons for distinguishing epistemic notions from semantic and metaphysical ones. And the traditional, Oxford ordinary language, Paradigm Case Argument about uniformity in ‘what speakers would say’ in applying the expression ‘analytic’ does not entail the existence of the distinction either. The uniformity of linguistic usage only gives, according to Soames – adapting a point of Gilbert Harman’s – a ‘presumption’ that there is a distinction, but it does not rule out the possibility that there is none (Soames 2003: 373) – contrary to the view of Grice and Strawson.8 Nonetheless, what Soames (375–7) accepts is the second of the GriceStrawson arguments, an allegedly reductio argument that Soames summarizes as ‘giving up synonymy only if one is willing to give up meaning and translation entirely’. Soames briefly glosses Quine’s view in Word and Object as – ‘disastrously’ – doing just that. My discussion of Quine above suggests that Quine’s argument is not quite what Soames supposes. Quine has a scientific (and metaphysical) objection to propositions as abstract objects: the meanings of sentences. If Soames thinks of his task as saving ordinary intuitions about synonymy from Quine’s alleged rejection of their existence, as contrasted with their utility in constructing a scientific theory of what a meaning is, Soames and Quine are talking past one another. In anticipation of a discussion of the Grice-Strawson Reductio Argument Version Two, I discuss a notion of explication, due to Carnap and Quine, that I shall use in my critical discussion. It will allow us to provide some common ground between Quine and Grice, so that there is less immediate begging of the question in each other’s arguments.
4. Quine on explication In his Word and Object (1960; see pp. 258–9), Quine asks ‘[W]hat [are we] most typically up to when in a philosophical spirit we offer an “analysis” or “explication” of some hitherto inadequately formulated “idea” or expression[?]’ As an elegant example he offers the way in which ordered pairs in mathematics are treated as objects, as values of individual variables of
52
Jay David Atlas
quantification. In order to explain relations, and to quantify over them, viz. over classes of ordered-pairs, one needs ordered-pairs as objects. Yet in a set theory in which individuals, sets of individuals, etc. are objects, and relations are classes of ordered-pairs, etc., one does not welcome as an additional category of object, ordered n-tuples of objects. One early solution, that of Norbert Wiener in 1914, is to identify the pair <x,y> with the class {{x},{y,Ø}}, whose members are the class {x}, whose sole member is x, and the class {y,Ø}, whose members are y and the empty class Ø. This construction satisfies the single axiom governing the identity conditions for an ordered-pair: (1)
if <x,y> =
, then x = z and y = w.
It is, he remarks: paradigmatic of what we are most typically up to ... We do not claim synonymy. We do not claim to make clear and explicit what the users of the unclear expression had unconsciously in mind all along. We do not expose hidden meanings, as the words ‘analysis’ and ‘explication’ might suggest; we supply lacks. We fix on the particular functions of the unclear expression that make it worth troubling about, and then devise a substitute, clear and couched in terms to our liking, that fills those functions. Beyond those conditions of partial agreement, dictated by our interests and purposes, any traits of the explicans come under the head of “don’t cares.” Under this head we are free to allow the explicans all manner of novel connotations never associated with the explicandum. This point is strikingly illustrated by Wiener’s {{x},{y,Ø}}. Our example is atypical in just one respect: the demands of partial agreement are preternaturally succinct and explicit, in (1) [my emphasis]. Quine (1960: 259) also anticipates a familiar line of objection to such explications. He writes that the notion that analysis must consist somehow in the uncovering of hidden meanings underlies also the recent tendency of some of the Oxford philosophers to take as their business an examination of the subtle irregularities of ordinary language. And there is no mistaking the obliviousness of various writers to the point about the don’t-cares. If nobody has objected to Wiener’s definition as falsifying the ordinary notion of ordered pair, e.g. in that it makes x and y members of members of <x,y>, the reason is perhaps that here the relevant considerations are so clearly on view; or perhaps only that ‘ordered pair’ is not ordinary language. Analogous objections to other and more classical philosophical analyses, at any rate, are not wanting. Russell’s theory of descriptions has been called wrong because of what it does to the truth-value gaps [my emphasis].9
Paradigm Cases, Analytic-Synthetic, Essentialism
53
In light of this paradigmatic example, he mentions that ‘Philosophical analysis, explication, has not always been seen in this way.10 Only the reading of a synonymy claim into analysis could engender the so-called paradox of analysis, which runs thus: how can a correct analysis be informative, since to understand it we must already know the meanings of its terms, and hence already know that the terms which it equates are synonymous?’11 Quine’s comments in Word and Object (1960: 257–66), on the complex relationship between explication and elimination, should be kept in mind when assessing the consequences of White’s and Kripke’s separation of semantic analyticity from metaphysical necessity and epistemic a prioricity. Had the logical empiricists’ explication of necessity and a prioricity by analyticity succeeded, they could have eliminated necessity and a prioricity in favour of Hume’s ‘relations of ideas’ – Kant’s analyticity. Even though on Quine’s view (ibid.: 260) explication is elimination ... (but not all elimination is explication – see ibid.: 261), explication does not necessitate elimination; it is also a defence of domestication. If there is a ‘striking if partial parallelism of function between the old troublesome form of expression and some form of expression figuring in the new method ... we are likely to view the latter form of expression as an explicans of the old, and, if it is longer, even abbreviate it by the old word ‘[my emphasis; see Quine (1960: 261)]. What was once inexplicable is now familiar; it is explicable and, though eliminable, no longer in need of elimination – it is just a sheep in wolf’s clothing. If Soames were right (even for the wrong reasons) that Quine fails to demolish the analytic/synthetic distinction, and the logical empiricists were wrong to try to reduce necessity and a prioricity to analyticity, necessity and a prioricity once again become independent notions. They become once again sources of philosophical puzzlement different from the puzzlement of being analytic – true or false solely by virtue of meaning – which was the puzzlement that disturbed White and Quine about the elimination strategy of the logical empiricists. So there had better be good arguments against the a priori/a posteriori knowledge distinction besides the attack on the analytic/ synthetic distinction. And, of course, there are.12
5. The fallacy in the Grice and Strawson Reductio Argument Version Two The original Grice-Strawson argument states that giving up a concept of ‘having the same meaning as’ entails giving up a concept of ‘having a meaning’, or ‘being significant’. But does it? If ‘x has the same meaning as y’ is a two-place relation predicate that is reflexive, symmetric, and transitive, it is an equivalence relation predicate. Sentences that have the same meaning are members of the same equivalence class. Then ‘meaning’ in the expression ‘x has the same meaning as y’ may be explicated (in Quine and Carnap’s sense of ‘explication’) by ‘equivalence class of the relation expressed by ‘x
54
Jay David Atlas
has the same meaning as y’. Let that equivalence-class notion of meaning be expressed by ‘Meaning’. Synonymy-equivalence classes of sentences yield an abstraction of the synonymy relation predicate, an abstract noun ‘Meaning’ expressing this special, equivalence-class-notion of Meaning. One can talk about the Meaning of a sentence, since the denotation of ‘Meaning’ is just the equivalence class of sentences in the language under the synonymy relation. If sentences belong to the same equivalence class, they Have (i.e. Belong to) the same Meaning. If we give up the relation of synonymy, we will give up the introduction of the equivalence class defined by it; we will give up the abstract entity Meaning, viz. the equivalence class of sentences under that synonymy relation. But will we in consequence give up meaning? Will we give up the notion of a sentence being meaningful, or having-a-meaning, as Grice and Strawson suggest we would? If we think of a sentence being meaningful as a sentence Having a Meaning, an explication of ‘meaningful’ that Carnap and Quine as well as Grice and Strawson could accept without begging any questions against the other side in the controversy, as it relies only on a defensible mathematical use of an intuitive, ‘ordinary’ synonymy relation, I see only two ways – one minimal, one maximal, both peculiar – in which Grice and Strawson’s conclusion against Quine would follow. Let us consider a language L in which no meaningful sentence is synonymous with a sentence other than itself and so belongs to no non-minimal equivalence class of synonymous sentences.13 The synonymy-equivalence class [S] for a sentence S is just, minimally, the singleton class {S}. The synonymy relation is reduced to the identity relation among meaningful sentences of L. In this language a sentence S that is meaningful is only self-synonymous and so Belongs to a Meaning {S}. If we gave up the notion of synonymy, and so gave up the concept of ‘selfsynonymy’ for this language, we would give up a predicate of sentences that was co-extensive with the predicate ‘meaningful’. For such a language, and such a Quinean explication of sentence-meaning, an extensional version of Grice and Strawson’s conclusion against Quine would follow. To give up the extension of the synonymy relation in this language, i.e. to give up selfsynonymy, would be to give up the extension of the predicate ‘meaningful’. But that does not entail, even for this language, that the intensional version of the conclusion follows. If we give up the concept of ‘synonymy’ for this language, we do not thereby give up the concept of ‘meaningfulness’ for this language. Now consider a language L′ in which every sentence is synonymous with all sentences.14 The synonymy-equivalence class [S′] for a sentence S′ is, maximally, the language L′ itself, viz. the set of all well-formed sentences. The synonymy relation is reduced to the universal relation among meaningful sentences of L′, e.g. that multiply-denoted by ‘((x=y) v (x≠y))’. In this language a sentence S′ that is meaningful is universally synonymous and so Belongs to a Meaning that is L′, viz. the Language itself. If we gave up the
Paradigm Cases, Analytic-Synthetic, Essentialism
55
notion of synonymy for this language, and so gave up the concept of universal synonymy, by giving up ‘the Meaning of x’ for this language, viz. giving up a predicate whose extension is L′ itself, we would give up a predicate of sentences that was co-extensive with the predicate ‘grammatical’. If all and only the grammatical sentences of L′ were meaningful in L′, we would be giving up the extension of the predicate ‘meaningful’ for L′. But that does not entail, even for this language, that the intensional version of the conclusion follows. If we give up the concept of ‘synonymy’ for this language, we do not thereby give up the concept of ‘meaningfulness’ for this language. For languages in which the synonymy relation is either not restricted to self-synonymy or not extended to universal synonymy, the Grice-Strawson argument, explicated in the rigorous Quinean way as about Meaning, is not valid under either the extensional or intensional interpretation. If the meaningfulness of a sentence were the same as the Meaning of a sentence, or if a sentence being meaningful were the same as the sentence Having (or Belonging to) a Meaning – i.e. a sentence ‘having-a-meaning’ were the same as the sentence Having (or Belonging to) a Meaning, an extensional version of Grice and Strawson’s conclusion would, or could be made to, follow. But those tacit identity assumptions are false, and in any case they would not be acceptable to Grice and Strawson or consistent with Version Two of their argument. Even in the best case for their argument, in which Carnap and Quine’s commitments to explication suggest the use of the Meaning abstraction from the synonymy relation as an explication of intuitive sentencemeaning, giving up the concept of ‘synonymy’ does not entail giving up the concept of ‘meaningfulness’. ‘Being in an equivalence class of synonymous sentences’ does not express the same concept as ‘being meaningful’ does, even if the predicates were co-extensive in a language, e.g. one as peculiar as the one in which the synonymy relation is the self-synonymy relation, i.e. is the identity relation for sentences of this language, or as peculiar as the one in which the synonymy relation is the universal synonymy relation, i.e. is identical with the extension of ‘well-formed sentence-pairs’ for the language. Strawson and Grice’s reductio argument against Quine fails. Since Soames (2003: 280) uses the same form of argument to show that Quine’s position must also be eliminativist about ‘ordinary’ notions of sense, reference, and truth of sentences or utterances – which like Grice and Strawson he takes to be an absurd consequence – Soames’s argument against Quine fails as well. Of course, it is important to note, the failure of Grice and Strawson’s argument does not show that Quine is correct to reject the semantic and linguistic analytic/synthetic distinction even if he were correct to reject the epistemic a priori/a posteriori knowledge distinction. Furthermore, Quine admitted in Word and Object (1960) the acceptability of the intuitive notions of synonymy and meaningfulness for purposes other than defining a scientifically
56 Jay David Atlas
acceptable notion of propositional identity. But I have illustrated why accepting an intuitive notion of synonymy allows a coherent definition of an abstract entity the Meaning of a sentence, an abstract entity that even Quine could accept. So there must be more to Quine’s attack on the analytic/synthetic distinction.
6. Quine’s attack on synonymy For Quine there is the semantic question whether two sentences A and B are synonymous, and there is the epistemic question how to decide whether two sentences A and B are synonymous. The latter question is answered by the mechanism of translation. Quine’s account of translation depends essentially on verbal behaviour and on dispositions to verbal behaviour. Theoretical sentences, unlike ‘observation sentences’, do not merely report observational evidence. So Quine inferred that translation of theoretical sentences would not be reducible to translation of sentences describing their evidential content. The crucial claim was that there could be empirically adequate translation functions m and m∗ that disagreed on the translation of theoretical sentences. If in translation manual m the translation of α under m was β, i.e. Tm (α,β), the translation of α under manual m∗ might not be β, i.e. ¬ Tm∗(α,β). Since the verification of sentence synonymy takes place by translation, the epistemic version of the transitivity of synonymy, i.e. of the condition ‘if Syn(α,β) & Syn(β,γ), then Syn(α,γ)’, is the transitivity of translation, i.e. if Tm(α,β) & Tm (β,γ), then Tm (α,γ). But for theoretical sentences, Quine’s thesis of the indeterminacy of translation is that Tm and Tm∗ are equally good translation relations. Substituting in the above, our principle could read: if Tm (α,β) & Tm (β,γ), then Tm∗(α,γ). Since ex hypothesi ¬ Tm∗(α,γ), the transitivity principle is false for translation. So, for Quine, it is unjustified for synonymy. But then if the synonymy relation cannot be transitive, it cannot be an equivalence relation, and so it cannot define an equivalence class. Hence it cannot justify an abstraction M allowing us to express the equivalence relation by ‘Having the same M’. In that case, by Quine’s verificationist account of synonymy, it is not an equivalence relation, and the Meaning of a sentence S – a Proposition – cannot be defined as the equivalence class of S under the verificationist synonymy relation – the epistemic version of synonymy, which is translation. There is no such equivalence relation. So there is no Meaning, no Proposition. In the 1960s this view seemed to me implicit in Word and Object and other writings, including the essay ‘Propositional objects’ (Quine 1969), which I had heard Quine deliver in the spring of 1965, when I was a 20-year-old undergraduate.15 Thirty-four years later I finally had confirmation of this interpretation in Quine’s (1999: 76) cryptic comment on Davidson, in
Paradigm Cases, Analytic-Synthetic, Essentialism
57
Quine’s essay ‘Where do we disagree?’, when he wrote: My own thought experiment of radical translation had a narrower purpose: I was challenging the notion of propositions as meanings of sentences. Indeterminacy of translation meant failure of sentence synonymy to qualify as an equivalence relation, hence failure of individuation of propositions; and there is no entity without individuation, without identity.16
7. Grice’s criticism of the Paradigm Case Argument for the analytic/synthetic distinction, his suggestions for a new defence, and Atlas’s version of the defence Grice’s argument was what he admitted was a sophisticated version of the Paradigm Case Argument. He commented (1986: 54): ... the attempt by Strawson and myself to defend the distinction by a (one hopes) Sophisticated form of Paradigm Case argument fails to meet, or even to lay eyes on, the characteristic rebuttal of such types of argument, namely that the fact that a certain concept or distinction is frequently deployed by a population of speakers and thinkers offers no guarantee that the concept or distinction in question can survive rigorous theoretical scrutiny. To my mind, the mistake by both parties [Quine and Grice] has been to try to support, or to discredit, the analytic/synthetic Distinction as something which is detectably present in the use of natural language; ... This last self-critical comment by Grice is precisely the criticism that I shall make later in this chapter of Saul Kripke’s attempt to justify the intuitive ‘ordinariness’ of the distinction between essential and accidental properties from the evidence of the use of ordinary, natural language. What, then, would be Grice’s latest thoughts on a way to defend the analytic/synthetic distinction? He continues (1986: 54–5): ... it would have been better to take the hint offered by the appearance of the ‘family circle’ of concepts pointed out by Quine, and to regard an analytic/synthetic distinction not as a (supposedly) detectable element in natural language, but rather as a theoretical device which it might, or again might not, be feasible and desirable to incorporate into some systematic treatment of natural language. The viability of the distinction, then, would be a theoretical question which, so far as I can see, remains to be decided; and the decision is not likely to be easy, since it is by no means apparent what kind of theoretical structure would prove to be the home of such a distinction, should it find a home.
58 Jay David Atlas
These suggestions by Grice in 1986 were an articulation of a strategy that had been employed independently by Atlas (1984). The analytic/synthetic distinction was a theoretical posit in the description of the logical relation of ‘analytical entailment’ between lexical items in a Horn Scale of degree adverbials from which conversational implicatures would be generated, e.g. . Employing as well the distinctions among gradable predicates, gradable accomplishment and gradable achievement predicates (Vendler 1967: 103–7), it was possible to show: (a) For gradable F, x does not F analytically entails x does not quite F. (b) Conversely, for gradable accomplishment and achievement predicates G, x does not quite G analytically entails x does not G. (c) The Horn Scale shows that for gradable F asserting x almost F’s conversationally implicates x does not quite F. (d) For gradable accomplishment and achievement predicates G, asserting x almost G’s implicates x does not quite G, which analytically and “directly” entails x does not G. Thus, “x almost G’s” ||- D • ||->> x does not G, viz. the composition of conversational implicature with direct entailment (roughly, the entailment of a sub-formula) takes the assertion of x almost G’s into the sentence x does not G. Note that this contrasts with the commonly accepted but incorrect thesis that x almost G’s entails x does not G. What are the commitments of the kind of explanation that I gave in 1984 for inferences among adverbials? First, I relied on analytical entailments among adverbial predicate and sentence expressions in a natural language; e.g. John is not wholly as tall as Brian analytically entails John is not as tall as Brian, and upon lexical relations underlying the analytical entailment from John is taller than Brian to Brian is shorter than John.17 Second, I have relied on conversational implicatures among assertions of sentence expressions in a language. The mechanism of implicature requires that there be lexical Horn Scales of lexical items from the language, like the adverbial scale discussed above, in order for there to be adequate theoretical explanations of utterance-interpretations. That is, a speaker of English must know that almost belongs to such a scale, and hence that almost bears a special semantic relation, not to every word in the language but to an identifiable finite set of adverbial expressions whose semantical relations to each other in the scale are known to competent speakers. If one takes the phenomena of generalized conversational inferenda and implicata seriously, one cannot escape an explanation that makes essential use of lexical meaning, and hence a semantical distinction between the analytic and synthetic, and essential use of an-atomic properties (in Jerry Fodor and Ernie Lepore’s sense; 1992: 1), like belonging to a particular Horn Scale of lexical items. Meaning holism
Paradigm Cases, Analytic-Synthetic, Essentialism
59
is inadequate if, as has been argued by Fodor and Lepore, it is inconsistent with a principled distinction between the analytic and synthetic, which is required by the explanatory use of analytical entailment in linguistic theory; semantical atomism is inadequate if it is inconsistent with the essential use in linguistic explanation of an-atomic semantic properties, which are required by linguistic explanation of generalized conversational inferenda and implicatures. Linguistic theory requires the philosophically unexciting middle ground – semantic molecularism or semantic corporatism. Thus linguistic science deflates the views of the exciting philosophical extremes, e.g. Quine’s Word and Object (1960: 15–16) and Richard Rorty’s Philosophy and the Mirror of Nature (1979: 319), where philosophical battles are typically fought – on the margins of science. Neither semantical atomism alone nor semantical holism alone have any explanatory value for empirical linguistic theory.18 I now turn to another application of a version of the Paradigm Case Argument, an argument for the existence of essential properties due to Saul Kripke. The faults of the argument, independently of the later view of Paul Grice discussed above, reside first in the mistaken reports of the data of ordinary use of natural language.
8. Soames’s defence of the intuitive appeal of essential properties, non-specificity, Locke, and Chomsky Scott Soames is enough of a realist about possible worlds to say things like, ‘We might try to specify possible world-states in which Nixon has a certain property P, but fail to do so because, in fact, Nixon couldn’t have had that property. For example, we can’t successfully stipulate a possible situation in which Nixon is an inanimate object. In this sort of case there is no such possible world-state corresponding to our specification. ... [F]acts about possibility are not created or determined by our stipulation’ (Soames 2003: 356).19 Well, how about a dead Nixon lying on the mortuary table just minutes after his brain activity ceases? Nixon is an inanimate object, right? Or suppose that Pat, Julie, Tricia, and David had decided, just after Air Force One had landed in San Clemente after the trip from Washington, DC on 8 August 1974, cryogenically to freeze Nixon. Nixon would be an inanimate object, right? Most philosophers have failed to appreciate that like other nouns, proper nouns might have had syntactico-semantic features, e.g. [+ABSTRACT] corresponding to a notion of abstractness, or [–ABSTRACT] corresponding to a notion of non-abstractness, e.g. concreteness, and similarly for [+ANIMATE] and [–ANIMATE], that, contrary to fact, would have excluded the grammatical use of a proper name like ‘Nixon’ in the sentence of the preceding paragraph. One might have thought that it is grammatical to say ‘A dead body is lying on the table’ but ungrammatical (semantically anomalous) to say ‘A dead Nixon is lying on the table’, were ‘Nixon’ to have the
60
Jay David Atlas
feature [+ANIMATE], being a Kripkean rigid designator for the person Richard M. Nixon or for the ego. As in many other such cases, I believe, as just indicated in the previous paragraph, that the proper name is semantically non-specific between such features.20 This is yet another example of the way in which linguistic properties constrain the uses made of a lexical item but do not semantically determine those uses.21 Trying to capture a notion of semantical non-specificity in the case of ‘abstract Ideas’, John Locke (Essay, IV. vii. 9) notoriously remarks of ‘triangle’ that the ‘general idea’ that it expresses ‘must be neither oblique, nor rectangular, neither equilateral, equicrural, nor scalenon; but all and none of these at once’. Commentators have not been kind to Locke, taking his words ‘all and none of these at once’ to indicate a blatant literal contradiction, when it is clear that he meant them figuratively, trying to capture the technical notion of ‘imperfection’ in Medieval logic, as it appears in the Draft A version of his Essay, a logical notion corresponding to the contemporary linguistic notion of ‘semantically non-specific’.22 But with historical distance, commentators are uncharitable, thinking Locke an idiot. What I would be surprised to find them doing is thinking the same of Noam Chomsky. When Chomsky writes, in a recent essay ‘Language as a natural object’ (2000: 126): A city is both concrete and abstract, both animate and inanimate. ... The abstract character of London is crucial to its individuation. If London is reduced to dust, it – that is, London – can be rebuilt elsewhere and be the same city, London. ... Pronouns involve dependency of reference, but not necessarily to the same thing; and both referential dependence and the narrower notion of sameness involve roles in a highly intricate space of human interests and concerns. it would be unusual for philosophers to think that Noam Chomsky did not notice what he was writing when he wrote ‘both animate and inanimate’ and that he was an idiot. Like Locke, Chomsky was speaking in Carnap’s ‘material mode’ rather than his ‘formal mode’. I would have said that the noun ‘city’ is semantically non-specific between [+ABSTRACT] and [–ABSTRACT], and between [+ANIMATE] and [–ANIMATE]. 23 In the case of Nixon, I would have said that ‘Nixon’ is semantically non-specific in the same way. Wittgenstein would have said, in the ‘material mode’, something gnomic-sounding like ‘Nixon is not the same as his body, but not another thing either’. But the point would have been the same. For Soames an essential property of an object ‘is a property the object could not lack in any circumstance in which it existed at all’. He writes, ‘... Uncontroversial examples of essential properties of mine are rarer [than contingent properties], but the following seem to be good candidates: the property of being a human being, the property of having a brain, the property of having a body made up of molecules, the property of being mortal,
Paradigm Cases, Analytic-Synthetic, Essentialism
61
the property of not being identical with Saul Kripke’ (2003: 347). Somewhere in Zettel Wittgenstein asks: Were we to open an ordinary, functioning man’s skull and discover that it was empty, what would be the consequence for our conceptions of human thought and action? And he answers that there would be none. So at least for Wittgenstein, the property of having a brain is not a likely candidate for an essential property of a thinking human being. (There’s always the stomach, the liver, or the heart, depending on the century you’re in.) Why, then, is Soames tempted to think that the property expressed by the predicate ‘x has a brain’ is an essential property of a human being? Wittgenstein remarks (Zettel: 605), ‘One of the most dangerous of ideas for a philosopher is, oddly enough, that we think with our heads or in our heads’. Soames believes that the distinction between contingent and essential properties is an ‘intuitive’ one, one that ‘we all recognize the intelligibility of ... ’ (2003: 350). For Soames it is ‘intuitive’ that ‘being raised in Seattle’ is a contingent property of himself, by contrast with ‘being a sentient creature’. We have a ‘pre-philosophical conviction that essentialist claims are intelligible ...’ (ibid.: 351); ‘we begin with an intuitive distinction’ between properties’ (ibid.: 350). Quine is sceptical about the intelligibility of this distinction between properties (1960: 199), but Soames takes Quine’s objections to hinge on the existence of rigid designators in Kripke’s sense (2003: 347) – a singular term that designates the same individual in every possible world in which it exists.
9. The origin of rigidly designating singular terms: Quine’s condition on the quantified variable in modal logic Quine has a view that might be glossed by the following: however things are referred to, if a thing must be a so-and-so, it must be a so-and-so however referred to. That is, Quine (1960: 197–8) objects to quantification into modal contexts unless the variables are so interpreted so that no open sentence ‘Fx’ is such that (∀x)(∀y)(x = y → (䊐Fx & ¬䊐Fy)). Equivalently, the variables are interpreted so that every ‘F’ is such that (∀x)(∀y)(x = y → (䊐Fx → 䊐Fy)).24 The trouble is that Quine thinks that it follows from this that if in fact one has two properties that each ‘specify’ uniquely the same object, it must be the case that they specify the same object. That is, the trouble is that Quine then characterizes his modest condition on quantification as requiring that for any predicates ‘F’ and ‘G’ such that for some a, {a} = {x : Fx} = {x : Gx }, i.e. both ‘F’ and ‘G’ specify the same object a, it is the case that 䊐(∀z)(Fz ↔ Gz), i.e. the predicates are necessarily equivalent. From this latter condition he readily deduces an absurdity for a predicate ‘Fx’ =df ‘(P and x = a )’ , where P is an arbitrary, true but not necessarily true, sentence, and a predicate ‘Gx’ =df ‘(x = a)’; he deduces the conclusion that 䊐 P, i.e. a merely contingently true P is deduced to be necessarily true – an absurdity. His claim is then that any quantified modal logic must presuppose his latter
62 Jay David Atlas
condition on predicates and so cause modal distinctions among sentences to collapse. The alternative is to specify an object only by predicates that one has already separated into ones that are contingent or into ones that are necessary. Then his objection (1960: 199) is that he cannot make sense of the distinction between attributes (predicates) that are essential (necessary) and those that are accidental (contingent).
10. Quine’s failed counter-example to the necessary/contingent distinction His ‘intuitive’ counter-example is the case of the cycling mathematician. If mathematicians are necessarily rational but not necessarily two-legged, and cyclists are necessarily two-legged but not necessarily rational, Quine asks whether the individual who is a cycling mathematician [a mathematically proficient cyclist?] is necessarily rational and contingently two-legged, or contingently rational and necessarily two-legged. Either option seems equally good, but that symmetry seems puzzling. Shouldn’t there be a definite answer? So Quine does not think this Platonic/Aristotelian distinction between contingent and necessary properties makes any philosophical or scientific sense. They make seemingly decidable questions undecidable. One might have observed that Quine’s presumptively exhaustive ‘or’formulation of his question about the properties of the cycling mathematician implies that the cycling mathematician is both contingently rational and contingently two-legged. From that, when combined with Quine’s prior assumptions about mathematicians and cyclists, one infers that either it’s not the case that mathematicians are necessarily rational or it’s not the case that cyclists are necessarily two-legged (or both). So the premises of his example cannot both be true, or at least it is presumed that the individual is contingently rational and contingently two-legged. In either case his example creates no ‘paradox’ for the distinction between contingent and necessary. None of Quine’s concerns in Word and Object is directed to the problem of rigid designators. Quine’s problem is not even rigid predicates. Quine’s problem, if there is one, is his view that the semantics of modal logic makes a special class of equivalent predicates necessarily equivalent. It is the analogous principle to Quine’s for equivalent singleton predicates that Kripke accepts for the identity expressed by individual variables, viz. that if x = y then 䊐 (x = y). But, as Bishop Butler would have said, equivalence of predicates is not the identity of individuals. The semantic issues are not the same. If there is direct conflict between Quine and Kripke, it is on matters of essence and accident, not on matters of rigid designation.25 Neither Kripke nor Quine has offered a demonstration or a refutation of the acceptability or even plausibility of essential properties that settles the metaphysical issue. Quine’s objections are intuitive, with a tacit appeal to the claim that no contemporary scientific theory needs any metaphysical
Paradigm Cases, Analytic-Synthetic, Essentialism
63
machinery borrowed from Aristotle. Soames’s (2003: 350) defence of Kripke is an intuitive appeal to a list of two essential attributes of Scott Soames that ‘everyone’ understands – everyone except Wittgenstein and Albritton, at least: being a sentient being and not being identical to Saul Kripke. I am dubious of the necessity of the first, as I have suggested above in the case of Nixon, and I am dubious of the physical properties that Soames also uses as examples (2003: 347), viz. being molecular and being mortal. Were I immortal, would I not be me? Would I not exist? Surely, if the test is the view of the ‘ordinary person’, the answer is ‘No; I’m still me’.26 (Note: Soames did not ask the theological question whether immortality were a necessary property of my soul.) And if my body were not molecular, well, it would have to be something else, e.g. the tiniest, 10-dimensional superstrings, or maybe just pneuma, or bits of F.H. Bradley’s Absolute, or a mode of Spinoza’s God (i.e. Nature). The second example of non-identity is a logical property of Soames that does not have much metaphysical force. But the merits of the case aside, what I find most arresting – some might say ‘startling’ – about Soames’s views is his appeal to intuition, to ‘what everyone understands’, to the claim – like Grice’s in his defence of the analytic by paradigm cases – that what he is saying is just the ‘ordinary view’ of the man on the Clapham omnibus. As far as intuition goes, Soames’s appeal to ‘what everyone understands’ about essential predications and Quine’s appeal to ‘what every scientist understands’ about the scientific irrelevance of the essence/accident distinction are equally matters of claims for intuition, rather than knock-down arguments, in their writings on this subject. Still, Quine has a chance of substantiating his claim, by examining the character of scientific theory and practice. On its face Quine’s claim seems right. The question for him is always the later one: how shall the construction of a metaphysics be guided by scientific theory, and why should it be? On its face Soames’s claim does not seem right, despite the passages from Kripke’s Naming and Necessity (1980) that he quotes in support of it. For the question is whether Kripke and Soames are arguing for the same claim, and, if they are, whether what is intuitive about their claims actually shows what they think it shows about the intuitiveness of necessary, essential properties. So I shall discuss briefly the famous passage from Kripke, quoted by Soames (2003: 352–3), in defence of the intuitiveness of essence.
11. The linguistic defects in Kripke’s arguments for the ‘ordinary’ intuitive notion of essential properties The crucial passage from Kripke is this (1980: 41–2): ... it is very far from being true that this idea [that a property can meaningfully be held to be essential or accidental to an object independently
64
Jay David Atlas
of its description] is a notion which has no intuitive content, which means nothing to the ordinary man. Suppose that someone said, pointing to Nixon, ‘That’s the guy who might have lost’. Someone else says ‘Oh no, if you describe him as “Nixon”, then he might have lost; but, of course, describing him as the winner, then it is not true that he might have lost’. Now which one is being the philosopher, here, the unintuitive man? It seems to me obviously to be the second. The second man has a philosophical theory. The first man would say, and with great conviction, ‘Well, of course the winner of the election might have been someone else’ [sic]. The actual winner, had the course of the campaign been different, might have been the loser, and someone else the winner; or there might have been no election at all. So, such terms as “the winner” and “the loser” don’t designate the same objects in all possible worlds. On the other hand, the term “Nixon” is just a name of this man. When you ask whether it is necessary or contingent that Nixon won the election, you are asking the intuitive question whether in some counterfactual situation, this man would in fact have lost the election. If someone thinks that the notion of a necessary or contingent property (forget whether there are any nontrivial necessary properties [and consider] just the meaningfulness of the notion) is a philosopher’s notion with no intuitive content, he is wrong. (pp. 41–2) I want to focus on the crucial claim of Kripke’s: When you ask whether it is necessary or contingent that Nixon won the election, you are asking the intuitive question whether in some counterfactual situation, this man would in fact have lost the election. There is a natural, ordinary language paraphrase of Kripke’s use of the modal verb ‘would’ in ‘This man would have lost the election (in some possible but non-actual situation)’, which is ‘It was possible for this man to have lost the election’.27 In Kripke’s ‘would’ sentence and in the ‘possible for’ paraphrase, the referring term ‘this man’ is syntactically placed so that it has an obvious referential interpretation (a transparent occurrence, in Quine’s language). Though in logical syntax it is possible to use Peano-abstraction to create a predicate, re-casting ‘It is necessary that Fa’ as the Peano-English sentence ‘a is such that it is necessary that it is F’, in ordinary English neither of the ordinary ‘would’ or ‘it is possible for’ sentences are paraphrasable by a ‘possible that’ sentence ‘*It was possible that this man did/does lose the election’ – the verb tenses are grammatically wrong. There is a related ‘possible that’ sentence in English that is grammatical: ‘It is possible that this man did lose/lost the election’, but this use of the English word ‘possible’ in this ‘possible that’ sentence is only paraphrasable by a ‘not certain that’ sentence ‘It is not certain that this man did not lose the election’; unfortunately ‘not
Paradigm Cases, Analytic-Synthetic, Essentialism
65
certain’ does not give the meaning intended. So, contrary to Kripke’s claim, the ‘would’ or ‘possible for’ sentence in which ‘this man’ occurs transparently, viz. ‘It was possible for this man to have lost the election’, does not paraphrase the ‘possible that’ sentence ‘It is possible that this man did lose/ lost the election’. Thus the negative sentence ‘It was not possible for this man to have lost the election’ does not paraphrase the negative sentence ‘It is not possible that this man did lose/lost the election’. So, assuming that ties in elections are not possible, Kripke’s original sentence ‘It is necessary that this man won the election’ is not paraphrasable by ‘It was necessary for this man to have won the election’, i.e. in no possible situation, counterfactual or otherwise, would this man have lost the election. But that is just to say that the ‘that’ sentence ‘It is necessary that proper name/demonstrative N is F’ does not have the paraphrase in ordinary English that would express the de re interpretation in which the proper name or demonstrative term occurs transparently, hence in which the referent of the proper name possesses F-ness necessarily. Thus Kripke’s claim that there is a natural, ordinary paraphrase by a ‘would’ or by a ‘possible for’ sentence, in which a referring term occurs transparently, of a ‘necessary that’ sentence is just a linguistic falsehood! It hardly suffices as a defence of the intuitiveness of an essentialist, de re interpretation of ‘necessary that’ sentences like ‘It is necessary that proper name is F’, since his linguistic evidence is actually an incorrect paraphrase of ‘necessary that’ using ‘would’. This is not to deny that in logical syntax one can use Peano-abstraction and give the appropriate model theory. It is to say that a defence of the intuitiveness of predicating necessary properties of an individual by appeal to the naturalness and linguistic obviousness of ordinary paraphrases of ‘necessary that’ sentences by ‘would’ sentences in which a singular term occurs transparently depends on whether there is such a paraphrase. In Kripke’s case, there is no such ordinary, intuitive paraphrase. 28 Kripke’s language here is, in any case, different from his report of the ordinary man, who said ‘That’s the guy who might have lost the election’. This special use of ‘might’, the past of ‘may’, indicates a possibility just as ‘could have’ does. What Kripke’s ordinary guy cannot felicitously assert is ‘? That’s the guy who may have lost the election – but he didn’t’. What Kripke should have said, less misleadingly, is that the ordinary guy says, ‘That’s the guy who could have lost the election’, paraphrasable by ‘It was possible for that guy to have lost the election’ or by ‘It would have been possible for that guy to lose the election’, as discussed above. And as above, the evaluation of ‘could have’ statements is not restricted to counterfactual situations.29 What the ordinary guy can felicitously assert is not a paraphrase of ‘It is necessary that Nixon won the election’. So the latter sentence does not express, in ordinary language, a necessary or essential property of an individual. Kripke has incorrect linguistic evidence for the alleged intuitiveness and correctness of ‘It is necessary that P’ and ‘It is possible/contingent that P’ to express
66
Jay David Atlas
necessity or possibility predicates in English. How, then, is he in a position to assert the following? If someone thinks that the notion of a necessary or contingent property ... is a philosopher’s notion with no intuitive content, he is wrong. ... [P]eople who think the notion of accidental property unintuitive have intuition reversed, I think. Property? And necessary and contingent ones, at that? The incorrect linguistic judgments supporting Kripke’s previous remark are now supposed to defend the intuitiveness of the concept of a necessary property? Really? The sentence ‘It was possible for this man to have lost the election’ – not, notice, ‘It is possible that this man ...’ – is supposed to justify intuitively the ascription to this man of the property expressed by ‘is such that x possibly has lost the election’, or the ungrammatical ‘*is to have possibly lost the election’, when concatenated with ‘this man’? The former, grammatical expression does not paraphrase the original sentence ‘It was possible for ... to have ...’, and the latter expression, since it is ungrammatical, is hopeless as a paraphrase. Are these the ‘ordinary meanings’ of ‘ordinary language’ that make the ‘ordinary person’ an intuitive understander of necessary and contingent properties? I think not. I asked whether Kripke and Soames are arguing for the same claim, and, if they are, whether what is intuitive shows what they think it shows about the intuitiveness of necessary, essential properties. I shall merely assume that Soames agrees with Kripke’s formulation of the problem of showing the intuitiveness of the distinction between necessary and contingent properties. I think it clear from linguistic considerations that Kripke’s linguistic claims do not show what he thinks they do about the ‘ordinary’ intuitions of the man on the Clapham omnibus about essences.
12. Conclusions Where Richard Rorty in Philosophy and the Mirror of Nature (1979) took Quine to have questioned successfully the intelligibility of the eighteenth-century epistemological question, ‘How is synthetic a priori knowledge possible?’, a question that needed to be answered if Kant was to refute Humean scepticism, and thus make unnecessary our preoccupation with this budget of eighteenth-century philosophical problems, Soames takes Kripke to have re-established, not the intuitive content of contingent a priori knowledge, which on Soames’s account (2003: 418–21) turns out not to be very intuitive (e.g. propositions expressed by ‘P if and only if actually P’ – of which contingent a priori knowledge is possible but not very interesting to a Kant wanting to refute Hume’s scepticism) but rather, the intuitive content of necessary a posteriori knowledge.
Paradigm Cases, Analytic-Synthetic, Essentialism
67
For Soames, as for most philosophers, the metaphysical interest of the view is the account of ‘theoretical identities’ employing natural kind predicates. With some amendments and extensions of Kripke’s views, Soames argues for the following to be necessary a posteriori (2003: 438): ‘For all x, x is a drop of water if and only if x is a drop of a substance molecules of which contain two hydrogen atoms and one oxygen atom’. On the left-hand side of the biconditional is a simple natural kind predicate ‘drop of water’ and on the right-hand side is a compound natural kind predicate. The former are like proper names; the latter are like singular descriptions. By contrast, ‘For all x, x is a woodchuck if and only if x is a groundhog’ expresses the necessary, a priori proposition that ‘For all x, x is a woodchuck if and only if x is a woodchuck’ expresses; one just need not know that the original sentence expresses that proposition. Soames seeks to disarm objections to a Kripkean account of necessarily true theoretical ‘identities’ by positing that not all epistemic possibilities are metaphysical possibilities; some of them are impossible (ibid.: 455). This and the distinction between sentences and propositions keep at bay the worst consequences of rationalism’s theory of knowledge: the view that if a metaphysical truth is necessary it must be known a priori (or known innately, like Descartes’s knowledge of the meaning of ‘jaune’, or remembered, like Plato’s knowledge of the Forms) while still preserving the thrill of acquiring knowledge of necessary truths. For Dewey, at first blush, this would still seem to be the ‘quest for certainty’ unabated. But Dewey would think, once understanding that these identities are necessarily true if true actually, that the metaphysical certainty of necessity is earned on the cheap. One still has to do the labour of empirical investigation to figure out what is true actually. Then, given names and natural kind terms, one gets the metaphysical necessity of the truth for free. So water is necessarily H2O, and maybe Mind is necessarily not Matter. But for Dewey the thrill of necessity would be gone. It’s now only a semantic thrill, despite the semantic’s nom de plume ‘metaphysical’; it’s the semantics of a modal language giving an interpretation of individual constants, and the posit (the ‘persuasive definition’) that a ‘substance’ is a natural kind specified by how it is constituted physically (ibid.: 441), where, for Soames, a natural kind is by stipulation (ibid.: 436) the unique entity described by metaphysically necessarily co-extensive predicates (ibid.: 440). Dewey would think that all the interesting work is now done by (a) the posit, (b) the uniqueness stipulation, and (c) the state of one’s scientific theory of ‘the physical’. If one abandons, as Hilary Putnam has now done, the metaphysically necessary in the case of water, much of this work cannot be done.30 Soames concludes with the claim that Kripke’s discovery of metaphysically necessary a posterori truths is one of ‘the great philosophical achievements of the twentieth century; it has transformed the philosophical landscape, recalibrated our sense of what is possible, and reshaped our understanding of our own philosophical past’ (ibid.: 456). Dewey would have
68
Jay David Atlas
applauded the brilliance of Kripke’s semantics for modal logic while thinking, with A.N. Whitehead, that the ways around the counter-intuitive aspects of Kripke’s, Salmon’s, and Soames’s views – propositions vs. sentences, epistemically possible but metaphysically impossible states – are not inventions mothered by necessity but futile dodges instead. For the interesting philosophical work is now to be done elsewhere: understanding (or failing to understand) natural kinds and a scientific theory of the physical while encumbered by notions of propositions, possible worlds, etc. The next ‘Quine’ will write a book ‘Thoughts and Things’ dismantling all this confident talk of metaphysical necessity with the same brio with which Morton White, Nelson Goodman, and W.V.O. Quine attacked semantical analyticity and a priori knowledge. Only this time, instead of psychological behaviourism as the scientific paradigm, there will be some form of cognitive science and neuroscience lurking in the background of the young philosopher who will abandon all these metaphysical necessities. T.H. Green thought Hegel’s Absolute Idealism was the culmination of a dialectical development through the rationalists, the empiricists, and Kant. Scott Soames thinks that Kripke’s logical rationalism is the culmination of a dialectical development. The theses of A.J. Ayer’s logical empiricism and the anti-theses of ‘meaning as use’ in the later Wittgenstein and Austin (though in rather different ways in their two cases) lead to the ‘transcendental’ synthesis of meaning as ‘said’ meaning (truth-conditions) combined with ‘implied’ meaning (Gricean conversational inferenda). Kripke, playing Hegel to their Kant, takes the step beyond Grice and Davidson’s adherence to the austere idiom of an extensional, non-modal, first-order language by holding the thesis that identity statements in which the identity sign is flanked by names or variables are metaphysically necessarily true if they are true. Then, just as T.H. Green extended, corrected, and fulfilled Hegel, Soames extends, corrects, and fulfils Kripke’s picture of ‘theoretical identities’ as metaphysically necessary. Despite Soames’s differences with Rorty, their preoccupation with forms of scepticism about perceptual knowledge and about linguistic meaning remain common to their historiographies of the achievements of twentiethcentury analytical philosophy. Both books focus on the same preoccupation of philosophy 1930–1970: justifying forthright anti-scepticism about knowledge and meaning. In Soames’s history, Grice’s architectonic of Maxims of Conversation preserves truth-conditional meaning against Wittgensteinian scepticism about meaning (a scepticism that Kripke takes seriously in his Rules and Private Language (1982), and Kripke preserves knowledge of necessity against Quinean scepticism about modality. The justification of all this necessity – as Soames’s appeal to Grice’s reductio argument against Quine’s attack on analyticity suggests – is Grice’s earlier and Kripke’s later insistence on the strength of ordinary semantic intuitions about analytic statements and about judgments of synonymy, whether in
Paradigm Cases, Analytic-Synthetic, Essentialism
69
Grice’s defence of the analytic/synthetic distinction against Quine’s criticisms or in Kripke’s defence of essential properties.31 The deep irony is that careful examination of the linguistic evidence to which Kripke appeals in supporting his allegedly ‘intuitive’ claims about essence show that the linguistic intuitions actually do no such thing. Incorrect linguistic intuitions are no help in philosophy. Kripke has no defence of the intuitiveness of essential properties. The other irony is that if one takes seriously the linguistic intuitions about meaning that Grice does – intensions, not the extensions of Quine’s equivalence-class Meaning, Grice’s own reductio argument against Quine collapses. Grice’s, Strawson’s, and Soames’s reductio of White and Quine’s attack on synonymy is a fallacious, though ‘intuitively’ plausible, argument. But (a) the failure of Grice’s reductio argument against Quine, and (b) the presumption that there is an intuitively projectible linguistic property of ‘analytic’ sentences, do nothing either to subvert or to justify twentieth-century empiricism’s epistemological use of the analytic/synthetic distinction. We do not need – and we cannot get – necessity and synonymy to protect us against Scepticism’s Attack on Human Knowledge or prevent the Disappearance of the Human Mind.
Notes 1. I am grateful to Matthew Davidson, Al Martinich, Hilary Putnam, Tony Roy, Paul Saka, and Morton White for comments on earlier versions of this chapter. This chapter is dedicated to mathematical logician Dana S. Scott, a long overdue ‘thank you’ to my teacher and friend. A version of this chapter was given in the Philosophy Colloquium, California State University, San Bernardino, on 21 November 2008. 2. See pp. 10, 13, 20–1, 35ff., 105, 114ff., 118ff., 122ff., 132, 137ff., 149, 191ff. 3. These two footnotes were omitted from the version reprinted in the first edition of From a Logical Point of View in 1953 and from all subsequent editions. In the Preface to From a Logical Point of View, Quine acknowledges the stimulus of discussions with White and Nelson Goodman, as well as Carnap, Tarski, and Alonzo Church, dating from 1939, but he fails to mention the proximate intellectual origins of the attack on the analytic/synthetic in the summer of 1947, which is found in the Appendix, ‘Nelson Goodman, W. V. Quine, and Morton White: a triangular conference’, to White (1999). The three-way discussion of the arguments that led to ‘Two dogmas’ began with a letter of Morton White’s to Quine, which included the manuscript of White’s paper ‘On the Church-Frege solution of the paradox of analysis’ (1948). 4. Grice and Strawson (1956), reprinted in Harris and Severens (1970: see p. 58). See also H. P. Grice, ‘Reply to Richards’ in Grandy and Warner (1986; see especially p. 54). 5. See Harris and Severens (1970: 61). 6. See Harris and Severens (1970: 62). Note that in this claim, what I call ‘Reductio Version Two’, Grice and Strawson have moved from the language of ‘having a meaning’ or ‘meaning something’ to the language of ‘sentence significance’ or ‘having meaning’ – not the phrase ‘a meaning’. 7. Goodman (1983). 8. For a criticism of the Paradigm Case Argument see Passmore (1961: chapter 6); and for a defence, see Hanfling (1990–91). For a canonical use of the argument, see Strawson (1952: chapter 9).
70 Jay David Atlas 9. See Atlas (1989) for a discussion of the arguments among Quine, Russell and Strawson on truth-value gaps. 10. And here he mentions the admirable exception of Carnap in Meaning and Necessity (1956). 11. Here he refers the reader to the essay by Morton White that I mentioned in note 3. 12. The best interpretation of Quine’s attack on the analytic/synthetic distinction was, in my view, articulated by Hilary Putnam in ‘ “Two dogmas” revisited’ (1976/1983). Hilary Putnam (personal communication, 28 February 2006) recommends the discussion of the analytic/synthetic distinction by Charles Pigden (1987). Putnam treats Quine’s arguments as best understood as an attack on the epistemological distinction between a priori and a posteriori knowledge, not on the semantic distinction between analytic and synthetic sentences. As I was recently reminded by my student Gavin Davis, it is crucial to Putnam’s 1976 version of Quine’s argument that replacement of Classical Logic by Quantum Logic is a viable option for a revision of our web of belief. In post-1976 work Putnam himself gives up on the replacement; and Quine has been of two minds about the status of the meanings of the connectives of Classical Logic. These difficulties lay the argument for epistemological holism open to question, and to modification into Morton White’s corporatism; see Morton White (1998). In Toward Reunion in Philosophy (1956: 25) Morton White points out that it was James and Dewey who first clearly criticized G.E. Moore’s and B. Russell’s defence of the a priori/a posteriori knowledge distinction. This criticism by James, Dewey, White, and Putnam stands even if Quine’s arguments against the analytic/synthetic distinction are undermined by Soames’s or others’ criticism. The a priori/a posteriori knowledge distinction does not stand or fall with the analytic/synthetic distinction. [I defend the analytic/synthetic distinction in Atlas (2005:181–4) as a distinction required for adequate explanation in linguistic theory.] The criticism of the epistemological distinction is what William James, Pierre Duhem, John Dewey, Bertrand Russell (of Philosophical Essays , 1910), Alfred Tarski, L. Wittgenstein (of On Certainty, 1969), Morton White (of A Philosophy of Culture: the Scope of Holistic Pragmatism, 2002), Hilary Putnam (of Pragmatism, 1995), and, on Putnam’s view, Quine, were up to. 13. Benson Mates and Nelson Goodman argued that American English was such a language. 14. Some, though not I, would take an example of such a language to be {p, (p & p), (p & (p & p)), ... ,(p v p), (p v (p v p)), ... }. 15. It takes one to know one. When I was a 17-year-old, I believed Quine’s Second Dogma of Empiricism, the reduction of an empirical sentence to observation sentences as its empirical ‘meaning’; in fact, even worse, I was a phenomenalist, who in the fashion of Russell and Carnap wanted to construct physical objects from sense-data. 16. For Quine’s distinction between identity and individuation, see Quine (1985). 17. Noam Chomsky remarked (1996: 52) that among ‘the intrinsic semantic relations that seem well-established on empirical grounds are analytic connections among expressions’. See also Chomsky (2000: 46–7, 61–5). 18. The paragraph above summarizes the arguments of Atlas (2005: 181–4), reflecting on Atlas (1984). 19. How about Soames’s impossible world-states? Is there an impossible world-state in which Nixon exists and is, say, to use a famous example of Rogers Albritton’s a filling-station? Soames is committed (2003: 378, n.6) to Nixon’s being in one of these impossible world-states – or in The Great Impossible Pumpkin World-State, if there is only one. So why isn’t he a filling-station? If Nixon couldn’t be a filling-station
Paradigm Cases, Analytic-Synthetic, Essentialism
20.
21. 22. 23.
71
in an impossible world – in fact if he must be a human being (2003: 375) even with respect to that impossible world – one wonders what could logically limit the extension of an allegedly essential property like ‘humanity’ if contradiction or metaphysical impossibility couldn’t. Perhaps Soames is thinking of an ‘impossible world’ of the Kripkean sort, in which Nixon was not the son of Hannah Nixon but was still a human being necessarily. But as I understand Nathan Salmon’s views, in ‘The logic of what might have been’ (1989: 3–34), even if Nixon could not be a filling-station, the possibility of the possibility of his being a filling-station cannot be ruled out, since Salmon rejects C.I. Lewis’s system S4, and the transitivity of the accessibility relation, as characterizing necessity. (Hilary Putnam, in his ‘Is water necessarily H2O?’ [1990: 55–6 – see n.30 below], points out that Kripke has held of the Albritton cases [glass bottles might turn out to be organisms] and the Putnam case [tigers might turn out to be the glass bottles that are frightened] that what has been shown is not that it is conceivable that tigers are glass bottles but only that it is conceivable that it could become conceivable. Putnam notes that if we construe ‘conceivable’ as ‘possible to conceive’, then Kripke was denying the characteristic axiom of S4: ◊◊p → ◊p, for ‘conceivability’). But then why is humanity an essential property of Nixon? And how about a filling-station – call it ‘Bert’? Could it be a human being in that impossible world? (On Salmon’s view of an impossible world that would be possible were the actual world different, where the question whether a filling-station is a human being in that ‘contingently impossible’ world is the same question as for a possible world – the possible world merely gets called ‘impossible’ – the issue becomes less thrilling.) And Soames thinks that these are ‘modest’ theoretical commitments? Even if one takes seriously modal metaphysics, or technical notions of the designation by proper names of possible individuals, or virtual individuals in Dana Scott’s manner (in his ‘Advice on modal logic’, 1970) an essence cannot merely be any necessary property; see R. M. Adams, ‘Introducing note to *1970’ and K. Gödel, ‘Ontological proof (*1970)’ in Feferman et al. (1995: 388–402, 403–4). Hence it would be a bad idea to create the Person-Body Problem by a Kripkean argument of the sort Kripke used to resurrect the Mind-Body Problem in Naming and Necessity. Here’s the bad idea: Let’s stipulate that the matter of Nixon’s body (at a time and place) is to be named ‘Fred’. If Nixon = Fred, then ☐(Nixon = Fred). But, one intuitively thinks, surely there are some possible worlds in which Fred exists, in some fashion or other – let us suppose the Conservation of Mass/ Energy – but not as Nixon’s body. Nixon just does not have that body in those worlds. So, in conclusion, Nixon is not identical to Fred in the actual world. So Nixon is not identical to an actual aggregate of matter. He is not a material object. Hence he is a mental object, or at least he is a person, thought of as an independent ontological category. QED. If we make the usual philosophical correspondence of the semantic categories with the ontological ones, the linguistic version of the reasoning is: ‘Nixon’ is not [–ANIMATE]; hence ‘Nixon’ is [+ANIMATE]. But that, as I have suggested, is just fallacious linguistic reasoning. ‘Nixon’ is semantically non-specific; the noun is neither [–ANIMATE] nor [+ANIMATE]. The Kripkean identity-sentence argument will not get you the thrilling ontological conclusion. See note 30. See Atlas (2007: especially p. 218). See Nidditch (1980: 47–8). For an historically accurate and theoretically substantive interpretation of Locke’s Abstract Idea passage, see Atlas (1989: 10–16). Paul Benacerraf, Dana Scott, and Robert Breusch have encouraged my natural tendency to be ‘formal’, for which I am grateful.
72 Jay David Atlas 24. What Quine thinks is required for the modal interpretation of individual variables is just what Kripke adopts for the identity symbol flanked by variables or by individual constants – rigidly designating proper names, as one can see by substituting in Quine’s formula, letting ‘Fu’ =df ‘x = u’. 25. As noted above, what Quine thinks is required for the modal interpretation of individual variables is just what Kripke adopts for identity, as one can see by substituting in Quine’s formula, letting ‘Fu’ =df ‘x = u’. 26. Was Soames thinking that immortality is sufficient to make a being ‘divine’ rather than ‘human’ – ‘non-human’ is not a kind – so that I would change from one kind to another were I to become immortal? Even if I were to alter in kind, would that entail that I would not be Jay Atlas, and not correctly so named? Does that mean that Soames thinks that proper names are semantically relativized to kinds? (That would be reminiscent of Peter Geach on Identity and contrary to the official dogma on individual constants of both Quine and Kripke. For criticism of Geach, see ‘Identity’, chapter 1 in McGinn (2000: 1–14). 27. Alan R. White (1975) offers the further paraphrase ‘It was possible that this man should have lost the election’, but his suggestion raises many questions that I cannot pursue here, so I shall forbear. 28. There is a trivial objection to an unintended ambiguity in Kripke’s wording, but I shall mention it anyway. It is linguistically false that ‘when you ask whether it is necessary or contingent that Nixon won the election, you are asking the intuitive question whether in some counterfactual situation, this man would in fact have lost the election’. Who says that it is sufficient to determine either the necessity or the contingency that Nixon won the election that ‘This man would have lost the election’ or ‘It was possible for this man to have lost the election’ is to be evaluated only with respect to counterfactual situations? One would have to think that ‘would have’ entails ‘did not’ to suppose the worlds for evaluation are only counterfactual worlds. But it is linguistically obvious that an assertion of ‘He would have got in by the back door’ does not even conversationally implicate, much less entail, that he did not get in by the back door. 29. For further interesting discussion, see Alan R. White (1975). 30. To see just how problematic natural kind terms become, both semantically and scientifically, consider Atlas (1980), reprinted in Davidson (2007: 495–6), Donnellan (1983), Atlas (1989: 136–9), Katz (1990: 144), Donnellan (1993), Chomsky (2000: 164–94). For Hilary Putnam’s abandonment of metaphysical necessities in the case of water, see his ‘Is water necessarily H2O?’ in Putnam (1990: 54–79). 31. As I have discussed above, I have theoretical reasons for supporting an analytic/ synthetic distinction. I do not rely on merely ‘ordinary intuitions’. Putnam reads the defence of the distinction by Grice and Strawson in a similarly ‘theoretical’ way. In his famous paper ‘The analytic and the synthetic’ (Putnam 1975: 33–69) he remarks (ibid.: 35) that Strawson and Grice ‘offer theoretical reasons for supposing that the analytic-synthetic distinction does in fact exist, even if they do not very satisfactorily delinleate that distinction or shed much real light on its nature. Indeed, the argument used by them to the effect that where there is agreement on the use of the expressions involved with respect to an open class, there must necessarily be some kind of distinction present, seems to me correct and important. Perhaps this argument is the only one of any novelty to have appeared since Quine published his paper’. This comment by Putnam is what Grice noted in 1986 as what made his Paradigm Case Argument a ‘sophisticated’ one. As I have indicated in the text above, this Grice argument that Putnam admires is one that Soames, Harman,
Paradigm Cases, Analytic-Synthetic, Essentialism
73
and Grice himself have rejected. Soames does not address Quine’s metaphysical objection – what Quine called ‘The fallacy of subtraction’; see section 2, above.
References Adams, R.M. (1995) Introductory note to *1970. In S. Feferman et al. (eds.) Kurt Gödel, Collected Works, Volume 3: Unpublished Essays and Lecture (New York: Oxford University Press), pp. 388–402. Atlas, J.D. (1980) Reference, meaning, and translation. Philosophical Books, 21: 129–40. Atlas, J.D. (1984) Comparative adjectives and adverbials of degree: An introduction to radically radical pragmatics. Linguistics and Philosophy, 7: 347–77. Atlas, J.D. (1989) Philosophy Without Ambiguity (Oxford: Clarendon Press). Atlas, J.D. (2005) Logic, Meaning, and Conversation: Semantical Underdeterminacy, Implicature, and their Interface (New York: Oxford University Press). Atlas, J.D. (2007) Meanings, propositions, context, and semantical underdeterminacy. In G. Preyer and G. Peter (eds.) Context Sensitivity and Semantic Minimalism: New Essays on Semantics and Pragmatics (Oxford: Oxford University Press), pp. 217–39. Carnap, R. (1956) Meaning and Necessity, 2nd edn (Chicago: University of Chicago Press). Chomsky, N. (1996) Language and nature. In Powers and Prospects: Reflections on Human nature and the Social Order (Boston: South End Press), pp. 31–54. Chomsky, N. (2000) New Horizons in the Study of Language and Mind (Cambridge: Cambridge University Press). Davidson, M. (ed.) (2007) On Sense and Direct Reference (Boston: McGraw Hill). Donnellan, K. (1983) Kripke and Putnam on natural kind terms. In C. Ginet and S. Shoemaker (eds.) Knowledge and Mind (New York: Oxford University Press), pp. 84–104. Donnellan, K. (1993) There is a word for that kind of thing. In J. Tomberlin (ed.) Philosophical Perspectives, 7 (Atascadero, CA: Ridgeview Pub.), pp. 155–71. Fodor, J. and Lepore, E. (1992) Holism: A Shopper’s Guide (Oxford: Blackwell). Gödel, K. (1995) Ontological proof (*1970). In S. Feferman et al. (eds.) Kurt Gödel, Collected Works, Volume 3: Unpublished Essays and Lectures (New York: Oxford University Press). Goodman, N. (1983) Fact, Fiction, and Forecast (Cambridge, MA: Harvard University Press). Grice, H.P. (1986) Reply to Richards. In R. E. Grandy and R. Warner (eds.) Philosophical Grounds of Rationality: Intentions, Categories, Ends (Oxford: Clarendon Press), pp. 45–106. Grice, H.P. and Strawson, P. F. (1956) In defense of a dogma. Philosophical Review, 65: 141–58. Reprinted in Harris and Severens (1970), pp. 54–74. Hanfling, O. (1990/1991) What is wrong with the Paradigm Case argument? Proceedings of the Aristotelian Society, 91 (1990–1). Harris, J.F. Jr and Severens, R. F. (eds.) (1970) Analyticity (Chicago: Quadrangle Books). Katz, J. J. (1990) The Metaphysics of Meaning (Cambridge, MA: MIT Press). Kripke, S. (1972/1980) Naming and Necessity (Cambridge, MA: Harvard University Press). Kripke, S. (1982) Rules and Private Language (Oxford: Blackwell). Linsky, L. (ed.) (1952) Semantics and the Philosophy of Language (Urbana, IL: University of Illinois Press). McGinn, C. (2000) Logical Properties (Oxford: Clarendon Press). Nidditch, P.H. (ed.) (1980) Draft A of Locke’s Essay concerning human understanding. The earliest extant autograph version, transcribed with critical apparatus (Sheffield: Sheffield University Press).
74 Jay David Atlas Passmore, J. (1961) Philosophical Reasoning (London: Duckworth). Pigden, C. (1987) Two dogmatists. Inquiry, 30 (1): 173–93. Putnam, H. (1975) Mind, Knowledge, and Reality. Philosophical Papers, Volume 2. (Cambridge: Cambridge University Press). Putnam, H. (1976) ‘Two Dogmas’ revisited. In G. Ryle (ed.) Contemporary Aspects of Philosophy (Stocksfield: Oriel Press), pp. 202–13. Reprinted in H. Putnam (1983) Philosophical Papers, Volume 3: Realism and Reason (Cambridge: Cambridge University Press), pp. 87–97. Putnam, H. (1990) Realism with a Human Face (Cambridge, MA: Harvard University Press). Putnam, H. (1995) Pragmatism (Oxford: Blackwell). Quine, W.V.O. (1951) Two dogmas of empiricism. Philosophical Review, 60: 20–43. Quine, W.V.O. (1953) From a Logical Point of View (Cambridge, MA: Harvard University Press; Harper Torchbooks edition 1968). Quine, W.V.O. (1960) Word and Object (Cambridge, MA: MIT Press). Quine, W.V.O. (1969) Ontological Relativity and Other Essays (New York: Columbia University Press). Quine, W.V.O. (1985) Events and reification. In E. Lepore and B. McLaughlin (eds.) Actions and Events: Perspectives on the Philosophy of Donald Davidson (New York: Blackwell), pp. 162–71. Quine, W.V.O (1999) Where do we disagree? In L.E. Hahn (ed.) The Philosophy of Donald Davidson. (LaSalle, IL: Open Court), pp. 73–9. Rorty, R. (1979) Philosophy and the Mirror of Nature (Princeton, NJ: Princeton University Press). Russell, B. (1910) Philosophical Essays (London: Longmans Green). Salmon, N. (1989) The logic of what might have been. Philosophical Review, 98 (1): 3–34. Scott, D. (1970) Advice on modal logic. In K. Lambert (ed.) Philosophical Problems in Logic: Recent Developments (Dordrecht; Reidel), pp. 143–73. Soames, S. (2003) Philosophical Analysis in the Twentieth Century, Volume 2 (Princeton, NJ: Princeton University Press). Strawson, P.F. (1952) Introduction to Logical Theory (London: Methuen). Vendler, Z. (1967) Linguistics in Philosophy (Ithaca, NY: Cornell University Press). White, A.R. (1975) Modal Thinking (Ithaca, NY: Cornell University Press). White, M.G. (1948) On the Church-Frege solution to the paradox of analysis. Philosophy and Phenomenological Research, 9 (2): 305–8 White, M.G. (1950) The analytic and the synthetic: an untenable dualism. In S. Hook (ed.) John Dewey: Philosopher of Science and Freedom (New York: Dial Press), pp. 316–30. White, M.G. (1956) Toward Reunion in Philosophy (Cambridge, MA: Harvard University Press). White, M.G. (1998) Normative ethics, normative epistemology, and Quine’s holism. In L. E. Hahn and P. A. Schillp (eds.) The Philosophy of W.V.Quine, expanded edn (LaSalle, IL: Open Court), pp. 649–62. White, M.G. (1999) A Philosopher’s Story (University Park, PA: Pennsylvania State University Press). White, M.G. (2002) A Philosophy of Culture: The Scope of Holistic Pragmatism (Princeton, NJ: Princeton University Press). Wittgenstein, L. (1969) On Certainty (Über Gewissheit), ed. G.E.M. Anscombe and G.H. von Wright; transl. D. Paul and G.E.M. Anscombe (Oxford: Blackwell).
4 Grice on Presupposition Anne Bezuidenhout
Grice’s writings on the notion of presupposition are not extensive. Grice (1989: 269–82) contains an essay titled ‘Presupposition and conversational implicature’ that discusses some issues related to presupposition.1 In this essay, henceforth referred to as P&CI, Grice is concerned with the debate between Russell and Strawson as to whether the implications of existence and uniqueness associated with the use of definite descriptions in sentences of the form ‘The F is G’ are best thought of as presuppositions or as entailments. Since Grice sides with Russell, he is inclined to say that they are entailments.2 But then he has to explain why those implications survive even in negative sentences of the form ‘The F is not G’, which do not entail the existence of a unique F. Grice accepts that these implications do survive, but he argues that they are conversational implicatures rather than Strawsonian presuppositions. In the first section, I lay out Grice’s arguments for this conclusion. In the following section, I argue that all of Grice’s observations can be accommodated within a presuppositionalist framework. However, we will see that there are two rather different ways of characterizing Grice’s insights in presuppositionalist terms. I will give reasons for favouring one of these over the other. It is one that treats definite NPs as ‘soft’ presupposition triggers. In the final section, I give some reasons for thinking that the presupposition account advocated in section 2 is to be preferred to Grice’s implicature account.
1. Grice’s implicature account Despite the fact that Grice had earlier collaborated with Strawson (e.g., on their 1956 Philosophical Review article on Quine’s two dogmas of empiricism), in P&CI he is inclined to come down on Russell’s side in the RussellStrawson debate about definite descriptions. He wants to accept the Russellian quantificational analysis of sentences of the form ‘The F is G’, according to which such sentences entail that there is a unique F. Strawson on the other hand regards definite descriptions as singular referring expressions that are 75
76 Anne Bezuidenhout
associated with existence and uniqueness presuppositions. On his account, the felicitous use of ‘the F’ requires that there be a unique F in the conversational context. If there is not a unique F in the context, then the assertive enterprise fails, and it is not possible to use the sentence ‘The F is G’ to make a truth-evaluable claim. One consideration apparently in favour of Strawson’s view is that the negative sentence ‘The F is not G’ seems to carry the same presuppositions as the positive sentence ‘The F is G’. Thus Strawson would say that in the negative case, if there is no unique F in the conversational context, then once again the assertive enterprise fails, and one cannot make a truth-evaluable claim by uttering ‘The F is not G’. Now Grice cannot, in the negative case, account for the implications of existence and uniqueness as entailments, as he can in the positive case – at least not as long as negation is treated as semantically unambiguous and as corresponding to wide-scope sentence negation or to what Horn (1992: 262) calls contradictory negation. So Grice argues that in the negative case the implication of unique existence is a conversational implicature. 3 He goes on to support this claim by showing that this implication has all the hallmarks of a conversational implicature, namely it is non-detachable, cancellable, and calculable. To support the claim that the unique existence implication is non-detachable, Grice notes that this implication remains even if we use close paraphrases of the ‘The F is not G’ form. For instance, all the following forms seem to carry the same implication of the existence of a unique F: It is not the case that the F is G. It is false that the F is G. It is not true that the F is G. Grice claims that the only way to detach this implication is to explicitly use the Russellian expansion ‘It is not the case that there is an F, there is only one F and all Fs are G’. As he remarks, so long as the surface form uses the definite article, the implication of unique existence seems to be triggered (P&CI: 271). To support the claim that the existence assumption is cancellable, Grice imagines a situation in which there is some dispute as to whether or not the British government employs a covert agent – a loyalty examiner – whose job it is to investigate citizens to make sure they are loyal to the government. If Paul is well known to his friends to be sceptical of the existence of such a person, but one of his friends is not so sure and is worrying about being summoned before the loyalty examiner to give an account of himself, Paul can reassure his friend by uttering (1): (1) Don’t worry, the loyalty examiner won’t summon you.
Grice on Presupposition
77
Grice thinks that it is clear in this situation that Paul is not assuming the existence of a loyalty examiner, and in fact his grounds for uttering (1) are precisely that he doesn’t believe in the existence of such a person. Similarly, against a background where it is known that Paul is sceptical of the existence of such a loyalty examiner, Paul can say to an obviously loyal person: (2) Well, the loyalty examiner will not be summoning you at any rate. Again, Paul will not be taken to imply the existence of a loyalty examiner by his utterance of (2). (Grice doesn’t give an example to establish that the uniqueness implication can be cancelled, but presumably he does think that this implication could be the specific target of cancellation, just as the existence implication can be.) It is worth noting here that if there is indeed implicature cancellation going on in these cases, it is what Grice (1989: 39) calls contextual cancellation. That is, the implicature is cancelled because it is inconsistent with common ground information. To support the claim that the implication of unique existence is calculable, Grice invokes his conversational maxim of Manner, and proposes to refine it by adding the following sub-maxim, which a little later (P&CI: 276) he calls the conversational ‘tailoring’ principle: Conversational Tailoring (CT): ‘Frame whatever you say in the form most suitable for any reply that would be regarded as appropriate’ or ‘Facilitate in your form of expression the appropriate reply’. (P&CI: 273) Grice’s reasoning in support of the calculability of the implication of unique existence is as follows: First, he assumes the correctness of the Russellian analysis according to which a sentence such as ‘The king of France is bald’ is a ‘definitional contraction’ (P&CI: 272) of There is at least one king of France, there is no more than one king of France, and nothing which is a king of France is not bald. He favours this expansion because ‘it sets out separately three distinct clauses, and each one of these can be false while both the others are true’ (P&CI: 273).4 Now, when someone makes a complex, conjunctive assertion, the CT principle instructs the person to pick a form of expression that makes it clear to his interlocutors what aspect of his conjunctive claim he takes to be most likely to be challenged. That is to say, the speaker must decide which of the conjuncts (if any) has what Grice calls ‘common-ground status’ (P&CI: 274) and thus is unlikely to be challenged by his interlocutors. Then the speaker should pick a form of expression that highlights the conjunct(s) that may be challenged. This is what happens when someone chooses to express his assertion using the contracted form ‘The F is G’ rather than the Russellian expanded form. The contracted form indicates that the speaker thinks the existence of a unique F is part of the common ground, and that what is new (and potentially contentious) information is the information
78
Anne Bezuidenhout
contained in the third conjunct of the Russellian expansion (and in the predicative part of the contacted form).5 All of this is something that the hearer can be expected to work out, which shows that the implicature of unique existence is calculable. In the negative case, the speaker will be construed as if he were responding negatively to someone who had uttered ‘The F is G’, and thus again the hearer will understand the speaker to be treating the assumption of the existence of a unique F as having common-ground status. Grice goes on to add a clarification as to what it means for something to have common-ground status. He notes that it would be perfectly natural for someone when discussing a certain concert with his friends to utter sentence (3) below, without assuming that his interlocutors know that he has an aunt, let alone that his aunt has a cousin (P&CI: 274): (3) My aunt’s cousin went to that concert. By the reasoning we’ve just been considering, Grice would have to say that the speaker of (3), by opting for the contacted form (3), is taking the existence of his aunt’s cousin as having common-ground status.6 If something’s having common-ground status was taken to mean that it was common knowledge, then taking the existence assumption as having commonground status would conflict with the obvious acceptability of (3) in a context where the existence assumption is not already common knowledge. So Grice notes that the ‘supposition must not be that it is common knowledge but rather that it is noncontroversial, in the sense that it is something that we would expect the hearer to take from us (if he does not already know)’ (P&CI: 274). These then are Grice’s arguments for claiming that the implication of unique existence is non-detachable, cancellable and calculable, and hence that this implication is a conversational implicature of utterances of sentences of the form ‘The F is not G’.
2. Reframing Grice’s observations in a presuppositionalist framework The lecture on which P&CI was based was given in 1970. Those familiar with the subsequent literature on presupposition will recognize that all of Grice’s arguments have echoes in this literature. In this section I will try to show how Grice’s observations can be handled in a presuppositionalist framework. 2.1 Non-detachability redescribed Firstly, with respect to his argument about non-detachability, Grice concedes that the appearance of the definite article in the surface form is what seems
Grice on Presupposition
79
to trigger the implication of unique existence. This is certainly compatible with thinking of ‘the F’ as a presupposition trigger. It was precisely this sort of regular behaviour that led people initially to suggest that definite descriptions are one of a variety of expression-types (along with factive verbs such as ‘know’, change of state verbs such as ‘stop’, adverbs such as ‘too’, etc.) that carry semantic presuppositions. Of course it is not solely because the implication of unique existence survives in all the variant forms of ‘The F is not G’ that makes it a presupposition. It is also the fact that the implication survives when the ‘The F is G’ is embedded under a variety of sentence operators besides negation. For example, ‘It is possible that the F is G’, ‘If the F is G, then H’ and ‘Is the F a G?’ all carry the same implication of the existence of a unique F. Chierchia and McConnell-Ginet (1990: 24) introduce the phrase ‘the S family’, for the variants of a positive indicative sentence S that are obtained by embedding S under such sentence operators. Thus we can say that any presuppositions of S are also presuppositions of its S-family.7 Grice presumably would not be happy to class the implications associated with the use of the definite article as semantic presuppositions, precisely because he was attempting to give a pragmatic explanation in opposition to what he took to be Strawson’s view.8 However, within the presuppositionalist framework, it has recently been suggested, by Abbott (2005), Abusch (2002) and Simons (2001) amongst others, that the expression-types that are presupposition triggers are of two sorts, hard versus soft triggers.9 Hard triggers are expression types that are invariably associated with certain presuppositions. On the other hand, the presuppositions associated with soft triggers are such that they can be ‘neutralized’, to use Abbott’s (2005) term. That is to say, in certain conversational contexts, the associated presupposition goes away. (Note also that neutralization is not the same as presupposition suspension, cancellation or filtering. See Abbott 2005: 6–8. I’ll return to this point in section 2.3 below). For example, factive verbs are presupposition triggers, which presuppose their complements. Thus, an out-of-the-blue assertion of (4) presupposes that Sigrun is in NYC: (4)
If Barry discovers that Sigrun is in NYC, he’ll want to see her.
However, cognitive factives such as ‘discover’ are soft triggers, since their presuppositions disappear in what Simons (2001: section 1.1) calls explicit ignorance contexts. So, (5) does not presuppose that Sigrun is in NYC: (5) I don’t know whether Sigrun is in NYC or not, but if Barry discovers that she is, he’ll want to see her. Soft triggers are such that pragmatic factors influence their behaviour. Thus if definite descriptions are soft triggers, we can offer an account within the
80
Anne Bezuidenhout
presuppositionalist framework that satisfies Grice’s desire to offer a pragmatic rather than a purely semantic explanation for the implications associated with definite descriptions. 2.2 Cancellability redescribed This issue of hard versus soft presupposition triggers is relevant also to the second strand of Grice’s argument. He argues that the implication of unique existence associated with the definite description in ‘The F is not G’ is a conversational implicature because it can be (implicitly) cancelled. But once we have admitted that there can be soft presupposition triggers, we have in effect allowed that presuppositions are defeasible, just as Grice takes conversational implicatures to be. Thus the behaviour of ‘the loyalty examiner’ in the contexts Grice imagines for (1) and (2) above isn’t by itself a conclusive reason for treating the implication of unique existence as an implicature. The contexts that Grice imagines for (1) and (2) could be construed as neutralizing contexts, similar to the contexts of ignorance that Simons (2001) argues neutralize the presuppositions associated with cognitive factives such as ‘discover’. This suggests that we can treat definite descriptions as soft presupposition triggers. But is it really plausible to think of definite descriptions as soft triggers? Abbott (2005), who is in favour of the idea that some presupposition triggers are soft triggers, claims that definite descriptions belong to the class of hard triggers. Firstly, she notes that, when functioning as predicate nominals, definite descriptions often seem to lack presuppositions. One might be tempted to use this in support of the idea that definite descriptions are soft triggers. Abbott offers the following pair of examples from Donnellan (1966: 284) as an illustration: (6) a. Is the king of France de Gaulle? b. Is de Gaulle the king of France? Example (6a) presupposes that there is a king of France, while (6b) could have been used simply to inquire whether France is a monarchy. As Abbott notes, several authors have claimed that it is the topic-focus structure of sentences that determines what is presupposed and what is not. With respect to definite descriptions, the idea would be that definite NPs in topic position carry presuppositions, while definite NPs in focus position do not. This topic-focus view is often, incorrectly in my view (see Bezuidenhout forthcoming), attributed to Strawson. It is discussed by Atlas (2004) and von Fintel (2004). The former tries to salvage some aspects of the view whereas the latter argues that it is mistaken. Presumably because of the contentiousness of this data, Abbott thinks it cannot be relied on to argue for the claim that definite descriptions are soft triggers.
Grice on Presupposition
81
On balance, Abbott is inclined to think that definite descriptions are hard triggers and she offers the following example in support of this claim (2005: 10): (7) # Possibly no one owns this book, but if I find the owner I will return it. The definite description ‘the owner’ in (7) does seem to fail the test that Abbott applies to see whether a presupposition can be neutralized. The possibility of the non-existence of an owner of the book in question is explicitly raised, but this does not seem to lift the presupposition associated with the ‘the owner’. Abbott’s intuitions are that the use of the description in this example is unacceptable, hence her use of the hatch mark ‘#’. My intuitions are not as robust as Abbott’s, since I think (7) might be acceptable. Certainly the following variant of (7) seems quite a lot better: (8)
I do not know whether or not anyone owns this book. However, if I ever find the owner I will return it.
Example (8) can be made even better if one imagines the following context. Suppose I find a copy of Pride and Prejudice tucked away in a corner of my attic that has every appearance of having been abandoned and so I start to carry it around and to use it. My friend notices its lovely cover asks me where I got it and I tell him the story of its discovery. He begins to berate me for assuming that such a valuable book is ownerless and suggests that I was wrong to use it without permission. In such a situation I might very well utter (8) and my interlocutor would know that I am not implying the existence of a unique owner of the book by my use of ‘the owner’. If anything, my interlocutor would know that I lean towards the view that there isn’t an owner. In examples (1) and (8) the definite NP is embedded under negation and in the antecedent of a conditional respectively. I would argue that neutralization is possible also where the definite NP occurs under question-formation. Consider the following example: (9) Is it such a good idea to ask his aunt for a reference? It is true that an out-of-the-blue utterance of (9) will presuppose that the person in question has an aunt. But consider the following scenario: We are trying to decide whom to ask for a job reference for John and someone suggests we ask his aunt. I might respond to this suggestion by uttering the following: (10)
I do not know whether or not John has an aunt. But is it such a good idea to ask his aunt for a reference? References from family members are not really appropriate.
82 Anne Bezuidenhout
The presupposition of existence associated with ‘his aunt’ has been neutralized in (10). As I will suggest in section 2.3 below, this is because the presupposition has been taken out of play. The speaker’s main point is premised on the trait or characteristic of being an aunt rather than on the existence of an individual aunt. I have been trying to clear the ground for the claim that Grice’s ‘loyalty examiner’ examples are cases of presupposition neutralization. However, one might argue that in example (1) the existence presupposition in fact survives, but that it is implicitly restricted to the fearful friend’s belief world. So, in uttering (1), Paul is tacitly conveying something like (11) below: (11)
Suppose that as in your belief-world there is a loyalty examiner i. Don’t worry, hei won’t summon you.
In this case it will still be clear that Paul’s grounds for uttering (1) are that he believes that the loyalty examiner is a figment of his friend’s imagination. On this way of thinking of things, definite NPs are hard triggers, but their presuppositions are sometimes accommodated within the scope of a (possibly implicit) belief operator and so not admitted to the main context (i.e., the common ground shared by the conversational participants).10 This idea that presuppositions are not always accommodated in the global context but are sometimes locally accommodated under some intensional operator, like a belief or modal operator, is more or less the standard view of matters amongst formal semanticists, especially those who accept something like Kamp’s discourse representation theory (DRT) or who advocate dynamic semantic frameworks like Heim’s file change semantics (FCS). See Asher and Lascarides (1998), Beaver (2001), Geurts (1996, 2000), Roberts (1996, 2004), Zeevat (1992). For example, Geurts (2000) articulates a principle that he calls the ‘Buoyancy Principle’ (BP), according to which presuppositions (and backgrounded assumptions more generally) tend to float up to the main discourse representation structure (DRS). That is to say, presuppositions triggered within some embedded context tend to be accommodated in the highest context possible. If they cannot be accommodated in the global context, they will be accommodated in the next highest context, and so on down to the most nested one. The BP has the virtue of being able to explain certain cases of intermediate accommodation. See Geurts (2000: 318, 325). Consider example (12a) below: (12) a. Most teachers spoil their children. b. Most teachers who have children spoil them. c. Most teachers have children and spoil them. The definite NP ‘their children’ in the nuclear scope of the quantifier in (12a) presupposes that the teachers in question have children. This assumption
Grice on Presupposition
83
cannot be accommodated in the global context (since it would then contain an unbound variable), and hence the BP predicts that it will be accommodated in the restriction of the quantifier, yielding the reading (12b). This is a case of intermediate accommodation. Local accommodation (accommodation in the DRS in which it is triggered) would yield the reading (12c), which is not the preferred reading of (12a). 2.3 Two competing presuppositionalist hypotheses So, within the presuppositionalist framework, we have two competing accounts of the status of definite descriptions as presupposition triggers: (i) Definite NPs are soft triggers and the presupposition of unique existence associated with such NPs is neutralized in certain contexts – especially contexts of explicit or implicit ignorance and/or indifference. (ii) Definite NPs are hard triggers but the presupposition of unique existence associated with such NPs is sometimes locally accommodated rather than admitted to the global context shared by the conversational participants. To decide between (i) and (ii) we would need to examine the notion of neutralization more carefully. I mentioned in section 2.1 that Abbott (2005) wants to distinguish presupposition neutralization from presupposition suspension, cancellation and filtering. It is worth briefly looking at this fourway distinction. Abbott distinguishes between cancellation and suspension on the grounds that in the former case the presupposition is denied, whereas in the latter, ‘its truth is left hanging’ (2005: 6). She offers the following example from Horn (1972) as an illustration of suspension: (13)
The milk train doesn’t stop here anymore, if it ever did.
The claim that the milk train no longer stops at the place in question presupposes that it used to stop there. However, the speaker goes on to cast doubt on this assumption, while not rejecting it outright. In the case of cancellation on the other hand, the speaker adds information to the context that explicitly signals that he doesn’t accept the presupposition. That is arguably what happens in (14) below (my own made-up example): (14)
Speaker A: The man at the ticket counter was so rude. Speaker B: No he wasn’t. That was a woman, not a man.
In this case speaker B’s rejection of A’s assertion is based on B’s rejection of the existence implication associated with the definite description.11 We have already seen that the presupposition of unique existence associated with ‘The F is G’ survives when this sentence is embedded under
84 Anne Bezuidenhout
negation and possibility operators and in the antecedents of conditionals, and it survives under question-formation. Such embedding constructions are presupposition ‘holes’, in the sense that presuppositions escape their scope. Other embedding constructions are presupposition ‘filters’, in the sense that the presuppositions triggered inside such constructions sometimes project and sometimes do not. That is, those presuppositions sometimes fall outside the scope of such constructions and sometimes do not. This is illustrated in (15) and (16) below (again my examples): (15) (16)
If it is ten years since I last saw John, John’s children must be grown up by now. If John has children, John’s children must be grown up by now.
In the case of (15), the presupposition that John has children triggered by the definite NP in the consequent of the conditional projects out of the consequent and is inherited by the compound sentence. This is not so in the case of (16). The presupposition is said to have been filtered out, because it is explicitly presented in the antecedent.12 Abbott distinguishes between presupposition filtering cases and presupposition suspension cases as follows: ‘Regular antecedent if clauses [such as in (16)] present hypothetical situations ... they introduce presuppositions of the consequent clauses for temporary consideration. However the suspender if clauses [e.g., the ‘if it ever did’ in (13)] suggest something about what actually is the case – namely, that the presupposition in question may not hold’ (Abbott 2005: 7–8).13 Thus we can summarize Abbott’s four-way distinction between types of presupposition modulation as in table 4.1. Given the way I have presented this four-way distinction in table 4.1, the line between suspension and neutralization may seem rather fine. Nevertheless, I believe it is a real and important distinction. I would say that the difference comes down to whether or not the presupposition is left in play in the conversation. I suggest that in the neutralization case the presupposition is taken Table 4.1 Types of presupposition modulation Type of modulation
Type of speaker commitment
Cancellation
Presupposition rejected by speaker
Suspension
Truth of presupposition left hanging by speaker
Filtering
Presupposition presented for temporary consideration by speaker
Neutralization
Presupposition lifted by speaker’s explicit or implicit avowal of ignorance and/or indifference
Grice on Presupposition
85
out of play, and not by being denied (as in the cancellation case) but by the speaker’s explicitly or implicitly indicating that his assertion is not premised on a commitment to the assumption’s truth. In the case of suspension on the other hand, the presupposition is left in play but the speaker indicates that he doesn’t want to be held accountable for its truth – he wants in some sense to be held blameless if it turns out to be false. To make it a bit clearer what I mean when I say that neutralization involves the speaker’s indicating that his assertion is not premised on a commitment to the assumption’s truth, consider again example (5), repeated below: (5)
I don’t know whether Sigrun is in NYC or not, but if Barry discovers that she is, he’ll want to see her.
