Palgrave Studies in Pragmatics, Language and Cognition Series Editors: Richard Breheny and Uli Sauerland Series Advisor...
16 downloads
818 Views
745KB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
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 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
Assertion Mark Jary Roehampton University, London, UK
© Mark Jary 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 author has asserted his right to be identified as the author 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-13: 978–0–230–57399–4 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 Jary, Mark, 1964– Assertion / Mark Jary. p. cm. — (Palgrave studies in pragmatics, language, and cognition) Includes bibliographical references and index. ISBN 978–0–230–57399–4 1. Grammar, Comparative and general—Mood. I. Title. P281.J37 2010 415 .6—dc22 2010002698 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
Dedicated to the memory of Greta Hutchings
This page intentionally left blank
Contents
List of Figures and Tables
ix
Acknowledgements
x
1 Introduction
1
2 Assertion in Speech-act Theory 2.1 Introduction 2.2 Two types of speech-act theorist 2.3 Speech-act fundamentalists and assertion 2.4 Conclusion
6 6 6 16 30
3 Assertion, Belief and Knowledge 3.1 Introduction 3.2 Assertion and belief: what is the correct order of explanation? 3.3 Assertion and belief attribution 3.4 Assertion and knowledge 3.5 Conclusion
32 32
4 A Sign of Assertion 4.1 Introduction 4.2 Frege’s assertion sign 4.3 Assertion: illocution and inference 4.4 A natural-language assertion sign? 4.5 Conclusion
52 52 53 56 59 82
33 38 46 50
5 Assertion and Common Ground 5.1 Introduction 5.2 Sense, force and representation in Stalnaker’s account 5.3 Assertion and presupposition 5.4 Comparison with Relevance Theory 5.5 Conclusion
84 84 85 92 99 103
6 Assertion and Mood 6.1 Introduction 6.2 Mood in formal semantics
105 105 105 vii
viii
Contents
6.3 Speech-act approaches to mood 6.4 Conclusion
120 128
7 Assertion and Main Point 7.1 Introduction 7.2 Non-asserted main point 7.3 Assertion, presupposition and conventional implicature 7.4 Ordered entailments 7.5 Conclusion
129 129 130
8 Assertion and Relevance 8.1 Introduction 8.2 Relevance in a context vs. relevance to an individual 8.3 Context types 8.4 Mood and force 8.5 Assertion and main point 8.6 Relevance maximising vs. relevance optimising 8.7 Comparison with other frameworks 8.8 Conclusion
150 150 151 152 155 173 177 192 194
9 Conclusion
196
Notes
199
References
211
Index
219
134 138 148
List of Figures and Tables
Figures 6.1 Fulfilment conditions 7.1 Ordered entailments 8.1 Types of assumption
110 139 154
Tables 4.1 Modes of commitment 8.1 Assertion and main point compared 8.2 Summary of relationship between interpretation strategies and metarepresentational requirements
ix
70 173 186
Acknowledgements
This book is the result of many years’ work, and many people have helped along the way. Foremost amongst those is Robyn Carston, from whose generous and rigorous supervision I benefited from greatly whilst studying at UCL. Another member of the UCL Linguistics Department faculty who has been a great source of support and intellectual stimulation is Deirdre Wilson, whose work with Dan Sperber provides the theoretical framework in which the ideas developed in the final chapters of the book are grounded. I have also benefited greatly from the support and encouragement of Rob Stainton, who generously read the manuscript and made many useful objections and suggestions. At UCL, I had the pleasure of interacting with a number of fellow students and members of the faculty. Those whose words and thoughts have certainly influenced my thinking include Nicholas Allot, Richard Breheny, Alison Hall, Corrine Iten, Milena Nuti, George Powell, Paula Rubio Fernandez, Neil Smith and Tim Wharton. Beyond UCL, I have also benefitted from discussions and email exchanges with Kent Bach, Stephen Barker, Coralie Chevallier, Manuel García-Carpintero, Mitch Green, Mikhail Kissine (who also read the manuscript and made very useful comments) and Agustín Vicente. Much of this book was written during a six-month period of study leave awarded by the School of Arts at Roehampton University, for which I am grateful. I also want to thank my friend Silvia Vidaurreta for the loan of her flat in Spain, where I worked on the final chapters of the book. The series editors Richard Breheny and Uli Sauerland have also helped with useful comments and encouragement. At home, my wife Helen and our children, George and Rachel, are a constant source of love and support, without which I would do very little, let alone write a book.
x
1 Introduction
This book has two main aims. One is to bring together and discuss in a systematic way a range of perspectives on assertion: philosophical, linguistic and psychological. As far as I know, this has never been done before.1 The other is to present a view of the pragmatics of assertion, with particular emphasis on the contribution of the declarative mood to the process of utterance interpretation. I hope that in laying the ground for the achievement of the second aim, I fulfil the first. As with so many topics relating to linguistic meaning, the study of assertion starts with Frege. Before Frege, the distinction between the conceptual content conveyed by a statement – what Frege referred to as ‘a thought’, but today would be called a proposition – and the act of putting forward and openly committing oneself to that content was not adequately appreciated. Frege rectified this, and gave assertion a key role both in his philosophy of language and in his logical symbolism. Reactions to Frege’s insight have differed. It has been taken by some to provide the grounds for a use-based theory meaning. Brandom, for example, seeks to explain meaning in terms of the consequences of assertion, understood in terms of the practical and inferential commitments and entitlements taken on by asserters. Others, however, have chosen to play down the significance that Frege gave assertion. Indeed, in most analytical philosophy of language since Frege, the tendency has been to factor out the act and focus on the content. On this view, propositions, not assertions, are the fundamental bearers of truth conditions. When force is discussed in this framework, it is generally seen as something that can be bolted on to truth-conditional content, rather than something that is integral to that content. Chapter 2 of this book looks at assertion from both these perspectives. It starts by looking at the approach to assertion in what I term 1
2
Assertion
‘traditional speech-act theory’, as typified by the work of Searle and Bach and Harnish. On this view, assertoric force is just one force among many that can be attached to propositional content, and is awarded no higher theoretical status than, say, directive force. This contrasts with the approach I call ‘speech-act fundamentalism’. One key characteristic of this view, championed by Brandom, is that assertion is the basic unit of analysis in the study of linguistic meaning. Other illocutionary forces are explained in terms of assertoric force, as, indeed, is conceptual content, though I do not go into that aspect of Brandom’s work. Rather, I compare his framework with that of Barker, who also takes a fundamentalist approach to speech-acts and gives assertion a central role in the analysis of other forces. Assertion is closely related to the notion of belief: assertions express beliefs. That is a platitude, but there is much more that can be said about the precise nature of the assertion-belief relationship, and this is far from platitudinous. First, there is the question of conceptual priority. Modern views on the pragmatics of linguistic communication might lead one to see belief as conceptually prior to assertions, for, due primarily to its Gricean legacy, modern pragmatics tends to view the interpretation of assertions as a matter of belief attribution. Add to this the widely-held view that linguistically encoded meaning radically underdetermines the content of our assertions, and one is easily led to the conclusion that assertions derive their content from the beliefs they express. However, if the question is considered from a perspective unburdened by consideration of the processes of utterance interpretation, the converse order of conceptual priority is possible. According to this view, as propounded by Dummett, assertion is conceptually prior to belief. Dummett’s arguments to this effect are discussed in Chapter 3, where it is argued that this position is compatible with the linguistic-pragmatic view that assertoric utterances ultimately derive their content from the beliefs they express. Despite Grice’s influence, not everyone accepts that the interpretation of assertoric utterances is essentially a matter of belief attribution. While it is widely agreed that assertions express beliefs, there is less consensus about whether hearers, when interpreting utterances, necessarily treat assertions as expressions of belief. There are those who argue that assertions are first and foremost treated as sources of information about the world, and constitute a form of perception-by-proxy. Holders of this view, such as Dummett, Millikan and McDowell thus stand in opposition to Grice and his followers, such as Sperber and Wilson and Bach and Harnish, for whom utterance interpretation is belief attribution. These
Introduction
3
positions and their motivations are also discussed in Chapter 3, where an ecumenical stance is advocated. Views on the relationship between assertion and belief have a bearing on views of the force-form relationship. For those who assign assertion a fundamental explanatory role in their theory of linguistic meaning, the notion of a format specified for assertion is crucial. This is because, for those taking this position, there must be a type of behaviour whose tokening counts as asserting, no appeal to a more basic meaningful entity, such as belief, being possible. For these authors, the sign of assertion is the declarative mood, and the production of this sign under the right conditions is constitutive of assertion. Those who see assertion as case of belief attribution, by contrast, are able to posit a much looser link between mood and force, and even to deny that there is any special relationship between the declarative and assertion. In Chapter 4, I look at arguments that have been made to this effect, and show that cases of non-assertoric use of the declarative are actually straightforwardly explained by the assumption that the declarative is indeed specified for the performance of assertions. This Chapter focuses heavily on Dummett’s claims concerning the assertion-declarative relationship, highlighting the crucial distinction he draws between the question of whether the positing of an assertion sign is necessary to explain what assertion is, and the question of whether an assertion sign is required in order that hearers can recognise the speaker’s assertoric intentions. Up to and including Chapter 4, the book examines assertion from a primarily philosophical perspective. However, as stated above, a key aim of the book is to provide an account of the mechanics of the interpretation of assertions, with particular emphasis on the role of the declarative mood in this process. From a philosophical point of view, it is sufficient to say that the declarative is a sign of assertion. From a linguistic perspective – especially one concerned with the pragmatics of utterance interpretation, this is not enough. Rather, one needs to identify the features of the declarative that make it apt for the making of assertions. This can only be done within a theoretical framework, the features assigned to the declarative being part of the vocabulary of that framework. Chapter 5 provides a bridge from the philosophical to the linguistic. It deals with Stalnaker’s common-ground view of assertion, a philosophical account that has provided the basis of a number of linguistic accounts of the declarative mood. These accounts, and others, are discussed in detail in Chapter 6, where formal-semantic accounts of mood are contrasted with varieties of speech-act accounts.
4
Assertion
Within linguistics, assertion is thought of not only in terms of truthcommitment, but also in terms of information structure. It is common to find asserted information identified as that which is the main point of the utterance. This aspect of assertion is discussed in Chapter 7, where it is shown that main-point status and assertion do not always coincide: main points need not be asserted by the speaker and what is asserted need not have main-point status. This Chapter also discusses the relationship between assertion and presupposition, a topic raised in Chapter 5 on Stalnaker. Following Stalnaker, presuppositions are commonly regarded as propositions treated as having common-ground status. I defend an alternative approach, on which presuppostional characteristics are indicative not necessarily of common-ground status, but of a particular functional role in the interpretation process. This view of presuppositional phenomena is found in the work of Sperber and Wilson, and it is their Relevance Theory framework which is employed both in this chapter and Chapter 8. Chapter 8 presents a novel account of the role of the declarative mood in the interpretation of assertions, and also puts forward an account of its non-assertoric uses. Because it is grounded in the notion of relevance, this account has the advantage of providing an explanation of why we attend to assertions: they come with the presumption that they will be relevant to us. However, this assumption is conveyed by all utterances, so what sets assertions apart? The claim I make is that assertions, unlike directives and questions, present the proposition expressed by the utterance as relevant in its own right. The role of the declarative is to mark the proposition expressed as potentially relevant in this manner. Nonassertoric uses of the declarative are explained in two ways: either the proposition is presented as relevant in its own right, but in an embedded context that represents, say, the content of a fiction or supposition; or it is relevant in its own right in a basic context that represents the actual world, but witnessing the performance of the act is sufficient grounds for accepting that proposition. If witnessing the act is sufficient grounds for acceptance, then the utterance does not count as an assertion for, as is explained in Chapter 8, assertion is characterised by certain cognitive and social safeguards that come into play when there is a possibility of the hearer being misled or deceived. Analysing assertion in terms of relevance has the further advantage of shedding light on the relationship between assertoric force and mainpoint status. On the account presented in Chapter 8, an assertion is a case of a proposition being presented as relevant in its own right to an individual. For a proposition to be relevant to an individual, in
Introduction
5
the technical sense employed in Chapter 8, it needs to be true. Thus the truth requirement on assertions follows from their special claim of relevance. Main-point status is also analysable in terms of relevance: the proposition communicated by an utterance that makes the greatest contribution to the overall relevance of the utterance is its main point. On this view, assertions often have main-point status because their point is to present the proposition expressed by the utterance as relevant to the individual. Under certain conditions, however, assertion and main-point status diverge. This is explained in Chapter 8. Chapter 8 also returns to issues raised in Chapter 3 concerning the relationship between assertion and belief. In Chapter 3, I argue that the interpretation of assertions need not necessarily involve belief attribution. In Chapter 8, I show that, of a range of possible relevance-driven utterance interpretation strategies, only the more sophisticated require belief attribution. This observation has a number of implications. One is that the case made by Sperber and Wilson for a pragmatics module is greatly weakened. Another is that, pace Sperber and Wilson, utterance interpretation is often a relevance-maximising, rather than relevance-optimising, procedure. The book ends with a short concluding chapter. Before starting on the book proper, both a note on terminology and an apology for a glaring omission are in order. In the literature on mood, the form associated with assertion is sometimes referred to as ‘the indicative mood’, and sometimes as ‘the declarative mood’. The former is most common among philosophers, but linguists are prone to object that the indicative is, strictly speaking, a verbal inflection rather than a clause type. I have used the term ‘declarative’ pretty much throughout this book, to refer to both embedded and unembedded instances of the form standardly associated with assertions – as opposed to commands and questions – in traditional grammars. The glaring omission is the almost complete absence of any discussion of the importance of intonation in relation to assertion. This is particularly important in languages, such as Spanish and Italian, where interrogative sentences are not marked syntactically or morphologically and are hence syntactically identical to declaratives. Clearly, when I make claims about the function of the declarative in relation to assertion, I mean when it is uttered with the appropriate intonation contour.
2 Assertion in Speech-act Theory
2.1 Introduction Although, as we will see later in this book, the notion of assertoric character is employed, tacitly or otherwise, in the study of inference and representation, assertion is first and foremost an action. Accordingly, we do well to start with an overview of assertion as it has been viewed in speech-act theory. This will serve a number of purposes. First, it will highlight issues relating to assertion to be discussed in later chapters, such as the relationship between assertion, truth and belief. Second, it will allow discussion of certain fundamental issues in some depth. One of these concerns which features of assertion – informativeness, truthcommitment, belief expression, explicitness – are central to its analysis. Another is whether assertion should be analysed as a sister of other illocutionary acts, or whether it is more fundamental, and therefore correctly seen as a superordinate species of act. The chapter begins by distinguishing two types of speech-act theorist. Their two approaches are then discussed, and the work of major players in each camp compared and contrasted in some detail. I end the chapter by highlighting the points raised in this chapter that will be important in the remainder of the book.
2.2 Two types of speech-act theorist Meaning theorists with an interest in speech acts tend to fall into two camps. For one, speech-act theory is a component of a theory of linguistic meaning, one that is designed to work in conjunction with a truth-conditional account of some notion of ‘core meaning’. As we will 6
Assertion in Speech-act Theory
7
see, such approaches tend to treat the proposition as this core element of meaning, viewing assertion as being just one of the many things that speakers can do with it. For these theorists, the key question concerning assertion tends to be how this act is to be distinguished from others such as ordering, promising and questioning. Probably the best known member of this group is Searle, who is justly considered one of the founders of speech-act theory, so we’ll call this camp the ‘traditional speech-act theorists’. But of course, while a founder, Searle is not the father of speech-act theory. That mantle rests with Austin (1962/1975), who, like Wittgenstein (1958), argued that philosophers of language had placed too much emphasis on the descriptive – or propositional – element of declarative sentences, with the result that they ignored the uses to which sentences of all types were put and hence failed to take into account a crucial social element of linguistic meaning. For those in the second camp, this social element of meaning is not merely crucial, but central: rather than viewing speech act-theory as a bolt-on to truth-conditional accounts, they argue that linguistic meaning should be explained in terms of the uses to which linguistic forms are put. Detailed and well articulated theories of this type have been proposed by Brandom, Alston and Barker. Because they see speech acts as fundamental to linguistic meaning, we’ll call them ‘speech-act fundamentalists’.1 They are of particular interest here because, unlike the traditionalists, they tend to give special status to assertion, claiming (in the case of Brandom and Barker, though not Alston) that other sentential illocutionary acts must be understood in terms of assertion. We begin by looking at the traditionalists. 2.2.1 Speech-act traditionalists and assertion 2.2.1.1 Individuating assertion: truth, belief and informativeness The view of assertion as a sister, rather than the mother, of other illocutionary acts has its roots in Austin’s rejection of the distinction he introduced, at the start of How to Do Things with Words, between constative and performative utterances. Austin first suggested that utterances might usefully be divided into two types, depending on whether they can be aptly judged true or false. Those that can, he termed ‘constatives’. Those that cannot, he suggested, have their meaning in the acts that they can be used to perform, hence the term ‘performative’. Examples of these included explicit performatives such as (1) and (2), which share with constatives declarative word order, and acts performed using
8
Assertion
other sentence-types, such as orders given using imperatives (3), and questions asked using interrogatives (4): (1) (2) (3) (4)
I warn you not to come any closer. I promise to be good. Stand at ease. What’s your name?
Rather than being true or false, utterances such as these were said by Austin to be felicitous or infelicitous according to whether certain conditions were in place. Thus one reason an order would be infelicitous, for example, is that the act ordered had already been carried out. Austin’s eventual rejection of the constatives/performative distinction was the result of two observations (Lecture XI). First, he noted that to state2 – the prototypical constative – is as much to do something as is to promise or to warn, as evidenced by the coherence of discourse such as (5): (5) When I said that the house was old, I wasn’t warning you against buying it, I was just stating a fact This suggests that merely being amenable to judgments of truth and falsity does not preclude an utterance from being also a case of doing something with words. Austin’s second observation was that statements were subject to the same sort of infelicities as performatives: for example, just as a promise can be insincere if the speaker has no intention of carrying out the act described, so a statement can be insincere if the speaker does not believe the content of her assertion. These considerations led Austin to conclude that ‘Stating, describing etc. are just two names among a very great many others for illocutionary acts; they have no unique position’ (Austin 1962/1975: 148–9, original emphasis). In seeking to develop Austin’s ideas, Searle (1969) set out to establish clear criteria for distinguishing illocutionary acts. Using a detailed analysis of promising as his starting point, he put forward four dimensions along which illocutionary acts can vary. The first is in terms of the propositional content that is acceptable for each act: whereas assertions and yes/no-questions can take any propositional content, some other acts are more restricted: for example, promises, according to Searle, can only express propositions denoting future acts performed by the speaker, while requests must concern future acts performed by the hearer.3 The second dimension relates to the conditions that must be in place for an act to be performed felicitously: for Searle, an assertion that P is
Assertion in Speech-act Theory
9
infelicitous if the speaker does not have evidence for the truth of P, or if it is clear to both speaker and hearer that the hearer already knows that P. Along the same dimension, a request is infelicitous if the speaker believes that the hearer is not able to perform the act requested. Illocutionary acts also differ in terms of the psychological states they are related to: assertion with belief, requests with desire, for example. Searle’s final dimension concerns what an utterance must count as if it is to be regarded as being of a particular type. Searle calls this its ‘essential condition’. The essential condition for assertion, according to Searle, is that it count as ‘an undertaking to the effect that P represents an actual state of affairs’, while requests, for example, must count as an attempt by the speaker to perform the act described by the utterance (1969: 66). Searle’s characterisation of assertion raises a number of issues. For him, the essence of assertion lies in an undertaking by the speaker that the proposition expressed by the utterance represents an actual state of affairs. However, the nature of this undertaking is not elaborated on. We will return to this issue, which is highlighted by Brandom (1994: 173), when we consider speech-act fundamentalist views of assertion in section 2.3.1. Another issue that will have to wait until later is the relationship between belief and assertion, which will be given a chapter of its own (Chapter 3). A point that can be considered now, though, is the fact that, for Searle, the intention to communicate the proposition expressed by the utterance is not essential to the notion of assertion. On Searle’s account, an assertion that P will be unhappy if it is made under circumstances in which it is obvious to both the speaker and the hearer that the hearer already knows that P, but it will still be an assertion, for it will still count as an undertaking that P is true.4 This is not the view taken by Bach and Harnish. Following Grice, these authors see linguistic communication as involving primarily the expression and recognition of speaker attitudes. On this view, assertion differs from other illocutionary acts primarily in terms of the attitude it expresses: belief, as opposed to desire, intention, gratitude, sorrow, etc. For Bach and Harnish, an utterance expressing P counts as an assertion only if the speaker intends the hearer to take her utterance as reason to believe both that she believes that P and that she intends the hearer to believe that P (1979: 16).5 Only the recognition of the intention is crucial: its fulfilment – so that the hearer actually believed that the speaker believed that P and/or came to believe P himself – would be a perlocutionary consequence of the act and hence not a requisite for its successful performance. Among the traditionalists, then, there is disagreement about the importance of communication to the analysis of assertion.
10
Assertion
That there is disagreement is not surprising, for there are good reasons both for wanting and for not wanting an informative element in a definition of assertion. We want to establish a set of criteria by which to identify an illocutionary act as an assertion. Such an endeavour will be successful to the extent which it classifies acts correctly. However, because there are cases of assertion in which the speaker does not intend to inform the hearer – perhaps because she does not expect the hearer to accept the assertion, because the hearer already accepts what is being asserted, or because there is no addressee – an account for which the presence of informative intentions is a defining feature of assertions will fail to classify some acts correctly. This is the problem faced by Bach and Harnish’s definition: it does not class as assertions those cases where the speaker has no intention of having his audience believe the content of his statement. Imagine someone accused of heresy being tortured so that she recant her heretical views. But she does not; rather she reaffirms them. In such a case, she cannot be said to intend to have her hearers believe the content of her utterance, but she surely makes an assertion. Or consider cases where the speaker is already sure that the hearer believes the proposition expressed (and that he does not need reminding of it), such as when a teacher asks a pupil ‘What’s the capital of France?’ In responding ‘Paris is the capital of France’, the pupil asserts that Paris is the capital of France, even though he can have no intention that the teacher should come to believe this.6 But attempting to come up with criteria for correctly classifying acts as assertions is not the only aim a theorist might have in developing a characterisation of assertion. It is equally legitimate – and necessary – to ask why the practice of assertion exists, and an adequate answer to this question must surely give informativeness centre stage. Assertions exist, and we are disposed to attend to them, because they are a source of information about the world beyond our perceptual capacities. This much seems obvious, and the burden of proof is surely on anyone who would deny it. Consequently, a complete understanding of assertion cannot be arrived at without considering its informative function. We appear, then, to be in a dilemma. On the one hand, we want an account of assertion that does not exclude non-informative assertions; on the other hand, we want that account to reflect the fact that the importance of assertion as a social activity lies in its function as a means of in information transfer. Fortunately, Millikan’s notion of proper function offers us a way out. According to Millikan (1984; 1993), a proper function of a behaviour is defined in terms of its history: ‘What caused that behaviour to be repeated?’ is the question that must be asked in
Assertion in Speech-act Theory 11
determining a function. It is surely uncontroversial to say that what caused the practice of assertion to be repeated is its potential for conveying information about how the world is beyond the perceptual reach of the hearer.7 On this understanding of ‘function’, assertion has an informative function even if many cases of assertion are neither informative, nor intended to be so, in much the same way that sex has a reproductive function even if many human sexual acts neither result in reproduction nor are intended to do so. Thus, while Bach and Harnish’s definition fails as a means of correctly classifying all acts of assertion, it is correct in seeking to include informativeness as a necessary element of a complete understanding of assertion. Searle’s definition of assertion as being essentially an undertaking to the truth of the proposition expressed might capture all cases of assertion and certainly identifies a feature central to the analysis of this act, it does not get at its function. Rather, Searle’s account reduces its central function to a subsidiary consideration. In sum, in order to arrive at a characterisation of assertion that acknowledges its informative function, in Millikan’s sense of ‘function’, whilst allowing that not all cases of assertion seek to inform, some qualification is needed. We would do well, then, to follow Sperber and Wilson (1986/1995: 181) and call an assertion that P which is intended to inform the hearer that P an ‘ordinary assertion’. This allows for the existence of unordinary assertions where there is no intention to inform the speaker of the proposition expressed, whilst also recognising that the informativeness is essential to an adequate characterisation of assertion.
2.2.2 The assertive family and direction of fit As well as being an act in its own right, assertion has also been viewed by speech-act theorists as prototypical of a group of illocutionary acts. It is clear that asserting is not distinct from, say, informing, reminding and complaining in the way that it is distinct from promising and ordering, and this observation underlies attempts within speech-act theory to taxomonise illocutionary acts. This practice was started by Austin (1962/1975: Lecture XII), and various alternative systems have since been proposed by, among others, Searle (1975/1979), Bach and Harnish (1979) and Alston (2000). Those who put forward such taxonomies generally take whatever feature they see as central to assertion as a necessary condition for being a member of the group for which it serves as a prototype. Thus, to be
12
Assertion
member of Searle’s group of assertives, an illocutionary act must count as an undertaking to the truth of the proposition expressed by the utterance (1975/1979: 12), while to be a member of Bach and Harnish’s class of constatives, an act must express both a belief and an intention that the hearer ‘form (or continue to hold) a like belief’ (1979: 42). Members of the group then vary in a variety of ways, such as according to the strength of speaker commitment (e.g. guessing vs. insisting) or whether the speaker is responding to a query. Searle (1975/1979: 2–8) lists twelve criteria according to which speech acts might be classified, but considers three of these to be the most important. These are differences in the point or purpose of the act, in the psychological state expressed, and in the direction of fit between the words and the world. Direction of fit concerns whether the aim of the utterance is to get the words to fit the world or to make the world fit the words. An assertion is true if the proposition it expresses obtains in the actual world, and hence has word-to-world fit; a command, for example, is fulfilled if the world is changed such that the proposition expressed obtains, and hence has world-to-word direction of fit. Searle suggested that it would be elegant if the whole of his taxonomy could be built around this distinction, but found himself unable to do so, largely because certain expressive speech acts, such as greetings and congratulating (i.e. members of Searle’s class of expressives), have no direction of fit.8 However, Recanati (1987) has suggested that a taxonomy of speech acts can indeed be developed on the basis of direction of fit, as long as this is restricted to what he terms ‘content-conveying acts’, a restriction that excludes expressives. By ‘content-conveying’, Recanati does not mean informative, but refers rather to acts that have a central aim of describing either how the world is, or how it is to be as a result of the utterance. Reviving Austin’s terminology (though not the basis of his distinction), he calls the former ‘constative’ and the latter ‘performative’. Performatives, on Recanati’s account, are content-conveying utterances with world-to-word direction of fit, while constatives, of which assertion is prototypical, are individuated by word-to-world fit. Recanati does not deny that assertions have a special relationship with belief and truth, though he does not take a position on the precise nature of that relationship as he does not rely on it to distinguish the class of acts that assertion belongs to from other acts (1987: 137–8). As we have seen, this is in contrast to Searle and Bach and Harnish, who, respectively, see truth-commitment and belief expression as central.
Assertion in Speech-act Theory 13
The notion of direction of fit, though, is a problematic one, especially when applied to future-tense utterances. Take an utterance of (6), for example: (6) I’ll be at the party This could be either an assertion of future intent or a promise. In the case that the speaker turns up at the party, the assertion is judged true or the promise deemed to have been kept. In both cases the speaker has brought about the state of affairs described, though the direction of fit is said to be different in each case: word-to-world in the assertion case, world-to-word in the promise case. It is hard to see what motivates this claim, though, other than the stipulation that promises have the latter fit and assertions the former. To be sure, the promise can count as a reason for the speaker to be at the party in a way that the assertion cannot (Watson 2004), but it is not clear how the notion of directionof-fit captures this. As we will see in Chapter 4 the differences between assertions and promises are best understood in terms of the grounds the utterance provides for accepting the proposition expressed, rather than in the nature of the relationship between the proposition expressed and the world. Here we just need to note that the supposed direction-of-fit distinction is hard to maintain. Another problem is that direction of fit suggests symmetry where there is none. To have word-to-world direction of fit is, essentially, to be truth apt. To have world-to-word direction of fit is to be compliance or fulfilment apt. But these latter terms cannot be explicated except in terms of truth. World-to-word and word-to-world directions of fit are not equals, then. Rather, the latter is superordinate to the former. Direction of fit has also been applied to the analysis of belief and desire. Roughly speaking, the idea is while a belief that P should be cancelled in the case that not-P, a desire that P endures under such circumstances. However, as Sobel and Copp (2001) have shown, close application of this criterion proves difficult. As these authors suggest, then, direction of fit is best treated as a useful metaphor. Applied to speech acts, it has intuitive appeal but should not, I think, be used as a basis for distinguishing assertion (or assertives) from other acts (or classes of acts). 2.2.3 The explicit nature of assertions The final means of individuating assertion and other assertive acts that we will consider in this subsection is that proposed by Alston (2000).