By prefacing my remark with an explicit avowal of ignorance, I make it clear that the truth of the complement of the verb ‘discover’ is not relevant to my remark and thus needs to be taken out of play. Instead I am more interested in conveying something to you about Barry’s mental states and his dispositions to action. Thus what you are to take away from my utterance of (5) is something about Barry’s relationship to Sigrun. You should not take my remark to indicate anything about the way the world is or about how the world might impact on Barry.14 Remember, we are trying to decide between the two opposing presuppositionalist hypotheses (i) and (ii), repeated below: (i) Definite NPs are soft triggers and the presupposition of unique existence associated with such NPs is neutralized in certain contexts – especially contexts of explicit or implicit ignorance and/or indifference. (ii) Definite NPs are hard triggers but the presupposition of unique existence associated with such NPs is sometimes locally accommodated rather than admitted to the global context shared by the conversational participants. I think that the distinction between presupposition neutralization and presupposition filtering is most relevant to deciding this matter. Hypothesis (i) is clearly committed to the idea of presupposition neutralization. Hypothesis (ii) is best cashed out in terms of the notion of presupposition filtering. I say this because the analysis of example (1) that treats ‘the loyalty examiner’ as a hard trigger assumes that the utterance of (1) implicitly invokes the fearful friend’s belief world in which the loyalty examiner exists. That is, the presupposition of existence is accommodated within the scope of an implicit belief operator. One can think of this as a case of filtering in Abbott’s sense, according to which the speaker presents a presupposition for temporary consideration. We are to pretend that things are as the friend believes: ‘Suppose as you believe that there is a loyalty examiner i’. Then we
86
Anne Bezuidenhout
can talk about the things that populate that world: ‘Don’t worry, hei won’t summon you’. Thus we want in effect to know whether example (1) is a case where the presupposition is taken out of play, as in example (5) (‘I don’t know whether Sigrun is in NYC or not, but if Barry discovers that she is, he’ll want to see her’), or whether it is one in which the presupposition is presented for temporary consideration, as in example (16) (‘If John has children, John’s children must be grown up by now’). If we treat (1) as a case in which the existence presupposition is taken out of play, this would mean that the speaker wants somehow to be taken to be making a claim that has nothing to do with whether or not there is a loyalty examiner. The relevance of the speaker’s utterance would have to lie elsewhere, for example in its being produced with the intention to comfort the fearful friend. The problem with this is that Grice thinks that the basis for the speaker’s utterance in this case is precisely the speaker’s belief that the actual world is a world in which there is no loyalty examiner. On the other hand, it seems equally problematic to treat (1) as a case of pretence or intensional subordination. This is because if the speaker is temporarily putting himself in the fearful friend’s shoes, it is no longer clear what his grounds for reassurance are. In the fearful friend’s belief world there are presumably grounds for being fearful, and thus the worries about being summoned before the loyalty examiner cannot be dismissed without addressing those grounds directly. Some readers may be wondering at this point whether it wouldn’t be best to treat example (1) as a case of presupposition cancellation in which the presupposition is rejected by the speaker, as in example (14) (‘The man at the ticket counter wasn’t rude; that was a woman, not a man.’). Certainly this would be a way of treating example (1) in the presuppositionalist framework that (superficially) most parallels Grice’s treatment of example (1) in his implicature framework. Nevertheless, this suggestion has problems too. Presupposition-cancelling negation is taken by Horn (1972) and Abbott (2005) to be metalinguistic and by Carston (2002) to involve an echoic element. This metalinguistic or echoic element is relatively clear in example (14). One can easily understand the speaker to be objecting to his interlocutor’s use of the word ‘man’. In spoken discourse the speaker would likely make this clear by putting special stress on that word. Moreover, the speaker isn’t necessarily disagreeing with his interlocutor’s assessment of the ticket seller’s rudeness. The dialogue in (14) could very well have gone as in (17) below: (17)
Speaker A: The man at the ticket counter was so rude. Speaker B: Yes he was. But ‘he’ wasn’t a man; ‘he’ was a woman.
In the case of (1), the element within the scope of the metalinguistic negation would have to be the entire sentence ‘The loyalty examiner will
Grice on Presupposition
87
summon you’. The speaker would in effect be saying that this sentence couldn’t be felicitously used to make an assertion. This is more or less the standard Strawsonian view of what happens when the assumption of unique existence associated with the definite description is false. My objection to assimilating example (1) to standard cases of presupposition failure is that this makes the speaker’s words ‘Don’t worry’ seem rather anomalous. The speaker would be communicating that his interlocutor shouldn’t worry because the interlocutor’s words cannot be used to make an assertion. But that sounds precisely like something the interlocutor should worry about! On the other hand, Grice’s use of the phrase ‘Don’t worry’ in (1) sounds quite natural to my ears.15 Thus on balance I am inclined to come down in favour of the neutralization explanation of example (1). The speaker is taking the existence assumption out of play and doing something else with his utterance – presumably trying to comfort and reassure his friend. A similar explanation could be given in the case of example (2) (‘Well, the loyalty examiner will not be summoning you at any rate’, said to an obviously loyal friend). The existence assumption is being taken out of play and the speaker is instead attempting to communicate something about his belief in the friend’s loyalty and trustworthiness. I will return to this issue in section 3 below. But I conclude tentatively that if the neutralization explanation works, then since it is one that makes definite NPs into soft triggers, we would have a suitably ‘pragmatic’ rival to Grice’s implicature account of the implication of unique existence associated with definite NPs. 2.4 Calculability redescribed I take Grice’s calculability requirement to be a rather general requirement on any account that invokes pragmatically derived meanings. Grice was rightly worried that when a theorist appeals to dimensions of meaning that go beyond what is semantically encoded, it may seem mysterious how people could grasp such meanings. So the theorist who invokes such meanings owes us at least a sketch of how the audience is able to work out what a speaker is communicating when what is communicated goes beyond what lies on the surface. Since I want to treat definite NPs as soft presupposition triggers that are influenced by pragmatic factors, I do not want to explain away calculability. Rather, I want to show that the apparatus that Grice invokes (and in particular his ‘conversational tailoring principle’) can be accommodated in a presuppositionalist framework. Note that if one holds that definite NPs are hard presupposition triggers, then the calculability constraint will seem less pressing, precisely because one will hold that presupposition recovery is driven by semantic mechanisms. Of course, one is still obliged to give an account of these semantic mechanisms, but presumably whatever account is given will not invoke processes that are sensitive to wide pragmatic factors. For example, Burton-Roberts (1989)
88 Anne Bezuidenhout
argues that the presupposition of unique existence associated with the definite NP in ‘The F is not G’ is a default semantic inference that arises due to an interaction of the semantics of negation with what he calls the ‘cognitive principle of bivalence’ (CPB). The CPB states that human thinking is inherently bivalent; it operates in accordance with opposing categories of truth versus falsity, affirmation versus denial, etc. Semantically, the negation operator is unambiguous and neither cancels nor preserves the existence presupposition that is triggered within its scope.16 However, the CPB will demand that the status of this assumption be resolved. Since it occurs in a negative environment but is not itself explicitly denied, it will be treated by default as affirmed. (This account would have to be generalized to handle presuppositions triggered in environments other than negative ones, but I set that problem aside here.) I assume that other accounts that treat definite NPs as hard triggers would appeal to a similar interaction between semantically encoded meanings and cognitive principles. For example, the sort of account offered by Geurts within the DRT framework appeals to the ‘buoyancy principle’ (see section 2.2 above) to explain how and when presuppositions are either globally or locally accommodated. I assume the BP is meant to operate in an automatic manner, in the sense that global accommodation is the default. Certain attempts at global accommodation would be ruled out simply because they resulted in ill-formed DRSs, as is the case with the attempt to globally accommodate the existence presupposition in (12a). However, wide pragmatic factors may still have some role to play in this DRT framework, since there will be cases in which global accommodation is ruled out by information in the conversational common ground and not because of structural constraints on DRSs. Inasmuch as this is so, it will be important to give some account of the pragmatic principles that constrain presupposition accommodation. Grice appeals to his principle of conversational tailoring (CT) to explain how hearers are able to figure out that a speaker is conversationally implicating that there is a unique F when he utters the sentence ‘The F is not G’. The CT principle constrains a speaker to choose a form of expression that focuses his listeners on those aspects of his claim that he deems the most likely to be challenged. Those aspects that are deemed to have commonground status will not be emphasized. For information to have commonground status in Grice’s sense is just for it to be non-controversial, i.e., to be information that the listener will accept on the speaker’s say-so, without challenge. It needn’t be information that the hearer already possesses. Thus when a speaker chooses the contracted form ‘The F is not G’ as opposed to the full Russellian expansion, he implicates that there is a unique F, since the contracted form in effect treats the implication of unique existence as having common-ground status. This implication thus escapes the scope of the negation.
Grice on Presupposition
89
Abbott (2000) offers an account of presuppositions that treats them as background propositions. I believe that Abbott’s account has many parallels with Grice’s CT account. By adopting Abbott’s account of presuppositions as backgrounded information, we can recast some of Grice’s insights in a presuppositional framework. One of Abbott’s (2000) motivations is to account for cases of so-called informative presuppositions without appealing to what she takes to be a problematic (because completely unconstrained) notion of presupposition accommodation. In the presupposition literature, examples like Grice’s in which the speaker uses the definite NP ‘my aunt’s cousin’ in a context where it is not common knowledge that the speaker has an aunt or that the aunt has a cousin – see example (3) above – have been treated as involving informative presuppositions. Informative presuppositions will seem problematic to anyone who defines a (pragmatic) presupposition as something that is assumed to belong to the conversational common ground prior to a speaker’s utterance.17 Yet informative presuppositions occur frequently in ordinary discourse. See the frequency data cited by Abbott (2000: 1428). Lewis (1979: 340) proposed a ‘principle of accommodation’ which says: ‘If at time t something is said that requires presupposition P to be acceptable, and if P is not presupposed just before t, then – ceteris paribus and within certain limits – presupposition P comes into existence at t’. Abbott (2000) points out that Lewis does not do anything to clarify what these ‘certain limits’ are. She also argues that subsequent attempts to spell out such limits, such as Heim’s (1982) account that appeals to the idea of bridging, are less than satisfactory. Grice’s notion of non-controversiality is also frequently appealed to as a criterion for accommodation; see Soames (1982). Those who appeal to non-controversiality in this context typically contrast definite NPs like ‘my pet hamster’ with ones like ‘my pet giraffe’. Since hamsters are stereotypical pets and giraffes are not, one is much more likely to balk at accommodating the assumption that the speaker has a pet giraffe. Remember however that Grice introduces the notion of non-controversiality to explain what it means to treat something as having common-ground status. He does not invoke non-controversiality as a criterion for when information can be accommodated. For Grice there is no problem of informative presuppositions that needs to be solved. The existence implication associated with ‘my aunt’s cousin’ is for Grice an entailment of what is asserted if the definite phrase (DP) occurs in a positive environment, and an implicature of what is asserted if the DP occurs in a negative environment. Either way, Grice would be happy to allow that the existence assumption may very well be new information. What Grice is concerned with is what might be called the problem of information packaging. In cases where what we are attempting to convey is complex, there are different ways in which we can package the information that we are trying to convey. Grice’s CT principle can be thought of as
90 Anne Bezuidenhout
his attempt to explain the best way to package information. The principle instructs the speaker to choose a form of expression that focuses on the controversial aspects of what he wants to convey and puts the non-controversial aspects in the background. Backgrounded information is information the hearer will likely accept just on the speaker’s say-so, and hence one makes it easier for one’s interlocutors if one foregrounds those aspects they may want to challenge. Grice’s talk of controversiality and of what may be challenged should not, I believe, be taken to imply that he thinks of ordinary conversations as adversarial events. For example, a lecturer talking to students is just as bound by the CT principle as is someone on the debate floor, even if the students are not disposed to question what the lecturer says. In fact one might say that the lecturer is especially bound by the CT, because a good conversationalist makes it easy for his interlocutors to focus on what is important or relevant in what he has said – to focus on what is most likely to move the conversation forward. Abbott’s (2000) account of presuppositions as background propositions clearly shares many features with Grice’s account of conversational tailoring. Like Grice and unlike many presuppositionalists, Abbott is not bothered by the fact that the implication of unique existence triggered by a definite NP may be new information to the hearer. And like Grice (although presumably for slightly different reasons), she does not see the need to invoke a notion of accommodation. Finally, like Grice, she is concerned with issues having to do with information packaging. She writes: ‘I propose that grammatical presuppositions are a consequence of a natural limit on how much can be asserted in any given utterance, where what is asserted is what is presented as the main point of the utterance – what the speaker is going on record as contributing to the discourse’ (2000: 1431). Abbott does want to say that there are limits to the amount and kind of new information that can be backgrounded. She writes: ‘propositions that are believed by the speaker to be new information to the addressee should not be presupposed if they should be asserted instead’ (2000: 1433, her emphasis). This by itself seems rather tautological. However, she goes on to mention Grice’s notion of non-controversiality (which she treats as more or less synonymous with the notions of expectedness and/or nonnewsworthiness). To reiterate, non-controversiality is not a constraint on presupposition for Grice but a constraint on what has common-ground status. Nevertheless Grice’s non-controversiality constraint can be construed as a criterion for what can be backgrounded, and hence it is appropriate for Abbott to appeal to this notion, since she defines presuppositions as backgrounded propositions. She adds a list of other factors that might influence whether something can be presupposed/backgrounded, such as the relationship between speaker and addressee, genre, stylistic preferences, topic of conversation, and channel clarity. A little later she
Grice on Presupposition
91
mentions the importance of medium, viz. written versus spoken discourse (2000: 1435). Although there are similarities in Grice’s and Abbott’s views, there are also several ways in which their views differ. Firstly, with regard to what is foregrounded, Grice talks about those aspects of a speaker’s claim that a hearer may want to challenge, while Abbott talks about what a speaker wants to go on record as contributing to the conversation. These are somewhat different (if related) ideas. If we assume that typically speakers want to contribute something newsworthy to a conversation, these ideas will converge. A second difference is that Grice’s CT principle seems to be more hearer-oriented, while Abbott’s admonition not to presuppose what should be asserted is more speaker-oriented. Thirdly, the CT principle is a sub-maxim falling under the maxim of Manner and instructs speakers to express themselves in a way that helps hearers to follow up with an appropriate response. So what the speaker is being admonished to do is to think about how to direct the flow of conversation – how to ensure smooth turn-taking between the interlocutors. Abbott on the other hand seems more concerned with the substance of the conversation rather than its form. Speakers need to think about what they want to go on record as contributing and need to worry whether or not their main point gets across. Some of the factors affecting what can be backgrounded that Abbott mentions, such as stylistic preferences and channel clarity, will affect form. For example, if one is talking to someone over a bad phone line one may keep one’s utterances short and simple and place prominent accents on the main points. But other factors, such as the relationship between speaker and addressee, genre, and topic of conversation, are more relevant to substance and to what information can be backgrounded rather than to how it should be expressed. A full account of information packaging would need to combine both Grice’s hearer-oriented and Abbott’s more speaker-oriented points of view. Moreover, I am not suggesting that either of these authors, or even both together, offer anything like a complete theory of information packaging. There is a large literature concerned with the information structure component of language, and the Grice-Abbott view would have to be supplemented with insights gleaned from a review of this literature. (See for example Chafe 1976; Prince 1981; Heim 1982; Ward 1988; Lambrecht 1994; Birner and Ward 1998; Ward and Birner 2001.) However, the main point I wanted to make in this section is that Grice’s insights can be recast in a presuppositionalist framework, and I think that I have established that point.
3. Why the presuppositionalist account is to be preferred At the end of section 2.3 I said that I am in favour of the presuppositionneutralization explanation of Grice’s ‘loyalty examiner’ examples, which he
92
Anne Bezuidenhout
presents as cases in which an implicature of unique existence has been cancelled. I noted that since the neutralization explanation is one that makes definite NPs into soft triggers, we have a suitably ‘pragmatic’ rival to Grice’s implicature account. The notion of neutralization that I developed was inspired by Abbott (2005). However, as I noted in section 2.2, Abbott (2005) treats definite NPs as hard triggers, so in arguing that the class of soft presupposition triggers includes definite NPs, I am going considerably beyond anything she would be prepared to endorse. Also, my explication of the notion of presupposition neutralization in terms of taking a presupposition out of play goes beyond anything Abbott explicitly claims. Assuming that my arguments in section 2.3 in favour of treating definite NPs as soft rather than hard triggers are persuasive, I want in this section to give some reasons for preferring the soft trigger account to Grice’s implicature account. One general reason for preferring the presupposition account (which would apply whether one thought of definite NPs as hard or soft presupposition triggers) is that the implication of unique existence associated with ‘The F is G’ doesn’t seem much like the sorts of contextual implications that are paradigm cases of conversational implicatures. Grice might try to argue that the existence implicature is a generalized conversational implicature (GCI) rather than a particularized one. However, again, the existence implicature doesn’t pattern with paradigm GCIs, such as the GCI ‘not all’ triggered by ‘some’ or the GCI ‘not both’ triggered by ‘or’. Unlike the implication of unique existence, those GCIs are not entailments when their triggers occur in unembedded simple positive sentences. I assume it is clear that Grice would not want to class the existence implication as a conventional implicature, because conventional implicatures are supposed to be detachable, and Grice argues that the existence implication is non-detachable.18 In section 1, I presented Grice’s arguments for the claim that in certain cases, such as in his ‘loyalty examiner’ cases, the existence implication goes away. He takes this to show that a conversational implicature has been cancelled. Although he does not discuss this point, the cancellation here would have to be contextual or implicit (see Grice 1989: 39), since in the examples as he presents them the speaker does not supply an explicit cancellation clause, such as ‘and I do not mean to suggest that there is a loyalty examiner’. I assume that Grice is committed to saying that in his ‘loyalty examiner’ cases the implicature is actually derived (since the speaker has used the contracted form, which by Grice’s own reasoning and the CT principle would trigger the implicature). It is just that once the implicature is triggered it is seen to clash with other information already in the conversational context and thus it is repressed. One might object that these cases do not seem to give any hint of the sort of dissonance that Grice’s account predicts. Since the presupposition neutralization account predicts no such dissonance (given that the existence presupposition has been taken out of play) it seems better able to
Grice on Presupposition
93
account for the ‘phenomenology’ of these examples. Although Grice himself never gives any indication that he cares about issues having to do with language processing, one could invoke some sort of processing story that would help Grice avoid this ‘dissonance’ objection. One could argue that in these cases the implicature is triggered (because of the speaker’s use of the contracted form) but that before the information can reach conscious awareness, it is extinguished because the other information in the context (namely the information that the speaker does not believe in the existence of a loyalty examiner) is more highly activated. Since this happens before any interpretation reaches conscious awareness, the hearer is not conscious of any dissonance. The main problem with this response on Grice’s behalf is that it runs counter to his claims about the calculability of implicatures. Calculability suggests that the reasoning involved in implicature derivation is transparent to language users, and not something that takes place below the level of conscious awareness. My own examples (8) and (10) of presupposition neutralization, repeated below, were modelled after the sorts of examples given by Simons (2001) and Abbott (2005) involving cognitive factives. They present the speaker as expressing ignorance as to whether or not the existence presupposition is true: (8) I do not know whether or not anyone owns this book. However, if I ever find the owner I will return it. (10) I do not know whether or not John has an aunt. But is it such a good idea to ask his aunt for a reference? References from family members are not really appropriate. I do not think that it is necessary in cases of presupposition neutralization for the speaker to preface his remark with an explicit statement of ignorance. If this ignorance has already been established as part of the common ground, it can remain implicit. However, by making it explicit one can test to see whether one has a genuine case of presupposition neutralization. Using this test on Grice’s examples, example (2) emerges as a plausible case of neutralization as I have characterized it: (2′) I don’t know whether or not there is a loyalty examiner. But the loyalty examiner won’t be summoning you at any rate. Remember that the speaker here is supposed to be talking to a clearly loyal friend. It is not unreasonable to say that in (2′) the existence assumption has been taken out of play. Instead, the speaker is focusing on the ‘role’ property of examining people for their loyalty and expressing the thought that, due to the friend’s impeccable record of loyalty, the friend will never be the object of such examination.
94 Anne Bezuidenhout
It is more difficult to deal with example (1) in this way. In this case Grice does not specify that the individual being addressed is obviously loyal. We are to imagine just that the person is anxious about being summoned. Moreover, Grice says that the speaker can utter (1) with the aim of dispelling his friend’s worries precisely because he firmly believes that the loyalty examiner does not exist. Under these conditions, if we put in an explicit neutralization clause, we seem to get something unacceptable: (1′) # I don’t know whether or not there is a loyalty examiner. However, don’t worry, the loyalty examiner won’t summon you. I concede that the context Grice imagines does not seem to be one in which the presupposition is neutralized. This does not prove that (1) is better handled in the implicature framework. For the reasons given at the beginning of this section, it remains problematic to regard (1) as a case of implicature cancellation.19 More importantly, it does not prove that the definite NP in (1) is a hard trigger whose presupposition can never be neutralized. To see this, suppose we change the context for (1) in the following way. As before, we are discussing whether or not there is a loyalty examiner and you are worrying about being summoned. I express the opinion that, given the way the government operates and its desire for layers of secrecy, it is unlikely that such a person would ever reveal himself. Any summoning would be taken care of by a separate office, one with a separate chief officer and an ostensibly different mission. It seems to me that an utterance of (1′) in this context would be felicitous and that in this case the existence presupposition would be neutralized. I would in effect be communicating that summoning is not one of the role responsibilities of a loyalty examiner, so that you needn’t worry about being summoned by such a person.
4. Summary Grice claims that the implication of unique existence associated with ‘The F is not G’ is a conversational implicature because this implication is nondetachable, cancellable and calculable. After outlining Grice’s reasons for these claims, I showed that his arguments could be recast in a presuppositionalist framework. However, since one of Grice’s major motivations for giving an implicature account was a desire to give a pragmatic, as opposed to a semantic, account of this implication of unique existence, it was also important for me to argue for a suitably pragmatic presuppositionalist alternative. Thus I argued that definite NPs are soft presupposition triggers. In certain contexts the presupposition of existence associated with the definite NP ‘the F’ can be neutralized. After reviewing Abbott’s (2005) four-way distinction between presupposition cancellation, suspension, filtering and neutralization, I suggested that neutralization could be cashed out as a matter of taking a presupposition
Grice on Presupposition
95
out of play. If the speaker signals either implicitly or explicitly that he is ignorant as to whether or not the existence presupposition is true, then he signals that his utterance containing the presupposition trigger is to be interpreted in such a way that its main point is independent of the truth of the existence assumption. I provided several examples to illustrate such presupposition neutralization. In section 2.3, I considered some rival presuppositionalist accounts of the crucial ‘loyalty examiner’ examples that Grice discusses. I argued against treating these examples as cases of presupposition cancellation or of presupposition filtering (i.e., as cases in which the existence assumption is accommodated within the scope of an implicit belief operator) and in favour of the neutralization explanation. Then, in section 3, I argued that the presupposition neutralization account is superior to Grice’s account in terms of implicature cancellation, at least with respect to example (2). I admitted that example (1) remained problematic for my account. However, this cannot be used directly as an argument for the Gricean position. So long as there is some way of handling (1) in the presuppositionalist framework that is more satisfactory than Grice’s implicature account, one strand of my argument against Grice is secure. I could concede that (1) is best handled as a case of presupposition cancellation, even though I earlier argued against this treatment (because it makes the words ‘Don’t worry’ anomalous). Of course, a unified neutralization explanation of the Grice examples would have been desirable. Nevertheless, I think I have established that definite NPs are soft triggers, even if I haven’t been able to use Grice’s own example (1) to demonstrate my point. I have also argued that Grice’s insights about conversational tailoring that he used to argue for the calculability of the implication of unique existence can be reconstrued as insights about the problem of information packaging. Grice’s principle of conversational tailoring fits nicely with the account of presuppositions as background propositions that Abbott (2000) defends. I argued that she too is concerned with the problem of information packaging, but that she approaches the problem from the speaker’s point of view whereas Grice approaches it from the hearer’s point of view. Their views therefore nicely complement each other. Combining them provides us with the outlines of a very interesting framework for dealing with presuppositions, and in particular with the problem of informative presuppositions. Since informative presuppositions are in fact predicted on this account, we have an account that allows us to sidestep an appeal to the notion of presupposition accommodation and some of the thorny problems that the notion of accommodation has caused.
5. Addendum on Schlenker (2007) An anonymous referee pointed out that I do not discuss Schlenker’s (2007) transparency theory of presupposition. My chapter was written in 2006,
96
Anne Bezuidenhout
before Schlenker’s paper appeared and thus I cannot easily integrate a discussion of Schlenker’s views into my chapter. I would therefore like to make some remarks here as to how I see his views in relation to my own. I have adopted Abbott’s (2000) pragmatic account of presuppositions according to which presuppositions are backgrounded propositions that are expressed alongside the asserted content of an utterance. This view is opposed both to a semantic conception of presupposition according to which p presupposes q just in case p lacks a truth value unless q is true, as well as to a pragmatic conception according to which presuppositions are the assumptions that must be in the common ground in order for an assertion to be felicitous. Abbott’s view is intended to solve the problem of informative presuppositions. All the information conveyed by an utterance may be new information, even though a part of that information may be in the background relative to the main point of the utterance. Schlenker (2007) offers a pragmatic account of presupposition that, like Abbott’s proposal, is meant to remove the semantic difference between presuppositions and assertions. Moreover, in a section on problems and prospects, Schlenker (2007: 239) envisages applying his account to cases involving definite descriptions of the sort I’ve been discussing. Schlenker posits a ‘principle of division’ that separates the content of an utterance containing a presupposition trigger into its preconditions and its asserted content. For example, a sentence containing the factive verb ‘know’, which is a presupposition trigger, is generally agreed to presuppose the truth of its sentential complement. Thus by the principle of division, the sentence (a) below would be expanded as in (b): (a) John knows that his wife is cheating on him. (b) John’s wife is cheating on him and John knows that. Schlenker then posits a maxim, ‘Be Articulate!’, which enjoins a speaker to utter the explicit conjunction of precondition-plus-asserted content, unless there is some infelicity involved in asserting the articulated form, in which case the contracted form would be preferred. He suggests that there is a competing ‘principle of minimization’ that works to eliminate a conjunction if one of its conjuncts expresses information that is redundant – say because it expresses information that is already in the common ground. All of this is summarized in a ‘principle of transparency’ according to which the expanded form will be preferred in all cases except when the articulation of the precondition leads to redundancy because that information is already entailed by the common ground information. Transparency explains why the contracted form (a) is preferred when it is already part of the common ground that John’s wife is cheating on him. As applied to the specific case of definite descriptions, Schlenker suggests that by the principle of division, ‘The F is G’ is expanded into the conjunction ‘There is a unique F and all (or some) F’s are G’. By the maxim ‘Be
Grice on Presupposition
97
Articulate!’, a speaker is enjoined to use the expanded form unless it would be infelicitous to utter the conjunction, in which case the contracted form ‘The F is G’ is to be preferred. Schlenker then suggests that Grice’s CT maxim might be invoked to explain when it is or is not felicitous to use the expanded form. I have discussed Grice’s own views at length in my chapter, and I do not see that Schlenker’s suggestions provide any additional insights into the debate between Grice’s implicature view and my presuppositionalist view. If anything, Schlenker’s pragmatic account is less fleshed out than mine, since he simply gestures at Grice’s CT maxim, whereas I have used both Grice’s suggestions (about non-controversiality and what it takes for something to have common-ground status) and Abbott’s ideas about the factors that determine when something is appropriately backgrounded and when it should instead be asserted, in order to say something about when information needs to be made explicit/foregrounded and when it can be implicit/backgrounded. Additionally, I would say that Schlenker’s views support my claims, since he too is proposing to re-construe Grice’s claims about definite descriptions in terms of a pragmatic notion of presupposition, rather than by appealing to a notion of conversational implicature. Schlenker also says some things that might be construed as helping with the examples (1) and (2) from Grice that are discussed at length in my paper. Those examples involve the use of a non-denoting definite description, ‘the loyalty examiner’. Schlenker says that from a semantic point of view, nondenoting definite descriptions all refer to the non-existent object # and that therefore any sentence containing such a description is trivially false. He then writes: ‘two pragmatic violations are obtained when the expected presupposition is not satisfied: one is simply the violation of transparency; and an additional deviance might be obtained because the non-existent object # was allowed to play a role in the truth-conditions’ (2007: 239). Schlenker seems to be saying that Transparency demands the use of the expanded form ‘There is a unique F and all Fs are G’ when ‘the F’ is non-denoting, and that therefore (1) and (2) involve at least one pragmatic violation. However, Grice’s point is that (1) and (2) are perfectly felicitous, and I agree. Grice construes these as involving the cancellation of an implicature. I construe them as cases in which the existence presupposition has been neutralized. Thus Schlenker’s Transparency theory does not help decide between the Gricean view and the presuppositionalist view I have been advocating. I address the issue of presupposition failure at more length in Bezuidenhout (forthcoming).
Notes 1. This essay is based on a lecture given in 1970. The lecture was the fourth in a series of lectures referred to as the ‘Urbana seminars’ in the H. P. Grice Papers (BANC
98 Anne Bezuidenhout
2.
3.
4.
5.
6.
7.
8.
MSS 90/135 c, The Bancroft Library, University of California, Berkeley). It was the only lecture in the series that was published. See Chapman (2005: 124). On some accounts, the presupposition of unique existence carried by ‘The F is G’ is (also) an entailment. Abbott’s (2000) account according to which presuppositions are background propositions is one such account. So ‘The F is G’ actually expresses the proposition that there is a unique F, but this content is backgrounded. As Carston (2002: 281–4) notes, Grice’s position in P&CI is slightly more complicated than suggested in the text. Many subsequent writers have assumed that the Gricean view is that negation is semantically univocal (i.e., that the wide-scope sentence negation is what is semantically encoded) and that the narrow-scope predicate negation reading is a conversational implicature or pragmatic strengthening (just as ‘I don’t like you’ is pragmatically strengthened to ‘I dislike you’). This is how Atlas (1989), Horn (1989), Kempson (1975) and many others have understood the Gricean view. However, in P&CI (p. 272) Grice actually claims that ‘The F is not G’ is structurally ambiguous, with the narrow-scope predicate negation reading – the one he calls the strong reading – entailing that there is a unique F, and the wide-scope sentence negation reading – the one he calls the weak reading – merely conversationally implying that there is a unique F. Carston notes that this goes against the principle of ‘Modified Occam’s Razor’ in the sense that it appeals to both ambiguity and implicature in accounting for the data. In the case of other logical constants, such as ‘and’ and ‘or’, Grice argues for a univocal semantics and explains alternative ‘strengthened’ readings as implicatures. For example, ‘or’ is semantically inclusive, and its exclusive reading derived as an implicature. It is not completely clear how the first conjunct could be false while the other two remain true, unless we admit that there could be a non-existent unique bald king of France. This sounds rather Meinongian. There is some additional argumentation of Grice’s that I’m glossing over here. He spends some time arguing that one wouldn’t bother to focus on the generalization in the third conjunct of the Russellian expansion (or to make the prediction in the contracted form) unless one believed that there was an F which was an instance of the generalization (or believed that there was an F of whom the predication could be made). Although the possessive construction ‘my aunt’s cousin’ is itself a contraction of the definite description ‘the cousin of my aunt’, the contraction being referred to is the contraction represented by the whole of (3), which Grice assumes is a contraction of the Russellian three-conjunct analysis. In the earlier presupposition literature operators belonging to the S-family were called presupposition ‘holes’, since a presupposition triggered in a sentence embedded under such an operator projects out of this sentence and becomes a presupposition of the compound sentence. Thus it escapes the scope of these operators. In the more recent literature, such projection behaviour would be described as cases of locally triggered presuppositions becoming globally accommodated. I say more about this local/global accommodation distinction below. Strawson is often taken to be an advocate of the semantic conception of presupposition, according to which P is a presupposition of Q just in case Q is neither true nor false unless P is true. That is, false presuppositions lead to truth-value gaps. However, Soames (1982) notes that Strawson tends to run two ideas of presupposition together in his writings. One is the semantic/logical conception just mentioned, but another is what Soames calls the ‘expressive conception’, according to which P is a presupposition of Q just in case Q cannot be used to further
Grice on Presupposition
9.
10.
11.
12.
13.
14.
15.
16.
99
the assertive enterprise unless P is true. In Bezuidenhout (forthcoming) I suggest that the expressive conception is more important to Strawson’s concerns than the logical one, and is the one needed to explain why Strawson later conceded that there can be cases in which false presuppositions do not wreck the assertive enterprise. Abusch (2002), who also advocates the hard/soft trigger distinction (in fact the hard/soft trigger terminology was introduced by her), assumes in the body of her paper that definite descriptions are hard triggers, but she comments in a footnote: ‘Actually, it is not clear that the definite description is a hard trigger, since [its] presupposition is suspendable’ (2002: 18, n.10). However, she gives no examples of such presupposition suspension. And in any case, following Abbott (2005), I will be arguing below that presupposition neutralization and presupposition suspension are distinct phenomena. I do not mean to suggest that presuppositions generated within the scope of a belief operator are never admitted to the main context. For example, ‘John believes that Susan has stopped smoking’ would when said out-of-the-blue seem to presuppose that Susan once smoked, even though it is a statement about John’s belief-world. Abbott also goes on to endorse Horn’s view that presupposition-cancelling negation is not ordinary sentential negation, but an instance of metalinguistic negation. In (14) Speaker B would be objecting to Speaker A’s utterance on the grounds that it has a false presupposition and thus fails to make an assertion. This analysis of presupposition-cancelling negation as metalinguistic negation in Horn’s sense is controversial. For arguments against this view and an alternative analysis see Carston (2002: 291–302). In more recent parlance, cases where the presupposition fails to project would be treated as cases where the presupposition fails to be globally accommodated. In some cases however we may want to say that the presupposition survives but is instead accommodated in a local or intermediate DRS, as was argued above for Grice’s ‘loyalty examiner’ examples and for the quantifier restriction cases illustrated in (12). Abbott (2005: 7) lists a few other distinguishing characteristics of suspension clauses – ones that were noted by Horn (1972) – namely that such clauses use negative polarity items, that they cannot be preposed and that their work can be accomplished by ‘and maybe not’ clauses. For my purposes, the difference noted in the text is the most relevant one. My earlier example (8) would be explained along similar lines. If I am taking the presupposition that there is an owner of the book out of play, the main point of my utterance must be something that doesn’t depend on this existence assumption. I suggest that the main point is to reassure my interlocutor that I am not the sort of person who would hold onto property dishonestly. I have already indicated that an explanation along these lines can be given for my other example of neutralization, example (10). See section 2.2 above. If one wanted to maintain the cancellation account, one could claim that the words ‘Don’t worry’ are not being used sincerely in (1). Rather, since the speaker is in effect communicating that the hearer has no legitimate way of expressing his fears, he is being snide and condescending. That is, there is not an ambiguity between a wide-scope sentence negation and a narrow-scope predicate negation, or between a presupposition-preserving negation and a presupposition-cancelling one. However, Burton-Roberts does believe there is another metalinguistic negation operator, which is a non-truth-functional
100 Anne Bezuidenhout operator that can be glossed as ‘The speaker objects to u’, where u is an utterance of the corresponding positive sentence. 17. Stalnaker (1999) has suggested that to deal with informative presuppositions one needs to recognize that it is not what is in the common ground before the speaker’s utterance that matters but rather what is a part of the common ground after the utterance but before the point at which the hearer must decide whether to update the common ground with the information that the speaker has asserted. So by the time that update occurs, the presupposed information will no longer be new. I won’t pursue this line of reasoning any further in this chapter. See von Fintel (2000, 2006), Simons (2003) and Szabo (2006) for more discussion. 18. See Krause (2000) for further discussion of problems with Grice’s implicature analysis. Krause argues that it is implausible to think that a speaker who utters ‘The F is G’ has implicitly asserted something consisting of three conjuncts. He also thinks that Grice’s appeal to the Cooperative Principle in his argumentation is implausible, and that the non-detachability of the implicature is not predicted by the Russellian analysis of definite descriptions. 19. Remember that I have already argued against treating (1) as a case of presupposition cancellation or of presupposition filtering. See section 2.3 above. I say more about this apparent stalemate in the concluding section.
References Abbott, B. (2000) Presuppositions as non-assertions. Journal of Pragmatics, 32: 1419–37. Abbott, B. (2005) Where have some of the presuppositions gone? Ms. Abusch, D. (2002) Lexical alternatives as a source of pragmatic presuppositions. In B. Jackson (ed.) Proceedings of SALT XII. (Ithaca, NY: CLC Publications). Asher, N. and Lascarides, A. (1998) The semantics and pragmatics of presupposition. Journal of Semantics, 15: 239–99. Atlas, J. (1989) Philosophy without Ambiguity (Oxford: Oxford University Press). Atlas, J. (2004) Descriptions, linguistic topic/comment and negative existential. In M. Reimer and A. Bezuidenhout (eds.) Descriptions and Beyond (Oxford: Oxford University Press), pp. 342–60. Beaver, D. (2001) Presupposition and Assertion in Dynamic Semantics (Stanford: CSLI Publications). Bezuidenhout, A. (forthcoming) Presupposition failure and assertoric inertia. In M. Hackl and R. Thornton (eds.) Asserting, Meaning, and Implying, a volume in honor of Jay Atlas (Oxford: Oxford University Press). Birner, B. and Ward, G. (1998) Information Status and Non-Canonical Word Order in English (Amsterdam: Benjamins). Burton-Roberts, N. (1989) The Limits to Debate (Cambridge: Cambridge University Press). Carston, R. (2002) Thoughts and Utterances (Oxford: Blackwell). Chafe, W. (1976) Givenness, contrastiveness, definiteness, subjects, topics and point of view. In C. Li (ed.) Subject and Topic (New York: Academic Press), pp. 25–56. Chapman, S. (2005) Paul Grice: Philosopher and Linguist (Basingstoke and New York: Palgrave Macmillan). Chierchia, G. and McConnell-Ginet, S. (1990) Meaning and Grammar (Cambridge, MA: MIT Press).
Grice on Presupposition
101
Donnellan, K. (1966) Reference and definite descriptions. Philosophical Review, 77: 281–304. von Fintel, K. (2000) What is presupposition accommodation? Ms., MIT. von Fintel, K. (2004) Would you believe it? The king of France is back! (Presuppositions and truth value intuitions). In M. Reimer and A. Bezuidenhout (eds.) Descriptions and Beyond (Oxford: Oxford University Press), pp. 315–41. von Fintel, K. (2006) What is presupposition accommodation, again? Paper delivered at OSU Presupposition Accommodation Workshop, Columbus, Ohio, October 13–15, 2006. Geurts, B. (1996) Local satisfaction guaranteed: A presupposition theory and its problems. Linguistics and Philosophy, 19(3): 211–57. Geurts, B. (2000) Buoyancy and strength. Journal of Semantics, 17: 315–44. Grice, P.H. and Strawson, P. F. (1956) In defence of a dogma. Philosophical Review, 65: 141–58. Grice, P.H. (1989) Studies in the Way of Words (Cambridge, MA: Harvard University Press). Heim, I. (1982) The semantics of definite and indefinite noun phrases. PhD dissertation, University of Massachusetts, Amherst. Horn, L. (1972) On the semantic properties of logical operators in English. PhD dissertation, UCLA. Horn, L. (1989) A Natural History of Negation (Chicago: University of Chicago Press). Horn, L. (1992) Pragmatics, implicature and presupposition. In W. Bright (ed.) International Encyclopedia of Linguistics, vol. 3. (New York: Oxford University Press), pp. 260–6. Kempson, R. (1975) Presupposition and the Delimitation of Semantics (Cambridge: Cambridge University Press). Krause, P. (2000) On the relationship between the theories of presupposition and implicature. Paper delivered at the Workshop on Communication & Understanding – 2: Saying, Meaning, Implicating, University of Bielefeld, September 27–28, 2000. Lambrecht, K. (1994) Information Structure and Sentence Form: Topic, Focus and the Mental Representations of Discourse Referents (Cambridge: Cambridge University Press). Lewis, D. (1979) Scorekeeping in a language game. Journal of Philosophical Logic, 8: 339–59. Prince, E. (1981) Towards a taxonomy of given-new information. In P. Cole (ed.) Radical Pragmatics (New York: Academic Press), pp. 223–55. Roberts, C. (1996) Information structure: Towards an integrated theory of formal pragmatics. In J. Yoon and A. Kathol (eds.) OSU Working Papers in Linguistics, 48: 91–136. Roberts, C. (2004) Discourse context in dynamic interpretation. In L. Horn and G. Ward (eds.) Handbook of Contemporary Pragmatic Theory (Oxford: Blackwell), pp. 197–220. Sauerland, U. (2008) Implicated presuppositions. In A. Steube (ed.) The Discourse Potential of Underspecified Structures (Berlin: Walter de Gruyter), pp. 581–600. Schlenker, P. (2007) Transparency: An incremental theory of presupposition projection. In U. Sauerland and P. Stateva (eds.) Presupposition and Implicature in Compositional Semantics (Basingstoke and New York: Palgrave Macmillan), pp. 214–42. Simons, M. (2001) On the conversational basis of some presuppositions. In R. Hastings, B. Jackson and Z. Zvolensky (eds.) Semantics and Linguistic Theory: (SALT) 11 (Ithaca: CLC Publications).
102
Anne Bezuidenhout
Simons, M. (2003) Presupposition and accommodation: Understanding the Stalnakerian picture. Philosophical Studies, 112: 251–78. Soames, S. (1982) How presuppositions are inherited: A solution to the projection problem. Linguistic Inquiry, 13: 483–545. Stalnaker, R. (1999) Context and Content (Oxford: Oxford University Press). Szabo, Z. (2006) Comments on von Fintel. Paper delivered at OSU Presupposition Accommodation Workshop, Columbus, Ohio, October 14, 2006. Ward, G. (1988) The Semantics and Pragmatics of Preposing (New York: Garland). Ward, G. and Birner, B. (2001) Discourse and information structure. In D. Schiffrin, D. Tannen and H. Hamilton (eds.) The Handbook of Discourse Analysis (Oxford: Blackwell), pp. 119–37. Zeevat, H. (1992) Presupposition and accommodation in update semantics. Journal of Semantics, 9: 379–412.
5 Irregular Negations: Implicature and Idiom Theories Wayne A. Davis
Horn identified a number of unusual negations that are not used in accordance with the standard rules of logic. I call them irregular negations. We will study five different types: scalar, metalinguistic and evaluative implicature denials; presupposition-cancelling denials; and contrary affirmations. What do these negations mean or convey? How is their irregular interpretation related to their regular interpretation? Are they pragmatically or semantically ambiguous? These are the principal questions I shall address. I will here focus on one type of pragmatic theory: implicature theories. I argue that an implicature theory works well only for evaluative implicature denials. Other irregular negations, I argue, are semantically ambiguous in an unusual way. I stress the partial compositionality of irregular negations, while accounting for their distinctive features by making the case that they are idioms which plausibly evolved from generalized conversational implicatures.1
1. Irregular negations We will focus our attention on singular subject-predicate sentences, whose form we shall represent ‘s is P’. In this form, ‘s’ stands for any singular term, including proper names, definite descriptions, and pronouns. The negation of ‘s is P’ will be represented by ‘s is not P’ or its contraction. I refer to ‘s is P’ as the root of its negation. More generally, I will use ‘p’ as a place-holder for declarative sentences of any form, and ‘not-p’ for any negation whose root is ‘p’. By a regular negation, I shall mean a negation interpreted in such a way that it conforms to the logical rules of truth functionality and obversion: (1)
a. ‘s is not P’ is true (false) iff ‘s is P’ is false (true). Truth-functionality b. ‘s is not P’ is equivalent to ‘s is non-P’. Obversion
A negation interpreted in such a way that it does not conform to both rules is irregular. I use ‘non-P’ to represent any complement of ‘P’: any predicate that necessarily applies to everything ‘P’ does not apply to, and nothing 103
104
Wayne A. Davis
else. If ‘P’ is ‘blue’, complements include ‘non-blue’, ‘other than blue’, and ‘a thing that is not blue’. When regular, ‘The sky is not blue’ is equivalent to ‘The sky is non-blue’, and is true iff ‘The sky is blue’ is false. We will not restrict our attention to predicates with the grammatical form ‘is P’. The complement of ‘believed that p’, for example, is ‘failed to believe that p’ or ‘lacked the belief that p’. So we will take the obversion rule to say that ‘s did not believe that p’ is equivalent to ‘s failed to believe that p’.2 While I am using the term ‘negation’ to denote sentences, it can also be used in different senses to denote a propositional operator, a proposition with that operator applied, or a speech act. Now consider the following examples of what I call negation-correction conjunctions. (2) a. The sun is not larger than some planets: it is larger than all planets. Scalar Implicature Denial b. That’s not a tomäto: it’s a tomaˉto.3 Metalinguistic Implicature Denial c. Vulcan is not hot: it does not exist. Presupposition-cancelling Denial The negation clauses in these examples can be interpreted as regular negations, but so interpreted the conjunctions as a whole are contradictory. They would be equivalent to the sentences in (3): (3)
a. The sun is larger than no planets: it is larger than all planets. b. That’s a non-tomäto: it’s a tomaˉto. c. Vulcan is other than hot: it does not exist.
The conjunctions in (2) would most naturally be interpreted in a way that is consistent, however. On the non-contradictory interpretation, the negations in (2) are not equivalent to the corresponding clauses in (3), and consequently do not conform to the obversion rule. The negations in (2) diverge from the truth-conditionality rule too (with the possible exception of presuppositioncancelling denials). The first clause of (2a), (2a1), is true even though ‘The sun is larger than some planets’ is true, not false. Since the negations in (2) on this interpretation do not conform to (1), they are irregular. The traditional example of a presupposition-cancelling denial is: (4) The king of France is not bald: there is no king of France. This may suggest that the difference between the regular and irregular interpretation is a Russellian scope distinction: ∃x(F!x&–Gx) does, and –∃x(F!x&Gx) does not, entail ∃x(F!x).4 But there is little plausibility to the suggestion that the two interpretations of (2c1) arise because the negation operator can have two different sentential structures in its scope. The evidence against the description theory of names, and for the view that they
Irregular Negations: Implicature & Idiom Theories 105
are directly referential, makes it highly unlikely that proper names express concepts with any significant constituent structure. I have defined an irregular negation as one that diverges from either or both of the rules in (1). I believe that ‘s is not P’ conforms to the truthfunctionality rule if and only if it conforms to the obversion rule, but I wish to leave this issue open along with the bivalence issue. On the view I favour, ‘Vulcan is hot’ is neither true nor false because Vulcan does not exist. Hence (2c1) is true on its presupposition-cancelling interpretation even though its root is not false, violating truth functionality. Others maintain, however, that ‘Vulcan is hot’ is false, satisfying (1a). The presupposition-cancelling interpretation of ‘Vulcan is not hot’ would still count as irregular, though; for ‘Vulcan is non-hot’ also comes out false, and thus not equivalent to ‘Vulcan is not hot’, a divergence from (1b). Since irregular negations are defined as those diverging from either of two rules used to formulate laws of logic, it is not surprising that they may also diverge from others. Indeed, we will see that irregular negations often diverge from the following as well: (5)
a. ‘s is not not P’ is equivalent to ‘s is P’. Double Negation b. ‘s is not P and Q’ is equivalent to ‘s is either not P or not Q’. DeMorgan’s Rule c. ‘s is not P’ and ‘s is P or Q’ entail ‘s is Q’. Disjunctive Syllogism d. ‘s is not R to some Q’ is equivalent to ‘s is R to no Q’. Contradictory Opposition e. ‘s is P’ and ‘s is not P’ cannot both be true. Non-Contradiction f. Either ‘s is P’ or ‘s is not P’ must be true if s exists. Excluded Middle
For example, ‘The sun is not larger than some planets’ as most naturally interpreted in (2a) is not equivalent to ‘The sun is larger than no planets’, diverging from the contradictory opposition rule. Hence both (2a1) and its root can be true, violating non-contradiction. Neither (2a1) nor its root would be true if the sun’s diameter were just one metre, violating excluded middle.5 For a divergence from double negation, note that on its most natural interpretation, (6a1) is true for the same reason (2c1) is: because Vulcan is non-existent, it is not anything. But on this interpretation, (6a1) is obviously not equivalent to (6b1), since (6b) is contradictory: (6) a. Vulcan is not not hot: it does not exist. b. Vulcan is hot: it does not exist. Note that while (6a1) is irregular, its root is regular. When I say that an irregular negation fails to conform to the double negation rule (5a), I am not suggesting that there are any exceptions to the equivalence of the propositions ¬ ¬ P and P . My claim only implies that sentence (5a) does not
106
Wayne A. Davis
express the equivalence of ¬ ¬ P and P when ‘s is not not P’ is irregular. Another question that any theory of irregular negations needs to answer is: Why do the logical rules governing regular negations fail for irregular negations? The obvious explanation is that ‘not’ expresses a different operator in irregular negations. Whether this is the best explanation remains to be seen. Irregular negations have a number of additional common features. The first is some sort of ambiguity. Any irregular negation can also be interpreted as a regular negation. Some negations, however, are unambiguously regular (e.g., Vulcan does not exist). The ambiguity is especially striking in (2), because one interpretation makes the conjunction contradictory and the other does not. The most widely discussed question concerns the nature of the two interpretations: Are irregular negations semantically or pragmatically ambiguous? Do they strictly speaking have two linguistically encoded meanings, or is one interpretation pragmatically generated? If the sentences are semantically unambiguous, which interpretation is the semantic meaning? Is the answer the same for all cases? A second common feature is limited substitutivity of synonyms.6 ‘The sun failed to be larger than some planets, it is larger than all planets’ can only be heard with the regular, contradictory meaning of (2a). Replacing ‘not hot’ with ‘non-hot’ in (2c) has the same effect. As a final and more complex example, note that regular negation-correction conjunctions like (7) can be paraphrased either s is not P but Q or equivalently s is not P, but it is Q. (7)
That’s not a violin, it’s a viola.
In example (2), however, the first sort of paraphrase is possible, but not the second. ‘The sun is not larger than some planets, but it is larger than all planets’ is incoherent; to obtain a paraphrase, ‘for’ should replace ‘but’. A third common property is partial compositionality. The meaning of an irregular negation is partly but not completely determined by the meanings of its components and its grammatical structure. For example, replacing ‘larger than’ with ‘hotter than’ or ‘attracting’ changes the meaning of (2a1) in completely predictable ways. And its meaning on a particular occasion depends in predictable ways on whether ‘sun’ means “Sun” or “star”. Nevertheless, the irregular meaning of (2a1) is not what we would expect given the meaning of ‘not’, ‘some’, and its grammatical structure. As a result, substituting ‘any’ for ‘some’, or ‘fails to be larger’ for ‘not larger’, changes the meaning entirely. So we should ask the following questions: What interpretation rule governs an irregular negation? Does one rule cover all cases? These questions can be raised whether the irregular interpretation is linguistically encoded or pragmatically generated.
Irregular Negations: Implicature & Idiom Theories 107
2. Marking Horn observed that the irregular negations in (2) are normally marked in various ways: they typically have a distinctive intonation, exclude negative polarity items while permitting positive ones, are ‘echoic’, and resist morphological incorporation.7 Intonation. When (2a1) is regular, ‘not’ may be stressed but not ‘some’; ‘not’ is given the same pitch as ‘some’, or a higher pitch. When the negation is irregular, ‘some’ is stressed and given the higher pitch. Even more distinctively, the irregular negations in (2) typically end with what Ladd (1980: 146) called the ‘fall-rise’ contour, whereas regular negations typically end with a simple ‘fall’. Thus when (2a1) is irregular, ‘planets’ starts off at a lower pitch than ‘some’, and then rises back up. When the clause is regular, ‘planets’ starts off at the same pitch as ‘some’, and then falls. This distinctive intonation, however, is only loosely connected with irregularity. (2a1) may have its irregular meaning even if delivered in a monotone, or written in straight Roman type. Conversely, ‘violin’ in (7) may be given contrastive stress and the fall-rise intonation, even though the negation is completely regular.8 Polarity licensing. ‘Any’ is described as a ‘negative polarity item’ because it can occur in negations in ways it cannot occur in their roots. Thus ‘Mary does not have any money’ means the same as ‘Mary does not have some money’; but while ‘Mary has some money’ is grammatical, ‘Mary has any money’ is not. Negative polarity items generally cannot occur, however, in the scope of irregular negations like those in (2). Thus ‘The sun is not larger than any planets’ can only be heard as regular. Similarly, ‘yet’ can replace ‘already’ in negative sentences, though not in their roots. But whereas ‘Vulcan is not already hot’ can be heard as either regular or irregular, ‘Vulcan is not yet hot’ has only a regular interpretation. Conversely, we generally avoid positive polarity items like ‘pretty’ in regular negations, preferring ‘She is not tall’ to ‘She is not pretty tall’. But the latter is very natural as an irregular negation when ‘pretty’ is stressed, and a correction clause like ‘She’s really tall’ is waiting. Echoicity. The irregular negations in (2) are most naturally used in response to an assertion of their root, and followed by a correction clause. They are much less likely than regular negations to begin a conversation. And when they are encountered out of context, we tend to imagine them as responses to assertions of their root. This property too is only loosely connected to irregularity.9 An irregular negation can occur without any prior assertion, as in ‘If Vulcan does not exist, then it is not hot’. Furthermore, regular negations may also be echoic. Thus (7) is most naturally used in response to the assertion ‘That is a violin’. When read out of context, we tend to imagine such an antecedent assertion.
108 Wayne A. Davis
Morphological incorporation. ‘Not’ typically incorporates morphologically as the ‘un-‘ or ‘in-‘ prefixes, as in ‘unnoticed’ and ‘immaterial’. But ‘Vulcan is immaterial’ does not have the ambiguity that ‘Vulcan is not material’ has. ‘Vulcan is immaterial’ only has a regular interpretation. At least one large class of irregular negations is unmarked. I call them contrary affirmations, because the negation of ‘p’ is used to affirm a contrary of ‘p’ rather than simply its contradictory.10 (8)
a. The fire chief does not believe that anyone survived. b. It is not good that the ice caps are melting. c. After Stalingrad, Hitler was not happy. Contrary Affirmations
For example, (8b) is used to affirm that it is bad that the ice caps are melting, which is the contrary of its root. (8b) may look like a mere understatement, but is too conventional. It contrasts strikingly with ‘It is not bad that the ice caps are melting’, which could in appropriate contexts be used as an understated way of saying that it is good they are, but does not have this as one of its meanings. Moreover, (8b) can be used as an understated contrary affirmation to mean that it is very bad that the ice caps are melting. Different intonations do help signal the different interpretations of some of the examples in (8) (8c is an exception). But they are normally used without the fall-rise intonation characteristic of the examples in (2). Contrary affirmations are not at all echoic. They can be used in response to an affirmation of their root. But we do not automatically imagine such a prior utterance, and they can very naturally be used to begin a conversation. If the sentences in (8) are followed by a correction clause, they are automatically given their regular interpretation; consider ‘It is not good that the ice caps are melting, it is bad’. If the negation were read as a contrary affirmation, the second clause would be redundant rather than corrective. Negative polarity items are permitted, as (8a) illustrates, and the negation incorporates prefixally. Thus (8a) says that the chief disbelieves that someone survived, and (8c) says that Hitler was unhappy. The only marking we find is a limited restriction on negative polarity items. ‘Ralph does not believe in God any more’ has only a regular interpretation (it entails that Ralph is an agnostic, but not that he is an atheist). But ‘anyone’ is permitted in examples like (8a). Despite the lack of marking, contrary affirmations are as irregular as the denials in (2). For example, the mere falsity of ‘It is good that the ice caps are melting’ does not suffice for (8b) to be true on its stronger interpretation. Something may fail to be good without being bad. Similarly, (8a) says more than that the chief failed to believe that anyone survived – which might be the case if he had no opinion on the matter. The negations in (8) are all markedly ambiguous in some way, with a stronger and a weaker reading. They are only partially compositional on their stronger
Irregular Negations: Implicature & Idiom Theories 109
interpretation, and substitutivity of synonyms is limited. Indeed, whereas reformulating s is not P as It is not the case that s is P preserves the ambiguity in (2),11 it blocks the irregular meaning in (8). ‘It is not the case that Hitler was happy’, for example, does not imply that Hitler was unhappy. From a semantic point of view, the similarities between contrary affirmations and implicature and presupposition-cancelling denials are much greater than the differences.
3. Geurts’s theory Geurts (1998) holds that no single rule covers the interpretation of all marked negations. His discussion suggests, plausibly, that (2a) and (2b) mean the following when used with their irregular meanings:12 (9)
a. The sun is not larger than just some planets, it is larger than all planets. b. That’s not a thing called a tomäto, it is a thing called a tomaˉ to.
Geurts’s analyses easily explain why obversion, truth functionality, and the other standard logical rules fail for irregular negations. Geurts’s rule for presupposition-cancelling denials is very different. What the first clause of (2c) presupposes is “Vulcan exists”. What it asserts, according to Geurts, is “Nothing is Vulcan and hot”.13 Cancelling the presupposition leaves the assertion. So Geurts takes (2c) to mean: (9)
c. Nothing is Vulcan and hot: nothing is Vulcan.
While (9c) and (2c) may have the same truth-conditions, they seem far from synonymous. Even as an irregular negation, the singular subject-predicate sentence ‘Vulcan is not hot’ does not express the universal negative (or negative existential) proposition “Nothing is Vulcan and hot”. With the exception of logicians inspired by Boole or Russell, ordinary English speakers seldom even think such thoughts. Alternatively, we can propose that in a presupposition denial, ‘s is not P’ is used to express the thought or proposition that s is P and deny that it is true.14 Then (2c) says: (9)
c′. The proposition that Vulcan is hot is not true: Vulcan does not exist.
‘The proposition that’ is often omitted in colloquial speech. Since ‘The proposition that s is P is not true’ may be true when both ‘s is non-P’ and ‘s is P’ are neither true nor false, this analysis accounts for why presuppositioncancelling denials do not conform to the obversion and truth-functionality rules.
110
Wayne A. Davis
Geurts’s conclusion that no single interpretation rule governs marked negations holds for contrary affirmations as well. All the examples in (8) are used to affirm a contrary of their root. But different contraries are affirmed. (10)
a. The fire chief believes that no one survived. b. It is bad that the ice caps are melting. c. After Stalingrad, Hitler was unhappy.
These can be subsumed under a generalization by noting that to believe not-p is to disbelieve p. Then in each case, ‘V’ expresses positive values on some scale, and ‘not V’ expresses negative values. That is, ‘not V+’ means ‘V−’. Geurts (1998: 287) denies that the word ‘not’ is ambiguous.15 It expresses a single propositional operator. The differences in meaning between negations, on his view, are due to differences in the propositions to which that operator is applied: the proposition expressed by the root sentence in the regular case and a proposition implied by the root sentence in the irregular cases. I believe the evidence against a lexical ambiguity in ‘not’ is compelling. It is hard to see how an ambiguity in ‘not’ could account for all the different differences in interpretation illustrated by (2a–c) and (8a–c). The Aristotelian distinction between contradictory predicate denial and contrary predicate negation (Horn 1989: 107) might account for the ambiguity of (2c), but not that of (2a), (2b), or (8). Moreover, if ‘not’ were ambiguous, we would expect to find an ambiguity in ‘Vulcan does not exist, Not everyone died’, and (9c′); but we do not. We would also expect to find languages in which the difference is lexicalized (Gazdar 1979: 65–6; Horn 1989: 367). Geurts maintains instead that the irregular meaning of (2a1) and (2b1) is due to a semantic transfer, whereby ‘on a given occasion, a word acquires a contextual meaning that is not encoded in its lexical entry’ (1998: 288). His discussion suggests that while the sole lexical entry for ‘some’ is “at least some”, it acquires the meaning “just some” (i.e., “at least some but not all or many”) on occasions in which (2a1) has its irregular meaning. Similarly, ‘tomäto’ acquires the meaning “thing called a tomato” when (2b1) is irregular. Geurts’s paradigm for semantic transfer is provided by Nunberg’s (1978) example ‘The ham sandwich is waiting for his check’, used by a waitress to mean that the person who ordered the ham sandwich is waiting. Geurts says that ‘ham sandwich’ acquires the meaning “person who ordered a ham sandwich” in such contexts. Geurts does not explain why he does not hold instead that it is ‘not’ that acquires different contextual meanings, “not just” in (2a) or “not properly called” in (2b). If semantic transfer really occurred, then the evidence against a lexical ambiguity in ‘not’ would not refute the thesis that additional meanings are transferred to ‘not’ in certain contexts.
Irregular Negations: Implicature & Idiom Theories 111
I believe we need to distinguish between what a word means on an occasion (or in a context) from what a speaker means by it then (or there).16 If someone suffers a slip of the tongue and says ‘The precedent is conservative,’ the speaker may mean “president” by ‘precedent,’ but the word ‘precedent’ does not mean “president” even on that occasion. Similarly, ‘ham sandwich’ does not mean “person who ordered a ham sandwich” even on occasions in which a waitress uses it with that meaning. She is merely using a figure of speech called ‘metonymy’. What a word means on an occasion must be something it means in the language; it is always one of the word’s lexical entries. This goes for sentences as well as for individual words. It is incoherent to say that a word means something when used as a word of English that the word does not mean in English, or that it acquires something on an occasion that it does not have. There is no incoherence in saying that what a speaker means by the word differs from what the word means. I believe that ‘some’ has its standard meaning in (2a1), just as it does when ‘Some S is P’ is used to implicate “Not all S is P”. If ‘some’ meant “just some” in certain contexts, then (2a1) would be perfectly regular and completely compositional in those contexts. We cannot classify ‘There are not more than five words on the page’ as irregular or non-compositional on an occasion in which ‘word’ means “word types” on the grounds that on other occasions ‘word’ means “word tokens”. I think it is clear that speakers use (2a1) to mean “the sun is not larger than just some planets”. But they do not mean “just some” by ‘some’ any more than they mean “not just” by ‘not’. Irregular negations are not completely compositional. We seek to explain why.
4. I-Implicature theory Grice (1957) distinguished what a speaker means from what a sentence or other expression the speaker used means. Many good examples are provided by implicature, the act of meaning or implying something by saying something else.17 Thus in a letter of recommendation for a position in metaphysics, the evaluator may mean “The candidate is weak” by writing ‘She dresses neatly’, but the sentence he wrote does not. The distinction is less clear but still discernible in what Grice called generalized conversational implicature. I have so far considered only cases of what I might call “particularized conversational implicature” – that is to say, cases in which an implicature is carried by saying that p on a particular occasion in virtue of special features of the context, cases in which there is no room for the idea that an implicature of this sort is normally carried by saying that p. But there are cases of generalized conversational implicature. Sometimes one can say that the use of a certain form of words in an utterance would normally (in the absence of special circumstances) carry such-and-such an implicature or type of implicature.... (Grice 1975: 37)
112
Wayne A. Davis
While speakers conventionally use sentences of the form ‘Some S is P’ to mean “Not all S is P”, the sentences themselves do not. On the contrary, ‘All S are P’ entails ‘Some S are P’. There is some plausibility to the idea that ‘some’ has a generic and a specific sense, meaning either “at least some” or “just/only some”. If so, it would be like ‘animal’, which has a sense in which it applies to humans and a sense in which it does not. But there is no interpretation of (11) on which it is contradictory; no reading of ‘some’ would require replacing ‘indeed’ with ‘take it back’: (11) There are some who survived, indeed all did. (12) There is an animal in the basement, indeed a person. In contrast, (12) does have a contradictory interpretation; there is a reading of ‘animal’ that would call for ‘take it back’ in place of ‘indeed’. When I read (12), I generally hear ‘animal’ as excluding humans until ‘a person’ arrives. At that point my reading of ‘animal’ shifts unless I am forcing the contradictory interpretation on the sentence. I similarly distinguish between what a speaker implicates and what a sentence implicates (Davis 1998: 6ff.). The sentence ‘Some passengers died’ implicates “Not all passengers died” even though a given speaker may not have used it with that implicature on a given occasion. Conversely, Ann may use ‘There is a station around the corner’ to implicate “You can get gas there” even though the sentence itself does not have that implicature. Roughly, what a sentence implicates is what it may be used conventionally to implicate. This is not what the sentence means, although in some cases the implicature is part of the meaning of the sentence.18 We might therefore propose that the irregular use of negations involves a generalized conversational implicature – a sentence implicature – rather than a second sense. I call this the I-Implicature theory, because it takes the irregular interpretation of a negation to be an implicature. The regular interpretation is its meaning. The clearest advocate of I-Implicature theory, I believe, is Burton-Roberts (1989: esp. 119, fn.12). It is Geurts’s (1998) theory too, if ‘semantic transfer’ is understood to involve the speaker-meaning something different from what the expression used means. The I-Implicature theory also accords well with Horn’s (1985: 132, 1989: 370–7) thesis that negation is pragmatically rather than semantically ambiguous. Words are semantically ambiguous provided they have more than one sense. They are pragmatically ambiguous, Horn can say, if speakers conventionally use them to mean different things even though they are not semantically ambiguous.19 The I-Implicature theory is not the only pragmatic ambiguity theory, as we shall see. The I-Implicature account of why irregular negations do not conform to the regular rules of logic is similar to the ambiguity account, with a key difference. What negations mean conforms to the logical rules of regular negation. So strictly speaking, the rules are valid unambiguously. But the
Irregular Negations: Implicature & Idiom Theories 113
sentences implicate something different, with a different logical structure. Even though ‘The sun is not larger than some planets’ is equivalent to ‘The sun is larger than no planets’, the former implicates something that is not equivalent, namely “The sun is not larger than just some planets”. The ambiguity theory takes this to be a second sense. I believe the I-Implicature theory is correct for a class of irregular negations we have not yet considered. (13) a. The glass isn’t half empty: it’s half full.20 b. It isn’t partly sunny: it’s mostly cloudy. c. The nominee isn’t principled: he’s an ideologue. Evaluative Implicature Denials In addition to their regular interpretations, these all have an irregular interpretation on which they are used to deny an evaluative implicature of their root. Thus (13a) is used to deny that things are bad because the glass is halfway from being full, and to affirm that things are good because the glass is half-way towards being full. A positive evaluation can be denied too, as (13b) illustrates. Evaluative implicature denials are clearly heard as figures of speech. We take the speaker to be saying something that is literally false in order to convey something else that is true. What the speaker says may be contradictory, as in (13a). What he means is not. The sentences themselves have only the regular interpretation. Since what S means is something other than what S literally says, the irregular meaning of the negation is an implicature. One of Grice’s ‘glosses’ would be fitting: It is perfectly obvious to A and his audience that what A has said ... is something he does not believe, and the audience knows that A knows that this is obvious to the audience. So, unless A’s utterance is entirely pointless, A must be trying to get across some other proposition than the one he purports to be putting forward. This must be some obviously related proposition. (Grice 1975: 34) The denial of the evaluative implicature of the root of (13a1) is an obviously related proposition. The I-Implicature theory is untenable, however, for the other irregular negations. When speakers use a scalar implicature denial, a metalinguistic implicature denial, a presupposition-cancelling denial, or a contrary affirmation, they do not say what the regular negation says. Indeed, the regular interpretation is not intended or heard at all. Hence the indirection or two levels of meaning characteristic of implicature is absent. On the I-Implicature theory, when (2a) is irregular, what the speaker says is that the sun is both not larger than any planets and larger than all planets. But in fact, this is not something the speaker says. We normally do not hear or intend (2a) as a contradiction.
114
Wayne A. Davis
Indeed, it is remarkable how hard it is to force a contradictory reading. Even in (14), we tend to hear (c) as non-contradictory. (14) The following claims are contradictory: a. Some triangles do not have three angles. b. He does and he does not like it. c. The sun is not larger than some planets: it is larger than all planets. This use of (2a) contrasts markedly with ‘I do and I don’t’ given in response to ‘Do you like the President’s policies?’ Here what the speaker says is contradictory even though what he means is not. The use of (2a) contrasts in the same way with (13a), which is clearly heard as a literal contradiction and as a figure of speech. The irregular use of (2a) is also unlike the ironic use of ‘The sun is not larger than any planets’. When this is used ironically, the speaker means the opposite of what he says. The speaker says that the sun is not larger than any planets, but what he means is that it is larger than some planets, indeed many. We hear both the literal and the ironic meaning. With the exception of evaluative implicature denials, irregular negations are more like idioms, in which the regular meaning can be heard if we focus on it, but is not normally perceived or intended. As Carston (1996: 312–13) observed, the ‘double processing’ that goes on in ‘garden path’ examples like (15) is markedly unlike what normally happens with scalar implicature denials. (15) Front of card: This card is not from one of your admirers. Inside of card: It’s from two! Happy Birthday from Both of Us.21 Grice (1975: 39, 1978: 43) formulated two tests of implicature: cancellability and non-detachability. These give mixed results, even for evaluative implicature denials. The irregular interpretation of a negation is cancellable. In the following arguments, for example, the negations would most naturally be interpreted as regular. (16) a. As the square of opposition tells us, if the sun is larger than all planets, then it is larger than some. The sun is not larger than some planets. So, it is larger than no planets. b. The glass is half full if and only if it is half empty. It is not half empty, so it is not half full. Cancellability does not prove, however, that the irregular interpretation is an implicature rather than a sense. The context may be disambiguating the negation rather than blocking an implicature. Grice’s second test fails: The irregular interpretation is detachable. The substitution of synonyms may block the irregular interpretation of negations.
Irregular Negations: Implicature & Idiom Theories 115
This is to be expected with metalinguistic implicature denials. But it happens with other irregular negations too. Substituting ‘at least some’ or ‘any’ for ‘some’ in (2a) forces the contradictory regular interpretation. Even in (13a), substituting ‘is other than’ or ‘fails to be’ for ‘isn’t’ blocks the irregular interpretation. Grice himself observed that manner implicatures are exceptions to non-detachability, since they arise from the way something is said, rather than what is said. But I-implicatures are thought to arise from the maxims of Quantity and Quality instead, since contradictions in what is said are neither true nor informative. So the failure is problematic for I-Implicature theory.
5. Calculability What Grice (1975: 31) took to be essential to conversational implicatures is calculability, which is the ability to infer the implicature in a specific way, from the meaning of the sentence used, the Cooperative Principle (CP), and information from the context of utterance. The key premise in the ‘working out schema’ is S could not be observing the Cooperative Principle unless he believes that p, where ‘p’ expresses what S implicated. The irregular interpretation of negations is not calculable. Since scalar implicature denials seem to present the most favourable case for calculability, I will focus on them. Let us suppose, contrary to what was observed above, that when (2a) is irregular, the speaker said that the sun is not larger than any planets, it is larger than all planets. Let us also grant that despite having said something contradictory, the speaker is observing CP, and even the maxim of Quality. The first thing to observe is that it is possible for the speaker to be observing CP without believing that the sun is larger than all, not just some, planets. It is perfectly possible that the speaker was using irony; she may have meant and believed that it is larger than some but not all planets. The speaker may also have been engaging in hyperbole, meaning not that the sun is literally larger than all planets, but only that it is larger than many. The speaker might even have intended a literal contradiction. She might have intended (2a) to be blatantly contradictory in order to invite the question ‘What do you mean?’ for which she had an answer she wished to provide. Alternatively, she might believe she was forced to a contradictory conclusion by some philosophical arguments she had developed, leading her to give up the law of non-contradiction and adopt Hegelianism. In these last two cases, she would literally mean the regular interpretation of the negation, while implicating nothing. Grice cannot assume, as part of the working out data, that the speaker is not being ironic or hyperbolic, or meaning just what she literally said. For that would assume what Grice is trying to infer: the speaker’s implicature. Even if the speaker is not engaging in irony, hyperbole, or literal contradiction, she could conform to CP by implicating other things. The woman
116
Wayne A. Davis
who uses (17) could have meant and believed either of the alternatives to (17a) listed below, a small sample of the possibilities. She would have been no less cooperative, and would have obeyed the maxims at least as well. (17)
The sun is not larger than some planets (as in (2a)). a. The sun is not larger than all planets. b. The sun is not larger than eight planets. c. The sun is not larger than nearby planets.
It would be highly unconventional for the woman to mean either of these alternatives, of course. Indeed, there is little or no precedent for such implicatures. But nothing in CP or the other working-out data rules them out. Weakening the key working-out premise by replacing ‘could not’ with ‘would probably not’ will not eliminate the problem. For the cooperative presumption may not make the irregular interpretation any more likely than a figure of speech or a nonce implicature. The non-calculability of implicatures does not mean that we cannot infer them from facts about the context, only that the evidence and inference rule Grice focused on is not sufficient to do so. Parallel reasoning will show that no conversational implicatures are calculable. That was one of the principal conclusions of my Implicature (1998). I argued instead that generalized conversational implicatures arise from conversational implicature conventions: regularities in what speakers use sentences to conversationally implicate (mean or imply by saying something else) that are socially useful (serving among other things the common interest in efficient, polite, and stylish communication) and self-perpetuating (people continue doing it because people have done it before as a result of precedent, association, habit, tradition, social pressure, and normative force), but nevertheless arbitrary (alternative regularities could have perpetuated themselves and served the same interests).22 Conventions in this sense are not restricted to language. Driving on the right (or left) is conventional in this sense, as is wearing dresses (or saris). Linguistic conventions are not restricted to semantic or grammatical conventions. The convention of saying ‘Hello, my name is ___’ when meeting someone in person (but not when answering the telephone) is a convention of use that is not a semantic convention (see Morgan 1978: 250). Conversational implicature conventions are also non-semantic conventions of use. Speakers regularly use sentences of the form ‘Some/a S are P’ to implicate “Not all S are P”.23 This regularity is socially useful, serving our interest in efficient communication; it may serve other purposes too, such as politeness, style, and amusement. The regularity is self-perpetuating: we implicate this in part because people before us have done so. And as noted, the practice is arbitrary: we could, and sometimes do, use ‘Some/a S are P’ to implicate other things, such as “No S are P” (when we
Irregular Negations: Implicature & Idiom Theories 117
are using irony), or “All S are P” (when we are engaging in understatement), or “I do not know whether all S are P” (when using a different implicature convention). We could also implicate things no one actually uses it to implicate, such as “Not more than 50% of S are P”. Typically, the lexical conventions assigning meanings to individual words are completely arbitrary because there is no connection between the form and the meaning assigned. Implicature conventions differ because implicatures depend on our seeing some connection between what a sentence means and what we use it to implicate (Davis 1998: §6.5). But they are still arbitrary to a significant extent because the sentence could have been used to implicate many other things. Horn (1989: 343) took it to be true ‘by definition’ that ‘All conversational implicatures are in principle calculable’. Yet he never established calculability before asserting a conversational implicature. At best Horn argued that a practice was ‘pragmatically motivated’ or ‘natural’. In the case of contrary affirmations, for example, Horn (1989: 333) noted that people often choose ‘I don’t like it’ over ‘I dislike it’ because it seems more polite and guarded. He connected this observation with the fact that contrary affirmation is more common with positive terms such as ‘like’, ‘believe’ and ‘good’ than with negative terms such as ‘hate’, ‘doubt’ and ‘bad.’ Horn (1989: 321–4, 337–52, 354ff.) also reviewed extensive evidence that whether a term is subject to contrary affirmation or not is arbitrary and conventional. Intralinguistically, he noted that ‘hope’ and ‘rich’ are positive terms, but do not allow contrary affirmation. Inter-linguistically, while both ‘think’ and ‘believe’ allow contrary affirmation in English, the word for ‘think’ allows contrary affirmation in Hebrew but not the word for ‘believe’, and the reverse is found in Malagasy. Following Morgan (1978), Horn (1989: 345, 352) concluded that the irregular interpretation is an implicature because it is ‘in principle calculable’, but short-circuited because it is conventional and not actually calculated.24 However, if a practice is indeed conventional, then it has a significant element of arbitrariness, and so cannot be calculated. The candid concession Horn makes in his discussion of indirect speech acts is thus especially appropriate for irregular negations: [I]t must be conceded that ascribing some phenomenon to the presence of [a short-circuited implicature] may amount more to labeling than explaining that phenomenon. By pushing the problem of variation in indirect speech act potential back to the pragmatics, we (along with Searle and Morgan) have in some sense reconstructed Sadock’s speech act idiom analysis in different garb, rather than replacing it with a new, improved theory. (Horn 1989: 350) While calculability is not required for an implicature, the non-calculability of the irregular interpretation does undermine one of the main arguments
118
Wayne A. Davis
for I-Implicature theory over the semantic ambiguity theory. Horn (1989: 365, 383) and others cite Grice’s Razor. This modification of Ockham’s famous principle claims that it is more economical to postulate conversational implicatures rather than senses, because conversational implicatures are derivable from independently motivated generalizations, CP and its maxims.25 Burton-Roberts develops the argument quite explicitly: The natural language conjunction or, like the English noun pen, can be understood in two different ways.... Now in the case of pen, this dichotomy of understanding is an irreducible particular fact, not explainable by reference to any general principle. We have no alternative but to observe the dichotomy, as semantic, and record it as such in the lexicon – i.e. acknowledge the existence of a genuine ambiguity.... In the case of or, on the other hand, a general explanation is available for the exclusive understanding.... Not only is that exclusive understanding of or derivable from the inclusive understanding by an extremely plausible conversationally driven calculation involving a general Gricean maxim of quantity, but this same calculation underlies and is required for a host of other expressions, for example the derivation of partitive understanding of some from its existential understanding.... This pragmatic analysis results in a simplification of the semantics.... (Burton-Roberts 1989: 107ff.) The maxim of Quantity enjoins speakers to be as informative as required. Why should we assume that speakers need to convey anything more informative than “p or q or both” when they say ‘p or q’? And if they do have to convey more, why should the requisite additional information be “not both”? All of the following statements and many others are more informative than ‘p or q’ on its inclusive interpretation: (18)
p and q. p rather than q. p or else q. p, or equivalently q. p or q and I have no idea which.
Indeed, ‘p or q’ is conventionally used to implicate the last three propositions.26 The maxim of Quantity therefore provides no basis for predicting that people using ‘p or q’ will mean or imply “p or q but not both” rather than other more informative propositions. We usually do have good reason to infer that a waitress means ‘or’ exclusively. But the maxim of Quantity plays no role here. For ‘You may have steak or lobster but not both’ is neither more nor less informative than ‘You may have steak or lobster or both’ (the disjunctions are embedded). If the maxim of Quantity did explain why people commonly use ‘p or q’ to mean “p or q but not both”, then it should
Irregular Negations: Implicature & Idiom Theories 119
in exactly the same way predict that people commonly use ‘p or q or both’ and ‘p and/or q’ to mean “p or q but not both”. But these are never used with the exclusive interpretation. Sentences have generalized conversational implicatures because sentences with their forms are conventionally used with those implicatures. It is similarly true that words have senses because they are conventionally used with those senses. The postulation of a generalized conversational implicature is therefore no more economical than postulating a sense. The fundamental difference between generalized implicatures and senses is the directness of the use. “Not all S are P” is an implicature of ‘Some/a S are P’ because people conventionally use sentences of that form to mean “Not all S are P” by using them to say and mean “At least some S are P”. The latter is the sense or meaning of ‘Some S are P’ because people conventionally use sentences of that form to mean “At least some S are P” directly, not by using them to mean something else. The fact that meaning something directly is less complex than meaning something indirectly would if anything make the postulation of implicatures less economical than the postulation of senses. We should not postulate an implicature rather than a sense unless we have specific reason to. But that goes without saying. Burton-Roberts provided another argument for the I-Implicature theory based on the fact that conjunctions like those in (2) are contradictory when interpreted as regular negations: [T]he semantic analysis of these examples as being literal contradictions is not only correct, it is surely crucial if we are to provide an explanation of what a speaker must intend by his utterance of them and of the extreme ease with which this intention is recognized. It is clear, in (27) [“Max isn’t not very tall, he’s a dwarf!”] for example, that it is the immediate utterance of the contradiction-inducing second clause that prevents the co-operative hearer from adhering to the analysis indicated by the semantics and analysing “not not very tall” as meaning “very tall.” ... In the face of such blatant contradictions, the co-operative hearer – that is, the hearer who assumes that the speaker is being co-operative – must perform a re-analysis in order to recover from the utterance of these literal contradictions an intention to convey another, non-contradictory idea. This calculation is NECESSITATED by the contradiction induced in each case by the second clause. And it will be FACILITATED [better, ENABLED] by a context that includes an appropriate previous utterance by some other speaker (or an allusion to such a previous utterance). ... This is enabling of the required re-analysis in that it allows (if not obliges) us to construe the utterance of the first clause ... as a metalinguistic use of negation, operating in respect of a quotational allusion to the previous use of, and hence in respect of a mention of, the positive proposition. ... (Burton-Roberts 1989: 117ff.)27
120 Wayne A. Davis
It is generally true that unless we have reason to believe that a speaker intended a contradiction, we charitably strive to find an interpretation of what he says that is consistent. However, charity does not favour the implicature theory over the ambiguity theory. It is true for the same reason that if one interpretation of an ambiguous expression makes what a speaker says contradictory, we try its other meanings. Thus if someone points at the unusual U-shaped instrument the percussionist is playing and says ‘That triangle is not a triangle!’ we would assume that the speaker is not contradicting himself, and infer that the two occurrences of ‘triangle’ have different senses. Moreover, as Seuren (1990: 443) observed, the fact that a sentence is contradictory does not always enable or compel the speaker to find an alternative interpretation. The fact that ‘The car has blue paint: it has no paint’ is blatantly contradictory neither necessitates nor enables recovery of a consistent idea. Seuren’s example is particularly instructive. (19) a. John did not stay till the end: he sat through it all. b. John did not stay till the end: he was never there. Sentence (19b) has a contradictory interpretation, but would normally be interpreted in a way that is consistent. Sentence (19a), in contrast, is unequivocally contradictory. We are not compelled to find a consistent interpretation, because there is none. The interesting question Seuren’s example raises is why a ‘correction’ clause denying one entailment of the root allows a consistent interpretation, but not one denying another entailment. The fact that a consistent interpretation is available when the entailment is presupposed, but not when it is asserted, seems relevant. The general rule that an irregular negation is used to deny an implicature of its root explains why, given that presuppositions are implicatures (section 7). Finally, irregular negations are recognizable even in compounds that are not at all contradictory on their regular interpretation. Consider: (20) a. The next Prime Minister won’t be Wilson, it’ll be Heath or Wilson.28 b. She is not feminine but smart: there is nothing surprising about being both. c. She’s not the pineapple of politeness, Mrs Malaprop, she’s the pinnacle. The negations in these examples are naturally interpreted as irregular despite the fact that the examples are consistent on the regular interpretation of the negations (or perhaps nonsensical in the case of (20c)). Furthermore, we observed above that hearers typically do not even hear the contradictory reading of the sentences in (2). So for several reasons, the fact that irregular negation-correction conjunctions are contradictory cannot be crucial to their interpretation.29
Irregular Negations: Implicature & Idiom Theories 121
6. R-Implicature theory On the I-Implicature theory, the regular interpretation of a negation is its semantic meaning, and irregular interpretations are pragmatically generated implicatures. Grice (1981) and Horn (1989: 486–90, 514) reverse the pragmatic and the semantic for at least one case. The irregular presupposition-cancelling interpretation is the meaning of the negation in (2c), on their view, while the regular presupposition-preserving interpretation is a conversational implicature.30 I call this an R-Implicature theory, since it takes the regular interpretation to be an implicature. The two forms of the implicature theory are compared in table 5.1. In the table, ‘N’ stands for the negation (a sentence), ‘S’ for the speaker, ‘R’ for the regular interpretation, and ‘I’ for an irregular interpretation. Table 5.1
Implicature theories I-Implicature theory
R-Implicature theory
Sentence meaning
N means R but not I
N means I but not R
Irregular use
S says R and means I
S says and means I
Regular use
S says and means R
S says I and means R
R-Implicature theories could select a different irregular interpretation to be the meaning, but the presupposition-cancelling interpretation is the only live option. R-Implicature theory avoids one problem of the I-Implicature theory by predicting that when (2c) has its irregular interpretation, it is not heard as a contradiction; the speaker does not say that Vulcan is non-hot, and thus does not presuppose what the correction clause denies. What the R-Implicature theory gets wrong is the regular use of negations. It predicts that when we use ‘Pluto is not hot’ to mean “Pluto is non-hot”, what we mean is an implicature. All we say is that the proposition that Pluto is hot is not true (or that nothing is both Pluto and not hot). In fact, the two levels of meaning characteristic of implicature are not present here either. We do not mean that Pluto is non-hot by saying something else. That is, we do not merely imply that Pluto is non-hot. A well-known problem for any theory that takes the presuppositioncancelling interpretation to be the only meaning of negations is that it must either deny truth-functionality and double negation or affirm that sentences whose presuppositions are false are themselves false. A more serious problem is that some negations do not have such an interpretation at all. For example, the first clause of (21) unequivocally presupposes that someone assassinated Eisenhower, and thus has no interpretation on which it is true (see Seuren 1990: 447ff.): (21) It was not Oswald who assassinated Eisenhower; no one did.
122 Wayne A. Davis
If (21) had an irregular interpretation, it would mean something true like “The proposition that it was Oswald who assassinated Eisenhower is not true”. ‘Vulcan is not hot’ can also be interpreted as a scalar implicature denial, meaning that Vulcan is not just hot (but super hot). If the presupposition-cancelling interpretation of this sentence is its meaning, however, then it is hard to see how it could have such a scalar implicature. For other sentences meaning “The proposition that Vulcan is hot is not true” (or “Nothing is Vulcan and not hot”) do not implicate “Vulcan is not just hot”. R-Implicature theory has a difficult time explaining why sentences of the form ‘s is not P’ should ever be heard as presupposing or implicating the existence of s.31 If ‘s is not P’ had the meaning of ‘The proposition that s is P is not true’, it would be entailed by ‘s does not exist’ as much as by ‘s exists and is P.’ Neither Quantity nor Relevance discriminates between these.32 Furthermore, ‘s is P’ should not have any implicatures that the allegedly synonymous sentence lacks.