14
Assertion
To a certain extent, Alston straddles the two camps identified at the beginning of this chapter: with the traditionalists, he sees assertion as being a sister to acts such as ordering and promising, but, like the fundamentalists, he avoids affording propositions the role of bearers of core-meaning. Alston’s analysis of illocutionary acts is markedly different from those discussed so far. He takes as his starting point Austin and Searle’s observation that the successful and felicitous performance of an illocutionary act relies on certain conditions being fulfilled. For example, a request that someone open the door is defective if (among other things) the speaker does not want the door to be opened, believes that it cannot be opened, or if the door is already open. Alston’s suggestion is that in performing an illocutionary act, a speaker takes responsibility for the holding of those conditions that must be in place if the act is not to go awry. The notion of responsibility that Alston intends is not that the speaker must be the cause of the conditions – in the sense that it is the speaker who has brought them into existence – but that the speaker is open to rebuke if these conditions are not met (Alston 2000: 54). The reason that the speaker is open to rebuke, Alston argues, is that illocutionary acts are norm governed, the conditions for their felicitous and successful performance being rules which govern their use (2000: 58–63). Having established this notion of taking responsibility, Alston then distinguishes classes of illocutionary acts according to the types of responsibility taken. In issuing an order, for example, the speaker takes responsibility for it being possible for the hearer to do the act described, for the speaker having the authority to place on the hearer an obligation to do the act, and for placing that obligation on the hearer by uttering the sentence (Alston 2000: 102). Thus the speaker is open to rebuke if it is not possible for the hearer do the act or if, for example, the speaker does not have the requisite authority to require the hearer to perform the act. The possibility of the speaker performing any given illocutionary act is dependent in part on what Alston terms ‘conceptually necessary conditions’ (Alston 2000: 93, fn. 8). Thus, if the order is that the hearer open the door, then it is conceptually necessary that there exists a uniquely identifiable door, and that this door is closed. So, according to Alston, in uttering, (7), the speaker takes responsibility for it being the case that (8) and (9) obtain. (7) Open the door. (8) There is uniquely identifiable doorx . (9) Doorx is closed.
Assertion in Speech-act Theory 15
In issuing an order by uttering an imperative sentence, then, the speaker takes responsibility, on Alston’s account, for it being the case that the world is a certain way (and places the hearer under an obligation to change it). It is even possible that the hearer is informed, by the utterance of (7), that the doorx is closed. However, the speaker has not asserted that the doorx is closed, on Alston’s account, for the simple reason that the utterance does not explicitly express the proposition that the door is closed. Thus, for Alston, what is distinctive about the assertive class of illocutionary acts is that the speaker takes responsibility for the obtaining of the proposition explicitly expressed by the utterance. Assertions thus make explicit a proposition the obtaining of which the speaker takes responsibility for, and it is this, on Alston’s account, that is their defining characteristic (2000: 116–20). The explicitness criterion for distinguishing assertions has a number of advantages. It acknowledges that a speaker generally expresses belief in (and undertakes a commitment to the truth of) a number of propositions regardless of the illocutionary act she is performing. For example, utterances of (7), (10) and (11) would each commit the speaker to truth of (and indicate that she believes) (8) and (9). (10) Thank you for opening the door. (11) I promise to open the door. Alston’s characterisation of assertion also acknowledges that speakers, when making assertions, commit themselves to propositions other than that which is explicitly expressed. Thus, an assertion of (12) commits the speaker not only to the proposition expressed by that utterance, but also to (8), (9), (13) and (14). (12) My brother will close the door. (13) The speaker has a brother. (14) The speaker’s brother will do something. As noted by Barker (2002), the explicit nature of assertions is rarely commented upon in the literature. Indeed, some authors deny that what is asserted need be explicitly expressed. For example, Russell seems to have thought that all the elements of the logical form of a definite description are asserted by the speaker who asserts a sentence containing that definite description: ‘Thus when we say “x was the father of Charles II.’ we not only assert that x had a certain relation to Charles II., but also that nothing else had this relation.” (Russell 1905: 481–2) Geach (1965), on the other hand, has it that subordinate clauses introduced by ‘the fact
16
Assertion
that’ are asserted by the speaker, as are those introduced by factive verbs such as ‘point out’. In cases such as (15) and (16) the speaker is said by Geach to have performed a ‘double-barrelled assertion’, with one assertion committing the speaker to the proposition expressed by the main clause, and the other to that expressed by the subordinate clause. (15) Peter is aware of the fact that his wife has arranged a surprise party for him (16) Peter pointed out that Chomsky is an American citizen Geach also claims that such a double-barrelled assertion takes place with ‘under the illusion that’, as in (17): (17) Peter is under the illusion that he is the son of God In this case, the ‘embedded assertion’ is the negation of the embedded clause (i.e. that Peter is not the son of God), so, for both Geach and Russell, it appears that commitment to truth is all that is required for a proposition to count as asserted. Explicitness, for these authors, is not a criterion. The problem with this position, as Alston points out, is that it fails to distinguish asserted propositions from those for which responsibility is taken in the performance of non-assertive illocutionary acts. As we saw above, the speaker of (7) takes responsibility for the truth of (8) and (9). The utterance (7), though, is not an assertion. Do we want to say that when issuing directives containing, say, a definite description we assert all the propositions associated with the use of that description? This would make assertion a potential element of every illocutionary act. Alston’s explicitness criterion avoids this consequence.9
2.3 Speech-act fundamentalists and assertion In order to spell out what is distinctive about the proposals of Barker and Brandom, we first need to look briefly at the relationship between traditional speech-act theory and truth conditional semantics. Speech-act traditionalists, as was noted at the start of this chapter, see speech-act theory as a complement to truth-conditional accounts of meaning. Their explanatory strategy is to divide utterance meaning into two components: a propositional element and a force element. The force/content distinction was, of course, introduced by Frege
Assertion in Speech-act Theory 17
(1918–19/1997), a variant of it being adopted by Searle (1969: 31), who uses F(P) notation to distinguish the force and propositional content of an utterance. Employing this notation, (18) stands for an assertion that P, (19) a question whether P, (20) a command to P, and (21) a promise that P.10 The notation can be extended to cover all sentential illocutionary acts by introducing a symbol for each act. (18) (19) (20) (21)
(P) ?(P) !(P) Pr(P)
One role played by propositions in traditional speech-act theories, then, is to explain a putative core element of meaning common to a set of utterances such as (22) to (24): what these have in common, on this view, is that they express the same proposition, which we can show as (25): (22) (23) (24) (25)
Peter will come. Will Peter come? Come, Peter.
On the truth-conditional account of meaning generally assumed by traditional speech-act theories, propositions play the role of ultimate bearer of truth. As that account sees truth-conditions as the means by which linguistic meaning is to be explained, it follows that, in these theories, propositions play the role of the ultimate explanans in linguistic semantics. However, assigning this role to propositions qua truth-bearers raises the problem of how to explain those aspects of sentential linguistic meaning which are not amenable to analysis in terms of truth and falsity, notably the non-assertive illocutionary acts and the non-declarative sentence-types associated with them. Faced with the problem that only assertions can be aptly judged true or false, theorists generally take one of two paths: either they exclude non-declarative sentences and/or non-assertive utterances from their analysis, or they introduce a superordinate term that places being true on a level with being complied with, in the case of orders, and being answered, in the case of questions. Thus truth conditions are seen as a type of satisfaction condition, on a par with compliance conditions and answerhood conditions.11
18
Assertion
The first approach is clearly inadequate if one wants a complete account of linguistic meaning. The second has the disadvantage of reducing truth, which is supposed by truth conditionalists to be the bedrock on which the theory of meaning rests, to a type of satisfaction. But satisfaction, in this sense, cannot be defined except in terms of truth. To see this, consider under what conditions an order is said to be satisfied. An order is satisfied, on this account, if it is complied with. But under what conditions is it complied with? The answer must be that that it is complied with if the proposition it expresses is made true (by the addressee). But being true, on this account, is another form of satisfaction. So the term we are attempting to define gets re-introduced in the definitions. This is not a feature of acceptable definitions. It might be responded that having compliance explained in terms of truth is to be desired, as it shows that the fundamental notion of truth can be applied to the analysis of non-assertive illocutionary acts and linguistic forms associated with these. Thus it might be suggested that the superordinate notion of satisfaction be dropped, and that the notion of compliance be analysed in terms of making true, thus saving the truthconditional approach. But this is to accept that propositionhood alone is insufficient to qualify an object for the status of truth bearer, and that assertives – being the only class of illocutionary acts aptly described as true or false – have a special relation to truth, and thus a special role to play in both in a truth-based theory of meaning and in the analysis of illocutionary acts. It is, in effect, to reject the view that propositions are the ultimate bearers of truth and the fundamental explanans in a theory of linguistic meaning.12 The authors whose work we will look at in this section do indeed give special status to assertions in the theories of meanings that they propose. Assertions differ from propositions in that the former are acts, while the latter – whatever they are – are certainly not.13 The emphasis that Brandom and Barker place on acts in their analyses of meaning is reflected in their conception of truth. Whereas truth-conditionalists see truth as a property that propositions either do or do not have, Brandom and Barker see truth primarily in terms of judgement. To understand truth, they argue, we should not ask which objects can have the property of being true, but rather what class of phenomena can aptly be judged true. Judging, like asserting, is an act; and judging an assertion as true expresses an attitude of acceptance towards that assertion. Barker and Brandom, therefore, are expressivists about truth, for they see the attitude of acceptance that judgments of truth express, and the consequences of
Assertion in Speech-act Theory 19
such expressions, as key to its understanding. We now look at each of their frameworks in turn. 2.3.1 Brandom: assertion as discursive commitment Like Searle, Brandom sees assertion as essentially an undertaking of commitment. For Brandom, though, ascertaining the precise nature of this commitment is crucial, not just in order to understand assertion, but as the basis for a theory of linguistic meaning. Brandom’s view of assertion has its roots in Dummett’s characterisation of competence in this speech act as being a matter of grasping both of the conditions under which it is correct to make an assertion, and the consequences of doing so: Learning to use a statement of a given form involves [. . .] learning two things: the conditions under which one is justified in making the statement; and what constitutes acceptance of it, i.e. the consequences of accepting it. Here ‘consequences’ must be taken to include both the inferential powers of the statement and anything that counts as acting on the truth of the statement. (Dummett 1981: 453). The conditions under which you are justified in making an assertion are those which make the assertion true. The consequences are both practical and inferential. For example, if you assert that the beer is in the fridge, then you are committed to going to get the beer from the fridge if properly ordered to fetch it. This is a practical consequence of your assertion. If you assert that the dress the actress wore was red (all over), then you are committed also to the claim that the dress was not white. This is an inferential commitment. The commitments that any one speaker undertakes depend upon her prior commitments. Thus if she asserts both that the dress the actress wore was red and that red is the colour of blood, then she is committed to the claim that the actress wore a dress that was the colour of blood. Dummett distinguishes two types of assertion: those that have verifiable truth-conditions and those which do not. The latter group includes ethical statements and ascriptions of responsibility. The category that comprises of both type of assertion he calls ‘quasi assertion’. What unites utterances in this group is the following: all can be used to deceive, all employ declarative syntax, all are open to demands for justification, and all constrain the speaker’s subsequent verbal and/or non-verbal action (Dummett 1981: 356–7). The last two of these features are the consequences of assertion, and it is these that are at the heart of Brandom’s account.
20
Assertion
Brandom divides Dummett’s consequences into two types: commitments and entitlements (1994: 168). In the example just used, the claim that the dress was red prohibits you from asserting (i.e. commits you to not asserting) that it was white (or green, or blue . . .), but it entitles you to assert that the dress was scarlet, or vermillion. As we will see below, it is through these notions of commitments and entitlements that Brandom seeks to explain semantic relationships. In making an assertion, on Brandom’s account, speakers both assume authority and undertake a responsibility: the authority they assume is the right to pass on the entitlements and commitments of their assertion to their listeners; the responsibility is to justify their assertions if challenged.14 Assertions can be justified in three ways: by deferring to a previous asserter (who assumed the authority to pass on the commitment undertaken by the current asserter), by reporting reliable perceptual experience, or by giving reasons in the form of further assertions (1994: 172–4). The last of these is most pertinent to present concerns. In relation to reasons and reason-giving, Brandom draws an important distinction between perception, the act of assertion in particular and action in general. Actions, in general, can require the giving of reasons (‘Why did you shoot him?’) and perception can be employed in reason giving (‘I shot him because I saw him reach for his gun.’), but only assertions are capable both of being a reason and of justifying calls for reasons. You might assert that rain is likely to give a reason for carrying an umbrella, but that assertion also justifies your hearer in requiring that you give a reason for it. This reason might be a further assertion (‘It often rains in London in April’), which would again have the role of both standing as a reason and creating the responsibility to give additional reasons. Having the dual characteristics of both being able to serve as reason and ‘reason-requirer’ is, for Brandom, sufficient condition for being an assertion (1994: 167).15 The notion of giving and requiring reasons makes sense only in a social context. In giving a reason, on Brandom’s account, a speaker takes on the responsibility of giving further reason, if so required, and, in doing so, she implicitly endorses inferential relationships between the reasons she gives. For an audience to accept her reasons is to grant that she is entitled to them, to take on those entitlements themselves and thereby endorse that inference. Brandom sees assertions, then, as speech acts that can serve as premises and conclusions, and accepting an assertion that P as a reason for an assertion that Q is to accept the inference from P to Q. Thus assertions, for Brandom, are fundamentally ‘fodder
Assertion in Speech-act Theory 21
for inferences’ (1994: 168), and a key function of assertion is to make explicit inferential relations implicit in the commitment undertaken by interlocutors.16 Brandom explains force in terms of the different commitments and entitlements undertaken by interlocutors. In this respect, his account is similar to Alston’s, discussed above. Thus promises commit speakers to act; orders commit hearers to act, and so on.17 The role played by the common propositional core in traditional speech-act theories is here played by assertions. For Brandom, other illocutionary-act types make sense only in relation to a prior notion of assertion. This is because promises and orders, for example, rely for their explication on the notion of a described state of affairs being made true, and Brandom, as we will see below, views truth primarily in terms of judgements of assertions. Assertion, for Brandom, ‘is the fundamental speech act, defining the specific difference between linguistic practice and social practices more generally’ (1994: 173). The traditional semantic relationships of entailment, compatibility and contradiction are explained by Brandom in terms of compatibility and incompatibility of commitments and entitlements. P entails Q iff asserting that P prohibits one from asserting not-Q. P is compatible with Q iff asserting that P entitles one to assert that Q. P contradicts Q iff asserting that P removes entitlement to assert that Q (1994: 160). The propositional content of an assertion, then, consists in the commitments and entitlements that are concomitant with that act of assertion, and bestowed upon those who accept it. Propositional content is thus explained in terms of commitments and entitlements rather than in terms of truth conditions. One can say, according to Brandom, of anything which has propositional content that it has truth conditions, but to do so is just to ‘pay a harmless compliment’ to things whose contents can be expressed in declarative sentences or by the corresponding ‘that clauses’ (1994: 329). Thus Brandom allows the possibility of expressing semantic content in truth-conditional terms, but denies the possibility of explaining semantic content truthconditionally (loc. cit.). In this respect, Brandom differs from Dummett, who places greater emphasis on truth conditions. As we saw above, Dummett has it that assertions proper are those whose truth can be verified. Indeed, he (1993) sees truth-conditional accounts of meaning such as Davidson’s as potential components of a theory of meaning in that they provide, according to him, a theory of sense. One finds little mention, however, of the sense/force distinction in Brandom’s work.18
22
Assertion
For Brandom, truth is to be explained in terms of the commitments and entitlements that result from truth-judgements. Judging an assertion true entails taking on commitments and entitlements, and an understanding of this is all there is to an understanding of the notion of truth, according to Brandom. Hence he is a deflationist about truth, in that he thinks the notion can be explained in terms of prior notions of judgement, commitment and entitlement (1994: ch. 5).19 This is not to say, though, that he is a relativist about truth: there is, for Brandom, an objective fact of the matter. Essentially, his view is that our ability to challenge the commitments undertaken by others is restricted by how the world is. These commitments are the result of a combination of perception, inference and action, and two of these – perception and action – are cases of non-linguistically mediated interaction with the world. Consequently, ‘empirical and practical constraints get built into what commitments (including inferential commitments) one can sustain entitlement to’ (1994: 333). Before concluding this subsection, it is necessary to consider Brandom’s aim in developing his theory, so that it can be compared with the aim of the present project. The aim of Chapter 8 is to present an account of assertion from the perspective of a cognitive theory of utterance interpretation. Brandom’s aim is quite different. He wants an explanation of what it is to be a language user such that one could recognise this ability however it were manifested, a theory, that is, that applies to human verbal communication but would also enable us to say whether a newly discovered extraterrestrial gaseous life-form had language. As he himself puts it: The primary explanatory target is what it is to grasp a propositional content and to have and to attribute to others states and performances with such contents – in the sense of explaining what doing the trick consists in, what would count as doing it, rather than show how it is done by creatures wired up as we are (1994: 155). The aim of the project towards which the ideas developed in Chapter 8 seek to contribute is precisely to explain how linguistic communication is achieved by ‘creatures wired up like we are’. This project is complementary to – rather than in competition with – the type undertaken by Brandom. The standards by which they will be judged differ greatly. Brandom’s is a philosophical account and will be assessed in terms of its coherence, elegance and ontological economy. The cognitive account put forward in Chapter 8 will ultimately be judged in terms of the
Assertion in Speech-act Theory 23
predictions it makes and whether these are novel, interesting and confirmed. Elegance and ontological economy are considerations, but are not primary: it might turn out that we creatures are wired up pretty messily. However, to offer a cognitive account of assertion is to suppose that assertions can be distinguished from other acts and are there for the explaining. Brandom’s definition of assertion as an act which can both serve as a reason and require the giving of reasons is both original and useful in this respect as it makes no reference to the notion of belief. As we will see in the next chapter, introducing belief into a philosophical analysis of assertion is a problematic move as it assumes that belief can be explained independently of assertion, and – as will also be seen – it is far from obvious that this is the case.
2.3.2 Barker: a speech-act theoretic approach to meaning Like Brandom, Barker (2004) seeks to ground an account of linguistic meaning in speech acts. Unlike Brandom, however, he seeks to do so in overtly psychologistic terms. Psychology enters his story in two ways: on the one hand, he analyses linguistic forms as indicators of characteristic associated intentions, and the interpretation of speaker meaning as relying on the hearer deciding whether the speaker actually has the intentions indicated (2004: 7, 45–62); on the other, the expressive powers of natural language are argued by Barker to be grounded in a more rudimentary representational system of cognitive ‘tracking states’ (2004: 19–20, ch. 7). Barker develops a view on which language derives its meaning not from a mapping relationship with the world – as would be the case on a standard truth-conditional account expressed in model-theoretic terms – but from a linkage between linguistic forms and intentions. Linguistic types are associated by Barker with act types. Declarative sentences, for example, are associated with the act of asserting, while NPs are indicators of an intention to denote (where ‘denote’ is to be understood as ‘to have stand for’ (2004: 5)). According to Barker, by using a given linguistic form, a speaker advertises (i.e. presents herself as having) the intention associated with that form. It does not follow, though, that she in fact has that intention, nor that she intends the hearer to believe that she has it. Whether she has the intentions she advertises is something the hearer must either infer, or determine from the linguistic environment of the intention-indicator: if a declarative sentence, for example, is embedded in the antecedent of a conditional, this
24
Assertion
indicates that the speaker does not have the assertoric intention advertised. Linguistic forms are thus ‘proto-acts’, in Barker’s terminology: they are characteristic of speakers intending to perform certain acts, but their production does not, on its own, constitute performance of those acts (2004: 45–51). Barker rejects the sense/force distinction that is assumed by traditional speech-act theorists. In this, and in his psychologism, Barker is resolutely anti-Fregean in his approach. Indeed, his stated aim is to overthrow Frege’s semantic paradigm (2004: vii). For Barker, assertions – not propositions – are the primary bearers of truth. Assertions, Barker claims, are associated with intentions to express commitment states and have these states accepted, confirmed or rejected by hearers. For a hearer to judge an assertion true is for him to express the same commitment state in his own case.20 Barker analyses assertion as follows. In uttering a declarative sentence, the speaker advertises the intention to assert. This is itself formed of two intentions, one representational and one communicative. The representational intention is to represent a state of affairs;21 the communicative intention is an intention to defend a commitment. A speaker defends a commitment, in Barker’s terms, if the hearer believes that the speaker has that commitment, and either accepts, confirms or rejects that commitment in his own case (2004: 8). But to what exactly does the speaker defend a commitment? The answer Barker gives to this question depends on a distinction he draws between reportive and expressive assertions. A reportive assertion is, in essence, characterised by the speaker’s intention to defend a commitment that the state of affairs described by her utterance obtains, that it is an accurate reflection of how the world is. However, because, according to Barker, logical complexity is not a feature of the world (2007: 4–5), logically complex assertions, such as those containing negation and modal operators, cannot be analysed as involving intentions to represent how the world is. In these cases, Barker claims, speakers do not intend to defend a commitment to a state of affairs obtaining, but to being in a particular commitment state, one that it is correct to be in for a creature whose cognitive system is aimed at accurately representing the world. Thus, in asserting the logically complex (26), the speaker has the following intentions: (26) There is probably life on Mars
Assertion in Speech-act Theory 25
Representative intention: to represent a state of affairs in which she believes that there is life on Mars Communicative intention: (a) that her hearer believe that the speaker is in this state and (b) that he accept, confirm or rejects being in that state himself. (27) I believe that there is life on Mars For Barker, two assertions may have the same representational element but be judged true under different circumstances. For example, on his analysis, assertions of both (26) and (27) represent the world as being such that the speaker believes that there is life on Mars. However, because the communicative intention is different in each case, they will be judged true under different circumstances: (27) is a reportive assertion, and will be judged true just in case the hearer agrees that the speaker is in the state described; (26), by contrast, will be judged true if the hearer accepts that the state that the speaker represents herself as being in is a correct one to be in, thereby undertaking commitment to that state in his own case. (2004: 8–13) Thus truth-judgments are not, for Barker, judgements of correspondence, but are to be understood in terms of the attitudes they express, as in Brandom’s framework. To judge an assertion as true is to accept it. What one is accepting, on Barker’s view, depends on the speaker’s communicative intention: in the reportive case, one is accepting that the world is as described by the speaker; in the expressive case, one is agreeing that the state that the speaker is in a correct state to be in, given how the world is. This is not to say that Barker posits two forms of acceptance, though. Claiming that the world is a particular way by performing a reportive assertion is, in effect, on his account, to present oneself as being in a particular commitment state. What distinguishes the commitment states associated with reportive assertion from those associated with expressive assertions is that, in the former case, the state that the speaker is in maps directly to the world, while in the latter it does not. As noted above, Barker grounds his theory of meaning in a notion of basic cognitive states that he calls ‘tracking states’. These are states that represent – or reflect – how the world is (or, in the case of intentional tracking states, how the holder wants the world to be). For expository purposes (though Barker does not speak of them in these terms), we can think of them as logically simple sentences of mentalese with empirical content (i.e.
26
Assertion
expressing thoughts like ‘Rain is wet’ rather than ‘Justice is good’).22 They are not beliefs, because beliefs can be logically complex (one can believe that it is not raining, for example), but are a more basic species of thought. Logically complex beliefs are available only to linguistic animals, on Barker’s story. The basic ‘tracking states’ that we are thinking of as simple sentences of mentalese, by contrast, are common to humans and some non-linguistic animals (2004: 190). So, when performing a reportive assertion, the speaker in effect expresses a cognitive state of the basic tracking kind. In accepting this, the hearer is in only a derivative sense accepting that it accurately represents the world. At a more basic level, she is accepting that the state expressed is an appropriate one to be in given how the world is, and hence the acceptance is of the same type that one finds, according to Barker, with expressive assertions. Those, such as Barker, who deny propositions an explanatory role in semantics must find another means of explaining what is common to a group of sentences such as (28) to (30), such that (30) might be uttered to ask whether (28) had been complied with, and (29) uttered to assert that it had: (28) Peter, stand up! (29) Peter is standing up (30) Is Peter standing up?’ Barker explains the meaning common to these three sentences in terms of assertion. An order given using an imperative sentence is analysed as a case of a speaker advertising intentions to: a) represent a state of affairs in which she desires that the hearer sees to it that a certain assertion is correct (in the sense that it is correctly asserted by anyone with perfect information); b) have the hearer believe she has this desire. If the speaker has the advertised intention b), then her utterance is an order (2004: 82) Thus in uttering (28), the speaker is, according to Barker, advertising an intention to represent a desire she has that Peter make the world such that (29) can be correctly asserted. Note that in an order, on Barker’s account, there is no intention advertised to defend commitment to the state represented. This is what distinguishes (31) from (28): although both represent, on his account, the same state of affairs, only (31) advertises the intention that the hearer should accept or reject that representation. In his words, assertions are democratic; orders are not (2004: 44).23
Assertion in Speech-act Theory 27
(31) I want you to make it the case that Peter stands up In the case of questions asked using interrogative form, the intentions advertised are as follows: a) to represent a state of affairs in which the speaker desires that the hearer affirm or deny a certain assertion; b) to have the hearer believe that the speaker has this desire. Thus in uttering (30) the speaker advertises an intention to represent a desire she has that the hearer affirm or deny (29). If she has the intention advertised, then she asks a question. The assertion that the speaker wishes to see made correct, in the case of orders, or affirmed or denied, in the case of questions, is identified by the hearer as the assertion that would be performed by the use of a proto-assertion constructed from the lexical elements of the imperative or interrogative sentence. Thus it is not necessary, according to Barker, to assume that there is a common propositional core that (28) to (30) share in order to explain the sense in which they are about the same state of affairs. Assertion is basic, and orders and questions express desires about assertions. Barker is adamant that the assertion (29) is not a component of either (28) or (30): it cannot be, he says, because the imperative and interrogative mood of the sentences uttered means that the speaker ‘does not engage in the behaviour characteristic of someone with an assertive communicative purpose’ (2004: 83). Such behaviour is, for Barker, the uttering of a declarative sentence. Thus, Barker insists, although assertions are required in order to explain the other main illocutionary acts, they do not fulfil this role in the same way that propositions do in traditional speech-act accounts: they are not constituents of these acts. Barker uses the bipartite nature of his analysis of assertion (the distinction between representative and communicative intentions) to distinguish between those propositions that are asserted by a speaker and those to which she is committed, but does not assert. One example he gives are so-called ‘conventional implicatures’ (Grice 1989: 25–6) such as the implication (32) arising from the use of ‘therefore’ in Grice’s famous example (33): (32) Being English entails being brave (33) He’s English. Therefore, he’s brave Barker’s analysis of this case would go as follows: the speaker’s use of ‘therefore’ in (33) advertises her intention to represent herself as being
28
Assertion
in the state of believing (32) and expresses her intention that the hearer should appeal to his own possession of this state in processing (33). Thus, for Barker, putative conventional implicatures such as those associated with ‘therefore’ and ‘even’ are unlike assertions both in that they have common-ground status and in that, although they advertise the speaker as being in a particular state, they are not associated with any intention to defend that state. In other words, while they share with assertions a representational intention, they lack a communicative intention (2004: 42–4). While both Barker and Brandom are fundamentalists about speech acts, there are important differences between them. At the root of these lie their attitudes to psychologism. As was noted in the previous subsection, Brandom has no interest in explaining the cognitive underpinnings of human linguistic communication: how creatures ‘wired up like us’ come to be competent asserters. Barker, however, does not appear to recognise Brandom’s distinction between theory of assertion and a theory of the human cognitive capacity to engage in that practice. Rather, he makes direct appeal to human ‘wiring’ in his explanations of assertion. Their respective positions on psychologism are reflected in how they ground their theories. Because Brandom seeks to explain semantic relations such as entailment in terms of commitment, the basic requirement, on his account, for becoming an asserter is to be a member of a community, for commitment can only be made sense of in terms of one person having a commitment to another. Brandom’s is therefore a social theory of meaning, one grounded in the idea of a community of language users. Barker, by contrast, grounds his theory in intentions, but to have an intention is, at very least, to have a representation of how one wants the world to be, which in turn entails having a representation of how the world currently is. An intentional account therefore presupposes a level of representation at which intentions are formed. The most basic level of representation, for Barker, is the level of logically simple tracking states discussed above. So, grounding his analysis of assertion in intentions means that Barker must appeal to cognitive science to explain the representational states and computational processes that intention-having presupposes (2004: 74–5). However, as we will now see, this reliance threatens to undermine his anti-Fregean stance. Barker’s problem is that he is ultimately reliant on the very force/sense distinction that he seeks to reject, even though this now operates, for him, at the level of thought. This is because the level of tracking
Assertion in Speech-act Theory 29
states that Barker posits as the medium in which basic intentions are formed contains both doxastic and intentional tracking states (2004: 193). These might be distinguished by saying that the former, as they are correct if they match the world, are truth-apt, while the latter are not. This move is not open to Barker, though, as, for him, to be truthapt is to be associated with an intention to defend being in the state represented: what make assertions truth-apt, on Barker’s account, is that they come with an intention to defend a commitment state. Doxastic tracking properties, though, are the medium in which the commitment states represented by assertions are grounded, so they cannot be seen as intentions to defend those states: they are the objects of defence, not bearers of the defensive stance. Consequently, an alternative criterion to truth-aptness is required in order to distinguish doxastic from intentional tracking states. Barker’s solution is direction of fit: doxastic states are said to have word-to-world direction of fit, intentional states world-to-word (2004: 193). But, as was noted above, direction of fit is a metaphor that needs to be cashed out. Barker’s cashes it out (not in Renewing Meaning, but in a later publication) by saying that to have word-to-world direction of fit is to be a state or act which has a defensive component, and that to have a defensive component is to be truth-apt (Barker 2007: 200). But Barker is adamant that tracking states are not truth-apt, precisely because they lack such a defensive component. It follows, therefore, that they do not – on his definition of the notion – have direction of fit. Without direction of fit, however, Barker has no way of distinguishing doxastic from intentional tracking states, even though such a distinction is required by his theory. Barker’s problems in this respect stem from his desire to undermine the Fregean concept of semantics – especially the sense/force distinction – while at the same time appealing to cognitive science. Traditional cognitive science seeks to explain thought and behaviour in terms of computations over representations. This requires that there be at least two types of representation available to any cognitive system: one that represents how the world is and one that represents goals. Call the first type B-Reps and the second type G-Reps. Goal achievement is recorded by the system when its behaviour results in it tokening a B-Rep that matches a G-Rep. How does it match? It matches when the G-Rep and the B-Rep have the same content, or sense. Thus at the cognitive level, no matter how sub-personal you go, a means of distinguishing goal representations from state-of the world representations is needed, and so is a means of matching them up. This is precisely what the force/sense distinction does. Cognitive science is inherently
30
Assertion
Fregean, and it cannot be appealed to in order to bring down the Fregean project.24 In more recent work, Barker (2007) no longer relies on appeals to traditional cognitive science. To outline these developments would take us too far afield here, but the lesson from the previous discussion is important for the present project. The proposals developed in Chapter 8 are couched in a cognitive theory of communication that assumes that thought can be usefully modelled in terms of computations over representations (essentially the position defended by Fodor 1981; 1998; 2008). We can now see that any such project cannot do without the sense/force distinction, and hence a position will have to be taken on the nature of propositions. This is given below in section 3.
2.4 Conclusion Of the topics that have been discussed in this chapter, three require special attention. First, there is the issue of propositionhood. Is there a core meaning over which illocutionary force operates, as the traditionalists have it, or is assertion basic, as the fundamentalists insist? The position taken in this book will be that assertion is fundamental, primarily for the reasons given in section 3.3: in a truth-based theory of meaning, the ultimate explanans must be truth bearers; propositionhood alone is insufficient for truth aptness; consequently, propositions are not truth-bearers. However, it is convenient to be able to talk of assertions expressing propositions, and of an assertion sharing a propositional core with other illocutionary acts, such as an order or question. Fortunately, we can continue to make use of this idiom by making the following stipulations: 1. To say that an assertion A expresses a proposition P is to say that there is a set of possible worlds in which A would be correctly judged true. We can then, inspired by Stalnaker (1978/1999) identify, P with that set of worlds. 2. To say that A, an order O and a question Q express the same proposition P is to say that, in order to identify the set of worlds in which O is complied with or Q correctly affirmed, we must derive a declarative sentence from the lexical elements of O or Q (respecting contextual parameters). An assertion of this declarative sentence will pick out the set of worlds P in which A is true, O is complied with or Q correctly affirmed.25
Assertion in Speech-act Theory 31
On this view, then, each non-assertive sentential utterance is related, by virtue of its lexical elements, to a derivable assertive utterance. The satisfaction (e.g. compliance) conditions of the non-assertive utterance are the truth conditions of the derived assertive. Talk of propositions is a convenient abstraction from this relationship. The second point relates to giving due credit to the importance of the consequences of assertion to its analysis. This point, made initially by Dummett and later developed by Brandom, will be shown in Chapter 4 to shed light on the nature of the relationship between assertoric force and declarative mood. In particular, it will be shown how viewing assertion in these terms allows one to hold the view that the declarative is an indicator of assertoric force, while also making available straightforward analyses of non-assertoric uses of this form. The third point is related to the second. The expressivist approach to truth advocated by the fundamentalists offers an alternative perspective on truth judgements in linguistic theorising. On this view, to judge an assertion as true is to accept the consequences of that assertion. Acceptance is only an issue, however, when rejection is also possible. In cases where rejection is not possible, acceptance, and hence truth judgements, should no longer be an issue. In Chapter 4, it will be argued that in cases where witnessing a non-assertoric utterance of a declarative sentence that expresses P results in the same commitments that follow from accepting an assertion that P, then rejection is not an issue and truth judgements do not apply. This analysis is applied to promises, certain commands, and explicit performatives. So far, then, most of the implications for the remainder of this book come from the fundamentalists. What of the traditionalists? The traditionalists’ emphasis on assertion as an expression of belief certainly needs attention. Indeed, so much so that this is the topic of the next chapter.
3 Assertion, Belief and Knowledge
3.1 Introduction Perhaps the most natural way to think about the relationship between assertion and belief is to see former as the expression – or exteriorisation – of the latter. This is certainly how Frege saw it. For him, belief is the result of judging a thought true, and assertion is the manifestation of that judgement (1918–19/1997: 329). Viewed this way, assertion can only be analysed given a prior notion of belief. This order of priority is found throughout linguistics and philosophy of language. Bach and Harnish (1979), as we saw in Chapter 2, analyse assertion as involving the expression of belief and the intention to cause belief formation in others. In this, they take their cue from Grice, who saw the speaker’s intention that ‘the hearer should think that the speaker believes something’ (1989: 123, original emphasis) as key to the correct analysis of assertion. Furthermore, it is now widely agreed in linguistic pragmatics that the linguistically encoded meaning of an assertion dramatically underdetermines the proposition to which the speaker undertakes a commitment (Sperber and Wilson 1986/1995; Bach 1994; Carston 2002; Recanati 2004). On this view, assertions derive their full meaning from the beliefs they express, and hence this ‘underdeterminacy thesis’, as it is commonly called, appears to require that belief must be seen as more fundamental than assertion. It might be thought, then, that there is universal consensus on this issue. However, there are those, such as Dummett, who dissent, arguing that placing belief before assertion gets the order of explanation wrong: it is belief, they say, that should be analysed in terms of assertion, not vice versa. We will see in section 3.2 that Dummett’s arguments to this effect are strong, but we will also see that whether, and in what 32
Assertion, Belief and Knowledge 33
sense, assertion is correctly said to be prior to belief depends very much on what project one is engaged in: the philosophical undertaking of explaining what the practice of assertion and the notion of belief consist in, or the scientific task of explaining the cognitive mechanisms underlying the human capacity for assertion. In section 3.3, we will take as given that, from the point of view of the cognitive project to which the proposals developed in Chapter 8 seek to contribute, belief is prior to assertion and that assertions are expressions of belief. The question to be considered in that section is whether we should then see the interpretation of assertions as necessarily involving belief attribution on the part of the hearer. It is not obvious that this is the case. It is possible that assertions are caused by and express beliefs, but that belief attribution is not required for their interpretation. It is certainly the case that some forms of animal communication have these characteristics. For example, the vervet monkey’s alarm call (Seyfarth et al. 1980) is caused by the animal’s belief that a snake is present and can hence be seen as an expression of that belief, but it does not follow that interpreting the call requires that the monkey’s conspecifics attribute to it these beliefs.1 As we will see, a number of authors have argued that all that is needed for the practice of assertion to function in a large proportion of cases is that hearers take asserters to be sources of information – or knowledge – rather than expressers of belief, still others have argued that assertion should be analysed primarily in terms of its relationship to knowledge, rather than to belief. This knowledge account of assertion is discussed in section 3.4.
3.2 Assertion and belief: what is the correct order of explanation? This section is devoted to a discussion of Dummett’s view that belief should be explained in terms of assertion, and not vice versa. The consequences of this claim for cognitive science are discussed, as is some evidence that, from an empirical point of view, investigating Dummett’s claims might prove fruitful in terms of an understanding of how children develop the notion of representation that underlies their eventual grasp of belief. The argument put forward by Dummett (1981; 1993) for the priority of assertion over belief can be summarised as follows: 1. The notion of belief is reliant on a notion of truth. 2. Truth must be explained, in part, in terms of treating as true.
34
Assertion
3. Treating as true is a practice. 4. Given 2 and 3, it follows that truth cannot be explained without reference to a practice that exemplifies treating as true. 5. Assertion is the practice that best exemplifies treating as true. 6. Given 4 and 5, it follows that assertion cannot be explained in terms of belief, for the notion of belief is reliant on the notion of truth, which must itself be explained in terms of the practice that best exemplifies treating as true: assertion. Let’s look at each of these steps. Step 1 claims that belief cannot be explicated without recourse to the notion of truth. This is uncontroversial: belief is standardly characterised as an attitude to a proposition, that attitude being one of holding that its propositional object is true. Though Dummett (1993: 217) speaks of belief as holding a sentence – rather than a proposition – to be true, nothing hinges on this for present purposes. In discussing his argument, too will use the term ‘sentence’ where I would normally be more inclined to say ‘proposition’. Step 2 is to claim that truth must be explained, in part, in terms of treating as true (Dummett 1993: 221). This in opposition to the view that the notion of truth can be given an analysis solely in terms of the conditions that make a sentence, statement, belief or proposition (choose your favourite fundamental truth-bearer) true. Dummett does not deny, however, that the notion of truth-conditions has a role to play in explicating truth. His claim is, rather, that merely being able to divide a set of sentences into those whose truth conditions obtain, and those whose do not, shows only a partial grasp of the notion of truth, for it fails to demonstrate an appreciation of the point of truth judgements. To show that one has a grasp of the point of such judgements, Dummett argues, one must demonstrate a grasp of the consequences, both practical and inferential, of judging a sentence true (Dummett 1981: 453). It follows from this that a complete account of truth requires an explanation of what it is to treat as true a sentence so judged. To digress slightly, it is this claim that lies at the root of Dummett’s disagreement with Davidson over the adequacy of a disquotational theory of meaning (Dummett 1989/1993; 1993). To say, as Davidson (1967/2001) does, that the meaning of a sentence such as (1) can be captured by the T-sentence (2) is, according to Dummett, only partly correct. (1) Snow is white. (2) ‘Snow is white’ is true iff snow is white. (3) ‘Snow is white’ is labelled ‘true’ iff snow is white.
Assertion, Belief and Knowledge 35
What a disquotational theory of meaning captures is the competent speaker’s ability to divide sentences into those which are correctly called ‘false’ and those which are correctly called ‘true’. But competent language users know more than this, Dummett argues: they know that judging a sentence true carries with it certain consequences, and it is only when one grasps this, he claims, that one can be said to be a fully competent language user. Dummett’s point here can be illustrated by saying that, according to him, the T-sentence (2) should be rewritten with ‘true’ in quotation, as in (3), for what a T-sentence captures is the ability to label a sentence correctly. It says nothing about the language user’s grasp of the point of such labelling, but it is precisely this, for Dummett, that is the key to understanding the significance of the label applied. In sum, the ability to decide correctly whether or not a sentence is true is insufficient to indicate a grasp of truth, according to Dummett. To show this, he argues, one must show that one understands what it is to treat a sentence as true. This requires competence in a practice that exemplifies treating as true. The preceding discussion has brought us to step 4 of Dummett’s argument for the priority of belief over assertion. We must now ask which practice best exemplifies treating as true. As we have just seen, to treat a sentence as true is, on Dummett’s account, to grasp its practical and inferential consequences. What behaviour would count as evidence of grasping these? While other forms of behaviour might demonstrate a grasp of some of the practical consequences of treating a sentence as true, many, perhaps most, of its inferential consequences can only be demonstrated by asserting (or assenting to assertions of) those consequences and denying those statements which are not consequences. It is fair to say, then, that treating as true is best exemplified by competence in the practice of assertion.2 This is step 5 of the argument. From this and step 4, step 6 follows: assertion cannot be explained in terms of belief, for assertion is required to explain the notion of truth, which is in turn needed to explain belief. This is a strong argument for the priority of assertion over belief, one that I am inclined to accept. What is particularly attractive about it is that the tight relationship it posits between notions of truth and assertion affords an understanding of falsity that alternative accounts do not provide. For example, the notion of truth employed in model-theoretic accounts of meaning reduces falsity to a failure to map to the model. But this does not distinguish acts which are not truth apt, such as directives performed by uttering imperatives, from false assertions. Both fail to map to a model of the world, but only the latter are correctly described as false. Dummett’s account suggests the following explanation: to be
36
Assertion
aptly described as false, a propositional form must be the sort of thing that can be treated as true. To treat a propositional form as true, though, is not merely to tick a box indicating that it maps to a model: it is to express a judgement that has consequences, these consequences being dependent both on what other propositions the individual has judged true, and on her intentions and desires. To accept a false assertion, on this view, is to undertake commitment to the consequences of that assertion even though the appropriate conditions for making it do not obtain. The absence of the conditions for assertion is part of what falsity amounts to, but it is not the whole story. However, despite my acceptance of Dummett’s argument for the priority of assertion over belief, I am intending to put forward in Chapter 8 a cognitive account of assertion which sees assertions as deriving their meaning from the beliefs they communicate. Such a view gives explanatory priority to beliefs over assertions. It is possible, however, to maintain the position that, from a philosophical point of view, assertion is prior to belief, while in the terms of cognitive science, belief is fundamental to assertion in an important sense. The order of explanation with which Dummett is concerned is an issue of conceptual priority. On this view, if we want to give an analysis of the notion of belief, we cannot do so without reference to assertion. It does not follow from this, though, that the empirical order of explanation will be the same. It may well be that, armed with the concepts of belief and assertion, we can devise an accurate model of communication according to which assertions derive their meaning from beliefs that they express. Thus, as long as philosophical questions of conceptual priority are kept apart from empirical questions about the cognitive processes involved in utterance interpretation, there is no conflict between the view that the notion of assertion is conceptually prior to the notion of belief, and the view that assertoric utterances derive their content from the beliefs they express. This is not to say that issues of conceptual priority necessarily have no bearing on empirical matters. Indeed, the assertion-first view proposed by Dummett does suggest the empirical hypothesis that competence in assertion is the means by which one comes to understand the notions of truth and falsity that underlie a full-fledged conception of belief. On this reading, the assertion-first view predicts that the grasp of truth and falsity required by the notion of belief will only be available to competent asserters. In other words, pre-linguistic humans (and non-linguistic animals) should not be able to demonstrate a grasp of the concept of belief. We will briefly consider whether this prediction is borne out.
Assertion, Belief and Knowledge 37
To have a full-fledged conception of the notion of belief is, minimally, to understand that action is guided not by how the world is but by how it is represented as being, such that when there is a mismatch between how the world is and how it is represented, the agent will act in accordance with how he represents the world rather than how it actually is. Underlying this is the recognition that falsity is about more than a failure of a representation to map reality. As we have seen, to grasp falsity, one must comprehend the distinction between the conditions for correctly judging true and the consequences of doing so, and appreciate that the consequences of a truth judgement follow even if the conditions that would make such a judgement correct do not obtain. The acid test for a grasp of the causal powers of false belief is the ability to pass one of the battery of false-belief tests that have been developed by experimental psychologists. The most well known is probably the Sally-Ann test (Baron-Cohen et al. 1985), in which a child is asked to predict where an agent will look for an object about whose location the agent has a false belief. To pass the test the subject must indicate that the agent will look for the object where she believes it to be, rather than where the subject knows it to be. Applied empirically, Dummett’s assertion-first account of belief predicts that passing this test will be possible only for competent asserters, as it is through competence in this practice that the required understanding of falsity is said to develop. There is some evidence that this is the case. It is now widely accepted that a certain degree of linguistic competence is required to pass the standard false-belief tests (Astington and Jenkins 1999; de Villiers and de Villiers 2000; Ruffman et al. 2003). Moreover, there is also evidence that it is competence in the use of the syntax of ‘say’-complements (i.e. embedded declarative syntax) that is the single best linguistic predictor of success in these tasks (de Villiers and Pyers 2002; Hale and TagerFlusberg 2003; Lohmann and Tomasello 2003). ‘Say’-sentences are used to report assertions, and are found in discourse that aims to explain behaviour in terms of the commitments undertaken by those engaged in making and responding to assertions (e.g. ‘Johnny looked in the cupboard for the biscuits because that’s where Billy said that the biscuits were’). Competence in the production and reception of complex sentences with embedded declarative syntax is an indication that a child can identify assertions as a class of acts and that she is able to participate in discourse in which the commitments undertaken by asserters are discussed. Indeed, there is evidence that it is engagement in this discourse – rather than mere syntactic competence – that is the key
38
Assertion
to developing the ability to predict behaviour based on false belief (Slaughter et al. 2007). These speculations are expanded on in detail in Jary (2010), and to go into them now would take us too far afield.3 The crucial point for present purposes is that the philosophical view of assertion as prior to belief, as defended by Dummett, is not incompatible with an approach to the science of utterance interpretation that seeks to explain assertion in terms of the communication of belief, treating belief as prior. The first is a conceptual claim, the second an empirical one. The position taken is this book, then, will be that assertion is conceptually prior to belief: if we want to say what assertion is without reference to the cognitive capacities that allow us to engage in this practice, then we cannot do so by reference to belief expression. This is because, as Dummett argues, the concept of belief is reliant on the notion of truth; the notion of truth requires for its explication reference to the consequences of a truth judgement; and these consequences can be explained only by recourse to the practice of assertion. However, from the point of view of explaining how beings ‘wired up like us’ (to borrow Brandom’s characterisation of the cognitive-science enterprise) engage in assertion, the order is best reversed: assertion is most profitably analysed as the expression and communication of belief.4
3.3 Assertion and belief attribution The view that there is a correct order of explanation between belief and assertion grants that there is relationship between the two notions, and that there is such a relationship is not seriously disputed by anyone, as far as I know. It is generally accepted that a sincere assertion expresses a belief of the speaker’s and, when accepted, results in the adoption of that belief by the hearer – what debate there is concerns how kernel this is to the correct analysis of assertion. It is also widely accepted, though, that assertions are a source of information about how the world is beyond the internal life of the speaker. Much – perhaps most – of the knowledge we have of the world comes from testimony, i.e. what we have been told rather than from what we have observed.5 Where theorists differ is on the question of which of these two observations is the key to the correct analysis of assertion. For some, most notably Grice, assertion is to be analysed as being characterised by the intention that the hearer should come to think that the speaker believes that P, there often being an ulterior intention to the effect that the hearer should also come to believe that P himself (1989: 123). This view is accepted by Strawson (1971) and,
Assertion, Belief and Knowledge 39
as we saw in Chapter 2, by Bach and Harnish (1979), although the latter see the intention that the hearer believe that P as a primary, not ulterior, assertoric intention. Those who object to this view include Dummett (1989/1993), McDowell (1980/1998) and Millikan (2004: ch. 9; 2005: ch. 10). For these authors, assertions are primarily sources of information about how the world is, not how it is represented by the speaker. Assertions, on this view, are a form of perception by proxy, and only derivatively do they serve as indications of the speaker’s doxastic state. In this section, I will look at these views more closely before presenting a hybrid position. This sees assertion as primarily a phenomenon of information transfer that is underpinned by a special type of intention, the full recognition of which involves attributing the belief that P to the speaker. However, it will be claimed that this communicative intention does not necessarily play a direct role in the process of interpreting an assertion. Rather, it is there to be called upon when needed. The view of assertion as being first and foremost about the communication of speaker attitudes has its roots in Grice’s notion of non-natural meaning, of which utterance meaning is an instance. Gricean nonnatural meaning is characterised by its association with special type of intention: the intention to induce a certain belief in the audience via the recognition of that intention (1989: 218–19). On this view, identifying utterance-meaning necessarily involves attributing intentions, and the intentions that need to be ascribed concern propositional attitudes of the speaker. In the case of an assertion that P, the attitude that needs to be attributed to the speaker, according to Grice (1989: 123), is the intention that the hearer should think that the speaker believes that P. As just noted, Grice saw the intention that the speaker believe that P as a possible ulterior motive, one that is present in only a subset of assertions. Where present, though, this ulterior intention is parasitic, according to Grice, on the speaker’s prior intention to exhibit her belief that P. So, on the Gricean view, interpreting an assertion is a special type of belief attribution. For those who oppose this position, analysing assertions as cases of invited belief attribution gets things the wrong way round. Assertions, say critics of the Gricean view, are primarily a means of information transfer, and the fact that they can serve as indicators of the speaker’s belief state is a by-product of that function. The following quotation from Dummett is a forthright expression of this position: It is essential to our most basic acquisition of the practice of using a language that we learn to act on what others tell us; only as a more
40
Assertion
sophisticated by-product do we learn to use their utterances as revealing their beliefs, and even then, their intention is typically to inform us that what they say is so rather than to allow us a glimpse of their private convictions. Were it otherwise, it would be, as it were, an afterthought that we could, by making an unintended use of them, gain information from the assertions of others; and we should accept those assertions as true only on occasions on which we had particular reason to trust the speaker’s veracity, rather than, as now, whenever we have no particular reason to doubt it (1989/1993: 171, original emphasis). Further doubt is cast on the Gricean picture by the developmental evidence. The view that interpreting an assertion requires the attribution of a belief to the asserter assumes that belief attribution is within the capacities of all who respond appropriately to assertions. However, young children who cannot demonstrate (by passing one of the false-belief tests mentioned in section 3.2) a conception of belief appear to be able to make, and respond appropriately to, assertions (Breheny 2006). This suggests that the ability to attribute belief is not a prerequisite for at least novice-level competence in the practice of assertion. The alternative to the Gricean view of assertion we can dub the ‘perception-by-proxy view.’6 According to this picture, assertion is first and foremost a means of information transfer and a direct cause of belief formation, the beliefs formed having the same propositional content as the assertion. Dummett says that ‘we learn to react to the statements of others in the same way that we react to various observed features in the environment’ (1981: 355); while Millikan says: ‘Forming a belief about where Johnny is on the basis of being told where he is, is just as direct a process (and just as indirect) as forming a belief about where Johnny is on the basis of seeing him there’ (2004: 125). In addition to not requiring that young children be competent belief attributors, the perception-by-proxy view has the advantage of being phenomenologically compelling: it doesn’t feel, in most cases, as if we go through the process of attributing beliefs in the interpretation of quotidian assertions. When a friend tells us of her weekend, it does not seem as if we treat what she says as primarily an expression of her beliefs about what she did: rather, we treat it as information about what she did. That said, though, the perception-by-proxy view is not without its problems, and it is to these that we now turn. The first problem is that the perception-by-proxy view does not recognise a distinction between understanding an assertion and accepting it.
Assertion, Belief and Knowledge 41
While it is certainly the case that we unreflectively accept many of the assertions directed at us, it is also undoubtedly true that there are many assertions directed at us which we do not accept; yet we still understand these. Comprehension and acceptance, then, are different things, and what a theory of assertion qua speech act should do (especially if it wants to respect the illocutionary/perlocutionary distinction), is to explain the former, not the latter. The perception-by-proxy view, at least in its simplest form, misses this point. The second problem is that utterance interpretation just does seem to need recognition of a special kind of intention. What distinguishes an assertion from a case of recital, for example, is speaker intention, and hearers need to be sensitive to this. But once this point is conceded, then the question of what type of intention the hearer must recognise once again raises its head. An off-the-shelf answer is that an intention broadly of the Gricean type is needed, and this requires attributing an intention to cause the hearer to form a belief about the speaker’s belief. Belief attribution is then back in the picture and the comprehension/acceptance distinction comes for free: comprehension just is, on this story, attributing the correct intention about a belief, but this does not entail acceptance of the belief on the hearer’s part. If the perception-by-proxy view is to be saved, then, it must find a way of distinguishing communicative acts of language production from noncommunicative acts, and this can only be done by accepting that this requires that intention attribution plays a role in the interpretation of assertions. We might ask, though, whether the required intention attribution necessarily involves belief attribution: could it not be the case that the intention that is attributed is an intention to impart information? This would give us a perception-by-proxy account that is Gricean in the stress it places on the role of intention attribution, but without Grice’s insistence that the intentions in question concern belief attributions. Let’s spell this proposal out in more detail before considering objections. The idea is this: assertion comes with an overt intention to let the hearer know that P, where P is the proposition explicitly expressed by the utterance, (where ‘know’ is understood as having information rather than having a justified true belief). Correctly identifying the production of a declarative sentence as an assertion and not, say, a case of recital, requires attributing to the speaker the overt intention to let the hearer know that P. This solves the problem of the need to distinguish assertions from the non-illocutionary production of assertoric forms, but it does not solve the comprehension/acceptance problem: on this
42
Assertion
story, once the intention has been recognised, P has been accepted by the hearer. This is because ‘know’ is a factive verb, and assigning an intention let the hearer know that P is to accept thereby that P. Before we consider this problem, note first that the comprehension/acceptance distinction is only an issue for assertive speech acts: in the interpretation of, for example, orders, questions and promises, interpretation and acceptance of the thought that the speaker wishes to communicate coincide. Correctly interpreting these speech acts requires the hearer forming, respectively, beliefs of the following type: (4) The speaker wants me to make it the case that P (5) The speaker wants to know whether P (6) The speaker promises that P Unless the hearer forms belief of these types, as a result of interpreting the utterance, he has not comprehended the utterance. In these cases, then, comprehension is acceptance (cf. McDowell 1980/1998: 42). What distinguishes assertion is that acceptance involves more than the hearer correctly describing the goal of some observed behaviour. Acceptance of an assertion requires the hearer not only forming a belief of the form (7) but also of the form (8), when (in straightforward cases) he has no perceptual evidence that P beyond the assertion. What the non-assertive cases show, though, is that it is not a feature of linguistic communication in general that comprehension and acceptance must be kept apart. Rather, it is a feature specific to assertion. (7) The speaker as asserted that P (8) P But it is one thing to say that assertion allows for the possibility of acceptance without comprehension, and another to say that in, the process of interpreting an assertion, comprehension necessarily precedes and is distinct from acceptance. It is open to a holder of the perception-byproxy view to maintain that although assertion doesn’t necessarily result in acceptance, this is in some sense the normal or primary procedure, the bifurcation of interpretation and acceptance being a secondary phenomenon. The challenge is to cash out the ‘in some sense’ and say just how information transfer is primary, while also allowing for the comprehension/acceptance distinction to be maintained where necessary. In the remainder of this section, I will outline a view that meets these criteria, building on suggestions made by McDowell (1980/1998).
Assertion, Belief and Knowledge 43
McDowell argues that a knowledge transfer account of assertion can be seen as prior in both phylogenetic and ontogenetic terms. He suggests that protolinguistic communication in humans was of the instinctive sort typified by the vervet monkey’s alarm call, as mentioned above in section 3.1. This sort of communicative behaviour does not rely on intention recognition or belief attribution. Rather, ‘When the communicative process functions properly, sensory confrontation with a piece of communicative behaviour has the same impact on the state of a perceiver as sensory confrontation with the states of affairs that the behaviour, as we may say, represents; elements of the communicative repertoire serve as epistemic surrogates for the represented states of affairs’ (McDowell 1980/1998: 45). Communication of this type, then, has the same impact on the cognitive disposition of the receiver as direct perception of the state of affairs represented. It is plausible, he claims, that ‘the assertoric core of linguistic behaviour is a descendent, now under intentional control, of the sort of instinctive communicative repertoire [typified by vervet monkeys]’ (loc. cit).7 Such a story, he goes on to suggest, can be told at the ontogenetic level too: ‘The acquisition of linguistic competence by a contemporary human being may be expected to recapitulate, in some salient respect, the evolution of language from instinctive communicative behaviours’ (1980/1998: 47). What McDowell suggests, then, is that the child’s early competence in communciation involves treating this type of utterance as a source of information about the world, rather than a source of information about the speaker’s beliefs. The ability to see assertions as indicators of the speaker’s epistemic state is a later development, dependent on a grasp of the representational capacities of assertion: ‘It is not that their [i.e. assertions’] capacity for use as expressive of belief breathes into them the life that enables them to represent reality, but rather that their antecedent capacity to represent reality is what makes them capable of expressing beliefs’ (McDowell 1980/1998: 48). If it is accepted that, as far as pre-belief-attributing children go, utterance interpretation is a form of perception-by-proxy, in which interpretation and acceptance are one and the same process, the question that follows is this: what happens when these children do acquire the ability to attribute beliefs? Two possible answers present themselves. One is that the interpretation process undergoes a radical overhaul, such that children now react to assertions primarily as expressions of belief and only as sources of knowledge if the speaker is judged to be reliable, and what she says is plausible. The other is that interpretation proceeds
44
Assertion
as before, but that belief attribution can be brought in to assist when necessary. This could be for a number of reasons: it might be that the speaker is judged unreliable or untrustworthy; that what she asserts is implausible; that the significance of her utterance cannot be ascertained without taking into account her belief state; or that the nature of the interaction means that it is common ground that the speaker’s aim is to persuade, rather than to inform, as in academic discourse. Of the two possible answers, the second seems the most likely. It would be odd for a cognitive system to completely abandon procedures that have taken a significant amount of time and energy to develop. It is much more likely that the system would be augmented with new procedures, these being brought into play when things go awry. On this view, the hearer treats assertions as a direct source of knowledge that the state of affairs represented obtains unless he has reason not to. The idea, then, is that a complex intention of the type that Grice saw as constitutive of speaker meaning is attributable by the hearer to the speaker, but not necessarily made use of in every case of interpretation. Sperber and Wilson’s notion of manifestness is useful here. An assumption is manifest to an individual iff he is capable of representing it mentally and judging it as true or probably true (1986/1995: 39).8 Importantly, an assumption does not have to be mentally represented by an individual to be manifest to that individual: it must merely be capable of being made so. Equally important is that only assumptions that are mentally represented can play a causal role in the interpretation process. The fact that an assumption is made manifest to the hearer by a speech act does not therefore entail that it plays a causal role in the interpretation of that utterance. This notion of manifestness is useful in reconciling Gricean (in the sense that they involve belief attribution) and perception-by-proxy accounts of assertion, for it suggests a means by which a special type of intention can be characteristic of assertion without necessarily playing a role in the interpretation of every assertion. We can say that a complex intention about the belief states of the speaker and hearer, underlies assertion, and that acts of assertion make that intention manifest, without it following that a representation of that intention necessarily plays a role in the interpretation of every assertion. In other words, assertions necessarily make manifest, to a hearer capable of belief attribution, this sort of complex intention about beliefs, but this intention need not be taken into account in the interpretive process. Looked at this way, Grice-style intentions are a sort of meta-communicative insight that we employ in order to help us interpret cases in which basic
Assertion, Belief and Knowledge 45
interpretive processes either won’t work, or would be likely to lead to us acquiring false beliefs (cf. Millikan 1984: 69). The hybrid view I propose, then, is this: novice competence in assertion, of the sort developed by young children is indeed a form of perception by proxy in which comprehension and acceptance coincide. The onset of the ability to attribute beliefs allows the child to conceptualise assertions as expressions of belief, and as attempts to cause beliefs. This is the ability to attribute Gricean-style intentions to communicators. However, this new insight does not result in the wholesale rejection of the interpretative procedures that have served the child well up to that point. Rather, these are augmented by the child’s new capacities, and these are brought to bear on the interpretive process only when needed, such as when the speaker is judged to be incompetent, unreliable or untrustworthy, or when, for example, the nature of the interaction makes it obvious that the speaker is not intending to inform, but to persuade, as in academic discourse. This position does require that the ability to recognise the intention to communicate by the use of language (rather than to recite, say) is not equated with the ability to attribute Gricean intentions. But that is not a problematic requirement. Sperber and Wilson characterise overt communicative behaviour as ‘ostensive’. In communication of this type, the communicator produces a stimulus (linguistic or otherwise) that makes mutually manifest to communicator and audience that the communicator intends, by means of that stimulus, to make manifest a set of assumptions (1986/1995: 63). As theorists, we might gloss ‘make manifest’ as ‘cause to believe’ (though in doing so we would lose some of the attractive features of the notion of manifestness). From the perspective of the child who has no conception of belief, though, we can gloss it as ‘make known’. To recognise ostension, then, is just to recognise behaviour as being directed at providing information: we only need to think of it as aimed at expressing – and possibly instilling – belief if, whether as theorists or as sophisticated communicators, we adopt a meta-communicative stance. In sum, then: assertion is primarily a means of knowledge transfer in the sense that this is what the practice has evolved to achieve, and this is how it is treated by novice asserters. Moreover, even sophisticated communicators may treat it as such unless they have reason not to. When a hearer approaches assertion in this way, comprehension and acceptance coincide. Looked at from the outside, however, this knowledge-transfer process functions by means of belief expression. To view assertion this way is to take a meta-communicative stance towards it. Such a stance
46
Assertion
is available both to the theorist and to the sophisticated communicator, both of whom have the notion of belief in their conceptual repertoires. When the sophisticated communicator adopts this stance, comprehension and acceptance come apart. Sophisticated communicative interactions, such as academic discourse, rely on both speaker and hearer adopting this meta-communicative perspective. That said, questions of truth and falsity – and hence of acceptance and rejection – can be entertained only when the meta-communicative stance is taken. And as assertion is bound up with truth, full-fledged competence in assertion results only when such a stance can be taken. This is the point acknowledged by Dummett when he says, ‘we might say that the child cannot assert anything until he can lie. It is this insight which underlies Grice’s account of meaning’ (1981: 453).