7. Idiom theory Every irregular negation we have examined obeys what I call the first implicature denial rule: ‘Not-p’ is an irregular negation on a given interpretation iff it denies an implicature of ‘p’. Thus ‘The sun is larger than some planets’ implicates “The sun is not larger than all planets”; the negation clause of (2a) denies that implicature. ‘That’s a tomäto’ implicates “That is properly called a ‘tomäto’ ”; the negation clause of (2b) denies that implicature. ‘The chief believes that someone survived’ implicates “It is not the case that the chief believes no one survived”; (8a) denies that implicature. The implicature denial rule holds for presupposition-cancelling denials too, but the situation is more complex. Even on semantic theories, presuppositions are related propositions that speakers commonly mean or imply, but do not say or assert. So presuppositions are one kind of sentence implicature.33 But what presupposition-cancelling denials deny are other implicatures of their roots. For example, ‘Vulcan is not hot’ does not itself deny that Vulcan exists. That presupposition/implicature is denied by the correction clause of (2c). What the irregular negation clause denies is a different implicature, namely, “The proposition that Vulcan is hot is true”, whose denial is entailed by the denial of the presupposition that Vulcan exists. Presupposition-cancelling denials and contrary affirmations are alike in denying implicatures that are part of the meaning of the sentence, and therefore non-conversational. They are ‘conventional implicatures’ in Grice’s (1989: 25–6) technical sense. Scalar and metalinguistic implicature denials differ in denying conversational implicatures that are conventional in the broader, non-technical sense (‘customary’).
Irregular Negations: Implicature & Idiom Theories 123
Sentences usually have more than one implicature. Thus ‘The sun is larger than some planets’ also implicates “Some things the sun is larger than are properly called ‘planets’ ”, and so its negation can also be used to deny that metalinguistic implicature. Whether the negation is a scalar or metalinguistic implicature denial on any given occasion depends on which implicature of its root the speaker intends to deny. A semantic account of irregular negations has to explain how a mere implicature of a component clause could be involved in the meaning and truth-conditions of a larger compound.34 Horn (1989: 370) implies that this task is impossible when he says that ‘conversational implicata by definition are not part of logical form’. However, the task is not to explain how an implicature of ‘p’ could be part of the logical form of ‘p’, but how it could be part of the logical form of a compound containing ‘p’. The explanation will not be in terms of the regular compositional semantic rules, of course. For the meaning of a compositional compound is determined by the meanings, not the implicatures, of its components. Compounds whose meanings are not determined compositionally by the meanings of their components and their grammatical structure are called idioms. By the regular compositional rules, kick the bucket means “strike the bucket with a foot”. But ‘kick the bucket’ has another meaning, namely “die”, that is idiomatic. The facts we have reviewed make it evident that the irregular meanings of negations are idiomatic. The regular meaning of ‘That is not a tomäto’ is the one that is predictable from its grammatical structure and the meanings of its components. The predicate of this sentence means tomato, not thing called a tomäto. Evaluative implicature denials fail to qualify as idioms only because their irregular interpretation is an implicature rather than a meaning. Many idioms are known to have evolved from metaphors, euphemisms, and other figures of speech, which are forms of conversational implicature. 35 For example, cut and run originated as a nautical term in the eighteenth century meaning “cut the anchor cable and leave immediately”. It was subsequently used metaphorically to mean “make a quick escape”, and is now a completely dead metaphor. When the metaphor was alive, the now idiomatic meaning was conveyed indirectly, and the literal meaning directly. As metaphors gradually die through repeated use, less and less attention is paid to the literal meaning. When dead, the formerly metaphorical meaning is conveyed directly, without conveying the literal meaning. People who use the idiom ‘cut and run’ today are not saying that the subject is cutting the anchor and sailing away, and are rarely even thinking about ships. Similarly, sleep with was first used in the tenth century to imply sexual relations, and today is generally used with no thought about sleeping. It is still considered more polite in some circles than ‘have sex with’, but there is no indirection. Idioms also retain their literal meaning, of course, and thus are semantically ambiguous. Thus parents might say ‘Johnny was scared, so he slept with us last night’ with no sexual implication at all.
124 Wayne A. Davis
The lexicalization of generalized conversational implicatures is a similar and even more common phenomenon. To take just a few examples, the causal meaning of ‘since’ evolved from an implicature in Old English, where ‘since’ had only its temporal sense. The word ‘homely’ originally had the meaning “pertaining to the home”, acquiring the metaphorical meaning “plain, unadorned” in the fourteenth century, and eventually becoming a euphemism meaning “physically unattractive” in American English.36 Most readers of this article will have lived through the lexicalization of the computer virus metaphor. Implicature involves using words to mean one thing and thereby mean another. It involves both a direct meaning and an indirect meaning. As long as this indirection exists, the second meaning will remain an implicature rather than a sense. When it becomes conventional to express the second meaning directly, it becomes a second sense (Davis 2003: Part II, and 2005: chapter 5). I hypothesize that the irregular meaning of negations arose by this process. Given that speakers standardly used sentences of the form ‘s Vs some/a O’ to mean “s Vs just some O”, it would have been natural for them to begin using ‘s does not V some/a O’ to mean “s does not V just some O”.37 It is also plausible that with repetition, this usage became first conventional and then direct. When that happened, sentences of the form ‘S does not V some/a O’ acquired the irregular sense in addition to their regular sense. Since the irregular interpretation of evaluative implicature denials remains indirect, it can be classified as a generalized conversational implicature. The indirection or two levels of meaning definitive of implicature is absent in the case of scalar and metalinguistic implicature denials, presuppositioncancelling denials, and contrary affirmations. They are therefore meanings. These idioms could well have arisen as what were generalized implicatures, like today’s evaluative implicature denials, ‘died’ through repeated use. Direct irregular negations are idioms because, though conventional and direct, their meaning is not predictable from the meanings of their components and their grammatical structure. They are partially, but not fully, compositional. Their irregular meaning depends on the meanings of their parts, but is not determined by them. Irregular negations deny implicatures of their roots, which are not determined by the meanings of their roots. The idioms we have focused on here are listed in table 5.2, from which we can read off the conventions that give rise to the irregular meanings. The idiom theory is schematized in table 5.3. One of the typical properties of idioms is that we perceive a connection between their literal compositional meaning and their idiomatic meaning. ‘Sleep with’ is a paradigm example. The conventions assigning idiomatic meanings to sentences are like conversational implicature conventions, and unlike typical lexical conventions, in not being completely arbitrary. We always perceive some antecedent connection between the idiomatic meaning or implicature and the literal meaning (Davis 1998: §6.5). Given that
Irregular Negations: Implicature & Idiom Theories 125 Table 5.2
Irregular meaning conventions Form
Irregular meaning
Scalar implicature denial
s does not V some/a O
s does not V just some O
Metalinguistic implicature denial
s is not P
s is not properly called ‘P’
Presupposition-cancelling denial
s is not P
The proposition that s is P is not true
Contrary affirmation
s does not believe that p
s believes that not-p
Table 5.3
Idiom theory Idiom theory
Sentence meaning
N means both R and I
Irregular use
S says and means I
Regular use
S says and means R
they are implicature denials, the same is clearly true of the irregular and regular interpretations of negations. We noted in section 1 that ambiguity, limited substitutivity of synonyms, and partial compositionality are common features of irregular negations. These are all essential features of idioms in general. The idiom theory can thus easily respond to the following objection to the ambiguity theory: [T]he incorporation of a further (presupposition-cancelling) means of negation within the semantics leaves totally unexplained the special, marked and, in Kempson’s word, unnatural character of the negations it is designed to account for. (Burton-Roberts 1989: 100)38 Semantic, grammatical, and even phonological irregularity is part of what makes idioms idioms. Consider ‘He doesn’t know which end is up’. This is partially but not completely compositional. Replacing ‘doesn’t’ with ‘does’, ‘can’t’, or ‘may not’ changes the meaning in predictable ways; but replacing ‘up’ with ‘down’, or ‘end’ with ‘terminus’, does not. Standard transformations produce dubious results (e.g., ‘Which end is up is something he doesn’t know’ or ‘He fails to know which end is up’). Standard entailments fail (‘He either forgot or never learned’). And only some components can be emphasized (‘He doesn’t know which end is up’ is okay, but not ‘He doesn’t know which end is up’ or ‘He doesn’t know which end is up’.) Horn (1989: 392) sought to explain why marked negation fails to incorporate morphologically.39 ‘Immaterial’, for example, is synonymous with
126
Wayne A. Davis
‘not material’. Why then does ‘Angels are immaterial’ have only the regular meaning of “Angels are not material”, not the irregular meaning? Horn’s explanation was that the irregular meaning is metalinguistic: ‘Angels are not material’ means something like ‘ “Angels are material” is not assertable’. In the latter, the ‘not’ does not combine with ‘material’ because one is inside quotation marks, and the other is outside. The words are occurring at different levels, ‘not’ in the metalanguage, ‘material’ in the object language. This explanation is satisfactory for (2b1), which is genuinely metalinguistic, as captured by (9b1). But it fails for (2a1), which is not (section 6). The idiom theory can explain why morphological incorporation fails in some irregular negations but not others. In marked negations, the expression composed of ‘not’ and the predicate ‘P’ is not a compositional unit: it does not express a concept composed of the concepts expressed by ‘not’ and ‘P.’ That is, it does not mean “non-P.” For different reasons, this holds for ‘not larger than some planets’ in (2a1), ‘not a tomäto’ in (2b1), and ‘not hot’ in (2c1), as indicated by (9a1–c1). The word ‘immaterial,’ however, expresses the concept expressed by ‘non-material.’ Horn’s explanation worked for metalinguistic implicature denials, because their expressing concepts occurring at different levels is one reason why two terms do not combine to form a compositional unit. But it is not the only way. Contrary affirmations are exceptions that prove the rule. Despite being irregular, they allow incorporation because ‘not V+’ expresses the contrary concept “V−”, as we saw in section 2. Thus ‘does not believe’ often means “disbelieves”, and ‘is not good’ often means “is bad”. Bach (1994: 153–4) would argue that an irregular interpretation I is not a meaning of negation N because N does not literally say I, and because the use of N to say I is not fully explicit. It does seem fair to say, for example, that (2a) is literally a contradiction, and that someone who uses ‘The sun is not larger than some planets’ to say that it is larger than all planets, not just some, is not being fully explicit. Idioms have an idiomatic meaning in addition to their literal meaning, however. ‘He kicked the bucket’ literally means “He struck the bucket with his foot”, but it also means “died” in English. Similarly, ‘English horn’ literally means “horn that is English” even though it has a more common meaning on which it denotes a musical instrument that is ‘neither English nor a horn but a tenor oboe’ (Latham 2002: 309). In general, I believe we are using ‘literal’ in these cases to describe a fully compositional meaning – a meaning that results from the meanings of an expression’s parts and the syntactic rules of the language.40 Since an idiomatic meaning results from a convention to use a compound with a meaning other than its compositional meaning, it is a non-literal meaning. Idioms also show that expressions may have a meaning without being the most explicit way of conveying that meaning. ‘Slept with’ illustrates this nicely. In general, a sentence meaning that is not fully compositional is not fully explicit, because a fully compositional way of saying the same thing is more explicit.
Irregular Negations: Implicature & Idiom Theories 127
A common reason for rejecting ambiguity claims is what I call ‘gut incredulity’. People simply find it hard to believe – absurd even – that sentences have so many meanings.41 The question is, why should this be hard to believe? Everyone acknowledges that all the different readings of negations we have identified are possible, and that they are conventional. Once this much is acknowledged, the additional claim that the readings are meanings should be just a question of detail. Whether a reading is a meaning or a conventional conversational implicature depends on whether it is conventional to mean it directly or indirectly.42 The idiom theory differs from other ambiguity theories on the source of the irregular meaning. Ambiguity theories that attribute the ambiguity to a lexical ambiguity in the word ‘not’ or a syntactic ambiguity arising from scope differences are implausible for the reasons indicated in sections 1 and 3. On the idiom theory, the ambiguity is neither lexical nor scopal. The irregular meaning of a negation does not result compositionally from the meanings of ‘not’ and its root. Instead, the construction as a whole acquired additional meanings as implicatures became conventional and then direct. The ambiguity is not a standard syntactic ambiguity because the idiomatic meanings are not fully compositional.
8. Idiom constructions and productivity I have developed a foundational theory in which word meaning is defined in terms of conventional speaker-meaning, and speaker-meaning is defined in terms of intention and the expression of thoughts and ideas (Davis 2003, 2005). Simplifying greatly, the base clause of my recursive definition specifies that expressions have a meaning μ if they are conventionally and directly used to mean μ – that is, to express the idea μ. The compositional meanings of complex expressions are provided by the recursion clause, which specifies that expression structures express certain idea structures if they are conventionally and directly used to do so. The base clause accounts for the meanings of morphemes, including both the literal and the metaphorical meanings of ‘virus’, along with the meanings of completely non-compositional idioms like ‘kicked the bucket’. I have argued here that negations are ambiguous because the meaning of ‘not’ together with the meaning of ‘p’ generates the regular, fully compositional meaning of ‘not-p’. The irregular, partly compositional, idiomatic meanings result from special conventions to use more specific negative sentence patterns to directly express ideas distinct from the regular negation. Two interrelated aspects of the idiom theory need further development. First, the term ‘idiom’ is traditionally used for an expression: a particular phrase or clause like ‘kicked the bucket’. This compound is an idiom because it has a meaning (“died”) that is not predictable from the meanings of its components and its grammatical structure. I also apply the term ‘idiom’ to
128 Wayne A. Davis
constructions like ‘s does not V some O’, which have many instances. How can a general construction be an idiom? Second, scalar implicature denials are not limited to sentences of the form ‘s does not V some O’. The process of forming scalar implicature denials is highly productive. How is that possible if they are idioms?43 I describe constructions as idioms when their instances are idioms. Constructions are forms (structures, patterns), which many expressions share. They define a general sentence type. Some forms are purely syntactic, such as ‘NP VP’. Other constructions are defined in part by their components having particular meanings. ‘Hot N’ is a partially semantic form if the occurrence of ‘hot’ in any instance must have its taste rather than its temperature meaning. Then ‘hot soup’ has that form on a particular occasion if it means “spicy soup”, but not if it means “high temperature soup”. If forms were exclusively semantico-syntactic, then constructions could not be idioms. For the meanings of their instances would be predictable from their syntax and the meanings of their components. If negations had two different syntactic forms, one determining the regular meaning and other determining an irregular meaning, then irregular negations would not be idioms any more than regular negations are. The same would be true if the forms were differentiated partly by different meanings of ‘not’. But forms need not be defined by syntax and semantics alone. Pragmatic features may define the type. We have been focusing, for example, on sentences that not only contain the word ‘some’ with its standard existential meaning, but also have a scalar implicature. ‘Johnny ate some of the cookies’ has the relevant form. ‘The U.S. used some atomic bombs in 1945’ does not have such an implicature, and so would not have the relevant form. We are concerned, as I mentioned above, with ‘generalized’ rather than ‘particularized’ implicatures – what a sentence implicates without reference to a context rather than what a speaker uses it to implicate on a particular occasion. What a sentence implicates is determined by what speakers conventionally use instances of its semantico-syntactic form to implicate. In describing the convention whereby ‘Johnny did not eat some of the cookies’ gets its irregular, scalar implicature denial meaning, I focused on the form ‘s does not V some/a O’. This form has an idiomatic meaning because it is conventional for speakers to use instances to directly express the proposition expressed by the related sentence ‘s does not V just some O’. The irregular meaning of a sentence like ‘Johnny did not eat some of the cookies’ fails to be predictable from the meanings of its components and its grammatical structure not only because none of its components means “just”, but also because its having that meaning depends on the fact that its root (‘Johnny ate some of the cookies’) has a scalar implicature, namely, “Johnny did not eat all of the cookies”. The relevant form is partly pragmatic. Since the irregular meaning of ‘s does not V some/a O’ is not predictable from the meanings of its components and its grammatical structure alone, it is an idiom.
Irregular Negations: Implicature & Idiom Theories 129
This explanation accounts for the irregular meanings of an infinity of novel sentences. But the process of forming scalar implicature denials is even more productive. The negations below, for example, all have an irregular interpretation as a scalar implicature denial, but none has the specific form s does not V some/a O: (22)
It is not the case that some passengers died. It is not the case that if exactly three died, then some died. The boys did not miss some points on the exam. Some of the cookies weren’t eaten by Johnny.
Geurts’s thesis that ‘some’ also means “just some” is superior in this respect because it easily accounts for all these scalar implicature denials and more. To account for the full generality of scalar implicature denial, I need to identify a more general form that is shared by all negations with a scalar implicature denial interpretation. Let ‘Not-(...some__†)’ designate the characteristic form. Given a sentence of the form ‘Not-(...some__†)’, let ‘Not-(... just some__)’ be the result of replacing ‘some’ by ‘just some’. Then we can describe the general convention as follows: (23)
The scalar implicature denial convention for ‘some’: It is conventional to use an expression of the form ‘Not-(...some__†)’ to mean “Not-(... just some__)”.
To say that a sentence is used to mean “Not-(...just some__)” is to say that it is used to directly express the proposition expressed by ‘Not-(...just some__)’ on its regular interpretation. The question now is how to define the characteristic form ‘Not-(...some__†)’. I believe the following is at least a first approximation: (24) A sentence is of the form ‘Not-(...some__†)’ iff it is the negation of some sentence Σ such that Σ contains ‘some’ and has an existential meaning on which it implicates the proposition expressed by the sentence Σ ′ that is obtained from Σ by replacing ‘some’ with ‘just some’. The form ‘Not-(...some__†)’ is partly pragmatic because it is defined in terms of an implicature of its root. Sentence (25) is an expression of this form: (25) The sun is not larger than some planets. For (25) is the negation of ‘The sun is larger than some planets’, which contains ‘some’ and has a salient existential meaning. On that meaning, the root of (25) implicates the proposition expressed by ‘The sun is larger than just some planets’. So in accord with the scalar implicature denial convention,
130 Wayne A. Davis
(25) can be used to mean “The sun is not larger than just some planets”. Any expression of the form ‘s does not V some/a O’ has the form ‘Not-(... some__†)’. The same goes for ‘It is not the case that some N V’ and ‘The Ns do not V some/a O on the M’. Not all negations of sentences containing ‘some’ are of the form ‘Not-(... some__†)’. Consider: (26)
The air has some mass. The U.S. used some atomic bombs in 1945. The president made some mistakes.44
(27) The sun is larger than at least some planets. The sun is larger than just some planets. The sun is larger than some but not all planets. The sentences in (26) and (27) do not have scalar implicatures. Consequently their negations are not of the form ‘Not-(...some__†)’ and do not have an irregular interpretation. There is another way to describe the scalar implicature denial convention: using ‘not some’ as elliptical for ‘not just some’ in expressions of the form ‘Not-(... some__†)’. Just as (28) is elliptical for (29), so (30) is elliptical for (31) on its irregular interpretation: (28) Bob gave the apple to Mary and the peach to Jane. (29) Bob gave the apple to Mary and Bob gave the peach to Jane. (30) The sun is not larger than some planets. (31) The sun is not larger than just some planets. ‘Bob gave the apple to Mary and the peach to Jane’ expresses the same proposition as ‘Bob gave the apple to Mary and Bob gave the peach to Jane’ even though ‘the peach’ does not mean “Bob gave the peach”. Similarly, ‘The sun is not larger than some planets’ expresses the same proposition as ‘The sun is not larger than just some planets’ even though ‘not some’ does not mean “not just some”. Sentence (28) is not an idiom because its meaning is predictable from its syntax and the meanings of its components. The irregular meaning of sentence (30) is not similarly predictable because it depends on an implicature of its root. Metalinguistic implicature denials involve ellipsis in the same way. Why shouldn’t we adopt Geurts’s view that ‘some’ means “just some” in (30) when it is irregular? First, ‘some’ is still in opposition to ‘all’, ‘no’, ‘many’ and so on. We understand ‘some planets’ as weaker than ‘all planets’ and contradictory to ‘no planets’. That is, (30) is understood as the denial of ‘The sun is larger than some planets’ when understood as entailed by ‘The
Irregular Negations: Implicature & Idiom Theories 131
sun is larger than all planets’ and contradictory to ‘The sun is larger than no planets’. Second, if ‘some’ meant “just some” in (30), (30) would not be heard as irregular. Instead, (30) would be heard as ambiguous with two regular interpretations. Third, there is no more reason to think that ‘some’ acquires the meaning “just some” in (30) than there is to think that ‘not’ acquires the meaning “not just”, or that ‘the peach’ in (28) acquires the meaning ‘Bob gave the peach’. Fourth, if ‘some’ means “just some” in a large and open class of English sentences, then why doesn’t it mean that in English? Why isn’t this meaning ‘encoded in its lexical entry’? Geurts cannot say that a meaning is encoded in a word’s lexical entry only if it has that meaning in every sentence. For that would entail, given Geurts’s assumptions, that ‘some’ does not have a meaning in English. According to Geurts, the meaning ‘some’ has in ‘s does not V some/a O’ on its regular interpretation is not the meaning it has in that sentence on its irregular interpretation. Geurts’s thesis seems to collapse into the discredited ambiguity theory. If ‘some’ were ambiguous, it should mean either “at least some” or “just some” in a sentence like (32): (32) If the sun is larger than all planets, then it is larger than some planets. (33) If the sun is larger than all planets, then it is not larger than some planets. But (32) does not have the contradictory meaning “If the sun is larger than all planets, then it is larger than just some planets”. And (33) does not have a tautological reading. The ambiguity theory would make the process of forming scalar implicature denials too productive. Geurts’s theory can avoid overgenerating by denying that ‘semantic transfer’ occurs in a sentence like (32). But without an explanation of how semantic transfer differs from ambiguity, and an account of why semantic transfer does not occur in (33), or in the negations of the sentences in (26) or (27), Geurts’s theory is just ad hoc. Since expressions that can be used elliptically are not thereby ambiguous, my theory does not entail that ‘not some’ means “not just some” as well as “not at least some”, nor a fortiori that ‘some’ means “just some” as well as “at least some”. My account makes presupposition-cancelling denials like contrary affirmations in not being elliptical. ‘Vulcan is not hot’ is not elliptical for ‘The proposition that Vulcan is hot is not true’ or even ‘Vulcan is hot is not true’. Simply omitting words from the second or third sentence will not produce the first. The ‘not’ also needs to be moved to a different clause. Presupposition-cancelling denials are also like contrary affirmations in another respect: the relevant implicatures of their roots are non-detachable entailments, and thus part of sentence meaning. I believe this is why contrary affirmations and presupposition-cancelling denials seem intuitively to be more meaning-like than scalar and metalinguistic implicature denials.
132 Wayne A. Davis
9. Questions for further study It is my hypothesis that direct irregular negations are idioms that evolved from a variety of conversational implicatures the way other idioms evolved from metaphors. Since their idiomatic meaning is in addition to their regular compositional meaning, they are semantically ambiguous the way dead metaphors are. If this is correct, a number of questions arise. When and why did such constructions acquire their irregular sense? When did something that was implicated come to be meant directly? Why did an idiom arise rather than a new word meaning? Why in some cases didn’t the implicature of the root similarly become semanticized? Did the irregular conventions arise independently for each negation form, or generalize from one to another? In general, how can we best describe and explain a system of idioms? These are questions for historical linguistics as well as pragmatic and semantic theory. The fact that conventions are arbitrary means that the answers may cite the vagaries of historical contingency rather than systematic linguistic or social factors.
Notes 1. I would like to thank Bart Geurts, Kent Bach, Stephen Gross, an anonymous referee, and participants in the 2008 University of Michigan Philosophy and Linguistics Workshop for helpful comments on earlier versions of this article. 2. For more on obversion, see Davis (1986: chapter 5). 3. ‘ä’ represents the vowel sound in ‘ah’ and ‘ma,’ ‘a ˉ ’ the vowel sound in ‘pay’ and ‘ate’. 4. ‘F!x’ means “x is uniquely F”, i.e. Fx & ∀y(Fy ⊃ y = x). 5. When s does not exist, ‘s is P’ and ‘s is not P’ are jointly exhaustive only if the latter is irregular. 6. See Seuren (1988: 183, 1990: 443) and Geurts (1998: 279). Contrast Grice (1981: 271). 7. Horn (1989: 368, 374–5, 392–413). See also Kempson (1986: 88), Burton-Roberts (1989: 118), Seuren (1990: 449–52), Chapman (1996: 390–1), Israel (1996: 621n.1), Carston (1998: 332ff.), Geurts (1998: 275, 278–80, 303) and van der Sandt (2003: §7). 8. Ladd (1980: chapter 7) shows that the same difference in intonation marks other differences in meaning, such as the different interpretations of ‘All S are not P’. 9. See my ‘Irregular negations and denial’ (forthcoming). 10. For a comprehensive examination of contrary affirmations, see Horn (1989: 340ff., 352–6). 11. See Atlas (1989: 71) and Horn (1989: 365). 12. See Seuren (1988: 191, 1990: 444), Horn (1989: 385, 424–5, 1990: 497–501) and Bach (1987: 71, 1994: 153–4). 13. See Levinson (1983: 171). 14. An equivalent but metalinguistic formulation would be ‘Vulcan is hot’ is not true: Vulcan does not exist. See Linebarger (1981), Burton-Roberts (1989: 120–1), Horn (1989: §6.5.1). Contrast Burton-Roberts (1989: 118ff.). Horn correctly observes that Linebarger’s analysis does not work for implicature denials.
Irregular Negations: Implicature & Idiom Theories 133 15. See also Gazdar (1979: 65–6), Levinson (1983: 201), Burton-Roberts (1989, 1999), Horn (1989: 366), Atlas (1989: 69), Seuren (1990: 449ff.), van der Sandt (1991: 333) and Carston (1996: 327). Contrast Seuren (1988: 196ff., 222). 16. See Grice (1968, 1969), Kripke (1977), Davidson (1978), Bach and Harnish (1979: 20–3) and Davis (2003: §7.9). 17. I defend this definition against Saul’s (2001, 2002) objections in Davis (2007). 18. For example, sentences of the form ‘p but q’ have a contrastive implicature as part of their meaning that distinguishes them from ‘p and q’. Grice (1989: 25–6) referred to such non-conversational implicatures as ‘conventional implicatures’. Generalized conversational implicatures are conventional not in Grice’s narrow technical sense, but in the broader non-technical sense explained in the text that applies to everything from linguistic meanings to dress customs. 19. Contrast Burton-Roberts (1989: 114), van der Sandt (1991: 332–3) and Carston (1996: 311), who profess to find Horn’s thesis confused or hopelessly vague. 20. See Horn (1989: 372), Burton-Roberts (1989: 109, 1999: 349). 21. Horn (1992b), Carston (1996: 312) and Chapman (1996: 395, 401–2). 22. I modified Lewis’s (1969) definition to make it fit paradigm examples better in Davis (1998: chapter 5, 2003: chapter 9). 23. The convention that gives ‘Some cats are black’ its scalar implicature is actually more specific. Contrast ‘Some typos are on the first page’ with ‘Some of the typos are on the first page’ and ‘Some typos are harmless’. Except in special contexts, the first lacks the “not all” implicature. I believe this is because the result of replacing ‘some’ by ‘all’ has no salient interpretation. Whereas ‘All of the typos are on the first page’ and ‘All typos are harmless’ have a salient interpretation in any context, ‘All typos are on the first page’ does not except in special contexts in which a limited class of typos is of interest. I use ‘Some/a S are P’ to represent the relevant form. 24. See Bach’s (1995: 683) notion of an inference ‘compressed by precedent’. 25. See for example Wilson (1975: 99), Kempson (1975: 142), Grice (1978: 44–7), Sadock (1981: 258), Wilson and Sperber (1981: 155), Levinson (1983: 97–100, 132), Bach (1987: 69, 77–9), Burton-Roberts (1989: 107ff.), Horn (1989: 213–14, 377ff., 1992a: 263, 266) and Neale (1990: 80–1, 90–1, 1992: 535). Contrast Geurts (1998: 298) and Davis (1998: 18–27). 26. See Ball (1986), Horn (1989: 378) and Davis (1998: 146–7). 27. See Horn (1989: 391, 444), Carston (1998: 340, 1999: 374), Burton-Roberts (1999: 355, 360) and Levinson (2000: 214). 28. This example comes from Grice via Wilson (1975: 150), Burton-Roberts (1989: 116ff.) and Horn (1989: 377ff., 396). Example (20c) is from Burton-Roberts (1999: 362). Note that (20a) shows that irregular negations need not conform to Disjunctive Syllogism. Neither the speaker nor the audience would take ‘It will be Heath’ to follow. 29. See Wilson and Sperber’s (1981: 160–2) critique of Grice’s thesis that the obvious falsity of what is literally said is somehow crucial to the existence or interpretation of figures of speech. See also Levinson (2000: 215–16). 30. See also Kempson (1975), Atlas (1975, 1979: 252–4), Gazdar (1979), Boër and Lycan (1976) and Horn (1990: 496ff.). 31. For attempts, see Wilson (1975: 99–100, 106), Atlas (1975, 1979: 273), Kempson (1975: 178–9), Boër and Lycan (1976: 278); but contrast 60–1), Grice (1981: 273–6), Levinson (1983: 218, 222), Lycan (1984: 84) and Carston (1999: 368–71). 32. See Kroch (1972) and Burton-Roberts (1999: 357).
134 Wayne A. Davis 33. Whether presuppositions are conversational or semantic implicatures is a complex and controversial issue we need not address here. 34. See Horn (1989: 370, 384), Burton-Roberts (1989: 122) and Levinson (2000: 212–13). Contrast van der Sandt (1991: 332–4), Geurts (1998: 294) and Levinson (2000: 214). 35. Ammer (1997), Siefring (2004) and ‘Euphemism’ (Wikipedia) are filled with fascinating examples. See also Sadock (1972), Grice (1975: 58), Searle (1975: 76ff.), Morgan (1978), Horn (1989: 344–5), Hopper and Traugott (1993: 75–93), Cowie (1994) and Davis (1998: chapter 6, 2003: chapter 8). Contrast Bach (1987: 71, 1994: 144). 36. Hopper and Traugott (1993: 76–7), The Online Etymology Dictionary and Geurts (1998: 297). 37. As explained in note 23, ‘/a’ indicates that a sentence containing ‘some’ is an instance of the form only if the result of replacing it with ‘all’ has a salient interpretation. 38. See also Kempson (1975), Horn (1989: 487) and Burton-Roberts (1997: 68, 1999: 348). Compare and contrast Carston (1998: 339, 346–90), Seuren (1990: 439). 39. See also Horn (1985, 1990: 497), Burton-Roberts (1989: 235–6) and Carston (1996: 322). 40. ‘Literal’ is also used to describe the etymology of a word, as when it is said that ‘hippopotamus’ literally means “river horse”. In this case, the literal meaning is not something the word means in English at all. 41. See for example Katz (1972: 92), Sperber and Wilson (1986: 188), Kittay (1987) and Bach (1994: 150). Contrast Davis (2003: §10.6). 42. I defend the claim that negations are ambiguous rather than general in ‘Irregular negations: explicature theory’ (forthcoming). 43. I am grateful to Barbara Abbott and Zoltan Szabó for pushing these questions at the 2008 Michigan Philosophy and Linguistics Workshop. 44. As explained in note 23, these instances of ‘s Vs some O’ are not of the more specific form ‘s Vs some/a O’.
References Ammer, C. (1997) The American Heritage Dictionary of Idioms (Boston: Houghton Mifflin Co.). Atlas, J. D. (1975) Presupposition: A semantico-pragmatic account. Pragmatics Microfiche, 1.4, D13-G9 Atlas, J. D. (1979) How linguistics matters to philosophy: Presupposition, truth, and meaning. In C.-K. Oh and D. A. Dinneen (eds.) Syntax and Semantics, 11: Presupposition (New York: Academic Press), pp. 265–81. Atlas, J. D. (1989) Philosophy Without Ambiguity (Oxford: Oxford University Press). Bach, K. (1987) Thought and Reference (Oxford: Oxford University Press). Bach, K. (1994) Conversational impliciture. Mind and Language, 9: 124–62. Bach, K. (1995) Standardization vs. conventionalization. Linguistics and Philosophy, 18: 677–86. Bach, K. and Harnish, R. (1979) Linguistic Communication and Speech Acts (Cambridge, MA: MIT Press). Ball, C. (1986) Metalinguistic disjunction. Penn Review of Linguistics, 10: 1–15. Boër, S. and Lycan, W. (1976) The Myth of Semantic Presupposition (Bloomington, IN: Indiana University Linguistics Club).
Irregular Negations: Implicature & Idiom Theories 135 Burton-Roberts, N. (1989) On Horn’s dilemma: Presupposition and negation. Journal of Linguistics, 25: 95–125. Burton-Roberts, N. (1997) On preservation under negation. Lingua, 101: 65–88. Burton-Roberts, N. (1999) Presupposition-cancellation and metalinguistic negation: A reply to Carston. Journal of Linguistics, 35: 347–64. Carston, R. (1996) Metalinguistic negation and echoic use. Journal of Pragmatics, 25: 321–40. Carston, R. (1998) Negation, ‘presupposition’, and the semantics/pragmatics distinction. Journal of Linguistics, 34: 309–50. Carston, R. (1999) Negation, ‘presupposition’ and metarepresentation: A response to Noel Burton-Roberts. Journal of Linguistics, 35: 365–89. Chapman, S. (1996) Some observations on metalinguistic negation. Journal of Linguistics, 32: 387–402. Cowie, A. P. (1994) Phraseology. In R. Asher (ed.) The Encyclopedia of Language and Linguistics (Oxford: Pergamon Press), pp. 3168–71. Davidson, D. (1978) What metaphors mean. In S. Sacks (ed.) On Metaphor (Chicago: University of Chicago Press) pp. 29–46. Reprinted in S. Davis (ed.) Pragmatics: A Reader (Oxford: Oxford University Press, 1991), pp. 495–506. Davis, W. A. (1986) An Introduction to Logic (Englewood Cliffs, NJ: Prentice-Hall). Davis, W. A. (1998) Implicature: Intention, Convention, and Principle in the Failure of Gricean Theory (Cambridge: Cambridge University Press). Davis, W. A. (2003) Meaning, Expression, and Thought (New York: Cambridge University Press). Davis, W. A. (2005) Nondescriptive Meaning and Reference (Oxford: Oxford University Press). Davis, W. A. (2007) How normative is implicature? Journal of Pragmatics, 39: 1655–72. Gazdar, G. (1979) Pragmatics: Implicature, Presupposition, and Logical Form (New York: Academic Press). Geurts, B. (1998) The mechanisms of denial. Language, 74: 274–307. Grice, H. P. (1957) Meaning. Philosophical Review, 66: 377–88. Grice, H. P. (1968) Utterer’s meaning, sentence meaning, and word meaning. Foundations of Language, 4: 225–42. Reprinted in H. P. Grice (1989) Studies in the Way of Words (Cambridge, MA: Harvard University Press), pp. 117–37. Grice, H. P. (1969) Utterer’s meaning and intentions. Philosophical Review, 78: 147–77. Reprinted in H. P. Grice (1989) Studies in the Way of Words (Cambridge, MA: Harvard University Press), pp. 86–116. Grice, H. P. (1975) Logic and conversation. In P. Cole and J. Morgan (eds.) Syntax and Semantics, 3: Speech Acts (New York: Academic Press). Reprinted in H. P. Grice (1989) Studies in the Way of Words (Cambridge, MA: Harvard University Press), pp. 22–40. Grice, H. P. (1978) Further notes on logic and conversation. In P. Cole (ed.) Syntax and Semantics, 9: Pragmatics ( New York: Academic Press), pp. 113–28. Reprinted in H. P. Grice (1989) Studies in the Way of Words (Cambridge, MA: Harvard University Press), pp. 41–57. Grice, H. P. (1981) Presupposition and conversational implicature. In P. Cole (ed.) Radical Pragmatics (New York: Academic Press), pp. 183–98. Reprinted in H. P. Grice (1989) Studies in the Way of Words (Cambridge, MA: Harvard University Press), pp. 269–82. Hopper, P. J. and Traugott, E. C. (1993) Grammaticalization (Cambridge: Cambridge University Press).
136 Wayne A. Davis Horn, L. R. (1985) Metalinguistic negation and pragmatic ambiguity. Language, 61: 121–74. Horn, L. R. (1989) A Natural History of Negation (Chicago: University of Chicago Press). Horn, L. R. (1990) Showdown at truth-value gap: Burton-Roberts on presupposition. Journal of Linguistics, 26: 483–503. Horn, L. R. (1992a) Pragmatics, implicature, and presupposition. In W. Bright (ed.) International Encyclopedia of Linguistics, Vol. 2 (New York: Oxford University Press), pp. 260–6. Horn, L. R. (1992b) The said and the unsaid. Ohio State University Working Papers in Linguistics, 40: 163–72. Israel, M. (1996) Polarity sensitivity as lexical semantics. Linguistics and Philosophy, 19: 619–66. Katz, J. J. (1972) Semantic Theory (New York: Harper and Row). Kempson, R. (1975) Presupposition and the Delimitation of Semantics (Cambridge: Cambridge University Press). Kempson, R. (1986) Ambiguity and the semantics-pragmatics distinction. In C. Travis (ed.) Meaning and Interpretation (Oxford: Basil Blackwell), pp. 77–103. Kittay, E. (1987) Metaphor: Its Cognitive Force and Linguistic Structure (Oxford: Oxford University Press). Kripke, S. (1977) Speaker reference and semantic reference. Midwest Studies in Philosophy, 2: 255–78. Kroch, A. (1972) Lexical and inferred meanings for some time adverbs. Quarterly Progress Report of the Research Lab of Electronics, MIT, 104. Ladd, D. (1980) The Structure of Intonational Meaning (Bloomington, IN: Indiana University Press). Latham, A. (ed.) (2002) The Oxford Companion to Music (Oxford: Oxford University Press). Levinson, S. C. (1983) Pragmatics (Cambridge: Cambridge University Press). Levinson, S. C. (2000) Presumptive Meanings: The Theory of Generalized Conversational Implicature (Cambridge, MA: MIT Press). Lewis, D. (1969) Convention (Cambridge, MA: Harvard University Press). Linebarger, M. (1981) The grammar of negative polarity. PhD Dissertation. MIT. Lycan, W. (1984) Logical Form in Natural Language (Cambridge, MA: MIT Press). Morgan, J. L. (1978) Two types of convention in indirect speech acts. In P. Cole (ed.) Syntax and Semantics, 9: Pragmatics (New York: Academic Press), pp. 261–80. Reprinted in S. Davis (ed.) (1991) Pragmatics: A Reader (Oxford: Oxford University Press), pp. 242–53. Neale, S. (1990) Descriptions (Cambridge, MA: MIT Press). Neale, S. (1992) Paul Grice and the philosophy of language. Linguistics and Philosophy, 15: 509–59. Nunberg, G. (1978) The Pragmatics of Reference (Bloomington, IN: Indiana University Linguistics Club). Sadock, J. M. (1972) Speech act idioms. In J. Peranteau, J. Levi and G. Phares (eds.) Papers from the Eighth Regional Meeting of the Chicago Linguistic Society (Chicago: Chicago Linguistic Society), pp. 329–39. Sadock, J. M. (1981) Almost. In P. Cole (ed.) Radical Pragmatics ( New York: Academic Press), pp. 257–72. Saul, J. (2001) Review of Implicature: Intention, Convention, and Principle in the Failure of Gricean Theory by Wayne Davis. Noûs, 35: 630–41.
Irregular Negations: Implicature & Idiom Theories 137 Saul, J. (2002) Speaker meaning, what is said, and what is implicated. Noûs, 36: 228–48. Searle, J. (1975) Indirect speech acts. In P. Cole and J. L. Morgan (eds.) Syntax and Semantics, 3: Speech Acts (New York: Academic Press), pp. 59–82. Reprinted in S. Davis (ed.) (1991) Pragmatics: A Reader (Oxford: Oxford University Press), pp. 265–77. Seuren, P. A. M. (1988) Presupposition and negation. Journal of Semantics, 6: 176–228. Seuren, P. A. M. (1990) Burton-Roberts on presupposition and negation. Journal of Linguistics, 26: 425–53. Siefring, J. (ed.) (2004) The Oxford Dictionary of Idioms (Oxford: Oxford University Press). Sperber, D. and Wilson, D. (1986) Relevance: Communication and Cognition (Cambridge, MA: Harvard University Press). Van der Sandt, R. A. (1991) Denial. Papers from the Chicago Linguistics Society: The Parasession on Negation, 27: 331–44. Van der Sandt, R. A. (2003) Denial and presupposition. In H. R. P. Kühnlein and H. Zeevat (eds.) Perspectives on Dialogue in the New Millennium (Amsterdam: John Benjamins), pp. 59–77. Wilson, D. (1975) Presupposition and Non-Truth-Conditional Semantics (New York: Academic Press). Wilson, D. and Sperber, D. (1981) On Grice’s theory of conversation. In P. Werth (ed.) Conversation and Discourse (New York: St. Martins Press), pp. 155–78.
6 A Gricean View on Intrusive Implicatures Mandy Simons
1. Introduction This chapter will explore one of the long-standing objections to Grice’s account of conversational implicature: the case of purported implicatures which are apparently generated by subordinate clauses, or which fall under the scope of a logical operator (typically both). Such cases, for reasons to be detailed below, pose a challenge to Grice’s account. While those who have posed the challenge, ranging from advocates of truth-conditional pragmatics to strict compositionalists, have a wide variety of views as to the correct account of the data, they are united in reaching the same negative conclusion: that Grice’s account cannot be extended to intrusive implicatures. In this chapter, I will argue for a different conclusion. I will suggest that there is a natural modification of Grice’s model which allows for the generation of implicatures from non-asserted sentence-parts. The goal of the chapter is to articulate this modification and apply it to some sample cases. This will be done in Part Two of the chapter. In Part One, I will introduce the cases to be investigated and explain in a little more detail what issues they raise.
Part One: The Issues 2. Two illustrative cases We begin by introducing two cases which are taken by many as paradigm cases of conversational implicature: conjunction buttressing, and scalar implicatures.1 Conjunction buttressing is illustrated in (1) and (2): (1) The old king has died of a heart attack and a republic has been declared. (2) Jane was in a car accident and she broke her leg. 138
A Gricean View on Intrusive Implicatures
139
(1) is most naturally understood as claiming that the declaration of the republic followed the death of the king; there is perhaps a further implication that the death of the king facilitated the declaration of the republic. Similarly, we understand (2) as saying that Jane broke her leg as a result of the car accident. Various arguments have been offered to establish that these implications of temporal ordering and of causal relations are not part of the meaning of the sentence itself, and in particular that they are not determined by the lexical meaning of and. Instead, these can be accounted for in Gricean terms as implicatures generated by the maxim of Manner.2 Sentence (3) is a typical example of scalar implicature: (3) The guests stole some of the teaspoons.3 The sentence has a natural reading that is stronger than that provided by the literal meaning alone. While (3) seems to be strictly speaking true if in fact the guests stole many, most or all of the teaspoons, an utterance of this sentence would normally convey that the guests stole some of the teaspoons and no more than some. The implication of upper boundedness can be accounted for as a Gricean implicature, this time generated by the first part of the maxim of Quantity (Quantity 1). This particular case of Quantity implicature is known in the linguistics literature as a scalar implicature. This name is given to any cases of Quantity implicatures that are generated by virtue of the use in a sentence of a lexical item which is a member of a quantitative scale, a concept originally invoked by Horn (1976), defined as follows (following Gazdar 1979): Let Q be an n-tuple of expressions <e0, e1, ... en>. Let Si be a sentence containing the expression ei ∈ Q, and let Si+1 be a sentence just like Si except that ei is replaced by the subsequent element of Q, ei+1. Then if Q is a quantitative scale, Si entails Si+1, as long as ei and ei+1 are not within the scope of an operator. Sentences containing scalar items give rise to implicatures in a very systematic way: an utterance of a sentence S containing a scalar item α typically implicates that, as far as the speaker knows, any sentence S’ in which α is replaced by a stronger element in the same scale is not true.4
2. Embedded implicatures Starting with Cohen (1971), various authors have pointed out that there are cases in which the purported implicature which would normally be associated with an assertoric utterance of a particular clause remains when that clause is uttered as an embedded constituent in a complex sentence. Here, I focus on cases of apparent embedding of the two types of implicature
140
Mandy Simons
introduced above; and I examine three different embedding configurations: antecedents of conditionals, complements of an attitude verb, and disjunctions. So, consider: (4) If the old king has died of a heart attack and a republic has been declared, then Tom will be quite content. (Cohen 1971) (5) Bill believes that some of his students are waiting for him. (Chierchia 2004) (6) Either Kai ate the broccoli or he ate some of the peas. (Sauerland 2004) As argued by Cohen (1971), the implication of temporal ordering and causal connection observed in (1) above is part of the content of the antecedent of (4). Chierchia (2004) argues that an utterance of (5) attributes to Bill the belief that some and not more than some of his students are waiting for him. Similarly, he argues that the second disjunct in examples like (6) implicates that Kai did not have more than some of the peas.5 These claims have been disputed. The debate is difficult, because it often turns on questions about speakers’ ability to distinguish literal meaning from a nonliteral interpretation assigned to an utterance.6 My goal here is not to resolve the dispute, but to ask whether a resolution of the dispute in favour of the claims of Cohen, Chierchia and others who share their intuitions forces us to abandon a generally Gricean view of the inferences in question. For the sake of that argument, I will simply accept the claims that these inferences indeed fall under the scope of the relevant operators and contribute to the literal meanings of the sentences in question.
3. Two problems for the Gricean perspective The examples under consideration raise two distinguishable problems for the standard Gricean view. The first problem I call the calculation problem, and it is this: Grice’s account of the calculation of conversational implicatures does not provide a means for the calculation of implicatures from non-asserted clauses. For Grice, conversational implicatures are aspects of a speaker’s meaning derivable by an interpreter, not from the conventional content of the words and sentences that the speaker used, but from the observation that the speaker has said what she said in the particular conversational context in which she said it. When Grice characterizes the notion of conversational implicature, he begins: ‘A man who, by (in, when) saying (or making as if to say) that p has implicated that q ... ’ (Grice 1989: 30). Later, when he provides a general schema for the calculation of conversational implicature, he similarly begins: ‘[The speaker] has said that p; there is no reason to suppose that he is not observing the maxims ... ; he could not be doing this unless he thought that q ... ’ (ibid.: 31). So, saying what you say comes first; conversational implicatures come after.
A Gricean View on Intrusive Implicatures
141
Grice had something very specific in mind in his use of the term ‘saying’, something, as he puts it (1989: 25), ‘closely related to the conventional meaning of the words (the sentence)’. Neale (1992), in his careful analysis of Grice’s work, concludes that Grice nowhere provides a definitive formulation of his notion of what is said. However, Neale extracts from Grice’s writings the following ‘preliminary’ definition: (7)
By uttering X, U said that p iff: a. by uttering X, part of what U meant was that p. b. X consists of a sequence of elements (such as words) ordered in a way licensed by a system of rules and c. X means “p” in virtue of the particular meanings of the elements in X, their order and their syntactical structure.7
Clauses (b) and (c) serve to connect what is said to what is conventionally meant, what is conveyed by virtue of ‘the particular meanings of the elements in X’. Clause (a) ties what is said to what the speaker meant. Now, recall that for Grice, in the case of an indicative utterance, what a speaker means is, roughly, what she intends to get the hearer to believe.8 It is thus a notion closely aligned with what is asserted. So, connecting this back to the definition of conversational implicatures, we see that a conversational implicature generated by an indicative utterance is calculated on the basis of what a speaker has asserted. This understanding of Grice’s account of conversational implicatures leads to the following conclusion: non-asserted sub-parts of an indicative utterance should not themselves give rise to conversational implicatures. Recanati (2003) cites an argument given by Anscombre and Ducrot (1983) for this conclusion: (a) Conversational implicatures are pragmatic consequences of an act of saying something. (b) An act of saying something can be performed only by means of a complete utterance, not by means of an unasserted clause such as a disjunct or the antecedent of a conditional. (c) Hence, no implicature can be generated at the level of an unasserted clause. In light of this, consider again the scalar implicature examples (5)–(6) above, repeated here: (5) Bill believes that some of his students are waiting for him. (6) Either Kai ate the broccoli or he ate some of the peas. Let’s focus first on (6). According to Sauerland, an utterance of (6) implicates that Kai did not eat all of the peas. Intuitively, that implication is generated
142
Mandy Simons
by the second disjunct – in particular, by a comparison of that disjunct (or its content) with (the content of) “Kai ate all of the peas”. But as a speaker of (6) does not say that Kai ate some of the peas, there are no grounds for generating any implicature from that clause (or its content). Nor can the implicature be generated from the content of the utterance as a whole by standard assumptions.9 The same is true for (5). As Chierchia points out, standard accounts predict that if this sentence gives rise to a scalar implicature, it is this: (8) It is not the case that Bill believes that all of his students are waiting for him. This, though, could be true if in fact Bill had no beliefs at all as to whether all of his students were waiting for him. The observed implicature, that Bill believes that not all of his students are waiting for him, cannot be generated directly from the asserted content of (5), but only from the clause embedded under believe. But, as the content of this clause does not count as said, then it should not trigger any implicature calculation. This, then, is the calculation problem. Sentences (4) and (5) also demonstrate the second problem, which I call the compositionality problem. According to Grice, what is said is supposed to be determined by conventional content (in addition to reference fixing and disambiguation).10 But in examples like (4) and (5), the implicatures apparently generated by embedded clauses seem to fall under the scope of the embedding operators, and thus to contribute to the truthconditional content expressed, that is, to what is said. Consider (4) again, repeated here: (4) If the old king has died of a heart attack and a republic has been declared, then Tom will be quite content. As Cohen points out (1971: 58), the truth of this sentence seems compatible with Tom being not at all content if a republic had been declared first and then the old king had died of a heart attack. If this is so, then the temporal ordering of the events is part of the truth-conditional contribution of the antecedent. So, whether or not one is concerned with Grice’s conception of what is said, intrusive implicatures pose a problem for the thesis that the determination of truth-conditional content is strictly compositional. There is thus a good deal at stake in the investigation of these rather commonplace examples. There has been a wide range of responses to the phenomena discussed here.11 On the one hand, advocates of truth-conditional pragmatics, including the relevance theorists (following Sperber and Wilson 1986),
A Gricean View on Intrusive Implicatures
143
reject Grice’s conception of what is said and any view which excludes the possibility of nonconventional content contributing to literal meaning. In particular, the interpretation of scalar items is typically taken by relevance theorists to be semantically indeterminate. The actual interpretation in a given case is argued to be determined by contextual factors. In a very different approach, Gennaro Chierchia (2004) attempts to maintain (a version of) semantic compositionality by arguing that scalar implications are not in fact Gricean implicatures, but are generated by grammatical rule in the course of semantic composition. This allows him to maintain an overall Gricean perspective: first, conventional content and grammatical rules provide a literal content for an utterance. This content can then provide the input to a Gricean process of implicature calculation. Falling somewhere between these two views is Levinson’s default interpretation view (Levinson 2000). Like the relevance theorists, Levinson argues that nonconventional content can make a truth-conditional contribution. However, he argues that such cases typically involve what he calls generalized conversational implicatures (GCIs), implicatures which, although explicable in terms of general conversational heuristics, are associated by default with particular expression or construction types. Default interpretations can always be overridden by global conversational considerations of a Gricean sort; but their initial generation is noninferential and automatic. As noted, what is common to these responses is an assumption that the general framework provided by the Cooperative Principle and the maxims of conversation cannot be modified to accommodate the troubling data. In the next section, I will suggest that they can. In doing so, I will offer a solution to the calculation problem posed by intrusive implicatures. The solution I propose does not attempt to preserve Grice’s notion of what is said or the compositionality of literal content. The goal is only to make plausible the idea that Gricean principles can apply to non-asserted sub-parts of sentences.
Part Two: The Proposal 4. Implicatures from non-asserted clauses: Walker (1975) Grice’s account of the workings of cooperativity, combined with his views on the determination of ‘what is said’, indeed seems to fail to account for cases of embedded implicatures. One response to this conclusion is to reject Grice’s account altogether, or at least reject it as being in any way applicable to these difficult cases. Another response is to consider modifications of Grice’s view that might allow resolution of the difficulties. The latter seems a perfectly reasonable approach. There is no reason why, as theorists, we have to take Grice’s views as an unseparable whole. The fundamental ideas underlying the notion of implicature are clearly separable from Grice’s expressed views about the nature of ‘what is said’ and about the
144
Mandy Simons
semantic content of particular items. For example, Grice argued that natural language conditionals have the semantics of material implication, and used the notion of conversational implicature to try to make this view plausible. Formal linguists have by now (I think) universally rejected this account of conditionals; but this does not prevent us from maintaining a Gricean account of conversational inference. Similarly, the Cooperative Principle and the associated maxims are a first stab at formulating the appropriate principles, not the last word on the matter.12 One can maintain a broadly Gricean outlook while modifying the details of the account of how implicatures are generated. Recanati (2003) suggests (and ultimately rejects) one such modification. The idea is to ‘construe ... implicatures as pragmatic implications of something other than a self-standing speech act’ (p. 311). As Recanati reports, a version of this modification is proposed in Walker (1975).13 Walker writes: [The Conversational Hypothesis] holds that by a particular utterance on a particular occasion the speaker can convey more than his utterance strictly means through relying on a general recognition of Grice’s Cooperative Principle. It is therefore concerned with utterances, whether they constitute self-standing speech acts or not; an utterance of a subordinate clause, as in the antecedent of a conditional, is still an utterance, and therefore may convey conversationally more than it literally means. It may convey, for example, a further condition on which the consequent is taken to depend. (Walker 1975: 151; also cited by Recanati) Walker takes this approach in response to Cohen’s (1971) objections to Grice’s proposal. In his response, he takes on both the calculation problem and, in a rather roundabout way, the compositionality problem. The passage quoted above constitutes the response to the calculation problem. He provides a more detailed discussion of this problem as it arises in Cohen’s examples of coordination in the antecedent of a conditional. The central example is repeated in (9) below: (9)
If the old king has died of a heart attack and a republic has been declared, then Tom will be quite content.
Walker argues as follows: The implication of temporal ordering between the conjuncts is generated by the requirement to be orderly. This requirement is in force in any conversational context where the order of the events is of conversational relevance. He continues: In the normal conversational context of [(9)] the order of the events is as important to the purposes of the talk-exchange as it is when ‘The old king
A Gricean View on Intrusive Implicatures
145
has died of a heart attack and a republic has been declared’ is asserted by itself. (Walker 1975: 138) So, the maxim applies just as well to the antecedent of the conditional, and the implication is generated. Recanati extends Walker’s account to the case of clauses embedded under believe, as in (10): (10) Paul believes that the old king has died of a heart attack and a republic has been declared. Echoing Walker, we can say that in a normal conversational context in which (10) might be uttered, Paul’s beliefs about the temporal ordering of the events are relevant, and so should be part of what is conveyed by the belief claim. Thus, as Recanati writes, ‘the speaker’s describing the two events in that order suggests that, according to Paul, they took place in that order’ (ibid.: 312). So, Walker addresses the calculation problem by arguing that the purposes of a talk exchange may impose constraints on the formulation of subordinate clauses as well as main clauses, and thus that the conversational maxims are applicable to subordinate clauses too.14 Walker is not the only theorist to have suggested this strategy. Green (1998: fn.40) argues that ‘nothing in the Gricean apparatus precludes calculation of what a speaker could have intended to convey in uttering a clause that is not put forth with any illocutionary force, say by virtue of that clause occurring in the consequent of a conditional or under the scope of negation.’15 The goal of this second half of the chapter is to try to extend Walker’s proposal into a general solution to the problem of embedded implicature.
5. Developing the account: the status of embedded clauses Walker’s argument is founded on intuitions about specific cases. In these specific cases, and given the specific conversational principle involved (i.e. orderliness), the claim that the principle applies to the embedded clause seems quite justifiable. But there is, I think, a more general case to be made for the position. The basis for the general case is that subordinate clauses do not serve merely to contribute to the propositional content expressed in an utterance. Typically, these clauses themselves serve identifiable discourse functions. Cooperativity requires these functions to be fulfilled as well as possible. To put this a different way: interpreters can pay attention to parts of sentences independently of the containing sentence, and can reason about why the speaker produced just that sentence-part in attempting to convey her communicative intention. This reasoning, I suggest, is what gives rise to ‘local’ conversational inferences.
146
Mandy Simons
In support of the claim that non-asserted clauses can have a certain independence, consider the discourse behaviour of sentences containing main verbs with sentential complements, such as propositional attitude verbs and verbs of saying. I have argued elsewhere (Simons 2007) for the conjunction of two claims: (11) Given an utterance of a sentence of the form x V that p: a. What is asserted is that x V that p. b. The proposition that p may be proffered as main point content. First, note that I distinguish what is asserted from the utterance’s main conversational point. This means that a speaker may contribute the proposition that p to the conversation without actually asserting p or becoming committed to its truth. The clearest example of this is in answers to questions. Consider any of the sentences in (13) offered as answers to the question in (12): (12)
Where did Jane go last week?
(13)
a. b. c. d.
Henry believes she was interviewing for a job with Google. Henry thinks she was interviewing for a job with Google. Henry said she was interviewing for a job with Google. Henry hinted she was interviewing for a job with Google.
It is the content of the subordinate clause which, in each case, constitutes an answer to the question. This in turn provides us with evidence that this content constitutes the main point of the utterance. The main clause serves an evidential function, indicating the source of the proffered information and the degree of confidence of the source in the truth of the claim.16 By so embedding the proffered content, the speaker avoids committing herself to its truth. An interpreter can, nonetheless, take up this embedded content in her response. She might, for example, say: ‘But she wasn’t. I know, because Bob already told me who they’re interviewing’. These examples show that an interpreter can ‘access’ the content of a non-asserted sentence-part (at least, of a clausal complement) and treat it in some sense as if it were independent of the larger clause in which it is embedded. There are two, not mutually exclusive, ways to think about how the interpreter ‘retrieves’ the embedded content and responds to it as if it were independent. We might tell a purely inferential story along the following lines: I have asked a question about where Jane went. The speaker responds with an assertion about Henry’s belief. This does not answer my question. But the content of the belief is relevant to my question. Therefore, the speaker intends me to pay attention to the content of Henry’s belief.
A Gricean View on Intrusive Implicatures
147
Alternatively, we might suppose that access to the embedded proposition is more direct: that the interpreter immediately recognizes the utterance as having two parts: a part which mentions that Henry has a particular belief, and a part that characterizes that belief. Now, the inferential story is still needed to explain how the interpreter knows, furthermore, that the intended answer to the question is provided by the content of the belief, while the main clause content serves an evidential function. But here we add the idea that the embedded clause is immediately recognized has having a particular function within the sentence itself, namely, the function of characterizing Henry’s belief. I don’t see any reason to think that competent speakers of a language couldn’t do this – in fact it just seems rather obvious that they can. So I will make this assumption in the discussion that follows. This does have a consequence: given this assumption, I allow that recognition of the function of the subordinate clause does not require the presence of conversational factors that force this clause into main point status, or anything similar. We can now reconsider Walker’s account of the Cohen example (9) in this light. First, let us see whether there is evidence that the antecedent of a conditional can serve a discourse function independent of the full sentence in which it appears. There is indeed a robust intuition that it does: that an interpreter recognizes the antecedent of a conditional as presenting the hypothetical circumstances in which it is claimed that the consequent would (or might, should, and so on) hold. This intuition is reflected in standard contemporary accounts of the semantics of conditionals, in both static and dynamic semantics. In dynamic semantics (see e.g. Stalnaker 1974, for an early informal formulation; Heim 1992 for a formalization), a standard approach involves a stage at which the context is updated by the content of the antecedent.17 In static semantics, for example Kratzer (1991), the antecedent of a conditional serves to restrict the set of worlds in which the consequent must be true. A formal semantics for conditionals of course may not correspond to what an interpreter actually retrieves in understanding a conditional. However, it is pretty clear that in ordinary talk, interlocutors can respond to the content of an antecedent independently of the total assertion. A discourse such as the following is totally natural: (14)
A: If Jane comes later, we can fill her in. B: She won’t be coming.
Note that in this case, speaker B seems to take speaker A’s utterance as raising the question of whether Jane will come later. Although conditionals can serve a range of discourse functions, it seems generally true that their antecedents have the potential of introducing a new question, or topic, for
148
Mandy Simons
discussion in the discourse. As noted by Groenendijk and Stokhof (1984), this is what makes sequences such as the following felicitous: (15)
A: What’s the weather going to be like today? B: If Bill took his umbrella, it’s going to rain.
Clearly, the conditional is not an answer to the question; but it provides a strategy for arriving at an answer, namely, by answering the question raised by the antecedent of the conditional. What examples such as these show is that, at the very least, an interpreter is able to identify the content of an antecedent clause and make use of it or respond to it independently of the conditional that the speaker asserts. Returning then to the Cohen example, we can say just what Walker does: the interpreter can be assumed to recognize that the antecedent clause provides a representation of the circumstances in which it is claimed that Tom will be happy. The interpreter can further assume that the speaker has given the best, most cooperative representation of those circumstances that she can. A good representation is an orderly one; therefore, the interpreter can presume that the speaker has abided by the orderliness requirement of the maxim of Manner. It is perhaps entirely obvious that an interpreter has access to the content and function of the antecedent of a conditional. I belabour the point only because the standard Gricean account of conversational inference seems to ignore it, assuming, as we have already noted, that only the total content of what is said can serve as the input to conversational reasoning. But if in fact an interpreter recognizes simultaneously that a speaker has asserted a proposition p with form F and has also expressed proposition q in form F9 to do it, I see no reason to rule out the possibility of reasoning conversationally about this second observation.
6. Disjunction We can make the following generalization. If a given clause gives rise to a scalar implicature when it occurs unembedded, it will typically give rise to the same implicature if it occurs as a disjunct in an or coordination. Similarly, a conjunction that gives rise to an implicature of temporal ordering or causal relation between two events when it occurs unembedded will give rise to the same implicature when it occurs as a disjunct. Let’s look again briefly at the data.18 First, a conjunction case: (16)
I’m not going to go to the party. It won’t turn out well for me. Either I’ll get drunk and no-one will talk to me, or no-one will talk to me and I’ll get drunk. (Modification of an example due to Wilson and Sperber 1998)
A Gricean View on Intrusive Implicatures
149
As Wilson and Sperber point out, the disjunctive sentence in (16) is not redundant, as it would be if the conjunctions in each disjunct were given a purely truth-functional reading. Clearly, the actual reading we get is one which takes into consideration the temporal and causal implications of the ordering of the conjuncts in each of the disjuncts. And of course, these implications are just what we would get if the disjuncts occurred as independent sentences, as in: (17) a. I’ll get drunk and no-one will talk to me. b. No-one will talk to me and I’ll get drunk. So, the most straightforward way to understand the interpretation is that the same principles which give rise to the temporal and causal implications in (17) also apply to the disjuncts in (16). Example (18) below illustrates the scalar implicature case: (18) Either the guests stole some of the teaspoons, or we didn’t have many to begin with. The speaker of this sentence clearly rules out the possibility that the guests stole all of the teaspoons. But how? The literature provides two different answers. Sauerland (2004) offers a refinement of the Gricean procedure for calculating scalar implicatures, on the basis of which he argues that (an utterance of) the sentence as a whole implicates (19): (19) The guests did not steal all of the teaspoons. We’ll look at Sauerland’s proposal in more detail below. Others (including Chierchia 2004 and Levinson 2000) argue instead that the first disjunct receives the interpretation in (20): (20) The guests stole some but not all of the teaspoons. In what follows, we’ll assume the position that is most difficult for the Gricean, that is, that the implicature is local to the disjunct. As before, we begin by gathering evidence that the disjuncts in an unembedded disjunction have a sort of conceptual independence. First, we note that interlocutors clearly can respond to disjuncts independently of the disjunction as a whole, for example, by rejecting a particular alternative. This is nicely illustrated by Grice’s notion (1989: 64) of ‘substitutive disagreement’, illustrated by the following exchange: (21) A: Either Wilson or Heath will be the next Prime minister. B: I disagree, it will be either Wilson or Thorpe.
150 Mandy Simons
Here, speaker B disagrees with the first speaker’s claim that Heath is even a possibility, and substitutes Thorpe. This is possible only if, in interpreting A’s utterance, B is able to recognize that part of what is accomplished by it is to offer Heath as a possible winner. Similarly, speaker B could have said: (22)
... No, it won’t be Heath
which again would not deny the disjunction, but only the reasonableness of one of the disjuncts. Or, again: (23)
... Yes, it really could be Wilson.
One could multiply examples, but it hardly seems necessary: it seems hard to deny that when a speaker utters a disjunction, we as interpreters retrieve the content of individual disjuncts, and can agree, disagree, question, or respond in any other way to that content independently of a response to the disjunction as a whole. There is a very different kind of data which has been used to argue that even at the level of semantic analysis, the content of individual disjuncts is accessible. Eggert (2000) demonstrates, contra earlier claims, that the adverb respectively can be used with or coordinations in certain environments. Here are two of his examples (all of which are culled from web searches): (24) In most carbon compounds, an adjacent atom will contribute one to three electrons, which are matched by an equal number from carbon to form a single, double or triple bond, respectively. (25) Choose the snowflake or sun for winter or summer information, respectively. These data are relevant because they show (although admittedly not for the clausal disjunct case) that the denotation of a disjunct must be independently retrievable in the course of interpretation. I made the case with respect to disjunctions embedded under modals (Simons 2005a,b) as in, for example, (26): (26)
Jane may sing or (she may) dance.
I argue that the problematic free-choice readings of sentences such as this, as well as other complications involving the interaction of or and modals, require a semantics for or which allows the denotations of the disjuncts to be ‘visible’ throughout the process of composition. My initial claim, though, was not merely that the content of disjuncts is retrievable by an interpreter, but that the disjuncts have a sort of conceptual independence, that they are recognized as having a particular discourse
A Gricean View on Intrusive Implicatures
151
function. Grice himself offers the following view of the function of unembedded disjunctions: A standard (if not the standard) employment of or is in the specification of possibilities (one of which is supposed by the speaker to be realized, though he does not know which one), each of which is relevant in the same way to a given topic. “A or B” is characteristically employed to give a partial (or pis aller) answer to some [wh]-question to which each disjunct, if assertable, would give a fuller, more specific, more satisfactory answer. In earlier work (Simons 2000), I argued that in fact an (unembedded) disjunction is felicitous only if it is so construable: that is, only if each disjunct is construable as an answer to a given question (either explicit or implicit), with each disjunct constituting a distinct answer to the given question. This view provides us with a route to the application of the conversational maxims to individual disjuncts. If we take a very strong view of the ‘possible answer’ requirement, we might suppose that each disjunct must be such that if its content were assertable, the disjunct could be felicitously uttered at that point in the conversation. (Obviously, a much more careful version of this view is needed to accommodate non-clausal disjuncts.) To be felicitously uttered, it would have to be consistent with the maxims, including orderliness and Quantity 1. If the interpreter assumes that the disjunct: (27) I’ll get drunk and no-one will talk to me could be felicitously uttered, she will further assume that the speaker intends to describe a situation in which the speaker first gets drunk and then, as a consequence, no one will talk to her. Similarly, on the presumption that the disjunct: (28) The guests stole some of the teaspoons if true would be felicitous if uttered, the interpreter can assume that the speaker’s belief is that if (28) is true, then the stronger claim: (29)
The guests stole all of the teaspoons
would not be true. Hence, the interpreter can take the intended content of (28) to be ‘the guests stole some but not all of the teaspoons’, even when it occurs as a disjunct. Even if one takes a weaker view of the ‘possible answer’ requirement, a case can still be made for local applicability of the maxims. Unembedded disjuncts are routinely taken as characterizations of states of affairs that the
152 Mandy Simons
speaker considers possible. Cooperativity presumably requires that each disjunct should be the best characterization available. The maxim of Manner then applies straightforwardly, just as in the case of the antecedents (or consequents) of conditionals. In the case of (27), extracted from (16) above, we have a description of a situation which contains two eventualities. These eventualities can be temporally ordered and can plausibly stand in a causal relation. The interpreter can therefore assume that, following the orderliness requirement, the speaker has presented the eventualities in each disjunct in a way which reflects the envisioned temporal and causal relations between them.19 Now consider (28) as it occurs in (18) above. Here again, the interpreter first recognizes that the first disjunct constitutes a description of a state of affairs which the speaker considers possible (and a possible explanation for the shortage of teaspoons). She is entitled to assume that this description constitutes the best, or most accurate, or most specific characterization of this state of affairs. If the situation the speaker has in mind as a possibility is one in which the guests stole all of the teaspoons, then, given that there is a straightforward way to characterize this state of affairs without unduly complicating the utterance, she should have done so. Therefore, the interpreter can assume that the speaker does not have this state of affairs in mind, and is offering as a possibility that the guests stole some but not all of the teaspoons.20 This understanding of the applicability of Quantity involves a rather liberal reading of the maxim. Quantity 1 as formulated by Grice enjoins speakers to provide ‘as much information as is required’. Linguists have typically taken this to mean: ‘Make the strongest relevant assertion licensed by your beliefs’, where entailment is taken to be the relevant notion of strength. Recanati (2003) is indeed sceptical of the idea that Quantity 1 can be applied directly to non-asserted utterance parts. He writes: ‘The scalar reasoning appeals to the idea that the speaker respects the maxim of quantity i.e. gives as much (relevant) information as possible; now it is far from obvious that the notion of “giving information” can be divorced from that of asserting (or from similar notions)’ (Recanati 2003: 313). But suppose we read Quantity 1 this way: Provide as much information as is required about the situation you are describing. Or, utilizing the notion of strength: Provide the strongest description of the situation you aim to describe compatible with the requirements of relevance. The idea is this: a speaker’s choice of words is always an indication of some belief she has about the situation she is describing. In the case where the utterance describes this situation as actual, the beliefs in question will be beliefs about what is the case. In the case where the utterance describes the situation as hypothetical, or as merely possible or probable, or as the content of someone’s propositional attitude, the beliefs will be beliefs about what is possible or probable, or about another agent’s beliefs. Quantity 1, I suggest, enjoins the speaker to give the best – in many cases, strongest – characterization of the envisioned
A Gricean View on Intrusive Implicatures
153
situation that is consistent with these beliefs. If the interpreter assumes that the speaker is abiding by this requirement, then she can apply Quantity 1 to the interpretation of non-asserted clauses. This reinterpretation of Quantity requires the assumption that speakers begin with something in mind that they intend to talk about, some situation that they intend to characterize. I am uncertain whether or not this is controversial, but it does seem in keeping with Grice’s understanding, where a speaker has a communicative intention which underlies her communicative acts. 6.1 A comparison with Sauerland (2004) We have already noted that Sauerland (2004) offers his own neo-Gricean response to the challenge posed by the generation of scalar implicatures from disjuncts and other embedded positions.21 Sauerland’s account has two parts. The first part is a revision of the Horn/Gazdar method for generating scalar alternatives. This revision begins with the definition of a onestep scalar alternative, as follows: (30)
A sentence ψ is a one-step scalar alternative of φ if the following two conditions hold: a. φ ≠ ψ b. there are scalar expressions α and α’ which both occur on the same scale C such that ψ is the result of replacing one occurrence of α in φ with α’.
Sauerland then uses this as the basis of a definition of a scalar alternative (simpliciter): (31) A sentence χ is a scalar alternative of a sentence φ if there is a sequence (φ0 ... φn) where φ0 =φ and φn=χ and for all i such that 1≤i≤n, φi is a one-step scalar alternative of φi21. In other words, χ is a scalar alternative of φ if one can arrive at χ by replacing the scalar items in φ one by one with other elements from their scales. The second part of Sauerland’s analysis is the part of most relevance to this discussion. To account for the scalar implicatures generated by scalar items within disjuncts, he proposes a modification of the scale of which or is supposed to be a member. At least since Gazdar (1979), it has been presumed that or is an element of the scale ; this assumption has formed the basis of a standard account of the exclusive reading of or coordinations as a scalar implicature.22 On this standard view, the sole scalar alternative to a sentence A or B is the conjunction A and B. Sauerland proposes that the scalar alternatives should
154
Mandy Simons
also include the disjuncts themselves.23 Thus, sentence (33) counts as a onestep scalar alternative to sentence (32) (by replacement of a disjunction by a disjunct),24 and sentence (34) as a one-step scalar alternative to (33) (by replacement of some by all). (32) Either the guests stole some of the teaspoons, or we didn’t have many to begin with. (33) The guests stole some of the teaspoons. (34) The guests stole all of the teaspoons. Thus, applying the definition of scalar alternatives, (34) is a scalar alternative of the disjunction (32), and the utterance of (32) is predicted to implicate that (as far as the speaker knows) it is not the case that the guests stole all of the teaspoons.25 This is undoubtedly an elegant solution. In particular, it allows for the apparently local implicature to be derived using a global mechanism. The calculation begins from the observation that the speaker has uttered A or B, and the implicature that the guests did not steal all of the teaspoons is generated as an implicature of the utterance as a whole, not of its sub-part. But this may also be a weakness of the account, for there are cases where the apparent local implicature could not plausibly be understood as an implicature of the utterance as a whole. Consider: (35)
Bernice is in trouble again. Either she was supposed to visit each of the senators personally but she only CALLED some of them, or else it was okay for her to call them and she did in fact call all of them but she wasn’t friendly enough. In any case, her performance is rated unsatisfactory.26
The first disjunct of the target sentence in (35) is naturally taken as meaning that Bernice called (and failed to visit) some but not all of the senators. On Sauerland’s account, though, there are no local implicatures. Rather, if there is an implicature, it must be of the utterance as a whole. Presumably, it would be this: (36)
It’s not the case that Bernice called all of the senators.
But of course this cannot be an implicature of the utterance as a whole, because the second disjunct allows that Bernice did in fact call all of the senators. Sauerland’s account would seem to predict that there should be no implicature at all, as the account does not allow for the generation of implicatures that are in conflict with the speaker’s explicitly demonstrated knowledge or ignorance.27 Yet the disjunction certainly is understood in a way consistent with a strengthened interpretation of the first disjunct.
A Gricean View on Intrusive Implicatures
155
Sauerland’s account seems to share a fundamental intuition with mine: that to get the right implicature, one has to be able to ‘get at’ the disjunct independently of the disjunction. But Sauerland does this via an algorithm for the generation of scalar alternatives. This leads to another concrete difference in the predictions of our accounts. First, I see the possibility of generating implicatures from disjuncts as part of a general phenomenon of applying conversational reasoning to embedded clauses (i.e. to non-asserted content). So, to the extent that my proposal is successful, it can be applied equally to implicatures generated by disjuncts and implicatures generated by the complements of propositional attitude verbs or the antecedents of conditionals. Second, my proposal offers a way to take disjuncts (and other embedded clauses) as input to any kind of conversational reasoning. So, parallel accounts can be given for local scalar implicatures and local implicatures of other kinds. Sauerland’s proposal provides no basis for an extension to other cases of pragmatic intrusion.
7. Back to the antecedents of conditionals: scalar implicatures The Walker example that we began with involved conjunction buttressing in the antecedent of a conditional. We now turn to the case of scalar implicatures in antecedents. The data in this case are more complex than in the disjunction case, for scalar items in the antecedents of conditionals sometimes give rise to (local) scalar implicatures, and sometimes don’t. Compare: (37)
If I give an extension to some of my students, the others will be annoyed.28 (38) If some of my students fail the course, I’ll be unhappy. In example (37), some is interpreted as “some but not all”; so here, the antecedent seems to get the interpretation we expect if a scalar implicature is generated locally. Example (38), on the other hand, does not require this local strengthening; there is certainly no implication that the outcome would be different if in fact all of my students failed. Here are two attested examples illustrating the same variation: (39)
You have undoubtedly heard the saying ‘Don’t put all your eggs in one basket’? The same principle applies to investing – put your investments into different baskets. If some of the baskets should fail then your losses will still be manageable. [http://www.abcsofinvesting.net/ investment-diversification/, Dec. 23 2009] (40) If some of the cylinders aren’t getting spark, then you have an ignition problem, most likely your distributor cap ... [http://www.barneymc.com/toy_root/techneek/mtr_test.htm, Dec. 23 2009]
156
Mandy Simons
The difference between the interpretations in each pair has a common-sense explanation, the sort that would be obvious to anyone unencumbered by theory. The sentence in (37) only makes sense on the assumption that by some the speaker meant “only some” (or at least, “not all”). In a situation in which an extension was given to all students, there would be no ‘others’ to be annoyed. Wanting to attribute to the speaker an intention to make a sensible claim, the interpreter so understands her. In (38), on the other hand, no such constraint is in place. Indeed, one assumes that the consequent is even more likely to hold if in fact all of the students fail the course. So here, the unstrengthened “some or more” interpretation makes the most sense. Can the theory-encumbered adopt this sensible account? It depends. The principal stumbling block is that the implication, when present, indeed is local – that is, it contributes to the interpretation of the antecedent, just as in the case of conjunction buttressing. Yet the common-sense account appeals to the question of whether or not the conditional as a whole makes sense – that is, to utterance-level considerations. The common-sense account remains compelling nonetheless, and different theorists find different ways to accommodate it. For advocates of the Levinsonian default interpretation view, cases like (37) and (39) show the default interpretation of the quantifier. Cases like (38) and (40) where the default is suspended show the context-sensitivity of the default: where the default is in conflict with other considerations, it can be overridden. It is not, though, completely clear that the global ‘sensibleness’ considerations adduced above would really suffice to override the default. Suppose a speaker had uttered (41) rather than (38): (41)
If some but not more of my students fail the course, I’ll be unhappy.
While it is hard to know why anyone would feel this way, the utterance is in no way incomprehensible. Similarly, in the case of (40), it is plausible that if more than some of the cylinders fail to get a spark, then some different problem exists. In other words, if we assume that the strengthened interpretation is the default, it is not clear that we have here considerations which are strong enough to override it. On the other hand, if there is no default, but only a process which opts for the interpretation most compatible with background assumptions, the unstrengthened readings are predicted. Chierchia (2004) also adopts a version of the common-sense account. His account predicts that scalar implicatures will in fact fail to arise in the antecedents of conditionals.29 So for him, the problematic cases are those where the strong readings appear nonetheless. He suggests that these interpretations involve a non-local process of accommodation, driven by the sort of ‘sensibleness’ considerations adduced above. 30 This applies at the sentence level, and does not involve any change in the interpretation of the
A Gricean View on Intrusive Implicatures
157
antecedent itself. Hence, it involves no pragmatic intrusion. As Chierchia notes: The effect of this accommodation is the same as the computation of an implicature. But if we are right, the mechanism through which this happens is very different from how implicatures come about normally [in Chierchia’s system – ms.]. (ibid.: 67) I am inclined to simply accept the common-sense explanation at face value. We assume, as above, that the interpreter of a conditional recognizes the antecedent of the conditional as a linguistic unit with a specific function within the discourse. She seeks to assign to it an interpretation which maximizes the cooperativity of the speaker. But interpretation is not necessarily linear. We need not assume that the interpreter assigns an interpretation to the antecedent entirely independently of the attempt to make sense of the utterance as a whole. Rather, it seems plausible that the interpreter assigns to the antecedent an interpretation which improves the overall sensibleness of the utterance. So, the antecedents of (37) and (39) get a strengthened interpretation; those of (38) and (40) do not. Let’s consider how the interpretation might proceed. We begin with (37), assuming our new understanding of Quantity 1. The interpreter reasons that the speaker intends her to consider a situation in which some of the students are given extensions. She reasons that the speaker does not (in particular) intend her to consider a situation in which more than some are; if she did, she should have offered a different characterization of the hypothetical situation. The interpreter can indeed make sense of the utterance as a whole only if she restricts her attention, in considering the antecedent, to situations in which only some of the students get extensions: so, such are the situations she considers. In (38), as before, the interpreter reasons that she is to consider situations in which some but not more of the students fail the course. She can make sense of the utterance given this understanding; but her world knowledge tells her further that if the conditional is true under the stronger interpretation of its antecedent, it is (probably) also true under the weaker interpretation of its antecedent. So in this case, the interpreter does not have a strong reason, as she does in the case of (37), to believe that the speaker intends her to consider only hypothetical situations in which only some of the students fail. In these cases, I am allowing for interaction between local pragmatic considerations (application of Quantity 1 to the antecedent of the conditional) and global ones (finding a plausible reading of the conditional as a whole). However, I am suggesting that where strengthening takes place, it is indeed a local effect. This observation points to a distinction that should be made
158 Mandy Simons
between two different notions of locality. In saying that an implicature is local, we may mean that the implicature is computed locally: that is, that the implicature (or other pragmatic effect) is the result of a process which takes into account only some sub-part of the entire utterance. My proposal as to how strengthened interpretations of disjuncts arise is an example of local computation. Levinson’s default interpretation view and Chierchia’s grammatical view are also proposals that allow for local calculation of scalar inferences. But there is a second thing we can mean when we say that an implicature is local: namely, that it shows its effects at a level below that of the maximal sentence. This is how we defined pragmatic intrusion at the start of the chapter. Perhaps the most typical view is that local effects must be the result of local computation. Both Chierchia’s and Levinson’s systems accomplish this, although both systems also allow for global considerations to override the generation of local effects. Sauerland, on the other hand, aiming to provide a non-local account of apparent embedded implicatures, argues both that the implicatures are computed globally (i.e. with the total utterance as input) and that the resulting implicatures are also global (i.e. the disjuncts themselves do not receive modified interpretations). In relevance theory and various realizations of the truth-conditional pragmatics programme, computations and effects are not assumed to go hand in hand. In particular, it is common in these accounts to argue that global pragmatic considerations give rise to local effects, including the assignment of specific interpretations to lexical items. We must also allow that in principle it might be possible for inferences that are locally computed – that is, generated from an embedded linguistic expression – to have a global effect. I would suggest tentatively that this is happening in examples like (38). While in this case the local effects of Quantity 1 seem to be undermined by the fact that the overall reasonableness of the utterance does not require them, the maxim seems not entirely without force. All other things being equal, the speaker at least weakly implies that she expects no more than some of her students to fail. Similarly, (40) appears to tell us something about the writer’s expectations about what proportion of cylinders are likely to fail to get spark. Let me then summarize briefly the conclusions from this discussion of scalar implicatures in the antecedents of conditionals. We see that scalar items in this position sometimes do and sometimes do not lead to strengthened interpretations of the antecedent. The most natural explanation for this variability is based on global considerations: an interpreter chooses a strengthened interpretation if this is required to make sense of the conditional as a whole. To accept this sort of explanation, however, we must accept that global considerations can have local effects.
A Gricean View on Intrusive Implicatures
159
8. Embedding under propositional attitudes By now the general strategy will be clear, so I will deal more briefly with this third case. The sort of examples we are now concerned with are repeated in (42) and (43) below: (42) Bill believes that Sally sold her apartment and left town. (43) Bill believes that some of his students are waiting for him. The application of Manner to the embedded clause in (42) is fairly unproblematic. Let’s consider the more difficult issue of applying Quantity, reinterpreted as above, to (43). The basic idea is as follows: the speaker, in uttering (43), has in mind a particular belief which she intends to attribute to Bill. Let’s characterize this belief as a belief that a situation with certain properties is actual. The speaker uses the words ‘some of his students are waiting for him’ to describe that situation. If she intended to characterize Bill as believing as actual a situation in which many, or most, or all of Bill’s students were waiting for him, she could have used a different formulation – embedded a different proposition – in order to provide a more accurate representation of that situation. As she didn’t do so, she must intend to convey that the situation Bill believes to hold is one in which some but not more than some of his students are waiting for him. First question: how is this different from the standard Gricean account? The standard account says that the speaker should make the strongest assertion compatible with her beliefs and with relevance. An utterance of: (44) Bill believes that many of his students are waiting for him would have been a stronger assertion (as (44) entails (43)). So, doesn’t the account above boil down to the same thing as the standard account? If the speaker were in a position to make a stronger assertion, she should have done so. The accounts are obviously closely related, but they differ in their predictions. The standard account, in its usual implementation, allows for the generation of an implicature which consists of the negation of (44). As noted above, this implicature does not produce a strengthening of the content of the embedded clause. My account, which focuses on the embedded clause as a characterization of a particular state of affairs, does exactly this. But there are two problems: possible ignorance of the speaker; and possible ignorance of the subject. It is standardly recognized that in cases where the speaker is understood to be under-informed about the situation at hand, no scalar impli-
160
Mandy Simons
cature is generated by the use of a scalar item. Consider, for example, the following: (45)
We have begun to review the evidence and have already determined that some of the results were fraudulent.
Here, the first clause makes clear that the speaker has incomplete information about the situation. In this case, the ‘some’ in the embedded clause is not interpreted as “some but not all”. Rather, what is understood is that the evidence so far has shown that some results are fraudulent; it is possible that further evidence will show that more than some are. Now, suppose the speaker of (44) were known to have only partial information about Bill’s beliefs, as is suggested, for example, in the following context: (46)
Bill said he had to go to his office hours. Apparently he believes that some of his students are waiting for him.
It is not entirely clear to me what, if anything, is locally implicated in this case. Perhaps the speaker still attributes to Bill the belief that some but not more than some of his students are waiting for him. If so, the reason is as before: if the speaker intended to attribute to Bill a stronger belief, there are linguistic resources for her to do so easily. However, unlike the case where the speaker is taken to be well-informed, we, as hearers, may be unwilling to give full credence to the strong belief claim. We might decide that the speaker’s evidence (that Bill said he had to go to his office hours) suffices for the weak claim, but not for the stronger one. On the other hand, perhaps the speaker is simply taken as attributing to Bill the belief that at least some of his students are waiting, and possibly more. This would be natural if we take the weakness of the speaker’s evidence as indicating that she does not have in mind a fully specified state of affairs, belief in which she intends to attribute to Bill. Alternatively, perhaps we recognize that while the speaker’s evidence suffices for her to make the weak belief claim, her evidence does not suffice for her to make the strong belief claim. Assuming the speaker to be acting in accord with the requirements of Quality, which supercede those of Quantity, we attribute the weaker claim to the speaker. And perhaps this is a case where the speaker neither implicates nor is explicitly recognized as not implicating. Remember that conversational implicatures arise only where they are required to make sense of the speaker’s utterance, to render it fully in conformity with the requirements of cooperativity. Sometimes, it simply doesn’t matter whether the speaker had in mind something more or less specific. This may be one of those cases.