3.4 Assertion and knowledge If proto-assertion were, as the McDowell view suggests, primarily perception by proxy, then it would not be surprising to find that some means of ensuring the reliability of this source of information were to develop. One possibility is that a cost because associated with passing on misinformation, such that an individual who did so became liable to receive some form of social sanction. This would explain how, in full-fledged assertion, speakers come to take responsibility for what they assert, such that they can be rebuked if what they say is false. For Williamson (1996), assertion is essentially an activity in which asserters take responsibility for the quality of the information that they pass on, this responsibility being such that speakers should only assert what they are certain of, i.e. what they know. In developing his position, Williamson seeks a constitutive rule for assertion that will distinguish it form other speech acts. A constitutive rule is, as the name suggests, a means of isolating the essence of an activity, and whatever rule is constitutive of assertion, Williamson argues, can be expressed in the form given in (9):9 (9) One must ((assert that P) only if certain conditions pertain) Formulating the constitutive rule for assertion requires specifying what the ‘certain conditions’ are. Williamson argues that the condition cannot simply be that P be true, for this would fail to individuate assertion, there being other speech acts, such as conjecturing and swearing, which are also subject to the rule that they should only be performed if the
Assertion, Belief and Knowledge 47
speaker believes them to be true, or likely to be so. One might want to argue that swearing that P is merely a strong form of asserting that P, but this, Williamson argues, would not undermine his case: for him, the key point is that assertion lies on a cline between conjecturing and swearing, each point on the cline having different standards of warranting evidence. Someone wanting to maintain that truth individuated assertion would have to argue that the standard of evidence required for assertion was ‘more intimately related to truth’ (1996: 497) than other truth-aiming speech acts. For Williamson, what distinguishes assertion from other truth-aiming speech acts is that the standard of evidence it requires is different from that required by the others: the evidence required for assertion, Williamson argues, is knowledge. On this view, the rule constitutive of assertion is: (10) One must ((assert that P) only if one knows that P) It is important to note that the claim is not that people only assert what they know, rather that they are open to rebuke if it turns out that they do not know what they assert. As García-Carpintero puts it: It is of course no objection to such an account that there are assertions whose producers lack knowledge of the contents they assert. The claim is not that knowledge by the asserter [. . .] of the asserted proposition is essential; the claim is rather that being subject to blame if knowledge [. . . is] missing is essential (2004: 145). In his attempt to show that knowledge is what warrants assertion, Williamson (1996) imagines a lottery in which the chances of winning are very low, say 14 million to 1, and in which there can only be one winner. A friend of yours has only one ticket and you say to him, after the draw has taken place but without either of you being aware of the outcome (but both aware of the odds): (11) Your ticket didn’t win The crucial point about (11) is that one can be very sure that it is true (the odds on it not being true are 14 million to 1), but it is still faulty in the circumstances described. If assertion were warranted by truth, though, then (11) should be OK, for it is highly likely to be true. The fact that (11) is not acceptable leads Williamson to conclude that
48
Assertion
what warrants assertion is not the very probable truth of P, but that the speaker knows that P (as opposed to her justifiably believing that P). (12) Your ticket is almost certain not to have won Moreover, Williamson argues that there is a difference in effect between (11) and (12) which would not exist if assertion were warranted by truth. (11) would be warranted, according to a defender of the truth-theory of assertion, by its almost certain truth. One should therefore expect, on this view, that (12) (which explicitly expresses the almost-certainty of the proposition expressed by (11)) would have the same effect and therefore be faulty in the same way. However, while (11) could be responded to, in the circumstances under consideration, by asking ‘How do you know?, (12) could be responded to by saying ‘I know’ (1996: 499–500). Or, to make the point another way, while (11) would be felt deceitful in the situation described, (12) would merely be uninformative. Indeed, the fact that one can respond ‘How do you know?’ to an assertion, Williamson argues, indicates that assertion presupposes knowledge. Moreover, ‘Do you know that?’ is an aggressive response to assertion, and the aggressiveness can be explained by the fact that it calls into question the speaker’s warrant to assert (1996: 505–6). The fact that speakers often assert when they do not have knowledge, Williamson goes on to argue, is not a problem for his account, as breaking a constitutive rule (a norm) does not mean that the activity it constitutes does not take place. There are different standards of rules for different occasions (the offside rule is not generally observed in informal games of football). Thus, in gossip, although one may assert without knowledge, it is not that the knowledge rule is not in force, but that breaking it is of little consequence (1996: 511). As noted above, the reason there exists a speech-act that is warranted by knowledge rather than (highly probable, belief-warranting) truth, Williamson suggests, is that, when one asserts, one takes responsibility for the truth of what one says, a responsibility one discharges by epistemically ensuring the truth of the content. It is because humans have a use for relations of responsibility, then, that the act of assertion exists, according to Williamson (1996: 521–2). Williamson’s claim that assertion is primarily a means of taking responsibility is supported by the variety of means a speaker has available to avoid taking epistemic responsibility for the information she seeks to communicate, even when she is inclined to take it to treat it as true. Evidential particles are the obvious example, but the epistemic use of ‘must’ and ‘can’t’ are
Assertion, Belief and Knowledge 49
notable as means of presenting information as certain without claiming knowledge. While generally very sympathetic to Williamson’s account of assertion, García-Carpintero (2004) argues that it fails to reflect the social function of assertion as a means of transferring knowledge: someone performing soliloquies in the absence of an audience could be complying with the knowledge rule, even though she would not be communicating anything to anybody. And surely, if one wants to explain assertion in terms of responsibility for information garnered via communication, then it must be viewed as primarily a communicative act. Moreover, from this perspective, Williamson’s knowledge rule can be argued not to individuate assertion, as someone who claims that knowledge is a prerequisite for assertion is also likely to view it as a prerequisite for presupposition: the grounds for one’s presuppositions must be the same as those that warrant one’s assertions. García-Carpintero (2004: 159) therefore proposes that the knowledge rule be replaced by the transfer of knowledge rule (13): (13) One must ((assert P) only if one’s audience comes thereby to be in a position to know P) It should be noted that Williamson is not the only author who sees assertions as knowledge claims. Brandom, whose work we considered in some detail in Chapter 2, argues that this view of assertion follows from his notion of discursive commitment. According to Brandom, to accept an assertion is to do three things: it is to ascribe to the asserter certain commitments; it is to judge that she is entitled to those commitments, and it is to acknowledge that those commitments apply in one’s own case. Each of these corresponds to one element of the standard view of knowledge as justified true belief: to ascribe practical and inferential commitments is to ascribe belief ; to grant entitlements is to view the belief as justified; and to undertake those commitments oneself is to assign the assertion the status of truth. It follows from this that an assertion, on Brandom’s account, is knowledge claim (Brandom 1994: 201–4; 2000: 168–9). For Brandom, then, assertions are knowledge claims because they are claims of justified true belief. Williamson (1995) does not accept the thesis that knowledge is analysable as justified true belief, so would not agree with Brandom’s reasoning. Where both authors concur, though, is on the importance of the notion of responsibility in understanding
50
Assertion
assertion. We have already seen in Chapter 2 that Brandom sees assertion as essentially involving undertaking the responsibility to provide reasons for the claims one makes, and we have seen in this chapter that Williamson also views the existence of the practice of assertion as explainable in terms of our use for notions of responsibility. Some have suggested that Williamson’s account offers nothing that could not be derived from a belief expression account of assertion. For example, Bach (forthcoming: n. 22) and Hindriks (2007) argue that the relationship between knowledge and assertion can be explained by viewing knowledge as the norm for belief. Looked at this way, they claim, assertion can be explained in terms of belief expression without having to posit a special relationship with knowledge. My own view is that the knowledge account is useful in the importance it places on epistemic responsibility. Belief-expression models of assertion can lose sight of the fact that an assertion is fundamentally faulty if it is false – regardless of of whether it expresses a sincerely held belief – and that the asserter can be rebuked for having made a false no matter how sincerely she holds the belief it expresses. If the effect of the knowledge account is to encourage proponents of belief-expression accounts to emphasise the importance of the beliefs expressed by assertions being justified and true, then it is to be welcomed. In order to establish whether Williamson’s knowledge account of assertion has anything to offer other than already made available by belief-expression accounts, we would have to ask whether he is right in denying that knowledge really is justified true belief, but that question would take us far beyond the scope of this book.10
3.5 Conclusion In this chapter, we have been concerned with the relationship between assertions and the mental states that this act expresses. We have seen that, from the perspective of conceptual priority, there is a good case to be made for putting assertion before belief. We also noted, though, that accepting this does not preclude us from reversing the order when adopting a scientific approach to the study of human communication. From the scientific perspective, we may see assertion as primarily about the communication of beliefs, but, again, perspectives must not be confused. From the claim that assertion is primarily about belief communication, it does not follow that interpreting assertions is necessarily about belief attribution. Belief attribution will certainly have a role to play in explaining sophisticated assertive behaviour, but it need not
Assertion, Belief and Knowledge 51
be brought in to explain every instance of successful interpretation. Rather, assumptions about the speaker’s complex intentions concerning the interlocutors’ belief states can be thought of as merely manifest to the sophisticated hearer, there to be drawn upon if needed, but not necessarily playing a causal role in interpretation. We also looked at Williamson’s knowledge account of assertion, noting that the extent that it could be said to be an alternative to a belief expression account depended ultimately on whether knowledge is explainable in terms of justified true belief, an issue that is beyond the scope of this book. In all of our discussion so far, however, we have not yet paused to ask why hearers might be interested in the beliefs (or, for that matter, knowledge) of others. The fact that someone expresses a belief, even a justified true one, is not in itself reason to attend to it. We are disposed to attend to assertions, surely, because we expect to derive cognitive benefit from them. It might be thought, then, that assertions are primarily relevance claims. This is the position that will be defended in Chapter 8, where it will be shown that their status as truth-claims follows from this, given the notion of relevance employed there. Before this position can be elaborated on, however, there is still substantial ground to be covered. In particular, we need to investigate the linguistic forms employed in the making of assertions. We begin looking at this topic in the next chapter, though it will be an issue that dominates the remainder off the book.
4 A Sign of Assertion
4.1 Introduction This chapter considers the place of assertoric forms in the analysis of assertion. It is often claimed that such forms are redundant: the sign that Frege employs in his logical symbolism to mark an assertion has been argued to be logically superfluous, and a number of authors have claimed that the declarative mood in natural language has no special relationship with assertion. The general thrust of this chapter will be that such views are misguided. In defence of Frege’s assertion sign, it will be argued that if one wants to use a logical system to build up a representation of how the world is, then such a sign is required on pain of infinite regress. Regarding the declarative mood, we will see how Dummett’s arguments (presented in the previous chapter) concerning the relationship between assertion, truth and belief require that a form specified for the performance of that act be posited. We will also see how a conception of assertion that stresses the commitments undertaken by those who make and assent to assertions leads to a straightforward analysis of non-assertoric uses of the declarative mood, uses which have been put forward as counterexamples to the view that the declarative is a marker of assertoric force. The chapter is organised as follows. Section 4.2 discusses Frege’s assertion sign and, in doing so, introduces the notion of assertion as a move in inference rather than, as we have been considering it so far, an illocutionary act. Consequently, in section 4.3, the differences between the two conceptions of assertion are considered. Finally, in section 4.4, I give considerable attention to the claim made by Dummett that the declarative mood is a natural-language equivalent of Frege’s assertion sign. 52
A Sign of Assertion 53
I show that the arguments levelled against this position can be resisted as long as the rendering one gives it is no stronger than that intended by Dummett.
4.2 Frege’s assertion sign Early in his presentation of his logical symbolism in the Begriffsschrift, Frege introduces a symbol ‘’, which, when prefixed to a string of symbols, indicates that that string expresses a content which has been judged true. The symbol itself is in fact a combination of two elements: the horizontal line indicates that the symbols that follow are bound together as the content of a potential assertion, the addition of the vertical line indicates that that content is asserted. Frege named the horizontal line ‘the content stroke’ and the vertical line ‘the judgement stroke’. Russell and Whitehead, in the Principia Mathematica, call the combined symbol ‘’ ‘the assertion sign’. In introducing this sign, Frege recognises that the same proposition – ‘thought’, in his terminology – can occur, in discourse and in logic, both with and without assertoric force.1 Geach (1965: 449) dubs this ‘the Frege point’, and points out that it is enlightening as to the nature of modus ponens. Expressed as in (1), one might ask why P is not redundant in the second line of the argument, for it is already to be found in the first premise. The Frege point is that, although present in both premises, P is only judged true by the introduction of the second premise. Frege’s assertion sign serves to make this explicit, as in (2): (1)
(2)
P→Q P Q (P → Q) P Q
But if all Frege’s judgement stroke offers is greater perspicuity, then it has no defence against the criticism that it is logically superfluous, nothing more than a punctuation mark. This is the complaint made by Wittgenstein, who says that ‘Frege’s assertion sign marks the beginning of the sentence. Thus its function is like that of a full-stop’ (1958: §22, original emphasis). Another criticism is that the judgement that lies behind assertion is a psychological phenomenon, and, as such, should have no place in logic (see, e.g. Anscombe 1959/1996: ch. 8). Indeed,
54
Assertion
Frege himself was adamant that psychology should play no part in logic (1897/1997: 243–50), so it is not immediately clear why he included in his symbolism a sign that signifies the psychological act of judging true. To see why he did so, we need to consider his motivation for developing his system of logical notation. Despite Frege’s immense influence on modern logic, his assertion sign is not commonplace today (the symbol ‘’ standardly means ‘is a theorem’ these days). The reason for this is that what follows logically from a set of premises is not determined by the attitudes towards those propositions: you and I may agree that an argument is valid, while disagreeing that it is sound; I may accept premises that you reject, for example. If logic is conceived of the study of valid arguments, then, the assertion sign is indeed redundant. However, logic has not always been conceived of in this way. Green points out that Frege was part of a tradition that saw the refinement of logical systems as having as one of its goals the development of a language that would, among other things, ‘serve as an organon for the discovery of new truths or for the systematization of what is already known’ (Green 2002: 207). This point, Green argues, is key to an appreciation of Frege’s reasons for employing his assertion sign, in spite of his staunch anti-psychologist stance. Both Green (2002) and Smith (2000) argue that Frege draws a distinction between inference and pseudo-inference, and that appreciating this point is the key to understanding the central role he gave to the assertion sign. Inference, for Frege, is a process governed by the laws of logic that is aimed at increasing a stock of knowledge. The application and study of the rules of logic for purposes other than the confirmation and extension of knowledge is, for Frege, mere pseudo-inference. When engaging in pseudo-inference the truth or falsity of the premises is indeed immaterial. When engaging in inference proper, however, the truth-status of the premises employed is of great consequence. Green puts it like this: if we wish to indicate which thoughts are known to be true and thus capable of funding further discovery or reconstruction, a device for marking these thoughts will reflect a distinction relevant to inference and thereby redound to the logical adequacy of a notation (Green 2002: 214). Frege, then, employs the assertion sign in his logical symbolism because he envisages its eventual use as a tool for the perspicuous expression and subsequent extension of knowledge: it is to be employed ultimately
A Sign of Assertion 55
in developing an accurate representation of the world, not merely as an object of study for logicians. That said, Wittgenstein is right that the assertion sign does not add to the expressive power of the symbolism. There is no proposition that can be expressed using this sign that cannot be expressed without it. Rather, the assertion sign is best thought of as providing a format for the expression of propositions judged true. The alternative to specifying such a format is to build into the conceptual repertoire of the system a predicate, such as ‘. . . is true’, that takes propositions as its subject. This would be applied to all propositions judged true, so that if one judged P true, then ‘That P is true’ would be added to one’s stock of knowledge, but then so would ‘That P is true is true’, and so on, ad infinitum. The use of an assertion sign avoids this problem. As Frege (1918–19/1997: 330) puts it: ‘We express acknowledgement of truth in the form of an assertoric sentence. We do not need the word ‘true’ for this.’ The point I take from this is that if one wants to use an inferential system as a means of world representation, then one needs to specify a propositional format whose use counts as acceptance of the truth of the proposition expressed. Failure to do so will result in regress. We can say that such forms, while not specified for the performance of a speech act, nevertheless have ‘assertoric character’. The term comes from Geach, who observes that ‘The assertoric force of a sentence is [. . .] shown by its not being enclosed in the context of a longer sentence’ and then goes on to speculate that: Possibly there is something corresponding to this in the realm of thoughts; possibly a thought is assertoric in character unless it loses this character by occurring only as an element in a more complicated thought [. . .] This would be a neat solution to how thought is related to judgment (Geach 1960: 456–7). What representations with assertoric character have in common with assertions qua speech acts is that their employment results in commitment. One who entertains a thought in a format which has assertoric character can neither reject its inferential consequences nor fail to act upon it. In a parallel fashion, one who makes an assertion is committed to the inferential consequences of that assertion and to acting on it. Thus if a representation has assertoric character, then its use results in assertoric commitment. As we will see in more detail in Chapter 8, the framework employed there – Sperber and Wilson’s Relevance Theory – posits a
56
Assertion
representational format at the level of thought which has assertoric character in Geach’s sense. Sperber and Wilson call a thought expressed in this representational format a ‘factual assumption’. To stipulate such a format in a model of cognition, they note, is to posit that ‘a fundamental propositional attitude of belief or assumption is pre-wired into the very architecture of the mind. As a result, a representation can be entertained as an assumption without the fact that it is an assumption being explicitly expressed’ (1986/1995: 74). Because both Frege and Sperber and Wilson – despite the great differences in the overall goals of their projects – have in common the aim of developing an inferential system of world-representation, both are compelled to stipulate a format whose employment in the expression of a proposition counts as acceptance of that proposition. Is there an equivalent format in natural language? And if so, what is it? One possibility is that there is a convention (where in Frege and Sperber and Wilson’s systems there is a stipulation) such that uttering a declarative sentence that explicitly expresses the proposition that P is to assert that P. But, as we will see in section 4.4, it has been objected that such convention could never hold sway in human linguistic communication, for it would be always open to abuse. Before considering this issue, though, we need to dwell a little on the move we have made from treating, in the previous two chapters, assertion as an illocutionary act to considering it, in this section, a move in inference. This raises the issue of the relationship between assertions qua speech acts and assertions qua premises and conclusions. This issue is discussed in the next section.
4.3 Assertion: illocution and inference As noted in Chapter 2, Brandom says that assertions are ‘fodder for inferences’ (1994: 168). But what exactly does that mean? There is one sense in which not only assertions, but all illocutionary acts are fodder for inference, for linguistic communication relies on inference and thus all speech acts provide, in that sense, input to an inferential process. Brandom’s point, though, is that assertions are a source of premises that can be used in inference beyond the identification of the speaker’s meaning. He means that assertions are a source of premises that enable one to enlarge, through inference, one’s stock of knowledge. But it is one thing to say that assertions qua speech acts are sources of premises and another to say that they can serve as premises in inference. To be sure, when we present inferences using natural language, we use
A Sign of Assertion 57
declarative sentences (i.e. assertoric forms) to represent premises and conclusions: (3) If Peter does not study, he will fail the exam Peter does not study Peter will fail the exam But an ‘inference’ such as (3) is very informal. It relies heavily on human cognitive dispositions. For example, we are disposed to assume that the individual denoted by ‘Peter’ is the same in both the initial and second premise, and that ‘he’ in the consequent of the conditional is co-extensive with ‘Peter’ in the antecedent. It also relies on acceptance of modus ponens as a valid rule of inference. A formal inference system must make explicit all of these assumptions if it is not to be reliant itself on the inferential abilities of its users. Thus pragmatic inference plays a significant role in the interpretation of ‘inferences’ such as (3). That is to say, we are inclined to say that the argument in (3) is valid because we bring to bear on it processes we employ in utterance interpretation. We are disposed to treat (3) as a stretch of discourse, and in finding it valid we are in fact judging that the discourse is coherent. In Brandom’s terms, we are saying that one who asserts both ‘premises’ in (3) is entitled also to assert the conclusion. That we can represent an inference as in (3) does not show that natural-language assertions can serve as premises. Rather, it shows that declarative sentences can be used to represent and communicate the premises and conclusions of an inference (cf. Stainton 2006: 183–5). There is a difference between what is required of a form used to represent a premise that plays a role in an inference and what is required of a form that is employed as a vehicle of inference.2 To serve as a vehicle of inference, a propositional form must meet two criteria. First, it must encode determinate truth conditions. That is to say, it must encode the proposition it expresses, for otherwise inference will itself be required in order to determine that proposition, and we will have a problem of regress. Second, a propositional form must be expressed in a format recognised by an inferential system as being of a type that can serve as a premise in that system. This second point is emphasised by Millikan, who illustrates it thus: suppose that the first premise of a would-be inference is stored in a Venn diagram and the second in a circuit diagram, or on a map, or
58
Assertion
in the notation of Principia Mathematica. No rule of inference could combine these premises directly, without some kind of translation (2004: 217). Natural language assertions fail to meet both of the criteria for serving as a vehicle of inference: the sentences that they employ underdetermine the truth conditions of the propositions asserted; and they are not designed to serve as premises in a formal inference system, but as stimuli in a process of inferential communication. The sense in which an illocutionary assertion is fodder for inference, then, is that it expresses a proposition and provides a warrant to introduce that proposition as a premise in inference. Because such an assertion does not, though, encode a proposition in a format apt for inference, it is important to distinguish between assertions qua premises in inference and assertions qua speech acts. What the two have in common is that they express propositions and carry with them a commitment to consequences of acceptance (or judgment). How they differ is that the latter do not encode the propositions that they express and cannot therefore serve as vehicles of inference. Because illocutionary assertions have a special relationship with inference in that they represent premises and provide warrant for their introduction, one way we can show that an utterance is an assertion is to show that it can be used with others as a representation of an inference, such as in (3). Pagin calls this the ‘inferential integration test’ and he uses it in an attempt to show that in uttering an explicit performative of the form ‘I hereby assert that P’, the speaker does not assert that P (2004: 855):3 (4) If 73 is a prime number, we cannot share the stones equally. I hereby assert that 73 is a prime number. So, we cannot share the stones equally. However, as we will now see, care is needed when applying this test. The claim that Pagin makes is that the explicit performative does not introduce the premise that 73 is a prime number. This, the claim goes, shows that in uttering ‘I hereby assert that 73 is a prime number, one does not assert that 73 is a prime number. But while (4) is certainly odd, it can easily be made into a coherent stretch of discourse by presenting it as a conversation, and having the addressee of the explicit performative explicitly accept the assertion that 73 is a prime number:
A Sign of Assertion 59
(5) Mary: If 73 is a prime number, we cannot share the stones equally. Peter: I hereby assert that 73 is a prime number. Mary: I agree. So, we cannot share the stones equally. The problem with the rendering of the inference in (3) arises because, when evaluating such inferences, we tend to treat the premises and conclusions as a soliloquy. In a soliloquy, assertions are not directed at an audience and hence there is no need for acceptance to be signalled: instead, assertion and acceptance coincide. In a dialogue, by contrast, acceptance of an assertion must be signalled by the hearer (though perhaps only by the absence of an objection) before the asserted proposition can become part of the conversational record. Introducing explicit performatives into an ‘inference’ such as (4) makes it difficult to interpret as a soliloquy, for such utterances are intrinsically audience-directed acts, whether they are considered declarations (Recanati 1987; Searle 1989) or acts of showing (Jary 2007).4 This is the source of the oddness of (4). By providing a context in which the explicit performative is clearly directed at an audience, the dialogue in (5) rectifies this oddness and demonstrates that the inferential integration test cannot be used to show that by uttering ‘I hereby assert that P’ one cannot thereby assert that P.5 To sum up this section, assertions qua speech acts and assertions qua premises both have assertoric character. But while both express a proposition and commit the user to the truth of that proposition, they are not identical: a premise must encode the proposition it expresses, so that it can be a vehicle of inference, whereas an assertion qua speech act is not subject to this requirement. Assertions qua speech acts express a proposition and grant warrant for the introduction of that proposition as a premise. Before it can be used as premise, though, the proposition must be encoded in a format recognised by an inferential system. The inferential integration test, while a useful tool for identifying assertions (or, at least, truth-apt speech acts), needs to be applied with care: the fact that it relies on our pragmatic intuitions regarding discourse coherence needs to be borne in mind in order to ensure that any apparent failure to pass the test is not in fact a result of pragmatic infelicity.
4.4 A natural-language assertion sign? As noted earlier, when constructing a formal inference system such as the Begriffsschrift, it can be stipulated that writing a proposition P in a specified format imbues P with assertoric character. We need now to
60
Assertion
ask whether the declarative mood is a natural-language equivalent of Frege’s assertion sign. That is to say, does its use result in a commitment to treat the proposition expressed as true? Dummett has argued that unembedded declarative syntax has just this function.6 But where Frege stipulates, Dummett relies on convention: for him, assertion is to be explained in terms of the conventions associated with the use of assertoric forms (1981: 311), declarative mood being ‘ordinarily’ a sign of assertoric force (i.e. an assertoric form) (1981: 357; 1993: 208). In this section, we will consider this position in some detail, paying particular attention to objections that have been levelled against it. Because Dummett sees assertion in terms of conventions associated with the use of the declarative mood, his position can be stated either from the perspective of mood, or from the perspective of force. From the former perspective, we can say that Dummett holds the Declarative Thesis: The Declarative Thesis: The declarative mood is a conventional sign of assertoric force. From the perspective of force, we can say that he holds the Assertion Thesis: The Assertion Thesis: Assertion is to be explained in terms of conventions associated with the use of declarative sentences. It is important to stress that, as far as Dummett is concerned, the notion of assertion is to be explained wholly in terms of convention. Thus, in saying that the declarative mood is a conventional sign of assertoric force, he does not mean that it is merely a reliable indicator of that force, but that due to the conventions associated with it, its use under appropriate conditions constitutes assertion. It is, for Dummett, a fully fledged Fregean assertion sign. As we will soon see, though, this does not commit him to the view that all assertions must be performed using the declarative mood, nor to the view that only the declarative can be used to make assertions. This section is organised around the challenges that might be levelled at the two theses. The Declarative Thesis has been objected to by pointing out cases where this mood is used to perform acts other than assertives. Such objections and the responses that are open to Dummett (some of which he himself makes) are discussed in 4.4.1. The Assertion Thesis has been objected to by pointing to cases in which forms other
A Sign of Assertion 61
than the declarative are used to perform assertions. Dummett’s response to this type of objection is discussed in section 4.4.2. The section concludes by returning to the issue of whether, in the light of the data discussed, assertion can be explained purely in terms of convention. 4.4.1 Challenges to the Declarative Thesis We begin this section by looking in detail at Davidson’s claim that, due to the fact that linguistic conventions do not constrain the use to which linguistic forms can be put, there could be no such thing as a naturallanguage equivalent of Frege’s assertion sign. We then go on to look at the use of the declarative to perform acts other than assertion. We will see that detailed consideration of these cases actually strengthens, rather than weakens, the claim that the declarative is a marker of assertion, as long as due emphasis is given to the consequences of assertion in the analysis of this act, as Dummett and those who follow him, such as Brandom, suggest. 4.4.1.1 ‘The autonomy of linguistic meaning’ and the declarative as assertion sign Arguing against the view that the declarative mood is a conventional sign of assertion, Davidson makes the strong claim that the nature of language is such that conventions of meaning can never fully constrain use, and that consequently there could never be a natural-language equivalent of Frege’s assertion sign. A fundamental feature of language, according to Davidson, is what he calls ‘the autonomy of linguistic meaning’: Once a feature of language has been given conventional expression, it can be used to serve many extra-linguistic ends; symbolic representation necessarily breaks any close tie with extra-linguistic purpose. Applied to the present case, this means that there cannot be a form of speech which, solely by dint of its conventional meaning, can be used only for a given purpose, such as making an assertion or asking a question (1979/2001: 113–14). Davidson’s point here is that a conventional relationship between a linguistic form and a meaning cannot guarantee that, on every occasion of use, that form is used in accordance with that convention. In terms of the relationship between mood and force, Davidson’s thesis has the consequence of ruling out the possibility of a natural-language equivalent of Frege’s assertion sign: in other words, if this thesis is correct, no
62
Assertion
propositional form in a language such as English could possibly have assertoric character. Hare (1989) objects, arguing that for any law, there could be a convention to the same effect. And, he says, there is no a priori reason that there could not be a law such that whoever uttered a certain propositional form was thereby committed to the proposition she expressed. Such a law, he points out, certainly governs the giving of testimony under oath, and one can imagine a possible world in which the same law applied to all utterances of declarative sentences, so that speakers were subject to rebuke if they uttered a falsehood under any conditions. As Hare notes, such a law would make acting and storytelling impossible (or at least, we might add, a risky business – there is a law against murder, but that does not prevent it from happening), but there is nothing conceptually incoherent about such a world. Nevertheless, the world we are in does have actors and storytellers, and they do manage to employ declarative sentences without fear of rebuke, so the fact that a world in which the declarative is a failsafe sign of assertion is a conceptual possibility is of little help in helping us to understand the role of declarative syntax in the practice of assertion in our world. Green takes a different tack. He notes that common counterexamples to the proposal that use of the declarative is a guarantee of assertoric force, such as fiction and acting, have in common the feature that the speakers are not performing illocutionary acts at all (1997: 222). However, one can allow that there is no form such that its production guarantees that the speaker is performing a particular illocutionary act without also giving up the possibility of there being a form that, granted that the speaker is engaged in one illocutionary act or another, indicates which act she is performing. Green calls such an illocutionary force indicator a ‘weak IFID’, a strong IFID being one that guarantees that the speaker is performing the associated act without the need to first ascertain that she is performing an illocutionary act of one type or another. We might agree that the use of the declarative mood does not commit a speaker to the truth of the proposition expressed regardless of the activity in which she is engaged, but still allow that it indicates which illocutionary act she is performing if it is first granted that she is performing one or another such act. This would make the declarative a weak IFID, in Green’s terms. There is one sense in which declarative mood clearly is not a weak IFID: it does not convey assertoric force if it is embedded in the antecedent of a conditional. But Dummett maintains that illocutionaryforce indication is a feature of sentences as a whole, not of clauses that serve as constituents of sentences (1981: 304), so this observation
A Sign of Assertion 63
cannot serve as an objection to his claim that the declarative is an assertoric-force indicator. However, as Green points out, there are cases where syntactically unembedded declarative sentences are uttered in the performance of an illocutionary act, yet do not carry assertoric force. This is the case, for example, in a discourse such as the following (1997: 227): (6) Let us suppose the next sentence for the sake of argument. Snow is white. By this light, then, a declarative sentence is not even a weak indicator of assertoric force, for it can be used in a speech act without thereby committing the speaker to the truth of the proposition expressed. However, we should note that, while not linguistically embedded, the utterance of ‘Snow is white’ is ‘cordoned off’ by the previous utterance. If ‘Snow is white’ were uttered, unembedded, in the performance of a speech act and in the absence of an introductory edict such as we find in (6), then the speaker would be taken to have asserted that snow is white, and this would be as a result of her choice of linguistic mood. Had she chosen to utter, under the same circumstances (and non-rhetorically), ‘Is snow white?’, for example, she would not have been taken to have made an assertion.7 In other words, cases such as (6) do not rule out the view that assertion is the default interpretation for utterances of declarative sentences, in that the utterance of a syntactically unembedded declarative sentence counts as an assertion unless there are contextual indications to the contrary. I think this is the correct way to view the relationship between the declarative and assertion. It is also a fairer rendering of Dummett’s view than the one attributed to him by those, such as Davidson, who object to the Declarative Thesis on the grounds that there are non-assertoric use of declarative sentences. In saying that the declarative is a sign of assertoric force, Dummett does not rule out the possibility of it being used, even unembedded, without the speaker thereby asserting. Indeed, he allows that if declarative syntax is used by an actor engaged in dramatic representation (a case that is commonly put forward as evidence against the view that the declarative is a conventional force marker), then it conveys something other than assertoric force. He does not accept, though, that when a declarative used in such a way it is completely devoid of assertoric force. Rather, he holds that what happens in such a case cannot be understood without recourse to the notion of assertion. In such a case, he says, the actor does more than assert: she
64
Assertion
asserts in a context that limits the force of her utterance. But one cannot characterise her behaviour without reference to the notion of assertoric force, for in order to understand what the actor is doing, Dummett argues, we need to treat her utterance as an assertion made under special circumstances. In other words, it is not the case that the actor is not, in any sense, asserting: it is that she is doing more than asserting. Were we to employ symbols to indicate what she is doing, rather than omit the assertion sign altogether, we would use the assertion sign prefixed by another sign that indicated ‘dramatic force’ (1981: 311; 1993:212). It is worth spending some time considering this point and the objections that have been made against it, for it has considerable bearing, as we will see, on the correct analysis of declarative mood. If it can be shown that non-assertoric uses of the declarative require, for their correct analysis, recourse to what we might term ‘modified’ or ‘restricted’ assertoric effects, as Dummett claims is the case with stage assertions, then it becomes more reasonable to claim that the declarative has a special relationship with assertion, rather than being, as some have maintained (Davidson 1979/2001; Recanati 1987), force-neutral. We will therefore now give some consideration to an objection to Dummett’s analysis of stage assertions made by Recanati (1987). Recanati claims that Dummett’s analysis of stage assertions fails to recognise a distinction between the force indicated by a form and the presence of that force in fact. Recanati’s objection relies on an observation made by Hare (1970/1971: 89–93; 1989) that the Fregean assertion sign, rather than signalling merely completeness and judgement, actually contains three elements. Hare shows that the assertion sign functions, in the work of Frege and Russell and Whitehead, also as a mood marker (a ‘tropic’ in Hare’s terminology). That is to say, it specifies that the descriptive content it prefixes (Hare’s ‘phrastic’) is expressed in a format that renders it assertable, i.e. in an assertoric format. For it to be asserted, though, another element is needed to signal judgement (Hare calls this a ‘neustic’). Thus what distinguishes the antecedent of a conditional premise and the minor premise in a case of modus ponens is, in Hare’s terms, that the latter contains a neustic which is absent in the former. This is why, for Hare, uttering a conditional sentence does not commit the speaker to the truth of the proposition expressed by the antecedent. The need to distinguish between neustic and tropic (the logical correlates of mood and force) is obscured in the work of Frege, and that of Russell and Whitehead, because they are concerned with inferences involving only assertoric forms. If one wants to develop an inference system that guides action, however, then a means of representing states
A Sign of Assertion 65
of affairs that are goals, rather than actual, is needed. This requires that two distinctions be marked. On the one hand, a distinction is needed between those propositional forms which are assertoric in potential and are in fact asserted, and those which have assertoric potential but are not asserted. On the other hand, a distinction must be marked between those forms that have assertoric potential and those (such as imperative syntax in natural language) which do not. Such a distinction is an issue for Hare, who is concerned with ethics, and for Grice (2001) in his work on practical reasoning. According to Dummett, if one actor says to another (whose character is named ‘Jim’, let’s say) ‘You are an imbecile’, we can represent her utterance as (7), where ‘’ is a marker of dramatic force, ‘’ is the assertion sign (both neustic and tropic) and angle-brackets contain the propositional content:8 (7) Recanati, however, argues, contra Dummett, that in this case, ‘’ is not functioning as a marker of assertoric force, but merely as an indicator of assertoric potential, i.e. as a mood marker (i.e. merely a tropic). However, Recanati (1987: 263) also uses the distinction between indicated force and actual force (i.e. between neustic and tropic) to explain the difference between A’s utterance and B’s utterance in (8): (8) A: You are an imbecile B: Oh, I am an imbecile! Thank you very much. According to Recantati, the force in A’s utterance is both indicated and actual (i.e. A is asserting that B is an imbecile), while in B’s it is merely indicated. The problem for Recantati is that a dialogue such as (8) could easily arise in a play. Some means must therefore be found of explaining the difference between the two utterances in such a context. Dummett’s claim (that A is asserting the proposition he expresses while at the same time doing so only within the conventions of dramatic representation) explains how the audience recognises the difference between A’s utterance and B’s. Recanati, though, is committed to the view that both A’s and B’s utterances, if they occur in dramatic representation, only indicate force, and that in neither utterance is the potential force realised. He therefore has no account of how the difference between the two. This point can be illustrated as follows. First, we separate tropic and neustic: let ‘’ be unambiguously a sign of a propositional form’s
66
Assertion
assertoric potential (Hare’s ‘tropic’) and ‘’ be a sign that the proposition is uttered with its indicated force (Hare’s ‘neustic’). We can now represent the dialogue in (8) as (9): (9) A: P B: P According to Recanati, in order to represent the dialogue as spoken by actors, we must remove the neustic. But this gives us (10), and we are no longer able to distinguish the two utterances. (10) A: P B: P The conclusion we must draw is that the speech of actors engaged in dramatic representation cannot be explained in terms of their merely indicating linguistically the force potential that their utterances have. Rather, as Dummett advises, we have to view a stage assertion as an assertion made under special circumstances. These circumstances have the effect of restricting the consequences of the assertion, so that the commitment undertaken is limited to the confines of the dramatic representation. Our understanding of that representation, however, relies on us treating the actor as one who has asserted, rather than one who has simply indicated that she has asserted by her use of a form with assertoric potential. Pace Recanati, there is more to stage assertions than the mere indication of force potential: there is also force of a kind. How does this relate to current concerns? It was suggested earlier that the declarative is a default maker of assertoric force, so that it counts as an instance of assertion unless there are overriding contextual conditions. What the discussion of the acting case indicates, though, is that even when overriding contextual conditions pertain the resulting act is one whose assertoric force is modified in some way, rather than completely nullified. In Hare’s terms, the neustic is not absent – as is the case when the declarative is found in the antecedent of a conditional – but prefixed by another sign that limits the consequences of its application. It is a mistake, then, to view non-assertoric uses of the declarative as being on a par with its use in the antecedent of a conditional. In the latter case, the neustic is indeed missing (and Dummett agrees with this). A neustic of some sort, however, must be present in those cases in which an actor is acting an assertion, otherwise we would have no grounds for attributing to the character whom the actor portrays the
A Sign of Assertion 67
commitments that she thereby undertakes within the confines of the dramatic representation in which she is engaged. Thus the correct analysis of stage assertions requires recourse to the consequences argued by Dummett to be associated with the use of the declarative mood by virtue of its function as a marker of assertoric force. This supports the view that the declarative is indeed an assertion sign (or, at least, shows that these cases cannot be used as objections to this view). However, the existence of stage assertions is not the only objection that has been made against this position. In the remainder of this section, we will consider other cases that are commonly put forward as instances of the non-assertoric use of an unembedded declarative in order to see whether these can also be explained in terms of the assertoric consequences associated, on Dummett’s view, with this form. In the following subsections, then, we will look at the use of the declarative in cases of supposition and fiction, in the performance of directives and commissives, and, finally, in the performance of explicit performatives and declarations. We will see that in each case a straightforward analysis is afforded by the assumption that the declarative is marker of assertion. 4.4.1.2 Discourse embedding and fiction What I am calling ‘discourse embedding’ occurs when an utterance of a syntactically unembedded sentence is governed by previous discourse that precludes an assertoric interpretation. My reasons for discussing it under the same heading as fiction-creation will become clear as we progress. One example of discourse embedding already given is (6); another is (11). In these cases, the speaker undertakes a non-assertoric mode of commitment to the proposition expressed. The speaker of (6) is committed to the proposition that snow is white as a supposition for the sake of argument, while the speaker of (11) is committed to it as a conjecture. In Dummettian terms, these commitments are the inferential consequences of supposition and conjecture. (6) Let us suppose the next sentence for the sake of argument. Snow is white. (11) I therefore put forward the following conjecture: snow is white The modes of commitment associated with assertion, conjecture and supposition differ in terms of their relationship to truth and evidence (or warrant). As we saw in our discussion of the work of Brandom and Williamson in previous chapters, a speaker who undertakes assertoric
68
Assertion
commitment to P thereby undertakes responsibility, if challenged by being asked, for example, ‘How do you know?’, to provide evidence that establishes that P (perhaps by citing a source from whose prior assertion she has received warrant to assert that P). Commitment under the mode of conjecture, however, does not involve undertaking this degree of responsibility. A speaker who conjectures that P is merely obliged to show, if challenged, that P is compatible with the set of propositions to which she is assertorically committed. What both modes of commitment have in common is that the speaker is required to reject P if it is shown that P is false. This does not hold for supposition: a speaker may legitimately suppose a false proposition (Green 1999). But while they differ greatly in their relationship to truth and in the grounds on which they can be challenged, assertion, conjecture and supposition have in common a commitment to consistency. Although one is entitled to suppose a falsehood, one cannot simultaneously suppose two contradictory propositions: for example, one cannot simultaneously suppose for the sake of argument that fairies exist and that fairies don’t exist. Importantly, this commitment to consistency does not hold for all propositional attitudes: I can both want not to damage the environment and want to take a long-haul flight. In wanting both these things, I am not being inconsistent, even though the desires are incompatible. There is a difference, then, between contradiction and incompatibility. P and Q are incompatible if both cannot be true simultaneously. A contradiction results, however, only if P and Q are objects of a propositional attitude that requires consistency. One way of thinking about the distinction is this: if P and Q are incompatible propositions, they are precluded (under ideal conditions) from inclusion in your belief box, but can be included in your desire box, for only the former has a consistency requirement. When one has incompatible desires, the result is a dilemma, not a contradiction. It is obvious why consistency is a requirement of belief: belief aims at truth and consistency is a means of preserving truth. As assertion also aims at truth, the mode of commitment undertaken by asserters also entails commitment to consistency, and the same goes for conjecture, which is also truth-aiming. But not all consistency-aiming endeavours aim at truth. As noted above, a speaker cannot be criticised for supposing a falsehood. She can, however, be rebuked if she supposes an inconsistent set of propositions. Similarly, although fiction is not truth aiming, it does create expectations of consistency (cf. Chierchia and McConnellGinet 2000: 216–17): if a character dies early in the narrative and then
A Sign of Assertion 69
reappears later, we are entitled to demand an explanation, to require that reason be given.9 Recanati (2000: 49–62) analyses supposition and fiction as a form of simulation. In supposing for the sake of argument or in recounting a story we engage, he says, in simulation. What we are simulating is assertion. On this view, then supposition and fiction are pretend assertions. In pretending to assert, as when we suppose, conjecture or create fiction, we undertake a mode of commitment to the consequences that would follow from actually making an assertion. Thus what unites these modes of commitment is that they all involve treating as true without claiming truth. Stalnaker’s term for this, which I will adopt henceforth, is ‘acceptance’: ‘To accept a proposition is to treat it as a true proposition one way or another’ (Stalnaker 1984: 79). Acceptance thus defined applies both to assertion and to the forms of simulated assertion we are considering here. What assertions and simulated assertion have in common, then, is acceptance of the inferential consequences of asserting that P. Acceptance involves a commitment to consistency: one cannot accept incompatible propositions for to do so would be a failure to treat them as true. How assertions and simulated assertions differ, is, in part, explicable in terms of how they relate to the conditions for asserting that P. While asserting that P is an undertaking that the conditions for asserting that P obtain (or, in the case of past tense and future tense assertions, did or will obtain,), conjecture, supposition and fiction involve no such undertaking. Conjecture, however, differs from supposition and fiction in that, while it does not involve a claim that the conditions for asserting that P obtain, demonstration they do not obtain obliges the speaker to withdraw her conjecture; no such requirement holds for supposition and fiction. The similarities and differences between these modes of commitment are summarised in Table 4.1.10 Thus discourse embedded and fictional uses of the declarative have in common an undertaking of commitment to the inferential consequences of acceptance of their propositional content. They are better analysed, therefore, as cases of simulated assertion, rather than as cases where the assertoric potential of the form employed is indicated but not fulfilled, i.e. where the assertoric tropic is present but the neustic is absent. The problem with viewing these as neustic-less uses of an assertoric form is that this fails to explain the commitments taken on by speakers. The neustic is certainly completely absent when an assertoric form is embedded in the antecedent of a conditional, for no commitment to the consequences of asserting that proposition is undertaken.
70
Assertion
Table 4.1
Modes of commitment Assertion
Conjecture
Supposition
Fiction
Is a claim that the conditions for asserting that P obtain
Yes
No
No
No
Must be withdrawn if it is shown that these conditions do not obtain
Yes
Yes
No
No
Is an undertaking of commitment (of some kind) to the inferential consequences of asserting that P
Yes
Yes
Yes
Yes
But discourse-embedded and fiction-creating uses of the declarative do involve taking on commitments and hence have little in common with its use as the antecedent of a conditional. They have much more in common with declaratives used as a means of undertaking assertoric commitment and therefore support, rather than undermine, Dummett’s claim that assertion is an act that is to be explained in terms of the use of the conventions associated with the use of the declarative. For some, but not all, of the conventional significance of the assertoric use of declarative syntax carries over to these non-assertoric uses. How discourse-embedded and fiction-creating uses of the declarative differ from assertion is in terms of their relationship to the conditions for assertion: only assertion is a claim about the obtaining of truth conditions.11 4.4.1.3 Directives and commissives Another observation that is often put forward as evidence against the claim that the declarative is an assertion sign is that it can be used to perform acts other than assertives. One example that has been given is the use of future-tense declaratives to issue directives, such as (12), which is adapted from Recanati (1987: 138): (12) You will clean the latrines. For Recanati, the existence of directives, such as (12), that employ declarative syntax is evidence that the declarative is force-neutral
A Sign of Assertion 71
(1987: 163–9). However, it is possible to provide an account of these directives that treats the declarative as primarily an indicator of assertoric force. Moreover, this can be done without the need to treat (12) as an indirect speech act, in the sense of Searle (1975/1979). First, note that future-tense declaratives are only acceptable as directives in a situation in which the speaker has authority over the hearer, and when the act predicated of the hearer is within the boundaries of what it is permissible for the speaker to require of him. Thus a beggar cannot issue a plea (another type of directive) by uttering (13). The use of future-tense declarative syntax is thus more limited than the imperative in terms of its potential for the performance of directives: both (12) and (13) could be performed using the imperative mood, as (14) and (15) show: (13) You will spare me some change (14) Clean the latrines (15) Spare me some change This point established, we can then ask what would be the effect of an utterance made using assertoric syntax that predicated a future act of an addressee over whom the speaker had authority. Future-tense use of assertoric syntax that expresses the proposition that P, on Dummett’s view, involves the speaker undertaking a commitment to the truth of P, and the putting forward of P for acceptance or rejection by the hearer. In the absence of a countermanding directive from a higher authority, the hearer will have no choice but to accept that P (unless he is prepared to accept the consequences of insubordination), for, as far as he is concerned, the speaker’s decisions about the hearer’s future actions determine those actions. He will have no grounds, therefore, for rejecting P, given that the speaker had expressed, by means of that utterance, her decision concerning his future behaviour. And, for the speaker to accept that P is in a case such as (12), in effect, for him to undertake to perform the act predicated of him. In practical terms, then, the use of assertoric syntax in such a case functions to convey directive force. However, the directive force of the utterance results from the potential of the form employed to commit the speaker to the proposition expressed and to put that proposition forward for acceptance or rejection by the hearer. In other words, it results from the form’s assertoric potential. Consequently, (12) does not count as an example against the view that declarative syntax is, under less narrow conditions, a marker of assertoric force. Rather, it is just this assumption that underlies this explanation of how an utterance such as (12) comes to have the force of directive.
72
Assertion
Note that this account does not treat (12) as an indirect speech act: for a speech act to be performed indirectly, two acts must be performed, the indirect one by means of a direct act. In the case just discussed, though, the hearer does not assign assertoric force to the utterance, find this interpretation unsatisfactory, and then go on to infer that the speaker’s intentions are in fact directive. Rather, the conventions associated with utterance of declarative syntax (i.e. that the speaker undertakes a commitment to the truth of P and puts P forward for acceptance) result – given the relationship between the interlocutors and the nature of the future act predicated of the hearer – in directive force. The strong content and context constraints that such an interpretation relies on, however, do not apply to the interpretation of the utterance of an imperative sentence as directive, thus allowing the imperative form to be used for a wider class of directive acts.12 A similar explanation can be put forward for the use of future-tense declaratives used to make promises. Like future-tense declaratives with directive force, promises have been seen to pose a problem for accounts, such as Dummett’s, which treat the declarative as a conventional marker of assertoric force. Recanati (1987: 167), for example, is worried that such a view leads to the position that an utterance such as (16) is an indirect promise. The problem with this position, he says, is that it implies that promising is never direct, for there is no commissive mood. (16) I’ll be there But, if indicating assertoric force is understood as consisting in undertaking commitment to the truth of the proposition expressed, we can treat (16) as a direct promise without abandoning the claim that declarative syntax is a marker of assertoric force. Promises and future tense assertions (such as (16) spoken as a statement about the speaker’s intentions) have much in common: both commit the speaker to the truth of the proposition expressed and both present themselves as grounds for the hearer to believe that the state of affairs described will obtain. How they differ is in the type of grounds they provide for this belief. An assertion requires that the speaker, if required to do so by the hearer, give justification beyond the utterance itself. In the case of (16) spoken as an assertion, the speaker might offer as justification the claim that she has decided to go there, (16) being in that case a report of that decision. A promise differs in that the utterance itself is presented as sufficient reason for the hearer to accept that the state of affairs described by the
A Sign of Assertion 73
utterance will obtain: no further reason for acceptance of the proposition expressed can be required. Indeed, the question of acceptance does not arise in the case of a promise. To be sure, we might judge that the speaker is unlikely to keep her promise. But this is quite distinct from rejecting an assertion: an assertion is put forward for acceptance or rejection; a promise is not. The possibility of rejection has an important role to play in the analysis of the activity of assertion, but none to play in the analysis of promising, for how hearers respond to a promise has no bearing on the success of that promise. The reason that the declarative is apt for promising is that it is a form whose use commits the speaker to the truth of the proposition expressed. When that proposition is future tense and relates to a state of affairs over which the speaker has influence, then the question arises of what the speaker is doing: is she undertaking to bring about that state of affairs, or is she, say, predicting that a certain state of affairs will hold? What the hearer needs to determine, in order to resolve the ambiguity, is the justificatory status of the utterance: is it presented as sufficient reason in itself to believe that P or can further reason be demanded? In the former case, the utterance counts as a promise; in the latter case, it is an assertive speech-act. To grant that the declarative is force ambiguous, though, is not to accept that it is not a conventional marker of assertoric force. Indeed, it is the potential of the form to commit the speaker to the proposition expressed that marks it for use in promising.13 Moreover, as is the case with directives, the use of the declarative to make promises is limited to certain contents: the utterance must describe a future situation the bringing about of which the speaker can reasonably be expected to see to. In both the directive and the commissive cases, then, the function of the declarative as a conventional marker of assertoric force can be exploited to explain these other – content- and context-constrained – uses. In both cases, the reason that assertoric force does not result, despite the use of the declarative, is that rejection of the proposition expressed is not presented as an option for the hearer. In the directive case, the speaker’s authority precludes this; in the commissive case, having witnessed the utterance is presented as sufficient grounds for the hearer to adopt the proposition expressed as a belief. 4.4.1.4 Explicit performatives and declarations Whether you regard explicit performatives as a counterexample to the claim that declarative syntax is a conventional marker of assertoric force will depend on whether you view them as assertions. The view that
74
Assertion
utterances such as (17) and (18) are in fact assertions is held by, among others, Bach and Harnish (1979; 1992) and Ginet (1979). On this view, these utterances present no threat to Dummett’s view of the conventional significance of the declarative mood as a marker of assertoric force. (17) I (hereby) accept your apology (18) I (hereby) promise to be there Others, however, reject this view and claim that explicit performatives are not assertions, primarily on the grounds that they cannot be true or false (Austin 1962/1975; Recanati 1987; Searle 1989). I side with this second camp (Jary 2007). Dummett’s position is not entirely clear: he opposes explicit performatives to ‘ordinary assertions’, the former being characterised by their ‘not stat[ing] something as being the case independently of the act of uttering it’ (1981: 334). This suggests that he does see them as a non-ordinary sub-species of assertion. However, explicit performatives fail to meet one of his conditions for assertionhood: they do not require justification. Moreover, they are not put forward for acceptance or rejection by the audience. A felicitous explicit performative such as (17) or (18) results in the proposition expressed becoming a belief of the hearer’s by virtue of his having witnessed it: he can no more reject the proposition that the speaker accepted his apology or promised to be there than he can reject the proposition that the speaker has spoken (Jary 2007: 210–17). Explicit performatives, then, are not assertions by Dummett’s criteria. It does not follow, though, that they undermine his claim that unembedded declarative syntax is a sign of assertoric force. An analysis parallel to that given above for promises is available. Whether uttered as a performative or as an assertion, an explicit performative sentence commits the speaker to the truth of the proposition expressed. However, only if an assertion is made is the option of rejecting that proposition open to the hearer. An explicit-performative utterance of a declarative sentence precludes rejection, for, in her utterance, the speaker gives the hearer perceptual evidence that the state of affairs described by her utterance obtains. As there is no option for rejection in the performative case, no assertion is made. That said, the very fact that we can talk about the utterance describing a state of affairs stems from the assertoric potential of the linguistic form employed. The special circumstances of use (to describe a state of affairs constituted by the utterance itself), precludes that utterance from having assertoric force, but if the form did not have
A Sign of Assertion 75
assertoric potential, then an explicit performative interpretation would not be possible. This is the point made by Dummett in the following quotation: It is essential to the notion of a performative [. . .] that related forms of words, with a different tense, a different subject, or perhaps even with only a different inflection of the verb, should be capable of being used as assertions of the ordinary kind [. . .] The reason [. . .] is that otherwise the phrase ‘what it signifies’ would be without content. Any significant utterance of a sentence is used to effect something, to perform some (conventional) act; but we can without triviality specify what it is that it effects by saying that it is ‘what it signifies’ only if there are other uses of it which are used to signify the performance of the act in the sense of stating that it is performed (Dummett 1981: 334 original emphasis) Thus Dummett’s position on explicit performatives is that cases such as (17) and (18) can be said to serve to effect the state of affairs described only because of the existence of minimally different, non-performative sentences, such as (19) and (20), that can be used to assert that that state of affairs obtains. I think, however, that what is required is not such minimally distinct sentences but an alternative use of the very sentence that is used in performative utterances, such as when a priest, explaining the marriage procedure prior to the ceremony, explains to the couple ‘. . . and then I pronounce you husband and wife’. In this case, all formal features and contextual parameters are the same as in the performative case, except that the utterance does not refer to itself (i.e. there is no implicit ‘hereby’). (19) She accepts your apology. (20) He promises to be there. On this analysis, although explicit performative utterances are not assertions, they do rely on the assertoric potential of declarative syntax for their success: if the same form of words could not be used merely to describe that state of affairs, then it could not be said that they bring about the state of affairs that they describe. Declarations such as (21) also rely on the assertoric potential of the declarative form employed, but, once more, there is no option for rejection. This is because the speaker has the authority to bring about the state of affairs described by the utterance. However, as in the
76
Assertion
performative case, it only makes sense to talk of declarations bringing about the state of affairs they describe if there is another, descriptive use of the same sentence. Thus the analysis of declarations requires that the forms used have assertoric potential, even though declarations themselves are not assertions. (21) This meeting is now closed This analysis of explicit performatives and declarations is strengthened by the fact that the inferences that are licensed by a felicitous declaration or explicit performative utterance are those that are licensed by an assertion with the same propositional content. Thus you are entitled to draw the same conclusions from an assertion of (19) as you are from witnessing the utterance of (17), for example. So, as in the previous cases that we have looked at in this section, the use of the declarative for explicit performatives and declarations can be analysed in terms of the consequences associated with the assertoric use of that form. The difference lies in your grounds for drawing those conclusions: in the case of declarations and explicit performatives, the grounds are that you have witnesses the act; in the case of an assertion, the grounds are that the speaker has provided you, by means of the conventions associated with the use of declarative sentences, with warrant to employ in inference the proposition expressed. Moreover, it is only in the latter case that is it open to you to deny that the state of affairs described obtains, or to require of the speaker further reason to accept the content of her utterance. 4.4.1.5 Conclusion to section 4.4.1 What unites the non-assertoric uses of the declarative discussed in sections 4.4.1.3 and 4.4.1.4 is that the successful performance of each act has the same inferential and practical consequences for its audience that an acceptance of an assertion expressing the same proposition would have. This explains the aptness of the declarative for the performance of these acts. The reason that they do not count as assertions is that they are not put forward for acceptance or rejection by the hearer. The directive cases exploit the authority the speaker has over the hearer such that rejection is not an option for him. Promises present the act itself as sufficient grounds for acceptance. Similarly, explicit performatives and declarations rely on the act in question being witnessed for their effect. Because hearer acceptance is not an issue in any of these cases, they
A Sign of Assertion 77
are not the sort of thing that can be judged true or false, because the function of such judgements is to signal acceptance. The discourse-embedding and fiction cases discussed in section 4.4.1.2, by contrast, do put themselves forward for acceptance. It is this they have in common with assertion and it is this which motivates the use of the declarative in such cases. To be acceptance-apt is to be treatable as true. To treat as true is to accept, under a certain mode of commitment, the consequences of assertion. It should be no surprise, then, that acceptance apt-speech acts employ the declarative if this is the mood that is specified for the making of assertions. Thus Dummett’s Declarative Thesis is not undermined by the alleged counterexamples that we have discussed in this section. It is possible to hold that the declarative sentence-type is an indicator of assertoric force and still allow that it has other uses which to do not result in an utterance being classified as an assertion. These other uses, moreover, can be analysed in terms of the consequences that result from the assertoric use of the declarative. Their existence therefore strengthens, rather than weakens, the claim that the declarative a default marker of assertoric force.