A Gricean View on Intrusive Implicatures
161
Russell (2006) offers an explanation of cases like (43) within the standard Horn/Gazdar framework for scalar implicatures. Let us make the standard assumption that an utterance of (43) implicates the negation of the stronger scalar alternatives. So, one of its implicatures will be: (47) It is not the case that Bill believes that all of his students are waiting for him. Chierchia (2004) argues that this implicature is too weak, as it could be true if Bill had no belief at all about whether his students were waiting for him. Russell points out, though, that if Bill does have a belief about whether all his students are waiting for him, then, if (47) is true, it must be the case that Bill believes that it is not the case that all of his students are waiting for him. Russell argues that the apparent strengthening of the belief clause is not itself an implicature, but is an inference generated from the actual implicature (47) plus the assumption that Bill has a belief about the relevant proposition. In the case in question, where it is asserted that Bill has a belief about some of his students waiting for him, it is not implausible that we assume that he also has beliefs about whether many, most or all of his students are waiting for him.31 Russell further supports his account with the observation that if the subject is known to lack an opinion about the stronger proposition, no implicature (or pseudo-implicature) is generated. His example: (48)
George has not yet formed an opinion about all of his advisors, but, at this point, he believes some of them are crooks.
The observation seems correct: there is no implication here that George believes that not all of his advisors are crooks. However, the explanation seems incomplete. Note first that the following is perfectly coherent: (49)
George has not yet formed an opinion about all of his advisors, but at this point, he believes most of them are crooks.
So, knowing that George lacks an opinion about all of his advisors does not rule out the possibility that he has an opinion about most of his advisors. This being so, Russell’s story doesn’t explain why (48) lacks the implicature that George believes it is not the case that most of his advisors are crooks. The solution here may be that the implicatures in question are not as finely differentiated as the standard story suggests. As interpreters, we take the subject to be either well-informed or not well-informed. In the latter case, we do not look for a strengthened reading. The same kind of explanation can be extended to my account.
162
Mandy Simons
It may be possible to adjudicate between my view and Russell’s if it can be established with certainty whether or not the observed implication is part of the content of the asserted belief claim. However, it should also be noted that my view and Russell’s are not incompatible. I am not claiming that implicatures cannot be calculated on the basis of the total asserted content, only that in some cases they may – perhaps in addition – be calculated on the basis of the form and content of an embedded clause. In some cases, multiple lines of conversational reasoning will converge on the same conclusion.
9. Local relevance implicatures The cases of intrusive implicature which we have discussed to this point are instances of generalized implicatures, cases where the use of a particular form typically gives rise to a given implicature. It is on cases such as these that the recent literature dealing with intrusive implicatures has focused. A good deal of this recent debate has been around the question of whether these generalized implicatures are indeed implicatures at all in Grice’s sense, or are rather some kind of conventional but defeasible content of particular expressions. If all of the relevant cases can be accounted for as cases of conventional content, then there is no need to find a way to account for local generation of conversational implicatures. To bolster the case that such an account is needed, I turn to relevance implicatures. Such implicatures are uncontroversially non-conventional and also non-generalized; contextual considerations must always play a role in their generation. As far as I am aware, no one has yet argued for the existence of local Gricean implicatures on the basis of local relevance implicatures.32 I believe that examples such as the following make the case: (50) A: How will you get to the party? B: Either I’ll borrow a car, or I’ll take the bus and walk from the bus station. B1: If I can’t borrow a car, I’ll take the bus and walk from the bus station. Consider the disjunctive response B. The first disjunct is incomplete as an answer to the question. To understand it as an answer, the interpreter must assume that the speaker will use the car she borrows to drive to the party. This is one of those almost undetectable cases of implicature – the inference is so immediate and obvious it is hard to see – but it is a relevance implicature nonetheless: something clearly intended to be conveyed, and conveyed on the assumption that the speaker intends her conversational contribution to be relevant to the question asked.
A Gricean View on Intrusive Implicatures
163
Now, we should contrast (50)B with (51) below: (51)
A: Did they have a picnic? B: Either it rained or it was cold.
B’s utterance here clearly implicates that there was no picnic. But this implicature can be understood as an implicature of the utterance as a whole. Either of the circumstances mentioned would (presumably) rule out a picnic; so, the interpreter can reason that if the disjunction is true, there was no picnic. But (50)B above is different. Here the implicature is generated by the first disjunct alone, and is required to establish its relevance. The implicature is not needed to establish the relevance of the utterance as a whole, except insofar as the relevance of the whole requires the independent relevance of each disjunct. The conditional (50)B1 demonstrates the same point. Here, the implicature must be internal to the antecedent of the conditional. Here is another example which illustrates the same point: (52)
[Context: scrabbling sounds are heard coming from overhead] A: What’s making all that noise? B: Either there’s a nest in the eaves or a family of squirrels has moved in there.
The first disjunct of B’s reply can be relevant only if she is taken to implicate that the nest, if there, is in use: an abandoned nest would not make any noise.
10. Conclusion I have argued here for a reconceptualization of conversational implicatures. Grice saw implicatures as arising primarily as a consequence of observations about what the speaker has said. The exception, for him, were Manner implicatures, which are generated by observations about how the speaker said what she said. I have suggested that we should instead see conversational implicatures as inferences generated by observations about what a speaker has (merely) expressed: this includes observations about sentence parts whose content is not part of what is asserted. The justification for this shift in view comes from the recognition that interpreters can, and do, identify the content, form and discourse function of non-asserted sentence parts and can reason about why the speaker has chosen just that content or just that form to carry out the particular discourse function. The maxims as formulated by Grice are not clearly inapplicable to nonasserted content. However, reformulations may be helpful in understanding
164
Mandy Simons
how they can be so applied. I have offered a reformulation of the first part of the maxim of Quantity, as follows: Quantity 1: Provide as much information about the situation you are describing as is required by current conversational purposes. In some cases (not considered here), implicatures generated by observations about non-asserted clauses may have ‘global’ effects, that is, be understood as commitments of the speaker.33 The cases examined here, however, involve locally generated implicatures which also have local effects. That is, they become part of the understood content of an embedded clause, falling under the scope of operators. This claim stands in conflict with Grice’s view that ‘what is said’ can be identified without recourse to conversational reasoning. It also stands in conflict with a compositional semantics in which only conventionally encoded content is ‘visible’ to sentential connectives and other embedding expressions. But perhaps this conclusion departs less from current semantic theory than one might expect. Semanticists have for some time countenanced different types of semantic content which can have truth-conditional effects. For example, semanticists (following Rooth 1985) allow that expressions can have, in addition to their ordinary semantic value, also a ‘focus semantic value’ which depends upon the placement of focus in the expression. Some sentences (for example, those containing focus-sensitive adverbs such as too and even) have truth values which are determined by both the ordinary semantic values and the focus semantic values of their parts. In the recent influential work by Chierchia alluded to several times in this chapter, sentence parts are assigned both an ordinary semantic value and a ‘strengthened’ or ‘scalar’ semantic value. Sometimes one and sometimes the other is relevant for determining the semantic content of the whole. One might object that moves of this sort are quite different from a move allowing particularized conversational inferences to contribute to truth-conditional content. But focus semantic values and scalar values are not themselves clearly context-independent. For example, although the focus semantic value of an expression is algorithmically determined given a set of alternatives, the identification of the set of alternatives is a contextual matter. There is, though, no doubt that the job of the formal semanticist becomes even harder once conversational contributions to ‘what is said’ are acknowledged. For the job becomes one of providing precise formulations of the semantic contributions of linguistic operators which operate on multiple levels of content simultaneously. This is undoubtedly complicated. But as Austin says at the conclusion of ‘Performative utterances’ (1970): Well, it is complicated a bit; but life and truth and things do tend to be complicated.
A Gricean View on Intrusive Implicatures
165
Notes 1. I choose these two cases because they have been very widely discussed in the literature. In particular, scalar implicatures are at the heart of a very vigorous current debate about the structure of the interpretative process. However, even if one is convinced that there are conversational implicatures of the standard Gricean sort, one might not be convinced that these two cases are best explained in Gricean terms. The enriched interpretation of coordination, for example, might seem better explained within a general theory of coherence. Questions can also be raised about the standard account of scalar implicatures. Happily for me, the argument I will present below is unaffected by the question of whether these particular cases are correctly analysed in Gricean terms. The argument I will give is an argument that the principles for the calculation of conversational implicatures are equally applicable to complete utterances with speech act force and to sub-parts of utterances. Whether these principles are actually applied in any specific case is tangential to my argument. 2. For details of the Gricean account, see Levinson (1983: 107–8). 3. I avoid using examples involving numerals, because of debate as to their truthconditional content. See for example Sadock (1984), Koenig (1991), Horn (1992). I also have some doubts about purported entailment relations in other cases of scalar implicature (e.g. that hot entails warm); so I will use quantifier examples throughout. 4. This description involves simplifications. For a more careful presentation, see either of the references in the text, or Sauerland (2004). 5. As we’ll see below, Sauerland considers this an implicature of the sentence as a whole, and therefore not a challenge to the standard picture. 6. For a thorough review of the debate, see Recanati (2004). 7. As Neale observes, this definition fails to distinguish what is said from what is conventionally implicated. This distinction need not concern us here, however. 8. The definition of this notion, too, is fraught with difficulty. See in particular Essays 5 and 6 in Grice (1989), Studies in the Way of Words, and Neale’s (1992) explication. 9. For details as to why this is so, see Sauerland (2004). In section 6.1. below, we will discuss Sauerland’s own account, which proposes a modification of standard assumptions to allow for the generation of this implicature from the utterance as a whole. But we will also raise doubts as to whether it is appropriate to understand the implicature in this way. 10. As many authors have pointed out, this caveat is non-trivial. For one discussion of this point, see Levinson (2000: section 3.2). 11. For a helpful summary and discussion of responses, see Recanati (2003). 12. Thomason (1990: fn.20), recalling that the proposal in Grice (1967) was offered as a rather preliminary outline and with some tentativeness about details, writes: ‘On the whole, I believe that linguists and computer scientists have taken the details of Grice’s theory more seriously than they perhaps should have’. 13. Walker himself doesn’t seem to have considered his proposal a modification of Grice’s views, but merely an articulation of them. However, in this he is at odds with the generally accepted interpretation of Grice. 14. Walker’s discussion also addresses the compositionality problem; but here, I think, his proposal is off-target. 15. Green takes issue with Recanati’s claim, reiterated above, that Grice himself conceived of conversational implicatures only as ‘pragmatic consequences of acts of
166
16. 17.
18.
19.
20.
21.
22. 23.
Mandy Simons saying something’. Green points out that in the ‘Retrospective epilogue’, Grice wrote: It certainly does not seem reasonable to subscribe to an absolute ban on the possibility that an embedding locution may govern the standard nonconventional implicatum rather than the conventional import of the embedded sentence ... (Grice 1989: 375) But it must also be acknowledged, I think, that in his primary presentations of conversational implicature, Grice writes only of them being generated on the basis of observations about what is said. In retrospect, he may indeed have thought that a more flexible theory was required. Note also that in this remark, Grice is envisioning embedding of ‘standard’ implicatures – presumably, generalized ones – and also thinking of cases where the operator operates on the implicature instead of the standard content. The proposal laid out below does not limit itself in this way. The observation that embedded clauses can have a main point content and embedding clauses an evidential function goes back at least to Urmson (1952). Recanati (2003) suggests that this approach, at least as articulated by Stalnaker, constitutes a commitment to the idea that (the utterance of) the antecedent constitutes an independent speech act. This idea, in turn, leads to an alternative solution to the problem of embedded implicatures. I am not here advocating this approach. I restrict myself to examples with clausal disjuncts. Non-clausal disjuncts can also generate scalar implicatures, as in: (i) The guests stole some of the teaspoons or the bathroom towels. Examples like these raise additional technical questions, which I don’t want to become embroiled in here. Of course, in the case of (16) there is extra motivation for taking the description to include particular temporal and causal relations: if the description is not so taken, then the two disjuncts describe the same state of affairs, and the utterance would be infelicitous (for general Gricean reasons). So, the interpreter is compelled to interpret the disjuncts differently. As the only difference between the disjuncts is the order of the conjuncts, then that difference must provide the difference in intended content of the disjuncts. An anonymous reviewer points out a possible problem for this account, which becomes evident in examples like (6) above. The reviewer observes that my account predicts for (6) the interpretation: Either Kai ate the broccoli or he ate exactly some of the peas. He presumes that the account would also allow for the derivation of the “but not both” implication, but worries that this would still allow for the possibility that Kai ate the broccoli and all of the peas. While I have not had time to work out a solution to this problem, I suspect that this reading will be ruled out by the requirement that the disjuncts be exhaustive relevant alternatives with respect to the question under discussion (see Simons 2000). Additional neo-Gricean proposals are made by van Rooij and Schulz (2004 and 2006) and by Spector (2003). Both make use of the notion of exhaustivity to account for strengthened readings, and in turn explain exhaustivity in Gricean terms. See also Groenendijk and Stokhof (1984) for an early formulation of this idea. For a summary of this account and objections to it, see Simons (2000: section 2.5, and references). To make this work technically, Sauerland is forced to introduce two novel binary L and
R , which constitute the lexical alternatives to or along with and. operators,
A Gricean View on Intrusive Implicatures
24. 25.
26.
27.
28. 29. 30. 31.
32.
33.
167
These operators, which do not correspond to any lexical items, are defined as follows: L B = A, A
R B = B. (i) For any two propositions A, B: A
Sauerland himself describes this move as ‘more of a technical trick than a real solution’. L . Technically, by replacement of or by
Sauerland also makes explicit the steps that are required to derive a ‘strong’ scalar implicature (of the form ‘K S(¬p)9 from the weaker implicature licensed by Quantity of the form ‘¬K s(p)’. The point can perhaps be made with a simpler example, such as: ‘Either she called some of the senators or she called all of the senators’. But in such examples, the quantifiers typically bear some kind of focus, and it has been suggested (King and Stanley 2006) that focus can be used to explain away some cases of apparent local implicatures. Hence, the more complex example. This is in fact a complicated example, as the first disjunct contains multiple quantifiers, a conjunction and a focus operator. So I am not entirely sure what implicature Sauerland would predict. But if in fact his account predicts a different implicature or no implicature at all, the problem remains: for the first disjunct does get the “some but not all” interpretation. See Levinson’s example: If John owns two cars, the third one in the driveway must be someone else’s (Levinson 2000: 206). This is because these are (roughly) downward entailing environments. His account is tailored to eliminate standard scalar implicatures in these environments. Chierchia does not say what he thinks is being accommodated: presumably, some sort of speaker presupposition. Russell discusses the specific example in (i): (i) George believes some of his advisors are crooks. He says ‘Speakers I’ve consulted readily agree that “Either George thinks all his advisors are crooks or he thinks they’re not all crooks” has to be true, suggesting that this is a “default” background assumption.’ I’m not sure what kind of default assumption Russell has in mind. Certainly, it seems implausible to assume that for every proposition p and every agent a, either a believes p or a believes not-p; that assumption would eliminate agnosticism. However, it may well be plausible that in a situation where it is known that a believes some proposition p, and q is a scalar alternative to p, that a is typically assumed to have a belief about q. To be clear: many people have argued that there are nonce inferences which become part of sentence content. See in particular, again, the work of the relevance theorists (Sperber and Wilson 1986), Robyn Carston (1988, 2002), Kent Bach (Bach 1994) and Francois Recanati (especially Recanati 2004). As far as I know, though, no-one has used such cases to advocate for a modified Gricean view. I have suggested elsewhere that this may be the correct account of presupposition projection in those cases where the presuppositions are best understood as arising from conversational inference.
References Anscombre, J.-C. and Ducrot, O. (1983) L’Argumentation dans la Langue (Brussels: Mardaga). Austin, J. L. (1970) Performative utterances. In J.O. Urmson and G.J. Warnock (eds.), Philosophical Papers, 2nd edn (Oxford: Oxford University Press), pp. 233–52.
168
Mandy Simons
Bach, K. (1994) Conversational impliciture. Mind and Language, 9: 124–62. Carston, R. (1988) Implicature, explicature and truth-theoretic semantics. In R. Kempson (ed.), Mental Representations: the Interface between Language and Reality (Cambridge: Cambridge University Press), pp. 155–82. Carston, R. (2002) Thoughts and Utterances: The Pragmatics of Explicit Communication. (Malden, MA: Blackwell). Chierchia, G. (2004) Scalar implicatures, polarity phenomena and the Syntax/ Pragmatics Interface. In A. Belletti (ed.) Structures and Beyond: The Cartography of Syntactic Structures, Volume 3 (Oxford: Oxford University Press), pp. 39–103. Cohen, L. J. (1971) The logical particles of natural language. In Y. Bar-Hillel (ed.) Pragmatics of Natural Language (Dordrecht: Reidel), pp. 50–68. Eggert, R. (2000) Grammaticality and context with respect to and ... and or ... respectively, Proceedings of Chicago Linguistics Society 36. Gazdar, G. (1979) Pragmatics: Implicature, Presupposition and Logical Form (New York: Academic Press). Grice, H. P. (1989) Studies in the Way of Words (Cambridge, MA.: Harvard University Press). Green, M. S. (1998) Direct reference and implicature. Philosophical Studies, 91: 61–90. Groenendijk, J. and Stokhof, M. (1984) Studies on the Semantics of Questions and the Pragmatics of Answers. PhD Dissertation, University of Amsterdam. Heim, I. (1992) Presupposition projection and the semantics of attitude verbs. Journal of Semantics, 9: 183–221. Horn, L. R. (1976) On the Semantic Properties of Logical Operators in English. PhD Dissertation, Yale University. Reproduced by Indiana University Linguistics Club, Bloomington. Horn, L. R. (1992) The said and the unsaid. Proceedings of Semantics and Linguistic Theory (SALT) 2: 163–92. King, J. C. and Stanley, J. (2006) Semantics, pragmatics and the role of semantic content. In Z. Szabó (ed.), Semantics vs. Pragmatics (Oxford: Oxford University Press), pp. 111–64. Koenig, J.-P. (1991) Scalar predicates and negation: Punctual semantics and interval interpretations. Papers from the 27th regional meeting of the Chicago Linguistics Society (CLS 27), Part 2: The Parasession on Negation: 140–55. Kratzer, A. (1991) Modality. In A. von Stechow and D. Wunderlich (eds.) Semantik/ Semantics: An International Handbook of Contemporary Research (Berlin: de Gruyter), pp. 639–50. Levinson, S. (1983) Pragmatics (Cambridge: Cambridge University Press). Levinson, S. (2000) Presumptive Meanings: The Theory of Generalized Conversational Implicature (Cambridge, MA: MIT Press). Neale, S. (1992) Paul Grice and the philosophy of language. Linguistics and Philosophy, 15(5): 509–59. Recanati, F. (2003) Embedded implicatures. Philosophical Perspectives 17(1): 299–332. Recanati, F. (2004) Literal Meaning (Cambridge: Cambridge University Press). Rooij, R. van and Schulz, K. (2004) Exhaustive interpretation of complex sentences. Journal of Logic, Language and Information, 13: 491–519. Rooij, R. van and Schulz, K. (2006) Pragmatic meaning and non-monotonic reasoning: the case of exhaustive interpretation. Linguistics and Philosophy, 29(2). Rooth, M. (1985) Association with Focus. PhD Dissertation, University of Massachusetts, Amherst. Russell, B. (2006) Against grammatical computation of scalar implicatures. Journal of Semantics, 23(4): 361–82.
A Gricean View on Intrusive Implicatures
169
Sadock, J. (1984) Whither radical pragmatics? In D. Schiffrin (ed.) Meaning, Form and Use in Context: Linguistic Applications (GURT ‘84) (Washington: Georgetown University Press). Sauerland, U. (2004) Scalar implicatures in complex sentences. Linguistics and Philosophy, 27(3): 367–91. Simons, M. (2000) On the Semantics and Pragmatics of Natural Language Disjunction. (New York: Routledge). Simons, M. (2005a) Dividing things up: The semantics of or and the modal/or interaction. Natural Language Semantics, 13: 271–316. Simons, M. (2005b) Semantics and pragmatics in the interpretation of or. Proceedings of Semantics and Linguistic Theory (SALT) 15: 205–22. Simons, M. (2007) Observations on embedding verbs, evidentiality and presupposition. Lingua, 117(6): 1034–56. Spector, B. (2003) Scalar implicatures: exhaustivity and Gricean reasoning? In B. ten Cate (ed.) Proceedings of the ESSLLI 2003 Student Session (Vienna). Sperber, D. and Wilson, D. (1986) Relevance: Communication and Cognition (Oxford: Blackwell). Stalnaker, R. (1974) Pragmatic presuppositions. In M. K. Munitz and P. K. Unger (eds.) Semantics and Philosophy (New York: New York University Press). Thomason, R. (1990) Accommodation, meaning and implicature: Interdisciplinary foundations for pragmatics. In P. R. Cohen, J. Morgan and M. E. Pollack (eds.) Intentions in Communication (Cambridge, MA: MIT Press), pp. 325–64. Urmson, J. O. (1952) Parenthetical verbs. Mind, New Series 61 (244) (Oct.): 480–96. Walker, R. (1975) Conversational implicature. In S. Blackburn (ed.) Meaning, Reference and Necessity (Cambridge: Cambridge University Press), pp. 133–81. Wilson, D. and Sperber, D. (1998) Pragmatics and time. In R. Carston and S. Uchida (eds.) Relevance Theory: Applications and Implications (Amsterdam: John Benjamins), pp. 1–22.
7 Speaker-Meaning, Conversational Implicature and Calculability1 Jennifer Saul
It is widely believed that H.P. Grice’s notion of speaker-meaning divides exhaustively into what is said and what is implicated. 2 Call this ‘Speaker Meaning Exhaustiveness’. And yet, as I have argued elsewhere (Saul 2002) this seems to be in clear conflict with Grice’s famous three-clause characterization of conversational implicature (1989: 30–1), which is also widely accepted. Legitimate worries can be raised, however, about the correct interpretation of the three-clause characterization, ones that may seem to cast doubt on the conflict just described. However, as I will argue here, there is also a conflict between Grice’s Calculability Criterion and Speaker Meaning Exhaustiveness. And the worries that arise for the three-clause characterization do not apply to the Calculability Criterion. We are, then, still left with a conflict between Speaker Meaning Exhaustiveness and key Gricean claims about conversational implicature. At the end of this chapter, I explore ways of resolving this conflict.
1. Initial conflict: Speaker Meaning Exhaustiveness and the three-clause characterization 1.1
The conflict3
Speaker Meaning Exhaustiveness is the claim that speaker-meaning divides exhaustively into what is said and what is implicated. Grice’s three-clause characterization of conversational implicature is as follows: A man who, by (in, when) saying (or making as if to say) that p has implicated that q, may be said to have conversationally implicated that q, provided that (1) he is to be presumed to be following the conversational maxims, or at least the Cooperative Principle; (2) the supposition that he is aware that, or thinks that, q is required to make his saying or making as if to say p (or doing so in those terms) consistent with this presumption; and (3) the speaker thinks (and would expect the hearer to think that the speaker thinks) that it is within the competence of the hearer to 170
Speaker-Meaning, Conversational Implicature, Calculability
171
work out, or grasp intuitively, that the supposition mentioned in (2) is required. (Grice 1989: 30–1) A conflict between Speaker Meaning Exhaustiveness and the three-clause characterization can be generated, we will see, by focusing either on (1) or (2). 1.1.1 Conflict with Clause 1 First, imagine that I am writing a letter of reference for my student Amanda, a very poor philosopher. I want to make this clear to her prospective employers without explicitly saying it. So I write a letter consisting solely of (a): (a)
Amanda has never been more than half an hour late to her appointments with me, and her dissertation fits nicely within the university’s word limit.
In writing this, I mean that Amanda is poor at philosophy. I fully expect the hiring committee to realise this. However, unbeknownst to me the hiring committee has been falsely informed that I always write pointless and uncooperative letters of reference. As a result, they do not take me to be implicating this. They take my apparently irrelevant letter to be a product of my uncooperativeness rather than supposing that I am trying to convey something that I don’t want to say. Clause 1 has, it seems, not been met: There is no presumption of cooperativeness.4 According to the three-clause characterization, then, I have not conversationally implicated that Amanda is poor at philosophy. Nonetheless, I surely meant this. According to Speaker Meaning Exhaustiveness, anything that the speaker means but does not say must be implicated. There are two main kinds of implicature, conventional and conversational. Conventional implicatures are wholly non-cancellable, always carried by a particular form of words. Grice gestures at a third sort of implicature (1989: 28), which is meant to function like conversational implicature but to be arrived at by reference to different maxims. This example seems to show that Speaker Meaning Exhaustiveness conflicts with the three-clause characterization. I clearly did not say that Amanda is poor at philosophy. Amanda is poor at philosophy is definitely not a conventional implicature of (1). If the three-clause characterization holds, this cannot be a conversational implicature of (1) either. Since the problem for considering it a conversational implicature involved Clause 1 rather than anything about specific maxims, there is no room for it to be an implicature of the ill-defined third sort. Nonetheless, I did mean it. It is, then, something I mean but neither say nor implicate: a counter-example to the conjunction of the three-clause characterization and Speaker Meaning Exhaustiveness.
172
Jennifer Saul
1.1.2 Conflict with Clause 2 Imagine that I am writing a letter of reference for another inadequate student, Beau. I take it that Beau is applying for a philosophy job, and I want to convey that Beau is a poor philosopher without saying it. So I write only (b): (b) Beau is always punctual and he is very good with word-processing software. By this, I mean that Beau is a poor philosopher. I expect the hiring committee to realise this. What I failed to realise, however, is that Beau is actually applying for a low-level secretarial job, requiring only punctuality and knowledge of word-processing software. His prospective employers find my letter very helpful, and have no reason to suppose that I meant to convey anything about Beau’s philosophical abilities. Clause 2 was not met, so according to the three-clause characterization I have not conversationally implicated that Beau is a poor philosopher. Once more, we get a failure of Speaker Meaning Exhaustiveness (on the assumption that the three-clause characterization holds). I clearly meant that Beau is a poor philosopher. And yet I neither said, conventionally implicated, nor conversationally implicated this. Nor does it seem likely to be an implicature of the third sort: it is hard to imagine any maxims, conversational or otherwise, that my audience might take the letter to violate. 1.2
The problems
The arguments above seemed very convincing to me in 2002. In fact, I still think that they are right. But three criticisms that I have received have shown that these arguments depend on interpretations of Grice that may be contentious. 1.2.1 ‘Provided that’ The three conditions of Grice’s three-clause characterization are introduced as follows (emphasis mine): A man who, by (in, when) saying (or making as if to say) that p has implicated that q, may be said to have conversationally implicated that q, provided that I have interpreted each of these conditions as a necessary condition for conversational implicature, and the two arguments given above depend on this interpretation. But ‘provided that’ is standardly understood by logicians as introducing sufficient conditions.5 If that is in fact what Grice was doing, then my arguments don’t work. As a matter of fact, in ordinary language ‘provided that’ does often introduce necessary conditions. A person who says ‘you may vote, provided that
Speaker-Meaning, Conversational Implicature, Calculability
173
you are a citizen and properly registered’ is offering a necessary condition for voting. Most commonly, it is used to introduce conditions which are both necessary and sufficient.6 (Indeed, this is a natural reading of the voting example.) In addition, it is standard to interpret Grice’s conditions as both sufficient and necessary for conversational implicature.7 But neither of these considerations is wholly decisive. It remains possible that Grice did intend his conditions to be sufficient rather than necessary. And if this is right, then I have not yet shown there to be a problem with Speaker Meaning Exhaustiveness. The examples above are both cases in which something I mean fails to be a conversational implicature because one of the three conditions is not met. If these conditions are not necessary, then I have given no reason to suppose that what I mean is not conversationally implicated. 1.2.2 Who presumes? Who must suppose? In the examples in section 1.1, I make much of the fact that, in the first example, the audience does not presume the speaker to be cooperative; and the fact that, in the second example, the audience does not need to assume the speaker to believe that Beau is a poor philosopher. These are the reasons that I take conditions 1 and 2, respectively, not to be met: 1 2
he is to be presumed to be following the conversational maxims, or at least the Cooperative Principle; the supposition that he is aware that, or thinks that, q is required to make his saying or making as if to say p (or doing so in those terms) consistent with this presumption; (Grice 1989: 30–31)
However, conditions 1 and 2 make no mention of the audience. In fact, they say nothing about who is doing the presuming, or who must need to make the assumption mentioned in 2. I assumed, in both cases, that the answer was ‘the audience’. Another seemingly reasonable answer, however, would be, ‘the omniscient theorist’.8 The omniscient theorist, of course, would know that I am in fact being cooperative in the first example; and also that the best way to make sense of my utterance in the second example is to take me to think that Beau is a poor philosopher. If this interpretation is right, then, my counter-examples to Speaker Meaning Exhaustiveness don’t seem to work. In fact, I think there is good reason to reject this interpretation. It seems to me very strange to cast conditions 1 and 2 as claims about the omniscient theorist, rather than the audience. It is strange because an omniscient theorist has no reason to suppose or to presume anything – she simply knows all the facts, since she is omniscient. It is hard to see how to even begin to make sense of the idea that there could be things that the omniscient theorist is
174
Jennifer Saul
required to suppose, given her lack of need for suppositions. Nevertheless, if I can provide an argument that does not rely on my interpretation of 1 and 2, that seems worth doing. 1.2.3 ‘To be presumed’ Clause 1 is easy to misread.9 Indeed, I did so myself for many years. I have even misquoted it, as I did in my (2002).10 Here is how I (mis)quoted it in that paper: (1M) he is presumed to be following the conversational maxims, or at least the Cooperative Principle (Saul 2002: 231). Here is what it actually says (emphasis mine): 1 he is to be presumed to be following the conversational maxims, or at least the Cooperative Principle (Grice 1989: 31). What is the difference between (1M) and 1? Well, one difference is surely the strangeness and awkwardness of 1. But we can also discern a more substantive difference. (1M) is a simple descriptive condition – that the speaker is in fact presumed to be cooperative. Clause 1 is more puzzling. It seems to be a more normative condition. The most natural reading, it seems to me, is that the condition holds if the speaker should be presumed to be cooperative. This seems at first a bit strange, as we are all meant to be presuming speakers to be cooperative pretty much all the time, according to Grice. But we can make sense of it by focusing on the ‘pretty much’ in my previous sentence: the presumption of cooperativeness can be cancelled. If there has been some clear indication that the speaker is not being cooperative it is no longer the case that we should presume the speaker to be cooperative. The most natural reading of (a), then, is that there has been no indication of non-cooperativeness. If this is right, then my example in section 1.1 survives the correction of my misreading of 1: The audience had been given a clear (albeit false) indication that the speaker was uncooperative. If this reading is right, then there is no problem here for my earlier arguments. However, I am not wholly sure of my reading of this clause.
2. Speaker Meaning Exhaustiveness and the Calculability Criterion There is no doubt whatsoever (at least among Griceans)11 about the status of the Grice’s Calculability Criterion as a necessary condition for conversational implicature. Moreover, there are no strange phrases like ‘to be presumed’ (although there, are of course, some interpretational issues lurking).
Speaker-Meaning, Conversational Implicature, Calculability
175
The following passage from Grice gives what I will call the Calculability Criterion: The presence of a conversational implicature must be capable of being worked out; for even if it can in fact be intuitively grasped, unless the intuition is replaceable by an argument, the implicature (if present at all) will not count as a conversational implicature; it will be a conventional implicature (1989: 31). The Calculability Criterion requires, then, that in order for Q to be a conversational implicature of some utterance U, it must be possible to work Q out from U. Grice continues: To work out that a particular conversational implicature is present, the hearer will rely on the following data: (1) the conventional meaning of the words used, together with the identity of any references that may be involved; (2) the Cooperative Principle and its maxims; (3) the context, linguistic or otherwise, of the utterance; (4) other items of background knowledge; and (5) the fact (or supposed fact) that all relevant items falling under the previous headings are available to both participants and both participants know or assume this to be the case (ibid.: 31). It is natural to read the above as an explanation of how it is that it must be possible to work Q out from U, in order for the Calculability Criterion to be satisfied.12 If this is right, then in order for Q to be a conversational implicature of U, it must be possible to work Q out by relying on items (1)–(5), above. 2.1
The conflict
The Calculability Criterion clearly generates a conflict with Speaker Meaning Exhaustiveness. And we can show this using one of the examples we have already seen. Recall my letter of reference for Beau, a bad philosophy student. In that letter, I truthfully wrote (b): (b) Beau is always punctual and he is very good with word-processing software. I wrote this thinking that my letter was going to a philosophy hiring committee, and therefore that it would convey my view that Beau is a terrible philosopher. Unbeknownst to me, however, Beau was applying for a lowlevel secretarial position so his new employers took my letter to be highly relevant praise for Beau’s abilities. We’ll begin with the wrong way to generate a clash. It is clear that Beau’s new employers failed to work out my attempted implicature that Beau is
176
Jennifer Saul
a terrible philosopher. One might (mistakenly) read Grice’s Calculability Criterion as requiring that a putative implicature actually be calculated in order to count as a conversational implicature. If this were right, then the mere fact that my attemped implicature was not calculated would be enough to show that it was not a conversational implicature. Since I meant it, and since it was neither said nor conventionally implicated, this might seem to show already that there are things that are meant but neither said nor implicated. But this is a mistaken interpretation of the Calculability Criterion. Calculability very clearly does not require that an implicature actually be calculated. Instead it requires that it be possible for an implicature to be calculated. And Grice makes this abundantly clear by noting that even if the implicature can be intuitively grasped, the intuition must be ‘replaceable by an argument’ (1989: 31, emphasis mine). The right way to generate a clash is to notice that Beau’s employers not only didn’t work out my attempted implicature, but also that they couldn’t have done this. To see this, consider the information that Grice says audiences are to rely on in working out implicatures: (1) the conventional meaning of the words used, together with the identity of any references that may be involved; (2) the Cooperative Principle and its maxims; (3) the context, linguistic or otherwise, of the utterance; (4) other items of background knowledge; and (5) the fact (or supposed fact) that all relevant items falling under the previous headings are available to both participants and both participants know or assume this to be the case. (ibid.: 31) The conventional meanings of the words in my brief reference (1) yield a useful and relevant claim for someone seeking a low-level secretary. The Cooperative Principle (2) gives no further insights, since what I wrote seems very cooperative. The context (3) is one in which a hiring committee is reading a letter of recommendation for a secretarial post – again no further insights come from this. They do not have any background knowledge (4) which could alert them to the fact that I may have meant something else –they have no idea that I was misinformed. They take all these facts or supposed facts (5) to be known to me as well. There is nothing, then, that would enable the hiring committee to work out my intended implicature that Beau is a terrible philosopher.13 Although I meant this, then, I neither implicated nor said it. We have, therefore, a clear conflict with Speaker Meaning Exhaustiveness. 2.2 An objection: who should be able to calculate? Despite other ways in which the Calculability Criterion is admirably clear, it shares a frustrating unclarity with Clause (2) of the three-clause characterization: it does not specify who must fulfil it. That is, it does not say
Speaker-Meaning, Conversational Implicature, Calculability
177
who must be able to calculate the implicature. I have so far assumed that it is the audience who must be able to do this. Since they cannot, in the example above, there is a clash between Calculability and Speaker Meaning Exhaustiveness. However, just as with Clause (2), one could suggest that the relevant person is not the audience but instead the omniscient theorist. The omniscient theorist, one might suppose, is easily able to calculate the implicature: unlike the actual audience, she knows that I thought I was writing for a committee of philosophers. She will, then, realise that I meant to convey that Beau is a terrible philosopher. My attempted implicature will in fact be a conversational implicature, and there will be no clash with Speaker Meaning Exhaustiveness. But this interpretation of the Calculability Criterion is difficult to support by looking at Grice’s text. Immediately after stating the Calculability Criterion, after all, Grice goes on to list what information is to be used by the audience in calculating conversational implicatures. This strongly suggests that the audience is the one doing the calculating. This interpretation also yields a very unappealing theory. The omniscient theorist, after all, knows everything. This means that she knows what the speaker intends to communicate. No matter how poorly chosen an utterance is, the omniscient theorist has no trouble calculating it. If it is the omniscient theorist’s calculating abilities that are relevant for Calculability, then it is not a very revealing test. To see this, imagine that I am writing a reference by which I mean to communicate that Charla is a terrible philosopher. The reference is for a philosophy job, and I know this. But I am very poor at writing bad references, and somehow always find myself writing glowing, positive references. I am firmly convinced, however, that the glowing reference I have written will communicate my low opinion of Charla. As it happens, I am of course wrong: there is nothing that would alert the audience to my low opinion of Charla. This is just the sort of case that should clearly fail the Calculability test. And yet, if ‘calculable’ means calculable by an omniscient theorist, it does not. The omniscient theorist has no difficulty working out the low opinion of Charla that I am trying to communicate – since she is omniscient. If Calculability is to be of any use at all as a test, then, it must be understood as calculable by the audience.14 This objection to my use of the Calculability requirement, then, fails.
3. Resolving the conflict We have seen that there is clearly a conflict between Speaker Meaning Exhaustiveness and other claims that Grice makes. In this chapter, however, my argumentative focus has been on the conflict between Speaker Meaning Exhaustiveness and Calculability. Obviously, there is more that one way to resolve such a conflict. We could give up either Speaker Meaning
178 Jennifer Saul
Exhaustiveness or Calculability. Both of these claims have been quite fundamental to standard discussions of Gricean theory. 3.1 Abandon Calculability? If we abandon Calculability, then (obviously) it is no longer the case that any putative conversational implicature must be calculable by the audience. Does this mean that we can preserve Speaker Meaning Exhaustiveness? It depends. The issue is the status of the three conditions for conversational implicature discussed earlier: A man who, by (in, when) saying (or making as if to say) that p has implicated that q, may be said to have conversationally implicated that q, provided that (1) he is to be presumed to be following the conversational maxims, or at least the Cooperative Principle; (2) the supposition that he is aware that, or thinks that, q is required to make his saying or making as if to say p (or doing so in those terms) consistent with this presumption; and (3) the speaker thinks (and would expect the hearer to think that the speaker thinks) that it is within the competence of the hearer to work out, or grasp intuitively, that the supposition mentioned in (2) is required. (Grice 1989: 30–1) In order to preserve Speaker Meaning Exhaustiveness, we must either abandon conditions 1 and 2 above, or make use of interpretations of them on which they do not conflict with Speaker Meaning Exhaustiveness. 3.1.1
Abandoning conditions 1 and 2
Suppose that we abandon Grice’s 1 and 2. What are we left with? We might think at first that Grice’s third condition remains. But this condition would be very strange without condition 2 in place, so presumably it too should be dropped. If we do this, we are left with a notion much like Wayne Davis’s speaker implicature: ‘S implicates that p iff S means or implies that p by saying something other than p’ (Davis 1998: 13). If we use this as our definition of ‘conversational implicature’ (which Davis does not do), Speaker Meaning Exhaustiveness is indeed preserved. The price of preserving Speaker Meaning Exhaustiveness is that we must give up much that has been central to discussions of conversational implicature. We give up not only Calculability, but also (in all likelihood) Grice’s three conditions. In so doing, we give up the idea that there is more to being a conversational implicature than being meant but not said by some particular utterance. Whatever speakers mean by their utterances but do not say, they conversationally implicate. When we think about how idiosyncratic speakers may be, we get some very implausible implicatures. To see this, imagine that I am obsessed with the 2000 US Presidential Election. I am convinced that Al Gore won the election, and I think about
Speaker-Meaning, Conversational Implicature, Calculability
179
this nearly all the time. (So far, we’re pretty true to the facts of my sad life.) Further (and here we thankfully depart from the facts), every time I make an utterance containing the words ‘count’, ‘vote’, or ‘Florida’, I mean to convey what (c) expresses: (c) Al Gore won the 2000 election. So for example, when I utter any of (d)–(f), I mean to convey (c): (d) My son is learning to count. (e) Women in France got the vote very late. (f) Trudy lives in Florida. How do I mean to convey (c) by uttering (d)–(f)? It doesn’t matter. Perhaps I stress the key words in an angry tone of voice and expect this to remind my audience of the travesty that took place, also expecting them to realise that this is my intention. Perhaps I think everyone I talk to agrees with me and is as obsessed as I am, and that they will automatically remember the recount when they hear the trigger words, and know that this is what I intended them to do. Perhaps I think that everybody knows my psychological tendencies and therefore knows what I mean by these things. All of these options are crazy. None of them looks very much like a conversational implicature. But according to the sort of theory we need in order to preserve Speaker Meaning Exhaustiveness, these are indeed conversational implicatures. And we really do need to say this in order to preserve Speaker Meaning Exhaustiveness: after all, these are all claims that I mean but do not say or conventionally implicate. Nor do I convey them in a manner analogous to conversational implicature, by exploiting maxims other than conversational ones. They must, then, be conversational implicatures. This seems deeply counter-intuitive. It also robs the notion of conversational implicature of much of its theoretical interest. If this is all there is to conversational implicature, why not just talk about what speakers mean and what speakers say? Why bother with the jargon, the diagrams, and the theory? Finally (though perhaps least significantly), it makes Grice’s discussion of the three conditions look like simply a huge error. 3.1.2
Reinterpreting Grice’s conditions
Reinterpreting Grice’s conditions in light of the worries raised earlier may initially seem appealing. However, it turns out that only one of these reinterpretations could help us to save Speaker Meaning Exhaustiveness. We’ll take the worries in reverse order. First, there is the fact that Grice’s first condition pertains to what is to be presumed about the speaker, rather than to what is presumed about the speaker. But we already saw that this made no difference to my argument (section 1.1.1) making use of this condition. Next,
180
Jennifer Saul
there is the possibility that Grice meant the person doing the presuming in condition 1 to be the omniscient theorist, rather than the audience. We saw, however, that this left us with a rather unappealing theory. The possibility of reading Grice’s conditions as merely sufficient for conversational implicature is, however, an intriguing one. On this reading, what Grice is doing is telling us one particular way of producing a conversational implicature, rather than giving us necessary conditions for doing so. On this reading, the cases discussed in sections 1.1.1 and 1.1.2 are no longer ruled out as cases of conversational implicature, and so no longer counter-examples to Speaker Meaning Exhaustiveness. If we have already abandoned Calculability, then there is no reason to reject Speaker Meaning Exhaustiveness. The price of this move is slightly less than the price of simply abandoning Grice’s conditions: we don’t have to treat Grice’s discussion of these conditions as simply a mistake. But we do still face the problem of defining ‘conversational implicature’. Unless some more restrictive understanding of ‘conversational implicature’ can be offered, we will face the same problems that we saw in section 3.1.1. 3.2 Abandon Speaker Meaning Exhaustiveness On this alternative, we keep Calculability and Grice’s three conditions for conversational implicature, at the cost of abandoning Speaker Meaning Exhaustiveness. But we may seem to lose some of the motivation for a notion of conversational implicature. It is easy to understand the need for a notion of what speakers mean but neither say nor conventionally implicate. If this could be systematized in the sort of way that Grice suggests, using his three conditions and his Calculability test, then this would be really interesting and theoretically fruitful. But we have seen that the notion – conversational implicature – that Grice describes using Calculability and his three conditions cannot capture all the things that speakers mean but neither say nor conventionally implicate. So what can it do? What is this notion good for? Conversational implicature, I have argued elsewhere, is a normative notion (see Saul 2002). Conversational implicatures are, among other things, claims that audiences are required15 to assume the speaker to believe, in order to make sense of the speaker’s utterances. Because of this, they are claims that the audiences should arrive at, but may not. One way of putting this is that they are claims that the speaker makes available to the audience. The best way to see the utility of this notion is to look at cases where audiences do not work out something that the speaker means, but where they should do so. Take, for example, the following fictitious letter of reference, from an article in The Onion: Karyn has an unusually insistent style of conducting business, undeniably effective in both achieving her goals and giving those she works
Speaker-Meaning, Conversational Implicature, Calculability
181
with a greater awareness of her value, the value of their lives, how much they love their wife and two daughters, and how desperately they want to live. (Anonymous 2007) The author of the above letter seems to conversationally implicate that Karyn has threatened them. If we assume that this was the writer’s plan, then we have a clear case of conversational implicature. The author is to be presumed to be cooperative (condition 1 is met); the audience must assume the author to believe that Karyn threatened them in order to understand the letter as cooperative (condition 2 is met); and, finally, the author knew that the audience could work this out (condition 3 is met). The implicature is clearly calculable. And yet, someone might miss it. Perhaps a careless reader might focus only on the praise early in the sentence, without reflecting much on what comes later. Perhaps a reader might think: ‘How nice, I like being aware of how much I love my family and life itself!’, without pausing to consider how it is that Karyn makes people aware of this. Understanding conversational implicature as a normative notion allows us to say that the author did implicate that Karyn had threatened him, even though the audience failed to pick up on it. We can say that because the conversational implicature was there, the audience should have picked up on it. Note that on the unrestrictive understanding of ‘conversational implicature’ we considered in section 3.1.1, it is simply not very interesting to say that a claim is conversationally implicated. According to that understanding, to say that a claim is conversationally implicated is simply to say that the speaker means it by what she says. This is very easy to do, and does not involve any requirement that the claim be made available to the audience. On this understanding, the fact that a claim is conversationally implicated does not mean that the audience should pick up on it. Recall my Gore 2000 examples in section 3.1.1: It is certainly not the case that the audience should realise that I mean Al Gore won the 2000 election every time I say ‘my son is learning to count’, and yet on this unrestrictive understanding it is conversationally implicated. The normative understanding of ‘conversational implicature’, based on Grice’s three-clause characterization, allows us to make claims about what intended messages the audience should pick up. In so doing, it gives us something that the unrestrictive understanding does not.
4. Conclusion It is clear that Speaker Meaning Exhaustiveness conflicts with Grice’s Calculability Criterion. Although both of these claims are very widely attributed to Grice and widely accepted by Griceans, they cannot be true at once.16 I have argued that abandoning Calculability will not on its own be enough to preserve Speaker Meaning Exhaustiveness. In order to
182
Jennifer Saul
do this, other significant changes would need to be made as well – either abandoning Grice’s conditions for conversational implicature, or treating them as merely sufficient. If either of these options is taken, Speaker Meaning Exhaustiveness can be maintained, but I have suggested this will come at a significant cost. It seems to me that abandoning Speaker Meaning Exhaustiveness is the preferable alternative. This leaves us with a plausible, interesting theory of conversational implicature as a normative notion.17
Notes 1. I am very grateful to many people for discussion of this paper and related issues, including especially: Kent Bach, David Braun, Wayne Davis, Ray Drainville, Chris Hookway, Larry Horn, and Marina Sbisa. I am also very grateful for my Philip Leverhulme Prize, which helped fund my work on this contribution. 2. See, for example, Bach’s rendition of the received view in his (1994), Horn (1992: 165), Levinson (1983: 131), and Neale (1990: 73–83, 1992). 3. Arguments for this conflict can be found, in more detail, in Saul (2002). 4. I question this interpretation of condition 1 below, in section 1.2.3. 5. See, for example, Lepore (2000: 83); I thank David Braun for emphatically impressing this upon me. 6. As David Braun has pointed out to me, one might take ‘provided that’ to express if but implicate if and only if. If this is right, then Grice merely implicated that the three conditions were both necessary and sufficient. 7. For example, all of the following replace ‘provided that’ with ‘iff’ or ‘just in case’: Davis (1998: 13), Levinson (1983: 113, 2000: 15), Green (1996: 95). 8. I thank Larry Horn for pointing this out. 9. Marina Sbisa first pointed this out to me. The importance of ‘to be presumed to be’ is discussed in Green (2002) and Davis (2007). 10. Others who neglect the extra ‘to be’ include: Davis (1998: 13), Davis (2005), Levinson (The very notable exception to this chronicle of neglect is Mitch Green, who correctly renders the criterion in his 1995 paper (1995: 98) and who discusses the importance of this correct rendering in his (2002) review. 11. Wayne Davis argues strongly against it in his 1998 paper and 2007. 12. And, indeed, this is how Grice is commonly interpreted. 13. As Wayne Davis has pointed out to me, Beau’s potential employers could work out my implicature if they had more information available to them – for example, if they knew that I thought I was writing a letter for a philosophy job. But if the Calculability Criterion is to yield any substantive results, it seems to me that it must focus on the information actually available to audiences. If instead we focus on what they could work out if they had all relevant information, every putative implicature will pass the test (since some highly relevant pieces of information are what the speaker believes, assumes, or wants to communicate by her utterance). If instead we consider what the audience could calculate if they had some, but not all information available to them, we need some non-questionbegging way to specify which information we are to consider as available. 14. An alternative view (suggested to me by Wayne Davis) would posit not an omniscient theorist, but instead one who knows all the data relevant to her theory. Such a person, perhaps, would know whether or not a speaker is being cooperative, but would still need to work out exactly what was being implicated. This
Speaker-Meaning, Conversational Implicature, Calculability
183
seems potentially promising, but I am not sure that this theorist’s knowledge could be specified in a non-question-begging manner. One might suppose that she should have available all the items Grice specifies as involved in the working out of conversational implicatures. But one of these items is ‘relevant background knowledge’. Why shouldn’t what the speaker believes, or what the speaker wants to communicate, be such an item? Finding a specification of the theorist’s knowledge which includes just enough but not too much information seems to me quite a daunting task. 15. There are, of course, serious problems with the notion that any such claims are required. See, for example, Atlas (2005: 70–3), Davis (1998). I don’t yet have a solution to these problems, sadly. 16. At least not without making significant changes elsewhere in the theory. 17. Though of course this notion too faces problems, most notably that mentioned in note 15.
References Anonymous (2007) Letter of reference clearly written under duress. The Onion, 43: 02. Atlas, J. (2005) Logic, Meaning, and Conversation: Semantical Underdeterminacy, Implicature, and Their Interface (Oxford: Oxford University Press). Bach, K. (1994) Conversational impliciture. Mind and Language, 9:124–62. Davis, W. (2005) Implicature. In E. N. Zalta (ed.) The Stanford Encyclopedia of Philosophy (Summer 2005 edn). URL: . Davis, W. (1998) Implicature: Intention, Convention, and Principle in the Failure of Gricean Theory (Cambridge: Cambridge University Press). Davis, W. (2007) How normative is implicature? Journal of Pragmatics, 39 (10): 1655–72. Green, G. (1996) Pragmatics and Natural Language Understanding, 2nd edn. (Mahwah, NJ: Erlbaum). Green, M. (1995) Quantity, volubility, and some varieties of discourse. Linguistics and Philosophy, 18: 83–112. Green, M. (2002) Review of W. Davis, Implicature: Intention, Convention, and Principle in the Failure of Gricean Theory (Cambridge: Cambridge University Press 1998). Grice, H. P. (1989) Studies in the Way of Words (Cambridge, MA: Harvard University Press). Horn, L. (1992) The said and the unsaid. In SALT [Semantics and Linguistics Theory] II, pp. 163–92 (Columbus: The Ohio State University). Lepore, E. (2000) Meaning and Argument (Oxford: Blackwell). Levinson, Stephen C. (1983) Pragmatics (Cambridge: Cambridge University Press). Levinson, Stephen C. (2000) Presumptive Meanings: The Theory of Generalized Conversational Implicature. (Cambridge, MA: MIT Press). Neale, S. (1990) Descriptions (Cambridge, MA: MIT Press). Neale, S. (1992) Paul Grice and the philosophy of language. Linguistics and Philosophy, 15: 509–59. Saul, J. (2002) Speaker meaning, what is said, and what is implicated. Noûs, 36(2): 228–48.
8 Some Aspects of Reasons and Rationality Judith Baker
1. Introduction In the 1970s Paul Grice and I discussed, in private conversations as well as joint seminars, a number of questions whose focus was rationality, most prominently, questions about wanting, evaluation, reasons, and incontinence. We compiled the notes for these conversations and entitled one selection of them ‘Some aspects of reasons and rationality’. It looks most closely at the relation of practical reasons to rationality, and provides the basis for the ideas developed here. This chapter takes as its focus one particular claim and the arguments that support it: rationality is displayed in a wider variety of actions than those done for reasons. One of the most common assumptions of contemporary writers on rational action is that the central, paradigm case of rational action is acting on reasons. Philosophers have taken opposing sides, however, on the question whether reasons for acting are provided by the fact that the agent wants something, or would want it under certain conditions. Many, if not most, Anglophone philosophers accept the desire-based account of reasons. For them, a typical reason for going downtown, for example, is given by the fact that I want to see a film. Some philosophers present a dramatically opposed story: (all) reasons for action are provided by facts about what is valuable. And being of value is not simply being desired (Scanlon 1988). The source of justification lies in facts about what one wants, not in the fact that one wants something. Much of the debated ground needs a new look. Grice and I challenged the assumption that is common to both sides of this argument, that all rational action is action based on reasons. We argued that a great many actions are not best understood as ones in which the agent acted for reasons – and yet they can be understood as rational, intelligible, and open to rational criticism. One may hope to do justice to the two opposed positions in this way: for, as the rationalists believe, reasons are provided by facts about what is valuable. 184
Some Aspects of Reasons and Rationality
185
Desires may, however, rationalize actions; that is, some desires, while not being reasons, weakly justify actions. We started with examination of specimens of practical thinking and attention to a variety of uses of the term ‘reasons’. I have included some observations about the language of reasons because of their suggestiveness with respect to philosophical problems, although these ideas were not developed. To some degree the present chapter tracks our explorations. It exhibits a methodology that was not uncommon at the time, one which Grice so clearly mastered. He called it ‘linguistic botanizing’. It was intended neither to catalogue nor to analyse ordinary usage. The point was, to use our mastery of language to help explain theoretical distinctions. Grice was also happy to invent his own technical vocabulary. Two concepts that he introduced proved most fertile: the ideas of a ‘thought transition’ and that of a class of motivations he called ‘propensions’.
2. Three types of reasons Grice and I did not formulate a systematic classification of reasons, but made use of distinctions he later presented more formally, citing detailed linguistic differences, in Aspects of Reason.1 In that work, he reaffirmed one of his central positions, that reason is univocal, that there is a common structure of theoretical and practical reasons. He also clearly distinguished the two important features of reasons for action, that reasons explain and that they justify. Grice proposed three classes of both practical and theoretical reasons: a purely explanatory, a purely justificatory, and a hybrid of the two. First, there is a purely explanatory use of reasons, primarily offered for events, e.g. reasons for the collapse of the Roman Empire, but sometimes for actions, e.g., ‘the reason for N’s sudden departure was that a remark was made about his suit, and all of his family are sensitive about their appearance’. But unless attention is specifically directed to this kind of case, philosophers who discuss reasons for action do not include as reasons for action (purely) explanatory reasons. While one can speak of the reasons for the collapse of the Roman Empire, (or why the Empire collapsed) we don’t evaluate those reasons as good or bad. We cannot say, ‘There was a bad reason for the collapse’ (or ‘There was a bad reason why the Empire collapsed’). We can say how favourable or good a case the reason makes. A teacher may say ‘Give me some good reasons’, but it is understood that good reasons here means ones that explain, not good reasons for the event. The second kind of reason is purely justificatory- it doesn’t explain, but it justifies. We may attribute to a given agent a justificatory reason to believe p, without implying that she believes p, or if she believes p, that she does so for that reason. Equally, we may attribute to an agent a justificatory reason for acting, for-x-ing or to-x, where we do not imply that the agent so acted,
186
Judith Baker
or would be moved to act on the basis of the consideration. Even if she did perform the act in question and was aware of the consideration, it need not be the case that she was moved to act for that reason. Even if an individual believed p, it might be for some other reason. Some examples: ‘The fact that the invited speakers to the conference were a day late was some (a) reason for thinking that the bridge had collapsed’. That may be a reason even if no one thought the bridge had collapsed, or someone did but didn’t connect the delayed speakers with it. In the realm of action, it may be true that ‘The fact that they were so late was a reason for postponing (or wanting to postpone) the conference’ (pp. 38–9). or ‘The fact that they were so late was a reason for a given individual to postpone the conference’. The consideration is at least a provisional justification for doing or wanting to perform the action; it goes some way towards making it the case that an appropriate person should act or should have so acted. But given that the individual need not have acted, or need not have acted for that reason, the reason cannot explain. In Aspects Grice distinguished between two uses of reason in English: ‘There was a reason for N to ...’ and ‘N had a reason to ...’ That there is a reason for Susana to believe something, or to do such and such, is paradigmatic of the purely justificatory type of reason, Grice’s second type. But to say that Susana has a reason is paradigmatic of the hybrid, Grice’s third type of reason. Grice’s third category is that of a Justificatory-Explanatory hybrid, or Personal Reasons; what philosophers call reasons for action are of this third type. They both explain (they are at least potentially explanatory), and justify. Justify in the sense that we see what there is to commend the action, from the agent’s point of view. They are relative to a given person. In Grice’s humorous examples, John’s reason for thinking Samantha to be a witch was that he had suddenly turned into a frog. John’s reason for wanting S to be thrown into the pond was that (he thought) she was a witch. John’s reason for denouncing S was that she kept turning him into a frog. When the reason is practical, a reason for action, it explains why someone acted as she did. Hence to say an agent’s reason for acting was such and such clearly implies that the agent was motivated, to say she has a reason is to imply, she is motivated, ceteris paribus to act for the reason given. Let me, however, add a note here. One of the standing controversies about reasons (a problem referred to as: Are there external reasons?) is whether a consideration can only be someone’s reason if motivation is presupposed. Grice’s classification is neutral in this respect, because the reason itself might be thought to motivate. The claim that a reason motivates, ‘all things being equal’–ceteris paribus – can be read to mean that the agent, insofar as she was rational, would be motivated. The word reason is used to explain and to justify (sometimes only one or the other) but current philosophical talk of reasons for action understands reasons to refer to a Justificatory-Explanatory hybrid. A reason ‘justifies’
Some Aspects of Reasons and Rationality
187
in the sense that we see what there is to commend the action, from the agent’s point of view. To cite an individual’s reasons for acting is, at least potentially, explanatory of his action. Potentially, because it is important to be able to use the idea of personal reasons, reasons which have both justificatory and explanatory functions, in some situations where an individual is not aware of the consideration in question (where reasons do not actually explain an individual’s action), because we want to be able to correct mistakes of deliberation. ‘There was a reason for a to x’ is understood to mean ‘Had a reflected properly he would have had reason r for x-ing, or, r would have been his reason for x-ing’. R may be attributed to a where he or she has failed to draw an obvious inference or is negligent in cases where knowledge is accessible, e.g. if a simply had neglected to read a letter he received and which, had he read it, would have informed him of the consideration in question. But the consideration must be accessible to be a reason; my great-grandfather did not have a reason to refrain from smoking.
3. Rationality without reasons Type 3 reasons for action are understood to be considerations that count in favour of an action in what would be the proper reflection or deliberation of an agent. Reasons justify because they are essentially tied to the agent’s evaluation of what is good/ required/ obligatory (and the like) to do. Reasons need not be conclusive, they may be outweighed; but it would be an error of the agent not to think they mattered. When they are not outweighed, reasons justify. In contrast, an agent who simply finds something appealing may not, and rationally need not, think that going for what appealed to her would on that account be justified. She need not be prepared to defend or commend her action,2 although she may defend it in a weaker sense by countering the claim that her action is bad or irrational. On this view, an agent simply moved by a desire, and a belief that a given act would satisfy that desire, does not thereby act for a reason. In some cases, however, her action may be rational and ‘weakly’ justified. I return to weak justification later in the chapter. Deliberation or proper reflection refers to argument. An agent who acts for a reason need not have rehearsed the argument before acting, but a reconstruction of the agent’s thought in terms of the argument must not distort the motivation attributed to the agent. This is essential to this chapter’s distinction between acting for a reason and actions that, while rational, are not done for a reason. Given their link to deliberation, it is not surprising that reasons cannot be attributed to an agent without taking into account their role in a specific context of guiding action. My arguments rest on attending to the context in order to determine whether deliberation and reasons may be attributed to an agent. What counts as a reason for action in one situation may not
188
Judith Baker
be a reason in another. While it seems clear that reasons are essentially connected to evaluations, it is important to recognize that what matters is whether the evaluation is used by the agent in deliberation. Whims present the strongest contrast. Warren Quinn (1999) has noted that, despite some philosophical examples, it is extremely hard to find whims that lack any evident appeal. What one does on whim ‘is usually something whose value (or apparent value) can either be discerned or made the object of intelligent speculation. In some cases, only the timing or means is capricious’. But that does not imply that acting on a whim is a case of acting on a reason. What one does may be of value, but whether one acts on a reason depends on whether an evaluation moves an agent, whether a consideration is present to the agent in the form of a principle for action, or as justification for her action. Even if I typically visit my hometown out of duty, or because I view it as a good thing to visit relatives, a visit might also just appeal to me as something to do one day. I do not then go because of my evaluation. Most philosophers have treated the incontinent individual as, typically, acting for a reason, although Grice and I argued (Baker and Grice 1985) that the incontinent individual typically does not act for a reason, in our article on Davidson’s account of incontinence. This chapter also challenges the accepted understanding of the incontinent, in attending to the situation in which familiar considerations motivate the agent. In many cases of incontinence, the thought of what is pleasant disrupts rational deliberation, and comes with no commendation, without authority. In other cases of incontinence, what starts off in reflection as a consideration may be either explicitly or implicitly silenced, ruled out-of-court; the agent cannot then think it justifies his action, unless he evades full knowledge or deceives himself. What is true of reasons is true of practical rationality more generally. One cannot delineate kinds of rational action without understanding the role of argument, what can count as argument (and its steps), or can be reconstructed in the form of argument, and, when argument is appropriate, for guiding action. What is evident when one begins to look for examples of argument leading to action, is the relative paucity of explicit practical arguments in human life. This presents a very general philosophical problem in understanding reasons. I see this lack as reflecting the existence of a larger domain for reason’s regulation of action than that of producing reasons or applying principles. The next section argues that much of practical reasoning takes the form of what Grice called ‘thought transitions’. These are neither in the form of standard practical arguments, nor can they be so reconstructed without distorting the reflection of the agent. I find thought transitions, reasonings that do not rely on reasons, in theoretical thinking as well. The next sections consider what makes thought transitions rational.
Some Aspects of Reasons and Rationality
189
4. Thought transitions Thought transitions in practical affairs do involve thinking before acting, but reasons are neither invoked nor relied on. Here are some examples: I don’t like the look of that, so I’m going back. I feel like doing something for the drought victims, so I’ll send a check to Oxfam. It would be fun to go to dinner tonight, so let’s go to O Mei’s; that’s a good place. While these are not explicit arguments, it might be thought that they could be put in argument form. Perhaps that they must be reconstructed in a standard argument form if the action is to be rational. Thus the third argument might read, starting with any one of the four premises (1a)–(1d): (1a) (1b) (1c) (1d) (2) (3)
Given my present state, I should do what is fun. Given my present state, the best thing for me to do would be to do what is fun. For me in my present state it would make for my well-being, to have fun. Having fun is good (or, a good). Going to dinner would be fun. Going to O Mei’s would be/make for dinner fun.
So, I’ll go to O Mei’s. None of the candidates for the first premise is plausible. It is evident that the agent who feels like going to dinner could easily reject (1a)–(1c), thinking them false or having no reason to assert them. She just finds going out to dinner appealing, in a specific way. The only plausible candidate is (1d). This specific proposition, and the more general one, ‘Pleasure is good’, may be thought by the agent to be true, but (1d) is not an acceptable reconstruction of the agent’s thinking ‘It would be fun to go out to dinner tonight’. Going out to dinner looks like fun. It has a specific appeal. An appeal different from not going where it looks dangerous, or feeling like doing something for a friend, or for a worthy group – and so on. But the agent does not represent her subjective state as a good to pursue, or an evil to avoid. The contrast I wish to highlight is between acting on a project one finds appealing/unappealing, on the one hand, and acting because of, on the basis of, one’s evaluation. The evaluation may target a particular subjective state as something to pursue or to avoid rather than assess actions. My focus in this chapter is those actions where we do what appeals to us, including those examples where the appeal is of what is pleasant or enjoyable. This is
190
Judith Baker
not to deny that on some occasions the desire for pleasure may be a reason or be offered as a reason. I will look at such a case below. Even if a critic insisted that (1d) is the way to represent our finding something to be appealing, and that it should be regarded as a respectable evaluation, the assembled propositions will not do the work of a standard argument. The premises do not support or yield the conclusion as in a standard argument. (1d)–(3) above can be said to yield the conclusion, or directive, for the particular agent whose thought process it is, only on the basis of a subjective condition: that the agent is in a certain subjective state, namely, feels like going out for dinner-fun. Rational beings (the agent at some other time, or other individuals) who don’t have that feeling, will not accept the conclusion. They may well accept as true ‘It is fun to dine out’, but will not accept it as a directive unless they feel like it now. Someone wondering what to do for the evening might think that if she were to go to dinner she would find it fun or pleasant, but right now she does not feel like going out. That is in general the end of the matter. The alleged argument lacks normativity: it is not authoritative or directive unless there is a supportive argument that she needs/ought to do something diverting/pleasant tonight. Practical arguments are different. Even if an agent did not feel like going to the dentist, an agent would think ‘I ought to have a dental check up yearly, now is the time, so I should see my dentist’ to be a directive with some force. It articulates a practical argument. Perhaps the strongest attempt to reconstruct (acceptable or rational) thought transitions as standard arguments is to treat the subjective condition, ‘I feel like having dinner-fun tonight’, as a premise, for then the premises would support the conclusion. But the individual, whose thought transition we are examining, does not regard a description of her inner state as a consideration that supports the conclusion. Casting it in this role is a distortion or a falsification of the thinking that guides her action. It is the pleasant prospect of going out to dinner that moves her to think of O Mei’s and to go there. Reflection on her state may come later, if challenges prompt her to reflect and deliberate. I come upon my houseguest ironing the sheets she has washed and vigorously protest that there is no call on her to do what I would not do myself. She replies that ironing is a pleasure for her, that it puts her into a kind of relaxed reverie. We cannot yet tell how best to reconstruct the practical thinking that led her to act. If she adds ‘I just felt like ironing’, or ‘It appealed to me to iron the sheets’, she has not acted on a reason. But suppose instead she adds that she anticipated my friendly protest, but thought I would not seriously object, and she wanted the dreamlike state, a time out of thinking, that ironing provides for her. This exchange makes us reconstruct her motivation in the form of a standard argument, with the prospective pleasure as a premise, and her reason for action. The houseguest need not have thought specifically of my response. The fact that she has reflected on this kind of situation in the past may be enough to attribute a reason to her.
Some Aspects of Reasons and Rationality
191
It will be useful to look more closely at a pair of closely related examples to note when it is appropriate to reconstruct thinking in the form of argument. Cynthia, hiking with a friend in the mountains, comes to a difficult spot and says, ‘I don’t like the look of that, I am frightened. I am going back’. That is usually enough for Cynthia to return, and for the friend to turn back with her. Cynthia’s action of turning back, admittedly motivated by fear, is, while not acting on reasons, nonetheless rational unless we judge her fear to be irrational. Cynthia’s subjective condition can serve as a premise, but only in a very different situation. Cynthia resorts to reasons. Suppose that, while her friend does not think Cynthia’s fear irrational, she still attempts to dissuade her from going back. After listening and reflecting, Cynthia may say ‘I am so frightened it is not worth it. I am not enjoying this walk anymore’. Or ‘I am too frightened to be able to safely go on’. Or ‘I often walk in these mountains and don’t usually get frightened. The fact that I am now is a good indication that this is a dangerous trail and I should turn back’. These are reasons, considerations implicitly backed by principles, and they could be the initial motivations of someone. But in Cynthia’s case they emerged when she was challenged. They do not express her initial practical reasoning. She was frightened by the trail ahead, wanted to go back, and didn’t have any reason not to.3 Faced with her friend’s objections, however, she needed justification for acting on her fear. She reflected and found reason(s) to act on her fear.
5. Thought transitions in theoretical reasoning Do thought transitions play a role in theoretical reasoning, in reasoning to statements of fact, rather than what to do?4 There are certainly appraisals in the world of action that support strategies rather than intentions, and which resemble thought transitions: ‘That looks dangerous, so it would be better to try it differently’, ‘That’s a significantly bigger than normal chair – it will take two men to move it’. More important, we see what look like thought transitions in the purely theoretical realm: ‘Those cells look abnormal. It is likely that the patient’s leukaemia is no longer in remission’. Not all rational beings who accept the ‘premise’ need accept the terminus, even supposing there are no counter-indicating additional premises in mind. Unlike the practical transitions discussed earlier, however, the appraisals above are licensed on the basis of experience, or expertise. The statements of appearances yield licensed ‘conclusions’ only if made by an (acknowledged) expert. This makes it look as if we can reconstruct the thought process, adding ‘The judgment that ... (repeating here the premise) is affirmed by an expert’ to yield a genuine argument. But the proposed reconstruction does not seem to be a correct representation of the individual’s thinking. One typically relies on one’s past success, rather than arguing on the basis of it. If queried,
192
Judith Baker
or contested, an individual would argue by appealing to past experience and success. At the moment of the alleged thought transition, however, the individual isn’t concerned to strongly justify, to answer possible objections to her observations. She is nonetheless engaged in rational formation of belief. Thought transitions do have a part to play in theoretical reasoning, but differences from their practical counterparts remain. It is expertise that licenses transitions in the theoretical realm. The conditions under which further justification or scrutiny is appropriate also differ. The move to genuine argument in the non-practical domain is accepted as a move to what is more reliable, even if it is not an accurate representation of the individual’s thinking process. The agent often, and appropriately, relies on her expertise without using it as a premise to support a conclusion. Individuals engaged in laboratory work and in general the work of discovery need to rely on their expertise before they turn to justification. Nonetheless, when questioned by themselves or others, ‘How do you know?’ or ‘Why do I believe ... ?’, individuals will turn to argument proper and to reasons. They will provide reasons based on their past success or on further determination of evidence. The turn to reasons proper in the theoretical domain may be a result of our contemporary social practices; it is nonetheless now judged always appropriate that rational enquirers should scrutinize their judgments. This may not mark a significant difference in the two realms, since the reasons not to subject one’s practical thought transitions to such scrutiny are strongly tied to certain contexts and activities. Contrast the thought transition that leads one to buy a gift for a friend and the argument one presents in a letter of reference. If some thought transitions leading to action are found to be rational, we may suppose that there are three classes of action: (1) actions strongly justified, which are supported by (good) reasons, (2) weakly justified actions, motivated by thought transitions that are judged rational,5 and (3) actions motivated by thought transitions that are without that commendation. ‘Weakly justified’ is a new term, meant to indicate a positive form of permission, although there is no one who permits. Instead of the dichotomy obligatory/forbidden (or good/bad) where permitted is equivalent to not-forbidden, there is introduced the idea of an action that is positively ‘OK’. To say that Cynthia’s action, motivated not by a reason but by fear, is rational,6 is to say it was OK for her to do it. Her action is weakly justified. The claim that some actions may be weakly justified must be supported, but I want to forestall a misunderstanding that might arise from the language of strong and weak justification. In cases of conflict, weakly justified actions are generally, but not always, inferior to strongly justified actions, actions supported by reasons. To meet the challenge of her friend, who proposed an alternative backed by reasons, Cynthia defended her action. She called on principles, finding reasons that supported her initial motivation to turn
Some Aspects of Reasons and Rationality
193
back. In another situation, a different principle might have convinced her that she ought to overcome her fear. But the mere fact that an acceptable principle opposes an agent’s motivation does not ensure the rational superiority of the principle. The agent who wants to go to O Mei’s may be challenged by herself or another person, ‘You always want to go to O Mei’s. You are in a rut’. It is not evident that the implied principle ‘It is bad to be in a rut’ ought, rationally to win out. It is clear that in practice it sometimes does, and sometimes does not.
6. Propensions What makes us say that some, but not all, thought transitions in the practical realm are rational? For the beginning of a short answer, note that some kinds of motivation are good to have. Those are the ones involved in thought transitions that we call rational. Grice introduced the idea of ‘propensions’, inclinations or motives he thought could be established as ones that are good to have. The idea of evaluating motives in this way was central to, and seemed to make sense of, a notion of weak justification of actions. A propension is to be thought of as a more, or relatively, primitive capacity that has survived in humans, and that continues to guide action. Grice was well aware that this was a sort of evolutionary fantasy. He thought it a good tool to think about the fundamentals of rational action. Impolite critics of the more enthusiastic wings of current evolutionary psychology accuse them of telling Just-So Stories. Grice was deliberately telling Just-So Stories. He had not the slightest intention of advancing an evolutionary ethics founded on known biological facts. There is, for rational beings, another capacity, on top of this, of acting on principles. Propensions should not, however, be thought of as necessarily non-rational. While behaviour prompted by what is pleasant may be an activity of non-rational, as well as rational beings, feeling like helping someone who helped you may be an activity peculiar to rational beings, possibly including non-humans. Thought transitions ‘certified’ as OK or rational may be considered the articulation of those inclinations or motives Grice called propensions. Propensions and their articulation in thought are certified together. Here are some examples, and possible groupings, of propensions. Doing what one believes to be enjoyable or pleasant, relaxing, avoiding what is unpleasant or merely uncomfortable, acting out of curiosity, pursuing something intriguing, going for what has some kind of appeal – these are all familiar human dispositions and, I will argue, have a common justification. This group would also include acting on all or most whims, or caprices, but exclude mere urges or the kind of bare ‘plumping-for’ illustrated by simply opting to bet on red rather than black. A second common group of propensions include: acting to avoid what ‘looks dangerous’, avoiding or taking
194
Judith Baker
cautionary steps regarding a person who ‘looks suspicious’, and in general, feelings that have cognitive content regarding our safety. Another important group includes the dispositions that express emotional feelings such as acting out of anger, gratitude, sympathy or fellow feeling, and the like. The approvability of a thought transition depends on the propension that it expresses. The kind of evaluation or endorsement that is connected to propensions may be made clearer by saying what it is not. Approval does not rest on the value – even what might be called the ‘modest’ good – of what is attained. In the case of the first group above, endorsement does not take as its target the future enjoyment, or whatever makes for the direct appeal of the action. Rather, the positive evaluation of propensions takes as its object the propension or inclination itself. Acting in line with some feelings is endorsed, or thought good. Although it is my particular individual promptings that count, this is not a matter of illegitimately counting my preferences twice.7 For while reasons often rest on appraisal of the end or object of the action, the aesthetic or prudential or moral value of what one seeks, this is not true in the case of propensions. To license the thought transition for what is pleasant or fun is to think, it is OK to be moved by what is pleasant. It may be thought that approval presupposes that pleasure is, in some sense, a good. I do not wish to deny, or to affirm, that pleasure is a good. The evaluation relevant to propensions is not of the form ‘To have x is a good thing’ but ‘The inclination to have x is a good inclination, or one we ought to have’. ‘I’m uncomfortable in this chair, so I’ll move’ is an OK transition, but not because one thinks that the world would then be a better place or that a man is maximizing his happiness by moving, or that being a bit uncomfortable is bad. Rather, one thinks that an inclination to remove one’s discomfort is a good thing with which people should be equipped. While I have yet to characterize the considerations that back up the endorsement of an inclination, the position outlined so far says something more than that as individuals we go for, or pursue, what we like, or what has a certain affective feel. For, as rational, we subject our motivations and our actions to evaluation. The question raised here is what is to be said for doing what we feel like doing or what is appealing, and so on. I have already argued, with respect to thought transitions, that representation of the felt appeal as a premise in a standard argument is a distortion of the thought that issues in action. To recall: the agent doesn’t regard doing what I feel like as a reason or justification for going to O Mei’s in the example given. It is the attractiveness of O Mei’s, just as in other examples it is the scariness of the road, the generosity of N, which gives rise to the concluding action. Whether or not desires on their own give rise to reasons, the observations made by Ruth Chang, which draw our attention to the affective feel of desires, are important. The felt character of those desires called propensions matters.8 Insofar as propensions are substitutes for principles and worth preserving, they must guide and not merely cause human actions. That an
Some Aspects of Reasons and Rationality
195
action is represented as appealing or as avoiding what is unappealing may be essential to what it is to ‘guide’ action (at this primitive level) rather than simply cause it. Emotional feelings, including non-hedonic ones, if there are any, show the same feature. To guide requires that one can exercise some control or monitoring. An individual who does what she enjoys continues to do it, as long as she enjoys the activity and nothing else intervenes. Similarly, she may slow down, or stop doing what ceases to appeal. Moreover, appeal allows the agent to represent the action in imagination and retrospection in a form that has motivational force. The discussion of thought transitions has assumed the commonality of a variety of actions described as doing what is appealing, a class that includes going for what looks pleasant or enjoyable. It includes actions done on impulse and its sub-class of whims or caprice. Acting on a whim has an added feature: the idea has a specific sort of appeal, which might be called aesthetic, which involves the stance of spectator as well as agent. Acting on impulse is one specific kind of response to what is appealing. An idea occurs to an agent and appeals to her. Sometimes the appeal requires no explanation, as in a sudden impulse to go off to the mountains for a day’s walk in solitude. When the appeal is indirect, as a sudden impulse to touch someone’s third toe, explanation is needed for this to count as acting on impulse: perhaps that the action was seen as an expression of affection. Because it is action on impulse, the desire to touch the other person’s toe must have been felt straightaway as a desire to express affection, not arise as a result of reflection, or a search for the means to some end, e.g. seduction. Where there are actions that are means to the action x, which, on impulse, an agent resolves or plans to do, those actions count as acting on the impulse to x only if they are seen by the agent straightaway as somehow constitutive of x-ing and not as a result of a search for the best means. Acting on impulse is like going for what is pleasant without deliberation about means, except that when one acts on impulse the idea must ‘strike’ the agent. It may help to counter possible objections to classifying acting on impulse as rational, to note that an impulse is not a compulsion. Impulse speaks in the voice of ‘Why not do such and such?’ of what appeals, not in the voice of command: ‘Do such and such!’ or ‘You must do such and such’. There is no general rule or principle that states grounds for the appeal of what one does on impulse, as is the case in general for doing what appeals to one. There may occasionally be such a principle but it is not a standing practical principle for the agent; it does not regulate her activity. Another way to put this, is to say that acting on impulse excludes acting in accordance with a principle (even a very general ‘maximize pleasure’ principle). It seems one couldn’t have a sudden impulse to feather one’s nest, advance one’s career, improve conditions of child labour in the developing world, or serve the nation. This is not because acting on impulse is ‘emotional’ in some way. Individuals may act in the heat of a passion elicited by gifted speakers or
196
Judith Baker
skilful rhetoric, but when enthusiasm is whipped up and individuals manipulated to accept principles, the individuals do not act on impulse.9 The class of rational actions that is under discussion in this chapter is potentially very wide, but there are many kinds of voluntary actions that are neither weakly justified nor done for reasons. Some of them may seem close to acting on a whim, perhaps because they are thought of as diversions, such as running one’s hand across fence rails. But the agent does not imagine or envisage them as appealing, and so they are pointless in that way. Some actions are not easily classifiable: making stones skip on water is typically not done for some purpose, but has an inner purposiveness: one tries to make the stone skip as many times as possible. Whether or not skipping stones and similar actions are candidates for being weakly justified, depends on whether they are envisaged as appealing or are merely secondary activities. Our claim is that actions are weakly justified in so far as a rational individual would endorse the motivation, or the propension to such acts, if he subjected them to evaluation. Grice proposed an argument for this kind of reflective endorsement that built on his playful but very serious fiction. In Method in Philosophical Psychology (1975), Grice made use of a fiction, a genitor building more and more complex creatures, in order to introduce and then explain psychological concepts. He wished to apply this fiction once again in order to analyse justification and weak justification. Aspects of behaviour, or human traits, could be ‘justified’ if one could support two claims. First, that a (fictional) genitor interested in constructing self-maintaining individuals would build such dispositions or traits into the things created. Second, that self-reflective humans would endorse these dispositions if they subjected them to evaluation,10 taking an evaluative stance analogous to the genitor’s. Acting on a propension is rational because, were an agent to reflect at some time regarding this disposition, she would endorse it. Grice thought that propensions are justified in so far as it is good for creatures like us to be equipped with them. He also thought that rational reflection on our propensions would refer to what is good for creatures like us. Both ideas need to be supported. The genitorial story of creature construction is obviously a colourful myth about the development of species in terms of what makes them flourish. Applied to an individual’s justification of actions, it has both the advantages and disadvantages of naturalism. This is not, however, the place to offer a serious examination of naturalism.
7. An extended use of ‘reasons’ The language of reasons is richer, however, than the typology above indicates. In particular, while reasons for action take as their focus an agent who considers the pros and cons of an undertaking, there are linguistic variants that point
Some Aspects of Reasons and Rationality
197
to the role of an enlarged cast of characters. One may recognize individuals who influence a given agent and while the language refers ‘officially’ to the reasons for action of the agent, the reasons are those of the person whose role is to persuade or influence. In another variant, people who care for an agent seem to have special linguistic license in the attribution of reasons. Grice noted an ambiguity of reference in the attribution to reasons to perform a given action, phi. ‘There is a (good) reason to phi (or, not to phi)’ may refer to a reason for one, anyone. for an agent who is under discussion. for the person who wants to get the agent to phi. ‘There is a good reason not to smoke’ may mean ‘Everyone has a good reason not to smoke’, given the link between respiratory diseases and smoking. Even when one refers to an agent, as in ‘There is a (good) reason for a to phi’, the statement is ambiguous. Although one usually refers to the agent’s reasons, it need not always be so. Someone may be speaking to an individual, who is not the agent, but is thought to be in a position to provide for the agent or to get the agent to do phi. Where an individual says to Oliver’s mother, ‘There is a good reason for Oliver not to smoke’, he may be referring to Oliver’s mother, to her having a good reason (namely, Oliver’s health) to influence him not to smoke. In this case, the parent’s reason is usually the good of the child, and it can be assumed that if the child were able to properly reflect, the same consideration would be his. But that is not a necessary feature for this kind of case. The Dean may say to the Chair of the department ‘There is a reason for Martin to teach medical ethics’, where the head of the department needs to increase enrolment. The Dean refers to the Chair’s reason to influence Martin. Managers, trainers and bosses speaking of an agent a may refer to their own (good) reasons to influence the agent to perform the action. This does not, of course, determine what considerations they will offer the agent to perform the action in question. It may seem that this is not a philosophically interesting variant, that such examples only show an extended range of agents. We are still interested in the agent’s reasons, what from his or her point of view is a consideration favouring an action. This variant makes us see more than one kind of agent, that an individual’s action may be that of persuading, manipulating, or ‘getting’ another agent to do something. But the second agent’s reasons are reasons in the standard sense of what, to him or to her, are considerations for pushing another to act. Still, one could have directly referred to the reasons of the boss or parent for influencing another. What is the point of ambiguity of reference, of portraying the reasons as those of the actor, who will perform the ultimate
198
Judith Baker
action, phi? Grice and I did not pursue the question, but here are some candidate explanations: First, in the given contexts, the ultimate actor or agent of phi-ing is supposed to accept the reasons offered him by the ‘influencer’. He ought to accept the authority of the other person. A second possibility recognizes that the influencer may be a simple manipulator. The ‘influencer’s’ reasons are what justify the action according to the speaker; it is a matter of indifference what actually motivates the agent to phi. We are not, however, forced to accept the second explanation insofar as we recognize the work of manipulators. To speak of increasing enrolment as Martin’s ‘reason’ for teaching medical ethics when it is known that this is not the sort of consideration Martin recognizes, may be to use the term ironically. If one accepts the need to explain the ambiguity of reference, something of philosophical interest emerges. For what is common to both candidates is this: the authoritative voice regarding what would justify doing phi need not be the agent’s. Grice and I discussed a second variant, or extended, use of reasons: we admired the great heavyweight boxer Floyd Patterson. We asked whether, referring to Patterson late in his career, we could intelligibly say ‘There was a (good) reason for Floyd Patterson to quit the ring’, knowing that Patterson did not care about his reputation, while we, the speakers do. We did not, however, accept the linguistic variant, that we held to be a different claim, ‘He had a reason to quit the ring’.11 We thought Patterson was going downhill and should have got out while he was still a great boxer. The proposed ‘reason’ does not fit easily into the types explored in Aspects. This is not a type 3 use of reason, for we, the speakers, did not believe the consideration motivated Patterson. But it is not a typical justificatory reason either. Attributing, as we did, a good reason for Patterson to quit, may involve thinking that a good reputation is something to be valued. But while reputation is a value to be recognized by any boxer, it need not, perhaps, enter seriously into his deliberations about what to do. Not all values can be espoused. We need not think that all (equally good) boxers ought to especially care for their reputation. Not, for example, boxers less noble, in our opinion, than Floyd Patterson. But type 2 (justificatory) reasons are considerations for any rational agent in the relevant circumstances. (A truly noble agent can hardly think his virtue a relevant circumstance.) We thought that our statement about Patterson’s reason was OK because we, the speakers, cared about the fighter’s reputation, although the boxer himself didn’t. An additional difficulty in classifying this type of reason stems from the role of reasons in deliberation. Granted that the reason we attributed to Patterson is not universally justifying, while we thought he ought to have valued his reputation, we could not have thought he ought to have also considered his ‘nobility’ in giving weight to his reputation.12
Some Aspects of Reasons and Rationality
199
Fans do judge that there are reasons for the persons they admire, and to whom they are loyal, and do so without presupposing the agent’s motivation. The discourse of individuals loyal to a sports team or to a star is partly constitutive of the community of those individuals, analogous in some ways to the connections between intimates. Society seems to recognize that caring for someone matters, that it gives one permission to speak for the other’s good. Of course, it does not make it true that the agent, in this case, Patterson, has the reasons in question. A type 3 reason cannot be attributed to him, since Patterson did not much care about remembered glory. But neither is it a type 2 reason. (Recall: we did not think it a reason all fighters of Patterson’s ability ought to have.) The speakers were, however, committed to thinking ‘The agent ought to care about the given consideration, here his reputation’. While this kind of example may seem at first a mere oddity, I think its import is more general. Both variants suggest that there are competing sources for the justificatory authority of reasons: parents, bosses, friends may have special status. If this is accepted, it may help explain the transmission or giving of reasons to another. Insofar as an individual N, recognizes the authority of another person to identify what would justify N’s acting in a certain way, he will be prepared to accept a reason from him or her.13
Notes 1. See Grice’s classificatory scheme in Grice (2001). 2. The distinction I wish to draw is between, on the one hand, judging something to be good, or evaluation proper, and, on the other, finding something appealing or desirable. Evaluation proper requires that one be prepared to defend one’s evaluation. It may well be true that we think of and speak of what is appealing or pleasant as good. ‘Good’ has a number of uses and may be like ‘the thing to do’. That is, we express our favouring something, whether in the mode of wanting or of opting for, in language that is close to evaluation. But, strictly speaking, judging something to be good, requires that (as a rational being) one be prepared to defend one’s judgment, perhaps as conforming to some general position or standpoint. I think this position is close to that proposed by Gary Watson (1987). He wished to distinguish (p. 150) between finding something pleasant and evaluating it. But since he distinguishes between valuing and evaluating, he thinks an agent may (perhaps perversely) ‘value’ what she is not prepared to defend. 3. Note that there is no general rational requirement to always act on reasons, and no general truth that a rational individual would be better off the more often she acted on reasons. 4. I was prompted to consider a possible role for licensed thought transitions in nonpractical reasoning by Phil Clark and Jennifer Whiting. Conversations with Clark and Whiting have contributed to the development of this chapter in many ways. 5. There are similarities here to Jonathan Dancy’s category of enticing reasons (2004 91–118). Unlike preemptory reasons, Dancy claims that enticing reasons, such as prospective pleasure, do not yield ‘oughts’; it is not wrong, although it may be silly, not to act on an unopposed enticing reason. While what I call weakly justified
200 Judith Baker
6.