4.4.2 Challenges to the Assertion Thesis The cases that we will discuss in this section are those in which assertions are made by means other than the unembedded use of declarative sentences. These cases do not threaten the Declarative Thesis: the claim that there is a conventional means of doing something does not exclude the possibility of it being done by unconventional means. On one reading of it, though, the Assertion Thesis – which states that assertion is to be explained in terms of conventions associated with the use of declarative sentences – is threatened by some of these cases. However, as we will see in this section, this is not the reading intended by Dummett. There are two main types of case that threaten the Assertion Thesis: sub-sentential assertions and alleged cases of assertions performed using non-declarative sentences. The first class have been discussed extensively by Stainton (1993; 1997; 2005; 2006), who gives a number of examples of phrases used in isolation to make assertions. An indicative list, taken from Stainton (1997: 61), is given below:
(22) (Two people are talking at a party. One points to a man near the door and says) John’s father.
78
Assertion
(23) (A student is receiving instruction in painting. The teacher looks at the current canvas and says) Nice work. (24) (A boat speeds by. A spectator says) Very fast. (25) (A letter arrives. The recipient looks at the envelope and says) From Spain. The crucial point about these cases is that the speaker may lie or mislead using these phrases, indicating both that assertoric commitment is being undertaken by the speaker and that a proposition is being put forward for acceptance by the hearer. It is seems clear, then, that an assertion is being made in each of these cases. Examples of the second class come from Davidson (1979/2001: 110), who says that utterances such as (26) and (27) may be used to assert that Joan is wearing her purple hat again: (26) Did you notice that Joan is wearing her purple hat again? (27) Notice that Joan is wearing her purple hat again. Dummett has two responses to these cases. The first relies on a distinction he makes between the force of the utterance and the speaker’s point in making the utterance, and is itself enough to deal with Davidson’s alleged counterexamples. This distinction, though, will not deal with the sub-sentential cases, as Stainton notes (1997: 61). In order to deal with these, we have to invoke Dummett’s deeper claim that the point of positing a conventional relationship between declarative syntax and assertion is required not to determine whether a given utterance counts as an assertion, but to make sense of the notion of assertion in the first place. We deal with each response in the next two sections. 4.4.2.1 The tactical role of mood and force In response to Davidson’s cases, Dummett grants that the point of these utterances might well be to alert the hearer to the fact that Joan is wearing her purple hat again, but says that to accept this is not to accept that (26), say, is not a question but an assertion. That (26) is a question, he notes, is clear from the fact that it calls for answer. Viewing (26) as an assertion, he says, conflates force and speaker’s point. Force is best seen as a tactical notion: a means by which speaker’s achieve their strategic aims. It relates not to the speaker’s long-term aim, ‘but to the form of linguistic or non-linguistic response, if any, which is called for, and the commitment, if any, undertaken by the speaker’ (Dummett 1993: 223). On this view of force, Davidson’s cases are not counterexamples to the
A Sign of Assertion 79
view that assertion is to be analysed in terms of conventions associated with declarative and the Assertion Thesis survives. 4.4.2.2 The role of convention in understanding assertion Dummett does not consider in any great detail sub-sentential assertions of the type identified by Stainton. He makes the brief comment that an utterance of ‘the highest mountain in the world’ uttered out of the blue, has sense and reference but no force: the speaker cannot be said to have done anything right or wrong. Spoken in response to a question such as ‘What is Mount Everest?’, though, the utterance has a force and can be counted as an assertion, he says (Dummett 1981: 297–8). Stainton’s claim, though, is that even without prior discourse to indicate its force, a sub-sentential utterance can serve as an assertion. The examples he gives, listed above, are designed to show how non-linguistic context alone is enough to imbue an utterance of a sentence fragment with assertoric force. These cases, then, might look to be counterexamples to Dummett’s Assertion Thesis, i.e. that assertion must be explained in terms of conventions associated with the use of declarative sentences. What I want to argue, though, is that this conclusion is based on a flawed interpretation of the Assertion Thesis. The mistake is to view this thesis entailing what we can call ‘Not-Dummett’:14 Not-Dummett: In the absence of a discourse context (such as a question-answer pair) that makes assertoric force otherwise apparent, assertoric force must always be signalled by the use of declarative syntax. Not only does the Assertion Thesis not entail Not-Dummett, Dummett does not hold Not-Dummett, as he makes very clear in Mood, force and convention. In that paper, Dummett (1993: 202) poses two questions about force. I have rephrased them below as questions specifically about assertion. Question 1: What is it to attach assertoric force to an utterance? Question 2: Under what conditions is it correct to say that someone has made an utterance with assertoric force? Question 1 asks what it is to make an assertion. It requires an explanation of what assertion consists in. Question 2 asks how one is to recognise an utterance as conveying assertoric force. Now, in order to take Stainton’s examples as evidence against Dummett’s position on the
80
Assertion
relationship between assertion and the conventional use of declarative syntax, we must read Dummett as proposing this relationship as a means of answering Question 2. That is, we must take him as meaning that it is correct to say that someone has asserted only if they have uttered a declarative sentence. This would indeed commit him to Not-Dummett. However, this is not what Dummett means. Rather, the conventional relationship between mood and force that he proposes is aimed at answering Question 1: his view is not that utterances cannot be assertoric unless they involve the unembedded use of declarative syntax, but that unless we posit a conventional relationship between assertion and declarative syntax, we have no means of explicating assertion. In other words, his Assertion Thesis is not intended as an answer to Question 2, but to Question 1: The conventions to which I intended to allude are not those which determine whether, for example, an utterance is to be taken as an assertion, but those without which there would be no such thing as the assertoric use of sentences, those by reference to which an account is to be given of what it is to make an assertion (Dummett 1993: 221). On this view, we need to posit a conventional relationship between declarative syntax and assertion in order to say what assertion is, but once we have this notion, we can identify assertions as acts performed with the intention to assert, regardless of whether this intention is signalled with declarative syntax. To show that Dummett is wrong, then, it is not enough to point to assertions such as (22) to (25): one must also show that it is not necessary to posit a conventional relationship between assertion and declarative syntax in order to explain what assertion is. To do this, one might argue, à la Grice, that assertion is best explained in terms of the expression of belief. But Dummett, as we saw in Chapter 3, holds the view that assertion is conceptually prior to belief, and has strong arguments to this effect. Consequently, to argue against him, it would be necessary to show that his views concerning this order of explanation are wrong. Another route suggests itself. One might want to characterise assertion in terms of its relationship to truth, rather than belief, but without positing a conventional link between a particular linguistic form and truth. One might want to claim, for example, that to assert is to speak with the intention of saying something true. This though, requires that truth be explicated. As we saw in Chapter 3, for Dummett, truth is to be
A Sign of Assertion 81
explicated in terms of treating true. We must then ask what practice best exemplifies treating as true, assertion being, for Dummett, the answer to that question. Circularity results: in attempting explain assertion, we make use of the notion of truth which in turn, for its explication, requires recourse to the notion of assertion. Again, Dummett forces any prospective opponent to tackle him on the nature of truth, rather than on the seemingly firmer ground of the mood/force relationship. None of the above should be taken to imply, however, that Dummett rules out intention attribution playing a role in the identification of assertoric force on a particular occasion of utterance. His claim is, rather, that having an intention to assert presupposes a notion of assertion, and it is this prior notion that must be explained in terms of convention, not intention. He draws a parallel with the use of ambiguous words such as ‘slip’: it may be necessary to take into account speaker intentions in order to determine whether, in uttering ‘I didn’t notice the slip’, the speaker meant an error or a garment, he says, but the possible meanings of the ambiguous word are determined by convention, not by intention (Dummett 1993: 215). In short, then, the existence of sub-sentential assertions does not undermine Dummett’s Assertion Thesis. Dummett’s point about the relationship between assertion and declarative mood is a deep one concerning the role of treating as true in explicating truth. To challenge it, this deep point must be targeted. Alleged counterexamples of the type considered in this section will not do the job.
4.4.3 Conclusion to section 4.4 In this section, I have considered, on the one hand, whether the declarative should be thought of as a marker of declarative force, and, on the other, whether an affirmative answer to this question means that assertions must always be performed using the declarative form. The answer has been that the declarative should be thought of as a sign of assertion, but that this does not entail the view that assertion cannot be performed by other means. It might still be asked, though, whether, given the cases discussed in sub-section 4.4.1, assertion can be fully explained in terms of the conventions associated with the use of the declarative mood, as Dummett holds. The objection would centre on the fact that where there is ambiguity in the use of the declarative, as in the cases discussed in that sub-section, this ambiguity is resolved by considering speaker intentions: what determines, for example, whether an utterance of ‘I promise never to drink again’ counts as a promise or as
82
Assertion
a statement concerning the speaker’s habitual response to a hangover is the speaker’s intention, not a convention. This would be taken to show that there is an essential intentional ingredient to assertion, and that convention cannot be the whole story. I think, though, that the cases discussed in sub-section 4.4.1 do not undermine the conventionalist analysis. It is open to the conventionalist to respond, I think, that those cases are marked by particular context and/or content factors. For example, on the context side, we noted how the use of the declarative in directive speech acts required the speaker to have authority over the hearer; on the content side, explicit performatives are characterised by the fact that they are about an illocutionary act. Thus, it is only under certain conditions, the conventionalist can argue, that speaker intentions come into consideration. Moreover, the issue of speaker intentions only arises, it can be claimed, due to the conventional significance of the form employed as a marker of assertoric force. On this issue, perhaps the parallel with lexical ambiguity (as in Dummett’s ‘slip’ example) is misleading: whereas ‘slip’ has (at least) two conventional meanings, the declarative is a conventional marker only of assertoric force. The potential that it has, under certain conditions, to convey other forces results not from conventions relating the declarative to these forces, but from its conventional role as a means of conveying assertoric force and further context and content considerations. Furthermore, against the intentionalist analysis, it could be argued that one can assert without intending to. Imagine, for example, a speaker who makes a sexist statement with ironic intentions, but whose audience has little appreciation of – or sensitivity to – irony. She would be taken by her audience to have asserted, and might even agree herself that she had unwittingly asserted the offensive proposition. Overall, then, I think that the view that the concept of assertion can ultimately be analysed in terms of the conventions associated with the utterance of declarative sentences can be maintained, though some nuance must be introduced in order to handle the cases discussed section 4.4.1.
4.5 Conclusion A number of significant points can be drawn from the preceding discussion. First, if an inferential system is intended as a means of representing the world, then a sign or format that imbues a propositional form with assertoric character must be specified. Failure to do so will result in infinite regress. Second, from the fact that assertions qua speech acts can present propositions as suitable for use as premises in inference, it
A Sign of Assertion 83
does not follow that assertions qua speech acts can serve as premises. Moreover, there is good reason to think that they cannot: assertions qua premises in inference must encode their truth conditions, something that assertion qua illocutionary acts rarely do. Third, we have seen that non-assertoric uses of the declarative that have been put forward as evidence that this form is not a marker of assertoric force can be straightforwardly explained on the assumption that this is indeed its role, even while allowing that assertions can be performed by other means. Generally, the proposed counter-examples can be explained in terms of their relationship to an assertion made using the same form: in some cases, witnessing the act performed has the same consequences as accepting an assertion employing the same form would have; in others, the consequences of the act are an acceptance, albeit in a restricted domain, of the inferential consequences of the associated assertion. In Chapter 6, we will look more closely at the role of mood in assertion, considering first just what the declarative encodes that makes it so suitable for the making of assertions. Before that, though, we need to consider one final perspective on assertion: Stalnaker’s common ground view. We look at this now because it has been very influential on some of the linguistic accounts of the declarative that we will consider in Chapter 6.
5 Assertion and Common Ground
5.1 Introduction Linguistic communication generally takes place against a background of assumptions held in common by those participating in the conversation. Call this the common ground. These assumptions represent the state of the world as it is taken to be by those participants. Depending on the interlocutors and the topic of discussion, these common assumptions might range from facts concerning who is participating in the conversation to shared beliefs about quantum physics. Assertion can be viewed as a means of adding to that stock of shared assumptions: when one asserts that P, one puts P forward for inclusion in the common ground, and P will acquire common-ground status unless it is rejected by the other participants in the conversation. On this view of assertion, what one is entitled to assert is constrained by the common ground: one should normally only assert a proposition that is compatible with, but not entailed by, the common ground. To do otherwise would either be contradictory (in the case that the proposition asserted were incompatible with the common ground, i.e. with propositions accepted as true by the speaker), or uninformative. A much-discussed model of assertion along these lines has been developed by Robert Stalnaker (1975/1999 inter alia). As noted in the introduction, this is the philosophical analysis of assertion that has most influence on formal-semantic approaches to mood. It is thus discussed here with a view to the subsequent discussion of these analyses of mood, and serves, to some degree, as a bridge from the philosophical concerns of the previous chapters to the linguistic and cognitive topics that will dominate during the remainder of the book. 84
Assertion and Common Ground
85
For Stalnaker, a proposition is a function from possible worlds into truth values. The set of propositions that make up the common ground will thus pick out a set of possible worlds consisting of all those worlds in which all the propositions in the common ground return a positive truth value. These possible worlds will be the live options for how the world might actually be. Stalnaker calls this set of worlds ‘the context set’. An assertion, if accepted by the addressee, adds the proposition expressed to the common ground. This proposition will return a positive truth value in some, but not all, of the possible worlds in the context set. Those worlds in which it returns a negative truth value will be removed from the context set, and so some of the candidate worlds for the actual world will have been eliminated. On this view, a central aim of conversation is to reduce uncertainty about the way the world is. Reducing the context set by means of assertion achieves this aim, and it is in these terms that Stalnaker seeks to capture the informative nature of assertion.1 A major concern of Stalnaker’s in developing this view of assertion is to provide a pragmatic account of presupposition phenomena. In linguistics, it is this aspect of his work that has had the most impact, particularly among formal semanticists. Stalnaker’s account of presupposition is discussed in section 5.3 of this chapter. From a speech-act theoretic perspective, Stalnaker’s account has received considerably less attention. However, as we will see in the section 5.2, his account does reflect the traditional speech-act distinction between sense and force in the dual way it employs the notion of truth: both to capture the content of assertions and the consequences of making them. Another feature of Stalnaker’s account that has received little attention is the features it shares with Sperber and Wilson’s Relevance Theory. Although they are perhaps not immediately obvious, we will see in section 5.4 that there are significant points of contact. I will argue there that some of the challenges faced by Stalnaker’s account, particularly the problem of logical omniscience that results from his possible-worlds framework, can be addressed by adopting elements of Sperber and Wilson’s framework.
5.2 Sense, force and representation in Stalnaker’s account In this section, we consider Stalnaker’s account from the perspective of speech-act theory. We will see that assertoric force is explained, on the common-ground account, in terms of a consistency requirement
86
Assertion
on the basic context set. Other speech acts can be distinguished from assertion both in terms of the type of context set they seek to modify and whether they are best analysed as attempts to reduce a context set, basic or otherwise. This discussion brings to light two conceptions of truth at play in Stalnaker’s account, and this is in turn related to two conceptions of representation.
5.2.1 Consistency and assertoric force As we noted in the previous chapter, to make an assertion is to undertake a commitment to consistency. This is captured in Stalnaker’s account by characterising the context set as the set of worlds in which all the propositions in the common ground are true in each world in that set. It is important to stress that this consistency requirement does not follow automatically from the fact that an assertion expresses a proposition. A set of propositions, thought of as functions from possible worlds to truth values, will also pick out the set of possible worlds consisting in all those worlds in which at least one proposition in that set is true. This, though, will not be a consistent set, for propositions true in some worlds in the set will be false in others. In characterising the context set as the set of worlds in which all propositions in the common ground are true in all members of that set, Stalnaker requires that this set be consistent and thereby reflects the fact that assertion is a truth-aiming activity. Notice, though, that truth plays two very different roles here. On the one hand, we have truth employed as a means of characterising propositions. A proposition is thought of as means of partitioning sets of possible worlds according to whether or not the proposition is true in each of the worlds in that set. On the other hand, we have truth as an aim of assertion. This is truth thought of in terms of treating as true. The assertions we can make are constrained, on Stalnaker’s account, by their compatibility with the propositions already in the common ground, and if an assertion is accepted, it will constrain what further assertions can be accepted. This latter – pragmatic – conception of truth draws on the significance of the everyday notion of truth employed in human affairs. By contrast, the conception of truth employed in characterising propositions is highly abstract and not open to analysis in terms of the consequences of making a truth judgement. The distinction between truth as applied to propositions and truth as applied to assertions calls to mind Dummett’s criticisms of Davidson’s disquotational theory of meaning, discussed in Chapter 3 section 3.2. Dummett’s point, recall, is that what an approach such as Davidson’s
Assertion and Common Ground
87
describes is the ability of a language user to distinguish sentences that should be called ‘true’. Such accounts take for granted, he argues, the point, or the consequences, of such a distinction. Stalnaker’s account seeks to capture the point of this distinction – that is, the point of truthjudgements – in terms of how they reduce uncertainty about the nature of the world by limiting the range of permissible assertions. This is why the context set must be one in which all propositions in the common ground are true in all the possible worlds in that set, for only through the requirement of consistency can the informative nature of assertions (and hence their point) be captured. The two conceptions of truth at play in Stalnaker’s account also relate to the traditional speech-act distinction between sense and force (as reflected in the F(P) notation discussed in Chapter 2). One could just as easily use the notion partitioning possible worlds – and thus the propositional notion of truth – in an analysis of what it is to understand an utterance of an imperative: you understand an utterance of an imperative sentence if you can distinguish all the worlds in which that utterance is complied with from all those in which it is not. Utterances of different imperative sentences will partition possible worlds in different ways. If the worlds selected as the compliance worlds of an imperative are the same as those worlds for which an utterance of declarative sentence selects returns the value ‘true’, then it is surely fair to say that the two utterances express the same proposition, that they have the same sense. But, as Dummett counsels, we must say more if we are to capture the difference between utterances of imperative and declarative sentences. We need to say what their point is – what they are used for. This, of course, is where force comes in, and the effect of reducing – while maintaining the consistency of – the context set is Stalnaker’s means of adding assertoric force to the propositional element of assertions. The point of an utterance of an imperative is surely not to remove all non-compliance worlds from the common ground. Imperatives are not informative in this sense. Imperatival – or ‘directive’ – force will have to be explained by other means.2 So while it might be correctly said that propositional content, and therefore sense, determines a sentence’s truth value at any given possible world, it is force that determines aptness for truth-judgement. Any sentential utterance can be said to express a proposition, but only assertives (and perhaps some rhetorical questions, see Barker 2004: 37; Pagin 2004: 851) are truth apt. As we saw in Chapter 2, the significance of the consequence of truth judgements for semantic theory is a major fault line amongst theorists. Those who think the analysis of meaning
88
Assertion
boils down to the analysis of truth conditions will be content with the abstract notion of truth as a value of propositions. Taking this stance does postpone having to consider questions of use and force, but it also, according to Dummett (1993), limits such theories to accounting only for sense, rather than for linguistic meaning in its totality. Employing a notion of truth that gives credence to the significance of truth judgements, however, requires giving explanatory priority to assertions, à la Brandom and Barker, for it is assertions which are the primary objects of truth judgements. In certain respects, through his dual use of possible worlds, Stalnaker straddles the two camps. On the one hand, he employs possible worlds to characterise the content of assertions. On the other, he employs possible worlds to characterise the information state of the participants in the conversation. The latter move enables him to say something about the informative purpose of assertion in terms of removing possible worlds that are candidates for the actual world from the context set. What enables him to do this, though, is the consistency requirement on the context set. Without this requirement, adding propositions to the common ground would extend, rather than reduce, the context set. 5.2.2 The problem of representation The two ways in which the notion of truth is employed in Stalnaker’s account also manifests itself in a dual use of the notion of representation. Discussing the relationship between propositions and representation, Stalnaker says that: A proposition – the content of an assertion or a belief – is a representation of the world as being a certain way. But for any given representation of the world as being a certain way, there will be a set of all the possible states of the world which accord with the representation – which are that way. So any proposition determines a set of possible worlds. And, for any given set of possible worlds, to locate the actual world in that set is to represent the world as being a certain way (1978/1999: 79 original emphasis). Discussing accounts, such as Stalnaker’s, which analyse assertions and beliefs as expressions of propositions, in turn understood as sets of possible worlds, Brandom comments: The question that then arises is what expressing a proposition has to do with representing anything (1994: 665, fn. 32 original emphasis).
Assertion and Common Ground
89
Brandom’s complaint against possible-world accounts of representation centres on the alleged inability of such theories to explain the significance of such representation in human affairs (echoes again of Dummett’s complaint against Davidson). This significance is, for him, the key to understanding the nature of linguistic representation. In this sub-section, we look at Brandom’s complaint and ask whether Stalnaker has any response available to him. Implicit in the above quote from Stalnaker is the view that (a) the content of an assertion or belief can be abstracted away from other elements of this act or state; and (b) such abstraction delivers up the representational core of that assertion or belief. This is evident in the fact that Stalnaker says that it is the proposition expressed by the assertion or belief, rather than the assertion or belief itself, which is ‘a representation of the world as being a certain way’. Note, though, that in the above quotation, the notion of representation is employed twice: first, a proposition is characterised as a representation of the world as being a certain way; then, locating the actual world in a set of propositions is described as a means of representing the world as being a certain way. Locating the actual world in a set of possible worlds is what one does, on Stalnaker’s account, when one asserts. So, just as we found two notions of truth at play in Stalnaker’s model, here we find two notions of representation. On the one hand, propositions represent the world as being a certain way; on the other, speakers represent the world as being a certain way when they assert. Moreover, these two notions are closely related to those two conceptions of truth: a proposition represents the world as being a certain way by dividing a set of possible worlds into those in which it is true and those in which it is not; a speaker and hearer coincide in representing the world as being a certain way if she makes an assertion and it becomes mutually accepted that he judges it true. The abstract notion of truth employed in the analysis of propositions is thus related to the notion of representation at play when isolating the contents of assertions and beliefs, while the practical notion of a truth judgement employed in Stalnaker’s analysis of assertions is related to acts of representing the actual world as being a particular way by locating the world in a set of propositions. Just as the two notions of truth at play in Stalnaker’s account correspond to the distinction between sense and force, so do the two notions of representation. A proposition simpliciter is the sense of a representation, but representation of the world as being a particular way – in the sense of locating the actual world in a set of possible worlds – is achieved only when a representation is employed with assertoric force.3
90
Assertion
Brandom’s objection to Stalnaker’s characterisation of the representational capacity of assertions and beliefs is aimed at the first of these notions of representation, i.e. the idea that a proposition, conceived of as a set of possible worlds, can capture the representational nature of these acts and states. For Brandom, an adequate analysis of representation must make reference to the use to which representations are put and what counts as making proper use of them. That is, representation only makes sense in relation to an individual or system whose behaviour will be guided by the ‘consumption’ of those representations (Brandom 1994: 72–4). This is not a view that is unique to Brandom: essentially the same position is held by a number of authors who have worked on representation. For example, Sperber says that ‘[a] representation involves a relationship between three terms: an object is representation of something, for some information-processing device’ (1996: 61, original emphasis), while Dennett (whom Brandom cites) says that ‘nothing is intrinsically a representation of anything; something is a representation only for or to someone’ (1981: 122). On this view of representation, abstracting representations such as assertions and beliefs to sets of possible worlds removes elements from these acts and states that make them representational in the first place. This, I think, is the crux of Brandom’s complaint: if we analyse representation in terms of sets of possible worlds, then we lose our hold on its significance in human activity. But this significance, Brandom would argue, is just what gives our assertions and beliefs representational content. It is only because they constrain and guide our behaviour (i.e. they are representations for us) that we have grounds to call them representations. However, the notion of representation that is employed in the latter part of the above quote from Stalnaker, whereby we represent the world as being a certain way by locating it in a set of possible worlds, is not open to this criticism. To locate, by means of assertion, the actual world in a set of possible worlds is to undertake a commitment to consistency, so that neither the speaker nor those who accept her assertion can go on to make another assertion that is incompatible with what has been asserted. As we noted in section 5.2.1, this commitment does constrain behaviour, so representation thus construed does meet Brandom’s requirement that representation be explained in terms of uptake, of ‘taking, treating or using it in practice as a representation’ (Brandom 1994: 74, original emphasis). Brandom is right, then, to worry that viewing propositions as sets of possible worlds raises the question of how they ever come to represent
Assertion and Common Ground
91
anything. But Stalnaker’s use of the possible worlds framework in analysing assertion also offers a means by which to explain how assertions manage to constrain our behaviour and hence of how they come to represent the world. Assertions, on his account, come with a commitment to consistency and this commitment constrains the behaviour of those who make, and accept, assertions. That said, a commitment to consistency alone will not explain why assertions are representations of the actual world. As was noted in the discussion of fiction and supposition in Chapter 4, not all consistency-aiming endeavours aim at representing the actual world. Other means are needed, then, in order to distinguish assertion as a species of representation. How this might be done is the topic of the next sub-section.
5.2.3 Distinguishing assertions from other acts aimed at reducing a context set Stalnaker stresses that in characterising assertion in terms of an attempt to reduce the context set, he does not mean to give a definition of assertion, but only to make ‘a claim about one effect which assertions have, and are intended to have – an effect that should be a component, or a consequence, of an adequate definition’ (1978/1999: 87). One of his reasons for not claiming that putting forward for context-set reduction is a sufficient condition for an act to count as an assertion is that there are, he says, other acts that have an are intended to have the same effect (loc. cit.). The act he mentions in this light in Assertion is the making of a supposition. However, it is likely that he had in mind a whole range of acts that are related to acceptance-apt attitudes. As noted in the last chapter, Stalnaker defines acceptance as treating a proposition as true in one way or another – as ignoring, perhaps only temporarily, the possibility that it is false (1984: 79). As well as supposing, he mentions presupposing, presuming, postulating, and assuming. We will leave presupposing until the next section as it is distinct from the others in that it cannot be used in an explicit performative (i.e. one cannot say ‘I hereby presuppose that P’ and thereby presuppose that P), suggesting that, unlike the others, it is not an illocutionary act. What we will see is that the remaining acceptance-apt acts can in fact be distinguished from assertion in terms of the type of context set that they act upon. In his work on belief attribution and conditionals, Stalnaker (1975/1999; 1988/1999) distinguishes between a basic context and a derived context. A derived, or embedded, context is one at which an
92
Assertion
embedded sentence such as we find in a belief report should be interpreted. The set of possible worlds that make up the basic context may be distinct from those that make up the embedded context: the set of worlds that speaker believes to be the live options for the actual world may be distinct from those that she takes the subject of the belief report to be treating as candidate worlds. Similarly, subjunctive (i.e. counterfactual) conditionals create a context which is usually disjoint from the basic context in which the conversation is taking place, hence the implication that follows from a subjunctive conditional that the negation of the antecedent is true. A fundamental difference between basic and derived contexts is that in the former, but not in the latter, it must be true that the participants of the speech event exist and that the speech act in question takes place. This need not be the case in derived contexts. One can say, for example, ‘Let’s suppose we weren’t here’ (Stalnaker 1988/1999: 156–9). This is not to say, though, that all derived contexts must be disjoint from the basic context. Presuming is an attitude that involves accepting a proposition as true while also accepting that one does not have knowledge of its truth. It differs from supposition in that one cannot presume what one knows to be false. The set of worlds (i.e. derived context set) picked out by acceptance of a presumption, then, will be all those worlds in the basic context set less those in which the proposition presumed is also true. Thus while, on the Stalnakerian account, assertion cannot be distinguished from other acceptance-apt acts in terms of the type of effect it has on a context set, it can be distinguished in terms of the type of context set it aims at affecting. Assertions are aimed at modifying the basic context; other acceptance-apt acts are aimed at modifying a derived context that is embedded within the basic context. This derived context set may be, but need not be, disjoint from the basic context set. Stalnaker’s conception of a derived context set therefore offers a means of distinguishing – within his general framework – assertion from other acceptance-apt speech acts, and belief from other acceptance-apt propositional attitudes. This distinction between basic and embedded contexts will be central to the analysis of assertion put forward in Chapter 8.
5.3 Assertion and presupposition As noted earlier, Stalnaker’s interest in assertion is motivated to a large extent by a desire to put forward a pragmatic account of what are generally termed presuppositional phenomena. Indeed, in his earliest
Assertion and Common Ground
93
writing on this topic, what he would later term ‘the context set’ is called ‘the presupposition set’ (Stalnaker 1973: 450). In this section, we will look closely at the assertion/presupposition contrast in Stalnaker’s account, concentrating in particular on the issue of presupposition accommodation. We will see that, while Stalnaker draws a useful distinction between those propositions that can be inferred from the fact that an assertion has been made and the proposition that the assertion has the overt goal of communicating, the identification of presuppositions as common-ground propositions is highly problematic. 5.3.1 Presupposition: the Stalnakerian view It has long been felt, among linguists and philosophers of language, that there is a class of implications that follow from certain statements which behaves differently to straightforward entailments. For example, (1)a entails both (1)c and (1)d. However, there appears to be a difference in the way the two entailments respond to the negation of (1)a. If we negate (1)a we get (1)b. A speaker who uttered (1)b, however, would not, it is widely held, generally be committed to (1)c, though she would normally be committed to (1)d. (1) a. b. c. d.
John gave his wife roses John didn’t give his wife roses John gave his wife flowers John has a wife
This illustrates, of course, the widely-drawn distinction between entailment and presupposition: (1)a entails (1)c but is said to presuppose (1)d. Generally, a statement S is said to presuppose P if P can usually be inferred both from S and its negation. Other examples of presuppositions are given in the pairs below. In each case, the b-member of the pair is an example of what is commonly considered a presupposition of the a-member.4 (2) a. b. (3) a. b. (4) a. b.
Peter hasn’t quit smoking Peter smokes Perhaps John realises that the exam will be difficult The exam will be difficult Peter wasn’t sad to see her leave She left
94
Assertion
(5) a. b. (6) a. b.
Bill might not know that the bar is shut The bar is shut It was Bill who wrote the book Someone wrote the book
As the name suggests, there is a widespread feeling that presuppositions should normally be part of the common ground of the conversation in which an utterance takes place. Stalnaker seeks to exploit this in an attempt to subvert the prevalent view that presuppositions, like entailments, are part of what is encoded by a sentence.5 Taking his cue from Grice’s conversational analyses of the apparent ambiguities of natural language conjunctions such as ‘or’, Stalnaker attempts to explain presuppositions in terms of expectations that hearers and speakers could reasonably have about each others behaviour, granted their rationality and common goals. Presupposition, in Stalnaker’s framework, is a propositional attitude of speakers: to treat a proposition as common ground is to presuppose it. Certain linguistic structures, on his account, should be used only if certain propositions are already part of the common ground. For example, a definite description should only be used if it is common ground that the referent exists, thus explaining why both (1)a and b presuppose (1)d. Change of state verbs such as ‘quit’ are to be used only if the state of affairs that is described as undergoing change obtains. Similarly, the Stalnakerian view would have it that factive predicates such as ‘know’, ‘realize’ and ‘be sad that’ should normally be used only if it is common ground that the proposition expressed by the complement clause is true, and cleft sentences such as (6)a only if it is common ground that the event described by (6)b has taken place. The justification for these stipulations is rational conversational efficiency. Take the factive verb ‘know’ as an example. As we have seen, Stalnaker claims that, generally, when one asserts that someone knows that P, one treats it as common ground that P. This can be explained, he says, by considering what would happen if one were to assert someone knows that P against a conversational background in which P were in doubt or dispute. In such a case, one would be making a statement that could be challenged in two ways: either by denying that the individual in question knows that P or by denying that P. Consequently, the point of the assertion would be obscured. In the same situation, to assert that someone does not know that P would be gratuitously weak. For in the case that not-P, someone’s knowing that P is already ruled out (Stalnaker 1974/1999: 55).
Assertion and Common Ground
95
On this account, then, assertion is an attempt to have the proposition expressed added to the presuppositions of the participants in a conversation. The presuppositions associated with utterances of declarative sentences are distinguished from the proposition expressed by those sentences in terms of their relationship with the common ground: the former are, under ideal conditions, already part of the common ground while the latter are not. However, there is a fly in the ointment. As Stalnaker has always been aware, linguistic forms are often not used in the way that his account, on a simple reading, predicts. For example, (1)a could be uttered quite felicitously in conversation in which (1)d was not common ground. However, if (1)a were accepted (and possibly even if it were not) then (1)d would also be common ground as a result of the utterance. It would appear that a definite description can be used – without giving rise to intuitions of infelicity or inappropriateness – to communicate that the referent exists. This creates a problem for Stalnaker, as his basis for distinguishing between assertion and presupposition appears to be undermined by such cases: both presuppostions and assertions can serve to extend the common ground. As just noted, Stalnaker was aware of this problem from early on (see Stalnaker 1974/1999: 51–53, especially fn. 2). He attempts to deal with it by appealing to Lewis’s (1979) notion of accommodation. The idea is this: when a linguistic form that is generally only appropriately used if P is common ground is employed in a conversation in which P is not common ground, then the participants of the conversation tend to accommodate that use by quietly adding P to the common ground. This will happen if the proposition is uncontroversial and it is common ground that the speaker is in position to know that P. Stalnaker implements this idea in an ingenious manner by distinguishing two the ways an assertion can modify the common ground. On the one hand, as we have seen, an assertion, if accepted, will modify the common ground by having its content added to that set of propositions, thereby reducing the context set. This Stalnaker calls ‘the essential effect’ of assertion. It is essential not because an act must have this effect if it is to count as an assertion, but because it grants an insight into the nature of assertion, in that this effect should be, according to Stalnaker, a consequence of any satisfactory definition of that act. However, an assertion will also affect the common ground in manner not essential to assertion but common to all speech acts: the fact that the speech act has taken place will become common ground. Thus if somebody asserts, promises or commands that P, the fact that she has asserted, promised or commanded that P will become common ground. In the case of assertion,
96
Assertion
this effect on the common ground differs from its essential effect in that it is automatic and unavoidable: it is not dependent on acceptance by the addressee.6 This is because the production of the speech act is ‘a manifest event’ (Stalnaker 2002: 708), one which has been mutually witnessed by all participants and hence cannot be rejected. Stalnaker introduces this distinction between the essential and nonessential effects of assertion in Assertion (1978/1999: 86), but it is in two later papers (1988/1999; 2002) that he details how, he claims, it can be used to account for presupposition accommodation while still maintaining the distinction he wishes to draw between assertion and presupposition in terms of the latter’s alleged common-ground status. What he says is that the distinction between the two effects makes it possible to idealise a moment between comprehension and acceptance. At that point, it will be common ground that the speaker has asserted that P, but P will not yet have been accepted into the common ground. Given that the speaker has asserted that P, it will be possible for the hearer to infer that the speaker assumes that certain propositions are common ground from the fact that the she has used linguistic forms – such as definite descriptions and factive verbs – which, according to Stalnaker, considerations of communicative efficiency decree should only be used if a certain proposition is common ground. The hearer can then add to the common ground any propositions that are not in fact presupposed before accepting (or rejecting) the asserted proposition. It is the common ground at this stage of the interaction – once the utterance has been comprehended and any presuppositions made by the speaker that were not already common ground have been incorporated into the common ground – that an assertion has the essential effect of updating. As the quote below shows, although elaborated on in later papers, this aspect of his model was present from quite early on. However, although ingenious, it is open to criticism, as we will see in the next sub-section. [T]he context on which an assertion has its essential effect is not defined by what is presupposed before the speaker begins to speak, but will include any information which the speaker assumes his audience can infer from the performance of the speech act (Stalnaker 1978/1999: 86). 5.3.2 Criticising the Stalnakerian view of presupposition accommodation One criticism that can be levelled against the Stalnakerian notion of presupposition accommodation is that, as Abbott (2000) and Simons (2007)
Assertion and Common Ground
97
have shown, the putative phenomenon of presupposition accommodation is too widespread to be seriously considered an exploitation of a norm such that certain constructions should only be used if the proposition whose acceptance is signalled by that use is part of the common ground. We will not go into this objection here, because the data will be discussed in Chapter 7 at some length. The problem that we will consider here lies in Stalnaker’s grounds for claiming that presuppostions are best explained as common-ground assumptions. His reasons for adopting this position are, recall, that he believes it would go against norms of rational communicative efficiency to put forward more than one proposition per utterance for inclusion in the common ground. For to do so, he says, would leave the audience in doubt about the speaker’s main point, ‘thus leaving unclear what direction he wanted or expected the conversation to take’ (1974/1999: 55). This assumes, though, that there are no means available to simultaneously both put a set of propositions forward for inclusion in the common ground and indicate which of these is the speaker’s main point. In other words, Stalnaker assumes that to utter, say, ‘x knows that P’ against a common ground which does not include P is equivalent to asserting ‘P and x knows that P’. Such a conjoined assertion might indeed leave the audience perplexed about the speaker’s intentions concerning the direction of the conversation. However, this analysis ignores the possibility that a function of main-clause declarative syntax is to endow the proposition expressed with main-point status. If that is indeed a function of this form, then there is no need to claim that it is only rational to assert ‘x knows that P’ in a context in which P is common ground, for the confusion that Stalnaker claims would result if P were not common ground is avoided by other means, viz. by not giving P main-point status through one’s choice of the means by which one expresses P (cf. Abbott 2000). A Stalnakerian might want to respond, of course, that it is just this main-point effect that the view of presuppositions as being associated with common-ground status seeks to explain. Common ground assumptions, the response would go, are by their nature, in the background, hence the non-main-point status of non-asserted propositions. However, by its insistence that conversations can proceed efficiently only if presupposed propositions are common ground, the Stalnakerian position opens itself to accusations of being self-contradictory: on the one hand, it identifies a presupposition as a proposition P associated with a linguistic form whose use is only acceptable, on grounds of rational communicative efficiency, if P is common ground; on the other hand, it
98
Assertion
argues that is sometimes acceptable, presumably on the same grounds, to use that form if the P is not common ground. Thus the reasoning process that is claimed to underlie presupposition accommodation goes as follows: 1. 2. 3. 4. 5.
S’s utterance contains L It is communicatively efficient to utter L only if P is common ground P is not common ground S is not behaving in a communicatively efficient manner If the hearer accommodates P, S will be behaving in a communicatively efficient manner 6. P is not controversial and S is in a position to know that P 7. Accommodate P! The problem with this inference is that accepting the premise at line 5 requires rejecting the premise at line 2: accepting that is communicatively efficient to accommodate presupposition P is tantamount to denying that it is communicatively efficient to utter the linguistic form associated with P only if it is common ground that P. In other words, as long as the alleged common-ground status of presuppositions is derived from considerations of rational communicative efficiency, the phenomenon of presupposition accommodation is incompatible with their having that status. The Stalnakerian view of presuppositions is thus undermined by cases in which it is communicatively efficient to use a form associated with a presupposition even if that presupposition is not common ground. The mechanism of accommodation cannot rescue this conception of presupposition because the same considerations that are claimed to underlie the alleged common-ground status are employed to explain those cases in which this status is absent.7 Modifying the common-ground requirement so that a presupposed proposition must have common-ground status only at an idealised point between comprehension and acceptance/rejection simply removes the intuitive appeal of the original conception. What we are left with is a distinction not between what is asserted and what is common ground, but between propositions whose addition to the common ground is the raison d’être of the speech act, and propositions which can be justifiably added to the common ground given that the speech act has been performed. This, I think, is a very useful and theoretically important distinction which has considerable explanatory potential: it might explain, for example, why ‘the family of utterances’ (to use the felicitous phrase coined by Chierchia and McConnell-Ginet 2000: 28–32) related to an utterance U (i.e. its negation, modal form etc.)
Assertion and Common Ground
99
share presuppositions. Because the speaker, in each case, is talking about the same objects and situations, we can draw similar conclusions from each of the utterances (such as the existence of the objects referred to). Our ability to draw such inferences can be exploited be speakers.8 The tough question then becomes explaining why it is acceptable to communicate a certain proposition without asserting it. The general intuition is that this has to do both with considerations of main-pointhood and controversy. What is needed is a framework that can give flesh to these intuitive bones. In Chapter 7, an account is proposed that seeks to do just this. This, though, makes use of a conception of common ground that better reflects the realities of linguistic interaction and the limitations of human cognition: Sperber and Wilson’s notion of mutual manifestness. This is compared with Stalnaker’s notion of common ground in the next section of this chapter.
5.4 Comparison with Relevance Theory The analysis put forward in Chapter 8 employs Sperber and Wilson’s Relevance Theory as a means of bringing together disparate perspectives on assertion. In order that the view developed there can be related to the Stanakerian picture, it will be useful here to compare the two frameworks. We will see that both models allow for a range of propositions to be a communicated by an assertion, and that both see the enlarging of a set of mutual beliefs as central aim of linguistic communication. I will argue, however, that from the perspective of a theory of utterance interpretation, the Relevance Theory notion of mutual manifestness offers a solution to certain problems raised by Stalnaker’s conception of common ground. One similarity between the two frameworks lies in the distinction that Stalnaker makes between the essential and non-essential effects of assertion. The essential effect of assertion, recall, is the reduction of the context set by the addition of the proposition explicitly expressed by the assertion to the common ground. This type of effect can be avoided if the assertion is rejected. Non-essential effects are reductions to the context set that result from the fact that the assertion has been made. These effects are not dependent on acceptance. One non-essential effect of an assertion will be to add (7) to the common ground, another, assuming the speaker is judged sincere, will be to add (8): (7) The speaker has asserted that P (8) The speaker believes that P
100
Assertion
This distinction is captured in Relevance Theory by the distinction between basic and higher-level explicatures. An explicature is a communicated assumption that is derived by developing the encoded meaning (logical form) of the utterance (Sperber and Wilson 1986/1995: 182). Higher-order explicatures are derived by embedding, under a propositional-attitude or speech-act description, the proposition derived by developing the encoded meaning of the sentence. In RelevanceTheory terms, then, (7) and (8) are higher-level explicatures of an assertion that P. Another similarity is an emphasis on the role of linguistic communication in co-ordinating behaviour. Whereas Stalnaker sees assertion as an act aimed at reducing the context set by adding the proposition expressed to the common ground, Sperber and Wilson see a central aim of linguistic communication as being to expand the mutual cognitive environment of the participants in a conversation (1986/1995: 64). A mutual cognitive environment is a set of shared assumptions whose shared nature is manifest to both the speaker and hearer. As noted briefly in Chapter 3, an assumption is said to be manifest to an individual to some degree if she is capable or representing it mentally and judging it as true or probably true (1986/1995: 39). As assumptions are propositional entities, they will pick out the set of possible worlds in which they are true. Moreover, because the aim of a cognitive environment is to provide an accurate representation of the world (i.e. it is consistency aiming), the set of worlds it picks out can be identified as those in which all mutually manifest assumptions are true in all possible worlds in that set. As in Stalnaker’s model, this is an idealisation, but it is a reasonable one in that consistency is what cognitive environments aim at, even if they never achieve it. The set of worlds picked out by the mutual cognitive environment is thus equivalent to Stalnaker’s context set, and extending the mutual cognitive environment entails reducing the context set. However, Sperber and Wilson’s notion of manifestness has the advantage of accounting for a phenomenon that poses a challenge for a possible-worlds account of mental representation such as Stalnaker’s. This is the problem of logical omniscience (Stalnaker 1991/1999). Analysing beliefs as sets of possible worlds, as Stalnaker does, assumes that the beliefs of an individual are deductively closed: i.e. that the holder of a set of beliefs believes all that that set of beliefs implies. This is also an idealisation: we are clearly ignorant of the consequences of many of our beliefs. Moreover, working out these consequences is often a difficult reflective process. These facts, though, are hard to explain at
Assertion and Common Ground 101
the level of abstraction at which Stalnaker’s account operates: there is no obvious reason, on the version of the possible-worlds account of propositions defended by Stalnaker, why the consequences of an individual’s beliefs should not be available to her. The Relevance Theory notion of manifestness, by contrast, reflects this fact about beliefs. As noted briefly in Chapter 3, Sperber and Wilson draw a distinction between those assumptions – to use their terminology – that are mentally represented by (or entertained) by an individual and those that he is capable of entertaining as a result of inference or perception. The former – those which are entertained and can therefore play a causal role in cognition – they refer to as the assumptions of an individual; the latter they call assumptions which are manifest to an individual (1986/1995: 40). This terminology is confusing, however, in that entertained assumptions are also manifest to the individual (one is clearly capable of representing assumptions that one has in fact represented). Consequently, I will from now on refer to the former as ‘mentally represented assumptions’ and the latter as ‘merely manifest assumptions’. By distinguishing between mentally represented assumptions and merely manifest assumptions, Sperber and Wilson avoid the problems of logical omniscience that Stalnaker encounters. Belief, on Stalnaker’s account, is a disposition to act a certain way, given certain desires (1984: 15). On this view of belief, the problem of logical omniscience reveals itself when (a) we observe that a certain belief B is entailed by a set of beliefs which, by virtue of his behaviour, we feel justified in attributing to an individual, but (b) the individual does not demonstrate a disposition to act on B to fulfil his desires. On Sperber and Wilson’s account, by contrast, this observation can be explained: the individual does not act on B because it is a merely manifest assumption, and hence has no cognitive powers of causation. Of course, buying this solution to the problem of logical omniscience requires accepting that for a belief to have causal powers it must be physically instantiated in some manner in the mind/brain of an agent. Sperber and Wilson follow Fodor (1975; 1998) in viewing such mentally represented assumptions as syntactic objects, or sentences of the language of thought. On this view, it is by virtue of their formal properties that beliefs have causal powers, so merely manifest beliefs will play no causal role in cognition. Now, Stalnaker has argued against what he calls the linguistic picture of mental representation, on which beliefs are viewed as mental sentences with formal properties (1984: chs 1 and 2). However, his objection is that, according to him, such an account cannot serve as a theory of content, of what is represented by thought.
102
Assertion
When it comes to the issue of how content is mentally represented, he is less vehemently opposed to the idea that thoughts may have form and structure akin to that we find in the sentences speaker’s use to express them. The difference between his pragmatic view – on which content is to be explained in terms of dispositions to act – and what he calls the linguistic view is that the pragmatic view sees the form in which beliefs are represented as inessential to their content. The views he seeks to reject, by contrast, attempt to explain content in terms of a causal relationship between objects and states of affairs in the world and formal objects in the mind/brain. However, as Stalnaker acknowledges, the pragmatic picture that he propounds does not preclude positing such formal mental objects in an explanation of how representation is achieved (1984: 23) Thus it is open to Stalnaker to adopt the distinction between mentally represented assumptions and merely manifest assumptions as long as his theory of content does not rely on this distinction. Indeed, Stalnaker (1991/1999) himself argues that the solution to the problem of logical omniscience lies in an understanding of accessibility (as opposed to implicitness). And accessibility is a notion open to analysis in terms of manifestness. A further advantage of adopting the notion of mutual manifestness is that it affords a more nuanced picture of the relationship between assertion and the mutual information state of the interlocutors than the more abstract Stalnakerian common ground. Rather than a proposition either being common ground or not, as is the case on Stalnaker’s model, we have degrees of manifestness, and a mutually manifest assumption may be more manifest to the hearer than to the speaker. An assertion may thus serve to make more manifest to the hearer what, in Stalnaker’s framework, is already a common-ground assumption in that it is true in all possible worlds in the context set. More generally, such a conception of the mutual background of beliefs against which linguistic communication takes place predicts that there should exist means by which speakers not only add to this background, but also modify the degree to which the propositions that constitute it are accessible. As we will see in Chapter 7, a number of putative presuppositional phenomena – particularly those associated with grammatical and phonological form – can be analysed as means by which to achieve this effect. On occasions, Stalnaker does appeal to functions other than informativeness that an assertion might have. In particular, he notes that they have the effect of influencing the direction that a conversation may take. One example is in his discussion, mentioned above, of the
Assertion and Common Ground 103
rationale for not asserting ‘x knows that P’ in a context in which P is in doubt or dispute. It would be communicatively inefficient to do so, he says, because it would ‘leave unclear what direction he intended or expected the conversation to take’ (1974/1999: 55). Elsewhere, he says that: There may be other constraints on appropriate assertion – other considerations that count in favour of stating that φ rather than informing the addressee that φ by manifestly presupposing it. A successful assertion may change the context in other ways than by simply adding its content to the context, for example by influencing the direction of the subsequent conversation (2002: 710, original emphasis). Stalnaker’s conception of common ground, though, does not allow for these other effects of assertion to be modelled. From the point of view of a theory of utterance interpretation, this is major shortcoming. It is likely that Stalnaker would respond that, for his purposes, his highly abstract notion of common ground is sufficient. For some of his concerns, such as the analysis of conditionals, belief reports and certain problems of reference, that may well be the case. For a thorough analysis of assertion, however, a framework that better reflects the constraints of human cognition is to be preferred.
5.5 Conclusion One of Stalnaker’s basic claims is that presuppositions are common ground assumptions. This view, however, cannot be maintained, at least not if it is motivated by considerations of rational communicative efficiency. Nevertheless, Stalnaker’s model offers important insights that will inform the account put forward in Chapter 8. First, it draws an important distinction between those propositions whose communication is the raison d’etre of assertion and those whose which can be inferred from the fact that this act takes place. Second, it shows how world representation can be modelled using the possible-worlds framework. A proposition, considered in isolation, is a representation of a world as being a certain way. However, it cannot yet serve as a representation of a particular world, for many worlds will be as represented by that proposition. To represent a particular world, we need a consistent set of propositions, so that alternative candidate worlds can be successively eliminated and we can move closer to an accurate representation
104
Assertion
of the target world. For the target world to be the actual world, however, some form of stipulation is needed. What Stalnaker does is treat the basic context set as having the function of representing the actual world. In Chapter 8, I argue that Sperber and Wilson’s notion of a factual assumptions can serve as the representational format for those propositions that pick out the basic context set. Before doing so, however, more needs to be said about the relationship between assertion and linguistic form. This is the topic of the next two chapters.
6 Assertion and Mood
6.1 Introduction Taken together, the conclusions drawn from the discussion in Chapter 3 of the relationship between assertion and belief and Chapter 4’s consideration of the role of an assertion sign in the analysis of assertion lead naturally to the view that declarative syntax has a special relationship with assertion. In this chapter, we begin to consider more closely the nature of that relationship. Our main concern will be how the declarative is to be distinguished from the other major moods, such that its assertoric role is explained. The approach taken will be to critically survey a range of proposals that have been made concerning the semantics of the major moods. First, we consider approaches to mood that take a formal semantic perspective, before going on to look at speech-act approaches.
6.2 Mood in formal semantics Within formal semantics, two general approaches to mood can be discerned. On the one hand, there have been attempts to integrate mood into a Davidsonian account of linguistic semantics by searching for the equivalent of T-sentences for non-declaratives; on the other, efforts have been made to account for mood distinction in terms of possible worlds, often employing Stalnaker’s common-ground framework. In this section, we look at each in turn. 6.2.1 Davidsonian accounts What I am calling Davidsonian accounts offer an analysis of nondeclarative moods that parallels the T-sentence analysis of declaratives 105
106
Assertion
championed by Davidson. Such theories have been put forward by, among others, McGinn (1977), Segal (1991) and Boisvert and Ludwig (2006). Davidson himself, though, favoured a different approach to mood, so we will briefly consider his own account before looking at the attempts made by others to incorporate mood into his general semantic programme. Davidson’s paratactic account of mood divides the non-declarative moods into a declarative core and a mood-setter. Thus, according to him, an imperative such as (1) encodes the information in (2), and an utterance of (2) is composed of two speech acts corresponding to the two sentential elements of (1) (1979/2001: 119): (1) Put the kettle on. (2) My next utterance is imperatival in force. You will put the kettle on. Davidson treats the declarative as force-neutral and thus the declarative core element of non-declaratives carries no indication of associated force. This is important, because analysing assertion as an act performed by uttering a form specified for that purpose, as Dummett seeks to (see Chapter 4), would create a problem of infinite regress for Davidson. This is because Davidson has it that when a non-declarative is used with its specified force, this is achieved by the speaker truthfully asserting the content of the mood-setter. But if assertion were analysed by him in terms of a form-force relationship, then a mood-setter would be needed for each assertion, and, as each mood-setter must be asserted if the sentence is to be uttered with its specified force, then a mood-setter will be needed for each mood-setter, ad infinitum. Thus a simple assertion such as (3) would have to be analysed as (4), and (1) as (5) (see Dummett 1993: 206): (3) The Earth is flat (4) . . . My next utterance is assertoric in force. My next utterance is assertoric in force. My next utterance is assertoric in force. The Earth is flat (5) . . . My next utterance is assertoric in force. My next utterance is assertoric in force. My next utterance is assertoric in force. My next utterance is imperatival in force. You will put the kettle on By claiming that the declarative is force-neutral, Davidson avoids these problems. However, this commits him to the view that the declarative offers no indication of force whatsoever. We might ask, with Dummett,
Assertion and Mood 107
how reasonable a claim this is: the fact that the declarative stands in opposition to other forms that Davidson accepts as force indicators surely means that its use must give some indication of the force attached to the utterance: To make out that the indicative mood was really a non-mood, one would have to show that its use gave no indication whatever of the force attached to the utterance. This is impossible to do: at the very least, the use of the indicative mood is a prima-facie indication that the speaker is attaching to what he says a force distinct from any of those which the interrogative, imperative, and optative moods are typically used to convey (Dummett 1993: 207). Dummett also points out that Davidson’s paratactic account misanalyses his own data. Arguing against the view that the declarative is a marker of assertoric force, Davidson (1979/2001: 110) offers (6) and (7) as examples of non-declarative assertions: (6) Did you notice that Joan is wearing her purple hat again? (7) Notice that Joan is wearing her purple hat again (8) a. My next utterance is interrogative. You noticed that Joan is wearing her purple hat again b. My next utterance is assertive. You noticed that Joan is wearing her purple hat again (9) a. My next utterance is imperative. You notice that Joan is wearing her purple hat again b. My next utterance is assertive. You notice that Joan is wearing her purple hat again In these cases, Davidson would have it that the mood indicates a double utterance so that (6) and (7) would be interpreted, initially at least, as having the truth conditions in (8)a and (9)a respectively. However, as these are presented by Davidson as examples of an interrogative and an imperative being used to make assertions, the mood-setter would be judged false and the utterances reinterpreted as in (8)a and (8)b. But this is wrong: if these are indeed assertions, then what the speaker is asserting is the content of the embedded declarative sentence (that Joan is wearing her purple hat again), not the content of the whole sentence (that the hearer noticed that Joan is wearing her purple hat again). But this means that Davidson must provide an explanation of how the content of the embedded declarative is picked out as the content of the
108
Assertion
assertion, and, because he claims that the declarative is not a force indicator, he must do this without any appeal to a relationship between declarative mood and assertoric force (Dummett 1993: 208–9). Davidson’s account has also been criticised on other grounds (Hornsby 1986; Hare 1989; Segal 1991; Harnish 1994; Boisvert and Ludwig 2006), but we will not go into these criticisms here as they do not relate directly to the relationship between declarative mood and assertoric force. Rather, we will consider other approaches to mood that have been put forward within his semantic framework. Of attempts to integrate mood into the Davidsonian programme, the most developed and thorough is Boisvert and Ludwig (2006), so it is that work on which we will concentrate here. While they do not deny a link between illocutionary force and mood, Boisvert and Ludwig reject the view that mood is a conventional indicator of force. Their main reason for doing so is that, according to them, such an account fails to explain the contribution of mood to mixed-mood sentences such as (10) to (22) (their (7) to (11)): (10) (11) (12) (13) (14)
If the war goes badly, the President’s approval rating will drop If you’re going to the store, buy some milk If you’ll be in later, would you like to have lunch? Don’t cheat or you’ll get a failing grade for the course Is it a secret or can you tell me who it is?