7. 8. 9.
10.
11.
12. 13.
actions do not entail an ought judgment, they are not the same as actions done for enticing reasons. In those contexts where prospective pleasure is a reason, it is an enticing reason. But being attracted to what is pleasant and pursuing it, is not always acting for a reason. Thought transitions characterize prospective action in a favourable light, but what makes one favour them is not an independent contribution or consideration. That ‘pleasantness’ here counts in favour of acting depends on the agent’s subjective condition, of feeling like doing the act in question. Not being moved, on a given occasion, to act on what one judges to be pleasant is not silly. It is not, on its own, subject to rational criticism if a person does not feel like it. A person can be criticized, however, who never, or rarely, feels like doing what she judges pleasant, and never, or rarely, acts on what she feels like doing. One condition was noted earlier; the fear must not be irrational. A second qualification is harder to formulate. If we suppose that an individual is faced with a burning house, we judge that she, or any rational individual, would have a reason to turn away. The fear must not be such as to be warranted for anyone, nor can it be any old fear. It must be a fear that is appropriate to her. See the dispute between Joseph Raz and Ruth Chang, in Chang (2004). Chang’s article, just cited, has been important in drawing attention to the felt character of desires. Careful examination of the expression ‘acting on impulse’ sharpens our own judgment; it confirms the insight that important distinctions are to be found in ordinary descriptions, that careful characterization can lead to a theoretical distinction. Grice noted that the ‘voice’ of impulse is one, which appeals rather than requires. This might seem at first glance merely impressionistic, or, at best, descriptive. But it is used to explain. During the Vietnam War, young American men at anti-war rallies burned their draft cards. Grice thought such actions could not in general be classified as acting on impulse. The individuals represented their actions to themselves as obligatory. They were influenced by argument, or, some sceptics might say, their activity was ‘engineered’ by rhetoric. Grice thought that if a propension was significant in human flourishing, it was one we ought to have. Here, as elsewhere, he was thinking in terms of creatureconstruction. He asked: what psychological traits or dispositions would one build into a creature who was meant to flourish? This viewpoint of a creator or ‘genitor’ (a word close enough to ‘janitor’ to appeal to Grice) is one that a reflective individual can adopt insofar as she aims to retain or modify those aspects of herself she can change. In introducing the notion of a propension, Grice theorized that a more, or relatively, primitive activity has survived in humans, which continues to guide action. There is, on top of this, a rational capacity of acting on principles. While Grice here and later accepted the distinction between there being a reason for someone and someone’s having a reason as marking two uses, a problem was noted. For there may be no real difference in general between, ‘There was a reason for a to x’, and ‘a had a reason to x’; ‘have’ may function here merely as a syntactical auxiliary. Just as the passive transformation is the means of switching terms from, ‘Bill stole the bicycle of John’ to ‘The bicycle of John was stolen by Bill’. I am grateful to Phil Clark for this observation. On our view, reasons are justifying. Hence, we would qualify the extremely interesting account of proleptic reasons offered by Bernard Williams. Williams
Some Aspects of Reasons and Rationality
201
thought an individual whose approbation was desired by another could make it the case that the other person had a (motivating) reason by attributing a reason R to him. While it is evident that individuals are moved by considerations attributed to them in such a context, to say that they thereby have or act on reasons is to regard those considerations as justifying.
References Baker, J. and H. P. Grice (1985) Davidson on weakness of will. In B. Vermazen and M. Hintikka (eds.) Essays on Davidson (Oxford: Clarendon Press), p. 249. Chang, R. (2004) Can desires provide reasons for action? In R. J. Wallace, P. Pettit, S. Scheffler and M. Smith (eds.) Reason and Value: Themes from the Moral Philosophy of Joseph Raz (Oxford: Clarendon Press), pp. 62ff. Dancy, J. (2004) Enticing reasons. In R. J. Wallace et al. (eds.) Reason and Value (Oxford: Clarendon Press), pp. 91–118. Grice, H.P. (2001) Aspects of Reason, ed. Richard Warner (Oxford: Oxford University Press). Quinn, W. (1993) Putting rationality in its place. As reprinted in W. Quinn (1994) Morality and Action (Cambridge: Cambridge University Press), p. 249. Raz, J. (1999) Agency, reason, and the good. In J. Raz Engaging Reason (Oxford: Oxford University Press). Scanlon, T.M. (1988) What We Owe to Each Other (Cambridge, MA: Harvard University Press), Ch.1 and Appendix. Watson, G. (1987) Free action and free will. Mind, 96: 145–72.
9 Showing and Meaning: On How We Make Our Ideas Clear Mitchell Green
1. Introduction In discussions of speaker-meaning, natural meaning is often mentioned only to be set aside. That boulders mean glacial activity, and that the infant’s mottled skin means its core body temperature is low, seem far removed from communication, and intentions to communicate, of the sort that Grice and those who work in his tradition seek to elucidate. At the same time, recent work on speaker-meaning has shown that an intention to communicate is not a necessary component of this phenomenon. Furthermore, once we formulate a more accurate elucidation of speaker-meaning in light of the fact that it does not require communicative intentions, we see that speaker- and natural meaning are more complexly intertwined than current consensus acknowledges. In particular, certain phenomena found squarely within the realm of natural meaning help shed light on some features of communication as it is found within our own species. In the course of this investigation we will also discern a connection between Gricean themes and some recent developments in the evolutionary biology of communication. As a point of methodology, I should note that I am suspicious of the idea that we can get a complete account of speaker-meaning by conceptual analysis, and this for two reasons. One is that I have no way of being sure that our intuitions about this phenomenon are coherent. If they aren’t, then there will be no single, true elucidation of speaker-meaning encompassing all those intuitions. Another reason is that our intuitions about the intension of that notion as it applies to hard cases probably don’t distinguish between unusual applications of that notion and applications of that notion that are definitely incorrect. Intuitions are just too crude a tool to make this distinction reliably. For this reason, I’ll let conceptual analysis as based on our intuitions about cases be the first word, but will not assume that it is the last word in elucidating speaker-meaning. This implies that if we find a
202
Showing and Meaning: On How We Make Our Ideas Clear 203
strong, independent motivation for a view of speaker-meaning that nevertheless has an unintuitive, or not obviously intuitive, consequence, we can still be justified in adhering to that analysis.
2. What speaker meaning isn’t I put on an overcoat and grab an umbrella as I prepare to go outside. You infer from this that I believe that it is likely to be raining. However, there is no plausibility in the idea that in so doing I mean that precipitation is likely. That is, doing something that tends to produce a belief in you is not a sufficient condition for speaker-meaning. What about performing an action with an intention of producing a belief in you? Here too we do not have a sufficient condition: I might leave bits of your hair and fibres from your clothing at the scene of a gruesome murder. Perhaps I could even contrive to place some of your blood under the victim’s fingernails. However, here too I do not mean that you are the culprit. The most that can be said is that I’ve done something with an intention of getting the authorities to believe that you are the culprit. Perhaps what’s wrong with the last case is that the intention to influence the authorities is not itself open to view. That idea is reinforced by the fact that if I had marched into the crime scene and dramatically dropped your monogrammed handkerchief at the detectives’ feet, they would probably have concluded that I was trying to tell them something. Of course, they may not have been convinced of what I was trying to tell them; but it does seem plausible that we here have a case of speaker-meaning, namely that you’re the culprit in the heinous crime. Can we generalize from this case? Perhaps doing something with an intention of producing a belief, while also intending that this very intention be recognized, is sufficient for speaker-meaning? Grice thinks not, and gives the following example in support of that denial. Herod presents Salome with St. John’s severed head on a charger. Herod meets the conditions just mooted, but Grice remarks that we have here a case of showing Salome that St. John is dead, rather than telling Salome that he is (1989: 217–18). Grice’s observation about this case is correct, but I doubt that the conclusion he draws from that observation is. For he infers that the Herod case is not one of speaker-meaning. Yet that only follows if the only way of speakermeaning something is by means of a telling or an assertion (at least for the indicative case). Is that further premise true? There are good reasons to doubt it. You ask me to join you for a bicycle ride on this beautiful Fall afternoon and I show you my recently amputated legs, making clear my intention to show you them by gazing intently at you. Have I not also meant that I’m unable to join you for a ride (at least without a lot of prosthetic help)? Or in answer to your doubt about the proposition
204
Mitchell Green
that the three sides of right triangle add up in the form a2+b2=c2, I reply by giving you a proof. Here I’ve shown you that a2+b2 =c2, but surely I also speaker-mean that it is as well.1 As is well known, after denying that the Herod/Salome case is one of speaker-meaning, Grice suggested that it is necessary and sufficient for speaker-meaning that an agent do something with an intention of producing a belief in an audience, while intending further that recognition of this intention be part of the audience’s reason for accepting this belief (ibid: 219). Thus for instance had Herod drawn for Salome a picture of St. John’s severed head (intending that she believe that St. John is dead, while intending further that recognition of this intention be part of her reason for accepting this belief), Grice would consider this a case of Herod speaker-meaning that St. John is no longer. After the 1957 publication of Grice’s famous analysis of speaker-meaning, an industry of counterexamples and new definitions sprang up, spurring new activity for nearly two decades.2 But something went wrong somewhere. If we return to the examples that motivated Grice’s account, we can see that even the intention to produce a belief in an audience is not a mandatory part of speaker-meaning. After all, I talk to my car on a cold winter morning to urge it to start, and I talk to my infant daughter as I change her diaper. In both cases I do mean what I say – be it, ‘Be a good little car and start’, or, ‘Someday you might be changing me’. In neither case do I say what I do in an effort to produce a belief in anyone, even myself. (I already believe my daughter may be caring for me in my dotage, for instance.)3 Intentions to produce beliefs don’t seem to be required for speaker-meaning. On the other hand, when I unceremoniously go to the closet for my umbrella and overcoat, I don’t speaker-mean that rain is imminent. Might our intuitions about speaker-meaning be incoherent after all?
3. (Some of) what speaker-meaning is I suggest that speaker-meaning does require something reflexive or reflexive-like, but not in a way that involves intentions to produce effects on an audience. Compare the above-mentioned case of my taking an umbrella and overcoat from the closet, with another case in which after a heated argument about the weather, I march to the closet, all the while eyeing you beadily, and dramatically retrieve both coat and umbrella. Here I might very well speaker-mean that it’s raining outside. Yet while it also seems natural to suppose that I’m intending to make you aware of my belief, this too is not a necessary feature: I might just be indifferent as to what you think about my stand on the weather. What does matter is that I make my position overt: Not only open to view, but open to view in a way that makes my intention to make it open to view, itself open to view.
Showing and Meaning: On How We Make Our Ideas Clear 205
Other cases reinforce the point. You saunter in late to a crucial meeting of which I am in charge, where it is well known that I’ve been depending on your punctuality. I scowl at you, but do so in a way that makes my intention to show my anger also palpable. Here I make my anger overt, and here too it seems plausible that I mean that I am angry. Further, while in this case it’s likely that I intend to make you aware of my anger, that aspect is not mandatory. I gaze at the portrait of my rich, just-deceased uncle, who left me nothing from his estate but this very portrait. Presently I deliver a loud ‘Bastard!’ in the direction of the portrait. In this case it seems clear that not only have I made manifest my contempt for my uncle, I’ve also made manifest my intention to make that very attitude manifest. At the same time, not believing in an afterlife, I don’t intend to influence his or anyone else’s state of mind. But surely I speaker-mean here that my uncle is a bastard. We may crystallize these observations with the suggestion that it is sufficient for speaker-meaning that one show not only one’s state of thought or feeling, but also one’s intention to show that state. That might be a good first approximation of speaker-meaning.4 Yet speaker-meaning is usually construed as being about contents only, be they propositional or whatever are the content of other grammatical forms, such as sentences in the imperative or interrogative mood. That is, one speaker-means that P, or that a is to F, etc. But speaker-meaning is often more than a matter of expressing or otherwise putting forth a proposition. I might express a propositional content to you, but that might prompt you to ask, ‘How do you mean that – as an assertion, a guess, or what?’ Similarly in reply to my boss’s announcement, ‘You’ll be more punctual in the future’, I might wonder whether she meant it as a warning, a prediction, or a command.5 The fact that these questions are about how the speaker meant what she did, strongly suggests that they seek a further elucidation of what the speaker speaker-meant. If that is correct, then illocutionary force is a dimension of speaker-meaning in spite, at least typically,6 of being no part of the content expressed. This also enables us to mark a division between illocutions and perlocutions, the latter being the characteristic effects of speech acts. I might mean, by an assertion, to convince you of what I assert. However, we can only read ‘mean’ here as having to do with what I intend (as in, ‘He means to run to the top of that steep hill’), rather than corresponding to the use of ‘mean’ carved out by the notion of speaker-meaning. Force is a dimension of speaker-meaning. Yet how might we conceptualize force in terms other than that of content? I suggest that differences in illocutionary force make for a difference not in what is meant, but rather in how that content is meant. When I assert that P, I not only put forth that proposition, but also put it forth as true – unlike what is the case when I suppose it for the sake of argument or entertain it to get clear on its consequences. Yet putting forth as true does not distinguish what we do with contents we assert from contents that we conjecture or, for that matter,
206 Mitchell Green
guess.7 Instead, we need a more refined account of the various speech acts comprising the extended family of which assertion is a member – what Austin called ‘expositives’ (1962: 162–3), and what Vanderveken called ‘assertives’ (1990: 180). One who asserts that P opens herself to the challenge: ‘How do you know that?’, whereas it would be inappropriate to reply in this way to someone who has made a guess or conjecture having the same content. This is reflected in what I intend to make overt in my speech act: When I assert P, I intend to make overt my assertoric commitment to P; that is, a commitment to P having the following shape: a. I am right or wrong on the issue of P depending on whether P holds. b. I am committed to responding to challenges to my proffered conversational contribution of the form, ‘How do you know that?’ c. If I am unable to reply to the challenge in (b) appropriately, I am obliged to retract my assertion. Contrast this with my conjecturing that P. While I still stand to be correct or not on that issue, it would be inappropriate to challenge my contribution by insisting that I explain how I know. Rather, an appropriate challenge would require me to give some reason for conjecturing what I do. That is, when I conjecture that P, I intend to make overt my ‘conjectural’ commitment to P, to wit: a. I am right or wrong on the issue of P depending on whether P holds. b. I am committed to responding to challenges to my proffered conversational contribution of the form, ‘What is your reason for that?’ c. If I am unable to reply to the challenge in (b) appropriately, I am obliged to retract my conjecture. One natural way of making that commitment clear is by saying, and speakermeaning, ‘I conjecture that P’. Another way is to use a parenthetical addendum, as in, ‘P, I conjecture’. Yet another way is simply to let the context of utterance make clear that I am putting forth P as a conjecture rather than as an assertion or a guess. This might, for instance, be because our conversation is one in which interlocutors are invited to throw out conjectures rather than assertions. We’re all puzzled as to why the engine has stopped working, or what species of insect destroyed these trees. None of this is to say that when a speaker makes an assertion or conjecture she is consciously aware of the norms governing her act, any more than I am consciously aware of the principles governing such syntactic phenomena as preposing or fronting. Rather, assertion, conjecture and other members of their family are tools with which I gradually attain a competence which, once in place, is second nature. Once second nature, I may have difficulty
Showing and Meaning: On How We Make Our Ideas Clear 207
formulating explicitly how these tools work. I may nevertheless manifest my competence in their use by participating as an effective interlocutor.
4. An ethological perspective on communication It is one thing to claim that in speech acts we make overt our commitment to various contents. It is quite another to explain how this is possible. For instance, simply describing someone as committed in a certain way to a content is not enough to show that commitment. Description does not, per se, demonstrate what it describes. Indeed, even describing myself as being in a certain condition does not demonstrate what I have thus described. This last claim may not seem obvious. After all, in an utterance such as ‘This is just a guess, but perhaps your keys are still in your car?’, the first sentence indicates that the second is being put forth as a guess rather than as something that I claim to know. This is commonsense, and I have no wish to deny it. Rather, my point is that this ‘indication’ only succeeds given a background of norms and practices supporting them that go beyond the literal meaning of the words, ‘This is just a guess’. In the absence of such a background, utterance of ‘This is just a guess’, would have no more efficacy than if one had spoken those words while testing a microphone.8 Accordingly, we need some way of explaining how in a speech act I can show my psychological state, and in particular show the epistemic status of the content of that state – as justified, for instance, and to what extent. How might this be possible? Ethologists concerned with communication have shed light on this question. For those working in the tradition of Axelrod and Maynard Smith, a core concern is how signalling systems retain their stability over time. The problem that motivates this question is as follows. Biologists agree that one creature can send a signal to another. For instance, the bright colouration on an anuran can signal to potential predators that it is noxious. However, random mutations being what they are, it is always possible for a population of signallers/receivers to be invaded by a mutant who is ‘dishonest’: sporting bright colours, for instance, without having expended any resources in building poison. Such invaders have an advantage over ‘honest’ signallers, and are likely to exploit that advantage in differential reproductive success. However, if that happens, then as the ‘cheaters’ propagate, over time the recipients of signals, such as predators, will stop paying attention to skin colouration in choosing what to eat. As a result, skin colouration will lose all signalling value. The phenomenon of ‘crying wolf’, with its attendant loss of credibility in the cryer, has application well beyond our own species. Recent theoretical biology, and some parts of ethology as well, have attended to the ways in which signalling systems achieve diachronic stability. One way in which they do this is if the signals are hard to fake. When does this occur? One condition in which this occurs is if the signal is hard to
208 Mitchell Green
fake because of physical limitations on the signalling organism. An example is the funnel-web spider, Agelenopsis aperta, among whom there is competition for ownership of webs. Two spiders will vibrate on a disputed web. Reichert (1978, 1984) found that if two contesting spiders differ in weight by 10% or more, the lighter spider retreats 90% of the time rather than fighting. Furthermore, a losing spider can be made into a winner by placing a weight on its back. This strongly suggests that vibrating on a web is a spider’s signal of its size. Further, in the absence of scientists placing weights on their backs, funnel-web spiders can’t fake these signals. Signals that can only be faked with great difficulty as a result of physical limitations on the organism are indices. Another way of ensuring the difficulty of faking a signal is by making it more costly than is required just to produce a signal of that type. Male peacocks have flamboyant trains, making them less agile and easier for predators to spot; growing such feathers also costs extra calories (Zahavi and Zahavi 1997). Such trains nevertheless give males an advantage in sexual selection (Petrie, Halliday and Sanders 1991). An ostentatious display is like saying, ‘Just think of how fit I must be if I can survive with this baggage!’9 Or consider the male stalk-eyed fly, Cyrtodiopsis dalmanni. In choosing mates, females prefer males with long eye-stalks (Wilkinson and Reillo 1994; David et al. 1998). This is in spite of the fact that these appendages make the male slower and easier to spot for predators, and are costly to produce. Observing that these appendages provide evidence of the male’s viability – since he’d have to be unusually fit to survive with them – Maynard Smith and Harper (2004: 33–4) hypothesize that the eye-stalks are also signals of the male’s viability. Signals, such as these, that can only be faked with great difficulty as a result of being costly to produce are handicaps. A signal’s being costly to produce is thus one way of being difficult to fake because of physical limitations on an organism. A handicap is, accordingly, a special type of index. Indices, be they handicaps or not, secure diachronic stability for signalling systems. They do this by being hard to fake, for in so being they show what they signal. Showing, however, is not demonstrative proof. One can show concern for someone’s injury without turning aside all possible scepticism about one’s sincerity. Rather, ‘showing’ is here used in its ordinary way, to mean (non-conclusive) demonstration.10 The stalk-eyed fly’s eye stalks show his viability, and the extent of the funnel-web spider’s vibration shows his Resource Holding Potential – roughly, his ability to defend his territory against challengers.11 In showing what they do, these signals endow populations of signallers/receivers with ‘resistance’ against invasion by dishonest mutants. Do we find signals that are index-like in our own species? Certain facial expressions and related forms of facial behaviour are difficult to fake. For instance, few people can weep at will, and because of this, tears are normally thought of as showing one’s strong emotion – grief or elation as the
Showing and Meaning: On How We Make Our Ideas Clear 209
case may be. Again, while we can deliberately activate the muscle that pulls up the corners of our mouths (the zygomatic major) to make a smile, few people can activate the muscle orbiting the eye (the orbicularis oculi). As a result, what researchers have come to call the Duchenne smile, which involves activation of both the zygomatic major and the orbicularis oculi, is very difficult to fake (Ekman 1998). That in turn is good reason to think that the Duchenne smile shows one’s happiness.12 Similar things may well be true of facial expressions associated with other of the ‘basic’ emotions: disgust, fear, sadness, surprise, rage. In addition to expressive behaviour that is an index of our affect while also being due to our biological makeup, there may also be ‘indexical’ behaviour that is idiosyncratic, either due to individual characteristics or to a cultural milieu that we have internalized to the point of its becoming second nature. Perhaps, for instance, a culture uses a certain intonation contour as a conventional expression of surprise. In spite of its being conventional, that contour might be hard to suppress and hard to produce at will for members of the culture in whom it is ingrained. If so, then that intonation contour will be an index of surprise in that culture, though not elsewhere.
5. Speech-act norms as handicaps The notion of an index may help us to understand how we show, and thereby express, our emotions. However, as we shall see in a moment, emotional expression is in many respects less fine-grained than are expressions of cognitive states such as belief and knowledge. Can the notion of an index, or the subordinate concept of a handicap, shed any light on how we show the epistemic status of those contents that we proffer for conversational and other purposes? Calories and territory are not the only sorts of goods that social mammals value and fight over. Another kind of cost social mammals can incur is a loss of status. Primates in particular expend extraordinary effort maintaining social hierarchies. Baboons cultivate relationships that determine mating and grooming protocols. Who gets to groom whom, and when, are the foci of considerable attention and sometime the subjects of lethal disputes. For instance, one baboon can spend years acquiring the ‘rights’ to groom a high-ranking conspecific, and a false step in that process might prevent his passing on his genes (Barrett et al. 1999; Dunbar 1996; Cheney and Seyfarth 2007). Analogously, I take it as uncontroversial that within our own species, one’s status within the family, on the playground, in one’s professional community, and elsewhere are also of paramount importance. Although grooming and the hierarchies it engenders may be less crucial in our species than in other primates, other types of status loom large. These take many forms, but I shall focus on just one, credibility.13
210
Mitchell Green
Finding someone credible is a matter of tending to believe what they say to be reliable; it is also a matter of tending to believe them to be sincere if their utterance admits of sincerity.14 A credible assertor can be trusted to be right and sincere; a credible promissor can be depended upon both to intend to keep her promise and keep it; and so on. The more credible a person, the more likely we are to take their utterances at face value. Appreciating this permits a natural refinement of the well-known ‘scorekeeping’ model of conversational dynamics pioneered by Stalnaker (1972, 1973, 1974, 1984), Lewis (1979, 1980) and Thomason (1990). On that model, interlocutors perform speech acts whose content is accepted into ‘common ground’ so long as no one demurs from the proffered contribution. Conversational ‘score’ can be tracked by the contents of common ground (along, optionally, with other factors such as relations of salience, standards of precision, domain of discourse, and so on), which in turn determines what interlocutors may presuppose in subsequent speech acts; it can also determine the very content of those speech acts via the process of ‘impliciture’ (Bach 1994). A natural refinement of this picture would keep tabs on which interlocutors are credible and to what extent, and that will in turn determine the weight – as one might call it – of their conversational contributions. An expert’s pronouncements on her field of interest have more weight than do those of a dabbler, likewise for a The Times as compared with a supermarket-checkout magazine. It is, likewise, precisely the gravamen of ad hominem arguments to undermine the weight of a person’s remarks. One’s credibility, then, determines the weight of one’s pronouncements, and for this reason many of us jealously guard that credibility. A speaker whose remarks are, in our sense, weighty, enjoys a presumption in her favour. A proffered assertoric contribution to common ground puts the onus on one who would challenge her to give good reasons for that challenge. (‘I know the expert consensus is that p, but I submit that a superior alternative explanation is the following ...’) By contrast, one whose remarks are in the appropriate sense, light, creates no such burden, and may even be disallowed all contributions save those that are either self-evident or that can be substantiated by immediately available evidence. Our credibility comes under threat every time we perform a speech act whose content could turn out to be incorrect, or that could be insincere, or could be based on inadequate evidence. Friendships, marriages, careers, company stock values, even governments can be brought down through a loss of credibility, and I will suggest that it is the threat of such loss that enables us to discern a connection between speech acts and handicaps.15 A signal S’s being is a handicap may appear incompatible with a signaller O’s being able to choose whether or not to produce S. After all, O’s having a choice in the matter seems precisely to show that S is not a handicap – for how then could it be costly? This, I suggest, is where norms come in. A norm can make a signal costly to produce not by exacting calories or territory, but rather by making an agent subject to a loss of credibility. For a norm might
Showing and Meaning: On How We Make Our Ideas Clear 211
have the following schematic content; where S is a speech act and C a psychological state: One who produces S is to be in condition C; otherwise she is subject to a loss of credibility. A loss of credibility will, as we have suggested, reduce the weight of a speaker’s speech acts, thereby hobbling her ability to serve as a conversational ‘player’. As the likelihood and severity of a loss of credibility increase, so does the exposure to risk for those who produce S without being in condition C. Exposure to risk does not, of course, guarantee harm: I could go decades without a seatbelt or a scratch. However, even in such a case I have paid a price in the sense of having closed off a possibility, and as I reflect on those years I might shudder at my recklessness. If, then, a norm conforming to the above schema is in force, and it is common knowledge that it is, a signal S of condition C will be difficult to fake precisely because of the cost involved. For this reason producing S will show C by being a handicap. Further, when such a norm is in force the production of S will be a handicap even when producing S is subject to an agent’s choice. In addition, the condition of which S is an indicator need not be the organism’s overall fitness as that term is used in the ethological literature. Instead it could be any feature of the organism that it has reason to signal. For instance it could be something on which O can introspect if she is possessed of a capacity for introspection. As a result, if O can signal one of her introspectible states by incurring a handicap, then O may be able to show that state as well. Thus consider the one-word expression of gratitude, ‘Thanks!’ Its use is governed by the norm: One who utters and speaker-means ‘Thanks!’ is to be feeling gratitude toward her addressee; otherwise she is subject to a loss of credibility.16 No test, of course, is going decisively to settle whether a speaker has respected this norm. However, we have many ways of justifying conclusions about a person’s sincerity. My thanking a person and then sniggering behind their back; my failure to return the kindness for which I purport to show gratitude; my scoffing at others who show similar kindness, are all evidence that my ostensible gratitude is hollow. The hypothesis that some such norm as the above governs the use of ‘Thanks!’ explains how one’s sincere use of that word can show one’s gratitude. It does so by enabling the speaker to incur the risk of a loss of credibility if she is not feeling gratitude. As such it justifies the thought, whether consciously entertained or not, in the addressee: She wouldn’t say it if she didn’t feel that way. (Compare: That fly must be awfully fit to sport those flamboyant eye-stalks!)17
212
Mitchell Green
6. The power of force The use of a particular illocutionary force shows the nature of our commitment to a proposition or other type of content. To see how this occurs, consider a norm governing assertion. It might go somewhat as follows: One who asserts p is to believe p with justification adequate for knowledge; otherwise she is subject to a loss of credibility. I hasten to add, however, that for present purposes we need not settle all current controversies surrounding the norms governing assertion. Perhaps Williamson (1996, 2000) is correct that one is only to assert what one knows; perhaps on the other hand weaker norms are more appropriate. What matters for our purposes is that assertion is governed by a norm of sincerity; that the sincerity condition be a belief, and that the belief be justified. Compare assertion with conjecture. If I offer P as a conjecture, it is always inappropriate to reply with, ‘How do you know that?’ A legitimate challenge would instead involve showing that what I say is demonstrably wrong or very unlikely. Likewise, a conjecture should be backed with some justification or other; otherwise, put it forth as a guess. However, if you do put it forth as a conjecture, the required justification need not be as powerful as that for assertion (Green 1999). Assertions, conjectures, suggestions, guesses, presumptions and the like are cousins sharing the property of commitment to a propositional content. They differ from one another in the norms by which they are governed, and thereby in the nature of that commitment. Consequently, a speaker incurs a distinctive vulnerability for each such speech act – including a liability to a loss of credibility and, in some cases, a mandate to defend what she has said if appropriately challenged. These liabilities to error, exposed insincerity, and injunctions to defend put the speaker at risk of losing conversational ‘weight’ in the community in which she has a reputation (Brandom 1983; Green 1999, 2000). That is precisely what it is to stick one’s neck out. Assertions, conjectures, suggestions, presumptions, and the like thus carry a cost. Speakers pay it because the information they contribute to the group can help achieve its aims as well as enhance their own status therein. This is not to say that an individual speaker carries out such a calculation whenever she makes an assertion or other speech act. Rather, her long exposure to the norms of her milieu ingrains in her a disposition to make conversational contributions; that disposition is one the possession of which by many group members adds to the well-being of the group; however no individual need be cognizant of this fact. In spite of this, however, in the absence of some mechanism for vouchsafing honesty, proffered speech acts will be prone to abuse by those who take assertion and other speech acts lightly in the hopes of scoring epistemic points; after all, if you say enough
Showing and Meaning: On How We Make Our Ideas Clear 213
Table 9.1
Speech acts, what they show, and in what light they show it
speech act
shows
as
assertion that p conjecture that p educated guess that p
belief that p belief that p acceptance or belief that p acceptance of p acceptance of p
justified appropriate for knowledge justified Justified
guess that p presumption of p supposition of p (for argument)
acceptance of p
n/a justified for current conversational purposes aimed at the production of justification for some related content r
you’ll in all likelihood eventually get some things right. This temptation threatens to make assertors, conjecturers, etc., less worthy of our belief. Liars and those who say things on insufficient evidence threaten to undermine sincere speakers in the way that brightly coloured but non-noxious frogs threaten the credibility of signals sent by their ‘honest’ cousins. Enter speech acts and the norms governing them. In performing a speech act I incur a liability – a handicap in our technical sense. The nature of that liability determines not just what psychological state is shown, but also in what light it is shown. Just as different pictures of one and the same sitter might show him as brave, as menacing, as a fop, etc., different speech acts can show a single psychological state in different lights. The liability that an assertor undertakes by incurring commitments characteristic of that speech act give strong (not conclusive) evidence that the assertion is both sincere and justified. That liability is less stringent for the case of conjecture, with the result that a conjecture does not show one’s belief as justified in a way appropriate for knowledge; instead it merely shows one’s belief as having some justification or other. Standards are even less stringent for a so-called educated guess. Further, for a sheer guess these standards only put the speaker in danger of being wrong if her guess turns out to be incorrect. A table (9.1) may help to clarify these points. In the table, the second and third columns describe what felicitous speech acts show, and in what way they show it. While all six speech acts considered here involve commitment to a propositional content, only two require belief for their sincerity condition. Guesses, presumptions, and suppositions require only acceptance for their sincerity condition sensu Stalnaker (1984); educated guesses can go either way.18
7. The etiology of force After developing a conception of speaker-meaning that does not presuppose linguistic conventions, Grice suggests how such conventions might
214
Mitchell Green
have come about. The central thought is that they come into being as an ossification of a series of one-off cases of speaker-meanings.19 Thus suppose that in spite of our not sharing a language, I manage by means of some gesticulations to speaker-mean to you that there is quicksand nearby. Then on a future encounter you find yourself wanting to convey the same message to me. The thing most likely to succeed will be a perceptibly similar set of gesticulations. This pattern will constrain future interactions by being the most salient: any other way of indicating the message one is trying to get across will be less efficient than the one already established. In this way the original gesticulation might become a conventional way of signalling that quicksand is near. On this view, then, conventional meaning is an ossification of a series of instances of speaker-meaning. Such a view suggests, without quite implying, that the occurrence of such a series is necessary for the emergence of semantic conventions. Whether or not that is the case, what is doubtful is that a speech act could occur in the absence of some system for vouchsafing its reliability. It is one thing to give an indication of one’s belief that p: this can be done by acting as if one believes that p. As we have seen, however, it is quite another to assert that p, that is, to commit oneself to p’s truth in such a way as to present it as something one knows, or at least has strong justification for. Mere utterance of an indicative sentence will not suffice for this. Rather, there must have come about some means for indicating the way in which a speaker is putting forth a content. What could that means have been? This is a question that cannot be decided purely in the armchair. However, the Gricean model for such an historical account is not a plausible conjecture. On that model, there was long ago a one-off case of an illocution (an assertion, say), which then served as a model for later agents as they attempted to perform illocutions of their own. This is an unlikely model, however, since such a one-off case itself requires that a system of norms such as those depicted in the above table already be in place. Such a conjecture presupposes what was to be explained. Another conjecture that might come to mind also seems implausible on reflection. A speaker might try to make the force of her utterance clear simply by saying what that force is: ‘The following is an assertion: You’ll be more punctual from now on’. However, while it is certainly possible that ‘assertion’ and the like could be given meaning before systems of assertoric and related norms come into place, that’s not very likely. We are not in the habit of coining terms for systems of norms that are not yet extant. Furthermore, even if we did have such terminology in place, it is still doubtful that its use could effect a speech act in the absence of norms. A more plausible model is in the spirit of the ‘light dawns gradually over the whole’, slogan. That is, there need not have been in the history of our species a single definite point at which an illocution occurred, and before
Showing and Meaning: On How We Make Our Ideas Clear 215
which no illocutions had occurred. Rather, we start with patterns of behaviour in which speakers make some sort of commitment to the content they express, albeit a commitment whose details are not entirely perspicuous. This might be effected in an early instance by utterance of a proposition with a strident tone and a look of determination. (It’s no accident that cognate uses of ‘assert’ concern shows of force – for instance as we find in the verb phrase ‘assert oneself’.) Here, even though the speaker is manifesting her willingness to stand confidently behind that proposition, her behaviour will leave unanswered the question whether she has asserted, suggested, or conjectured the content she has expressed. This situation is problematic in that speakers are liable to be vexed by unclarity as to what a given act committed a speaker to: has she been shown up if the content she expressed turned out to be one for which she only had weak justification? Disputes over such questions as this might lead to frustration. To see how such problems can be addressed, observe that the development of norms governing and enabling speech acts is similar to the development of norms governing so-called social space. We operate with assumptions, usually unconscious, about how far away others are to stand when engaged with them in face-to-face conversation. We tend only to become aware of such norms when they are violated, for instance when we are visiting a country with different norms from our own. Yet in any location in which these norms are in force, there is little reason to think that they would have come into being as a result of conscious decisions. What is more, one can easily imagine a person implicitly learning to negotiate among different sets of such norms for different walks of life. With her Greek extended family she follows one set, while with her business colleagues in Singapore she follows quite another.20 So too the expression of a proposition might involve a person in some kind of commitment, even if it is not quite clear just what that commitment comes to. Over time, however, in a given speech community the nature of a person’s propositional commitment might be made more precise in light of such things as what counts as a legitimate challenge, and what sorts of things would cast the speaker’s credibility into doubt. Does the expression of a proposition oblige the speaker to have good reasons for what she says? If that proposition turns out to be incorrect, should we see the speaker’s credibility as being tarnished? A given linguistic community might answer questions like these in their practice rather than by means of decrees or other explicitly formulated policies. Another linguistic community might answer such questions differently, creating a different system of norms. Further, as with the case of personal space, a person might learn to move effortlessly between one system of norms and another without ever explicitly formulating for himself the norms at play in each case. A natural next step will be to name these different systems of norms: ‘assertion’, ‘supposition’, ‘educated guess’, as the cases may be. Further, once that step has been taken, one can then invoke such norms by saying that one is doing so.
216
Mitchell Green
8. Conclusion In the last two decades, illocutionary force has fallen behind semantics, as well as other areas of pragmatics, as a source of rich philosophical discussion. This may be due to the fact that after the rigorization offered by Searle and Vanderveken (1985), and Vanderveken (1990), it may seem that the most we may hope to do is taxonomize various kinds of speech act and study their relations to one another. This chapter has attempted to revitalize discussion of illocutionary force by relating it to social norms, and thereby to issues in the evolution of language. I have done that by arguing that force is an aspect of speaker-meaning, and then showing that speaker- and natural meaning are more complexly interrelated than the current consensus allows. It would be a welcome benefit if this chapter also stimulates more philosophers to learn about new research in the evolutionary biology of communication, and to consider the significance of that research for some of our classic questions.
Notes 1. The first example is inspired by Schiffer (1972); the second is from Vlach (1981). 2. Much of that literature is reviewed in Vlach (1981). 3. This line of objection is developed in further detail in Davis (1992b). Also, one might suggest that in such cases I am merely pretending to speaker-mean that my car should start (or that someday my daughter might be changing me). Perhaps actors on stage pretend to speaker-mean their lines; so too for those of us who sing along with the radio. However, in these cases there is no question of the speaker’s sincerity; it makes no difference to our assessment of the agent’s performance whether she is sincere. On the other hand, while intuitions might be less than vivid in the automobile case, it does seem clear that in the diaper-changing case we expect the speaker to be sincere: we expect what he says to be an expression of his state of thought or feeling. 4. It is only a first approximation because it is susceptible to a counterexample analogous to Strawson’s (1964) ‘river rat’ case. For an elucidation of speaker-meaning that is more cumbersome, but not susceptible to the river rat and related cases, see Green (2007c: chapter 4). 5. Davis (1992a) distinguishes between ‘cogitative’ and ‘cognitive’ speaker-meaning, where only the latter involves putting forth a content as true. The former is exemplified by such cases as expressing a proposition simply for the sake of examining its implications, and involves no commitment to the truth of that proposition. The cases that concern me in the text are closer to cognitive speaker-meaning than to cogitative speaker-meaning. 6. For the reasons for this qualification, see Green (2000). 7. Searle (1969) characterizes assertion of proposition P as an ‘undertaking to the effect that P’. That may be a necessary, but it is not a sufficient condition for being an assertion: My betting a large sum on P’s coming out true is an undertaking to the effect that P but is no assertion thereof. (For instance, if I bet on P without believing P true I am not a liar; if I assert P under those conditions, I am.)
Showing and Meaning: On How We Make Our Ideas Clear 217 8. This point was made forcefully by Davidson some decades ago. It has been clarified and defended in Green (1997). 9. We are not attributing any such thought to peahens, however. Ethologists instead merely take it that peahens act as if they harbour such thoughts by being selected for their ability to prefer flamboyant trains. 10. Green (2007c) distinguishes between three forms of showing: showing that something is the case; showing an object in such a way as to make it perceptible; and showing how something appears or feels. In this chapter I shall only be discussing the first of these three. 11. See Maynard Smith and Harper (2004) for a fuller explanation of this notion. 12. I leave as an open question, to be decided on the basis of empirical results from the evolutionary record, whether these forms of expressive behaviour are also signals of the emotions they indicate: Establishing that they are will require establishing that they were designed to show what they do. 13. Desalles (1998) offers a more general discussion of the relation of status to communication comporting with the spirit of the present proposal. Also, Stalnaker (2006) explores the notion of credibility in relating game theory to communication. 14. Not all utterances, not even all speech acts, admit of sincerity. For instance, the only infelicities to which resigning and appointing are prone are misfires, rather than abuses. 15. If a speech act is a handicap in any intuitive sense of ‘handicap’, that will not be because of physical limitations on the performer of the speech act. That is to say that ‘handicap’ in the present context should not be seen as a special kind of index, given our definition of the latter term. So we now define, as before, a handicap as a signal that can only be produced with great difficulty as a result of being costly to produce; however, that cost need not be construed in such physical terms as the expenditure of calories. Rather, ‘cost’ is now to be widened to include such things as one’s status within a social hierarchy. 16. It may be possible to speaker-mean one’s thanks without uttering any words, perhaps with a gesture. One can also utter the expression without speaker-meaning it, for instance on stage. That is why neither of the two conditions is redundant. 17. Unlike the female stalk-eyed fly, conspecifics in my linguistic community might well entertain such a thought. It does not follow that when they do so it must be consciously entertained. An emerging consensus in experimental psychology now posits a great many cognitive processes that may, but need not, be conscious. Instead they are automatic in a sense elucidated by Bargh and Chartrand (1999) and Wilson (2004). Pragmatic processes are a likely candidate for inclusion in this group. 18. A more formal rendition of these ideas would prove that a separating equilibrium (a Nash equilibrium in which signallers of different condition are reliably distinguished from one another) exists in which credible speakers may be distinguished from the non-credible, and in which agents who receive signals from credible speakers tend on the whole to add the contents of those signals to common ground while also registering the epistemic state from which those contents flow. While, as argued in Green (2009), the particular form of his proposal is not compelling, the game-theoretic tools van Rooij (2003) uses to make out his case for politeness as a handicap should give the flavour of how a mathematization of the present line of thought would go. 19. Grice’s suggestions on this matter are expounded and developed in Avramides (1987).
218 Mitchell Green 20. The view of conventional norms here is consilient with that summarized and defended in Gilbert (2008). In particular, the present view is most naturally developed in non–individualistic terms, and does well to eschew commitment to the idea that what perpetuates norms over time is rational agency on the individual level.
References Austin, J.L. (1962) How to Do Things with Words, ed. J.O. Urmson and M. Sbisá, 2nd edn (Cambridge, MA: Harvard University Press). Avramides, A. (1987) Meaning and Mind: An Examination of a Gricean Account of Language (Cambridge, MA: MIT Press). Bargh, J. and Chartrand, T. (1999) The unbearable automaticity of being. American Psychologist, 54: 462–79. Barrett, L., Henzi, S., Weingrill, T., Lycett, J. and Hill, R. (1999) Market forces predict grooming reciprocity in female baboons. Proceedings of the Royal Society, B, 266: 665–70. Brandom, R. (1983) Asserting. Noûs, 17: 637–50. Cheney, D. and Seyfarth, R. (2007) Baboon Metaphysics: The Evolution of a Social Mind (Chicago: Chicago University Press). David, P., Hingle, A., Greig, D., Rutherford, A., Pomiankowski, A. and Fowler, K. (1998) Male sexual ornament size but not asymmetry reflects condition in stalk-eyed flies. Proceedings of the Royal Society of London, B, 265: 2211–16. Davis, W. (1992a) Cogitative and cognitive speaker meaning. Philosophical Studies, 67: 71–88. Davis, W. (1992b) Speaker meaning. Linguistics and Philosophy, 15: 223–53. Desalles, J.-L. (1998) Altruism, status and the origin of relevance. In J. Hurford, M. Studdert-Kennedy and C. Knight (eds.) Approaches to the Evolution of Language (Cambridge: Cambridge University Press), pp. 130–47. Dunbar, R. (1996) Grooming, Gossip, and the Evolution of Language (Cambidge, MA: Harvard University Press). Ekman, P. (1998) Afterword to P. Ekman (ed.) Charles Darwin’s The Expression of the Emotions in Man and Animals, 3rd edn (Oxford: Oxford University Press), pp. 363–94. Gilbert, M. (2008) Social convention revisited. Topoi, 27: 5–16. Green, M. (1997) On the autonomy of linguistic meaning. Mind, 106: 217–44. Green, M. (1999) Illocutions, implicata, and what a conversation requires. Pragmatics & Cognition, 7: 65–92. Green, M. (2000) Illocutionary force and semantic content. Linguistics & Philosophy, 23: 435–73. Green, M. (2003) Grice’s frown: on meaning and expression. In G. Meggle and C. Plunze (eds.) Saying, Meaning, Implicating (Leipzig: University of Leipzig Press), pp. 200–19. Green, M. (2007a) Moorean absurdity and showing what’s within. In M. Green and J. Williams (eds.) Moore’s Paradox (Oxford: Oxford University Press), pp. 189–214. Green, M. (2007b) How do speech acts express psychological states? In S. L. Tsohatzidis (ed.) John Searle’s Philosophy of Language: Force, Meaning and Thought (Cambridge: Cambridge University Press), pp. 267–84. Green, M. (2007c) Self-Expression (Oxford: Oxford University Press).
Showing and Meaning: On How We Make Our Ideas Clear 219 Green, M. (2007d) Expression, indication and showing what’s within. Philosophical Studies, 137: 389–98. Green, M. (2007e) Speech acts. The Stanford Online Encyclopedia of Philosophy. Green, M. (2009) Speech acts, the handicap principle and the expression of psychological states. Mind & Language, 24: 139–63. Grice, H.P. (1989) Studies in the Way of Words (Cambridge, MA: Harvard University Press). Lewis, D. (1979) Scorekeeping in a language game. Journal of Philosophical Logic, 8: 339–59. Lewis, D. (1980) Index, context, and content. In Kanger and Ohman (eds.) Philosophy and Grammar (Dordrecht: D. Reidel). Maynard Smith, J. and D. Harper (2004) Animal Signals (Oxford: Oxford University Press). Millikan, R. (1998) Language conventions made simple. Journal of Philosophy, 95: 161–80. Petrie, M., Halliday, T.R. and Sanders, C. (1991) Peahens prefer males with elaborate trains. Animal Behaviour, 41: 323–31. Reichert, S. (1978) Games spiders play: behavioral variability in territorial disputes. Behavioral Ecology and Sociobiology, 3: 135–62. Reichert, S. (1984) Games spiders play III: cues underlying context-associated changes in agonistic behavior. Animal Behaviour, 32: 1–15. Schiffer, S. (1972) Meaning (Oxford: Oxford University Press). Searle, J. (1969) Speech Acts (Cambridge: Cambridge University Press). Searle, J. and Vanderveken, D. (1985) Foundations of Illocutionary Logic (Cambridge: Cambridge University Press). Stalnaker, R. (1972) Pragmatics. In D. Davidson and G. Harman (eds.) Semantics of Natural Language (Dordrecht: D. Reidel), pp. 380–97. Stalnaker, R. (1973) Presuppositions. Journal of Philosophical Logic, 2: 447–57. Stalnaker, R. (1974) Pragmatic presuppositions. In M. Munitz and P. Unger (eds.) Semantics and Philosophy (New York: New York University Press), pp. 197–13. Stalnaker, R (1984) Inquiry (Cambridge, MA: MIT Press). Stalnaker, R. (2006) Saying and meaning, cheap talk and credibility. In A. Benz, C. Jäger and R. van Rooij (eds.) Game Theory and Pragmatics (Basingstoke: Palgrave Macmillan), pp. 83–100. Strawson, P. (1964) Intention and convention in speech acts. The Philosophical Review, 73: 439–60. Thomason, R. (1990) Accommodation, meaning and implicature: interdisciplinary foundations for pragmatics. In P. R. Cohen, J. Morgan and M. E. Pollock (eds.) Intentions in Communication (Cambridge, MA: MIT Press), pp. 325–64. van Rooij, R. (2003) Being polite is a handicap: Towards a game-theoretical analysis of polite linguistic behavior. In M. Tennenholtz (ed.) Proceedings of TARK IX. (New York: ACM Press). Vanderveken, D. (1990) Meaning and Speech Acts, Vols. I and II (Cambridge: Cambridge University Press). Vlach, F. (1981) Speaker meaning. Linguistics and Philosophy, 4: 359–91. Wharton, T. (2003) Natural pragmatics and natural codes. Mind and Language, 18: 447–77. Wilkinson, G. and Reillo, P. (1994) Female choice response to artificial selection on an exaggerated male trait in a stalk-eyed fly. Proceedings of the Royal Society of London, B, 255: 1–6.
220
Mitchell Green
Williamson, T. (1996) Knowing and asserting. The Philosophical Review, 105: 489–523. Williamson, T. (2000) Knowledge and its Limits (Oxford: Oxford University Press). Wilson, T. (2004) Strangers to Ourselves: Discovering the Adaptive Unconscious (Cambridge, MA: Harvard University Press). Zahavi, A. and Zahavi, A. (1997) The Handicap Principle: A Missing Piece of Darwin’s Puzzle (Oxford: Oxford University Press).
10 Illocution, Perillocution and Communication Klaus Petrus
1. Introduction Much of the debate between conventionalists and intentionalists in speech act theory depends on the question of whether Paul Grice’s theory of meaning is suitable for an analysis of illocutionary acts. This debate goes back to Peter Strawson’s ‘Intention and convention in speech acts’, dating from 1964. In this essay he holds that John Austin’s characterization of illocutionary acts is too one-sided. Strawson concedes that there actually are essentially conventional speech acts, but this only applies to a certain sort, for instance to highly ceremonial acts like marrying, baptizing, or swearing. Apart from that there is a multitude of illocutionary acts like asserting, telling, or warning, whose performance does not require any particular conventional procedures. What makes up their kernel – and in a certain respect, as Strawson thinks, all illocutionary acts – is rather that inherent to them there is a specifically Gricean element. Since these days it has been intensely discussed in speech act theory wherein this Gricean element exactly consists, to which degree illocutionary acts are essentially intentional, or whether it is not conventions, rules, or norms whatever which prove to be indispensable after all. In this chapter, conventions shall be touched on merely once and even then only in passing. What rather matters to me is showing that Grice’s theory is not suitable for the analysis of illocutionary acts. It is, however, appropriate for one kind of speech acts which are easily confused with illocutionary acts, and which I would like to name perillocutionary acts. More concretely, I shall proceed as follows: section 2 deals with the central feature of illocutionary acts. For reasons of space I shall have to restrict myself to commenting on the definition of illocutionary acts only very roughly. In this chapter I shall not be able to say something about how this definition is embedded in a theory of illocutionary acts, and about how it is connected with the other two central subjects of speech act theory, i.e. meaning and communication (see, however, Petrus 2006). In section 3, I shall deal with the question of whether illocutionary acts are essentially intentional. As briefly 221
222
Klaus Petrus
mentioned, it is to be answered in the negative: illocutionary acts are not essentially intentional – at least not in the sense normally assumed. In section 4 I shall characterize a particular type of perillocutionary acts in more detail. Afterwards, in section 5, I will speculate on the question of whether in this regard we are dealing with communicative acts à la Grice. In section 6, I would like to highlight once again, wherein – to my mind – the absolutely decisive difference between illocutionary and perillocutionary acts consists.
2. Illocutionary acts First of all, let’s turn to the question of what illocutionary acts are. Illocutionary acts are essentially characterized by the fact that they can be performed in a special way. The corresponding definition, which basically goes back to Andreas Kemmerling, goes as follows (Kemmerling 1997, 2001; Siebel 2002): (ILL)
An act C is an illocutionary act if and only if in doing what one does making it clear that one wants to C in doing what one does is, by conceptual necessity, sufficient for therein C-ing.1
Let’s informally abbreviate (ILL): (ILL)
C is an illocutionary act iff 䊐C(∀S)(∀C)(∃ψ) CS ⇐ (ψS & MS(ψS (ψS, WS(CS))))
These abbreviations are to be read as follows: 䊐C: CS : ψS: M S: WS:
It is conceptually necessary that – S performs an act of type C S does ψ S makes it clear that – S wants / intends that –
For clarification, I shall now add a handful of comments concerning (ILL) (see in more detail Petrus 2003: 315–22, 2005: 110–18). •
(ILL) deals with the essence of illocutionary acts in a generic sense. This may sound old-fashioned, but it hits the point. The question to be considered is about what actually makes any act whatever an illocutionary act, and not about specifics of particular illocutionary acts, for instance about what makes an utterance an assertion, or even the assertion that Ferdinand is a dangerous bull (see also Kemmerling 2001: 95). In other words, by means of (ILL) the defining feature of all illocutionary acts is captured. No matter what may characterize assertions, warnings, or orders in detail, or distinguish them from one another, all illocutionary acts
Illocution, Perillocution and Communication
●
223
share the property that they can be performed (under appropriate conditions) by simply making it clear that one wants to perform them. Here is a list of paradigmatically illocutionary acts: asserting, telling, ascertaining, reporting, describing, promising, warning, requesting, advising, ordering. It is important to notice that in dealing with (ILL) we are concerned with the mode of performance and not with the conditions of the performance of illocutionary acts. The latter concern a specific property of certain speech act types. For example, who wants to give an order must be entitled to do this; i.e. he must possess the necessary authority. Otherwise what he does is not an order, but, for instance, a request. Now, the decisive point is that such conditions of the performance on their own tell us nothing about the question of whether or not the speech act under scrutiny is an illocutionary act. (In order to convince someone of something – a typically perlocutionary act – certain requirements must likewise be fulfilled.) Whether an order counts as an illocutionary act depends on the question of whether it is already performed by making it clear that one wants to give someone an order. To put it differently, whether or not a certain speech act is considered as being of the type of an illocutionary act at all holds independently of those conditions of the performance which are fulfilled in the individual case. That the respective conditions of the performance are fulfilled shall, in what follows, always be presumed. Accordingly, the question to which (ILL) knows the answer reads: Is the speech act under scrutiny an illocutionary act at all? Furthermore it should be noticed that in dealing with (I LL) we are concerned with types of actions (i.e. CS), and not tokens. Consequently, the definition merely stipulates in which cases, for instance, the speech act type ‘warning’ is an illocutionary act. From this we have to separate your concrete warning or mine, and even more what we both do respectively in order to warn someone (i.e. ψS). According to (ILL) you may pull a funny face, make a sweep of the hand, produce a drawing, or utter a sentence – the main thing is that you do something thereby making it clear that you want to warn someone in this way. Thus, (ILL) does not contain any restrictions worth mentioning with respect to what exactly someone does or has to do in order to perform an illocutionary act. By noting this it is of course not ruled out that there are more or less suitable means of making it clear that one wants to perform an illocutionary act, as well as which illocutionary act one wishes to perform. In Austin’s opinion for example, explicitly performative formulae of the kind ‘Hereby I x that –’ (in which ‘x-ing’ stands for a performative or illocutionary verb) are especially well-suited (Austin 1962: 32f., 73, 135). In more concrete terms, he thought that there is a convention according to which an illocutionary act – a warning for instance – can be performed by uttering a sentence by means of which the act itself
224 Klaus Petrus
●
●
is lexically topicalized – a sentence such as ‘I warn you: Ferdinand is a dangerous bull’. Thus considered, Austin’s idea – which to my mind is of great relevance – shows perfectly perceivable similarities with the mode of performance of illocutionary acts described in (ILL). Insofar as the fact that these speech acts can be performed in an explicitly performative manner makes them conventional actions, one could concede without any problems that the essence of illocutionary acts contains a conventional component – namely in the sense that now and then they are performed by simply making it conventionally clear that one wants to perform them (see in more detail, Petrus 2005). In (ILL) ‘making clear’ (i.e. MS) – or, as I shall also say, ‘making explicit’ – is to be taken as a factive verb. This means that by doing ψ S cannot make it clear that he wants to C if it is not the case that he wants to C by means of ψ. Yet, according to the definition the point is not merely that an illocutionary purpose is factually (always) fulfilled if one makes it clear that one wants to perform it, but rather that is so for conceptual reasons (i.e. 䊐C; Kemmerling 2001: 89; Siebel 2002: 147). Illocutionary acts are, as far as their generic essence is concerned and thus in conceptual respects, actions which can be performed (ILL)-like; i.e. they are already performed by making it clear that one wants to perform them. As regards the question of whether illocutionary acts are essentially intentional, this point is of prime importance. More about that shall be said in the next section. That S makes it clear that he pursues the illocutionary purpose in question does not imply that it becomes clear to the addressee A – that A actually understands – what S aims at, or that such an understanding is intended by S. By raising his hand S may make it explicit that he would like to greet A by means of this gesture – and he may actually have greeted A in this way – without this becoming clear to A. In other words, an analysis of ‘S makes it clear that p’ in the context of (ILL) does not yield any specification which would indicate ascriptions of the kind ‘To A it becomes clear / A understands that p’, or ‘S intends that it becomes clear to A / that A understands that p’. Thus, according to (ILL) neither factual understanding on the part of A, nor the intention of S that A should understand which illocutionary purpose he pursues, belong to the felicity conditions of illocutionary acts. Illocutionary acts are not necessarily (attempts at) communicative actions. Rather, they stand out owing to their peculiar mode of performance: they are actions which can be performed by simply making it clear that one wants to perform them. So much for the definition of illocutionary acts.
3. Illocutionary acts and intentions Let’s now turn to the question of whether and in which sense illocutionary acts are essentially intentional, and can therefore – as it is often assumed – be
Illocution, Perillocution and Communication
225
described by means of a Gricean theory of meaning. Since in the following I shall presume that the generic essence of illocutionary acts is correctly captured by (ILL), the discussion amounts to the question of whether this definition necessarily displays something Gricean. It is common knowledge that Grice’s original analysis aimed at explicating the fact that a non-natural sign can mean something, and what it means (Grice 1989: 213–23). His answer was as follows: That a sign has non-natural meaning is due to the fact that the speaker S means something by using it; and this again means that he has the intention of producing a particular reaction or effect in the addressee A – namely by wanting A to recognize that he entertains this particular intention. What the sign in question means correspondingly depends on what S means by it, or which reaction he intends to induce in A. It is quite easy to notice that while we are talking about Grice’s theory of meaning we are not talking about speech act theory. But still, not a few people think that his approach to a theory of the illocutionary can be made use of perfectly well – insofar as one is willing to pass at least three stages.2 Stage 1: Interpret Gricean cases of meaning as attempts to influence the addressee in a rational way (Kemmerling 1986). We are dealing with attempts at influence A insofar as S wants A to react in a certain way. According to Grice this reaction consists either in A’s believing something or else in A’s having the intention to do something. Such an attempt to influence A is rational in at least two respects: Firstly, S has reasons for the choice of a certain sign x, and it is part of these reasons that he can rely upon A’s recognizing in the face of x that S wants A to produce r, and that A indeed produces r owing to this recognition. Secondly, Gricean attempts to influence are rational insofar as the reaction desired by S is under the control of A, which means: S presupposes that A for his part has reasons for producing r, and it is part of these reasons that A recognizes that S wants him to react like that. Stage 2: Conceive of Gricean attempts to influence the addressee as attempts at rational communicative acts (Meggle 1981) and thus as cooperation games which require that either both, S and A, win, or else both lose: S only wins if A recognizes what S intends by x; and he only succeeds if it really becomes clear to A what S is up to (Petrus 1999). All this of course requires that we transform Grice’s analysandum ‘S meant something by x’ (more or less tacitly) into ‘S attempts to communicate something by x’ (see already Strawson 1964). Stage 3: Consider communicative acts à la Grice (i.e., stage 1 and 2) as performances of a certain sort of actions, viz. as illocutionary acts. This is roughly – very roughly – the path which leads from meaning via communication to the illocutionary. (Those who wish to know how demanding this route is may read Schiffer 1972 and Bach and Harnish 1979. In fact, the three stages run uphill, passing mountainous regions.) In the
226
Klaus Petrus
light of the remarks at the beginning of this section the thesis would now be more or less like this: who strives for an adequate analysis of illocutionary acts must take this route – and this even if he assumes the validity of definition (ILL). As far as I can see, this can mean two things. The first interpretation goes as follows: illocutionary acts necessarily involve Gricean intentions; therefore they are essentially intentional. Indeed, one could easily hit on the idea of reinterpreting (ILL) as follows: (GRI)
An utterance X is illocutionary if and only if for performing X it suffices to make it clear that one intends to perform X.
Thus understood, however, the thesis is not accurate. As we have seen, it is correct indeed that ‘making clear’ in (ILL) or, alternatively, in the Gricean version (GRI) is to be understood as a factive verb. This means that S is not able to make it clear that he intends to perform a certain illocutionary act without having this intention (Kemmerling 2001: 84f.; Siebel 2002: 148). But still it would be wrong to infer that illocutionary acts are actions which are necessarily performed with certain intentions. Even in a Gricean interpretation of (ILL), making it clear that one has this intention constitutes only a sufficient condition for the successful performance of an illocutionary act. Consequently, in (GRI) there is no indication at all that such an action can only be performed if the speaker makes the intention he entertains clear. Strictly speaking, (GRI) does not even stipulate that the performance of the action involves any intentions at all which would have to be made clear by the speaker. Even (GRI) ‘merely’ says that is sufficient for illocutionary acts to make the intention in question clear. Incidentally, the same applies even if – as Strawson and his followers have done – the above thesis is merely related to a certain sort of illocutionary acts, namely to so-called communicative illocutionary acts (Petrus 2010). In this case as well it should be taken into consideration that by (ILL) or (GRI) it is not excluded at all that, for example, assertions, or warnings are indeed performed with (Gricean) intentions now and then, or even quite often. Moreover, with respect to such a restricted thesis it is crucial to notice that at best it allows statements about the peculiarity of specific illocutionary acts. Against this, in (ILL) we are dealing (as emphasized before) with the generic essence of illocutionary acts – and this actually even if the definition is couched in Gricean jargon. The second interpretation, on the other hand, goes as follows: illocutionary acts are essentially intentional because performing them contains a genuinely Gricean element. Let’s call this element the ‘Gricean mechanism’ as it is usually done (Bennett 1976: §40ff.; Kemmerling 2001). It is this very mechanism on which the speaker relies if he seeks to influence his interlocutor in a rational way (compare stage 1 above). Concretely speaking, S is confident that should A recognize that S wants A to produce
Illocution, Perillocution and Communication
227
a certain reaction, exactly this recognition too is a reason for A to react in the way desired by S. Or more succinctly: S presupposes that the recognition of his intention leads to the fulfilment of the latter. Speaking by way of example, S assumes that A believes Ferdinand to be a dangerous bull because (among others) A has recognized that S wants him to believe this. Whenever S makes such an assumption he relies on the said mechanism. In fact, this is reminiscent of the mode of performance of illocutionary acts described in definition (I LL). At any rate they are actions which are performed by simply making it clear (in Gricean usage: by making it recognizable or giving to understand) that one wants to perform (or intends to perform) them. However, it is important to carefully keep apart two things once more. What is correct is: for an action of type C to be a candidate for the illocutionary it must be possible to perform C in such a way that it is already performed if one makes it clear that one wants to perform C. An act which cannot be performed (ILL)-like under any conceivable circumstances cannot, in other words, be an illocutionary act simply for conceptual reasons. This, however, does not mean that C is always or at least standardly performed in this way. To put it differently, it is not a necessary condition that the successful performance of illocutionary acts is based on the Gricean mechanism. And it is just as little part of the essence of such actions that their performance is normally based on this mechanism. Against that, illocutionary acts stand out owing to their necessary affiliation with the type of actions which can be performed by relying on the Gricean mechanism or by making it recognizable that one wants to perform these actions. Should these deliberations be correct, illocutionary acts neither involve Gricean intentions in any compelling way nor is the mode of performance of these actions necessarily, in every single case or merely usually, based on the Gricean mechanism. Grice’s theory of rational communication is therefore not suitable for illocutionary acts.
4. Perillocutionary acts Nevertheless, Grice’s account applies to a sort of speech acts which are strikingly similar to illocutionary acts and which I will call ‘perillocutionary acts’. In this section, I shall first of all characterize this sort of speech acts very briefly and then pursue the question about the extent to which we are concerned with Gricean communicative acts. In doing so I should emphasize that the structure of perillocutionary acts (against expectations) proves to be very complex and that here and now I have to be content with looking at merely one type of perillocutions in more detail. Roughly speaking, a speaker S performs a perillocutionary act if he performs a perlocutionary act by performing an illocutionary act. S, for example, may promise his interlocutor A to arrive on time (illocutionary), and by doing
228 Klaus Petrus
so reassure him (perlocutionary). Or he may ask after A’s aunt (illocutionary) and thus sidetrack from the topic (perlocutionary). As can easily be noticed, this characterization presupposes the (in speech act theory up to today intensely discussed) illocutionary/perlocutionary distinction. I consider it to be a further merit of definition (I LL) that it does not only capture the generic essence of illocutionary acts but at the same time functions as a prima facie criterion for distinguishing between illocutionary and perlocutionary acts. For as regards perlocutionary acts, it is simply not accurate to claim that they are (even under appropriate conditions) performed by simply making it clear that one wants to perform them. 3 In order to, for instance, convince someone or to encourage him, it typically requires something further which goes beyond the sheer clarification of these purposes. If we want to understand in which sense the performance of a perlocutionary act asks for the something ‘extra’, we have (among other things) to find out how perlocutionary purposes can be achieved. In this regard, at least two distinctions are to be taken into consideration.4 Firstly, we have to distinguish between direct and indirect perlocutionary acts.5 As regards the former, what S does is not relevant to S’s achieving the perlocutionary purpose in question, whereas it is relevant as regards the latter – and this even in the following sense: S assumes that should A recognize what S does (i.e. to which type of speech act his utterance belongs), this will give A a reason to produce the perlocutionary effect desired by S. Let’s from now on call this connection assumed by S the ‘perlocutionary nexus’. Furthermore, we shall abbreviate the perlocutionary effect intended by S by means of 'FA’ as well as (with Austin) assume that regarding S’s doing – the so-called ‘perlocutionary trigger’ – we are dealing with a phonetic, phatic, rhetic, or illocutionary act. Secondly, we have to distinguish between overt and covert perlocutionary acts. A perlocutionary act is overt if it is completely overt or partially overt. It is completely overt if S wants A to recognize that S assumes the above mentioned perlocutionary nexus. A perlocutionary act is partially overt if S really wants A to recognize what S does and wants A to recognize that S wants to achieve F, but wants as well that A does not recognize that S assumes the said perlocutionary nexus. A perlocutionary act may be regarded as covert if it is completely covert or else partially covert. It is completely covert if S wants A to recognize nothing of the things set out so far. It is partially covert if S definitely wants A neither to recognize that S assumes the perlocutionary nexus nor to recognize that S wants to achieve F, but nonetheless wants A to recognize which speech act type his acting belongs to. Against the background of these distinctions, we can formulate a typology of perillocutionary acts, even though in the following, as mentioned, only one sort will matter to me, namely perillocutionary acts which are indirect and overt. From now on, the label ‘perillocutionary act’ will only be
Illocution, Perillocution and Communication
229
used for this sort of speech acts which are, among others, characterized by the following features: (1) (2) (3) (4) (5)
CS WS(FA) WS(UA(WS(FA))) BS(CS → FA) BS(UA(CS) ⇒ FA)
These informal abbreviations are to be read as follows: WS : UA: B S: →: ⇒:
S wants / intends that – A recognizes (below: understands) that – S believes that / relies on – – leads to – – is reason for –
Again some comments on the features listed above. (1) states that as far as the perlocutionary trigger is concerned we are dealing with an utterance CS which counts as an illocutionary act in the sense of definition (ILL) in section 2. (2) states that S pursues a perlocutionary purpose, or wants to produce a corresponding effect in A, while S – see feature (3) above – also wants that this is recognizable for A. Fairly normal circumstances presumed, convincing, confirming, supporting, encouraging, cheering up, or reassuring should be exemplars of such ‘overt’ purposes. On the other hand, again normal circumstances presumed, persuading, deterring, intimidating, surprising, exposing or outwitting are more like ‘covert’ perlocutionary purposes or effects. (4) states that S regards C as a suitable means for achieving F. After Austin there have been a lot of speculations about the extent to which the connection between C and F is conventional (after all) (see, among others, Cohen 1973; Schlieben-Lange 1974, 1976; Davis 1979). I doubt that this question can be answered by way of conceptual analysis. At any rate, as regards perillocutionary acts we are allowed to presume that S relies on the assumption that should he perform an illocutionary act C A will recognize that S (by this act) seeks to achieve the perlocutionary effect F. Thus, in addition to (1)–(5), feature (6) holds: (6)
BS(CS → UA(WS(FA)))
(6) stands for the perlocutionary nexus assumed by S, and this feature presupposes that S wants A to recognize that S performs an illocutionary act C, i.e.: (7)
WS(UA(CS))
230 Klaus Petrus
Finally, in case of completely overt acts, the above list needs to be completed (at least) by addition (8): (8)
WS(UA(BS(UA(CS) ⇒ FA)))
So much for the rough description of perillocutionary acts.
5. Perillocutionary acts and communication à la Grice Let’s now ask whether perillocutionary acts, unlike illocutionary acts, are communicative acts à la Grice. In my opinion the answer must be yes, and this for at least three reasons. Firstly, perillocutionary acts are obviously attempts at influencing the addressee in a rational way. S wants to produce a certain (namely perlocutionary) effect in A (he wants, for instance, to convince A of something or prevent him from doing something). S is not merely concerned with achieving this effect by taking any path whatever, but – see (4) – by taking the (in his eyes) strategically best way. In other words, S has reasons to perform C while it is part of these reasons that S believes that A recognizes in the face of C that he wants to achieve F – see (6) – and that A – see (5) – owing to this recognition will react in the way expected by S. This means that A himself has reasons to react like that and it is, among others, part of these reasons that he recognizes that S pursues F. With respect to this, one should keep in mind that without exception we are concerned with A-reasons, and that S believes or presupposes that these are important to attaining his goal. Whether A really has these reasons or which reasons A has at all, is another question. Secondly, perillocutionary acts are clearly (overt) attempts at communicative acts. In this connection, two communicative goals can fundamentally be distinguished with respect to which S is overt or entertains no covert intentions as regards A. The one goal names the primary purpose of S’s venture – see (2): S wants to produce a perlocutionary effect F in A. Consequently, A has understood what S primarily aims at if (3) is fulfilled, i.e. if (α) holds: (α)
UA(WS(FA))
The other communicative goal could be called the strategic goal: S wants – see (7) – A to recognize that he performs an illocutionary act C as well as which act he performs. Accordingly, A has understood what S strategically aims at if (7) is fulfilled, i.e. if (β) holds: (β)
UA(CS)6
As said before, one of the reasons why S wants to achieve the strategic goal is that S relies on the perlocutionary nexus. This is what the third similarity
Illocution, Perillocution and Communication
231
with communicative acts à la Grice consists in. That S presupposes a perlocutionary nexus actually means that he relies on a kind of Gricean mechanism, insofar as he – see (5) – relies on the possibility that should A recognize that S performs C, this too will be a reason for A to really react in the way expected by S. Now, the following should be noticed: (α) and (β) do not belong to the felicity conditions of the performance of perillocutionary acts. An utterance counts as perillocutionary if (1)–(7) are given (and perhaps some further conditions). It is a necessary part thereof that S wants A to understand what he aims at in the strategic as well as in the actual sense. Thus considered, perillocutionary acts are essentially attempts at communication. By contrast, (α) and (β) specify conditions under which a perillocutionary act succeeds, or under which such a kind of communication comes off. This suggests seeing in (β), if not a guarantor (at least from the perspective of the speaker), at least a precondition for the successful performance of a perillocutionary act, since this act is successful if and only if A produces the desired perlocutionary effect in the way expected by S, i.e. if (χ) holds: (χ)
UA(CS) ⇒ FA
And this exactly requires (at the minimum) that A understands what S strategically aims at. So much (or better: so little) for perillocutionary acts and for three reasons why I consider them to be suitable candidates for communicative acts à la Grice.
6. Illocutionary vs. perillocutionary acts I would like to conclude by (once again) seizing the opportunity to go into the, in my opinion, central difference between illocutionary and perillocutionary acts. First of all, illocutionary acts are easily confused with perillocutionary acts (and the other way round). Yet illocutionary acts are never perillocutionary acts. This is true even – and especially – if they are used for perlocutionary purposes, since a speaker who does this always pursues a strategic goal in advance. This means that he wants his interlocutor to understand that he performs an illocutionary act as well as which act he performs. As seen, it is indispensable to the success of a perlocutionary act that this goal is achieved. Nothing of all of this, however, holds for illocutionary acts! This is so because illocutionary acts are acts which come about completely independently of the speaker’s pursuing such a strategic goal. In other words, whenever a speaker uses an illocutionary act for perlocutionary purposes he, strictly speaking, does not perform an illocutionary act anymore, but rather
232 Klaus Petrus
and in general a perillocutionary act. This, by the way, holds as well if S performs an illocutionary act for his own sake so to speak, i.e. merely with the intention that it should become clear to A that he, for example, wants to assert something (and does not pursue any further, i.e. perlocutionary purposes). The special feature of this kind of communicative act ‘merely’ consists therein, that if it is the case that it turns out well it has already been successful. Of course it is possible, if one really wishes to do so, to name the strategic goal the ‘illocutionary purpose’ and to call the corresponding effect, which is to be achieved on the part of the addressee (i.e. understanding), the ‘illocutionary effect’ (as has usually been done in certain circles since Searle 1969). Yet it should always be kept in mind that this effect is only called like this because A’s understanding is directed at an illocutionary act. And above all one should always, and recognizably for anyone, note that S’s intention to achieve such an effect does not belong to the felicity conditions of the performance of illocutionary acts. Whoever seriously wanted to maintain that illocutionary acts are essentially directed at achieving an illocutionary effect or a certain understanding on the part of the addressee would not only misjudge the (generic) essence of the illocutionary, but would also confuse the illocutionary with the perillocutionary. For illocutionary acts are not communicative acts à la Grice at all. Perillocutionary acts, on the other hand, are.7
Notes 1. For stylistic reasons I shall mostly use a shorter version of (ILL) which goes as follows: C is an illocutionary act if and only if making it clear that one wants to C is sufficient for C-ing. Kemmerling’s definition is geared to so-called ‘gricy actions’ and is meant to show that such actions (to which, in any case, certain illocutionary acts belong as well) are neither essentially intentional nor essentially conventional. 2. In the opinion of other authors the Gricean approach is directed at perlocutionary acts right from the start. Those who stick to the speech act-theoretic dogma, according to which the illocutionary should not be explained by means of the perlocutionary, will thus not want to pass the subsequent stages just for this reason (e.g. Alston 2000). Those, however, who look past this principle will do it readily and note that every effect which is to be achieved by means of a speech act is a perlocutionary effect anyway (e.g. Ulkan 1992). I shall keep out of this dispute until shortly before the end of this chapter. 3. Perlocutionary acts need not be intended. In the following I shall, however, assume that S pursues a perlocutionary purpose by his utterance. In this connection it holds for perlocutionary purposes (analogously to acts) that they are not fulfilled by simply making it explicit that one wants to achieve them. 4. What follows now is not a contribution to the specification of this rest. Quite a lot of things may be crucial to the success of perlocutionary acts, and most of them may not be accessible by conceptual analysis alone (Marcu 2005; furthermore Cohen and Levesque 1990; Gu 1993).
Illocution, Perillocution and Communication
233
5. It should not be confused with Cohen’s (1973) distinction of the same name. 6. In the case of completely overt perillocutionary acts (8) is part of the strategic goal of S. Accordingly, A understands what S strategically aims at if UA(BS(UA(CS) ⇒ FA)) holds. With respect to the following it is sufficient though if we concentrate on partially overt acts. 7. This work was supported by the Swiss National Foundation for the Promotion of Scientific Research (project-numbers 101511–113985 and PP001–114812/1).
References Alston, W. P. (2000) Illocutionary Acts and Sentence Meaning (Ithaca and London: Cornell University Press). Austin, J. L. (1962 [1955, 1975]) How to Do Things With Words (Oxford: Oxford University Press). Bach, K. and Harnish, R. M. (1979) Linguistic Communication and Speech Acts (Cambridge, MA: MIT Press). Bennett, J. (1976) Linguistic Behaviour (Cambridge: Hackett). Cohen, P. R. and Levesque, H. J. (1990) Persistence, intention, and commitment. In P. R. Cohen et al. (eds.) Intentions in Communication (Cambridge, MA: MIT Press), pp. 33–69. Cohen, T. (1973) Illocutions and perlocutions. Foundations of Language, 9: 492–503. Davis, S. (1979) Perlocutions. In J. R. Searle et al. (eds.) Speech Act Theory and Pragmatics (Dordrecht: Reidel), pp. 37–55. Grice, H. P. (1989) Studies in the Way of Words (Cambridge, MA and London: Harvard University Press). Gu, Y. (1993) The impasse of perlocutions. Journal of Pragmatics, 20: 405–32. Kemmerling, A. (1986) Utterer’s meaning revisited. In R. Grandy and R. Warner (eds.) Philosophical Grounds of Rationality (Oxford: Oxford University Press), pp. 133–55. Kemmerling, A. (1997) Der bedeutungstheoretisch springende Punkt sprachlicher Verständigung. In G.-L. Lueken (ed.) Kommunikationsversuche (Leipzig: Leipziger Universitätsverlag), pp. 60–106. Kemmerling, A. (2001) Gricy actions. In G. Cosenza (ed.) Paul Grice’s Heritage (Brepols: Turnhout), pp. 73–90. Marcu, D. (2005) Perlocutions: the Achilles’ heel of speech acts. Ms. Meggle, G. (1981) Grundbegriffe der Kommunikation (Berlin and New York: Walter de Gruyter). Petrus, K. (1999) Kommunikation als Kooperation. In M. Flügel et al. (eds.) Werte und Fakten (Stuttgart: Haupt Verlag), pp. 213–34. Petrus, K. (2003) À la Austin & à la Grice. In A. Bächli and K. Petrus (eds.) Monism (Frankfurt a. M. and London: Ontos-Verlag), pp. 315–32. Petrus, K. (2005) Illokution und Konvention, oder auch: Was steckt nun wirklich hinter Austins ‘securing of uptake’? Grazer Philosophische Studien, 70: 101–26. Petrus, K. (2006) Illokutionäre Akte. Alte Begriffe, ein neuer Weg. Linguistische Berichte, 206: 117–46. Petrus, K. (2010) Strawson, Austin & Grice, to appear in S.-J. Conrad and S. Imhof (eds.) Peter F. Strawson – Ding und Begriff (Frankfurt and London: Ontos-Verlag). Schiffer, S. (1972) Meaning (Oxford: Oxford University Press). Schlieben-Lange, B. (1976) Perlokution und Konvention. In K. Gloy and G. Presch (eds.) Sprachnormen III (Stuttgart: Frommann-Holzboog), pp. 58–66. Searle, J. R. (1969) Speech Acts (Cambridge: Cambridge University Press).
234 Klaus Petrus Siebel, M. (2002) Was sind illokutionäre Akte? In M. Siebel (ed.), Kommunikatives Verstehen (Leipzig: Leipziger Universitätsverlag), pp. 138–64. Strawson, P. F. (1964) Intention and convention in speech acts. In P. F. Strawson, Logico-Linguistic Papers (Hampshire: Ashgate), pp. 115–29. Ulkan, M. (1992) Zur Klassifikation von Sprechakten (Tübingen: Niemeyer).
11 Speaker-Meaning and the Logic of Communicative Acts Christian Plunze
1. Introduction It seems that someone, given appropriate circumstances, can communicate almost anything by almost any kind of doing. Of course, usually we communicate by means of language, but even the moving of furniture can be a communicative act (as used in this chapter, the expression ‘communicative act’ applies exclusively to concrete acts performed by someone at a particular time and at a particular place). Since the concept of a communicative act (or: the act type to communicate-something-to-someone) involves no reference to a particular type of doing, it seems that the only possibility to fix the concept of a communicative act is to focus on the effects brought about by the respective doings. If one restricts oneself, as I will do here, to cases where these effects have been intended, the concept of an intentional communicative act can be defined via the notion of a communicative intention. Let me say that an agent X communicates (intentionally) to an addressee Y that A only if X, by her doing (like an utterance or a gesture), intends to bring it about that it is communicated to Y that A. Given this condition, a first step to define the notion of a communicative act is to define the notion of a communicative intention without reference to the notion of communication. The way to do the latter is to characterize the goal of a communicative intention (i.e. the fact that X intends to bring about) without reference to the notion of communication – for example, by adopting the view that it is communicated to Y that A just if Y acquires the belief that A. Before I consider the question what else besides the presence of a communicative intention must be the case in order for a doing to be a communicative act, let me point out an outstanding feature of communicative acts, namely that the recognition of a communicative intention is already sufficient for communication. This is shown, for example, by the following situation: (S1)
Johanna (X), waiting at the airport in Lisbon for the departure of her flight to Frankfurt, wants to inform her husband (Y) about the time 235
236
Christian Plunze
of her arrival. Hence, she sends him the following text message on his mobile phone: ‘I will arrive at midnight’. – However, since her husband has been told that the airport in Frankfurt will be closed soon (due to bad weather), he does not believe her. Here it would be absurd for Johanna’s husband to say something like ‘She didn’t communicate to me that she will arrive at midnight because I didn’t believe her’. The opposite is correct: although he didn’t believe her, she has communicated to him the time of her arrival. Sufficient for this is, so it seems, his recognition of her communicative intention. Now the claim that the recognition of a communicative intention implies communication allows for two interpretations. The weak one can be expressed as follows: (CW ) By doing α, X communicates to Y that A, if Y recognizes that X intends to bring it about that it is communicated to Y that A. According to the strong interpretation it is even appropriate to speak of successful communication if the communicative intention is recognized. According to this view it holds: (CS)
By doing α, X communicates successfully to Y that A, if Y recognizes that X intends to bring it about that it is communicated to Y that A.
Usually this stronger interpretation is endorsed in the literature. For example, with respect to linguistic communication, Kent Bach writes: Communicative success is achieved if the speaker chooses his words in such a way that the hearer will, under the circumstance of the utterance, recognize his communicative intention. (Bach 1994: 9) Likewise, Searle: Human communication has some extraordinary properties not shared by most other kinds of human behaviour. One of the most extraordinary is this: If I am trying to tell someone something, then (assuming certain conditions are satisfied) as soon as he recognizes that I am trying to tell him something and exactly what it is I am trying to tell him, I have succeed in telling it to him ... In the case of illocutionary acts, we succeed in doing what we are trying to do by getting our audience to recognize what we are trying to do. (Searle 1969: 47) And with respect to a situation like (S1) Sperber and Wilson claim that: although [X] has failed to convince [Y] she has succeeded in communicating to him what she meant. (Sperber and Wilson 1986: 29)
Speaker-Meaning and the Logic of Communicative Acts
237
This strong interpretation needs, on my view, further justification. First of all, not everybody will share the view that someone communicates already successfully if the respective communicative intention is recognized by the addressee. After all, something has gone wrong in (S1): Johanna’s husband does not believe that she will arrive at midnight but surely, by her message, Johanna intended to bring about exactly this belief. Hence, this intention of hers is not fulfilled and this could be taken as a reason to say that Johanna has not communicated successfully to her husband that she will arrive at midnight (although she has communicated successfully other things; see section 3, below). Moreover, the distinction between successful and unsuccessful communicative acts has to be fixed by reference to the distinction between a fulfilled and an unfulfilled communicative intention in the following way: By doing α, X communicates successfully to Y that A:= (1) By doing α, X intends to bring it about that it is communicated to Y that A. (2) X’s doing α brings it about (in the manner expected by X) that it is communicated to Y that A. By doing α, X communicates unsuccessfully to Y that A:= (1) By doing α, X intends to bring it about that it is communicated to Y that A. (2) It is not the case that X’s doing α brings it about (in the manner expected by X) that it is communicated to Y that A. Now, given this characterization of the distinction between successful and unsuccessful communication, to assume that (CS) is true is to assume that the recognition of a communicative intention somehow leads to its fulfilment. But how should it be possible that a communicative intention is fulfilled if it is recognized by an addressee? In the next section I will discuss briefly several attempts to explain (CS). On my view, none of them is satisfactory. I take this as an indication that (CS) is too strong. Nevertheless, the weaker interpretation (CW) seems true. Therefore, I will put forward a different account of communicative acts which allows one to explain why (CW) holds. In the final section, I will consider a potential objection against the account of communicative act favoured here.
2. Speaker-meaning and communication Many authors share Strawson’s view that Paul Grice’s analysis of speakermeaning was: undoubtedly offered as an analysis of a situation in which one person is trying, in a sense of the word ‘communicate’ fundamental to any theory of meaning, to communicate with another. (Strawson 1971: 156)
238 Christian Plunze
According to Grice’s initial proposal, it holds that someone means something if he or she intends ‘to produce some effect in a audience by means of the recognition of this intention’. Because of this connection between the recognition of an intention and the production of certain effects, it is assumed usually that Grice’s analysis of speaker-meaning – or a suitable modification of it – somehow delivers an explanation for (C S). In this section I scrutinize how much truth lies in this assumption. In doing so, I will focus exclusively on cases of speaker-meaning where the intended effect is a belief on the part of the addressee. For these cases, Grice’s analysis looks (slightly modified) as follows:1 By doing α, X means that A if and only if there is a Y such that the following holds: (1) By doing α, X intends to bring it about that Y believes that A (2) By doing α, X intends to bring it about that Y recognizes that (1) (3) X believes that Y will believe that A on the basis of his recognition of X’s intention to bring it about that X believes that A According to Strawson, cases of speaker-meaning are attempts to communicate in a special sense of the word ‘communicate’. Hence, there is no reason to think that Strawson wants to deny that an act can be an attempt to communicate even if the agent does not mean something by his or her doing. According to him (and most of his followers), X tries or intends to communicate that A if he or she means that A, but the converse does not hold. Now, since someone’s doing is an attempt to communicate if (and only if) the agent does something with a communicative intention, one may wonder which intention mentioned in Grice’s analysis is supposed to be the communicative one. The natural choice is the intention mentioned in the first clause of Grice’s analysis. After all, the goal of this intention determines what S means, and it is usually assumed that X communicates that A if he means that A. However, this choice is obviously incompatible with the assumption that (CS) is true. Clearly, it is possible that Y recognizes that X intends to bring it about that Y believes that A without believing what he or she should believe. Normally, this will be the case if Y believes that X is insincere or if Y believes that X errs by believing that A (as is the case in situation (S1)). So, Grice’s initial analysis does not deliver an explanation for (C S). In order to justify (C S) by reference to the notion of speaker-meaning, two basic types of proposals have been put forward. Some authors assume that the perlocutionary intention mentioned in the first clause of Grice’s analysis should be replaced by a so-called ‘illocutionary intention’ (Bach 1994; Recanati 1987). It is held that (C S) comes out as true if one modifies Grice’s proposal along these lines. Others think that it is a mistake to identify the communicative intention with this perlocutionary intention. Instead, a communicative intention is taken as a second order intention with the goal
Speaker-Meaning and the Logic of Communicative Acts
239
that a first order intention is recognized (Sperber and Wilson 1986; Searle 1969). Let me discuss this latter view first. According to this view, cases of speaker-meaning are attempts to communicate because of the presence of the second order intention with the goal that the addressee recognizes the intention of the agent to bring it about that he or she believes that A (or any other first order intention). Accordingly, one can say that X communicates successfully just if his or her second order intention is fulfilled. This holds, for example, in situation (S1). Obviously, Johanna does not only communicate, but also means that she will arrive at midnight. Moreover, her intention to bring it about that her husband believe exactly this is recognized by him. Hence, her second order intention is fulfilled, and therefore one is entitled to say that Johanna has communicated successfully although her husband has not acquired the belief that she will arrive at midnight. But what was communicated successfully? That Johanna intends to bring it about that her husband believes that she will arrive at midnight? Or that she will arrive at midnight? Although the latter is presupposed with respect to situations of this kind (compare, for example, Sperber and Wilson 1986: 29), it is hard to see how this view could be justified. Obviously, the claim that X has communicated successfully to Y that A if his or her first order intention is recognized by Y, is at odds with the quite plausible principle that the communicative content of a communicative act is determined by the goal of the communicative intention. Hence, if it is assumed that communicative intentions are second order intentions, this principle forces one to say that we communicate exclusively that we have certain first order intentions. But this is quite implausible (Johanna does not communicate exclusively that she intends to make her husband believe that she will arrive at midnight. Her primary message is that she will arrive at midnight). Therefore, it seems more appropriate to assume that someone, who means that A, has two communicative intentions (a first and a second order one). However, even then it is still wrong to say with respect to situations like (S1) that X has communicated successfully that A, since the first order intention is not fulfilled. At this point, it is tempting to argue that Johanna has communicated successfully because by recognizing her first order intention her husband already knows what she communicates primarily to him (namely that she will arrive at midnight). But this is not true. A sentence like: (i) Y knows what X communicates to him is ambiguous. In a first reading it means something like ‘Y knows (already) the fact that X communicates to him’. In the second reading it means something like ‘Y knows what it is that X communicates to him’. One can find this ambiguity also in the sentence ‘Phillip knows what Emma suspects’. In a first reading this sentence means that Peter already knows the fact
240 Christian Plunze
that Emma merely suspects – for instance, the fact that Black has stolen some money. In this reading the sentence can be true even if Phillip has no opinion about Emma’s suspicions. In a second reading, ‘Phillip knows what Emma suspects’ means that Phillip knows a certain conjecture of Emma – for instance, her conjecture that Black has stolen some money. In this reading the sentence can be true even if Peter has absolutely no knowledge about Black’s thefts. Obviously, the pertinent reading of (i) is the second one. If Y recognizes the communicative intention, he knows what it is that Y intends to communicate to him. However, this does not imply that Y recognizes the goal of the communicative intention or the content of this intention. Hence, it is false to say that Y knows what X communicates to him if Y recognizes that X has such-and-such communicative intention – as it is false to say that Phillip recognizes that Black has stolen some money if Phillip recognizes that Emma suspects that Black has stolen some money. For these reasons, I do not think that this strategy to justify (CS) by reference to speaker-meaning is successful. Let me now look at proposals which rest on the assumption that (CS) can be justified by replacing the (first order) perlocutionary intention by an illocutionary intention and taking this illocutionary intention as the communicative intention. According to these proposals, the goal of an illocutionary intention is of such an extraordinary kind that (C S) comes out as true. Some advocates of this strategy hold that the fulfilment of (CS) can be secured by the assumption that communicative intentions are self-referential: part of the content of a communicative intention somehow refers to this very intention (Blackburn 1984; Harman 1974). Others think that (C S) is fulfilled because X intends to bring it about that it is communicated to Y that A just if X intends to provide a reason for Y to believe that A (Recanati 1987; similarly, Bach and Harnish 1979). Since a detailed discussion of these proposals is far beyond the scope of this chapter, let me restrict myself to two brief remarks. First, it is doubtful whether (CS) is really fulfilled within the illocutionary frameworks that have been put forward so far. Consider, for example, Recanati’s account of illocutionary intentions. According to him, an intention to bring it about that an addressee has a reason to believe that A is an illocutionary intention, and such an intention is fulfilled if it is recognized. Recanati holds this view for two reasons. On the one hand, it could be that such an intention is fulfilled although the addressee does not believe that A, because one can have a reason to believe that A without actually believing that A. On the other hand, the recognition of an intention to bring it about that an addressee believes that A provides – due to a ‘general presumption of truthfulness’ – always a reason for an addressee to believe that A. According to Recanati this holds even if in the given circumstances this reason is overruled by stronger reasons to believe the opposite – for example, by strong reasons to believe that X is lying. Therefore, Recanati concludes that
Speaker-Meaning and the Logic of Communicative Acts
241
the recognition of a such an intention ‘cannot but give a reason to believe’ something. But this is hard to believe. Consider the following situation: (S2)
Y, a police officer, who has stopped X’s car, asks whether X has drunk any alcohol. X denies this – but his breath reveals that this is not true.
Does X’s utterance in (S2) provide Y with a reason to believe that X is sober? I don’t think so. Moreover, I think the same holds in all situations in which the ‘general presumption of truthfulness’ is overruled. The reason seems to be that the assumption that the speaker is saying something that is true is the only reason an addressee can have for believing what was said. Hence, nothing is left over if this assumption is overruled, and we get a situation in which an illocutionary intention is not fulfilled although it is recognized. A second weakness of the illocutionary proposal is that it is also doubtful that someone intends to communicate that A if he or she intends to bring it about that an addressee has merely a reason to believe that A. Suppose, by saying ‘Probably, Smith has stolen the money’, John has expressed his conjecture that Smith is a thief. Given that John thinks that his addressee will believe that John is sincere and have reasons for his conjecture, it is plausible to assume that he intends to provide his addressee with a reason to believe that Smith is a thief. But it seems false to say that John intends to communicate this. What is communicated is that John conjectures that Smith is a thief. So, it seems that communicative intentions are not as weak as illocutionary intentions. If the considerations so far are roughly correct, there is little reason to assume that Grice’s analysis of speaker-meaning or something similar can deliver an explanation for the supposed feature (CS) of communicative acts. This is, I think, a strong indication that (CS) is simply too strong. It is simply not true that one communicates already successfully if the communicative intention (whatever it is) is recognized. However, as I try to show in the next section, the claim (CW), which captures also situations like (S1), can be justified – even if one assumes that communicative intentions are perlocutionary intentions.
3. Presence, success and failure of communicative acts Common to all proposals discussed so far is the assumption that X communicates successfully only if his or her pertinent communicative intention is fulfilled. There is no reason to doubt that this is true. However, as far as I can see, all authors suppose furthermore that X does not communicate at all if the pertinent communicative intention is not fulfilled. But this latter claim would be true only if the presence of a certain intention (or intentions) cannot be a sufficient condition for the exemplification of a corresponding act
242 Christian Plunze
type. But this assumption is false: in some cases a certain doing is a token of a certain act type just because the agent does what he or she does with a certain intention. As I will argue, the same holds for communicative acts: A doing is a communicative act just because the agent does what he or she does with a communicative intention. Consider the following two groups of sentences that are normally used to state the performance of an intentional act of a certain type:2 GROUP 1
GROUP 2
X opens a window X boils water X deceives Y X catches a fish X kills Y
X asserts that A X asks Y whether A is the case X lies X is fishing X searches for her sunglasses
A common feature of the acts in both groups is that someone performs such an act by doing something more basic: one kills someone by poisoning a Martini or by shooting a gun; and one asks someone something by uttering a sentence or making a gesture. However, there is an obvious difference between the act types of the first and the act types of the second group. For act types in the first group it holds that X’s doing must bring about a certain effect (or effects) in order to perform an act of this type at all. If a poisoning of a Martini does not bring it about that someone dies, the poisoning of the Martini is not a killing. Henceforth, I will call the effect (or the effects) that must obtain in order to perform an act of a certain type β the defining result (or results) of this act type. Accordingly, I will call act types of the first group result-defined act types. For these act types it holds: (R)
By doing α, X does something of type β if and only if there is an effect A such that the following holds: X’s doing α brings it about that A. (As before, I assume both that X does α intentionally and that X’s doing α does not analytically imply that X’s doing α is also a doing of type β.)
Act types of the second group I will call goal-defined act types. In order to perform a goal-defined act type it is not necessary that X’s doing bring about a certain result. Necessary and sufficient is just that the agent intends to bring about a certain effect. For example, X, by looking into the drawer, searches for her sunglasses if and only if X does what she does with the intention to bring it about that she finds her sunglasses. Obviously, her looking in the drawer can be a search for her sunglasses even if she does not find them. For goal-defined act types it holds: (G) By doing α, X does something of type β if and only if there is an effect A such that the following holds: By doing α, X intends to bring it about that A.
Speaker-Meaning and the Logic of Communicative Acts
243
On the basis of (G) the distinction between a successful and an unsuccessful performance of any goal-defined act type can be characterized as follows: (G-1) By doing α, X does something of type β successfully:= (1) By doing α, X intends to bring about that A. (2) X’s doing α brings it about that A. (G-2) By doing α, X does something of type β unsuccessfully:= (1) By doing α, X intends to bring about that A. (2) It is not the case that X’s doing α brings it about that A. In contrast to this, there is no reasonable distinction between a successful and an unsuccessful performance of a result-defined act type. Consider, for example, the difference between (1) and (2) on the one side, and (3) and (4) on the other side: (1) (2) (3) (4)
By uttering the sentence S, X asks Y successfully whether A. By uttering the sentence S, X asks Y unsuccessfully whether A. By poisoning the Martini, X kills Y successfully. By poisoning the Martini, X kills Y unsuccessfully.
Sentences (3) and (4) are – in contrast to (1) and (2) – hard to understand. This can be explained by the difference between goal- and result- defined act types. Given that the act type to-ask-a-question is defined by the goal that the agent intends to bring about (probably, the goal that the addressee answers), the first sentence states simply that the pertinent goal has been realized whereas the second sentence denies this. The presupposition involved in both cases – namely that X has asked a question – can be explained by the assumption that it is presupposed that X has the relevant intentions. The oddity of the last two sentences seems due to the fact that the act type to-kill-someone is a result-defined act type. Since it is possible to bring it about unintentionally that someone dies, it is possible to interpret the third sentence as a somewhat awkward way to express that the killing was intended. Hence, this sentence is not as bad as the last one. It is almost impossible to understand (4) because one presupposes (due to the use of ‘kills’) that a killing has occurred. But because of this it is unclear what could be meant by the qualification ‘unsuccessfully’. Surely not ‘unintentionally’ – an utterance of ‘By poisoning the Martini, X kills Y unintentionally’ gives no reason to cast doubt on the conceptual competence of the speaker. The only interpretation that remains is that by using ‘unsuccessfully’ the speaker claims that the defining result of the killing has not been realized. But this contradicts the presupposition that a killing has occurred, which explains why this sentence is extremely odd. So, the distinction
244 Christian Plunze
between a successful and an unsuccessful performance of an act of type β is only applicable if the act type is a goal-defined act type. Of course, with respect to result-defined and goal-defined act types it is possible to draw a distinction between successful and unsuccessful attempts to bring about the defining effect. All of the following sentences are fine: (5) X’s poisoning of a Martini is a successful attempt to bring it about that Y dies. (6) X’s poisoning of a Martini is an unsuccessful attempt to bring it about that Y dies. (7) X’s looking into a drawer is a successful attempt to bring it about that she finds her sunglasses. (8) X’s looking into a drawer is an unsuccessful attempt to bring it about that she finds her sunglasses. However, if (6) is the case, X’s poisoning of the Martini is not a killing, but if (8) is the case, X’s looking into the drawer is still a searching. Now, given that the mere presence of a certain intention can be a conceptual sufficient condition for the presence of a corresponding act of a certain type, one may wonder whether the presence of a communicative intention can be a conceptual sufficient condition for the presence of an act of the type to-communicate-something-to-someone. A first reason to think that this is really the case, is the fact that it is possible to distinguish without oddity between successful and unsuccessful performances of communicative acts: (9) By uttering ‘It is raining’, X has communicated successfully to Y that it is raining. (10) By uttering ‘It is raining’, X has communicated unsuccessfully to Y that it is raining. If communicative acts were tokens of a result-defined act type, one would expect that both sentences are as odd as sentences (3) and (4). The fact that (9) and (10) are fine, supports the view that communicative acts are goaldefined. A second reason for the same conclusion is the observation that only verbs that characterize goal-defined act types can be used in a so-called ‘explicit performative sentence’. It is not possible (for obvious reasons) to use in an explicit performative sentence a verb that characterizes a perlocutionary effect. Consider the contrast between (11) and (12): (11) I hereby assert that it is raining. (12) I hereby convince you that it is raining.
Speaker-Meaning and the Logic of Communicative Acts
245
Significantly, ‘to communicate’ is on a par with ‘to assert’: (13)
I hereby communicate to you that it is raining.
A final reason (probably, the most important one) for assuming that the presence of a communicative intention is sufficient for the presence of a communicative act is that this assumption delivers a straightforward explanation of the above-mentioned feature (CW ) of communicative acts. Henceforth, I will suppose that a communicative intention is an intention to bring it about that someone believes something. Then, communicative acts can be defined as follows: CA1 By doing α, X communicates to Y that A:= By doing α, X intends to bring it about that Y believes that A. On the basis of this definition, we get the following characterization of the distinction between successful and unsuccessful communicative acts: CA 2
By doing α, X communicates successfully to Y that A:= (1) By doing α, X intends to bring it about that Y believes that A. (2) X’s doing α brings it about that Y believes that A.
CA3
By doing α, X communicates unsuccessfully to Y that A:= (1) By doing α, X intends to bring it about that Y believes that A. (2) X’s doing α does not bring it about that Y believes that A.
Now, if we assume that someone who recognizes that Q also recognizes that P if the sentence that expresses Q is analytically equivalent to the sentence that expresses P, it follows from definition CA1 that Y recognizes that X communicates to him or her that A just if he or she recognizes that X intends to bring it about that he or she believes that A. According to this proposal, the recognition of a communicative intention is tantamount to the recognition that X communicates something because X communicates anyway: to recognize a communicative intention is just to recognize that someone already communicates. Hence, (CW ) is just a trivial result of the fact that communicative acts are tokens of a goal-defined act type. This explanation for (CW ) is not mysterious or unusual. Suppose, that Y has recognized that X is looking into a drawer with the intention to bring it about that she finds her sunglasses. Then, it is true that Y has recognized that X searches for her sunglasses. Why? Because the concept of a search involves the concept of a goal-defined act type. Likewise, it is true that Y has recognized that X communicates if Y has recognized X’s communicative intention because the concept of a communicative act involves the concept of a goal-defined act type.
246
Christian Plunze
Let me emphasize two further points. First, according to the account of communicative acts favoured here, the stronger claim (CS) is false since, as argued in the last section, Y can recognize X’s intention to make him believe that A without actually believing what he or she should believe. Second, according to CA1, the presence of a communicative intention is a conceptual-sufficient condition for the presence of a communicative act. Hence, it is strictly speaking false to say that X’s doing is merely an attempt to communicate if it holds both that X does what she does with a communicative intention and that this intention is not fulfilled – as it is false to say that X’s looking into a drawer is merely an attempt to search for her sunglasses if it is true both that X, by looking into the drawer, intends to bring it about that she finds her sunglasses and that this intention is not fulfilled. As far as I can see, the popular opposite view rests basically on a confusion between an unsuccessful performance of a goal-defined act type and the non-performance of a result-defined act type. Of course, it is possible to say something like this: (14)
X’s uttering of ‘It is raining’ is a successful attempt to communicate to Y that it is raining. (15) X’s uttering of ‘It is raining’ is an unsuccessful attempt to communicate to Y that it is raining. But this possibility does not militate against definition CA1, because the distinction between a successful and an unsuccessful attempt to communicate in (14) and (15) should be interpreted as the distinction between a successful and unsuccessful attempt to bring about the defining effect of a communicative act. Hence, (14) and (15) should be paraphrased as follows: (14a)
X’s uttering of ‘It is raining’ is a successful attempt to bring it about that Y believes that it is raining. (15a) X’s uttering of ‘It is raining’ is an unsuccessful attempt to bring it about that Y believes that it is raining. Now, as argued above, the distinction between a successful and an unsuccessful attempt to bring about the defining effect of an act type is applicable to goal-defined act types as well as to result-defined act types. Therefore, the possibility to say something like (14) and (15) does not prove that communicative acts are tokens of a result-defined act type.
4. Understanding According to the proposal favoured here, it could be that X communicates something to Y although X has not the intention that Y recognize his or her communicative intention. This contradicts the commanding view
Speaker-Meaning and the Logic of Communicative Acts
247
according to which it holds that X’s doing is a communicative act only if X intends that the addressee understand what X is doing. Since to understand what someone is doing is to recognize the intentions behind the doing, this assumption is tantamount to the assumption that a doing is a communicative act only if X intends that the addressee recognizes his or her communicative intention. In the remaining part of this chapter, I will defend my rejection of a conceptual connection between communicative acts and such an understanding-intention. The argument for the opposite view I will call the argument from nonunderstanding. Here is the version of this argument given by Wayne Davis: We seek to define communication, which is not the same as attempting to communicate. People often attempt to communicate but fail ... Intentions are not enough for communication, however. While attempting to communicate requires attempting to produce understanding, success in the former attempt requires success in the latter. That is, [X] cannot communicate with [Y] unless [Y] understands [X] ... An American tourist may very well mean the usual when he says ‘I want a cup of coffee’ in a Tibetan restaurant. But he will not communicate his desire unless the waiter understands him. (Davis 2003: 86) First, let me look at how the notion of an ‘attempt to produce understanding’ is involved in Davis’s definition of communicative acts. His definition reads as follows: X communicates M to Y if and only if X does something by which X expresses M and from which Y recognizes that X is expressing M. (Davis 2003: 94) The variable ‘M’ stands for a mental state like a belief or a desire. According to Davis, X expresses a certain mental state M just if X performs an observable act as an (undisguised) indication of the occurrence of M. The latter requires on Davis’s view that X performs her act with the intention that her act is an indication of M. As Davis’s definition reveals, Y’s understanding of X’s communicative act is characterized as the recognition of this ‘expressing intention’. This characterization is in accordance with the above-mentioned assumption that to understand an act (or an attempt to perform an act) is to recognize those intentions whose presence is necessary for the performance of this act (or necessary for the attempt to perform this act). Since Davis thinks that the mere presence of intentions is not enough for communication, the presence of the expressing intention must be assumed by him as a necessary and sufficient condition for the presence of an attempt to communicate. Hence, the expressing intention must be the communicative intention. But X’s intention to express a mental state is obviously
248
Christian Plunze
not X’s intention with the goal that Y understands the communicative act. This would be X’s intention that Y recognize that X intends his or her act as an indication that he or she is in a certain mental state. Since this intention is not included in Davis’s account of attempts to communicate, this account contradicts Davis’s claim that ‘attempting to communicate requires attempting to produce understanding’. So what? Should the definition be supplemented? Or is it false to assume that attempts to communicate are by definition attempts to produce understanding? There are two reasons for the latter view. First, the mismatch in Davis’s proposal between what an addressee has to recognize (namely the expression-intention) in order to understand and what the agent has to intend (namely to express the belief that A) in order for her doing to be an attempt to communicate, is inescapable. If it is assumed that communicative acts are tokens of a result-defined act type, then one has to assume that X’s doing is an attempt to communicate to Y that A if and only if X intends to bring it about that it is communicated to Y that A. But X’s communicative intention cannot be X’s intention that Y understands X’s attempt to communicate because an intention with this goal requires a further intention that should be recognized. But if the ‘understanding-intention’ is identical with the communicative intention there is nothing that could be recognized. Hence, X’s intention that Y understands X’s attempt to communicate can only be the intention that Y recognizes X’s communicative intention. But the presence of the latter intention is a conceptual sufficient condition for the presence of an attempt to communicate. Therefore, the presence of any further intention – in particular, X’s intention that Y recognizes that X has such-and-such communicative intention – cannot be a conceptual necessary condition for the presence of an attempt to communicate. As you may expect, I think, a second shortcoming of the argument from non-understanding is that it is supposed without argument that communicative acts are result-defined acts. Of course, if it is true that ‘attempting to communicate requires attempting to produce understanding’ it is true that ‘success in the former attempt requires success in the latter ’. But from this it does not follow that understanding is a necessary condition for communication. As emphasized above, the fact that Y does not recognize the communicative intention does not show that X does not communicate at all, but only that X does not communicate successfully. I don’t communicate because you don’t understand me? No. You don’t understand me although I communicate. This is not to deny that someone who communicates, in particular someone who communicates by means of language, intends quite often that his or her communicative intention be recognized. There is a good reason for this. If you assume that a communicative intention is an intention to convince someone of something, there is quite often no hope that this intention will be fulfilled if it is not recognized. As emphasized by Grice, if X’s
Speaker-Meaning and the Logic of Communicative Acts
249
doing is not a natural sign for its being the case that A, X will think that the addressee needs another reason for believing that A. Since X usually believes that Y’s recognition of X’s communicative intention in normal circumstances will actually give him a reason to believe that A (because X will believe that Y assumes that X is truthful), X will normally intend that Y recognize this communicative intention. However, this consideration shows only that the recognition of a communicative intention is an appropriate means for the fulfilment of this intention. It does not show that the recognition of a communicative intention is the only means for this purpose. In particular, it could be rational for X to disguise her communicative intention in order to bring about its fulfilment: (S3)
X is a spy who works for the Americans at Germany’s Ministry of Foreign Affairs. X believes both that her American boss Y recently has come to the belief that she is a double agent and that he believes that she will tell him in her reports exactly the opposite from what she thinks is true (her boss does not know that she believes this about him). However, both beliefs of her boss are false. Qua spy, X works only for the Americans. Now, X has been informed that Mr. Schwarz has been elected as the chief of the Federal Intelligence Service. In order to inform her boss about this situation, she sends him the following message: ‘Schwarz has lost the election’.
I see no reason to deny that X communicates here to Y that Schwarz has won the election. Nevertheless, X hides the pertinent communicative intention (namely, the intention to make Y believe that Schwarz has won the election). Probably, X will be successful: Y will assume that she has told him exactly the opposite from what she thinks is true. Hence, Y will conclude that X believes that Schwarz has won the election. Since there is no reason for Y to believe that X errs, he will actually come to the belief that Schwarz has won the election. So, why should one even deny that X has communicated successfully to Y that Schwarz has won the election? Of course, Y probably will believe that X has communicated this unintentionally – but this (false) belief presupposes that he thinks that she has communicated something to him. So, even if one restricts oneself to communication with non-natural signs, there are counterexamples to the claim that such acts involve essentially an understanding-intention. A final remark. According to definition CA1, it could be that X is communicating even if nobody recognizes her communicative intention. This happens in the following version of situation (S1): (S4) Johanna, waiting at the airport in Lisbon ... – Unfortunately, due to a momentary breakdown of the telephone network the message is not brought forward. Hence, Johanna’s husband does not recognize that
250 Christian Plunze
she intends to bring it about that he believes that she will arrive at midnight. It seems false to comment on this situation like this: ‘By uttering “I will arrive at midnight” Johanna has communicated to her husband that she will arrive at midnight although he didn’t recognize that she intends to bring it about that it is communicated to him that she will arrive at midnight.’ I will leave it open whether this statement is false or rather quite misleading. Be that as it may; the impression that this statement is false or misleading depends on the so-called ‘aspect’ of the simple present perfect tense which signals here the successful completion of the act under consideration. (Consider, for example, the contrast between ‘X had walked to the cinema when he was run over’ and ‘X was walking to the cinema when he was run over’.) If we replace the simple present perfect by a tense that does not carry a definite aspect, the sentence becomes fine: ‘By uttering “I will arrive at midnight” Johanna communicates to her husband that she will arrive at midnight although he does not recognize that she intends to bring it about that it is communicated to him that she will arrive at midnight’. And so it seems to me. Quite often we communicate and nobody understands.3
Notes 1. For the rationale for the modification of the third clause see Bennett (1976: chapter 5) 2. The following distinction is made (somewhat differently) also by Ryle (1949) and Kenny (1963). 3. I am grateful to Fabian Seitz for helpful comments and to the German Science Foundation for financial support.
References Bach, K. (1994) Meaning, speech acts and communication. In R. Harnish (ed.) Basic Topics in the Philosophy of Language (Englewood Cliffs, NJ: Prentice Hall). Bach, K. and Harnish, R. M. (1979) Linguistic Communication and Speech Acts (Cambridge, MA: MIT Press). Bennett, J. (1976) Linguistic Behaviour (Cambridge, MA: Hackett). Blackburn, S. (1984) Spreading the Word: Groundings in the Philosophy of Language (Oxford: Oxford University Press). Davis, W. (2003) Meaning, Expression, and Thought (Cambridge: Cambridge University Press).
Speaker-Meaning and the Logic of Communicative Acts
251
Grice, H. P. (1957) Meaning. The Philosophical Review, 66: 377–88. Harman, G. (1974) Review of Stephen Schiffer’s ‘Meaning’. Journal of Philosophy, 71: 224–9. Kenny, A. (1963) Action, Emotion and Will. [2003 2nd edn] (London: Routledge). Plunze, C. (2002) Why do we mean something rather than nothing? In G. Grewendorf and G. Meggle (eds.) Speech Acts, Mind, and Social Reality. Discussions with John R. Searle (Dordrecht: Reidel). Recanti, F. (1987) Meaning and Force. The Pragmatics of Performative Utterances (Cambridge: Cambridge University Press). Ryle, G. (1949) The Concept of Mind (Chicago: Chicago University Press). Searle, J. R. (1969) Speech Acts (Cambridge: Cambridge University Press). Sperber, D. and Wilson, D. (1986) Relevance (Oxford: Blackwell). Strawson, P. F. (1971) Intention and convention in speech acts. In Strawson LogicoLinguistic Papers (London: Ashgate), pp. 115–30.
12 The Total Content of What a Speaker Means Al Martinich
1. Preliminaries In ‘Meaning’ and subsequent papers, H. P. Grice took the idea of what a person meansNN to be the foundation for linguistic phenomena.1 It was to be the basis for semantics, where semantics is understood to be the study of linguistic meaning, that is, the study of what words and sentences mean. Rather than ‘basis’, perhaps, meaningNN should be said to be the ground upon which the foundation of linguistic meaning rests, because meaningNN is broader than the concept of meaning that applies to a speaker meaning something or to the concepts of word- and sentence-meaning. Moreover, Grice was primarily interested in this broader sense of ‘meaning’. This deserves to be mentioned because most philosophers of language continue to think, or at least give the impression that, John Searle’s alleged counterexample to Grice’s analysis of meaningNN is either decisive against Grice’s theory or at least problematic, when in fact it is beside the point. When in Searle’s supposed counterexample to Grice’s analysis, a captured American soldier utters, ‘Kennst du das Land ... ,’ he meansNN that he is a German soldier, just as Grice’s theory predicts. Although the spoken sentence, ‘Kennst du das Land ...’, is homophonic with the soldier’s utterance, that utterance has the occasion meaning of “I am a German soldier” (see Grice 1989: 100–3). Grice’s view was groundbreaking because it provided an alternative to the view of Gottlob Frege and Bertrand Russell, according to whom sentences or words were the basic semantic phenomena. Unfortunately, Donald Davidson’s project of developing a recursive definition for truth for natural languages marked a return to the spirit of Frege and Russell, and, I believe, has engendered confusion about the nature of semantics and the distinction between semantics and pragmatics, among other things (see Carston 2004: 633). One reason this confusion should never have occurred is that Grice and Davidson had very different projects in mind. Grice wanted to illuminate the nature of meaningNN. In contrast, Davidson wanted to eliminate meaning as a semantic concept and replace it with truth, and to show 252
The Total Content of What a Speaker Means
253
how human beings with finite intelligence and time are able to produce and understand an infinite number of sentences. Semantics plays only a subsidiary role in Grice’s theory of the total content of what a speaker meansNN. That theory occupies the central ground of what now goes by the name ‘pragmatics’ (‘conversation as such’, Grice 1989: 24). Instead of trying to prove that Grice’s view is right, I will show how a slight revision and an extension of his view results in a perspicuous view of linguistic communication. I say ‘communication’ advisedly because although Grice explicitly analyses phrases of the form, ‘an utterer U means’ and specifies ‘the total content of what a speaker means’, he would have been better served, I believe, by taking ‘an utterer U communicates’,2 as his primary analysandum and specifying ‘the total content of what a speaker communicates’. Communication has priority over meaningNN because the latter would be pointless if there were no chance of it being understood. That is to say, utterer’s meaning is one of two correlative components of communication, the other being the audience’s understanding. Something similar to meaningNN without communication3 can exist as derivative phenomena only because of meaningNN’s normal place in communication. Taking communication as the basic object of inquiry would have obviated a problem with Grice’s analysis of conversational implicature. When Grice says that a person has implicated something provided that ‘he is presumed to be observing the conversational maxims’ (Grice 1989: 30), the question of who is doing the presuming arises. It is natural to assume that the audience, not the speaker, is doing the presuming. This position makes sense at least insofar as a situation in which the hearers did not think that the conversational maxims were operating would be nonstandard. However, it is also a nonstandard case if the speaker did not himself presume or think that the conversational maxims were in effect. If the speaker did not need to think that the conversational maxims were in effect, then he would sometimes mean something that he was not aware of and would disavow if it were attributed to him. A speaker would conversationally implicate something in virtue of some maxim, even though he did not believe the maxims were in effect. The upshot is that the belief that the maxims are operating is something that both the speaker and the hearer share in paradigmatic cases of linguistic communication. Whether Grice should have identified (linguistic) communication as his primary object of analysis, in order to keep my language consonant with that of him and other commentators I will typically talk about what the speaker meansNN, except that hereafter the subscript ‘NN’ will be suppressed.
2. The total content of what a speaker means My main purpose here is to give a better account of the idea of the content (Grice: ‘signification’) of what a speaker means. The account is largely an
254 Al Martinich
extension of Grice’s view and in that way also a defence of it. Let’s begin with his understanding of the total content of what a speaker means: what S says conventionally
what a speaker S means what S implicates
nonconventionally (conversationally)
The distinction between saying and implicating or implying is not obviously a proper one. The natural antithesis to ‘imply’ is ‘express’, and it is not obvious why saying something should be the only way of expressing something. Verb tenses and the connotations of some words express times and attitudes, respectively, without saying what time or attitude is being expressed. To say-that4 ‘Ava is going to school’ is not to say-that ‘Ava is now going to school’, and to say-that ‘Beth is a police officer’ is not to say-that ‘Beth is a copper’ (‘copper’ being a slightly dismissive term for a police officer). In order for something to be linguistically explicit, there must be some syntactic or phonological feature that makes it so. That means that if in a report some unarticulated element is added to the words actually used, that addition is in fact not part of what was expressed. This obvious point needs to be emphasized because some otherwise careful philosophers and linguists say that some things are explicit that clearly are not. According to Robyn Carston, if a person says, ‘She has a brain’, or ‘Sam left Jane and she became very depressed’, then the expressed propositions are something like: She has a high-functioning brain and Sam left Jane and as a result she became very depressed, respectively (Carston 2004: 639; see also 641 and 643; also Carston 1991: 33b-4a).5 However, it is as clear as the words on this page that the italicized words do not represent something expressed by the original sentences or the speakers even though they represent part of what the speaker meant. In addition to distinguishing between what is explicit and implicit in what the speaker means, it also makes sense to eliminate Grice’s concept of conventional implicature, as Kent Bach has argued (Bach 1999).6 The words ‘but’ and ‘although’ indicate7 and thus express that some contrast or change in the direction of the discourse is occurring without saying so. Although Grice maintains that conventional implicatures never contribute to the truth-conditions of a sentence, it seems to me that some of them do,
The Total Content of What a Speaker Means
255
for example, the alleged conventional implicature associated with ‘therefore’. The statement made by ‘He is an Englishman; he is, therefore, brave’, is surely false if the implication from the first clause (not necessarily logical implication) to the second clause is not valid or in some way justified (see Grice 1989: 121). The word ‘therefore’ is not unique. The statement, ‘Ava is rich; contrariwise, so is Beth’, is strictly false even if both Ava and Beth are rich because the statement involves no contrariety. The phenomenon that Grice described as conventional implicature is not part of what is said, as he correctly believed, but is not part of what is implied either. It is part of what a speaker indicates. Finally, a speaker means many of the things that are logically implied by what he expresses in virtue of the meaning of the words used. So to the Gricean schema, we should add a branch that shows that in addition to conversational implication, a speaker semantically implies some things. So there is good reason to revise the Gricean schema as follows.
says what S expresses indicates what a speaker S means semantically what S implicates
nonsemantically (e.g., conversationally)
Two caveats need to be added. One is that in normal reports of what a speaker says, the elements that contribute to what is indicated are typically included. So, if a speaker said, ‘Ava is happy but she is a scoundrel’, one may say that the speaker said that Ava is happy but a scoundrel, and the time indicated by the tense of ‘is’ and the negative connotation of ‘scoundrel’ are not separated out as expressed or indicated. This fact about the idiomatic way of reporting what people express does not undermine the theoretical appropriateness of the concept. The other caveat is that there is good reason for dividing what a speaker expresses into two parts, one that takes what a speaker says and indicates as a single unit, and what the speaker ‘meta-says’ as the other part. A speaker often comments on his own words or the attitude expressed by those words, as in the sentences, ‘To digress for a moment, will we have time to visit George and Martha?’ and ‘Frankly, I don’t care what you think’. Bach calls these ‘second-order speech acts’ (Bach 1999: 356–60; see also Potts 2005). Although Grice’s original way of dividing up the total content of what a speaker means can be improved in the ways suggested, for the purposes of this chapter, I will not try to incorporate them in a diagram.
256 Al Martinich
Many philosophers think that the distinction between semantics and pragmatics is the distinction between what a sentence says and what a speaker conversationally implies, because they think that the primary source of what is said is the words themselves. They believe that in addition to the truth of this statement: ‘Snow is white’ is true if and only if snow is white, the following statement is also true, ‘Snow is white’ says that snow is white. They think that what a speaker says is derivative from what a sentence says. It seems to me that they are mistaking an instance of synecdoche for the literal truth. While the applied timeless meaning of words constrains and partially determines what the speaker said, the words themselves, being inert, cannot say anything. Following J. L. Austin, I believe that the latter statement is best construed as shorthand for: A speaker who uses ‘Snow is white’ properly says that snow is white. ‘Properly’ means something like for the purpose for which it is primarily designed. Words and sentences only do things in a derivative or instrumental way. For Grice, the distinction between pragmatics and semantics is the difference between locutions of the form ‘U (utterer) meant that ...’ and locutions of the form ‘X (utterance-type) means “... ” ’ (Grice 1989: 118). As he goes on to say, the difference between specifying meaning through the use of a ‘that’-clause and specifying meaning by using quotation marks ‘is semantically important’ (Grice 1989: 118). If an utterer meant-that something by uttering some words, then the utterer at least tried to say something (to perform some speech act) and the ‘that’-clause is required; but if an utterance means something, then it is not necessarily the case that anything was literally said and its meaning may be given by providing a synonymous expression. In short, both saying and implicating are properly and immediately part of pragmatics (see Carston 1991: 39b and 46a, and 2004: 633). Semantics gets introduced into the analysis of saying (saying-that) because that concept is partially defined in terms of the timeless meanings of words.
3. What a speaker says and speech act pluralism The adequacy of Grice’s schema of the total signification or content of what a speaker means depends on the coherence of the concept of what a speaker
The Total Content of What a Speaker Means
257
says. This concept has been criticized by Herman Cappelen and Ernie Lepore on the grounds that it is so unregimented as to have no theoretical use. They assert: (SPAP)
No one thing is said (or asserted, or claimed, or ...) by any utterance: rather indefinitely many propositions are said, asserted, claimed, stated. What is said ... depends on a wide range of facts other than the proposition semantically expressed. (Cappelen and Lepore 2003: 4; see also their pp. 194 and 199–206.)
When they say that an ‘utterance’ says many things, they mean that the speaker says many things.8 They call their view ‘speech act pluralism’ in contrast with speech act monism, the view that a speaker says just one thing ceteris paribus (see Grice 1989: 89). They also hold semantic monism, the view that a simple sentence says only one thing, on the assumption that the words are univocal and the syntactic structure is unique. So they hold that the sentence, ‘Liddy is unreliable’ says only that Liddy is unreliable even though a speaker in uttering this sentence may say that the speaker does not like Liddy or that Liddy should not be used for a certain task, as well as that Liddy is unreliable. Included in the ‘wide range of facts other than the proposition semantically expressed’ that determine what a speaker says are the knowledge, attitudes and purposes of the speaker and hearer and changes in the world (Cappelen and Lepore 2005: 194). For example, suppose Sarah says, ‘The end table is filled with books’. One straightforward way to report what she said is to say that (1)
Sarah said that the end table is filled with books.
However, if the hearer does not know the meaning of the word ‘table’, the reporter might also truly and correctly say (2)
Sarah said that the piece of furniture next to the sofa is filled with books.
Or suppose that the end table has been moved into a closet. Then what she said can be reported truly and correctly by saying that (3)
Sarah said that the table in the closet is filled with books.
Combining the hearer’s lack of knowledge with the change in the world, what Sarah said might be (4)
Sarah said that the piece of furniture in the closet is filled with books.
258
Al Martinich
Cappelen and Lepore maintain (i) that each occurrence of ‘Sarah said’ in (1)–(4) is a strictly correct way of describing what Sarah did, and (ii) that each variation in the form of the reports of (1)–(4) describes something different. They can be challenged on both points. Concerning (i), since the concept of what the speaker said is a technical one, ordinary uses of the phrase ‘S said’ need not indicate an instance of what the speaker said in the technical sense. The same is true of the ordinary and technical senses of ‘zombie’. In philosophy, a zombie is a hypothetical creature that is indistinguishable from a normal human being but lacks conscious experience. This technical concept is not undermined by ordinary truths of 1950s horror films, that zombies walk as if they are in a trance, that they always have a vacant stare, and that they were once human. Concerning (ii), multiple descriptions of an action need not indicate multiple actions. According to perhaps the dominant view of actions, the movement of a finger may be identical with the flipping of a switch, with the setting off of a bomb, with the killing of twenty people, with the beginning of a war, and so on. There are many descriptions of just one action. Similarly, the content of an utterance may be describable or reportable in many ways without there being more than one content. The appearance of multiple contents for an utterance dissipates if the content is reported in a regimented way. Reporting what a speaker said could be restricted to using the speaker’s own words, except for obvious changes required for demonstratives, indexicals and similar linguistic elements. Alternatively, what the speaker said could be represented by ordered n-tuples in which each member would be an ordered pair, consisting of a way of designating a referent or predication, plus the actual word or words used. So, (1) and (2) would be represented respectively by (R1) and (R2): (R1)
<, > (R2) << the piece of furniture next to the sofa, ‘the end table’>, >. The extension to (3) and (4) is obvious. Cappelen and Lepore think that their ‘speech act pluralism’ is justified by their being ‘naïve about speech act content’ in the sense that they take ‘nontheoretic beliefs and intuitions about what speakers say, assert, claim, etc. at face value’. They claim that ‘What speakers say ... is what we hear when we listen to them ... [I]t is not something requiring an investigation to uncover’ (Cappelen and Lepore 2005: 191). This is their characterization of what speakers say. But their naïve attitude does not count as an objection to speech act theorists having a different, more regimented characterization. The ordinary or intuitive concept of what a speaker says is the starting point
The Total Content of What a Speaker Means
259
for a more precise technical one. A large part of the terminology of the philosophy of language is like this: ‘statement’ and ‘proposition’ have technical definitions that make their ordinary meanings more precise. Why should ‘what speakers say’ be treated differently? Their justification for treating speech act content differently is that the content ‘isn’t deeply hidden somewhere, where only a theorist can excavate it’ (Cappelen and Lepore 2005: 191). By parity of reasoning, they should hold the same thing about semantic content. Here is a consequence of being naïve about semantics. Suppose that Ava is reading a novel in which one character asks, ‘Are we going to the opera tonight?’ Another character replies, ‘Is Paris Hilton a nun?’ Ava does not understand the reply and asks her friend Beth what it means. That is, Ava in effect asks, ‘What does “Is Paris Hilton a nun?” mean?’ Beth answers, ‘It means that they are not going to the opera’.9 So, on a naïve view of semantics, the semantic content of ‘Is Paris Hilton a nun?’ is that we are not going to the opera tonight. Or suppose Ava is reading Macbeth and asks her teacher, ‘What do these lines mean: “life’s ... a tale / Told by an idiot, full of sound and fury, / Signifying nothing”?’ Her teacher answers, ‘They mean that life is not worth living’. So, according to the naïve view of semantics, Macbeth’s words mean that life is not worth living. Cappelen and Lepore would surely not accept this data as evidence that one can be naïve about semantics and that often there is no one thing that a sentence means. They would distinguish, as Ava and Beth may not, between what the sentences mean and what the speaker meant; and they would restrict what the sentences mean to what some theory of meaning dictates. But to make this distinction and to restrict the boundaries of meaning to those dictated by some theory is to abandon naiveté. By the same token, they are not justified in remaining naïve about pragmatics simply because the relevant pragmatic distinctions are not observed in nonphilosophical discourse. Perhaps Cappelen and Lepore would say that they are naïve about pragmatics but not semantics because intuitions about semantics are overridden by ‘important theoretical considerations’. However, there are important theoretical considerations for pragmatics as well. As Austin (1965: 94–101) made clear, there are at least three different senses of ‘saying’ as applied to what speakers do, distinctions that are often blurred or not recognized in ordinary usage. So a technical sense of ‘saying’ is needed by pragmatics. Further, since there is a difference between what speakers say and what they imply when they ask, ‘Can you pass the salt?’ in normal circumstances, some sharpening of the concepts of saying and implying or implicating is needed. This need is not mitigated even if a report of what a speaker (who uttered ‘Can you pass the salt’) said can be reported as: She said that you should pass the salt. (Alternatively: She requested that you pass the salt.)
260 Al Martinich
Moreover, Cappelen and Lepore’s criterion for determining semantic content provides a good starting point for developing a criterion for determining what a speaker said: what a speaker says is the minimal proposition expressed by the sentence he uses. Alternatively, a speaker says what the meaning of his words is designed to enable him to say. Grice’s own tentative characterization of what is said was along these lines (Grice 1989: 25). Pace Cappelen and Lepore (2003: 201), I think ‘Original Utterance Centrism’ is true, and their objection to it is handled by having a standard form of representing what is said, such as that suggested by (R1) and (R2). Their claim that repeating a speaker’s own words may ‘grossly misrepresent what he said’ in certain contexts is true, but does not settle the issue of whether the speaker said one or more things or how what the speaker said should be represented in a theory of speech acts, for the reasons given above (Cappelen and Lepore 2005: 201). In short, they seem to be confusing the form of the words used to say something with what is said. It is no good to object that so many different definitions or characterizations of what a speaker says have been given that there is no hope that a formal theory could be developed, because dozens of philosophical terms have the same history of competing definitions. Since scepticism about these terms is not justified, neither is scepticism about the concept of what is said. Cappelen and Lepore try to undermine the usefulness of a theoretical concept of what a speaker says by giving examples, such as the ‘Smoking Gun’ utterance, from the Watergate Tapes of the Richard Nixon Administration (2003: 191–4). Because Nixon’s words are largely incoherent, they deny that there is any one thing that the speaker says. However, by the same token, Nixon’s words do not say anything. Just as that fact need not cast doubt on the concept of words saying something, it does not cast doubt on the concept of a speaker saying something. Nixon’s actual incoherent words and ungrammatical sentences could be converted to grammatical sentences in several ways (depending on assumptions about the context).10 If the unregulated multiplicity of what is said is the result of the unregulated multiplicity of the possible sentences meant, the concept of what a speaker says is no worse off than the idea of a grammatical sentence. Since Cappelen and Lepore think the latter concept is unproblematic, they cannot justifiably use their argument against the concept of what a speaker says. In another example from the Watergate Tapes, Cappelen and Lepore quote this passage. PRESIDENT NIXON:
He [Hunt] didn’t know how it was going to be handled though, with Dahlberg and the Texans and so forth? Well who is the asshole that did? Is it Liddy? Is that the fellow? He must be a little nuts. He is. H ALDEMAN: PRESIDENT NIXON: I mean, he isn’t well screwed on, is he? Isn’t that the problem?
The Total Content of What a Speaker Means
261
They then present the reader with numerous questions, such as these: Does Nixon say that Liddy is clinically insane, that Liddy has a low IQ, that Liddy handled the situation in an irresponsible way, that Liddy mishandled the situation because he is stupid, and that Nixon blames Liddy? They seem to think that these questions have no clear answer, when in fact they do. According to one version of speech act monism, the view espoused by Grice and me, the answer to each of them is ‘No’. Cappelen and Lepore may respond that if Liddy objected to Nixon’s remarks, it would be absurd for Nixon to defend himself by saying that he didn’t, strictly speaking, say that Liddy was insane or irresponsible or stupid. Indeed, it would be absurd for Nixon to defend himself in this way, not because he did say that Liddy was insane or irresponsible or stupid, but because it is obvious that Nixon did imply them. So Nixon’s ‘did-not-say’ defence would immediately collapse for being irrelevant, and would bring further attention to the fact that he was implying that Liddy was insane, irresponsible or stupid. Before moving on to a very different attack on Grice’s concept of what a speaker says, I want to discuss what appears to be a problem for Cappelen and Lepore’s view, because it reveals the need for a theoretical concept of what is said. According to them, the concept of what a speaker says is unregimented and consequently that any speaker says an innumerably large number of things, even with a simple sentence. Now consider this scenario: Ava says to Beth, ‘Carol is my best friend’. It is later discovered that Carol murdered Ava’s father. Beth then reports, ‘Ava said that her father’s murderer is her best friend’. It seems that Cappelen and Lepore are committed to maintaining that Beth’s report is true because their own theory seems to rely on substituting coreferential singular terms for each other and they think there is no principled way of restricting the substitutions. In other words, for them ‘S says that’ does not create an oblique context. However, Cappelen and Lepore (in conversation) deny that Beth’s statement is true. According to them, if people do not accept that Ava said that Beth’s statement is true in those circumstances, then it is not true that Ava said that her father’s murderer is her best friend. But the truth of what Beth said is determined by the facts and not by what people accept or believe the facts to be, except in special circumstances. Since those facts do not include the actual words Ava uttered or their sense, it would seem that the relevant facts are whether Carol killed Ava’s father and whether Carol is identical with Ava’s best friend (and she is). The point here is that since Cappelen and Lepore want some way to block substitutions of co-referring terms, they need some principles to do this and those principles will yield the resources for a theoretical distinction between what is said and what is not. To respond that no such explanation is possible since the concept of what a speaker says is unregimented, is to beg the question.
262
Al Martinich
4. What a speaker says and conversational implication A very different attack on the concept of the content of what a speaker means is that the concept of what a speaker says is often otiose. For example, suppose that in answer to the question of who is going to Switzerland, a speaker says: (5) Franz is going. By (5) the speaker supposedly implicates: (6) Franz is going to Switzerland. Since (6) entails (5), (5) seems to be superfluous. Since (5) represents what is said according to Grice’s view, the concept of what is said seems to be superfluous in a theory of conversation. Here’s a slightly more complicated example. Suppose that Ava and Bob need to meet someone for an appointment in a park in five minutes. Bob suggests that they walk. Ava replies: (7)
The park is far from where I live.
It is plausible that Ava has implied: (8)
The park is so far from where I live that we do not have time to walk to it.
Again, since (8) implies (7), (7) seems to be conversationally otiose; and since (7) specifies what Ava said, and (8) is a specification of what she conversationally implicated, what Ava said is otiose (Carston 1991: 39b–40a). I think that these arguments are weak because they depend on a paucity of resources in English and other natural languages for reporting what is implied.11 For many speech acts, it is not immediately obvious how to express what precisely is implicated without also mentioning what was said. For the Switzerland example, it’s not grammatical to say: (9)
*The speaker implied (or implicated) to Switzerland.
But this grammatical fact about the resources of some natural languages does not warrant elimination of the concept of what the speaker said. For one thing, with some circumlocution, the implied content can be reported without including any, or at least much, of what the speaker said. Concerning Switzerland, one could say: (10) The speaker implied (implicated) a reference to Switzerland
The Total Content of What a Speaker Means
263
or the implication could be combined with a report of what the speaker said: (11)
The speaker said that Franz was going and implied (implicated) that the place was Switzerland.
Further, the resources of English, like any natural language, can be supplemented, so that both what is said and what is implied are reported without repetition. One way to do this is to report or represent the total content of what is meant and designate what the speaker said in Roman type and what is supplied by boldface. So (5) is represented as (12): (12) Franz is going to Switzerland. And (7) is represented by (13): (13)
The park is far from where I live and that distance is so great that we do not have time to walk to it.
This kind of artificiality is justified by the technical nature of the subject. (See Bach 1994.) The representational device just recommended also takes care of utterances that are subpropositional. (See Bach 1994 on expansion.) Suppose that in answer to the question, ‘Who is going to Switzerland?’, the answer given is ‘Franz’, and to the question, ‘Where is Franz going?’, the answer is ‘Switzerland’. Then the specifications of what the speaker said are (14) and (15), respectively: (14) Franz is going to Switzerland. (15) Franz is going to Switzerland. This complication of Grice’s theory is needed because the class of utterances to be analysed has been expanded from the relatively complete speech acts that Grice considered to the larger group of incomplete ones. It is unfair to accuse Grice’s theory of being defective when, in the initial stages of theorizing about linguistic communication, he restricted the linguistic phenomena to relatively complete utterances. When the broader phenomena become the data set, what Grice’s theory needs is amendment, not abandonment. It may appear that my revision actually undermines the distinction between what is said and what is implicated, for two related reasons. First, what is represented is not only what is said, but also what is meant but not said. In other words, the category of what is said seems to have dropped out, just as relevance theorists claim it should. Second, none of the elements
264 Al Martinich
represented in boldface is either said or conversationally implied, again just as relevance theorists maintain. However, the appearance of (14) and (15) is deceiving. The category of what is said (or expressed) has not disappeared but is represented by the words in Roman type. Most cases of linguistic communication are, thankfully, not fully explicit, and the missing elements are silently supplied by the audience or the theorist as necessary. In addition to representing what the speaker meant, (14) and (15) can be understood as representing what the speaker said, with two provisos: that the concept of what is said is a technical one, and that the use of boldface trumps the literal meaning of ‘what the speaker says (or expresses)’ and signals that the element being represented is being supplied. This use of boldface is analogous to cancelling a conversational implication. We can also introduce another technical term, ‘what the speaker said-s’, which stands for what the speaker said plus its supplement. The supplement corresponds to what Bach (1994) calls ‘impliciture’. The Gricean diagram of what is meant might then be revised like this:
what S says, plus supplement what a speaker S means semantically what S implies
nonsemantically (conversationally)
5. What a speaker says and minimal propositions We have just indicated that the concept of what a speaker says is theoretically coherent, by solving a number of problems. One more problem related to the concept of what is said needs to be treated. Consider the contents of what is said by five utterances of the brief sentence (16) in circumstances (16a–e): (16) (16a) (16b) (16c) (16d)
The kettle is black It is made of normal aluminum but soot covered; It made of normal aluminum but painted black; It is made of cast iron but glowing from heat; It is made of white enamel on the outside but saturated in black light; (16e) It is made of cast iron with a lot of brown grease stains on the outside. What, if anything, was said? I agree with Cappelen and Lepore (2005: 63–4) in holding that the same thing was said, namely, that the kettle is black. They
The Total Content of What a Speaker Means
265
further hold that it is not necessary for the content of what is said by uttering (16) to be supplemented in order to have a truth value; moreover, given the normal way of talking about truth-conditions, the truth-conditions for (16) do not change from utterance to utterance. It seems to me, however, that in an intuitive sense of ‘truth-conditions’, the conditions under which each utterance of (16) is true are different. Cappelen and Lepore reject this view because they believe that different truth-conditions mean different meanings. So they think that if the truth-conditions of (16) change, the meaning of (16) changes. As I have argued elsewhere, truth-conditions may vary according to the criterion of applicability appropriate to the situation without the meaning of the relevant sentence changing (Martinich and Stroll 2007: 26–31). These criteria of applicability are not semantic. They vary depending upon the interests of the speech participants; they need not be known in order to know the meaning of the sentence; and they may be ad hoc. Cases that are more difficult than utterances of (16) are those that involve utterances of sentences like: (17)
He had enough
which many philosophers consider semantically incomplete because the meanings of the words of (17) are not sufficient to identify the relevant fact that makes it true or false. So Bach thinks that (utterances of) (17) do not express a proposition at all, but what he calls a ‘propositional radical’, a fragment of a proposition that can become a proposition by adding certain elements. Depending upon the context, (utterances of) (17) express propositions when the content is supplemented by the appropriate ‘implicitures’, such as money, food, air, or strength. I am neutral in this dispute because I reject a presupposition that the interested parties accept, namely, that (utterances of) sentences say things in the primary sense. As mentioned in section 1, it is speakers that say things in the primary sense. The presupposition that (utterances of) sentences say things, combined with the thesis that every grammatically complete sentence is also semantically complete and expresses a minimal proposition, causes another problem for Cappelen and Lepore. According to them, every utterance of ‘It is raining’ expresses the same minimal proposition and that proposition is true if and only if it is raining. Since it is always raining in some part of the world, every utterance of ‘It is raining’ is true. It seems to me that by parity of reasoning, an enormous number of sentences are virtually always true, such as, ‘John had breakfast’ and ‘Sarah is ready’, because John certainly had breakfast on some preceding day, and Sarah is ready for something. For me, their commitment to this large class of contingent truisms shows that something is wrong with their view and shows that it is not (utterances of)
266
Al Martinich
sentences that say something in the primary sense. Almost every utterance of ‘It is raining’ (and ‘It is not raining’) is used by speakers to communicate a different proposition, because what is meant is determined by the speaker’s communicative intentions, other contextual features and the meaning of the words used.
Notes 1. MeaningNN (meaning nonnaturally) is not itself linguistic meaningNN; it is broader than linguistic meaningNN, because a person can use gestures, sounds and other devices that are not part of a language to communicate. Linguistic meaning evolves out of the foundational case of a person meaning something by an utterance. See Grice (1989). The word ‘meansNN’ (means nonnaturally) has been explained by me to have the sense of ‘communicatively means’. If semantics is taken to be the study of communicative meaning, then meaningNN is the fundamental semantic phenomenon. However, meaningNN is not necessarily linguistic meaningNN. If semantics is restricted to linguistic meaning, then to be more precise, ‘means’ here has the sense of ‘communicatively means’. 2. Even better would be the more explicit formula, ‘When or by producing an utterance, a person communicates to an audience that p’. 3. I am alluding to some of the problem cases in ‘Utterer’s meaning and intentions’, in Grice (1989: 86–116). 4. I use ‘say-that’ here because I’m talking about words successfully used to perform a speech act. The use of ‘say-that’ avoids confusing it with the sense of ‘say’ when it is used to report a phonetic or phatic act. I disagree with Bach, who rejects Grice’s concept of making-as-if-to-say. See Martinich (1984). When there is little or no chance of confusion, I will use ‘say’. 5. Carston’s claim that ‘pragmatics contributes to the proposition explicitly communicated by an utterance although there is no linguistic element indicating that a contextual value is required’ is contradictory according to the ordinary sense of ‘explicitly’ and my technical use (Carston 2004: 639). 6. A decisive criticism of Grice’s concept of conventional implication is Bach (1999). Christopher Potts (2005) has an impressive treatment of ‘conversational implicature’. However, his concept is substantially different from Grice’s. For example, on Potts’s view, the words ‘therefore’ and ‘but’ do not generate conventional implicatures, even though these are Grice’s paradigmatic examples. For him, nonrestrictive relative clauses, such as the one in ‘My grandmother, who is an octogenarian, rides a bicycle’, and expressives, like ‘damn’ in ‘I have to cut the damn lawn again’ do generate conventional implicatures. Although I think that Potts’s conception of conversational implicature belongs with what is said. 7. Grice sometimes uses ‘indicate’ as I am using it here. He talks about words that have conventional implicatures ‘indicating’ something (Grice 1989: 121). But because he distinguishes between saying and implicating (instead of expressing and implying) he puts ‘indicating’ into a third category, which involves some indirectness or inexplicitness. This use has the odd consequence that a person who indicates where a ball is by pointing at it is not explicitly revealing the location of the ball, but indirectly revealing it. 8. It is not clear to me whether they think that the speaker says many things in virtue of the utterance’s saying many things or whether they think that the utterance says many things in virtue of the speaker’s saying many things.
The Total Content of What a Speaker Means
267
9. It is possible that the question, ‘What does “Was Mozart a musician?” mean?’ is asking about meaning as significance and not literal or semantic meaning. For the sake of the example, I want to deny that it is. 10. Cappelen and Lepore say that ‘our intuitions about what speakers say with their utterances are influenced by, at least, six considerations’ (2003: 193–4). But these same considerations influence what we would take the intended meaningful sentences to be. So parallel to the thesis of speech act pluralism, they should hold (minimalist) semantic pluralism. 11. This is not a sign that natural languages are defective. First, natural language is fine for the purposes it is designed to serve. Second, it can be expanded to serve the theoretical purpose philosophers of language need.
References Austin, J. L. (1965) How to Do Things With Words (New York: Oxford University Press). Bach, K. (1994) Conversational impliciture. Mind and Language, 9: 124–62. Bach, K. (1999) The myth of conventional implicature. Linguistics and Philosophy, 22: 327–66. Cappelen, H. and Lepore, E. (2005) Insensitive Semantics: a Defense of Speech Act Pluralism (Malden, MA: Blackwell Publishing). Carston, R. (1991) Implicature, explicature, and truth-theoretic semantics. In S. Davis (ed.) Pragmatics (New York: Oxford University Press), pp. 33–51. Carston, R. (2004) Relevance theory and the saying/implicating distinction. In L. Horn and G. Ward (eds.) Handbook of Pragmatics (Malden: Blackwell Publishing), pp. 633–56. Grice, H. P. (1989) Studies in the Way of Words (Cambridge, MA: Harvard University Press). Martinich, A. P. (1984) Communication and Reference (Berlin and New York: Walter de Gruyter). Martinich, A. P. (2006) On two kinds of meaning and interpretation. In Bo Mou (ed.) Davidson’s Philosophy and Chinese Philosophy (Leiden: Brill), pp. 207–27. Martinich, A. P. and Stroll, A. (2007) Much Ado about Nonexistence: Reference and Fiction (Totowa, MD: Rowman & Littlefield). Potts, C. (2005) Logic of Conventional Implicatures (New York: Oxford University Press). Searle, J. R. (1969) Speech Acts (Cambridge: Cambridge University Press).
13 On Three Theories of Implicature: Default Theory, Relevance Theory and Minimalism1 Emma Borg
Paul Grice’s distinction between what is said by a sentence and what is implicated by an utterance of that sentence is, of course, extremely familiar. It is also almost universally accepted. However, in recent literature, the precise account he offered of implicature recovery has been questioned and alternative accounts, emerging from different semantic programmes, have emerged. In this chapter, I would like to examine three such alternative accounts. My main aim is to show that the two most popular accounts in the current literature (the default inference view and the relevance-theoretic approach) still face significant problems. If this is right then there is a reason to look for a third alternative and in conclusion I’ll suggest that it is the approach emerging from so-called semantic minimalism which is best placed to accommodate Grice’s fundamental distinction between what a sentence means and what utterances of it implicate. To keep things relatively constrained I’m going to focus my attention on three desiderata which it seems plausible a successful theory of implicature should meet. I’ll assume (pretty much without argument) that any successful account should either accommodate or explain away (A)–(C): (A) the intuitive difference Grice identified between generalized and particularized conversational implicatures (B) the putative semantic (truth-conditional) relevance of some implicatures (C) recent experimental data concerning implicature recovery Condition (A) marks the fact that Grice’s distinction does seem intuitively appealing. Condition (B), as we will see, arises out of an apparent problem with the original Gricean framework; hence it would seem natural to expect any theory seeking to supplant Grice to avoid, or in some way defuse, this worry. Finally, the third condition (which I take to be the 268
On Three Theories of Implicature
269
strongest constraint here) is pretty self-evident: any theory of implicature which is concerned with modelling the actual processes underpinning implicature recovery in ordinary speakers must be open to confirmation or rejection in the light of empirical evidence. 2 Given these desiderata, then, the structure of the chapter is as follows: I’ll begin in section 1 by recapping on the Gricean approach and will outline one apparent problem that it faces (which yields desideratum (B) above). Sections 2 and 3 will then be concerned with the negative aim of showing that the two most popular accounts of implicature to be found in the recent literature fail to satisfy the desiderata given above. Finally, in section 4, I’ll sketch the approach of semantic minimalism and suggest that it provides a more plausible theory of implicature.
1. Grice’s theory of implicature According to Grice, two very different elements combine to make up the total signification of an utterance. As is well known, his fundamental divide came between what is said and what is implicated. ‘What is said (by a sentence)’ is a technical term for Grice and is usually held to be determined by the syntactic constituents of that sentence together with the processes of disambiguation and reference determination for context-sensitive expressions. Although Grice steered clear of the terminology of semantics and pragmatics, it is common to see this notion of what is said by a sentence as lining up with literal, semantic content for sentences, while what is implicated by an utterance is taken to line up with pragmatic meaning. Within the category of implicature Grice also recognizes distinct kinds, first distinguishing conventional (or non-conversational) and conversational implicatures, and then distinguishing generalized from particularized conversational implicatures. For the purposes of this chapter, it is the latter distinction which will be of special interest to us.3 Conversational implicatures in general are those propositions which a hearer is required to assume in order to preserve her view of the speaker as a cooperative partner in communication. Thus, to put things rather more formally, according to Grice, a speaker S conversationally implicates that q by saying that p only if: (i)
S is presumed to be following the conversational maxims, or at least the Cooperative Principle; (ii) the supposition that S is aware that (or thinks that) q is required to make S’s saying or making as if to say p consistent with this presumption; (iii) the speaker thinks (and would expect the hearer to think that the speaker thinks) that it is within the competence of the hearer to work out, or grasp intuitively, that the supposition mentioned in (ii) is required. (Grice 1989: 30–1)
270
Emma Borg
We can see the way in which the Gricean account of conversational implicature works, and appreciate the difference between generalized and particularized implicatures (GCIs/PCIs), by looking at an example. Imagine that Jack and Jill are planning a conference and Jack wants to know when the conference should start. Jill responds by saying: (1) Some delegates will attend at 8.30am. Here it seems that Jill’s utterance licenses both a generalized and a particularized conversational implicature. Her use of the term ‘some’ warrants a GCI because it is a weaker term on a scale (<some, all>) and, via observation of the maxim of Quantity, Jack is licensed to infer that, had Jill been in a position to assert the more informative proposition all delegates will attend at 8.30am she would have done this. Since she did not assert the stronger scalar term, Jack is licensed in inferring either that it does not hold or at least that Jill lacks evidence that it holds. Thus there is a pragmatically strengthened reading to be recovered from Jill’s utterance: Some but not all delegates will attend at 8.30am. This is a generalized conversational implicature because it attaches quite generally to the use of a scalar term like ‘some’.4 Although the pragmatically strengthened reading may be cancelled in a particular context (so that there is no contradiction in Jill saying ‘Some delegates will attend. In fact, now I think about it, all will’), it seems that we don’t need to know the particulars of this context of utterance to appreciate that, usually, an utterance of ‘some’ implies some but not all. Things are very different with the class of particularized conversational implicatures which are intimately bound up with specific contexts of utterance. For instance, in the example above, if it is common knowledge amongst the parties that not every delegate will attend every talk, then Jill implicates that we could start the conference at 8.30am. This further implicature is generated through observation of the maxim of Relevance: Jill’s utterance should be relevant to Jack’s inquiry about the start time for the conference. However, that this further proposition is conveyed is a feature of this particular context of utterance – were Jill to utter the same words in a different context the implicature might change or disappear. Intuitively, at least, this Gricean distinction between GCIs and PCIs seems an appealing one, capturing a real difference in the way that implicated propositions are conveyed. However, despite the intuitive appeal of the Gricean framework for implicature, it does face problems. One well-known objection to the Gricean framework (whereby what is said by a sentence is taken to be determined independently of what is implicated by an utterance of that sentence) concerns the claim that some implicatures can in fact contribute to what is said by a sentence (i.e. to the literal truth-conditions of the utterance). So for instance, consider (2): (2) Jill blew the whistle on poor practices at work and was sacked.
On Three Theories of Implicature
271
According to Grice, the meaning of ‘and’ is given by the logical conjunction, thus what is said by (2) is simply that two events took place; it does not assert any temporal or causal connection between them. Any such additional features arise only as a GCI connected to the use of the term ‘and’. Thus the GCI attaching to (2) is: (2*) Jill blew the whistle on poor practices at work and was sacked because of that. However, when we embed (2) in a larger context, it can seem compelling to think that the causal element is truth-conditionally relevant. Thus the truth-conditions of (3) and (4) seem to differ: (3) (4)
If Jill blew the whistle on poor practices at work and was sacked, then she is entitled to compensation. If Jill was sacked and blew the whistle on poor practices at work, then she is entitled to compensation.
Prima facie, at least, it is hard to see how to reconcile cases like this with the Gricean framework where literal meaning precedes implicature derivation, and thus where what is said by a sentence is independent of any implicature. Given this worry with the Gricean account, we might instead think to look for an alternative theory of implicature, one which can accommodate the apparent truth-conditional relevance of these generalized conversational implicatures. In the next two sections I will explore two such proposals.
2. The default interpretation view: Levinson (2000) In his book Presumptive Meanings, Stephen Levinson is concerned to draw apart three distinct levels of meaning. There is the level of sentence-type meaning (akin to the Gricean notion of what is said by a sentence) and there is utterance-token meaning (the full interpretation of a linguistic act), but there is also utterance-type meaning. This last category is the level of generalized conversational implicature. Levinson writes that GCIs ‘are carried by the structure of utterances, given the structure of language, and not by virtue of the particular contexts of utterances’ (Levinson 2000: 1). That is to say, GCIs form autonomous, default inferences, resting on quite general features of the usage of certain sub-sentential locutions. So, for instance, it is simply a feature of the default usage rule associated with the term ‘some’ that it licenses an automatic, heuristic-driven inference to “not all” (via what Levinson terms the ‘Q-heuristic’: what isn’t said isn’t the case). This is very different to the class of PCIs which are nonce inferences: potentially one-off, context-sensitive inferences, driven by close consideration of a particular speaker’s aims and
272 Emma Borg
intentions. Clearly, then, it is an in-built feature of Levinson’s account that it meets desideratum (A): it preserves Grice’s intuitive distinction between GCIs and PCIs. Furthermore, Levinson is at pains to accommodate desideratum (B): to give an account of the truth-conditional relevance of certain implicatures. His way of meeting this desideratum is to specify so-called ‘intrusive’ constructions, which allow embedded expressions to contribute not only their literal meaning to the truth-conditions of the whole but also any implicatures that the expressions carry (ibid.: 213–14). So for instance, conditional constructions may take account not only of the literal truth-conditions of embedded sentences but also the default implicatures they generate. In this way the intuitive difference between claims like (3) and (4) is explained, because the temporal/causal element associated with ‘and’ is incorporated into what is said by the larger conditional sentence. However, although Levinson does seek to address our second desideratum, it is not entirely clear that his account is successful in this respect. One initial worry with the account he provides is the extent to which it gives us a genuine explanation of the phenomenon in question versus a re-description of it. The worry is that, unless we have an independent characterization of what ‘intrusive constructions’ are, and how and why they function in the way supposed, it may appear that we are simply re-describing the problem phenomenon rather than taking a genuine step down the explanatory road. Be this as it may, however, the account also faces a much more concrete challenge. In her 2004 review of Levinson’s book, Robyn Carston (2004a) notes that, given Levinson’s notion of intrusive constructions, certain intuitively valid arguments will be rendered invalid. The problem is that the truth-conditions of the antecedent of a conditional like ‘p → q’ may differ from the truth-conditions of any unembedded instance of it, since in the more complex embedded setting the truth-conditions of the antecedent pick up on the GCIs of the unembedded ‘p’. Levinson’s account apparently predicts, then, that despite appearances to the contrary, many arguments which we take to be uncontroversial instances of modus ponens actually have the invalid argument form ‘p, r → q, therefore q’. As Carston notes, a theory which entails this outcome seems unacceptable. Finally, it is not clear that Levinson’s account can meet condition (C), concerning the experimental data relating to implicature recovery (this point is noted by Carston, but discussed at more length in papers by Ira Noveck). The problem is that experiments seem to show that the logical readings of scalar implicatures are accessed more quickly than pragmatic readings, yet such findings undermine the claim that it is the pragmatic reading which provides the default interpretation in these cases.5 So, for instance, in experiments conducted by Bott and Noveck (2004), where participants were given explicit instructions for interpreting scalar terms (e.g. ‘treat “some” as meaning “at least one and possibly all” ’ versus ‘treat “some” as meaning “at least one but not
On Three Theories of Implicature
273
all” ’), those who were required to give logical interpretations to scalar terms responded with truth-value judgments for sentences such as ‘Some elephants are mammals’ much more quickly than those required to give pragmatic readings. Furthermore, in experiments where no instruction was given and participants had free-choice over logical or pragmatic readings, the time-delay effects between the logical and the pragmatic readings remained. So, shown the sentence ‘some elephants are mammals’ a response of ‘true’, indicating a logical reading for ‘some’, was quicker than a response of ‘false’, indicating a pragmatic some but not all reading. Finally, if participants were placed under time constraints in providing truth evaluations for sentences containing scalar terms they were far more likely to respond with the logical reading, while when a delay prior to response was imposed they were more likely to respond with the pragmatic reading. What all these experiments point to, then, is the status of the pragmatic reading not as a default interpretation but as a delayed interpretation. As Noveck (2004: 314) notes: ‘there are no indications that turning some into some but not all is an effortless step’. Prima facie, then, Levinson’s default inference account faces problems with meeting both our second and third desiderata. Given this, many theorists have rejected the default inference view, preferring a more context-sensitive account of implicature recovery of the kind we will look at in section 3. However, before we move on to this kind of alternative account we should note that Levinson’s theory is not the only version of the default inference view currently on the table. For Gennaro Chierchia has proposed a radically different version of the default view which we need to consider. 2.1
The default interpretation view: Chierchia (2004)
Although Gennaro Chierchia is in agreement with Levinson concerning the default status of the enriched reading of scalar terms, the mechanism underpinning their default status is very different for Chierchia. He argues that ‘pragmatic computations and grammar-driven ones are “interspersed”. Implicatures are not computed after truth conditions of (root) sentences have been figured out; they are computed phrase by phrase in tandem with truthconditions (or whatever compositional semantics computes) . . . [C]ontrary to the dominant view, SIs [scalar implicatures] are introduced locally and projected upward in a way that mirrors standard semantic recursion’ (Chierchia 2004: 40).6 Chierchia retains the defeasibility of such implicatures by noting that they are cancelled only in certain semantic environments (roughly, those which are downward entailing). Thus, his idea is that the enriched reading is always calculated at the local level, though it may be subsequently removed at a more global level if the scalar term is found to be embedded in the wrong kind of semantic environment. Now the details of the account are fairly complex and we might question whether such a radical revision of the way in which pragmatics affects semantics is really motivated by the phenomenon of scalar implicature alone (see Horn 2006 for an argument
274 Emma Borg
along these lines). However, it is also clear that, on Chierchia’s version of the default inference view, many of the challenges from the previous section fall away. For instance, the intuitive worry that the notion of ‘intrusive constructions’ provides a re-labelling rather than an explanation of the semantic relevance phenomenon is avoided here. Clearly, if scalar implicatures build into the meaning of complex constructions in a way that parallels standard recursion for grammatical content, then we have a good explanation of why they are relevant to the truth-conditions of phrases in which scalar terms are embedded. Furthermore, Carston’s challenge that certain intuitively valid arguments get rendered as invalid also fails to hold against Chierchia’s position, because since implicatures build in to semantic content at the local level, there will be no difference in content between a premise containing a scalar term and an instance of that premise embedded in a conditional phrase. Clearly, then, Chierchia’s account fares better in these respects than Levinson’s. However, although he avoids Carston’s challenge, there are other argument forms where the account seems less compelling. For instance, since the pragmatically enhanced reading is generated at the local level, alongside grammatical content, and feeds into the compositional semantics in exactly the same way, it would seem that ordinary speakers should always work with the enriched reading (unless the context is a defeating one). Yet this would seem to suggest that ordinary speakers should find instances of the following argument valid: (5) Some elephants are mammals. (6) There is at least one elephant which is not a mammal. Yet it seems they do not. Now, there may be further moves for this version of the default view to make but, at least prima facie, the apparent invalidity of arguments like the above seem to suggest that we should keep literal semantic content and pragmatically determined implicatures further apart than does Chierchia’s account.7 Most seriously, however, it seems that Chierchia’s default view, just like Levinson’s, is at a loss to explain the time-delay data canvassed above. If the enriched reading is the default then it should be quicker to recover than logical readings, yet this is not what the experimental data shows.
3. Relevance theory: Sperber and Wilson (1986) According to relevance theory (RT) there are three possible stages or levels of utterance interpretation: the recovery of (usually) incomplete logical forms, the expansion or completion of these logical forms to yield explicatures (what is said by the utterance), and finally the further derivation of any implicatures. Both the construction of explicatures and the derivation of implicatures are driven by relevance-theoretic principles, i.e. the drive to recover the optimal interpretation in terms of a balance between cognitive effects and processing
On Three Theories of Implicature
275
effort (Sperber and Wilson 1986: 47–50). Once the hearer finds an interpretation which crosses the relevance threshold, she is licensed in taking it to be what the speaker intended to communicate. So, for RT, it is a mistake to think of either the logical or the pragmatic reading as the default interpretation of a scalar item. Although the lexical entry for the expression ‘some’ is given by the weaker logical interpretation, whether an utterance containing the expression is interpreted as conveying the logical or the pragmatic interpretation will be a matter of the relevance constraints in that particular context. Whichever interpretation crosses the relevance threshold in that context of utterance will constitute the correct interpretation. Now one feature of the RT view is that it rests no theoretical weight on the original Gricean distinction between GCIs and PCIs. All implicatures now stand together simply as further possible inferences to be drawn (via relevance-theoretic mechanisms) on the basis of the explicature; thus the account does not accommodate desideratum (A). However, the first of our conditions of success for an account of implicature is the weakest of the three; if an account deals well with requirements (B) and (C) without positing a difference in types of implicature, we should be prepared to relinquish Grice’s distinction, despite its intuitive appeal. So, how does RT fare with respect to (B) and (C)? At least initially, RT seems to deal very well with desideratum (B) – the semantic relevance constraint – since it is a key feature of the account in general that it allows for pragmatically contributed elements to contribute to the first truth-evaluable proposition to which the utterance gives rise (the explicature). Hence it should be no surprise at all to find that a pragmatically contributed strengthening of a scalar term contributes to the literal truth-conditions of an utterance.8 So, for instance, in (3) above, since the pragmatic reading of the antecedent is required to support the consequent, RT predicts that this will indeed be the recovered interpretation. What is said by this utterance is the pragmatically enhanced reading: if Jill blew the whistle on poor practices at work and was sacked because of that then she is entitled to compensation. However, although the solution is successful in this case, there do seem to be cases which cause problems for the relevance-theoretic account. For, as Larry Horn (2006) notes, the recovery of the stronger, pragmatic interpretation can occur as a ‘retroactive effect’, rather than something which occurs at the ‘first pass’ interpretation of the antecedent. Horn discusses the example: (7)
If some of my friends come to the party, I’ll be happy – but if all of them do, I’ll be in trouble.
and comments that ‘it’s only when the stronger scalar is reached that the earlier, weaker one is retroactively adjusted to accommodate an upper bound into its semantics, e.g. with some being reinterpreted as expressing
276 Emma Borg
(rather than merely communicating) “some but not” ’ (Horn 2006: 7). Yet this seems to suggest that a hearer will be unable to arrive at the explicature of the first sentence until she has waited to see if a stronger triggering item is delivered later. Now we should note that RT theorists do explicitly allow for a kind of retroactive effect in their account. For they do not posit the serial derivation of first what is said and then its implicatures, as we had on the Gricean picture; instead the two are derived in parallel and each may contribute to the other. Thus Carston (2004b: 8–9) notes: ‘The assumption . . . that “what is said” is determined prior to the derivation of conversational implicatures is [also] relaxed and the two levels of communicated content are taken to be derived in parallel via a mechanism of “mutual adjustment”, so that, for instance, an interpretative hypothesis about an implicature might lead, through a step of backwards inference, to a particular adjustment of explicit content’; see also Carston (2002: section 2.3.4). However, the cases Horn discusses cannot be treated as instances of mutual adjustment, for it is clearly possible that the trigger for Horn’s kind of retroactive reassessment could come some considerable time after the initial production of the scalar term. Yet it seems problematic to think that the explicit content of the original utterance cannot be determined until we have assured ourselves that no strengthening triggers will emerge anywhere further down the line. (And yet we know that the strengthened reading is not the one to be given by default, since RT opposes the default interpretation view.) In these kinds of cases, then, it seems that RT faces something of a dilemma: to accommodate truth-conditional relevance it seems that we need the pragmatically strengthened reading of the first antecedent in (7) to count as the explicature; but given RT’s independent acceptance of the idea that the logical reading gives the explicature in at least some cases, it seems quite possible that the context for the antecedent of (7) supports only the logical reading as an explicature.9 One option RT might pursue in the face of these kinds of cases is to hold that in retroactive cases like (7) the explicature of the utterance really is given by the weaker, logical reading, while the pragmatically strengthened reading is generated as some kind of (retroactively triggered) meta-representation of the original utterance.10 A worry, however, with any such move is that it threatens to undermine the RT account of putative truth-conditional relevance (since it seems that some additional account of the putative truth-conditional relevance of the meta-representation will then be needed). Even given this, however, it is not at all obvious that this line of response is in fact open to RT. The problem here turns on the conditions they offer for determining which pragmatic features count as part of the explicature and which contribute to the implicatures.11 It seems that there are two conditions which explicature content must fulfil: first, it must be an expansion of the logical form of the utterance and, second, it must satisfy the ‘scope principle’. According to this scope principle a pragmatically determined aspect of meaning is part of what
On Three Theories of Implicature
277
is said if it falls within the scope of logical operators such as negation and conditionals.12 Clearly, a pragmatic strengthening of a scalar term satisfies the first of these two conditions: it certainly seems to be an expansion of the logical form, since it enriches the interpretation of a lexical constituent of the utterance. However, it also seems that (at least on the current account) it satisfies the second condition: if a context is such as to give rise to a pragmatic enrichment of a scalar term in a given utterance then it seems that that pragmatically enriched reading will also fall under any logical operators. So, it seems that either a pragmatically enriched reading shouldn’t be generated at all (so that the explicature of ‘some’ is given just by the logical reading) or a context supports the pragmatically strengthened reading and this then constitutes the explicature of the utterance. It is thus not clear how, according to relevance theory, a context could be such as to support the pragmatic reading of a scalar term and yet that reading could then fail to count as an ordinary part of the explicature. Yet unless this is possible the option of treating the pragmatic reading of ‘some’ in (7) as some kind of meta-representation, distinct from the ordinary explicature, seems closed to RT.13, 14 Turning now to condition (C) – that the theory fit with experimental evidence concerning implicature recovery. Once again, it might seem that RT is in a good position to account for this condition. Recall that, for RT, it is perfectly plausible that in many contexts it is the logical interpretation which first crosses the relevance threshold. Thus, where hearers go on to recover a pragmatically enriched interpretation, this will be the result of further processing. In these cases, the pragmatic reading is effortful and thus the prediction that the pragmatic reading should take longer to recover seems to follow naturally. Thus as Bott and Noveck (2004: 454) suggest: ‘relevance Theory would argue that inferences arise as a function of effort . . . Thus, according to Relevance Theory, the logical response ought to be faster than a pragmatic response’.15 However, on closer inspection, it is not so clear that RT does fit well with the experimental findings. A first worry is that, on the RT view, the pragmatic reading can form the explicature of an utterance (i.e. the first interpretation which provides sufficient cognitive effects at the cheapest cognitive cost). Thus, if the recovery of enriched scalar terms is held to be effortful and time-consuming, so too should the recovery of other kinds of pragmatically enriched explicatures. Yet it seems unlikely that relevance theorists would want explicature recovery in general to be marked by timedelays. For instance, consider: (8) ‘The apple is red’ ⇒ (a) The apple is red (in some way) (b) The apple is red on most of its skin (9) ‘Everyone was sick’ ⇒ (a) Everyone was sick (b) Everyone at the party was sick
278
Emma Borg
In each of these cases, in the right context, the (b) interpretation would form the pragmatically enhanced explicature of the original utterance, with (a) providing an interpretation which is free from pragmatic effects. Yet it seems extremely unlikely that advocates of RT would want to predict that the (b) interpretations are recovered more slowly than the (a) interpretations; thus the explanation of the time-delay data on GCIs seems threatened. Of course, advocates of RT either need to hold that time-delay effects will be present in the recovery of all pragmatically enhanced explicatures or that the pragmatically enhanced explicatures in (8) and (9) are in some way different from those recovered via scalar implicatures.16 Pursuing this second disjunct might allow advocates of RT to accommodate the time-delay data appropriately, but someone pursuing this line obviously still owes us an account of what the relevant difference here is.17 A second worry with RT’s explanation of the experimental data concerns children’s recovery of scalar implicatures. As many theorists have observed, children are less likely to give pragmatic interpretations to scalar items; instead, in this respect at least, children turn out to be much more logical than adults (Rips 1975; Noveck 2001; Papafragou and Musolino 2003).18 Yet it is at least prima facie unclear how this data is to be accommodated on the current view, for RT predicts that both adults and children are in possession of the right reasoning processes to generate both explicatures and implicatures (i.e. both possess the same relevance-directed mechanisms). Furthermore, children are as good as adults in utilizing these processes in most other respects. For instance, it is the RT mechanisms which, ex hypothesi, explain how putative semantic underdetermination is to be overcome (i.e. how an utterance of ‘I’ve eaten’ gets interpreted in a context as explicitly meaning that I’ve eaten recently), yet children overcome semantic underdetermination just as well as adults. So it looks, prima facie, as if RT owes us an account of why children fail to use the comprehension mechanisms they do supposedly possess in just these cases. That is to say, why does a child who can move from ‘The apple is red’ to the apple is red on the outside via RT mechanisms usually fail to move from ‘Some children like sprouts’ to at least one but not all children like sprouts via identical mechanisms? To summarize: RT doesn’t place any theoretical weight on the distinction between GCIs and PCIs. While this was one of our original desiderata, it was the weakest of the three and if RT proved successful in accommodating the other two, this wouldn’t be much of a problem. However, I’ve tried to suggest that it is far from clear that RT does offer a successful account of the two remaining conditions. First, the account seems to face problems with explaining the putative truth-conditional relevance of implicatures due to the possible retrospective nature of pragmatic interpretation. Since we may end up revising our interpretation of a scalar term at any later point in the conversation it seems that either RT must maintain that the explicature of the utterance is not settled until we have made sure that no strengthening triggers occur at any later point (an unappealing suggestion), or it requires
On Three Theories of Implicature
279
some alternative account of the apparent truth-conditional relevance of the pragmatically strengthened reading in cases like Horn’s. Secondly, it seems that there are problems accounting for the experimental data, since RT predicts that where GCI effects occur they can contribute to the explicature of the utterance, yet it is at best a moot point as to whether RT wants to be committed to predicting similar time-delay effects for the recovery of all pragmatically enhanced explicatures. Furthermore, it is not immediately clear how the account copes with the difference between GCI recovery in adults and children, since it predicts that exactly the same mechanism is responsible for GCI and other forms of explicature recovery in adults and children, yet in many other cases of explicature recovery children seem unimpaired. Given these worries, then, I think we are licensed in turning to our third and final account.
4. Semantic minimalism According to semantic minimalism, there is, just as Grice predicted, a minimal sentence meaning to be recovered from a constrained set of pragmatic processes (such as disambiguation and reference assignment for indexicals and demonstratives). This minimal meaning (minimalists tend not to call it ‘what is said’ since they conceive of ‘saying’ as a thoroughly pragmatic notion) can diverge to a greater or lesser extent from what a speaker conveys by the utterance of the sentence in a given context, so minimalism fully embraces the fundamental Gricean distinction between literal sentence meaning on the one hand and speaker-meaning on the other. Furthermore, minimalists agree with Grice that an interpretation of ‘Some Fs are G’ as some but not all Fs are G lies on the implicature side of the divide, going beyond the content recoverable by lexical information and rules of composition alone. According to semantic minimalism, semantic content is marked out by one (or both) of the following features: (i)
every contextual contribution to semantic content is syntactically triggered; (ii) every contextual contribution to semantic content is formally tractable. Concentration on (i) yields the kind of minimalism defended by Cappelen and Lepore (2005), where any relevant feature of the context can be brought into the semantic domain by an appropriate syntactic cue.19 On the other hand, a combination of (i) and (ii) yields the kind of minimalism defended by Borg (2004a). On this view, even when a syntactic item requires a contextual contribution, the relevant contextual features must themselves be formally tractable, and speaker intentions, it is argued, lie outside this formal domain (the argument for this claim is the ‘Frame Problem’, discussed by Fodor 2000: 37–8
280 Emma Borg
and Borg 2004b). According to the former variety of minimalism, an interpretation counts as pragmatic just in case it contains elements not directly contributed by the syntactic constituents of the sentence. According to the latter, the difference between semantic and pragmatic features lies in the kind of comprehension processes underlying their recovery: semantic content can be recovered via computational operations alone, while pragmatic content relies on non-formal, abductive processes (capable of recovering any contextual feature, up to and including speaker intentions) for its recovery.20 Given the rendering of the semantics/pragmatics distinction within my variety of minimalism, however, it seems that GCIs end up straddling the divide. They cannot constitute semantic content, since they require information beyond the strictly semantic to recover (i.e. information about what speakers using the given terms commonly tend to convey). However, they do not seem to constitute full-blown pragmatic content, since they do not rely on access to the current speaker’s state of mind. A GCI can be recovered just by knowing that a speaker who says ‘some’ often conveys some but not all, but this information could be gathered as the result of exposure to past conversational exchanges, it would not require access to the current state of mind of the speaker.21 Thus recovery of a GCI, like the recovery of literal meaning on this picture, could be the result of a purely ccomputational-, syntax-driven reasoning process. This embeds the intuitive distinction Grice recognized between GCIs and PCIs, for it treats them as the results of different kinds of cognitive processing. This in turn might suggest the following kind of minimalist model for implicatures: there is assumed to be a modular language faculty, responsible for linguistic processing up to and including literal semantic content, and this content then feeds into two further systems – the holistic, general pragmatic system (responsible for recovering PCIs) and the more limited (and possibly modular in its own right) system for recovering GCIs which operates over statistical facts about what speakers have intended in past conversational exchanges. As noted, it seems such a model would meet desideratum (A); but what about (B) and (C)? Turning to (B) first: the advocate of minimal semantics will have to bite the bullet with respect to the apparent semantic relevance of some implicatures, claiming that in all cases it is an appearance of truth-conditional relevance not a genuine case of rich pragmatic input to semantic content. Thus, given (3) and (4) (repeated here): (3)
If Jill blew the whistle on poor practices at work and was sacked, then she is entitled to compensation. (4) If Jill was sacked and blew the whistle on poor practices at work, then she is entitled to compensation. the minimalist must hold that there is no semantic difference between the two. However, while this may seem simply unacceptable, we should note
On Three Theories of Implicature
281
that the current model can offer an explanation of why we hear a difference of meaning in such cases. For given the context-invariant aspects of GCI recovery, it is unsurprising that we have the intuitions we do about cases like the above. GCIs are formally derived, they are mechanical and habitual, just as is the recovery of semantic content for the minimalist; thus it is no surprise that we can’t help but hear a serious difference in cases like these. The way our minds work on this model explains why we find it so easy to hear what is really a difference in speaker-meaning as a difference in sentence meaning. Thus although the current account rejects (B) as a valid constraint on a theory of implicature, it can explain why it seems like a genuine condition. Furthermore, we might also note that this picture, whereby GCI recovery shares features with both the semantic and the fully pragmatic (PCI) side of the divide, might help to explain why GCIs display the kind of habitual nature- and context-independence stressed by Levinson, since they are indeed accessible without any knowledge of the aims and intentions of the particular speaker and they are generated by purely computational, deductive processes. Finally, then, turning to (C): the account seems to fit well with the experimental data canvassed earlier. It can predict the time-lag data of Noveck (2001, 2004), since a GCI is indeed a further interpretation above and beyond the semantic content of the sentence. Furthermore, the account does not need to predict a general time-delay in implicature recovery, for it may be that while further computational processing of an utterance is effortful, subjecting the literal content of the utterance to further non-computational processing is relatively effortless (to put things crudely, the thought would be that getting at what a speaker intends by her utterance comes naturally to us, or at least more naturally than does reflecting on what speakers tend to convey by specific terms).22 This minimalist account also explains the relative lack of GCI recovery in children, for it allows that though they have the cognitive resources for drawing both GCIs (which are computationally derived, as is semantic content) and PCIs (abductively derived), still they lack the necessary premises for (at least some cases of) GCI recovery, viz. information about the usual use of scalar terms in conversational exchanges. As exposure to normal language use increases so the drawing of implicatures based not directly on particular speaker intentions but on the characteristic use of words increases. 23 It seems, then, that this minimalist theory of implicatures fares reasonably well with the desiderata we set out in the introduction.
5. Conclusion I’ve looked at three different neo-Gricean accounts of implicature and tried to argue that it is the final version, drawn from minimal semantics, which gives the most satisfactory account of the phenomenon. The default interpretation view is to be commended for taking seriously the difference between
282
Emma Borg
GCIs and PCIs and recognizing the habitual and context-independent nature of the former. However, we saw that the approach ran into problems with its account of the apparent semantic relevance of GCIs and with the experimental data concerning time-delays for the recovery of the pragmatic reading of scalar terms in adults. Next we turned to relevance theory. On this account, the distinction between GCIs and PCIs disappeared, with all implicatures being further interpretations of an utterance driven by relevance-theoretic mechanisms. Whether or not this failure to meet desideratum (A) mattered turned on the account’s success with our other two conditions. However, despite initial appearances that RT coped very well with conditions (B) and (C), I suggested that there were underlying problems. First, questions were raised about RTs account of the putative semantic relevance of GCIs, due to the retrospective nature of implicature accommodation at the semantic level. Secondly, questions were raised about how the theory fits with experimental evidence: it doesn’t seem to accommodate the time-delay data from Noveck (since it doesn’t predict time-delays for explicature recovery in general); nor does it seem to accommodate the variation in GCI recovery between adults and children (since it predicts both have the right mechanisms in place for such recovery and predicts both make successful use of those mechanisms in other settings). Finally, I turned to minimal semantics, and suggested that on this account GCIs be treated as computationally based, habitual inferences, which contrast with the recovery of PCIs (where PCI recovery requires abductive reasoning based on hypotheses about the speaker’s intentions). I suggested that this minimalist account not only preserves the intuitive divide between GCIs and PCIs, but it is also able to explain the experimental data (both concerning time-delay effects in adults and the lack of GCI recovery in children). Furthermore, it offers at least some explanation of the apparent truthconditional effects of GCIs. Minimalism holds that these effects are indeed apparent since, as Grice held, GCIs do not contribute to the recovery of the semantic content of a sentence (a process which the minimalist can hold takes place within a computational language faculty, which does not have access to such contextual information as the further propositions standardly communicated by speakers uttering ‘p’). However, the account also predicts that the impression that GCIs do contribute to semantic content is pretty much unavoidable, given the nature of GCI derivation. For these reasons, then, I conclude that the third theory of implicature, drawn from minimal semantics, deserves further investigation.
Notes 1. Versions of this chapter were presented at a workshop on minimal semantics at the University of Valladolid, Spain, the 9th International Pragmatics Association
On Three Theories of Implicature
2.
3.
4.
5.
6. 7.
8. 9.
283
conference in Italy, the Centre for Linguistics, Philology, and Phonetics in Oxford, Glasgow University, and Trinity College Dublin. I’m grateful to the audiences on those occasions, particularly Robyn Carston and Jim Levine, for comments. Thanks also to Kent Bach, Larry Horn and an anonymous reader for very useful comments on draft versions of the chapter. This contribution was completed during an AHRC research leave award. We should note, as Bach (2006) stresses, that Grice himself is apparently offering a rational reconstruction of the recovery of implicatures by ordinary speakers, not a theory of psychological processing. So empirical data concerning implicature recovery does not tell directly for or against his theory. Yet it seems clear that the accounts which seek to supplant Grice are concerned with providing psychologically real accounts of implicature recovery, thus empirical evidence is relevant. Grice himself was somewhat doubtful about the former category, stating that ‘the nature of conventional implicature needs to be examined before any free use of it, for explanatory purposes, can be indulged in’ (1989: 46); and recently some theorists, such as Bach (1999), have argued that there are no such things as conventional implicatures. Although I don’t want to pursue this debate here, I do think there is enough doubt about the status of conventional implicatures to warrant our concentration on conversational implicatures. Though we should note that Bach (2006) lists the labelling of scalar inferences as GCIs in the ‘top ten misconceptions’ about conversational implicature. For him, these are clear cases of impliciture. The experimental findings here have been disputed, for instance by Bezuidenhout and Cutting (2002). However, as noted in Breheny, Katsos and Williams (2006), questions may be raised about the opposing form of experiments undertaken by Bezuidenhout and Cutting with respect to their relevance for scalar implicatures. While the status of recent empirical work is no doubt always open to question, I will assume in what follows that the findings of time-delays are significant. It should also be noted, as pointed out by an anonymous reader, that the experimental findings discussed here and below relate only to scalar implicatures and not to GCIs in general. Hence if there was a reason, as Bach (2006) argues, to hold that scalars are not GCIs, then it would be no objection to a theory of GCIs that it didn’t accommodate the time-delay data for scalar cases. However this move makes it imperative that scalars be shown not to be genuine GCIs, a claim which remains controversial and which, in fact, none of the theorists under discussion here (such as Levinson) embrace. We might note that the idea that scalar implicatures are derived locally is also supported by experimental findings from Storto and Tanenhaus (2004). One move an advocate of this position might make would be to suggest that intuitions about validity run on grammatically derived content only, ignoring locally derived pragmatic content. However, any such move would seem to sit uncomfortably with the original motivation for treating GCIs as contributing to semantic content, which turned on intuitions about the truth-conditions of conditional statements. A second option would be to claim that the invalidity is explained by the fact that the meaning of ‘elephant’ entails ‘is a mammal’. However, this move would depend on a potentially questionable view of meanings as definitions and might not help in all cases (e.g. the invalidity of ‘some penguins can’t fly, therefore there is at least one penguin which can fly’). Thus RT is a variety of a broader school of thought often termed ‘contextualism’. It might be thought that these retroactive cases simply constitute examples where hearers are forced, by subsequent conversational moves, to revise their initial
284
10. 11.
12. 13.
14.
Emma Borg judgments of explicatures. Thus we might contend that the speaker intended to assert the stronger interpretation of a scalar term but that the audience mistakenly only grasped the weaker reading, with this mistake coming to light later on. However, while this scenario is possible, it is not the one envisaged here: in this case relevance constraints at the time of utterance dictate that the most relevant interpretation is the weaker scalar reading. In this case, if we assess explicatures at the time of utterance, it simply is the case that the explicature is the weaker reading (so the audience are right in their assessment) and thus any later requirement for the stronger reading is at odds with this. Since it is relevance constraints, and not simply speaker intentions, which dictate explicatures, the only way RT can capture Horn’s retroactive cases is by predicting that we hold off on assessments of relevance until some terminal point in the conversation. But, as noted in the text, this seems problematic. The worry here is perhaps parallel to one noted by Levinson (2000: 247) and discussed further by Bezuidenhout (2002: 268–9), namely: what is the size of the constituent processed by Levinson’s heuristic-driven system? It appears that it must be above the lexical level but below the level of the whole discourse (since, as Bezuidenhout notes, Levinson ‘does not want the conclusion that we hold off on processing until we have all the information that could possibly be relevant’). Here we have an apparently similar worry for RT: what is the size of the constituent against which relevance is assessed? It seems perfectly possible that the relevance of a pragmatically strengthened reading may not become evident until the completion of the whole discourse, but surely hearers can’t hold off explicature derivation until they get to the end of a conversation? A possibility suggested by Robyn Carston in personal correspondence. Although this is a point of some debate within relevance theory. Carston suggests that there can be no failsafe tests for determining the explicature/implicature divide. Thus while the various tests proposed may provide useful heuristics, they do not constitute principles for determining explicature content. She writes (2002: 191): ‘I was (and still am) of the view that the communicative principle of relevance itself or, more particularly, the comprehension strategy that follows from it, effects a sorting of pragmatic inferences into contributions to the proposition expressed (explicature) and implicatures’. See Recanati (1993: 269–74). Another possible response here, raised by Daniel Watts (p.c.), would be to allow that hearers derive both the logical and the pragmatic readings as explicatures and then wait for later conversational developments to tell them which one to cancel. Such a move, however, would run counter to RT (which holds that we stop processing once we have one interpretation that crosses the relevance threshold). It would also seem to impose quite serious cognitive burdens on agents, requiring them to hold multiple interpretations for multiple sentences for an indefinite length of time. Finally, it might interfere with a hearer’s ability to draw PCIs from an utterance, since, on this account, they remain essentially ignorant of the explicature at first pass. A final point to note with respect to this argument concerns how the Gricean fares with examples like (7). On this approach, since GCI’s cannot infect semantic content, the claim must be that the utterances are literal contradictions: to say that I will be happy if some friends come to the party, but not happy if they all come, is to contradict myself, since all of them coming entails some of them coming. The explanation of why utterances like (7) seem acceptable then lies in the fact that hearers are concentrating on GCI content, not semantic content
On Three Theories of Implicature
15.
16. 17.
18. 19.
20.
21.
22.
23.
285
(this claim is fleshed out somewhat in section 4 on minimalism). Clearly, however, this kind of explanation is not available to advocates of RT, since explaining the apparent truth-conditional relevance of GCI’s in cases like (7) without positing actual truth-conditional relevance would undermine RT’s motivation for ever treating GCI’s as genuinely semantically relevant. Thanks to Jim Levine for stressing this point. The claim that experimental evidence of time-delay in scalar implicature recovery supports a context-driven, as opposed to a default, approach is also made in Breheny et al. (2006). Napoleon Katsos, in conversation, suggested pursuing the second disjunct here. Furthermore, one might think that, in answering this point, the account is very likely to be led towards reinstating something very like the GCI/PCI divide which we saw it rejected at the outset. Though see Pouscoulous et al. (2007) for evidence that four-year-olds can recover scalar implicatures. If we allow more syntactic context-sensitivity than is traditional, (i) also yields the syntax-driven approach of Stanley and Szabo (2000), and Stanley (2005). See Borg (2007, 2009), for discussions of the similarities and differences between minimalism and Stanley-style indexicalism. One reason for thinking that the more stringent version of minimalism is the one that we want is that it would allow semantic content to be recovered by a dedicated Fodorian language module (where the hallmark of a Fodorian module is that it is computational) and there is evidence to suggest that linguistic abilities are modular in this way; see Borg (2004a: chapter 2). The idea would thus be that GCIs are a kind of abstraction from, or ossification of, PCIs: an agent starts by grasping a pragmatically enhanced reading of a scalar term as a PCI but later comes to grasp such readings simply through knowing how words are commonly used. It is because GCI understanding is taken, on this view, to emerge from prior PCI understanding that the mechanisms underpinning GCI recovery are held to lie outside the language faculty proper. So does the account need to predict a general time-delay for all GCIs (a prediction which seems unlikely to be borne out since, e.g., hearing ‘and’ as ‘and then’ seems unlikely to be marked by any significant time-delay)? I think the answer to this question is ‘no’, for there are at least two moves available on the current model to explain differences in time-delay data for GCIs. First, it could be argued that, in at least some cases, the ossification of PCIs into GCIs is more extreme than for scalars, so that the time-delay reduces, in some cases yielding an insignificant degree of delay, or no delay at all. A prediction on this line of explanation would be that different subjects, with different degrees of familiarity with a language, should exhibit different degrees of delay (with the time-delay reducing as GCIs become more firmly embedded for the subject). Secondly, it might be argued that in some cases of apparent GCI recovery a hearer can’t in fact access the right content without reference to the specific context and what the speaker in that context intends, thus these inferences are in fact PCIs and thus no timedelay is expected. However I’m grateful to Manuel Garcia-Carpintero, in conversation, and an anonymous reader, for pressing me on this point. An additional piece of experimental evidence which might be thought relevant here comes from Storto and Tanenhaus’s (2004) observation that the exclusive interpretation of ‘or’ (often taken to be a GCI) is accessed locally. That is to say, hearers are able to utilize the information ‘one or the other but not both’ to help to instigate action on-line, at the point at which they hear the term ‘or’, rather
286
Emma Borg than waiting until sentence comprehension is completed. However, although this information is available very early, their experiments showed that it is not available quite as early as genuine lexical information. Although they note that this feature might be a result of problems with their experimental design, if it were to prove robust for scalar implicatures in general it might provide further support for a view of GCI derivation as midway between semantic and fullblown pragmatic processing.
References Bach, K. (1999) The myth of conventional implicature. Linguistics and Philosophy, 22: 327–66. Bach, K. (2006) The top ten misconceptions about implicature. In B. Birner and G. Ward (eds.) Drawing the Boundaries of Meaning: Neo-Gricean Studies in Pragmatics and Semantics in Honor of Laurence R. Horn (Amsterdam: John Benjamins), pp. 21–30. Bezuidenhout, A. (2002) Generalized conversational implicatures and default pragmatic inferences. In J. Campbell, M. O’Rourke and D. Shier (eds.) Meaning and Truth (New York: Seven Bridges Press), pp. 257–83. Bezuidenhout, A. and Cutting, J.C. (2002) Literal meaning, minimal propositions and pragmatic processing. Journal of Pragmatics, 34: 433–56. Borg, E. (2004) Minimal Semantics (Oxford: Oxford University Press). Borg, E. (2004b) Formal semantics and intentional states. Analysis, 64: 215–23. Borg, E. (2007) Minimalism versus contextualism in semantics. In G. Preyer and G. Peter (eds.) Context-Sensitivity and Semantic Minimalism: New Essays on Semantics and Pragmatics (Oxford: Oxford University Press), pp. 546–71. Borg, E. (2009) Meaning and context: a survey of a contemporary debate. In D. Whiting (ed.) The Later Wittgenstein on Language (Basingstoke and New York: Palgrave Macmillan). Bott, L. and Noveck, I. A. (2004) Some utterances are underinformative: the onset and time course of scalar inferences. Journal of Memory and Language, 51: 437–57. Breheny, R., Katsos, N. and Williams, J. (2006) Are generalised scalar implicatures generated by default? An on-line investigation into the role of context in generating pragmatic inferences. Cognition, 100: 434–63. Cappelen, H. and Lepore, E. (2005) Insensitive Semantics: A Defense of Semantic Minimalism and Speech Act Pluralism (Oxford: Blackwell). Carston, R. (2002) Thoughts and Utterances (Oxford: Blackwell). Carston, R. (2004a) Review of S. Levinson, Presumptive Meanings. Journal of Linguistics, 40: 181–6. Carston, R. (2004b) Truth-conditional content and conversational implicature. In C. Bianchi (ed.) The Semantics/Pragmatics Distinction. ( Stanford: CSLI Publications), pp. 65–100. Chierchia, G. (2004) Scalar implicatures, polarity phenomena, and the syntax/ pragmatics interface. In A. Belletti (ed.) Structures and Beyond (Oxford : Oxford University Press), pp. 39–103. Davis, W. A. (1998) Implicature: Intention, Convention, and Principle in the Failure of Gricean Theory (Cambridge: Cambridge University Press). Fodor, J. (1983) Modularity of Mind (Cambridge, MA: MIT Press). Fodor, J. (2000) The Mind Doesn’t Work That Way (Cambridge, MA: MIT Press). Grice, H. P. (1989) Studies in the Way of Words (Cambridge, MA: Harvard University Press).
On Three Theories of Implicature
287
Horn, L. (2006) The border wars: a neo-Gricean perspective. In K. Turner and K. von Heusinger (eds.) Where Semantics Meets Pragmatics (Cambridge: Elsevier), pp. 21–48. Langdon, R., Davies, M. and Coltheart, M. (2002) Understanding minds and communicated meanings in schizophrenics. Mind and Language, 17: 68–104. Levinson, S. (2000) Presumptive Meanings: The Theory of Generalized Conversational Implicature (Cambridge, MA: MIT Press). Noveck, I. A. (2001) When children are more logical than adults. Investigations of scalar implicature. Cognition, 78: 165–88. Noveck, I. A. (2004) Pragmatic inferences related to logical terms. In I. A. Noveck and D. Sperber (eds.) Experimental Pragmatics (Basingstoke and New York: Palgrave Macmillan), pp. 301–21. Papafragou, A. and Musolino, J. (2003) Scalar implicatures: experiments at the semantics-pragmatics interface. Cognition, 86: 253–82. Pouscoulous, N., Noveck, I., Politzer, G. and Bastide, A. (2007) Processing costs and implicature development. Language Acquisition, 14: 347–75. Recanati, F. (1993) Direct Reference: From Language to Thought (Oxford: Blackwell). Rips, L.J. (1975) Quantification and semantic memory. Cognitive Psychology, 7: 307–40. Sauerland, U. (2004) Scalar implicatures in complex sentences. Linguistics and Philosophy, 27: 367–91. Sperber, D. and Wilson, D. (1986) Relevance: Communication and Cognition (Oxford: Blackwell). Stanley, J. (2005) Semantics in context. In G. Preyer and G. Peter (eds.) Contextualism in Philosophy: Knowledge, Meaning, and Truth (Oxford: Oxford University Press), pp. 221–54. Stanley, J. and Szabo, Z. (2000) On quantifier domain restriction. Mind and Language, 15: 219–61. Storto, G. and Tanenhaus, M. (2004) Are scalar implicatures computed online? Proceedings of WECOL 2004 (Fresno: California State University).
14 Contextualism in the Philosophy of Language Nikola Kompa
A word is dead, when it is said, some say; I say it just begins to live, that day. Emily Dickinson
1. Contextualism vs. semantic minimalism Different things go by the name ‘contextualism’. Even when we focus on contextualism in the philosophy of language we will find that there is no single characterization accepted by all those who claim (or are said) to be contextualists. To a very rough first approximation, one may characterize the contextualist as someone who is sympathetic to pragmatically oriented, usage-based approaches to meaning. She will take sides with Robyn Carston when she claims that: [a] major development in pragmatics since Grice’s work is the recognition that linguistically decoded information is usually very incomplete and that pragmatic inference plays an essential role in the derivation of the proposition explicitly communicated. (Carston 2002a: 133) Her opponent, on the other hand, will advocate a formal semantic approach to meaning and insist that deriving the proposition expressed by a sentence is a purely syntactic-semantic enterprise. He will, therefore, tend to agree with Emma Borg that: in general, formal theories can be characterized as fundamentally syntax-driven theories, which claim that it is possible to deliver an account of the propositional or truth-conditional content of a sentence in natural language simply via formal operations over the syntactic features
288
Contextualism in the Philosophy of Language
289
of the sentence, that is, over the lexical items it contains and their mode of composition. (Borg 2004b: 3) More specifically, the contextualist and his opponent, the formal semanticist, negotiate what was traditionally taken to be key semantic notions: the notion of what is said in an utterance of a sentence, the notion of the proposition expressed by a sentence, the notion of a sentence’s truthconditions, and – at the most basic level – the notion of meaning itself. Semanticists try to save one or another of these notions from pragmatic intrusion. Kent Bach argues for a semantic notion of what is said. Herman Cappelen and Ernie Lepore defend the notion of the proposition semantically expressed by an utterance. Emma Borg tries to show that there is a level of semantic content free of contextual effects. Jason Stanley argues that the (intuitive) truth- conditions of an utterance are the result of purely syntactic-semantic interpretation. And they all presuppose a viable notion of linguistic meaning. In the remainder of the section, I will briefly review some of the more detailed characterizations of contextualism and its chief rival, also known as semantic minimalism. Semantic minimalism can be taken to be the ‘natural descendant of formal theories of meaning’ (Borg 2007: 355). Emma Borg, Herman Cappelen and Ernie Lepore all defend one version or another of semantic minimalism; and they also give more or less detailed characterizations of their opponent, the contextualist. Cappelen and Lepore, for example, distinguish between moderate and radical contextualists. Among the radical contextualists they count philosophers such as Charles Travis, John Searle, François Recanati, Anne Bezuidenhout and relevance theorists such as Dan Sperber, Deirdre Wilson and Robyn Carston. Their views can be traced back to the work of Ludwig Wittgenstein and John Austin, according to Cappelen and Lepore (see Cappelen and Lepore 2005: 5–6). ‘Moderate contextualism’, on the other hand, is Cappelen and Lepore’s label for all other opponents of semantic minimalism. The problem with this label is that, as Cappelen and Lepore acknowledge, most of the so-called moderate contextualists wouldn’t call themselves such (see ibid.: 7). So this label is bound to cause confusion. Charles Travis, for example, takes the moderate contextualists to be: in essence, followers of Carnap. Carnap’s idea was this: if an expression’s contribution to what is said in speaking it varies from speaking to speaking, find a variable with different values for those speakings; postulate a (tacit) device in the expression, referring, on a speaking, to the value of that variable for that speaking, find a function from that value to the contribution of the expression, on that speaking, to what was then said. (Travis 2006: 39)
290 Nikola Kompa
This view is more akin to the moderate version of a formal semantic approach put forward by Jason Stanley. He claims that any truth-conditional context sensitivity a sentence may exhibit can be resolved by assigning values to elements in the logical form of the sentence uttered. In all cases in which a sentence’s truth-conditions are affected by (extra-linguistic) context, this can be explained by pointing to an indexical, a demonstrative, a pronominal element or a variable in the logical form of the sentence. To the extent that context affects the truth-conditions of a sentence, this can be traced back to something like an open slot in the logical form of the sentence. Yet there need not be a corresponding open slot in the surface grammatical structure, as the logical form might differ from the surface structure. The logical form is the sentence’s ‘real structure’, though (Stanley 2000: 399–400). The effects of context on the truth-conditional interpretation of an assertion are restricted to assigning values to elements in the logical form of the expression uttered. And each context-sensitive element in logical form ‘brings with it rules governing what context can and cannot assign to it’ (ibid.: 396). Truth-conditional interpretation is purely syntactic-semantic interpretation, in that it is syntactically triggered and governed by semantic rules. While Cappelen and Lepore’s characterization of moderate contextualism is a bit off target, their characterization of radical contextualism is more to the point. The radical contextualist, they say, holds that ‘no English sentence S ever semantically expresses a proposition’ (Cappelen and Lepore 2005: 6). It is only utterances that can semantically express a proposition and, hence, have a truth-condition. The semantic minimalist disagrees; he endorses Propositionalism, the claim that every non-indexical sentence expresses a proposition (Borg 2007: 347–50). Emma Borg takes Propositionalism to be one of the defining features of semantic minimalism.1 And according to François Recanati, who holds a contextualist view, the basic question in the debate is: whether we may legitimately ascribe truth-conditional content (the property of ‘saying’ something, of expressing a thought or a proposition) to natural-language sentences, or whether it is only speech acts, utterances in context, that have content in a basic, underived sense. (Recanati 2004: 83) The contextualist holds that natural language sentences are contextsensitive through and through. It is utterances in contexts, not sentences, that have determinate and truth-evaluable content (ibid.: 154). A second defining feature of semantic minimalism is what Borg calls The Basic Set Assumption, the assumption that there is only a small set of contextsensitive expressions (Borg 2007: 350–51). Cappelen and Lepore agree with Borg’s characterization: The most salient feature of semantic minimalism is that it recognizes few context sensitive expressions, and, hence, acknowledges a very limited
Contextualism in the Philosophy of Language
291
effect of the context of utterance on the semantic content of an utterance. (Cappelen and Lepore 2005: 2) Competent speakers can easily be made to agree that what expressions such as ‘I’, ‘you’, ‘here’ or ‘that’ refer to is subject to contextual variation. But they won’t as easily agree that the same holds for expressions such as ‘black’, ‘flat’ or ‘ham sandwich’. And, as Borg emphasizes, the minimalist is anxious to ‘preserve this intuition’ (Borg 2007: 350). He takes an expression to be contextsensitive only if that is obviously so. The basic set of context-sensitive expressions comprises only obviously context-sensitive expressions – and they are few in number. According to Cappelen and Lepore, for example, the set comprises only personal and demonstrative pronouns, adverbs such as ‘here’ and ‘now’ and a handful of other expressions (Cappelen and Lepore 2005: 144; they devise a couple of tests in the attempt to distinguish genuine contextsensitive expressions from ‘pretenders’, see ibid.: chapter 7). Bach is even more restrictive and allows only indexical expressions to contribute to what is said. Stanley, on the other hand, is less restrictive as he promotes hidden indexicality. The contextualist, finally, ‘argues for the radical context-dependence of what is said’ (Bezuidenhout 2002: 106). She will happily agree that there is in principle no limit to the effects of contextual information on propositional content: not only are all expressions context-sensitive, but context may also add constituents to propositional content which are entirely unrepresented in syntax. (Carston and Powell 2006: 350) Emma Borg’s account of what is distinctive of semantic minimalism is similar to (though more detailed than) that of Cappelen and Lepore. Yet her characterization of the opponent position differs somewhat from their account. She contrasts semantic minimalism with a view she calls Dual Pragmatics (Borg 2004b). The dual pragmatist is so called because he allows pragmatic processes to play a double role in utterance interpretation. They are required not only in order to work out implicatures of a given utterance, but also in order to get at the truth-conditional content, i.e. the proposition expressed in the utterance of the sentence in question. She attributes this view to relevance theorists such as Sperber and Wilson or Carston, to Recanati (whom she calls a contextualist) and also to proponents of discourse representation theory such as Hans Kamp. But Borg doesn’t count Travis among the contextualists or even among the dual pragmatists. She sees him as advocating a much more radical view according to which meaning is essentially indeterminate. And while she is right that Travis’s views invite a more radical departure from traditional semantic views than more moderate versions of contextualism, I would nonetheless like to classify Travis as a contextualist, albeit a radical one. So I suggest we think of all these positions as ordered on a scale, with more moderate positions at one end and more radical positions at the other.
292
Nikola Kompa
Borg discusses another important difference between contextualism (or dual pragmatics) and minimalism. First she points out that ‘the contextualist is committed to rich pragmatic effects throughout what is said by a sentence, or throughout the proposition expressed, or throughout semantic content’. (Borg 2007: 340) These pragmatic effects have to be pragmatically inferred. That is what makes them pragmatic in the first place. And they may not be syntactically triggered (see, for example, Carston 2002b: 23). Also, these inferences are, like pragmatic inferences in general, defeasible. In contrast, Borg sees the semanticist as being committed to Formalism – ‘the idea that there is an entirely formal (i.e. syntactic) route to semantic content’ (Borg 2007: 355): After all, it seems that this is just what it is to pursue a formal account: to think that everything which can be found at the semantic level can be traced to the syntactic level (that we don’t get anything ‘for free’, as it were, at the semantic level). (Borg 2004a: 217) Yet the commitment to Formalism can be understood in two ways, she says. First, it could be understood to require that in order to get at the semantic content of the sentence uttered, appeal is to be made to the context of utterance only if it be syntactically triggered. Secondly, it could be understood to require that contextual contributions to semantic content be formally tractable (Borg 2007: 355). That in turn requires, according to Borg, that the contextual features in question be objective (non-perspectival) features of the context, for only objective features of context are formally tractable. Intentional states she takes to be non-objective features of context. They are not formally tractable, on her view. And: the main reason for rejecting appeals to speaker intentions from within a formal semantic theory concerns the clash between the deductive, computational nature of a formal theory [ ... ] and the abductive nature of inferences concerning speakers’ intentions. (Borg 2004a: 218–19) The emphasis is on the deductive, computational nature of a formal semantic theory, which is contrasted with the abductive nature of pragmatic inferences. Inferences concerning speaker intentions in particular and pragmatic inferences more generally are liable to be defeated by new information. And in pragmatic inference any piece of information may turn out to be relevant (Borg 2007: 356). So ‘there is no limit to the amount of contextual information that can affect pragmatic interpretation’ (Recanati 2004: 54). More fully, these inferences are non-monotonic in the following sense. It may be the case that although conclusion C intuitively follows from a set of premises P, it no longer follows from a set M, such that M ⊃ P. This feature of pragmatic inferences obviously doesn’t square well with the
Contextualism in the Philosophy of Language
293
allegedly deductive character of semantic interpretation. Here is an example of Borg’s to illustrate the abductive nature of intentional interpretation: So, imagine that you see Sally filling a glass of water from the tap. Then you might reason as follows: ‘Sally is getting a glass of water from the tap. The best explanation for this action is that Sally is thirsty and wants a drink; therefore Sally is thirsty and wants a drink’. Clearly this is a nondemonstrative piece of reasoning and it is susceptible to the influence of an open-ended range of contextual factors. For instance, say that you know that Sally has just come in with Sourav and that Sourav is wearing running gear and looks out of breath, then the best explanation for Sally’s action might be that Sourav is thirsty and Sally is getting a drink for him. Or imagine that Sally has just glanced at her potted plant, then the best explanation might be that she wants to water her plant. (Borg 2004b: 78–9) Yet the nature of pragmatic inferences which often seem to take the form of an inference to the best explanation not only conflicts with the deductive, but also with the computational nature of formal semantic theories. The semanticist is subscribing to something like the following (admittedly somewhat simplified) picture of how semantic interpretation proceeds. In a first step, the logical form of the sentence uttered has to be figured out. Then the elements of the logical form are each assigned a value in accordance with the respective linguistic conventions or assignment functions. In a third step, these values are combined in accordance with certain combination rules to yield the proposition expressed by the utterance of the sentence and its truth-conditions. Here is a somewhat lengthy quote from Stanley: First, a speaker makes an utterance, and her linguistic intentions uniquely determine a certain syntactic structure, or ‘logical form’, as it is known on syntax. [ ... ] Successful interpretation involves assigning denotations to the constituents of the logical form, and combining them in accord with composition rules that do not vary with extra-linguistic context. The denotations that successful interpreters will assign to constituents of a logical form will be constrained by the linguistic conventions governing those elements. In the case of certain elements, which wear their context-dependent nature on their sleeve, the linguistic conventions governing them are rather lax. [ ... ] What results from a successful application of this first stage of interpretation is a unique proposition, a fully truth-evaluable entity. (Stanley 2002: 149–50) More specifically, what the semanticist is looking for is a procedure that yields for every well-formed sentence (of finite length) of the language in question its meaning, i.e. the proposition expressed or the truth-conditions,
294
Nikola Kompa
after finitely many steps. But if in intentional interpretation ‘there is no determinate boundary at the outset on which facts could turn out to be relevant’ (Borg 2004b: 79), then there can be no such algorithmic procedure for pragmatic interpretations, as there is no guarantee that the task can be completed after finitely many steps. Now, as for the first rendering of Formalism according to which appeal to context has to be syntactically triggered, that is something all semantic minimalists seem anxious to subscribe to. Bach would happily do so, as the requirement strongly resembles the Syntactic Correlation Constraint which he approvingly quotes Paul Grice as endorsing: ‘what is said must correspond to “the elements of [the sentence], their order, and their syntactic character” (Grice 1989: 87). So if any element of an utterance, i.e. what the speaker intended to convey, does not correspond to any element of the sentence being uttered, it is not part of what is said.’ (Bach 2001: 15). Stanley would also subscribe to some version of the Syntactic Correlation Constraint; he would insist, though, that there are syntactic constituents that cannot be read off the surface grammatical form of the sentence in question. And Cappelen and Lepore too hold that: [a]ll semantic context sensitivity (i.e., context sensitivity that affects the proposition semantically expressed) is grammatically triggered, i.e., it is triggered by a grammatically (i.e. , syntactically or morphemically) articulated sentential component. (Cappelen and Lepore 2005: 144) The second rendering of Formalism, according to which any contextual contribution to what is said has to be formally tractable, is more problematic. If Borg is right – and I think she has convincingly argued the point – that intentional interpretation requires pragmatic (that is, defeasible) inferences and doesn’t therefore square well with the deductive nature of formal semantics, a semantic minimalist who wants to subscribe to Formalism cannot allow speaker intentions to play a role in the determination of semantic content or the proposition expressed. I will come back to that in the last part of the chapter. Let us take stock. I followed Emma Borg in characterizing the semantic minimalist as someone who subscribes (at least) to Propositionalism, Formalism (on at least the first reading) and The Basic Set Assumption. The contextualist, on the other hand, is commonly regarded as someone who rejects all three assumptions. In the remainder of the chapter, I will sketch some of the considerations contextualists commonly adduce in order to defend their position. (A full defence of contextualism is beyond the scope of this contribution, though.) More specifically, I will look at various forms of context sensitivity in natural language and try to clarify what exactly the alleged context sensitivity in these cases amounts to, i.e., which features are sensitive to contextual
Contextualism in the Philosophy of Language
295
variation, and what it takes to resolve the context sensitivity at issue in order to interpret a context-sensitive utterances. I will thereby argue mainly against The Basic Set Assumption, but I hope that it will become clear that the argument against The Basic Set Assumption also casts doubt on Formalism. The assumption of Propositionalism, on the other hand, will not be given much attention in what follows. The problem is that no one in the debate has a clear criterion of when something – be it a sentence or an utterance – expresses a proposition, i.e., is fully propositional or truth-evaluable. Consequently, we also lack a clear grasp of what the result of semantic interpretation is supposed to be. Space being limited, I will not go into that here. Instead, let’s get down to cases.
2. Cast of characters (I) There are various well-known cases of overt (as opposed to covert) context sensitivity. For a start, there is abundant ambiguity in natural languages. It comes in two varieties. There is lexical and syntactic ambiguity, the most widely discussed instances of the latter being scope ambiguities. Within lexical ambiguity it is common to further distinguish between homonymy and polysemy. The latter is more problematic but also more interesting from a contextualist point of view, and it will be discussed separately below. So for the time being, ‘ambiguity’ should be read as ‘homonymy’. In the resolution of ambiguity, context is often said to play a pre-semantic role: Sometimes we use context to figure out with which meaning a word is being used, or which of several words that look or sound alike is being used, or even which language is being spoken. These are pre-semantic uses of context. (Perry 1997: 593) Various tests for ambiguity have been discussed in the literature (see, for example, Lakoff 1970; Zwicky and Sadock 1975, 1987; Atlas 1989; Pinkal 1985, 1991; Quine 1960). Ambiguous expressions stand in need of precisification. The speaker has to have a particular reading in mind. The hearer too has to make up his mind as to which reading was the intended one. This relates to another feature of ambiguous expressions: they lack an encompassing reading. Put otherwise, homonymic readings of an ambiguous expression: strongly resist any kind of unification. Take the case of bank. It is hard to think of the different sorts of bank as parts of a whole, or as united into a global Gestalt. We might think of a very general category to which they both belong, such as ‘entity,’ or even ‘location,’ but this is not good enough, because it does not distinguish banks from non-banks. (Croft and Cruse 2004: 115)
296 Nikola Kompa
Also, ambiguous expressions are said to have only finitely many, discrete, and natural readings – as opposed to stipulated precisifications. The word ‘fast’ can mean ‘faster than 20 mph’, but that is not a natural reading of ‘fast’ (Pinkal 1985). For all that, it is sometimes difficult to decide where homonymy ends and polysemy, metonymy or metaphor begins (Aristotle, Top. I, 15). Note also that ambiguity is not the only case were context has to play a pre-semantic role. Even in order to phonetically recognize what has been said in a given utterance, appeal has to be made to context because: in normal speech it is physically impossible to hear each segment: speech is just too fast. Twenty segments a second is not unlikely, but the brain cannot distinguish even half that number of separate sounds at a time. (Aitchison 2003: 227) So word recognition is to a large extent guesswork. And it seems reasonable to suppose that in order to phonetically figure out what has been said, one draws on whatever contextual clue one can get hold of. (II) Next, there are well-known and much-discussed cases of overt context sensitivity where context is said to play a semantic role, as in the case of indexicality, demonstrative and anaphoric reference, and definite as well as indefinite descriptions. In all these cases one has to accord a role to context in general and, arguably, to speaker intentions in particular, in the resolution of the respective context sensitivity.2 By way of illustration, consider the case of anaphoric reference. As is well known, there are interestingly different cases of anaphoric pronouns (see, for example, Geach 1962; Evans 1977, 1980; and Neale 1990). Yet the important point for present purposes is simply that in order to fix the referent of an anaphoric pronoun one has to take speaker intentions, worldly knowledge, contextual clues as to the purpose or point of the conversation, and discourse or linguistic context (also called cotext) comprising the previous utterances or the enclosing text, into account. Consider this: (1)
The city councillors refused to issue the workers a permit for a demonstration because they ... feared violence. were communists. were preparing for an election. were from out of town. applied too late. (Pylyshyn 1987: viii)
The pronoun ‘they’ can be taken to refer either to the city councillors or to the workers. And hearers will have no problem deciding which referent
Contextualism in the Philosophy of Language
297
to choose on any given completion of (1). But they thereby obviously have to draw on their knowledge about linguistic as well as non-linguistic context. (III) Cases of obvious incompleteness give rise to context sensitivity too. Examples are provided by non-sentential and elliptic assertions: (2) From Spain (– said while holding up a letter). (Stainton, 1997: 61) (3) Bill likes working and so does Bob. (Bach 1994: 281) There is an ongoing debate about how to handle these cases, and in particular whether they should be treated semantically or pragmatically. Robert Stainton, for example, argues for a pragmatically oriented approach to nonsentential speech, while Stanley questions the assumption ‘that all alleged examples of non-sentential assertion must be treated by the same general strategy’ (Stanley 2000: 403; see also, for example, Elugardo and Stainton 2004; Ludlow 2004; Stainton 1995, 1997, 2005). (IV) More exciting, from a contextualist perspective, are somewhat subtler forms of underdetermination or incompleteness. Here are some examples: (4) Steel isn’t strong enough. (Bach 1994: 268) (5) You are not going to die (– said by a mother to her son who has cut his finger). (ibid.) (6) Mending this fault will take time. (Carston 2002b: 26) (7) Jill can’t continue. (Borg 2004b: 228) (8) The girl with the flowers is happy (Atlas 1989: 26) (9) Jack tells Jill, ‘I love you too’. (Bach 1982: 593) (10) Every boy (in the class) is seated. (Stanley 2005: 223/224) (11) John is tall (for a fifth grader) (ibid.) (12) Sam left Jane and she became very depressed. [and as a result] (Carston 2002a: 135) The contextualist will claim that these sentences are context-sensitive in that they can be used to say different things given different contexts of utterance. And the truth-conditions of these utterances will differ accordingly. And that is so because they allow (and even ask) for contextual completion. (That is not to deny that some of the sentences have preferred or even default completions.) Sentence (7), for example, can be used to say that Jill can’t continue school, or that she can’t continue dance classes, or university education, and so on. And sentence (11) can be used to say (as the material in brackets indicates) that John is tall for a fifth grader, or that he is tall for an NBA player, and so on.
298
Nikola Kompa
Consequently, the contextualist will also be inclined to deny that homophonic truth-conditions and propositions, as advocated by Cappelen and Lepore (see Cappelen and Lepore 2005), serve any theoretical or explanatory purpose. Suppose, for example, that you want to know what we have been talking about and I tell you: ‘John said that Tom is ready’. What does the indirect report of John’s utterance tell you in this case that I couldn’t have told you by directly reporting on John’s utterance? Sentences can do all the work homophonic propositions are supposed to do; the latter are therefore theoretically superfluous. Suppose John next says: ‘Tom is rich’. If all you know is that John’s utterance is true iff Tom is rich, then you won’t be able to tell whether John’s utterance is true – even if you know all there is to know about Tom’s financial situation. You also have to know what it takes for someone to count as reasonably rich in the context of John’s utterance, i.e., what the relevant comparison class is. Gilbert Harman gets it exactly right when he says that the fact that ‘A is F’ is true iff A is F ‘is a trivial point about the meaning of “true”, not a deep point about meaning’ (Harman 1987: 216). (V) Another interesting and very common form of context sensitivity is metonymy. We often use a term to refer not to its literal referent but to something that is somehow related to the literal referent. Here is François Recanati’s favourite example: (13) The ham sandwich left without paying. (Recanati 2004: 26) Of course the waiter didn’t complain about the ham sandwich but about the person who ordered or ate the sandwich. One may succeed in referring to a person or thing by using an expression for something only accidentally (but saliently) related to that person or thing. But note also that not just any relation will do, as only some relations between any two given things can be exploited for purposes of deferred reference (see Nunberg 1996 for an example). Metonymical reference usually follows certain patterns – it is not that anything goes. We often use a container for the content, a part for the whole (pars pro toto), or an attribute for the possessor of the attribute. These are common forms of metonymy. Here is another example. Think of reading a recipe that says: (14)
Now throw in a tablespoon full of chopped parsley and cook five minutes more.
Of course you are not asked to throw in the tablespoon itself. (It won’t get tender anyway.) But, as has just been said, we often use the container metonymically for the content. Or suppose you overhear someone say: (15) They drank a bottle of wine.
Contextualism in the Philosophy of Language
299
Most probably, they did not drink the bottle, but its content. This, again, is a case of a container/content metonymy. Yet not only singular terms but also demonstratives can be used metonymically. Here are two examples due to Emma Borg: (16) ‘That is my favourite author’ said while pointing at a book. (17) ‘This singer is great’ said while pointing at a record. (Borg 2004b: 175)3 The demonstrative ‘that’ in (16) is not exactly used figuratively. But is it used literally? Also, examples such as these seem to suggest that there is no clearcut distinction between deferred reference and non-deferred reference: one may refer to someone by pointing to his head in the crowd, or his arm visible over the wall, or to his shadow, or to a photograph depicting him, and so on (see Borg 2004b: 179). Moreover, the contextualist will emphasize the fact that the deferred referent is truth-conditionally relevant: an utterance of sentence (13) will, arguably, be true only if the customer who stood in the contextually salient relation to the ham sandwich left without paying – not if the ham sandwich itself left without paying. The metonymic referent enters into the truthconditions of the sentence. Also, whether the expression ‘the ham sandwich’ is used to refer to the ham sandwich or to someone who ordered or ate it depends on context. This kind of context sensitivity affects all expressions that can be used metonymically. And if someone doesn’t understand deferred uses of linguistic expression he can hardly qualify as a competent speaker. (VI) To my mind, the most interesting examples of context sensitivity are provided by cases of metaphor and polysemy, though. To that I now turn. Suppose a speaker says: (18) Sally is a block of ice. (Searle 1993: 83) (19) Robert is a bulldozer. (Carston 2002b: 350) As various philosophers have pointed out, the properties attributed to Sally or Robert in (18) and (19) respectively are not properties of blocks of ice or bulldozers. Which properties is someone who utters (19) thereby attributing to Robert? She is, presumably, attributing to him such properties as the property of being stubborn and persistent, of not looking left or right, of taking the risk of collateral damage, of egoistically pursuing his own interests, and so on. But none of these properties is a property of bulldozers. That is to say, these properties can be attributed to bulldozers only metaphorically. It should come as no surprise, therefore, that two of the most appropriate
300 Nikola Kompa
characterizations of metaphor are themselves metaphorical: Metaphor, it seems is a matter of teaching an old word new tricks. (Goodman 1976: 69) Briefly, a metaphor is an affair between a predicate with a past and an object that yields while protesting. (ibid.) Of course, metaphorical interpretation is context-sensitive and requires that contextual clues be taken into account. But it also requires a lot of background knowledge, knowledge about ‘the world’. For all that, metaphorical language use is a central feature of natural language – not just a linguistic deviation. It may even improve understanding: The good metaphor satisfies while it startles. Metaphor is most potent when the transferred schema effects a new and notable organization rather than a mere relabeling of an old one. (Goodman 1976: 79/80) Expressions are often used metaphorically in order to conceptualize abstract and mental phenomena. And the expressions we thereby use are often taken from the realm of sense experience. That is why we are feeling blue, or complain about someone being cold, and so on. And that also allows us to use metaphors in explaining human behaviour. We say that a friend of ours broke down under pressure – thereby exploiting the metaphor of the mind is a brittle object (see Lakoff and Johnson 1980: 28). Or we say of a friend that she is feeling up, that her spirits rose, or that, sadly, she sank into a coma. We use orientational metaphors, as George Lakoff and Mark Johnson call them (see Lakoff and Johnson 1980: chapter 4), to conceptualize mental phenomena by giving them a spatial orientation. Abstract phenomena, too, are commonly conceptualized by way of metaphors. We think of love as if it were a journey: Look how far we have come; yet now we are at a crossroad; from here on we have to go separate ways as this isn’t going anywhere; we are in a dead-end street. And we talk of theories and arguments as if they were buildings: The theory needs more support, it lacks a foundation; the argument collapsed, it was not solid enough (see Lakoff and Johnson 1980: chapter 10). We commonly use expressions from the concrete realm of sense experience to conceptualize abstract and mental phenomena (see also Keller and Kirschbaum 2003: 36, 99). Consequently, if metaphorical utterances were always strictly speaking false, that would render most of our talk about abstract or mental phenomena false. Moreover, metaphor is a driving force behind language change. We use language in order to classify things and events. When we encounter a new situation, we usually prefer to conceptualize it by using familiar vocabulary – even if we have to somewhat stretch the word’s ‘old meaning’ so that the word becomes applicable to the new situation. And some of these uses catch on.
Contextualism in the Philosophy of Language
301
(VII) Lexicalizing a metaphorical use of a word yields polysemy, where ‘[p]olysemy is understood here in a broad sense as variation in the construal of a word on different occasions of use’ (Croft and Cruse 2004: 109). Consider this: (20) (a) (b) (c) (d)
The janitor goes from top to bottom of the building. The staircase goes from top to bottom of the building. The river Ganges goes from the Himalayas to the Indian Ocean. The power of prayer goes round the world. (Aitchison 2003: 59)
Does the janitor go in the same way from top to bottom as the staircase does? Is that already a case of metaphor? The river Ganges, on the other hand, rather clearly goes to the Indian Ocean only metaphorically. But polysemy need not be the result of lexicalizing a metaphor. Many expressions can be used to describe different actions or entities without that being necessarily due to a kind of figurative use. They simply allow for different understandings or interpretations relative to different contexts. Compare the following: (21)
(a) Jane opened the window. (b) Bill opened his mouth. (c) Sally opened her book to page 56. (d) Mike opened his briefcase. (e) Pat opened the curtains. (f) The child opened the package. (g) The carpenter opened the wall. (h) The surgeon opened the wound. (Carston 2002b: 325)
It is something different to open a curtain than to open one’s mouth. And a wall is usually opened not in the way one opens a wound. The word ‘open’ can be used to describe a variety of different actions, and one has to contextually work out the appropriate understanding. Some philosophers take these examples to show that word meanings (or concepts) are not ready-made entities but have to be constructed in context. According to Robyn Carston, for instance, in many of the cases discussed so far, ‘an ad hoc concept is constructed and functions as a constituent of what is explicitly communicated’ (Carston 2002b: 357). The idea is that words encode concepts which ‘provide the starting point for a pragmatic process which results in a different concept, one which is narrower and/or broader than, or, perhaps in the case of some metaphors, different in some way from the lexical concept’ (ibid.). And it was also examples such as these that led John Searle to argue for the claim that the literal meaning of a sentence determines its truth-conditions only relative to a set of background assumptions, ‘and without some set of background assumptions the sentence does not determine a definite set of
302
Nikola Kompa
truth conditions at all’ (Searle 1978: 220–1). Moreover, a sentence’s truthconditions vary with differences in the respective background assumptions. (To speak of background assumptions is a bit misleading as the background not only comprises assumptions but also practices, according to Searle.) Here is another example to further illustrate the point: (22)
(a) (b) (c) (d) (e)
Bill cut the grass. The barber cut Tom’s hair. Sally cut the cake. I just cut my skin. The tailor cut the cloth. (Searle 1980: 221)
Searle takes it that the word ‘cut’ is used literally in these sentences. Yet the word makes different contributions to the truth-conditions of these sentences (see ibid.: 222–3). For what one does when one cuts the grass is different from what one does when one cuts the cake. If ‘someone tells me to cut the grass and I rush out and stab it with a knife [ ... ] I will have failed to obey the order’ (ibid.: 223). And the reason why the word ‘cut’, meaning what it does, fixes different truth-conditions in these cases stems from the fact that as members of our culture we bring to bear on the literal utterance and understanding of a sentence a whole background of information about how nature works and how our culture works. [ ... ] I understand the sentence “He cut the grass” differently from the way I understand “He cut the cake” [ ... ] because I know a lot of things about grass, e.g. what people have grass lawns for, what they do to their lawns, etc. [ ... ] and my knowledge that cutting grass is different from cutting cakes is part of this larger system of knowledge. (ibid.: 226–7) A somewhat similar idea is elaborated by Charles Travis (see also Moravcsik 1998 and Bezuidenhout 2002). He argued in a series of papers and books that whether a word can be used to describe a particular entity (or action) depends not only on how things stand (‘objectively’, if you like) with that entity but also on the participants’ interests, the purpose or point of the conversation and so on. Take the following sentences: (23) The kettle is black. (Travis 1985: 196) (24) The ball is round. (Travis 1996: 454) (25) The leaves are green. (Travis 1997: 89) These sentences don’t wear their context sensitivity on their sleeves. Yet Travis invites us to consider, for example, the following story: Pia’s Japanese maple is full of russet leaves. Believing that green is the colour of leaves, she paints them. Returning, she reports: ‘That’s better. The
Contextualism in the Philosophy of Language
303
leaves are green now.’ She speaks truth. A botanist friend then phones, seeking green leaves for a study of green-leave chemistry. ‘The leaves (on my tree) are green,’ Pia says. ‘You can have those.’ But now Pia speaks falsehood. (Travis 1997: 89) It is worth noting that Travis’s concern here is not with ambiguity in the sense of homonymy (nor with bio-chemical details). Nor is this a case of a metaphorical or otherwise non-literal use. Pia might, arguably, be said to have used the expression ‘green’ in two different ways in both cases – as might be witnessed by the fact that although she used the words to describe one and the same state of affairs, her words were true in one case and false in the other. Yet if these different ways of using ‘green’ were readings of a homonymic word ‘green’, then the list of readings would be very long indeed, for there are many ways to use (and understand) an utterance of ‘X is green’. It can be used to say that X is green all over, green only on the outside, painted green, green under normal lighting conditions, green given the particular lighting conditions obtaining now, and so on. And each of these different ways of using and understanding an utterance of ‘X is green’ would issue in yet another reading of ‘green’. So the readings a homonymic expression has differ from the ways a not-homonymic (or disambiguated) expression might be understood in different contexts of its use. Nonetheless, each particular understanding determines a different truth-condition: ‘Typically, an (e.g.) English expression is such that, with its meaning (unambiguously) fixed, there are a variety of distinct (perhaps better: distinguishable) things to be said in using it on some production of it or other.’ (Travis 1985: 187)4 Examples such as this one seem to show that even a non-indexical and non-ambiguous sentence need not succeed in determining all by itself, so to speak, a truth-condition. Trivially, the sentence ‘The leaves are green’ is true iff the leaves are green. But, again, the point is that even if one knows everything there is to know about leaves and their colour, one won’t know whether an utterance of ‘The leaves are green’, said of a particular tree, is true or false unless one also knows what the point and purpose of the utterance was supposed to be, what interests were operative in the context in question, and so on. So whether any such sentence can be correctly applied to a particular situation depends on the participants’ interests, purposes, background assumptions, and so on. Given such and such interests and assumptions, it may be correctly applicable; given different interests and assumptions, it may be not correctly applicable. Of course we could specify further and try to make some of these interests and assumptions explicit. But as Anne Bezuidenhout points out: There is no sentence that we can produce that can settle all questions about how some original sentence is to be understood, since language
304 Nikola Kompa
doesn’t function that way. It is not self-interpreting. (Bezuidenhout 2002: 113) Given all these examples of polysemy, one might, echoing John Austin, ask the question: ‘Why do we call different things by the same name?’ (Austin 1979a, 69). And a sensible answer seems to be that it allows us to use a finite vocabulary to talk about or describe an – in principle – infinite array of situations. That is just a case of what Jon Barwise and John Perry called the efficiency of language. (Barwise and Perry 1983: 5; see also Goodman 1976: 80) And Jonathan Cohen puts the point thus: Rather, it is an enormous convenience that the same word can often be uttered in one or other of several different though related senses. Instead of having to learn a very much larger number of words, each with fixed and context-independent meanings, we can learn a relatively small number of words with variable meanings and then exploit their verbal or situational contexts of utterance in order to disambiguate their actual occurrences. (Cohen 1985: 132) Our expressions can be made to fit all the various situations we encounter almost perfectly. They answer to our needs, i.e., they are context-sensitive. They can be adjusted, modulated, if need be. And need there is, as we never encounter exactly the same situation twice over. Given that linguistic expressions serve the purpose of classification (a function of language commonly ignored in the formal semantic tradition where the focus is on language’s function of describing some state of affairs or of stating facts), it is to be expected that they are context-sensitive and polysemous. So when we encounter a hitherto unknown situation, we are thereby able to conceptually structure it by using familiar vocabulary. That yields polysemy. But in order to be able to classify new situations by using familiar vocabulary, we sometimes have to ‘unduly’ stretch the meaning of a given expression so that it becomes applicable to the case at hand. This commonly results in metaphor. (VIII) Finally, sometimes we even create words.5 These nonce words are tailor-made for a particular purpose or situation, as when I want to take a picture of you and say: ‘Please do a Napoleon for the camera’ (Aitchison 2003: 171). Most likely, you will understand what I want you to do. As Aitchison puts is: ‘Humans are amazingly good at extending the application of words’ (ibid.: 162). And sometimes one of those nonce words catches on, as language is in a constant state of flux.
3. Context-sensitive extensions and pragmatic inferences What do the examples show? They show that there is, in natural language, context sensitivity in abundance. Of course, the semantic minimalist will
Contextualism in the Philosophy of Language
305
be impressed by these examples only if they affect semantics, that is, if the context sensitivity involved affects the expressions’ extension, the sentences’ truth-conditions or some such thing. And he will be worried only if it cannot be resolved by purely semantic interpretation. But that is exactly the contextualist’s point. The context sensitivity involved in these cases affects semantics, and it cannot be resolved by purely semantic interpretation. So there is reason for the semanticist to be concerned. Let us take the first point first. To what extent does the context sensitivity involved in the above examples affect semantics? In almost all the cases discussed above, it seems that what varies in accordance with contextual variation is something like the expressions’ extension or denotation, what the expressions are used to refer to, denote, or talk about. This is obviously so in cases of ambiguity, indexicality, demonstrative and anaphoric reference. But it is, though less obviously, also so in the case of polysemy and metonymy (and maybe even in the case of metaphor, but that depends on what our best theory of metaphor will turn out to be). If John says ‘Tom cut the cake’ then the verb ‘cut’ is used to pick out a particular action – an action that differs from the action performed by cutting the lawn. That is the point of polysemic expressions: they can be used to denote slightly different kinds of things, properties, actions or events. (This, again, results from the fact that language serves the purpose of classification.) Moreover, in all these cases context sensitivity seems to affect truth-conditions. Even in the case of metonymy the metonymic referent seems to be truth-conditionally relevant. And John’s utterance of ‘Tom is rich’ is true iff Tom is rich given the standard of richness operative in John’s context of utterance. This will be no problem for those semanticists who sever semantics from truth-conditions (as for example Kent Bach does). Yet most semanticists would, I think, agree with David Lewis that ‘[s]emantics with no treatment of truth conditions is not semantics’ (Lewis 1970: 190). Let us turn to the second point, the point about how to resolve the context sensitivity at issue. In all cases discussed above, appeal has to be made to context in order to properly interpret the respective utterances. One traditional approach to meaning and interpretation takes it that this is best explained on the assumption that the meaning of an expression provides us with an assignment function that delivers for any given context the value of the expression in that context. So the idea is that meanings are something like intensions, functions from Montague-style indices to extensions. Yet in light of the examples just discussed, it becomes questionable whether contextual interpretation can be semantically controlled in such a way. In many of the cases discussed, pragmatic inferences are required in order to figure out an expression’s extension and consequently the truth-conditions (or the proposition expressed) of the sentences containing the expression. A pragmatic inference is a sort of inference to the best explanation. It is nonmonotonic, defeasible. Consequently, as Emma Borg has rightly pointed
306 Nikola Kompa
out, the need for pragmatic inferences clashes with the allegedly deductive nature of semantic interpretation. Moreover, in performing pragmatic inferences one draws on contextual information concerning the circumstances of utterance, speaker intentions (to the extent that they are manifest), the participants’ interests and background assumptions, the purpose or point of the conversation, and so on. A formal characterization of an index as an n-tuple of contextual features will, therefore, hardly be available. Rather, context should be thought of as providing information but not as determining anything (a point Kant Bach and others have emphasized before). To think of context as a neat list of contextual features that determine the semantic values of context-sensitive expressions is to employ a metaphor, and a bad one at that. David Lewis, for example, was well aware of the difficulty: I emphasized that the dependences of truth on context was surprisingly multifarious. It would be no easy matter to devise a list of all the features of context that are sometimes relevant to truth in English. In [General Semantics, N.K.] I gave a list that was long for its day, but not nearly long enough. (Lewis 1980: 30) To sum up: I argued on behalf of the contextualist that there is ample context sensitivity in natural language. And what varies from context to context in these cases is the expression’s extensions and the sentence’s truth value. So The Basic Set Assumption has to be given up. Moreover, since the contextually appropriate extension of a given context-sensitive expression has to be pragmatically inferred, Formalism has to be given up, too. There is no purely syntactic-semantic route to extension and content. Pragmatic inferences are indispensable if one wants to figure out an expression’s extension. Yet the process of interpretation thereby preformed hardly deserves to be called (purely) semantic interpretation. I take that to be one of the basic contextualist insights.
Notes 1. Kent Bach rejects Propositionalism and thereby distinguishes his own position, which he calls Radical Semantic Minimalism (or Radicalism), from semantic minimalism. According to Bach, even indexical free sentences sometimes fail to express a complete proposition but express only a propositional radical, as he calls it. 2. The latter claim is contested. Christopher Gauker (2008), for example, holds that the referent of a demonstrative is the object that best satisfies certain wholly nonintentionally specifiable accessibility criteria. See also Marga Reimer (1991) and Kent Bach (1992) for two contrasting views on ‘intentionalism’. 3. Geoffrey Nunberg also discusses cases of deferred reference: ‘You can point at a girl child to identify her father (“He is in real estate”). You can point at a book to identify its author (“She was my chemistry teacher”), or at an author to identify a
Contextualism in the Philosophy of Language
307
book (“That is a wonderful autobiography”)’ (Nunberg 1993: 26; see also Kaplan 1977: 490, footnote 10). 4. Here is another example due to Anne Bezuidenhout: ‘My son comes into the kitchen from the backyard and when I ask him what he has been doing he replies: “I’ve been playing baseball.” Is what he says true? Well, the game he was playing resembles standard league baseball games only rather remotely. There certainly is no baseball diamond in a our backyard. In the game my son plays with his father and our dog in our backyard, the bases are marked by three trees that stand in a very rough diamond shape with respect to “home plate”, which is itself a rather poorly defined place somewhere at the fourth point of the rough diamond. The game is played only with a pitcher and a batter. When the batter makes it to the base, he leaves an “invisible man” on base and returns to bat. The dog plays in the outfield. Sometimes he returns the ball to the batter and sometimes he chases the runner round the base with the ball in his mouth, but not in any predictable way. Yet this joint activity counts as playing baseball, as playing baseball is understood in this context. So if my son was in fact playing baseball on this understanding, then what my son says is true. (This example is loosely inspired by Moravcsik 1998.)’ (Bezuidenhout 2002: 106) 5. The list of examples is not meant to be exhaustive. There are other interesting examples of context sensitivity. And of course there is vagueness. Yet I would like to distinguish vagueness from the various forms of context sensitivity just discussed, because the hallmark of vagueness, I think, is that even if all features of context were fixed, that is, even if all interests, purposes, background assumptions etc. were made explicit and agreed upon, two equally rational speakers may still disagree about whether a given vague predicate is applicable to a given object or not.
References Aitchison, J. (2003) Words in the Mind: An Introduction to the Mental Lexicon, 3rd edn. (Oxford: Blackwell). Aristotelis Opera, ed. by Immanuel Bekker (1831–1870) Preußischen Akademie der Wissenschaften, Berlin. Atlas, J. (1989) Philosophy Without Ambiguity (Oxford: Oxford University Press). Austin, J. L. (1979a) The meaning of a word. In Austin (1979b), pp. 55–75. Austin, J. L. (1979b) Philosophical Papers, 3rd edn (Oxford: Oxford University Press). Bach, K. (1982) Semantic nonspecificity and mixed quantifiers. Linguistics and Philosophy, 4: 593–605. Bach, K. (1992) Intentions and demonstrations. Analysis, 52: 140–6. Bach, K. (1994) Semantic slack. In S. L. Tsohatzidis (ed.) Foundations of Speech Act Theory: Philosophical and Linguistic Perspectives (London and New York: Routledge), pp. 267–91. Bach, K. (2001) You don’t say. Synthese, 128: 15–44. Barwise, J. and Perry, J. (1983) Situations and Attitudes (Cambridge, MA and London: MIT Press). Bezuidenhout, A. (2002) Truth-conditional pragmatics. Philosophical Perspectives, 16: 105–34. Borg, E. (2004a) Formal semantics and intentional states. Analysis, 64: 215–23. Borg, E. (2004b) Minimal Semantics. (Oxford: Oxford University Press). Borg, E. (2007) Minimalism versus contextualism in semantics. In G. Preyer and G. Peter (eds.), Context-Sensitivity and Semantic Minimalism: New Essays on Semantics and Pragmatics (Oxford: Oxford University Press), pp. 339–59.
308 Nikola Kompa Cappelen, H. and Lepore, E. (2005) Insensitive Semantics: A Defense of Semantic Minimalism and Speech Act Pluralism (Oxford: Blackwell). Carston, R. (2002a) Linguistic meaning, communicated meaning and cognitive pragmatics. Mind & Language, 17 (1&2): 127–48. Carston, R. (2002b) Thoughts and Utterances: The Pragmatics of Explicit Communication. (Oxford: Blackwell). Carston, R. and Powell, G. (2006) Relevance Theory – new directions and developments. In E. Lepore, and B. C. Smith (eds.) (2006) The Oxford Handbook of Philosophy of Language (Oxford: Clarendon Press), pp. 341–60. Cohen, J. L. (1985) A problem about ambiguity in truth-theoretical semantics. Analysis, 45 (3): 129–34. Croft, W. and Cruse, A. D. (2004) Cognitive Linguistics (Cambridge: Cambridge University Press). Elugardo, R. and Stainton, R. J. (eds.) (2004) Ellipsis and Nonsentential Speech. (Dordrecht: Kluwer). Evans, G. (1977) Pronouns, quantifiers and relative clauses (I). The Canadian Journal of Philosophy, 7, 3: 467–536. Evans, G. (1980) Pronouns. Linguistic Inquiry, 11 (2): 337–62. Gauker, C. (forthcoming) Zero tolerance for pragmatics. Synthese, 65: 359–71. Geach, P. (1962) Reference and Generality (Ithaca: Cornell University Press). Goodman, N. (1976) Languages of Art. An Approach to a Theory of Symbols, 2nd edn (Indianapolis: Hackett Publishing Company, Inc.) Grice, P. H. (1989) Logic and conversation. In P. H. Grice Studies in the Way of Words (Cambridge, MA: Harvard University Press), pp. 1–143. Harman, G. (1987) (Nonsolipsistic) conceptual role semantics. In E. Lepore (ed.) New Directions in Semantics (London: Academic Press), pp. 55–81. Reprinted in G. Harman (1999) Reasoning, Meaning, and Mind (Oxford: Clarendon Press), pp. 206–31; quotations are to Harman (1999). Kaplan, D. (1977) Demonstratives. An essay on the semantics, logic, metaphysics, and epistemology of demonstratives and other indexicals. In J. Almog, J. Perry and H. Wettstein (eds.) Themes from Kaplan (Oxford: Oxford University Press), pp. 481–563. Keller, R. And Kirschbaum, I. (2003) Bedeutungswandel: Eine Einführung (de Gruyter Studienbuch). (Berlin: de Gruyter). Lakoff, G. (1970) A note on ambiguity and vagueness. Linguistic Inquiry, 1: 357–9. Lakoff, G. and Johnson, M. (1980) Metaphors We Live By (Chicago: University of Chicago Press). Lewis, D. (1970) General semantics. Synthese, 22: 18–67. Reprinted in Lewis (1983) Philosophical Papers, Vol. 1. (Oxford: Oxford University Press), pp. 189–232; quotations are to Lewis (1983). Lewis, D. (1980) Index, context, and content. In S. Kangar and S. Öhman (eds.) Philosophy and Grammar (Dordrecht: Reidel). Reprinted in Lewis (1998) Papers in Philosophical Logic (Cambridge: Cambridge University Press), pp. 21–44; quotations are to Lewis (1998). Ludlow, P. (2004) A note on alleged cases of nonsentential assertion. In R. Elugardo and R. Stainton (eds.) (2004), pp. 95–108. Moravcsik, J. M. (1998) Meaning, Creativity, and the Partial Inscrutability of the Human Mind (Stanford: CSLI Publications). Neale, S. (1990) Descriptions (Cambridge, MA: MIT Press). Nunberg, G. (1993) Indexicality and deixis. Linguistics and Philosophy, 16: 1–43.
Contextualism in the Philosophy of Language
309
Nunberg, G. (1996) Transfers of meaning. In J. Pustejovsky and B. Boguraev (eds.) Lexical Semantics: Problems of Polysemy (Oxford: Clarendon Press), pp. 109–32. Perry, J. (1997) Indexicals and demonstratives. In R. Hale and C. Wright (eds.) A Companion to the Philosophy of Language (Oxford: Blackwell Publishers), pp. 586–612. Pinkal, M. (1985) Logik und Lexikon – eine Semantik des Unbestimmten (Berlin and New York: de Gruyter). Pinkal M. (1991) Vagheit und Ambiguität. In A. von Stechow and D. Wunderlich (eds.) Semantik: Ein internationales Handbuch der zeitgenössischen Forschung (Berlin and New York: de Gruyter), pp. 250–69. Pylyshyn, Z. (1987) Preface. In Z. Pylyshyn (ed.) The Robot’s Dilemma (Norwood, NJ: Ablex), pp. vii–xi. Quine, W. V. O. (1960) Word and Object (Cambridge, MA: MIT Press). Récanati, F. (2004) Literal Meaning (Cambridge: Cambridge University Press). Reimer, M. (1991) Do demonstrations have semantic significance? Analysis, 51: 177–83. Searle, J. (1978) Literal meaning. Erkenntnis 13: 207–24. Searle, J. (1980) Background of meaning. In J. Searle, F. Kiefer and M. Bierwisch (eds.) Speech Act Theory and Pragmatics (Dordrecht: Reidel Publishing Company), pp. 221–32. Searle, J. (1993) Metaphor. In A. Ortony (ed.) Metaphor and Thought, 2nd edn (Cambridge: Cambridge University Press), pp. 83–111. Stainton, R. J. (1995) Non-sentential assertions and semantic ellipsis. Linguistics and Philosophy, 18: 281–96. Stainton, R. J. (1997) What assertion is not. Philosophical Studies, 85: 57–73. Stainton, R. J. (2005) In defense of non-sentential assertion. In Z. G. Szabó (ed.) Semantics versus Pragmatics (Oxford: Clarendon Press), pp. 383–457. Stanley, J. (2000) Context and logical form. Linguistics and Philosophy, 23/4: 391–434. Stanley, J. (2002) Making it articulated. Mind & Language, 17: 149–68. Stanley, J. (2005) Semantics in context. In G. Preyer and G. Peter (eds.) Contextualism in Philosophy: Knowledge, Meaning, and Truth (Oxford: Oxford University Press), pp. 221–53. Travis, C. (1985) On what is strictly speaking true. Canadian Journal of Philosophy, 15: 187–229. Travis, C. (1996) Meaning’s role in truth. Mind, 105: 451–66. Travis, C. (1997) Pragmatics. In R. Hale and C. Wright (eds.) A Companion to the Philosophy of Language (Oxford: Blackwell Publishers), pp. 87–107. Travis, C. (2006) Insensitive semantics. Mind & Language, 21: 39–49. Zwicky, A. and Sadock, J. (1975) Ambiguity tests and how to fail them. In J. Kimball (ed.) Syntax and Semantics 4 (New York: Academic Press), pp. 1–35. Zwicky, A. and Sadock, J. (1987) A non-test for ambiguity. Canadian Journal of Philosophy 17: 185–8.
15 WJ-40: Issues in the Investigation of Implicature* Laurence R. Horn
Without actually employing the term ‘pragmatics’, Paul Grice laid out the map for modern pragmatic theory in his William James lectures of 1967 by setting out: ... a distinction ... within the total signification of a remark ... between what the speaker has said (in a certain favored and maybe in some degree artificial, sense of ‘said’), and what he has implicated (e.g., implied, indicated, suggested, etc.), taking into account the fact that what he has implicated may be either conventionally implicated (implicated by virtue of the meaning of some word or phrase which he has used) or nonconventionally implicated (in which case the specification of implicature falls outside the specification of the conventional meaning of the words used). (Grice 1989[1967]: 118) While each genus of implicature has undergone rigorous scrutiny from many directions over the ensuing four decades, our attention here will be focused on the non-conventional species, particularly the dominant breed of conversational implicature. Conventional implicature has been buried (Bach 1999) or radically reconfigured (Potts 2005) but is on the rise; see Horn (2007b, 2008) for one attempt at rehabilitation and Gutzmann (2007) for a valuable new extension.
1. The uses of Quantity It has been recognized for millennia that assertions based on the particular, existential, or weak scalar operator some can express true propositions even when the stronger value all is known to hold, although the result (Some dogs are mammals, It is possible that 2+2=4) may appear awkward or anomalous. For Aristotle, whatever holds of all dogs holds ipso facto of some (Topics 109a3), and this view largely prevailed until the mid-nineteenth century, when Sir William Hamilton of Edinburgh posited a semantic distinction 310
WJ-40: Issues in the Investigation of Implicature
311
between two senses of some, the indefinite (at least some) and the semidefinite (some but not all), with the latter as basic: ‘Some, if not otherwise qualified, means some only – this by presumption’ (Hamilton 1860: 254). On this reading of the particular, the statements Some men are bald and Some men are not bald are not only (as for Aristotle) compatible (or ‘merely verbally opposed’), given that their conjunction is logically consistent, but are in fact indistinct: ‘In reality and in thought, every quantity is necessarily either all, or none, or some. Of these the third ... is formally exclusive of the other two’ (ibid.: 261). Hamilton’s lifelong nemesis, Augustus De Morgan, was quick to attack this approach. While acknowledging the existence in ‘common language’ of Hamilton’s ‘presumption’ whereby some conveys some not (not all), De Morgan defended the standard practice of relegating this inference to an extra-logical domain. For both De Morgan and his fellow anti-Hamiltonian John Stuart Mill, the delimiting of some is subject both to the effects of context and speech level and to the speaker’s epistemic state (as signalled by the added emphases below): There are three ways in which one extent may be related to another ... : complete inclusion, partial inclusion with partial exclusion, and complete exclusion. This trichotomy would have ruled the forms of logic, if human knowledge had been more definite. De Morgan 1858: 121) No shadow of justification is shown...for adopting into logic a mere sousentendu of common conversation in its most unprecise form. If I say to any one, ‘I saw some of your children today’, he might be justified in inferring that I did not see them all, not because the words mean it, but because, if I had seen them all, it is most likely that I should have said so: even though this cannot be presumed unless it is presupposed that I must have known whether the children I saw were all or not. (Mill 1867: 501) The implied upper bound (=not all) associated with the ordinary language assertion of some is attributed here to a tacit principle that would need another century before its explicit formulation. An early stab at this principle is due to Strawson (1952: 178–9), who credits this ‘general rule of linguistic conduct’ to ‘Mr H. P. Grice’: ‘One should not make the (logically) lesser, when one could truthfully (and with greater or equal clarity) make the greater claim’. Grice’s own ‘first shot’ (1961: 132) at this ‘general principle governing the use of language’ was that ‘One should not make a weaker statement rather than a stronger one unless there is a good reason for so doing’, later refashioned as his (first) maxim of Quantity (Grice 1989: 26): ‘Make your contribution as informative as is required (for the current purposes of the talk-exchange)’. (For a survey of related, and in some cases independent, formulations of the principle within mid-twentieth-century
312
Laurence R. Horn
philosophy of language, see Horn 1990.) Note in particular the role of the boldfaced codicils above, which recognize the interplay of quantity with other factors, including clarity, quality (truthfulness) and relevance, represented by competing maxims in the Gricean framework. In neo-Gricean frameworks (Horn 1972, 1989, 2004; Gazdar 1979; Levinson 2000), the maxim of Quantity – is canonically1 induced by unilateral entailment relations between lexical oppositions – motivates the establishment of quantity scales such as those in (1): (1)
< and, or >
<no(ne), few/not many, not all> <excellent, good, OK>
Based on such scales, the speaker’s assertion of a relatively weak value Q(uantity)-implicates that she was not in the epistemic position to have asserted any stronger value (to its left) within the same scale (see Horn 2006b for a response). This accounts for the role of context in the cancellation and reinforcement of the upper bound of scalar predications, allows for generalizations across operator types (quantifiers, binary connectives, deontic and epistemic modals, non-embedding predicates), while obviating the need to invoke any lexical ambiguity for the relevant operators (e.g. inclusive vs. exclusive disjunction). On this approach, scalar values are lower-bounded by their literal meaning (‘what is said’) and upper-bounded by quantity-based implicature. Thus the ‘one-sided’ meanings delivered by the linguistic semantics may be pragmatically enriched to yield the ‘two-sided’ understandings typically communicated: (2)
a. Pat has 3 children. b. You ate some of the cake. c. It’s possible she’ll win. d. He’s a knave or a fool. e. It’s warm.
‘... at least 3 ...’ ‘... some if not all ...’ ‘...at least ◊...’ ‘... and perhaps both’ ‘... at least warm ...’
‘... exactly 3 ...’ ‘... some but not all ...’ ‘...◊ but not certain...’ ‘... but not both’ ‘... but not hot’
The alternative view on which each scalar predication in (2) is lexically ambiguous between one-sided and two-sided readings is ruled out by the Modified Occam’s Razor principle (Grice 1989: 47): ‘Senses are not to be multiplied beyond necessity’. (See Carston 2002 for a sceptical view of this heuristic and Bontly 2005 for an empirically supported defence of M.O.R. based on its role in language acquisition.)
[affirmations]
[particulars] Figure 15.1
distinction in QUALITY
313
[negations]
contraries
A
distinction in QUANTITY
[universals]
WJ-40: Issues in the Investigation of Implicature
E
contradictories
subcontraries
I
O
A: all/every F is G
E: no F is G
I: some F is/are G
O: not every F is G, some F is not G
Square of Opposition (Apuleius and Boethius, after Aristotle)
The neo-Gricean approach to the subcontraries allows us to reconstruct Aristotle’s notion of ‘merely verbal’ opposition between I and O vertices of the Square of Opposition as a relation of mutual quantity implicature, as shown in figure 15.1. Scalar implicature provides a natural account of the lexicalization asymmetry of the Square of Opposition, an asymmetry displayed as table 15.1.
Table 15.1
Lexicalization and the three-cornered square
Determiners/ quantifiers
Quant. adverbs
Binary quantifiers
Correlative conjunctions
Binary connectives
A:
all α, everyone
always
both (of them)
both...and
and
I:
some α, someone
sometimes one (of them)
either...or
or
E:
no α, no one
never
(=all¬/¬some)
(=always¬) (=both¬/¬either) (=[both...and]¬) (=and¬)
O:
neither (of them) neither...nor
nor
*nall α, *neveryone *nalways *noth (of them) *noth...nand *nand (= some¬/¬all) (= ¬always) (= either¬/¬both) (= [either...or]¬) (= and¬/¬or)
314
Laurence R. Horn
Although some does not contribute the same semantic content as some not (not all), the use of either of the two values typically results in communicating the same information in a given context, ‘some but not all’. The relation of mutual quantity implicature holding between positive and negative subcontraries results in the superfluity of one of the two for lexical realization, while the functional markedness of negation (see Horn 1989 for extensive documentation) predicts that the unlexicalized subcontrary will always be O rather than I. I have argued (Horn 1972: chapter 4; 1989: §4.5; to appear) that this pragmatic account of the ‘three-cornered square’ is more general and more explanatory than the rival theories that either dismiss the asymmetry as uninteresting or restrict it to the determiners and quantificational operators while bypassing other operator types (e.g. connectives, adverbs, and modalities) along with intermediate values that can be mapped onto the Square of Opposition. The neo-Gricean position on scalar predicates has been vigorously challenged by relevance theorists (e.g. Carston 2002, 2004, 2005 and work reviewed therein), who take scalar predications to involve not lexical but propositional ambiguity, with the pragmatically enriched two-sided meanings constituting ‘explicatures’. While the standard neo-Gricean line (Horn 1972, 1989; Levinson 2000) treats all the cases in (2) homogeneously, there is considerable evidence that this analysis is not actually tenable for number words, as in (2a). Rather, such predications are semantically underspecified, rather than assigned the weak, ‘at least n’ values by linguistic means; the propositional content is filled in only through reference to the context of utterance. Arguments for this position, originally given in work by Carston (1988) and Koenig (1991), are ratified and extended in later work including Horn (1992), Geurts (1998) and Bultinck (2005). But crucially, what’s sauce for the cardinals is not necessarily sauce for the other scalar values. Thus, while Ariel (2004, 2006) disputes an implicature-based account of the upper bound of most-statements (i.e. the move from most F are G to ‘not all F are G’) in part on the basis of a putative parallel between most and the cardinals regarding the status of the upper bound, I have argued (Horn 2006a: §4) that most is crucially distinct in behaviour from the cardinals and that its meaning, like that of some, should be assigned the standard neo-Gricean account (unilateral semantics cum upper-bounding scalar implicature). The strongest evidence for such a distinction is the fact that a simple negative answer to a general scalar question, as opposed to one involving a cardinal value, always returns a ‘less than’ meaning, since this context selects descriptive and not metalinguistic negation. If you ask me whether most of the students passed, my negative response commits me to the proposition that 50% or fewer passed, not to the disjunction that either 50%-or-fewer passed or else all of them did. Yet it is just such a disjunction that I must be asserting if my reply negates the proposition that ‘50% to 99% of the students passed’ as on Ariel’s semantic-upper-bound account.
WJ-40: Issues in the Investigation of Implicature
315
On the other hand, if you asked me whether ten students passed and I knew that fifteen did, I must first determine whether you were asking me if at least ten passed or exactly ten passed before knowing whether to answer positively or negatively; a simple ‘No’ response to ‘Did ten of the students pass?’ would commit me to either ‘Fewer than ten passed’ or ‘Either fewer or more than ten passed’, depending on the context. Similarly, I would report that ‘I am surprised that most of the students failed’ only if I had expected at least half of them to pass, while my report that ‘I am surprised that five students failed’ is consistent with either a more pessimistic or more optimistic expectation. In addition to the linguistic evidence collected in Horn (2006a) and other work, a now considerable body of empirical work has confirmed that the acquisition and processing of cardinals differs along a variety of parameters from that of other scalar values; note especially the studies in Papafragou and Musolino (2003) and Hurewitz et al. (2006), as well as Papafragou and Schwarz (2006) for empirical challenges to Ariel’s findings on most. Thus, while pride of place must be given to the important work of Carston, the behaviour of cardinals must somehow be distinguished from that of their inexact scalar cousins. It is not obvious to me how the unitary explicaturebased programme for all scalar operators is equipped to draw the necessary distinctions here, any more than is the approach in Levinson (2000: 87–90), which retains the original (Horn 1972) neo-Gricean line for both cardinal and general scalar predications. If (non-cardinal) scalar assertions are upper-bounded by scalar implicature, just what is it that a speaker quantity-implicates against? There is a major split on this issue between weak and strong treatments. Gazdar (1979) and Levinson (1983, 2000) are strong epistemic theorists, taking a speaker who asserts p(i) to implicate that she knows ¬p(j), where j>i. For weak theorists, following Mill (1867) above, this implicature is a two-stage process: in telling you some of the students in the class are freshmen, I directly implicate that I don’t know/believe all of them are; it’s only if I assume that you assume my full knowledge of the situation (e.g. that I’ve looked through all the pre-registration forms) that I will implicate, and you will be licensed to infer, that I know not all of them are freshmen. (This last step embodies what Geurts 2009 has called the ‘Competence Assumption’.) Weak epistemic theorists include Soames (1982), Hirschberg (1985), Horn (1989), Sauerland (2004), and Geurts (2009, to appear). But which weak treatment is warranted? Hirschberg takes the utterer of p(i) to implicate a disjunction: either the speaker knows the stronger proposition doesn’t hold or doesn’t know whether or not it holds (1985: 79–80). When S affirms that p(i) holds, ‘S believes that higher p(j) are false or S does not know whether higher p(j) are true or false’ (ibid.: 81). But this disjunction – S either knows p is false or doesn’t know whether or not p is true – in fact reduces to the proposition that S does not know (for a fact) that p(j) is true, the position
316
Laurence R. Horn
adopted in Horn (1989: chapter 4). Essentially, Hirschberg takes S’s implicature to be the disjunction (K¬p) ⋁ (¬Kp & ¬K¬p). This amounts to the claim that either nobody won or somebody-but-not-everybody won, equivalent in turn to the claim that not everybody won, since a disjunction of the E vertex of a logical square with the conjunction of the I and O vertices is equivalent to O, i.e. the contradictory of A (figure 15.2): Kp
K¬p
A
E
I
O
¬K¬p
¬Kp
(I ⋁ E, as a disjunction of contradictories, is true by definition.) This result is intuitively correct: In saying it’s warm, I implicate that I don’t know for a fact that it’s hot. If you believe I know the actual temperature, you will strengthen this to infer that I’m communicating that (I know) it isn’t hot, but since there’s no guarantee of my epistemic security, this can’t be a firstorder implicature. Hirschberg (1985: chapter 5) represents the content of the relevant scalar implicatures in the form ‘¬BEL (...)’, with a strengthening of ‘¬KNOW’ to ‘¬BEL’. But this strengthening is not warranted, for the following reason: I can say ‘It’s warm, and I believe it’s hot’, but not (without anomaly) ‘It’s warm and I know it’s hot’. If I knew it was hot, I should have said so. But if I just believe it’s hot, I can’t really assert that it is without violating the second Quality maxim. I can only assert what (I believe) I know, not what I (merely) believe. What the utterance of p(i) implicates, ceteris paribus, is just that the speaker doesn’t know that p(j), for any p(j) stronger than p(i). In sum, the use of a weaker value W (e.g. some, possible) suggests that for all the speaker knows no stronger value – and especially not the strongest value S – on the same scale (all, certain) could have been substituted salva veritate. a’s utterance of ... W ... implicates not K a¬(S), i.e. that a knows the stronger counterpart ... S ... is false, but only (ceteris paribus) that ¬Ka(S).
2. Implicatures and ‘strengthening’ In addition to the question of what is implicated, recent work has also addressed the issue of when and how what is implicated is implicated. In earlier work on the projection problem for implicatures, Gazdar (1979)
WJ-40: Issues in the Investigation of Implicature
317
argued that scalar implicatures (henceforth SIs) in particular are blocked in embedded contexts, based on the observation that a standard upper-bounding implicature like that from (the assertion of) (3a) to (3b) seems to disappear when the scalar predicate falls within the scope of a logical operator, as it does in (4a), which does not implicate (4b): (3) a. Paul ate some of the eggs. b. [For all the speaker knows] Paul did not eat all of the eggs. (4) a. It is not the case that Paul ate some of the eggs. b. [For all the speaker knows] It is not the case that Paul did not eat all of the eggs. [= He ate all of them.] As observed by Hirschberg (1985), however, this cannot be a fact about embedded environments in general, since substituting e.g. ‘It is true that’ for ‘It is not the case that’ in (4a/b) will restore the implicature. On Hirschberg’s account, SIs are blocked by overt negation alone. But while Gazdar’s approach blocks too many implicatures, Hirschberg’s blocks too few. Without fully making the case, I suggested (Horn 1989: 233–4) that SIs are blocked in downward entailing (DE) contexts. But, as Levinson points out, it is not really a matter of blocking the implicatures generated by the positive scale so much as predicting the implicatures induced by the inverse scale, given the scale-reversing properties of negation and other downward entailing operators described in Fauconnier (1975): ‘[T]he apparent blockage is due to the fact that negatives reverse scales and so we get different implicatures, which themselves survive negation’ (Levinson 2000: 254–5). Thus, in asserting that Paul didn’t eat many of the eggs, I implicate (ceteris paribus) not that he didn’t eat all of them (since the positive scale will not be relevant here), but that he didn’t eat none of them, i.e. that he ate some (since the scale <none, not many, not all> is now operative). More recently, Chierchia (2004) has argued, based on the interaction of negation and disjunction in complex sentences, that SIs are computed locally by semantic composition rules rather than read off utterances globally and that they are hence part of the grammar, instead of constituting ‘merely’ pragmatic effects derived from general principles of rational interchange. For Chierchia, the ‘suspension’ of SIs in DE contexts and the licensing of NPIs in the same environments (see Ladusaw 1980) represent a parallel effect of strengthening. I will not address here (but see section 3, below) the general issues posed by the compositionality problems raised by Chierchia, or by Fox (2006) in a related localist theory utilizing covert exhaustification; see Sauerland (2004), Spector (2006), Russell (2006), and Geurts (2009, to appear) for various Grice-compatible approaches to the locality data, Reinhart (2006) for a
318 Laurence R. Horn
mixed approach, and King and Stanley (2005) for a different route. (Geurts’s work is especially useful for its review of the comparison of the various menu options.) But the basis for Chierchia’s notions of ‘blocking’ and ‘strengthening’ are worth examining more closely, and it is to this examination that I now proceed. Cherchia’s claim that DE contexts block, rather than reverse, implicature derives from his perception that while positive scales induce ‘direct’ SIs, negative and DE contexts generate only ‘indirect’ SIs, which ‘appear to be generally somewhat weaker and flimsier than their positive counterparts’ (2004: 58). Thus for Chierchia, there is an asymmetry not acknowledged in the standard neo-Gricean account, which incorrectly predicts – on the basis of the scales in (1) – a parallel between the positive case in which the assertion of (5a) will (ceteris paribus) implicate (6a), resulting in the speaker’s communicating (5b), and the negative case in which the assertion of (6a) will (ceteris paribus) implicate (5a), resulting in the speaker’s communicating (6b). (5)
a. Some F are G. (6) b. Some Fs are Gs, but not all are.
a. Not all F are G. b. Not all F are G, but some are.
Similarly, Chierchia (2004: 69) maintains, ‘Our intuitions concerning the implicature of sentences like [I don’t have many matches left] are somewhat shaky. In particular, such a sentence may or may not implicate that I have some matches left.’2 Of course, the ‘may or may not’ codicil is endemic to implicature calculation, given the nature of cancellability and indeterminacy on the Gricean account. But is there any evidence that the implicature in the positive case is more direct or stronger than in the negative, or that the scalar effects in (8) are more robust than those in (9)? As it happens, the Oxford English Dictionary’s first citation for the term ‘scalar implicature’ in the literature is for a passage in which it is observed that ‘not all implicates not none, i.e. some’ (Horn 1972: 96). Nor is this premise restricted to neo-Griceans. In one of the earliest applications of Gricean reasoning in the linguistic literature, Chomsky (1972: 112) takes the inference from not many to some in contexts like (7) to be not weaker but actually stronger than a garden-variety conversational implicature. (7)
a. Not many arrows hit the target. b. Some arrows hit the target.
For Chomsky, ‘Sentence [(7a)] (equivalently, Few arrows hit the target) presupposes that some arrows hit the target.’ Similarly, (8a) is taken to presuppose (8b): (8)
a. {Not much/Little} enthusiasm was shown for that project. b. At least some enthusiasm was shown.
WJ-40: Issues in the Investigation of Implicature
319
I have argued (Horn 1972: chapter 2) that the relation in (7a) and (8a) must be (scalar) implicature, not presupposition. But there is no reason to believe that the inference with these pairs is on shakier or flimsier ground than that with (5) or other positive scalars. Absent the posited asymmetry between positive and negative scales, we can’t really claim that SIs are suspended in DE contexts, but only that (as Levinson points out) the SIs induced there are based on negative scales as opposed to positive ones. Given DE operators’ twin role as licensers of NPIs and putative blockers of SIs, Chierchia seeks to unify the two phenomena under the umbrella of compositional semantics. In particular, he argues, just as NPIs serve to strengthen a negative implicature (as in the widening-cum-strengthening account of any due to Kadmon and Landman 1993), so too ‘implicatures must lead to strengthening’ (Chierchia 2004: 70). But do NPIs invariably strengthen negative force? It is undeniable that any in both negative polarity and generic/nonepisodic (free choice) contexts effectively strengthens a simple indefinite (Kadmon and Landman 1993) or contributes an end-of-scale even-type meaning (see Lee and Horn 1994; Lahiri 1998; Horn 2000b) as in (9), and that ever – the temporal analogue of any – and minimizers such as those in (10) likewise serve to reinforce negation: (9)
a. I don’t have {potatoes/any potatoes}. b. {An owl/Any owl} eats mice.
(10)
a. Robin didn’t {eat a bite/drink a drop}. b. Dana isn’t saying a word about it.
However, other NPIs, in particular those not involving indefinites, do not obviously result in strengthening. In his valuable studies of the lexical semantics of polarity, Israel (1996, to appear) distinguishes emphatic NPIs (any, ever, at all, and the minimizers) from attenuating NPIs (e.g. much, overmuch, long, be all that, any too, great shakes, be born yesterday, trouble to, mince words); the attenuators do not strengthen negative force. Among the attenuating NPIs are negative modals (need, Dutch hoeven), yet, anymore (for the majority dialect), and until. While He won’t ever recover is a stronger negative than He won’t recover tout court, She hasn’t recovered yet mitigates rather than strengthens the negative force of She hasn’t recovered, just as He doesn’t read much weakens the force of He doesn’t read.3 If NPI licensing does not necessarily yield strengthening, what of the purported strengthening effect of scalar implicatures? To return to the examples in (5) and (6): Some but not all Fs are Gs is, to be sure, more informative and more specific than Some F are G sans implicature, but it does not result in a stronger positive assertion, nor does Not all F are G but some are result in a stronger (negative) claim than Not all F are G. By implicating the upper
320 Laurence R. Horn
bound, the speaker in effect weakens the positive or negative force of what is said. In the dualistic model of implicature I have been urging over the last two decades (Horn 1984, 1989, 2000a, 2007a), those implicatures based on the Q Principle (‘Say enough’, a generalization of Grice’s first maxim of Quantity) are distinguished from those based on the R Principle (‘Don’t say too much’, a correlate of Zipf’s principle of least effort encompassing the second Quantity maxim, Relation, and Brevity). While Q-based implicatures do not strengthen the force of an assertion, R-based implicatures do. These include the ascription of the ability to perform some action implicating the successful performance of that action (Dana was able to solve the problem), the ‘perfection’ of a sufficient if-condition to a necessary and sufficient iff-condition (If you mow the lawn, I’ll give you $5), and the conventionalized narrowing of a word’s extension from a set to a salient or prototype member or subset, including the use of vague expressions as euphemisms for what one would prefer to leave unsaid. (Compare the I [for Informativeness] heuristic of Levinson 2000 and the ‘inference to the best interpretation’ it invokes.) Evidence for a Manichaean pragmatics (Horn 2007a) is provided by the mirror image principles of synonymy-avoidance (an R-based tendency motivated by the speaker’s economy) and homonymy-avoidance (a Q-based tendency motivated by the hearer’s economy) as factors in language acquisition and linguistic change, as well as the complementary processes of linguistically motivated Q-based narrowing (rectangle +> ‘non-square’, finger +> ‘nonthumb’, friend +> ‘non-lover’) and socially motivated R-based narrowing (drink +> ‘alcoholic drink’, temperature +> ‘fever’, ‘umfriend’ +> ‘lover’). Related to lexical narrowing, and again motivated by social considerations – in particular, those relating to euphemism and respect for negative face as described in Brown and Levinson (1987) – is the R-based strengthening of negative expression. In Bosanquet’s words (1911: 281), ‘The essence of formal negation is to invest the contrary with the character of the contradictory’. Speakers across a wide range of languages tend to weaken the force of their intended negative judgments, counting on hearers to fill in the intended stronger negative evaluation. In English, the resultant contrary negatives in contradictory clothing include the varieties of affixal negation, simple litotes, and ‘neg-raising’ instantiated in (11a–c) respectively; see Horn (1989: chapter 5) for extensive discussion. (11)
R-based negative strengthening (a) contrary readings for affixal negation (conventionalized/lexicalized strengthening) He is unhappy. She was unfriendly. I disliked the movie.
(stronger than ¬[He is happy]) (stronger than ¬[She was friendly]) (stronger than ¬[I liked the movie])
WJ-40: Issues in the Investigation of Implicature
321
(b) litotes/understatement in simple denials (online/non-conventionalized strengthening) He’s not happy with it. (stronger than ¬[He’s happy with it]) I don’t like ouzo. (stronger than ¬[I like ouzo]) I’m not optimistic that φ. (stronger than ¬[I’m optimistic that φ]) (c) neg-raising effects (strengthened understanding as a convention of usage) I don’t believe it’ll snow. (≈ I believe it won’t) I don’t want you to go. (≈ I want you not to go) It’s not likely they’ll win. (≈ It’s likely they won’t) In each case a general, formally contradictory negation is strengthened to a specific, contrary understanding; where the constructions differ is in the degree of conventionalization of this strengthening inference. Since Aristotle, it has been recognized that affixal negatives conventionally strengthen to contraries, whence the readings in (11a). In litotes, I say that I don’t like ouzo (11b), or that I’m not exactly thrilled with your advice, precisely to avoid acknowledging my antipathy directly; at the same time, I count on your willingness to fill in my intended R-strengthened (contrary) interpretation rather than simply taking me at my (contradictory) word. In an embedding environment, this same practice is responsible for the negraising effect seen in (11c), where a negative operator with semantic scope over certain predicates of opinion, desire, or likelihood is understood as if it had lower-clause scope. The contrary meaning (‘x disbelieves that p’, ‘x believes that not-p’) is sufficient but not logically necessary to establish the truth of the contradictory (‘x does not believe that p’), yet it is treated as if it were necessary – not surprisingly, since it represents the inductively salient case that makes the contradictory true and since there may be social constraints against direct expression of the stronger contrary (see Horn 1989, 2000a). Carston (2002: chapter 3, 2005) rejects the distinction between R-based and Q-based implicature as illusory, on the grounds that ‘there is a strengthening of communicated content from “at least some” to “just some” ’ (Carston 2005: 314–15) that is entirely parallel to, say, the strengthening of not believing that p to believing that not-p. One question arising here is the extent to which relevance theory itself is truly unitarian, given the trade-off between effort and contextual effect: ‘Human cognitive activity is driven by the goal of maximizing relevance: that is ... to derive as great a range of contextual effects as possible for the least expenditure of effort’ (Carston 1995: 231). But in any event, does the upper-bounding effect of Q-based, in particular scalar, implicature amount to strengthening, as maintained by both Carston and Chierchia? Does a scalar implicature, by upper-bounding an assertion, in fact strengthen it? In particular, what exactly do we mean by ‘strength’?
322 Laurence R. Horn
In fact, while R-based implicature increases both the informative content and rhetorical strength (positive or negative) of the assertion, what is communicated as a result of Q-based upper-bounding, while it is more specific and hence informationally stronger than the unbounded utterance, is not rhetorically stronger than what is said (i.e. the basic utterance without the implicature). Thus, while some is consistent with all, some but not all (let’s call it some|) is inconsistent with all. Thus some| F are G, while unilaterally entailing some F are G, yields a more specific but not a stronger positive assertion. Further, as Michael Israel (p.c.) points out, a statement with some is clearly stronger than one with some| in the terms of Ducrot’s argumentation theory (see, e.g., Anscombre and Ducrot 1983). Thus, a sentence like (12a) represents a stronger argument for the (underlined) conclusion than does the more specific but rhetorically weaker (12b): (12)
a. I’ve graded some of the exams, so it’s time for a break. b. ?I’ve graded some, but not all, of the exams, so it’s time for a break.
As Israel also notes, already is possible in the former but not in the latter, given the incompatibility of directly expressing the Q-based implicature with the suggestion that things are ahead of schedule: (13) a. I’ve already graded some of the exams (so let’s go out for a beer). b. #I’ve already graded some, but not all, of the exams (so let’s go out for a beer). Another argument for distinguishing the two kinds of strength is provided by the distribution of rank orders (Lehrer 1974; Hirschberg 1985; Horn 1989, 2000b), which are related to, but distinct from, true scales; (14) incorporates a proposed notational differentiation and (15) provides additional examples of rank orders. (14)
True scales
Rank orders
<scalding, hot, warm>
<> <<win|place|show>> <<dead|sick>>
(15) <> <> <<senior|junior|sophomore| freshman>> <<α φ’d| α almost φ’d>>
<<married|engaged>> <> <> <<α didn’t φ|α barely φ’d>>
WJ-40: Issues in the Investigation of Implicature
323
In a scale of the form , ... Y ... unilaterally entails ... X ... : if it’s hot, it’s warm; in a rank order <>, ... Y ... unilaterally entails ... ¬X ... : if they’re married, they’re not engaged; if he’s a colonel, he’s not a lieutenant. If she’s a full professor, it’s false that she’s an assistant professor – but true that she’s at least an assistant professor. (S is at least P can be true while S is P is false.)4 Similarly, compare these exchanges between players in (non-wild card) poker, where having a full house outranks but precludes having a flush: (16) A: Do you have a flush? B: {No/#Yes} (in fact) I have a full house. (17) A: Do you have at least a flush? B: {Yes/#No} (in fact) I have a full house. Rank-ordered items essentially build in the upper bound: Chris has a full house and Chris has a flush are equally informative, in that neither entails the other. Similarly for Dr. Doolittle is a full professor vs. Dr. Doolittle is an associate professor. Yet the first member of each pair is rhetorically stronger in asserting that the higher rank holds. Once again, we see that rhetorical strength is distinct from informative strength. In addition to challenging – or at least problematizing – the claim that scalar implicature results in strengthening, these observations provide additional support for a dualistic or Manichaean approach to pragmatic inference; a minimal pair illustrating the difference between R-based and Q-based implicature is provided in (18): (18)
a. Not only was she able to solve the problem, (in fact) she solved it. (a was able to ϕ R-implicates a ϕ‘d) b: #Not only is it possible that she solved the problem, (in fact) it’s (possible but) not certain that she solved it. (it’s possible that p Q-implicates it’s not certain that p)
3. Pragmatic intrusion and ‘what is said’ Recent years have witnessed the formulation of a partial consensus regarding semantic underspecification and pragmatic enrichment, a consensus that agrees in rejecting the conception bequeathed by Grice that the pragmatics can be simply ‘read off’ the semantics, while disagreeing on the conclusions to be drawn for the Gricean notion of ‘what is said’. When we turn from the relatively straightforward cases of reference fixing and ambiguity resolution countenanced by Grice himself to the more problematic phenomena of completion and saturation or free enrichment (see Bach 2001; Recanati 2001, 2004; Carston 2002, and references therein, as well as the
324
Laurence R. Horn
relevant chapters in Horn and Ward 2004), it is clear we must accept what Bach (2005) terms the ‘contextualist platitude’: Linguistic meaning generally underdetermines speaker meaning. That is, generally what a speaker means in uttering a sentence, even if the sentence is devoid of ambiguity, vagueness or indexicality, goes beyond what the sentence means. Thus, the speaker uttering the non-bracketed material in each example in (19) may well communicate the full sentences indicated, enriched by the bracketed addenda. As seen from the cancellability diagnostics on display in (20), however, this process is pragmatic in character, even though its result is the computation of truth-conditionally relevant propositions that are not directly expressed: (19)
a. b. c. d. e.
I haven’t had breakfast {today}. John and Mary are married {to each other}. They had a baby and they got married {in that order}. Robin ate the shrimp and {as a result} got food poisoning. Chris is ready {for the exam}.
(20) a. John and Mary are married, but not to each other. b. They had a baby and got married, but not necessarily in that order. The demarcation of the explicit is no trivial matter; the subtitle of Carston (2002) is, after all, The Pragmatics of Explicit Communication. A faithfully Gricean theory can accept neither Levinson’s picture of implicatures as benignly informing literal content (what is said) nor the notion of explicature as applied by relevance theorists. As Bach has stressed, the typically communicated propositions in (19) cannot literally be ‘explicatures’ because they are not explicit; at the same time, they are not implicatures but ‘implicitures’, implicit in, rather than read off, what is said. Objecting to the premise that a pragmatically supplemented proposition is not communicated explicitly, Carston (2005: 311) observes that in many cases there will then be nothing (or, more exactly, no proposition) on our account that is communicated explicitly. But if we can agree that no fullfledged proposition is literally expressed and explicitly communicated, what do we gain by then taking the most nearly explicitly communicated proposition to constitute an explicature, with the acknowledged suggestion of explicitness? Carston defends the term as justified on the grounds that ‘this partially pragmatically-determined proposition is about as explicit as we ever are, or can be’ in (19)-type examples. But even if we grant that filling in the bracketed material in such cases yields the closest thing we
WJ-40: Issues in the Investigation of Implicature
325
can get to an explicitly communicated proposition, the theory of meaning is not an enterprise like horseshoes or hand grenades, where closeness counts. Whether or not we invoke ‘truth-conditional pragmatics’ (to adopt Recanati’s term), it must be borne in mind that implicatures – whether conventional or conversational – are propositions that have their own truthconditions. In uttering a given sentence in a given context, the speaker may communicate more than one truth-evaluable proposition, and what is said (in the Gricean sense) may correspond to a proposition radical that is not itself truth-evaluable, as with Chris is ready (Bach 2001, 2006). At the same time, there are the intuitions of the localists – Levinson (2000: §3.3) and Chierchia (2004), among others – and the problem itself, formulated through examples in which the implicatures licensed by a sub-expression seem to affect the truth-conditions of the larger expression in which they are embedded, as in the celebrated example from Cohen (1971) of conjunctions within conditional protases, where (21a, b) are intuitively assigned distinct truth-conditions. (21)
a. If the old king dies and a republic is declared, I’ll be happy. b. If a republic is declared and the old king dies, I’ll be happy.
Analogous cases noted by Wilson and Carston involve comparatives and negations: (22) It is better to meet the love of your life and get married than to get married and meet the love of your life. (23)
a. He didn’t drive home and drink 3 beers; he drank 3 beers and drove home. b. Driving home and drinking 3 beers is better than drinking 3 beers and driving home.
Even if we follow King and Stanley (2005) in explaining away (23a) as an instance of metalinguistic negation (Horn 1989), this won’t extend to the evidently parallel cases. Some concessions to the localists appear inevitable, despite the efforts of diehards like King and Stanley (2005), whose response to the evident presence of what they term strong pragmatic effects is to massage the logical form of the original sentences to reveal independent semantic motivation for what would otherwise require localist pragmatic accounts. As stressed by Geurts (2009), the key issues are the plausibility of that independent motivation and the nature of the explanation it provides. One factor on which there is some agreement is the role of focal stress (Horn 2004, 2006a; King and Stanley 2005). In the scalar-antecedent conditionals in (24), both
326 Laurence R. Horn
Levinson and explicature theorists would build the stronger (bilateral) meaning (e.g. some but not all, warm but not very warm) into what is said:5 (24)
a. If some of my friends come to the party, I’ll be happy – but if all of them do, I’ll be in trouble. b. If it’s warm, we’ll lie out in the sun. But if it’s very warm, we’ll go inside and sit in front of the air-conditioner. c. If you’re convicted of a felony, you’ll spend at least a year in jail. And if you’re convicted of murder, you’ll be executed. d. If you’re injured, the paramedics will take you to the nearest trauma centre. But if you’re fatally injured, you’ll be taken to the morgue.
But in each of these contexts, it’s only when the stronger scalar is reached that the earlier, weaker one is retroactively adjusted to accommodate an upper bound into its semantics, e.g. with some being reinterpreted as expressing (rather than merely communicating) ‘some but not all’ or injured ‘non-fatally injured’. The same issues arise for other applications of the pragmatic intrusion argument. Thus, Levinson (2000: 210) extends the argument from conditionals like (21) to because clauses, based on such examples as (25): (25)
a. Because he earns $40,000, he can’t afford a house in Palo Alto. b. Because he’s such a fine friend, I’ve struck him off my list. c. Because the police recovered some of the missing gold, they will later recover it all.
But Levinson’s cases are heterogeneous. Example (25a) involves a cardinal, which as we have seen is indeed plausibly taken to involve an adjustment of what is said. In (25b), such a fine friend involves conventionalization of the intended sarcasm; compare ?Because he’s so considerate, I fired him. Finally, the all in the main clause of (25c) forces the reprocessing of the some in the first clause as ‘some but not all’, a reading triggered by the focal stress on some. Without the all or another context-forcing continuation and without focus, this narrowing is difficult or impossible to obtain. In general, the distribution of such because clauses is quite constrained, in particular for the non-cardinal scalar cases in which the implicated upper bound is taken to be the reason for the truth of the second clause and in which no reprocessing is forced by the affirmation of a stronger value, whence the oddness of the examples in (26): (26)
a. Because the police recovered some [i.e. only some] of the gold, the thieves are expected to return later #(for the rest). b. #Because it’s warm out [i.e. because it’s warm-but-not-hot], you should still wear a long-sleeved shirt.
WJ-40: Issues in the Investigation of Implicature
327
c. #Because you ate some of your gruel [i.e. and not all of it], you get no dessert. Typically, both focus and contrast of scalar values are required, serving to convert a scale to a rank order of incompatibles, just as at least and yet coerce rank orders into scales, as we saw in section 2. So what can we conclude on the issue of implicatures and pragmatic intrusion? While scalar implicature is derived for the most part by the usual Gricean means, that original programme must evidently be modified to allow for a restricted range of cases of locality effects in which upper bounding can enter into the reinterpretation of what scalar operators express, although this reinterpretation is itself pragmatic in nature. (Geurts 2009 provides a useful survey of the domain, including a useful distinction between marked L[evinson]-type cases and unmarked C[herchia]-type cases of putative locality effects and convincing arguments on why only the former represent true problems for classical and neo-classical Gricean theories of implicature; see also Geurts to appear.) In any case, pace Levinson (2000), generalized conversational implicatures cannot be default inferences, both because they are not inferences – by definition an implicature is an aspect of speaker’s meaning, not hearer’s interpretation (see Schwenter 1999: 26; Bach 2001, 2006; and Saul 2002b) – and because they are not defaults. This last point is especially worth stressing in the light of much recent work in experimental pragmatics undertaken by Ira Noveck, Dan Sperber, Richard Breheny, and their colleagues (see e.g. Noveck and Posada 2003; Bott and Noveck 2004; Breheny, Katsos and Williams 2006) arguing that – contra ‘neo-Gricean theory’, in their styling – children and adults do not first automatically construct implicature-based enriched meanings for scalar predications and then, when the ‘default’ interpretation is seen to be inconsistent with the local context, undo such meanings and revert to the minimal implicature-free meaning. To the extent that the empirical work on the processing of implicature recovery can be substantiated and extended, this is a very interesting result, but any automatic enrichment or default interpretation accounts threatened by such work are not those of the actual Gricean tradition. I see no reason to revisit the distinction between generalized and particularized implicature as Grice originally formulated it (1989: 37, emphases added): I have so far considered only cases of what I might call ‘particularized conversational implicature’... in which an implicature is carried by saying that p on a particular occasion in virtue of special features of the context, cases in which there is no room for the idea that an implicature of this sort is normally carried by saying that p. But there are cases of generalized conversational implicature. Sometimes one can say that the use of a certain form of words in an utterance would normally (in the
328
Laurence R. Horn
absence of special circumstances) carry such-and-such an implicature or type of implicature. An implicature may arise in a default context without thereby constituting a default or automatic inference.6 Nor should this be surprising: I shave every morning unless it’s a weekend day at home, but this does not render my shaving on work days ‘automatic’ – I still have to get out the razor and actually scrape away.
4. On what is said (again): a defence of austerity A bone of perennial contention in neo- and post-Gricean pragmatics is the proper treatment of ‘what is said’. Relevance theorists (e.g. Carston 2002) have questioned the utility of this notion (to the extent that it cannot be identified with the RT notion of explicature); Recanati (2001) distinguishes a Gricean notion of what-is-saidmin from what-is-saidmax,with only the latter playing a significant role within his framework of truth-conditional pragmatics; and even the self-described ‘semantic minimalists’ Cappelen and Lepore (2005) endorse an inflationary view of what is said, incorporating pragmatically inferred expansions. To be sure, Grice’s notion of what is said cannot be accepted as is. For one thing, as Bach (2001) and Saul (2002a) have stressed, we can – and should – give up Grice’s overly restrictive condition that saying something entails meaning it, i.e. that we can’t say what we don’t mean. This constraint becomes implausible when we consider slips of the tongue, non-literality, and performances or rehearsals. (Recall in this connection the distinction between the locutionary and illocutionary acts from Austin 1962: saying ≠ stating.) Communicative intention does not determine what is said. This point aside, what is the fate of the Gricean notion of what is said? As argued by Bach (2001), Saul (2002b, 2006) and Borg (2004), the death-knell for a relatively orthodox or (in Saul’s terms) austere conception of what is said may be premature. Bach (2001) has advanced the ‘syntactic correlation constraint’, based on Grice’s position (1989: 87) that what is said must correspond to ‘the elements of [the sentence], their order, and their syntactic character’; aspects of enriched content that are not directly linked to the utterance cannot be part of what is said. As we have noted, many have been sceptical of this view, from Cappelen and Lepore (2005) to Recanati (2001, 2004) and the relevance theorists. In the words of Carston (2005: 310; see also Carston 2002: §2.5 and Carston 2004), ‘It is hard to see what this conception of “what is said” buys one.’ What does ‘what is said’ buy one?7 As the following travelogue suggest, it depends on who – and where – one happens to be. The first stop in our grand tour through the space–time continuum is twenty-first-century New York, with an episode from The White Rose
WJ-40: Issues in the Investigation of Implicature
329
(Korelitz 2005: 296), a reimagining of Der Rosenkavalier as a comedy of manners in modern Manhattan. In the local context, protagonist Oliver is disconsolate after a quarrel with his older lover Marian during a weekend in the Hamptons. After she drives them back to New York (supposedly for a faculty meeting she must attend) and drops him off in midtown, he impulsively heads uptown to Columbia to set things right with her, arriving at the history department only to discover that there is no faculty meeting, whereupon he happens to encounter Sophie, a grad student of Marian’s who had engaged the services of his florist shop for her upcoming wedding. So poor, distracted Oliver must make small talk with Sophie ... He hasn’t done the first thing about the flowers for her wedding. And now, just thinking about the flowers for the wedding fills him with abject misery. “I’m sorry”, he hears himself say, “I feel terrible about what happened”... “You came all the way up here?” Sophie asks, “To apologize?” “Yes,” Oliver says, with some relief, and truthfully enough. He has in fact come all the way up here to apologize, albeit not to her. What is SAID is true, as indicated in the boldfaced commentary, whence the feeling of relief (Oliver did in fact go up to Columbia to apologize, after all, albeit to Marian); what is COMMUNICATED is false (viz. that Oliver went to Columbia to apologize to Sophie). The manipulation of these two dimensions of meaning – what is (austerely) said vs. what is communicated – can be accomplished with more premeditation than it was for Oliver. We travel back now to seventeenthcentury Paris to visit the Jesuits’ take on applied pragmatics (Simon 1762; Fauconnier 1979). One expert, Sanchez [1614], instructs his charges on how to promise: A debtor can make a promise by reciting the formula ‘I promise to pay what I owe’ while silently adding that he will pay it not to his creditor, but rather to his confederate. More generally, the accused can avail himself of the doctrine of mental restriction, as expounded by Fillucius [1633] (Fauconnier 1979: 20; translation mine): [L]orsqu’on commence à dire je jure, il faut ajouter tout bas cette restriction mentale, qu’aujourdhui, & continuer tout haut, je n’ai pas mangé telle chose
When you begin to say ‘I swear...’, you must add under your breath this mental restriction, ‘that today...’, and continue out loud, ‘I have not eaten such and such’
In the same way, advises Tolet, the accused can truthfully swear ‘I have not killed anyone’ as long as ‘sa pensée soit de dire qu’il ne l’a pas fait depuis qu’il est en prison’ (see Fauconnier 1979: 17); this is legitimate since, as noted in the Fillucius passage above, such enrichments are entirely natural.
330
Laurence R. Horn
Nor is it only the (justly) accused who deserve recourse to such methods of hiding the truth: ‘La manière de cacher la verité doit être à la portée de tout le monde’ (Casnedi 1719 [in Simon 1762]). But hiding the truth is not tantamount to lying: crucially, ‘Il est licite d’employer toute ces manières de cacher la vérité non pas dans l’intention expresse de tromper les autres, mais seulement de les laisser se tromper eux-mêmes’ (Fegeli 1750 [in Simon 1762], emphasis added). The distinction between lying and hiding the truth – along with that between deceiving others and allowing them to deceive themselves – was artfully employed by the Jesuit-trained Bill Clinton in twentieth-century Washington, DC, who in turn transmitted his understanding of these methods to his disciple Monica Lewinsky. Asked under oath whether her testimony in re Paula Jones was false, Lewinsky truthfully testified that this testimony was ‘incomplete and misleading’, implicating (falsely) that it wasn’t false. In recounting this exchange, New York Times reporter Francis X. Clines describes Lewinsky as ‘exhibiting a Clintonian way with words’ (6 Feb. 1999), but as we can see, the skill involved might be indirectly attributed to seminary training.8 We now move from the Oval Office to the O.R./operating theatre, observing the exploitation of the difference between what is said vs. what is meant in twenty-first-century TV hospital dramas. First, a bit of background. The Jesuits’ disingenuous parallel between (27a, b) parallels our sense of the contrast between (28a) [see (19a) above], which favours a natural temporal restriction, as we have noted, and (28b), which does not: (27) a. I have not eaten such-and-such. b. I have not killed anyone. (28) a. Have you eaten breakfast? b. Have you had sex? This minimal pair is revisited by Taylor (2001: 46), who elucidates the distinction between having sex and having breakfast and contexts in which an utterance of [28b] amounts to a “so far today” question Taken on its own, independently of context and of the speaker’s communicative intentions, [28a], for example, expresses neither the question “have you ever in your life eaten breakfast?” nor the question “have you eaten breakfast yet this morning?” Moreover, neither the meaning of the word ‘sex’ nor the meaning of the word ‘breakfast’ forces one rather than the other temporal import on the relevant question. Indeed, without changing the meanings of either word, we can cook up contexts in which an utterance of [28a] amounts to an ‘ever’ question. Now consider a scenario developed in the 29 Sept. 2006 episode of ‘Grey’s Anatomy’ (ABC-TV). Young Benjamin, a patient who has a brain tumour
WJ-40: Issues in the Investigation of Implicature
331
causing him to blurt out his thoughts without restraint, is in his hospital bed beginning to undergo an examination by Dr. Meredith Grey, the intern who he has earlier witnessed exchanging smouldering glances with Dr Derek Shepherd (a.k.a. Dr McDreamy): Benjamin: ‘Did you have sex with that brain surgeon?’ Benjamin’s sister: ‘Benjamin!’ Meredith: ‘It’s OK. Nope, I haven’t. [PAUSE.] Not today, anyway.’ Moving from eros in Seattle to thanatos in New Jersey, we enter the domain of ‘House’ (FOX, 21 Nov. 2006). Once more we find a young (but doomed) male patient, his young (but precocious) sister, and the examining doctor (House’s associate Eric Foreman). Kama is grilling Dr Foreman about her brother Jack’s serious but as yet undiagnosed mystery disease from which he is suffering:9 Kama: Dr. Foreman: Kama: Dr. Foreman: Kama:
‘Is he gonna die?’ ‘No, no one’s gonna die.’ ‘In the whole world? Ever? That’s so great!’ ‘I meant ...’ ‘I know what you meant.’
Resetting the dials of our time machine back to fourth-century Egypt, we join Saul (2006) in borrowing a page from Alisdair MacIntyre that recounts an episode from the life of St. Athanasius to illustrate the distinction between lying (entailing the falsity of what is said) and misleading (allowing the falsity of what is implicated): Persecutors, dispatched by the emperor Julian, were pursuing [Athanasius] up the Nile. They came on him travelling downstream, failed to recognise him, and enquired of him: “Is Athanasius close at hand?” He replied: “He is not far from here.” The persecutors hurried on and Athanasius thus successfully evaded them without telling a lie. (MacIntyre 1994: 336, cited in Saul 2006: 3) This is all well and good, but given the content infallibly provided by wikipedia (http://en.wikipedia.org/wiki/Athanasius_of_Alexandria) revealing that St. Athanasius, ‘along with excommunication ... used beatings, intimidation, kidnapping and imprisonment to silence his theological opponents’, one wonders whether he really would have been all that bothered by the occasional outright lie. With due deference to Professor MacIntyre, we must seek our patron saint of truth concealment elsewhere. Some centuries after the trials and devices of St. Athanasius, Tristan and Iseult find themselves the quarry of the jealous Cornish lords Denoalen,
332
Laurence R. Horn
Andret, and Gondoïne, who are determined to trap the lovers in adultery and betray them to King Mark, Isolde’s husband and Tristan’s uncle. These ‘felons’, seizing their chance when Tristan is (supposedly) abroad in service to the King of Frisia, convince Mark to subject his queen to the Ordeal by Red-Hot Iron, as a result of which she will either clear her name forever or – as the villains confidently expect – burn to death for falsely swearing her fidelity. Before being put to the Ordeal by King Mark and his retainers, with King Arthur and his knights as warrantors, Iseult secretly sends word to Tristan instructing him to appear at the muddy river bank on the Day of Judgment in the garb of a miserable pilgrim. All proceeds accordingly, and now – after the ragged man carries her safely to shore before stumbling and collapsing into the mud – Queen Iseult addresses the nobles assembled by the river: In the words of Béroul, the twelfth-century Norman poet who first transcribed the Tristan legend of Tristan and Iseult, ‘Dieus i a fait vertuz’ (Bédier
Roi de Logres, et vous, roi de Cornouailles, et vous, sire Gauvain, sire Ké, sire Girflet, et vous tous qui serez mes garants, par ces corps saints et par tous les corps saints qui sont en ce monde, je jure que jamais un homme né de femme ne m’a tenue entre ses bras, hormis le roi Marc, mon seigneur, et le pauvre pélerin qui, tout à l’heure, s’est laissé choir à vos yeux. Roi Marc, ce serment convient-il? — Oui, reine, et que Dieu manifeste son vrai jugement! —Amen! dit Iseut. Elle s’approcha du brasier, pâle et chancelante. Tous se taisaient; le fer était rouge. Alors, elle plongea ses bras nus dans la braise, saisit la barre de fer, marcha neuf pas en la portant, puis, l’ayant rejetée, étendit ses bras en croix, le paumes ouvertes. Et chacun vit que sa chair était plus saine que prune de prunier. Alors de toutes les poitrines un grand cri de louange monta vers Dieu. (Bédier 1946: 137–8)
King of Logres and you, King of Cornwall, and you, Sir Gawain, Sir Kay, Sir Girflet, and all of you who are my warrantors, by all the holy things in the world, I swear that no man born of woman has ever held me in his arms other than my lord King Mark, and that poor pilgrim who just fell down before your eyes. King Mark, will that oath stand?’ ‘Yes, Queen’, he said, ‘and let God manifest his true judgment!’ ‘Amen!’ said Iseult. She approached the brazier, pale and staggering. All were silent; the iron was red hot. She thrust her naked arms into the coals, seized the iron rod, and bore it as she walked nine steps. Then, casting it off, she stretched out her arms in a cross, her palms open. Every witness saw her flesh healthier than a plum on a plum tree. Then, from the lungs of all assembled, a great cry of praise ascended to God. [translation and boldface mine]
WJ-40: Issues in the Investigation of Implicature
333
1946: 129). Rendering the sentiment loosely in modern language, it can only be concluded that God Himself is a semantic minimalist with an austere neo-Gricean conception of what is said.
5. Concluding remarks Like WD-40, the lubricant with over 2000 uses (cf. http://www.wd40.com/), WJ-40 – Paul Grice’s doctrine of conversational implicature that recently celebrated the fortieth anniversary of its unveiling in the William James lectures on logic and conversation – has displayed an impressive versatility along with a slippery reputation. Grice’s attempt to reconcile the logical elegance of the Russellians with the ordinary language insight of the Oxonians (see the valuable discussion in Chapman 2005) has won both converts and sceptics, often in the very same individual. Despite recent attacks on Gricean and neo-Gricean premises and modes of argumentation, the heart of his programme for meaning – in particular, the positing of a systematic (if often underdetermined) range of implications deriving from the interaction of opposed maxims grounded not just in cooperation but in rationality, retains its powerful explanatory force. The speaker’s and hearer’s joint (though tacit) recognition of the tendency to avoid unnecessary effort, and the inferences S expects H to draw from S’s efficient observance of this tendency, are more explicable directly from rationality than from cooperation as such.10 While Grice (1989: 28) recognizes that the maxims apply to cooperative ventures outside of language (baking a cake, fixing a car), cooperation is not a necessary condition, much less communication. It is plausible to take generalized forms of both Q and R Principles – ‘Do enough; Don’t do too much’ – to govern any goaloriented rational activity: a person brushing her hair, a dog digging a hole to bury a bone. In this way, the maxim of Quantity, in both its opposed (Q and R) subforms, is a linguistic instantiation of these rationality-based constraints on the expenditure of effort.11 Of course, as Grice recognized, the shared tacit awareness of such principles to generate conversational implicatures is a central property of speaker-meaning within the communicative enterprise. We have surveyed some of the ways in which Grice’s account of implicature leads to a simplification of the overall picture of meaning by transferring much of the burden of explanation from logical semantics to a general, independently motivated pragmatic theory that preserves the Modified Occam’s Razor principle and provides an explanatory treatment of the relation between the subcontraries. In sections 1 and 2, we addressed issues arising in the characterization of the epistemic conditions on scalar implicature and the two notions of informative and rhetorical strength within a Manichaean model of natural language inference that reconceptualizes the Gricean picture of maxim interaction. In addition to
334
Laurence R. Horn
rebutting some of the premises of Chierchia’s (2004) non-Gricean view of scalar implicature, I have made the case for a relatively orthodox Gricean notion of what is said, directly linked to the linguistic properties of the utterance, as an integral component of sentence meaning. Evidence for the utility of this notion ranges from the limited nature of pragmatic intrusion, as seen in section 3, and the need to invoke minimal meanings to account for intuitions of lying and misleading, as seen from our grand tour in section 4. Since the ancient rhetoricians first distinguished between what is said and what is meant – or, as the French put it, between ce qu’on dit and ce qu’on veut dire – the theory of meaning in natural language has focused on just where to draw the line separating the two and just how to draw the lines connecting them. My goal in this chapter has been to look at some of the ways in which a minimally modified Gricean model might provide a natural and intuitively satisfying approach to phenomena within this domain.
Notes * A version of this paper appears as Horn (2009). Thanks are due to those who attended and commented on earlier presentations of parts of this material in Milan, East Lansing, Oakland, Rochester, Sheffield, and Leysin, and in particular to Barbara Abbott, Mira Ariel, Kent Bach, Emma Borg, David Braun, Bart Geurts, Michael Israel, Manfred Krifka, Anna Papafragou, Klaus Petrus, Jennifer Saul, and Gregory Ward. The blame is all mine. 1. As stressed by Hirschberg (1985) and Geurts (to appear), entailment-based scalar implicature represents just one variety of quantity implicature. The prevalence of ad hoc (non-lexically generated) scales and the implicatures derived therefrom must also be taken into account in any comprehensive treatment of quantity. 2. Similarly, Chierchia et al. (2001: 160) observe that while John saw many students will normally implicate that he didn’t see all of them, no such implicature arises in e.g. There aren’t many students. From such observations follows the descriptive generalization: (i) Scalar implicatures do not arise in downward entailing environments. But in fact There aren’t many students does normally induce a scalar implicature, viz. that there are some students. Interestingly, it is noted in the same paper (Chierchia et al. 2001: 162) that ‘DE operators reverse canonical entailments’, but it is for precisely this reason that they also reverse scales and SIs, rendering the generalization in (i) untenable. 3. The generalization linking SI ‘blocking’ and NPI licensing is open to challenge on other grounds. Empirical work since Ladusaw (1980) has demonstrated that DE-ness is neither necessary nor sufficient for the occurrence of NPIs. Polaritylicensing environments (including subjunctives, imperatives, generics, and modals) are not always DE, and while some environments are subsumable under
WJ-40: Issues in the Investigation of Implicature
4.
5.
6.
7. 8.
9.
10.
11.
335
DE-hood by adjusting the definition (see Heim 1984; Krifka 1995), not all of them are (see Linebarger 1987; Israel 1996; and Giannakidou 1998 for the roles of pragmatic inference, lexical semantics, and non-veridicality in accounting for the distribution of polarity items). What is relevant to scalar implicature is not polarity licensing per se but the scale reversal triggered by DE-ness. Geurts and Nouwen (2007) offer a plausible account of at least that provides distinct truth-conditions for these two statements. Like at least, the effect of yet is to convert a rank order into an acting scale: (ii) and (iii), unlike (i), invoke a disjunction between the item in question and higher values: (i) a. Is your daughter a sophomore? b. No/#Yes, (in fact) she’s a junior. (ii) a. Is your daughter at least a sophomore? b. Yes/#No, (in fact) she’s a junior. (iii) a. Is your daughter a sophomore yet? b. Yes/#No, (in fact) she’s a junior. Even the retroactive adjustment of semantic content appears to be a pragmatic and hence defeasible process. Thus King and Stanley (2005) cite (i) (their (27)), borrowed from Robyn Carston, as a problematic case for the unadorned Gricean account: (i) If Hannah insulted Joe and Joe resigned, then Hannah is in trouble. But even here, a possible response is that in fact Hannah insulted Joe and Joe did resign, but Hannah has nothing to worry about because Joe was going to resign anyway. The causal connection, therefore, does not seem to be semantically written into the protasis of the conditional in (i). Geurts (to appear) argues against not only the ‘strong’ defaultism of Levinson and Chierchia, amounting to the automaticity of generalized conversational implicature, but even the ‘weak’ defaultism I have endorsed. I regard this as an open question, pending the development of more refined empirical evidence. See also Jaszczolt (2005) on defaults. Terkourafi (2009) provides empirical support for the utility of a conservative notion of what is said. For more on Clintonian applied pragmatics and its relation to perjury statutes, along with other real-life illustrations, see Solan and Tiersma (2004: chapter 11). If Dr. Grey was operating under the influence of Taylor (2001) and his Jesuit precursors, the dialogue in ‘House’ was obviously informed by the commentary in Bach (1994): The proposition being communicated is a conceptually enriched or elaborated version of the one explicitly expressed by the utterance itself. So, for example, if a mother utters (1) to her crying son upset about a cut finger, (1) You’re not going to die from that cut. she is likely to mean that he is not going to die from that cut, not that he is immortal. See Kasher’s principle of effective means (1982: 32), another minimax balancing act stipulating that ‘Given a desired end, one is to choose that action which most effectively, and at least cost, attains that end, ceteris paribus.’ Grice (1986: 83) proposes his own ‘Principle of Economy of Rational Effort’; for more on the central place of rationality in Grice’s philosophy of language, see Davies (2007).
336 Laurence R. Horn
References Anscombre, J. C. and Ducrot, O. (1983) L’argumentation dans la langue (Brussels: Pierre Mardaga). Ariel, M. (2004) Most. Language, 80: 658–706. Ariel, M. (2006) A ‘just that’ lexical meaning for most. In K. Turner and K. von Heusinger (eds.)Where Semantics Meets Pragmatics (London: Elsevier), pp. 49–91. Austin, J. L. (1962) How To Do Things With Words (Cambridge, MA: Harvard University Press). Bach, K. (1994) Conversational impliciture. Mind and Language, 9: 124–62. Bach, K. (1999) The myth of conventional implicature. Linguistics and Philosophy, 22: 327–66. Bach, K. (2001) You don’t say? Synthese, 127: 11–31. Bach, K. ( 2005) Context ex machina. In Z. Szabó (ed.) Semantics vs. Pragmatics (Oxford: Clarendon), pp. 15–44. Bach, K. (2006) The top 10 misconceptions about implicature. In B. Birner and G. Ward (eds.), Drawing the Boundaries of Meaning: Neo-Gricean Studies in Pragmatics and Semantics in Honor of Laurence R. Horn (Amsterdam: Benjamins), pp. 21–30. Bédier, J. (1946) Le roman de Tristan et Iseut, renouvelé par Joseph Bédier de l’Academie Française (Paris: L’Édition de l’Art H. Piazza). Bontly, T. (2005) Modified Occam’s Razor: Parsimony arguments and pragmatic explanations. Mind & Language, 20: 288–312. Borg, E. (2004) Minimal Semantics (Oxford: Oxford University Press). Bosanquet, B. (1911) Logic, Vol. 1, 2nd edn (Oxford: Clarendon). Bott, L. and Noveck, I. (2004) Some utterances are underinformative: The onset and time course of scalar inferences. Journal of Memory and Language, 51: 437–57 Breheny, R., Katsos, N. and Williams, J. (2006) Are generalized scalar implicatures generated by default? An on-line investigation into the role of context in generating pragmatic inferences. Cognition, 100: 434–63. Brown, P. and Levinson, S. (1987) Politeness (Cambridge: Cambridge University Press). Bultinck, B. (2005) Numerous Meanings: The Meaning of English Cardinals and the Legacy of Paul Grice (Oxford: Elsevier). Cappelen, H. and Lepore, E. (2005) Insensitive Semantics (Oxford: Blackwell). Carston, R. (1988) Implicature, explicature, and truth-conditional semantics. In R. Kempson (ed.) Mental Representations: The Interface Between Language and Reality (Cambridge: Cambridge University Press), pp. 133–81. Carston, R. (1995) Quantity maxims and generalized implicatures. Lingua, 96: 213–44. Carston, R. (2002) Thoughts and Utterances: The Pragmatics of Explicit Communication (Oxford: Blackwell). Carston, R. (2004) Relevance theory and the saying-implicating distinction. In Horn and Ward (eds.), pp. 633–56. Carston, R. (2005) Relevance theory, Grice, and the neo-Griceans: A response to Laurence Horn’s ‘Current issues in neo-Gricean pragmatics’. Intercultural Pragmatics, 2: 303–20. Chapman, S. (2005) Paul Grice, Philosopher and Linguist. (Basingstoke and New York: Palgrave Macmillan). Chierchia, G. (2004) Scalar implicatures, polarity phenomena, and the syntax/ pragmatics interface. In A. Belletti (ed.) Structures and Beyond (Oxford: Oxford University Press), pp. 39–103.
WJ-40: Issues in the Investigation of Implicature
337
Chierchia, G., Crain, S. Guasti, M. T. , Gualmini, A. and Meroni, L. (2001) The acquisition of disjunction: Evidence for a grammatical view of scalar implicatures. BUCLD 25 Proceedings (Somerville: Cascadilla), pp. 157–68. Chomsky, N. (1972) Some empirical issues in the theory of transformational grammar. In S. Peters (ed.) Goals of Linguistic Theory (Englewood Cliffs, NJ: PrenticeHall), pp. 63–130. Cohen, L. J. (1971) Some remarks on Grice’s views about the logical particles of natural language. In Y. Bar-Hillel (ed.) Pragmatics of Natural Language (Dordrecht: Reidel), pp. 50–68. Davies, B. (2007) Grice’s Cooperative Principle: Meaning and rationality. Journal of Pragmatics, 39: 2308–31. De Morgan, A. (1858) On the syllogism: III, and on logic in general. In De Morgan On the syllogism and other logical writings (London: Routledge & Kegan Paul), pp. 74–146. Fauconnier, G. (1975) Pragmatic scales and logical structure. Linguistic Inquiry, 6: 353–75. Fauconnier, G. (1979) Comment contrôler la vérité. Actes de la Recherche en Sciences Sociales, 25: 3–22. Fox, D. (2006) Free choice and the theory of scalar implicatures. Ms., MIT. Gazdar, G. (1979) Pragmatics (New York: Academic Press). Geurts, B. (1998) Scalars. In P. Ludewig and B. Geurts (eds.) Lexikalische Semantik aus kognitiver Sicht (Tübingen: Gunter Narr Verlag), pp. 95–117. Geurts, B. (2009) Scalar implicature and local pragmatics. Mind and Language, 24: 51–79. Geurts, B. To appear. Quantity Implicatures (Cambridge: Cambridge University Press). Geurts, B. and Nouwen, R. (2007) At least et al.: The semantics of scalar modifiers. Language, 83: 533–59. Giannakidou, A. (1998) Polarity Sensitivity as (Non-)Veridical Dependency (Amsterdam: John Benjamins). Grice, H. P. (1961) The causal theory of perception. Proceedings of the Aristotelian Society, Supplementary Volume 35: 121–52. Grice, H. P. (1986) Reply to Richards. In R. Grandy and R. Warner (eds.) Philosophical Grounds of Rationality (Oxford: Oxford University Press), pp. 45–106. Grice, H. P. (1989) Studies in the Way of Words (Cambridge, MA: Harvard University Press). Gutzmann, D. (2007) Eine Implikatur konventioneller Art: Der Dativus Ethicus. Linguistische Berichte, 211: 277–308. Gutzmann, D. (2008) On the Interaction between Modal Particles and Sentence Mood in German (Mainz: German Institute, Johannes Gutenberg University). Hamilton, Sir W., of Edinburgh (1860) Lectures on Logic, Volume I. (Edinburgh: Blackwood). Heim, I. (1984) A note on negative polarity and downward entailingness, NELS, 14: 98–107. Hirschberg, J. (1985) A Theory of Scalar Implicature. University of Pennsylvania dissertation. (Revised version published New York: Garland, 1991.) Horn, L. R. (1972) On the Semantic Properties of Logical Operators in English. UCLA dissertation. Distributed by Indiana University Linguistics Club, 1976. Horn, L. (1984) Toward a new taxonomy for pragmatic inference: Q-based and R-based implicature. In D. Schiffrin (ed.) Meaning, Form, and Use in Context (GURT ‘84) (Washington, DC: Georgetown University Press), pp. 11–42.
338
Laurence R. Horn
Horn, L. (1989) A Natural History of Negation (Chicago: University of Chicago Press). (Reissued with a new introduction, Stanford: CSLI, 2001.) Horn, L. (1990) Hamburgers and truth: Why Gricean inference is Gricean. BLS, 16: 454–71. Horn, L. (1992) The said and the unsaid. SALT II (Columbus: Ohio State University), pp. 163–92. Horn, L. (2000a) From if to iff: Conditional perfection as pragmatic strengthening. Journal of Pragmatics, 32: 289–326. Horn, L. (2000b) Pick a theory (not just any theory): Indiscriminatives and the free-choice indefinite. In L. Horn and Y. Kato (eds.) Negation and Polarity (Oxford: Oxford University Press), pp. 147–92. Horn, L. (2004) Implicature. In Horn and Ward (eds.), pp. 3–28. Horn, L. (2006a) The border wars: A neo-Gricean perspective. In K. Turner and K. von Heusinger (eds.) Where Semantics Meets Pragmatics (Oxford: Elsevier), pp. 21–48. Horn, L. (2006b) More issues in neo- and post-Gricean pragmatics: a reply to Carston. Intercultural Pragmatics, 3: 81–93. Horn, L. (2007a) Neo-Gricean pragmatics: a Manichaean manifesto. In N. BurtonRoberts (ed.) Pragmatics (Basingstoke and New York: Palgrave Macmillan), pp. 158–83 Horn, L. (2007b) Toward a Fregean pragmatics: Voraussetzung, Nebengedanke, Andeutung. In I. Kecskes and L. Horn (eds.) Explorations in Pragmatics: Linguistic, Cognitive, and Intercultural Aspects (Berlin: Mouton de Gruyter), pp. 39–69. Horn, L. (2008) “I love me some him”: The landscape of non-argument datives. In O. Bonami and P. Cabredo Hofherr (eds.) Empirical Issues in Syntax and Semantics, 7, pp. 169–92. Horn, L. (2009) WJ-40: Implicature, truth, and meaning. International Review of Pragmatics, 1: 3–34. Horn, L. (To appear) Histoire d’*O: Lexical pragmatics and the geometry of opposition. In J.-Y. Béziau (ed.) New Perspectives on the Square of Opposition (Bern: Peter Lang). Horn, L. and Ward, G. (eds.) (2004) The Handbook of Pragmatics (Oxford: Blackwell). Hurewitz, F., Papafragou, A., Gleitman, L. and Gelman, R. (2006) Asymmetries in the acquisition of numbers and quantifiers. Language Learning and Development, 2: 77–96. Israel, M. (1996) Polarity sensitivity as lexical semantics. Linguistics and Philosophy, 19: 619–66. Israel, M. (to appear) Pragmatics, Polarity, and the Logic of Scales (Cambridge: Cambridge University Press). Jaszczolt, K. (2005) Default Semantics: Foundations of a Compositional Theory of Acts of Communication (Oxford: Oxford University Press). Kadmon, N. and Landman, F. (1993) Any. Linguistics and Philosophy, 16: 353–422. Kasher, A. (1982) Gricean inference revisited. Philosophica, 29: 25–44. King, J. and Stanley, J. (2005) Semantics, pragmatics, and the role of semantic context. In Z. Szabó (ed.) Semantics vs. Pragmatics (Oxford: Clarendon), pp. 111–64. Koenig, J.-P. (1991) Scalar predicates and negation: Punctual semantics and interval interpretations. CLS, 27: 140–55. Korelitz, J. Hanff (2005) The White Rose (New York: Miramax Books). Krifka, M. (1995) The semantics and pragmatics of polarity items. Linguistic Analysis, 25: 209–57. Ladusaw, W. (1980) Polarity Sensitivity as Inherent Scope Relations (New York: Garland).
WJ-40: Issues in the Investigation of Implicature
339
Lahiri, U. (1998) Focus and negative polarity in Hindi. Journal of Semantics, 6: 57–125. Lee, Y.-S. and Horn, L. (1994) Any as indefinite + even. Ms., Yale University Lehrer, A. (1974) Semantic Fields and Lexical Structure (Amsterdam: North-Holland). Levinson, S. (1983) Pragmatics (Cambridge: Cambridge University Press). Levinson, S. (2000) Presumptive Meanings: The Theory of Generalized Conversational Implicature (Cambridge, MA: MIT Press). Linebarger, M. (1987) Negative polarity and grammatical representation. Linguistics and Philosophy, 10: 325–87. Macintyre, A. (1994) Truthfulness, lies, and moral philosophers: What can we learn from Mill and Kant? The Tanner Lectures. Downloadable at http://www.tannerlectures.utah.edu/lectures/macintyre_1994.pdf. Mill, J. S. (1867) An Examination of Sir William Hamilton’s Philosophy , 3rd edn (London: Longman). Noveck, I. and Posada, A. (2003) Characterizing the time course of an implicature: An evoked potentials study. Brain and Language, 85: 203–10. Papafragou, A. and Musolino, J. (2003) Scalar implicatures: Experiments at the semantics-pragmatics interface. Cognition, 86: 253–82. Papafragou, A. and Schwarz, N. (2006) Most wanted. Language Acquisition, 13: 207–52. Potts, C. (2005) The Logic of Conventional Implicatures (Oxford: Oxford University Press). Recanati, F. (2001) What is said. Synthese, 128: 75–91. Recanati, F. (2004) Literal Meaning (Cambridge: Cambridge University Press). Reinhart, T. (2006) The acquisition of scalar implicatures. §5.3 of Interface Strategies − Optimal and Costly Computations (Cambridge, MA: MIT Press). Russell, B. 2006. Against grammatical computation of scalar implicatures. Journal of Semantics, 23: 361–382. Sauerland, U. (2004) Scalar implicatures in complex sentences. Linguistics and Philosophy, 27: 367–391. Saul, J. (2002a) Speaker meaning, what is said and what is implicated. Noûs, 36: 228–48. Saul, J. (2002b) What is said and psychological reality: Grice’s project and relevance theorists’ criticisms. Linguistics and Philosophy, 25: 347–72. Saul, J. (2006) Lying, misleading, and accidental falsehood: The role of what is said. Ms., University of Sheffield. Schwenter, S. (1999) Pragmatics of Conditional Marking: Implicature, Scalarity, and Exclusivity (New York: Garland). Simon, P. G. (1762) Extraits des Assertions dangereuses et pernicieuses en tous genres soutenues et enseignées par les soi-disans Jésuites. (Tome III: Parjure, Fausseté, Faux Témoignage.) Paris. Soames, S. (1982) How presuppositions are inherited: A solution to the projection problem. Linguistic Inquiry, 13: 483–535. Solan, L. and Tiersma, P. (2004) Speaking of Crime: The Language of Criminal Justice (Chicago: University of Chicago Press). Spector, B. (2006) Aspects de la pragmatique des opérateurs logiques. PhD dissertation, Université de Paris VII. Strawson, P. F. (1952) Introduction to Logical Theory (London: Methuen). Taylor, K. (2001) Sex, breakfast, and descriptus interruptus. Synthese, 128: 45–61. Terkourafi, M. (2009) What use is ‘what is said’? In M. Kissine and P. de Brabanter (eds.) Utterance Interpretation and Cognitive Models (Amsterdam: Emerald), pp. 27–58.
Name Index Abbott, B. 79–81, 83–6, 89–96, 98–9n, 134n, 334n Abusch, D. 99n Adams, R. M. 71n Aitchison, J. 296, 301, 304 Albritton, R. 63, 71 Alston, W. P. 332n Ammer, C. 133n Anscombre, J. C. 322 Ariel, M. 314–15, 334n Aristole 63, 310–11, 321 Asher, N. 82 Atlas, J. 4, 25, 47, 57–8, 63, 66–8, 70–2n, 80, 89n,132–3n, 183n, 295, 297 Austin, J. L. 4, 16, 31, 35–9, 41–4, 68, 164, 206, 221, 223–4, 228, 256, 259, 304, 328 Avramides, A. 17, 127n Ayer, A.J. 35–7, 68 Bach, K. 10, 16, 19, 21–3, 41, 126, 132–4n, 167n, 182n, 210, 225, 236, 238, 240, 254–5, 263–5, 282n, 289, 291, 294, 297, 305–6, 310, 323–5, 327–8, 334n Baker, J. 3, 26, 188, 198 Ball, C. 133n Barrett, L. 209 Bartgh, J. 217n Barwise, J. 304 Beaver, D. 82 Bédier, J. 332–3 Benaceraff, P. 72n Bennett, J. 17, 226, 250n Bezuidenhout, A. 25, 80, 99n, 283–4n, 291, 302–4, 307n Bianchi, C. 21 Birner, B. 91 Blackburn, S. 240 Boer, S. 133n Bontly, T. 312 Borg, E. 27, 279–80, 285n, 288–94, 299, 305, 328, 334n
Bosanquet, B. 320 Bott, L. 272, 277, 327 Bradley, F. H. 63 Brandom, R. 212 Braun, D. 182n, 334n Breheny, R. 283n, 285n, 327 Breusch, R. 72n Brown, P. 320 Bultinck, B. 314 Burton-Roberts, N. 87–8, 99n, 112, 118–19, 125, 132–4n Cappelen, H. 21, 257–61, 264–5, 267n, 279, 289–91, 294, 298, 328 Carnap, R. 33–4, 51, 53–5, 60, 69n-70n, 289 Carston, R. 22, 86, 89–9n, 114, 132–4n, 167n, 252, 254, 256, 262, 266n, 272, 274, 276, 283–4n, 288–9, 291–2, 297, 299, 301, 312, 314, 321, 323–4, 328, 335n Cavell, S. 48 Chafe, W. 91 Chang, R. 200 Chapman, S. 2, 25, 34, 43, 98n, 132–3n, 333 Chartland, T. 217n Cheney, D. 209 Chierchia, G. 79, 140, 142–3, 149, 156–8, 161, 164, 167n, 271, 273–4, 317–19, 321, 325, 327, 334, 335n Chomsky, N. 39, 47, 59–60, 70n, 72n, 318 Church, A. 69n Clark, P. 199–200 Cohen, J. L. 139–40, 142, 144, 147–8, 304, 325 Cohen, P. R. 332n Cohen, T. 228, 233n Conrad, S.-J. 27 Cowie, A. 134n Croft, W. 295, 301 Cruse, A. D. 295, 301
341
342
Name Index
Cutting, J. C. 283n Dancy, J. 199 David, P. 208 Davidson, D. 56, 68, 72n, 133n, 188, 252 Davidson, M. 69n Davies, B. 335n Davis, G. 70n Davis, S. 228 Davis, W. 25, 112, 116–17, 124, 127, 132–4n, 178, 182–3n, 216n, 247–8 DeMorgan, A. 311 Desalles, J.-L. 217n Descartes, R. 67 Dewey, J. 67, 70n Donellan, K. 72n, 80 Drainville, R. 182n Dreben, B. 48 Ducrut, O. 322 Dunbar, R. 209 Eggert, R. 150 Ekman, P. 209 Elugardo, R. 297 Evans, G. 296
Green, T. H. 68 Groenendijk, J. 148, 166n Gross, S. 132n Gu, Y. 10, 332n Gutzmann, D. 310 Halliday, T. R. 208 Hamilton, W. 311 Hanfling, O. 70n Harman, G. 51, 73n, 240, 298 Harnish, R. 133n, 225, 240 Harper, D. 208, 217n Harris, J. F. 69n Hegel, G. F. W. 68 Heim, I. 89–90, 147, 335n Hirschberg, J. 11, 315–17, 322, 334n Hook, 48 Hookway, C. 182n Hopper, P. 134n Horn, L. 9, 27, 58, 76, 86, 98–9n, 107, 110, 112, 117–18, 121, 123, 125, 132–4n, 153, 161, 165n, 182n, 273, 275–6, 283–4n, 310, 312, 314–25 Hume, D. 66 Hurewitz, F. 315 Israel, M. 132n, 319, 322, 334n
Fauconnier, G. 317, 329 Fintel, K. 80, 100n, Fodor, J. 58–9, 279, 285n Fox, D. 317 Frege, G. 9, 69n, 252 Garcia-Carpintero, M. 285n Gauker, C. 306n Gazdar, G. 110, 133n, 139, 153, 161,312, 315–17 Geach, P. 296 Geurts, B. 82, 109–10, 112, 130–1, 132–4n, 314–15, 317–18, 325, 327, 334–5n Giannakidou, A. 335n Gibbs, M. 13 Gilbert, M. 218n Goodman, N. 51, 68, 69n Graeser, A. 27 Grandy, R. 3, 16, 69n Green, M. 26, 145, 165–6n, 182n, 212, 216–17n
Jaszczolt, K. 335n Johnson, M. 300 Kadmon, N. 319 Kamp, H. 291 Kant, I. 32, 42, 53, 66, 68 Kaplan, D. 307n Kasher, A. 3, 335n Katsos, N. 283n, 285n, 327 Katz, J. J. 72n, 134n Katzer, A. 147 Keller, R. 300 Kemmerling, A. 3, 17–18, 222, 224–6, 332n Kempson, R. 98n, 132–4n Kenny, A. 250n King, J. 167n, 318, 325, 335n Kirschbaum, I. 300 Kittay, E. 134n Koenig, J.-P. 165n, 314 Kompa, N. 27
Name Index Korelitz, J. H. 329 Krause, P. 100n Krifka, M. 334–5n Kripke, S. 47–8, 51, 53, 57, 59–69, 71–2n, 133n Kroch, A. 133n Ladd, D. 132n Ladusaw, W. 334n Lahiri, U. 319 Lakoff, G. 295, 300 Lambrecht, K. 91 Landman, F. 319 Lathan, A. 126 Lee, Y.-S. 319 Lehrer, A. 322 Lepore, E. 21, 58–9, 182n, 257–61, 264–5, 267n, 279, 289–91, 294, 298, 328 Levesque, H. J. 332n Levine, J. 283n, 285n Levinson, S. 9, 11, 133–4n, 143, 149, 156, 158, 165n, 167n, 182n, 271–3, 284n, 312, 315, 317, 320, 325–7 Lewis, C. I. 73n Lewis, D. 89, 133n, 210, 305–6 Linebarger, M. 132n, 335n Linzky, L. 48 Loar, B. 17, 41 Locke, J. 32, 59–60 Ludlow, P. 297 Lüthi, D. 7 Lycan, W. 133n MacIntyre, A. 331 Marcu, D. 332n Martinich, A. 26, 69, 266n Maynard Smith, J. 207–8, 217n McConell-Ginet, S. 79 McGinn, C. 72n Meggle, G. 3, 225 Mill, J. S. 311, 315 Montague, R. 305 Moore, G. M. 34, 70n Moravcsik, J.M. 302 Morgan, J. C. 116–17, 134n Mundle, C. W. K. 36 Musolino, J. 278, 315
343
Naess, A. 34 Neale, S. 9, 15–16, 23–4, 133n, 141, 165n, 182n, 296 Nidditch, P. H. 72n Nouwen, R. 335n Noveck, I. A. 272–3, 277–8, 327 Nunberg, G. 110, 298, 306–7n Papafragou, A. 278, 315, 334n Passmore, J. 70n Perry, J. 304 Peter, G. 21 Petrie, M. 208 Petrus, K. 3, 16, 18, 26, 221–2, 224–5 Pfister, J. 21, 27 Pinkal, M. 295–6 Plato 67 Plunze, C. 26 Popper, K. 33 Posada, A. 327 Potts, C. 255, 266n, 310 Pouscoulous, N. 285n Powell, G. 291 Preyer, G. 21 Prince, E. 91 Putnam, H. 67, 69–73n Pylyshyn, Z. 296 Quine, W. V. O. 6, 33–4, 36, 39, 42, 47–8, 50–7, 59, 61–3, 66, 68, 69, 69n–70n, 72–3n, 295 Quinn, W. 188 Raz, J. 200 Recanati, F. 13, 22, 141, 143, 145, 152, 165–7n, 238, 240, 284n, 289–92, 298, 323, 328 Reichert, S. 208 Reillo, P. 208 Reimer, M. 306n Reinhart, T. 317 Ripps, L. J. 278 Roberts, C. 82 Rolf, E. 11 Rooij, R. van 167n, 217n Rooth, M. 164 Rorty, R. 59, 66, 68 Roy, T. 69n Russell, Bertrand 4–5, 32, 70n,75–8, 88, 98n, 104, 109, 252, 317
344
Name Index
Russell, Benjamin 161–2, 167n Ryle, G. 35, 250n, Sadock, J. M. 133–4n, 165n, 295 Saka, P. 69n Salmon, N. 68 Sanders, C. 208 Sandt, van der R. A. 132–4n Sauerland, U. 140–1, 149, 153–5, 158, 165n, 167n, 315, 317 Saul, J. 9, 26, 133n, 170, 174, 180, 327–8, 331, 334n Sbisà, M. 3, 182n Scanlon, T. M. 184 Schiffer, S. 16–17, 214n, 225 Schlenker, P. 95–7 Schlick, M. 32 Schlieben-Lange, B. 228 Schulz, K. 167n Schwarz, N. 315 Schwenter, S. 325 Scott, D. S. 69–72n Searle, J. R. 16, 134n, 216n, 332, 236, 239, 252, 289, 299–302 Seitz, F. 250n Seuren, P. A. 120–1, 132–4n Severen, R. F. 69n Seyfarth, R. 209 Siebel, M. 222, 224, 226 Siegfried, J. 134n Simon, P. G. 329–30 Simons, M. 25–6, 79–80, 93, 100n, 146, 150, 166n Soames, S. 4, 47–8, 50–1, 55, 59–61, 63, 66–9, 71–3n, 89, 98n, 315 Solan, L. 335n Spector, B. 167n, 317 Sperber, D. 13, 133–4n, 142, 148–9, 167n, 236, 239, 274–5, 289, 291, 327 Stainton, R. J. 297 Stalnaker, R. 100n, 147, 210, 213, 217n Stanley, J. 167n, 285n, 289–90, 293, 297, 318, 325, 335n Stokhof, M. 148, 166n Storto, G. 283n, 285n
Strawson, P. F. 4–7, 15–16, 31, 41, 43, 47, 49–51, 53–6, 75–6, 79, 87, 98n, 221, 225–6, 237–8, 311 Stroll, A. 265 Szabo, Z. 21, 100n, 134n, 285n Tanenhaus, M. 283n, 285n Tarski, A. 69–70n Taylor, K. 330, 335n Terkourafi, M. 335n Thomason, R. 210 Tiersma, P. 335n Traugott, E. C. 134n Travis, C. 289, 291, 302–3 Truniger, J. 2, 14, 27 Ulkan, M. 332n Urmson, J. O. 36, 166n Vanderveken, D. 206, 214 Vendler, Z. 58 Vlach, F. 216n Walker, R. 143–8, 155, 165n Ward, G. 91, 324, 334n Warner, R. 3, 16, 69n Watson, G. 199 Watts, D. 284n White, M. 47–8, 51, 53, 68, 69–70n, 72n Whitehead, A.N. 32, 68 Whiting, J. 199 Wiener, N. 52 Wilkinson, G. 208 Williams, B. 200 Williams, J. 283n, 285n, 327 Williamson, T. 212 Wilson, D. 13, 133–4n, 142, 148–9, 167n, 236, 239, 274–5, 289, 291, 325 Wilson, T. 217n Wittgenstein, L. 32, 34, 39, 60–1, 68, 70n Zahavi, A. 208 Zeevat, H. 82 Zwicky, C. 295
Subject Index ambiguity 9, 21, 27, 98, 103, 106, 108–10, 112, 114, 118, 120, 125, 127, 131, 134, 197–8, 269, 279, 295, 303, 305, 312, 314, 323 analytic/synthetic distinction 6, 25, 32–4, 47–51, 53, 55–9, 63, 68–70, 72 assertion 58, 72, 203, 205–6, 210, 213–16, 222, 257–8, 312, 315, 317–18, 320, 322; see also illocutionary act, felicity condition atomism 58–9 Basic Set Assumption 27, 290, 294–5, 306 calculability / calculation 26, 77–8, 93–4, 115, 117, 140, 142–5, 162, 165, 170, 174–8, 180, 182 cancellability 76–8, 80, 83–5, 87, 94–5, 97, 99–100, 114, 116, 174, 264, 312, 324 communication 1, 3, 7, 12, 16, 26, 31, 40, 87, 94, 116, 177, 202, 207, 217, 221, 227, 230, 236–40, 245, 249, 253, 263–4, 266, 269, 276, 284, 312, 314, 318, 324, 326, 329, 333, 335; see also communicative act, communicative intention communicative act 26, 153, 224–7, 230, 231–2, 235, 239, 241–2, 244–8; see also communication, communicative intention communicative intention 26, 146, 153, 330; see also intention, speaker-meaning, communication, communicative act completion 21, 297, 323 conceptual analysis 4, 7 conditional 84, 140–1, 144–5, 147–8, 155–8 context 27, 77, 79, 82–3, 85, 94, 99, 110–11, 115, 128, 140, 144–5, 162–3, 187, 260, 265–6, 270–1, 274,
277–9, 281–2, 294, 296–7, 301, 303, 306–7, 314, 321, 326–7, 330 context-sensitivity 156, 269, 271, 273, 285, 300, 302, 304–5, 307 see also contextualism, semantic underdetermination contextualism 21, 27, 288–90, 292, 295, 297, 299, 324; see also context, semantic underdetermination convention 16, 116–17, 124, 126–7, 130, 132, 175–6, 209, 293, 320–1, 326; see also meaning: conventional meaning conversation 91, 108, 146, 253, 281, 333 conversational implicature: see implicature conversational maxims 7, 11–14, 26, 68, 116, 118, 143–4, 152, 158, 164, 170–6, 178, 180, 253, 269, 333–4 theory of conversation (theory of implicature) 2, 7, 9, 14, 18–19, 23, 25, 31, 40, 43, 120–1, 262, 269, 271, 281–2 see also implicature, Cooperative Principle, maxims Cooperative Principle 3, 10, 13–14, 100, 115–16, 118, 143–4, 170, 173–6, 178, 181–2, 225, 269, 333; see also implicature: conversational implicature, conversation: conversational maxims definite description 25, 75, 79–81, 83, 87, 97, 103, 296; see also theory of description demonstrative 64–5, 258, 279, 291, 296, 299, 305–6, 323; see also indexical designator (rigid) 61–2 detachability (non-detachability) 76, 78, 92, 94, 100, 114–15, 131 disjunction 148–51, 154–5, 163, 317 enrichment: see expansion essential property 59–66, 69
345
346
Subject Index
expansion (strengthening, enrichment) 22–3, 78, 98, 154–9, 161, 164, 166, 263, 270, 274, 276–9, 289, 291, 294, 312, 314, 316, 318–21, 323, 328, 355; see also pragmatic inference, pragmatic intrusion explication 51–5 explicature 134, 275–6, 278–9, 284, 314, 324, 328; see also expansion, impliciture explicit performative 223–24, 44; see also illocutionary act, speech act expression 254–5, 264–6 felicity condition 26, 65, 96–7, 148, 151, 166, 224, 231–2, 320; see also illocutionary act, speech act formal semantics 41, 82, 147, 164, 288–90, 292, 294–5, 304, 306, 333; see also semantic minimalism generalized conversational implicature (GCI): see implicature Gricean mechanism 17–18, 226–7, 231 Holism 58–9 homonymy 295–6, 303, 320 identity 57, 67, 70, 72 illocutionary act /force 16, 26, 39, 145, 205, 212, 214–16, 221–32, 236, 328; see also assertion, communicative act, speech act illocutionary intention 238, 240–1; see also illocutionary act, intention, speaker-meaning implication 6–7, 25, 40, 58, 78, 144, 255 implicature 17, 19, 21, 23, 26, 43–4, 78–9, 86, 92–5, 100, 103, 111–18, 120–4, 129, 131, 139–40, 139–45, 148–9, 153–8, 160–3, 165–7, 170–1, 177, 210, 255, 262–3, 266, 268–9, 272–4, 276, 281–3, 284, 310, 312, 314–16, 318, 320–5, 327–8, 331, 333–4 conventional implicature 7, 9, 10, 12, 16, 20, 92, 118–19, 122, 127, 133, 164, 171–2, 175–6, 180, 254–5, 266, 269, 283, 310, 325 conversational implicature 7, 10–12, 16, 25, 27, 58, 68, 72, 75–6, 78, 80, 88,
92, 94, 97–8, 115–18, 121–4, 127, 132, 138, 140–1, 143–4, 160–3, 165–6, 170–9, 181–3, 253–4, 256, 262, 264, 266, 269–70, 276, 283, 325, 333 embedded implicature 139–40, 142–8, 153, 155, 158–60, 317 generalized conversational implicature (GCI) 9, 11–12, 58–9, 111–12, 119, 124, 128, 133, 143, 162, 166, 268, 270–2, 275, 278–86, 327, 335 I-Implicature 111–13, 115, 118–19, 121 intrusive implicature 138, 142–3, 162 particularized conversational implicature (PCI) 9, 11, 128, 164, 268–72, 275, 278, 280–2, 285, 327 R-Implicature 121–2 scalar implicature 104, 113–15, 122–5, 128–31, 138–9, 141–3, 148–9, 153–5, 158–61, 164–7, 272–4, 277–8, 285, 310, 312–14, 317–19, 323, 326–7, 334–5 see also, conversation: conversational maxims, theory of conversation, Cooperative Principle impliciture 264 indexical 9, 21, 27, 258, 279, 290–1, 296, 305, 323; see also demonstrative indication 255, 266 intention 2–3, 15–18, 38, 86, 115, 120, 145, 153, 156, 159–60, 163, 202–5, 221, 224, 226–7, 230, 232, 235, 236–44, 248, 266, 271, 281, 284, 292, 296, 306; see also communicative intention, speakermeaning intrusive implicature: see implicature intuition 50, 55–6, 63–4, 66, 68–9, 72, 258–9, 267 irony 20, 41, 114–15, 117; see also metaphor irregular negation: see negation justificatory reason: see reason linguistic botanizing 4, 36, 42, 185 linguistics 1, 32, 50, 59, 65, 69, 315, 320 literal meaning: see meaning locutionary act 39, 256, 328; see also speech act
Subject Index logic 6, 32, 35, 41, 58, 61–2, 68, 98–9, 105, 109, 112–13, 138, 271–2, 276, 321, 333 logical form 123, 274, 276–7, 290, 293, 325 logical positivism 32–5, 41, 48, 53 maxims of Manner 11, 77, 99, 139, 148, 152, 159 Quality 11, 115, 160, 316 Quantity 10, 26, 115, 118, 121, 139, 151–3, 157, 159, 161, 164, 167, 311–12, 320–3, 333–4 Relation 11, 320–3, 333 see also conversation: conversational maxims meaning / use 4, 6–7, 14–15, 19, 39, 41, 50, 68 conventional meaning 11, 16, 20, 40, 112, 124, 127, 141–3, 162, 164 literal meaning 40, 58, 68, 106, 113, 115, 123–4, 126, 134, 139–40, 143–4, 207, 252, 256, 266, 271–2, 279, 284, 289, 312, 324 natural meaning 15, 26, 38, 225, 249 non-natural meaning 15, 26, 38, 252, 266 sentence meaning 7, 17, 20, 38–9, 41, 49–50, 54–5, 111, 121–2, 125, 131, 139, 156–7, 179, 271, 252, 256, 259, 264–6, 279, 281, 334; see also intention, pragmatics, semantic content, speaker-meaning, what is said metaphor 41, 123–4, 132, 296, 299–301, 303–5 metaphysics 2–3, 14, 24, 31, 42, 48, 51, 53, 62, 67–8 metonymy 111, 298–9, 305 minimal proposition 260, 264–5
347
43–4, 49, 52, 57–8, 65–6, 144, 185, 252, 259, 262–3, 267, 281, 290, 294–5, 300, 304, 311, 333–4; see also ordinary language philosophy ordinary language philosophy (OLP) 1, 4, 25, 31–2, 34–5, 38–40, 43, 51, 52; see also ordinary language
natural kinds 67, 70, 72 natural language: see ordinary language negation 25, 37, 75–6, 84, 86, 99, 103–4, 110, 112, 119, 120–1, 127–30, 160–1, 313–14, 317, 319–21, 325 irregular negation 25, 103, 105–9, 111–17, 120, 123–7, 129–30, 132–3
Paradigm Case Argument 49, 57, 63, 70, 73 Particularized conversational implicature (PCI): see implicature perillocutionary act 26, 221–2, 227–8, 230–3; see also speech act perlocutionary act 205, 223, 227–9, 231–2, 244 intention 238, 240–1 see also speech act person-body problem 59–61, 71–2 philosophical psychology 2–3, 14, 16, 24, 42, 196 philosophy of language 1–4, 24–5, 38–40, 48, 259, 288, 312, 335 polysemy 295–6, 299, 301, 304–5 practical reason: see reason pragmatic inference 304–5, 323, 335 intrusion 326–27, 334 see also expansion pragmatics: see semantics vs. pragmatics presumption 173–4, 179–81, 210, 212–13, 253, 303 presupposition 5–6, 10, 25, 75–6, 79–100, 103–5, 109, 113, 120–2, 124–5, 131, 134, 265, 319 proper name 59–60, 64–5 property 47, 57–64, 66, 299, 334; see also essential property proposition 20, 22–3, 32, 37, 50–1, 56–7, 67, 89–90, 104, 110, 118–19, 121–2, 129, 145–8, 159, 161, 167, 189, 203, 205, 212, 215–16, 257, 259, 263, 265–6, 269–70, 288, 290, 292–3, 295, 298, 306, 310, 314, 324–5, 335; see also semantic content, meaning: literal meaning, truth: truth-condition)
Occam’s razor 4, 98, 118, 312, 333 ordinary language (natural language) 5–6, 31, 35, 39–40,
rationality 1–3, 13–15, 18–19, 24–6, 42, 184, 187–8, 190–1, 193, 195–6, 200–1, 225–6, 230, 317, 333, 335
348
Subject Index
reason 17–18, 26, 42, 185–9, 192, 194, 196–200, 210, 215 justificatory reason 185, 187 practical reason 184–5 theoretical reason 185, 191–2 relevance 122, 152–3, 162–3, 282, 285; see also conversational maxims relevance theory (RT) 143, 158, 167, 268, 274–9, 263–4, 282, 284–5, 289, 291, 314, 321, 324, 328 scalar implicature: see implicature semantic content 144, 164, 269, 258, 260, 264, 281, 314, 335; see also meaning: literal meaning, meaning: sentence meaning, speakermeaning semantic minimalism 21, 26–7, 267–9, 279–82, 285, 289–91, 304, 306, 328, 333; see also formal semantics semantic underdetermination 21–3, 265, 278, 291, 297, 324; see also completion, context: contextsensitivity, contextualism, expansion semantics: see semantic content, semantic minimalism vs. pragmatics 6–7, 21, 27, 43, 58, 79, 87–8, 96–8, 103, 106, 112, 117–18, 121, 128, 132, 141, 158, 216, 252, 255–6, 259, 264, 269, 275, 280, 282, 297, 306, 317, 325, 335 showing 203, 208, 213, 217 speaker’s implicature 115, 178; see also implicature speaker-meaning 3, 7, 15–16, 23, 26, 38–40, 111–12, 128, 141, 147, 153, 163, 170–1, 177, 181, 202–6, 211, 214, 216–17, 225, 235, 237–8, 253–4, 262, 264, 281–2, 292, 324, 327, 330, 333; see also intention, meaning speech act 20, 39, 117, 144, 165–6, 205–7, 209–15, 217, 221, 223, 225, 227–8, 255–6, 259–63, 266 speech act pluralism 256–8, 267 see also communicative act, locutionary act, illocutionary
act, perillocutionary act, perlocutionary act Speaker Meaning Exhaustiveness 170–4, 177–81; see also speaker-meaning strengthening: see expansion synonymy 25, 32, 47–50, 53–7, 68–9, 256, 320 syntax 21–2, 64–5, 126–8, 130, 206, 257, 269, 279–80, 285, 288, 292, 294–5, 328 theoretical reason: see reason theory of conversation (theory of implicature): see conversation, implicature theory of description 4, 104; see also definite description theory of language 1, 7–9, 14, 23; see also philosophy of language theory of meaning 2–4, 7, 9, 15, 17, 19, 23, 25, 225, 259, 325, 334 Theory-theory 2–3, 7, 25 total signification of utterances 8–9, 23, 26, 39, 41, 253, 255, 269, 310 truth 48, 61, 67, 85, 252, 256, 260–1, 265, 326, 329–31 truth-condition 5–6, 10, 20, 23, 34, 68, 76, 98, 104–5, 109, 121, 138, 142–3, 149, 158, 164, 208, 254, 264, 271–6, 278–80, 282, 285, 288–91, 293, 295, 297–9, 301–3, 305–6, 316, 324–5, 328, 335; see also proposition understanding 156, 188, 224, 227, 229–32, 245, 246–9, 253 value 2, 13, 184, 188 what is said 6–7, 9–11, 17, 19–23, 26–7, 40–1, 43–4, 68, 113–15, 120, 140–3, 164–6, 170, 172–3, 178–9, 181, 254–67, 268–71, 276, 279, 289, 291–2, 310, 312, 323–4, 326, 328–31, 333–5; see also meaning:conventional meaning, meaning: sentence meaning