Regarding (10), they point out that while this contains two declarative sentences, neither is used assert to the proposition expressed. The obvious response for an adherent of the force-indicator approach is to say that only main-clause mood indicates force and that what is being asserted is the proposition expressed by the conditional sentence as a whole. Boisvert and Ludwig accept this objection as far as (10) goes, but claim that it does not explain examples (11) and (12). These sentences, they say, are not, respectively, in the imperative and interrogative moods because they do not issue a simple directive nor ask a simple question; rather, these sentences issue a conditional directive and ask a conditional question. Unfortunately, Boisvert and Ludwig do not consider the obvious response to this counterargument: conditional directives and conditional questions are still directives and questions, and how we identify (11) and (12) as such can be straightforwardly explained if we assume that main-clause mood indicates force.1 As regards (13) and (14), Boisvert and Ludwig rely heavily on punctuation in order to pass these off as evidence against the force indicating role of mood. The lack of a
Assertion and Mood 109
comma or semi-colon between the disjuncts might give the impression that we are dealing with single utterances here, but if read aloud, these examples require what is often termed ‘comma intonation’, indicating that each example actually consists of two illocutionary acts (a directive and an assertion/warning in the case of (13), two questions in the case of (14)). Thus analysed, (13) and (14) do not provide evidence against the force-indicating view of mood. This is not to say, though, that there is no evidence against the forceindicating view. The strongest case is put forward by Pendlebury (1986) who notes that mood is not semantically inert when embedded, but can contribute towards the sense of the sentences in which it occurs. He gives the following examples (his (5) and (6)): (15) Rick thinks he knows that Sam will play it again (16) Rick thinks he knows whether Sam will play it again The difference between the grammatical moods of the embedded sentences in these cases has truth-conditional consequences for the sentences as a whole. It is not only with ‘know’ that this occurs: note how the mood of the embedded clause leads to a truth conditional difference between the following examples: (17) Peter insists that they are here (18) Peter insists that they be here Because they deny that mood is a force indicator, accounts such as Boisvert and Ludwig’s might be better equipped to deal with this data. Therefore, despite the failures of Boisvert and Ludwig’s objection to the force-indicating approach to mood, their positive proposals still merit consideration.2 Davidsonian accounts such as Boisvert and Ludwig’s seek to account for mood as a feature of sentence types that can make a contribution to the compositionally derived meaning of a sentence type in which it occurs. The Davidsonian approach to linguistic meaning gives an extensional account of the meaning of declarative sentence types by specifying, compositionally, their truth conditions. Non-declaratives do not have truth conditions, however, so it is not immediately obvious that this account can be extended to account for the full range of sentence types. The solution proposed by Boisvert and Ludwig (following McGinn (1977) and Segal (1991)) is to generalise the notion of truth to a notion of fulfilment. They then identify two main types of fulfilment
110
Assertion
Fulfilment conditions
Truth conditions
Compliance conditions
Obedience conditions Figure 6.1
Response conditions
Fulfilment conditions
conditions: truth conditions and compliance conditions, subdividing the latter into obedience and response conditions. This is illustrated in Figure 6.1 (Boisvert and Ludwig’s Figure 34.1): However, despite the impression that Figure 6.1 might give, Boisvert and Ludwig do not claim that the notions of obedience and response conditions can be explicated without recourse to the notion of truth. An utterance of an imperative, on their account, is obeyed iff the audience makes it the case that the declarative core of the imperative is true (and does so with the intention of obeying that utterance). This highlights an important feature of Davidsonian treatments of mood that distinguishes them from force-indicating accounts: whereas the latter see the force indicator as operating over a proposition, or sentence radical, the former see imperative and interrogative sentences as being derived from a genuinely declarative core. In this respect, then, Davidsonians are closer to the speech-act fundamentalists discussed in Chapter 2 than the speechact traditionalists: recall that the traditionalists treat assertion as just one mood among many and treat propositions as the basic bearers of truth conditions; Davidsonians, by contrast, treat declarative sentencetypes as basic and analyse the other moods in terms of these, paralleling the speech-act fundamentalist strategy of analysing the other major illocutionary forces in terms of assertion. But speech-act fundamentalists explain the link between mood and force by defining the sentence typically used to perform an illocutionary act in terms of that act. So, for Brandom the declarative is whatever form is typically associated with the making of assertions (claims) within a language, and while Barker identifies declarative sentences as protodeclaratives. Davidsonian accounts of mood, by contrast, deny such a direct link between form and force. How, then, do they explain
Assertion and Mood 111
the systematic relationship that is commonly observed? The move that is made, both by Segal and by Boisvert and Ludwig is to claim that what relates a form with a force is a common type of fulfilment condition: according the these authors, declaratives and assertions both have truth conditions; imperatives and directives both have obedience conditions, interrogatives and questions both have response/answerhood conditions. Given this parallel, a pragmatic principle can be posited such that when someone utters a token of a sentence type that has a certain kind of fulfilment condition, they can thereby be taken to be performing an act which has the same class of fulfilment condition, unless there is some indication to the contrary (Segal 1991: 177 posits such a principle). This view has it that the declarative – our current concern – is apt for assertion because it has truth conditions, and assertions aim at truth. We might justifiably ask at this point why it should be the case that a sentence type should exist that has the same type of fulfilment conditions as speech-act type even though its function is not to indicate that an utterance of that speech act is intended. However, it is not the case that these authors see the supposed existence of sentence types and illocutionary-act types with matching fulfilment conditions as merely a fortunate coincidence. Indeed, Boisvert and Ludwig (2006: 878) speak of imperatives and interrogative as being ‘designed to help us perform’ certain speech acts. Rather, the aim is, on the one hand, to account for mood in a compositional theory of the semantics of natural language, and on the other to have a suitably weak link between form and force such that the commonly made observation that the moods are often used without their associated force can be explained. Accounting for the fact that mood is a defeasible determinant of force is a challenge that any theory of mood must meet. However, it is not certain that this requires abandoning the views that the function of mood is to indicate force, as Boisvert and Ludwig counsel. As we will see, there are other accounts that view mood as a force-indicating form but do not require that mood always determines force. Boisvert and Ludwig’s proposal must therefore be assessed in terms of how it meets its other aim of accounting for mood within a compositional semantic theory. Here, though, we have a problem. The sort of data that they seek to account for is not the truth-conditional consequences of mood alternation that we find in the Pendlebury type examples above ((15) to (18)), but the ‘mixed mood’ sentences exemplified by (10) to (14). But, as we have seen, these data can be handled quite straightforwardly by a force-indicating account.
112
Assertion
Discussing the Pendlebury-type cases, Boisvert and Ludwig (2006: 883) raise the issue of whether these would not be better analysed by treating ‘knows that’ and ‘knows whether’ as distinct compound verbs.3 If this is the case, then the complements of (15) and (16) have the same mood. (23) (24) (25) (26)
Rick thinks he knows that Sam will play it again Rick thinks he knows whether Sam will play it again Peter insists that they are here Peter insists that they be here
Pendlebury discusses this problem and argues that it is uneconomical, as it would also require treating each of ‘knows when/why/how/how often/what/who’ as a compound verb. Greater parsimony is achieved by treating ‘whether’ as a mood-marker than as a verb modifier, Pendlebury claims (1986: 364). Pendlebury could also point to cases such as (17) and (18): here the complementiser is the same and the difference in truth conditions must be explained in terms of the syntax of the subordinate clause rather than positing that the complementiser modifies the verb.4 Moreover, it is not clear how Boisvert and Ludwig’s account could explain this difference: the fulfilment of (18) does not depend on anyone’s compliance with the subordinate clause, but on whether Peter actually issues a directive with that content. Of course, Boisvert and Ludwig could deny that an account of imperatives should also account for apparent ‘embedded imperatives’. But if, as I have argued, their account has no advantages over force-indicating accounts in explaining the ‘mixed mood’ cases, and cannot explain the apparent contribution of mood to the Pendlebury-type cases, then it is fast running out of unique selling points and we are left with few reasons to buy it. One notable reason Boisvert andLudwig’s account will find it difficult to account for the Pendlebury cases is that it seeks to explain the non-declarative moods in terms of the perlocutionary consequences of the associated illocutionary forces: i.e. the hearer’s compliance with the speaker’ directive. It is hard to see how this could be extended to cover embedded clauses such as those we find in (15) to (18), for subordinate clauses of this type are generally associated neither with an illocutionary nor a perlocutionary point independent of that associated with an utterance of the main clause, and the truth of these sentences certainly does not depend on the fulfilment of any perlocutionary goals associated with the subordinate clause. Moreover, there is room for concern
Assertion and Mood 113
about the whole strategy of generalising the notion of truth conditions to a broader category of fulfilment conditions: this puts being true on a par with being complied with, but it is odd to say that an assertion is fulfilled if it is true, and being true is certainly not a perlocutionary goal of assertion. There is more that could be said about Boisvert and Ludwig’s approach to mood, but to do so would take us too far away from our current concern, which is the declarative and how it is to be distinguished from the other moods. As far as that question is concerned, the discussion in this subsection has shown that to attempt to do so in terms of distinct types of fulfilment conditions is, at the least, highly problematic.
6.2.2 Possible-world accounts of mood As noted at the beginning of this section, the alternative approach to mood that is favoured in formal semantics is to employ a possibleworlds framework. One of the earliest proposals to deal with mood in this way was put forward by Huntley (1984) in his analysis of the English imperative. Huntley argues that what distinguishes the declarative from the imperative is that the former, but not the latter, involves deictic linking to a particular possible world. Thus, he claims, declaratives are apt for assertions because they specify the world at which they are to be evaluated as true or false, while imperatives do not. The world to which declaratives typically establish a deictic link is, he argues, the actual world, and it is in these cases that assertoric force is conveyed. A similar claim is made by Farkas (1992). According to her, the declarative requires anchoring the proposition expressed to a particular possible world. A main-clause declarative is said to be anchored to ‘the world that models reality for the purposes of a conversation’ (p. 85), whereas embedded declaratives are claimed to anchor the proposition expressed to a world that represents the way reality is taken to be by the subject of the matrix verb, so that a belief report anchors the proposition expressed by the complement to a world that represents reality for that individual to whom the belief is attributed. The (Romance) subjunctive is chosen, on her account, when the anchoring relationship is to a set of worlds, rather to than to a particular world. This is intended to explain the use, for example, of the subjunctive as the complement of desire attributions such as (19): the proposition expressed by the complement of a desire report is not claimed by the speaker to be true in anyone’s representation of reality.
114
Assertion
(19) Silvia quiere que Juan se vaya ‘Silvia wants Juan to go-SUBJ’ In section 5.2.2, it was noted that possible worlds are used in two different ways in Stalnaker’s common-ground model of assertion: both as the content of the assertion and as the object of representation. An assertion picks out a set of possible worlds, and that is its content, but an assertion is a representation of a particular possible world, and that world is the object of representation. Huntley’s (1984) and Farkas’s (1992) accounts use possible worlds in this second way, as the object of representation: the anchoring relationship in Farkas’s account and the deictic reference in Huntley’s identify which possible world is being represented by the proposition expressed by the declarative. But, while this approach to mood is intuitively appealing, it is open to objection on the grounds that it mis-describes the phenomenon of belief attribution. It is fair to say that, in making an assertion, I am describing the actual world as being a particular way. However, it is misleading to extend this to the analysis of belief reports, and say that when I attribute a belief to another individual (using an embedded declarative sentence) I am describing another world as being a certain way. Rather, I am describing the actual world as being a certain way according to someone else. It is wrong to talk of the world represented by an individual’s beliefs, if what is meant is some world other than the actual world. The world represented by the individual’s beliefs is the actual world: it is that in the case of a false belief, the actual world is mis-represented. What an individual’s beliefs pick out, though, it not a particular world but a set of worlds: beliefs, even when consistent, are not comprehensive enough to pick out unique possible worlds. Hence the standard Hintikka (1962) semantics for belief reports has it that a belief ascription is true iff the proposition expressed by the embedded clause is true, not in a particular possible world, but in all worlds picked out by the attribuee’s belief set. But, if this is accepted, then the view that declaratives anchor a proposition to a particular world in which the proposition expressed is to be evaluated as true or false cannot be maintained. The embedded use of declaratives in belief reports cannot be said to anchor the proposition expressed to the particular world taken to model reality according to the individual to whom the belief is attributed, for no such world can be identified; and they cannot be said to be anchored to the actual world (i.e. the world represented, albeit indirectly, by the object of a belief report), because the truth of a belief report does not depend on the truth of the embedded proposition in the world in which the report is made.
Assertion and Mood 115
In later work, Farkas (2003) modifies her position and suggests that what distinguishes declaratives as used in belief reports from (Romance) subjunctives as found in desire reports is the type of context-change potential (‘CCP’, see Heim 1983/2002) associated with each. The declarative complements of belief reports have, according to Farkas, assertive CCP in that the proposition expressed is to be added to a derived context set (see section 5.2.3) that constitutes the content of the attribuee’s representation of the world. The subjunctive complements of desire verbs, by contrast, change the context set not by removing possible worlds but by ordering them, ranking those in which the proposition expressed by the complement clause is true higher than those in which it is not (as suggested by the analysis of desire predicates in Heim 1992).5 Although Farkas herself doesn’t make this point, it is worth noting that her later approach to mood manages to explain, without positing an anchoring relationship, the observation that declaratives are apt for representing particular worlds. We saw in Chapter 5 that the reason that assertions can be said, on the Stalnakerian picture, to represent the actual world is because they come with a commitment to consistency. Because declaratives, on Farkas’s (2003) account, have assertive CCP and hence involve a commitment to consistency, they can be used to represent particular worlds. However, we need not say that the object of a belief report represents the particular possible world anchored to the subject of a belief report. Rather, it represents the actual world for that individual according to the speaker. By contrast, because re-ordering a context set does not require removing worlds, consistency plays no role in this process and the forms related with it cannot be employed as a means of representing a particular world. Taking a slightly different approach, though still within the Stalnakerian framework, Schlenker (2005) argues that the declarative encodes the constraint that it can only be used if the world denoted by the proposition expressed is present in (i.e. the proposition is not ruled out by) the context set. This is not the same, however, as saying that the declarative encodes assertive CCP. Having assertive CCP requires that: a) the proposition expressed by the utterance is not already part of the common ground (i.e. is not true in all possible worlds in the context set); b) the proposition expressed by the utterance is not ruled out by the common ground (i.e. is true in some possible worlds in the context set);
116
Assertion
c) the utterance expresses a proposition in each of the possible worlds in the context set; d) the proposition expressed can be added to the common ground as a result of the acceptance of the utterance. Requirements a) and b) ensure, respectively, that adding the proposition expressed to the common ground will reduce the context set but not eliminate it altogether. Requirement c) ensures that the context set can be reduced in a determinate way: if a proposition cannot be given a truth-value at a possible world in the context set, it will be unclear whether it should be removed or not and the audience will be confused about the speaker’s intentions (on these first three requirements, see Stalnaker 1978/1999: 88–90). Requirement d), I argue, is needed to distinguish assertions from other speech acts that share the first three requirements but do not have the function of reducing the context set. For example, a command such as (20) requires that the proposition expressed is not part of the common ground ((20) would be odd if it were common ground that the hearer had finished his meal) and that the proposition expressed is not ruled out by the common ground ((20) would be infelicitous if the meal were so huge that it could never be finished by the addressee).6 Moreover, it is likely that the proposition expressed by (20) will enter the common ground if (20) is compiled with. What distinguishes (21) from (20) is that acceptance of (21) will allow the participants of the conversation to treat it as common ground that Peter will eat his dinner up (whether he eventually does or not), while the only means by which (20) will get the proposition expressed into the common ground is if Peter complies with the order and his compliance becomes common ground. Thus it is context change that results from acceptance which is the hallmark of assertive CCP. (20) Eat your dinner up, Peter! (21) Peter will eat his dinner up As, on the Stalnakerian model, presuppositions are common-ground assumptions and hence true in all possible worlds in the context set, and as directives with content P generally share presuppositions with assertions of P, a command cannot be distinguished from an assertion in terms of requirement c). Consequently, if assertions are to be distinguished from commands, it must be by some other means. Requirement d) does this: issuing a command is not an attempt to modify the common ground by getting P accepted into the common ground.
Assertion and Mood 117
According to Schlenker, the declarative encodes simply requirement b) (that the proposition expressed by the utterance is not ruled out by the common ground). This requirement, though, is also fulfilled by presuppositions: a presupposition, on the Stalnakerian model is by definition true in all possible worlds in the context set. Consequently, Schlenker’s account predicts that clausal presuppositions should be expressed by declarative clauses. In French (the language he considers), though, there are counterexamples: ‘be happy that’, which presupposes the proposition expressed by the embedded clause, takes the subjunctive. (22) Jean est heureux qu’il pleuve/∗ pleut ‘Jean is happy that it rains-SUBJ/∗ IND’ Schlenker attempts to deal with this case by first noting that so-called ‘factive emotive’ predicates such as this are best analysed as causal statements: that it is raining is presented as the cause of Jean’s happiness. This position is taken by a number of authors (Bosque 1990; Heim 1992; Giorgi and Pianesi 1997; Quer 1998; Jary 2009). Schlenker also argues, though, that causal statements require counterfactual reasoning, on the grounds that, according to him, they require recourse to a counterfactual statement to explicate their meaning (roughly, ‘Jean would be less happy if it were not raining’). As we have seen, on a Stalnakerian account of conditionals (Stalnaker 1975/1999), counterfactuals require recourse to worlds outside the context set. This removes the presupposition that the context set contains a world in which the proposition expressed by the subjunctive clause is true, and consequently, on Schlenker’s account, warrants the use of the subjunctive. While this is a neat manoeuvre (a similar one is made by Quer 2001), it is an expensive one: it requires buying a whole theory of causality (i.e. Lewis’s 1973) and thereby attempts to resolve semantic issues by appeal to metaphysics. Moreover, even if causality is correctly analysed in terms of counterfactuality, it is still a rather desperate move to say, when faced with a sentence that threatens an analysis, that we should instead consider a different sentence (a counterfactual, in this case) which would confirm that analysis. Given these concerns, we should reject the claim that the declarative is unique in requiring that the proposition expressed by the utterance not be ruled out by the common ground, for this is also a feature of any embedded non-declarative that expresses a presupposition.
118
Assertion
Perhaps the most fundamentally Stalnakerian account of mood is that put forward by Portner (2005; 2007), who attempts to give an account of mood in the terms of Stalnaker’s (1984) broader philosophical aim of explaining semantic content in terms of rational action. On this account, the content of desires is to be explained in terms dispositions to act given certain beliefs, and the content of beliefs in terms of dispositions to act given certain desires. Rational action is explained in terms of choosing between possibilities, and so meaning is characterised in terms of possible worlds. Portner seeks to mirror this view of meaning in an account of mood by positing that, in addition to Stalnaker’s common ground, there is associated with each participant in the conversation a ‘To-Do List’. The role of the To-Do List is to rank worlds in the common ground as more preferable than others, actions being rational if they tend to remove lowly ranked possible worlds from the context set.7 Portner defines a To-Do List not as a set of propositions, but as a set of properties. He then claims that what distinguishes imperatives from declaratives is that the former denote properties (derived by abstracting over subjects) and are thus apt for updating a To-Do List, while declaratives denote propositions and are therefore apt for updating the common ground. His reason for denying that imperatives denote propositions is the claim that this view fails to account for differences in acceptability between imperative sentences which might be claimed to express the same propositional content. The example Portner gives is (23) (his (8a & b)): (23) a. ∗ Mary kiss (you)! b. (You) be kissed by Mary! (24) Everyone sit down. Portner’s explanation for the unacceptability of (23)a and the acceptability of (23)b is that the property of being kissed cannot be derived from (23)a, because (23)a has a nominal subject: only null and pronominal subjects permit the required abstraction. In response to the obvious objection that what is wrong with (23)a as an imperative is that it has a third-person subject, Portner notes that imperatives can allow thirdperson subjects, hence (24). However, it could also reasonably be argued that the constraint that imperatives place on their subjects is not to do with grammatical person, but with thematic role. The constraint appears to be that the addressee of the utterance must be both the referent of the subject of the sentence and the agent of the act. This would explain
Assertion and Mood 119
why the only interpretation that comes to mind for the very odd, but marginally acceptable, (23)b is that the addressee should allow herself to be kissed by Mary, perhaps by offering her a cheek or removing an obstacle, such as a veil.8 (23)a is ruled out by the fact that although the agent, Mary is not the addressee. It is not necessary, then, to posit that imperatives denote properties in order to explain the data in (23). The only other motivation that Portner gives is to explain how imperatives are interpreted as having directive force: because the To-Do List is made up of properties, an utterance of a sentence type that denotes a property will indicate that the addressee should add the property (claimed to be) denoted to his To-Do List. This is not independent motivation for accepting the view that imperatives denote properties, though, for Portner has stipulated both that the ToDo List contains properties and that imperatives denote these. There is little reason to assume, therefore, that imperatives differ from declaratives in the way that Portner claims. Moreover, the ordering of possible worlds that his theory of mood requires can be done on the assumption that imperatives denote propositions, as on standard theories of modal semantics (see e.g. Kratzer 2001). The only remaining reason that one might not want to accept that, on the Stalnakerian picture, imperatives denote propositions is that imperatives cannot be true or false. But if one takes assertions to be the primary bearers of truth, another explanation is possible: imperatives cannot be true or false because, despite expressing propositions when uttered, they are specified for performing directives, and directives are not truth apt. Indeed, it is reasonable, on the Stalnakerian framework that Portner aims to implement, to assume that imperatives do denote propositions, despite the fact that they cannot be true or false. Consider the following quotation: There are just two truth-values – true and false. What are they: mysterious Fregean objects, properties, relations of correspondence and non-correspondence? The answer is that it does not matter what they are; there is nothing essential to them except that there are exactly two of them. We could formulate the definition of proposition in a way that did not mention truth values at all without changing its essential character: a proposition may be thought of as a rule for selecting a subset from a set of possible worlds. The role of the values true and false is simply to distinguish the possible worlds that are members of the selected subset from those that are not (Stalnaker 1984: 2).
120
Assertion
All a proposition is, for Stalnaker, is a means of identifying a subset of possible worlds. So, as understanding an utterance of an imperative sentence requires being able to divide a set of possible worlds into those in which it is complied with and those in which it is not, imperatives are just as aptly thought of, on Stalnaker’s model, as expressing or denoting propositions as declaratives. As noted in Chapter 5, the notion of truth that is employed in truth judgements, and which cannot be applied to imperatives, is distinct from the abstract notion employed in the characterisation of Stalnakerian propositions. The truth-judgement notion relates to acceptance and the consequences this has for rational action. It relates not merely to propositions but to the activity of assertion and its role in constraining and co-ordinating behaviour. Consequently, as mere propositionhood (as defined by Stalnaker) is not a sufficient condition for truth aptness, a lack of truth aptness should not be taken to be evidence that a form does not express or denote a proposition. And if it cannot be shown that imperatives do not express propositions, then Portner’s proposal affords no means of distinguishing declarative from imperative mood. That concludes our look at attempts to account for mood within a formal semantic programme. As we have seen, none of these attempts is unproblematic. We will now see how speech-act approaches fare.
6.3 Speech-act approaches to mood Not all speech-act theorists make detailed claims about what is encoded by mood. Among those who do make explicit claims, the approach they take depends largely on their perspective on assertion: if they are fundamentalists, in the sense of Chapter 2, then they tend to adopt a much closer link between mood and force than the traditionalists. This is hardly surprising: as they seek to explain meaning in terms of speech acts, fundamentalists need a means of identifying assertions that makes no reference to other meaningful acts, entities or states. Consequently, they need to posit a tight link between form and force in order that assertion can be identified as the production of this form under appropriate conditions. Traditionalists, however, do not take assertion as fundamental, it is thus open to them to explain the force-form relationship in other terms, such as the expression of belief. In this section, we will first look at the fundamentalist, forceindicating accounts before moving on to look at the attitude-indicating account put forward by Bach and Harnish (1979), and comparing
Assertion and Mood 121
this with the similar approach put forward by Wilson and Sperber (1986/1995; Wilson and Sperber 1988).
6.3.1 Force-indicating accounts Of the speech act fundamentalists discussed in Chapter 2, only Barker puts forward a detailed account of mood. Brandom (1994) says little about mood beyond claiming that declarative mood is to be identified, in any speech community, by looking for the form that is typically associated with the giving of reasons. Alston (2000) clearly takes mood to be an indicator of illocutionary force, as sentence meaning is analysed by him in terms of illocutionary act potential, but he does not offer an analysis of why the moods have the illocutionary potential that they do. Barker, by contrast, offers an account of what is encoded by the major moods and of what happens when these are embedded. Consequently, we will focus on his proposals in this section. Barker’s main claim is that tokens of the three major mood types are proto-acts: to produce a declarative, imperative or interrogative sentence is, for him, to perform, respectively, a proto-assertion, proto-command or proto-question. Recall from Chapter 2 that, for Barker, to perform a proto-act is to behave in the manner of someone who has the intentions constitutive of that (non-proto) act. In order to actually perform an act by uttering a sentence whose tokening constitutes a proto-act, one must have the intentions that her proto-act advertises. For those who posit such a tight link between form and force, the challenge is to explain in a straightforward manner non-assertoric uses of main-clause declaratives, as well as this form’s significance when embedded. Of course, it is always open to the theorist to dodge the second requirement by arguing that force is only ascribed to main clauses, thereby effectively rendering embedded clauses moodless. Barker does not take this route: he claims that embedded declaratives are also protoassertions. We will look first at how Barker explains embedding, and then at his account of non-assertoric uses of the declarative. Barker treats conjunctive and disjunctive pairs of declarative clauses as expressive assertions,9 in which the speaker performs two protoassertions in order to signal that she is in a certain commitment state. In the case of conjunction, this is the state of being disposed to perform both conjoined proto-assertions with the advertised illocutionary intentions;10 in the case of (exclusive) disjunction it is the state of being disposed to perform one of the embedded proto-assertions, given the negation of the other (Barker 2004: 87). Barker treats embedding in
122
Assertion
conditionals in terms of supposition. A conditional of the form ‘If S1 , then S2 ’ is, on his framework, is comprised of two acts (op.cit: 97): Act 1: In uttering ‘If S1 ’, the speaker performs the proto-assertion S1 (but without the advertised intentions) and signals, by using ‘if’, her acceptance of a suppositional state formed by adding proto-acts S1 to her doxastic background. Act 2: In uttering S2 , the speaker performs the proto assertion S2 , but without the advertised intention. By attaching ‘If S1 ’ to the proto assertion S2, the speaker implicates that S2 is performed within the scope of the suppositional state accepted by Act 1. To accept a suppositional state is, in Barker’s framework, to accept that a certain discourse rule holds, namely that, within the scope of a suppositional state formed by adding a proto-assertion to one’s set of beliefs, one can perform a given act only if the permissibility of that act is entailed by that suppositional state (op. cit: 59–62). It is important to stress that, for a conditional sentence to be appropriately uttered, what must be entailed by the suppositional state is not a proposition but the permissibility of a speech act. Thus (25) is acceptable, on Barker’s story, not because the proposition expressed by the consequent is true in all worlds in which that expressed by the consequent is true, but because one who supposes the antecedent is entitled to assert the consequent scope of that supposition. This entitlement may arise because the truth of the antecedent requires the truth of the consequent in all possible worlds, but it is the fact that this makes asserting the consequent permissible by one who accepts the consequent that is crucial to Barker’s analysis of the conditional. This allows him to extend his analysis straightforwardly to conditional questions, such as (26): (25) If he gave her roses, he gave her flowers (26) If Jenny has three pounds, how many onions can she buy? It is a felicity condition on questions that they should normally be asked only if the speaker can justifiably assume that the hearer can answer the question. Thus (26) is a permissible question only if the hearer knows the price of onions (modulo other felicity conditions on questioning being fulfilled). In Barker’s terms, the permissibility of the proto-question ‘How many onions can Jenny buy?’ is entailed by adding the information that Jenny has three pounds to a belief set that contains the price of onions.
Assertion and Mood 123
Though Barker does not discuss these cases, his analysis lends itself straightforwardly to relevance, or ‘biscuit’, conditionals, such as (27):11 there is clearly a relevance felicity condition on assertions that could be exploited to explain such cases. The analysis would go as follows: adding the proto-assertion S1 (‘You’re hungry’) to a background of beliefs about how hunger can be sated creates a background against which S2 (‘There’s some chicken in the fridge’) would be permissible on grounds of relevance. (27) If you’re hungry, there’s some chicken in the fridge. Barker treats ‘that’-clauses containing declarative sentences as denoting proto-assertions. A belief ascription thus relates an individual and a proto-assertion. What is attributed to the subject of the belief report is therefore a commitment state and a disposition to defend that state. This is a straightforward account of such embedding. He says nothing about non-declarative embeddings, so his position on whether imperatives and interrogatives embed as the objects of desire and speech-act reports is unknown. We will not speculate on this matter here. The previous discussion shows that Barker’s analysis of the declarative deals well with its free-standing, mixed mood and ‘that’-embedding uses. It is now time to consider how well it meets the other challenge of explaining non-assertoric uses. Barker analyses declaratives used to issue directives, such as (28) uttered to give an order, as a case of a protodeclarative being uttered without the intention advertised but with the intention that the hearer believe that the speaker desires that certain a state of affairs obtains (Barker 2004: 57). (28) You’re going to leave on that boat tomorrow This intention is made clear, Barker claims, ‘by context’. This is leaving a great deal to context, especially when one could easily make use of the analysis of this sort of case given in section 4.4.1, where, recall, it was noted that utterances of declaratives such as (28) can only be given a directive interpretation if the speaker is in a position of authority over the hearer. In such a case, the speaker acting in the manner of one who is making a claim about a future state of affairs which the hearer is in a position to bring about will be a de facto directive. Such an analysis is open to any account that takes the declarative to be an indicator of assertoric force and assertion to involve the putting forward of commitments for acceptance, rejection or confirmation. Barker’s is such an account.
124
Assertion
Barker also treats explicit performatives as cases where the speaker performs a proto-assertion without having the intentions advertised by this act. Rather, the speaker has the intention that the hearer believe that the speaker intends to bring about, by virtue of her utterance, the state of affairs represented by that proto-assertion. Because they lack assertoric intentions, explicit performatives cannot be true or false; because they are proto-assertions, though, the representational element of proto-assertion can be exploited. Barker’s account handles the data on mood very well. By treating tokens of sentential moods as proto-acts that advertise certain representational and communicative intentions, he is able to give an account of why these forms typically have the forms they have without also positing that they necessarily have these acts. In many ways, his account is an effective blend of Dummett’s conventionalism and Grice’s intentionalism. However, his is very much a philosophical account of mood, a work of conceptual analysis. The proposals to be made in Chapter 8, by contrast, are concerned with the mechanics of assertion, i.e. with the cognitive processes underlying the interpretation of assertions and the role of declarative mood in those processes. As such, it should be seen as complementary to, rather than in competition with, a conceptual analysis such as Barker’s.
6.3.2 Attitude-indicating accounts As noted in Chapter 2, while speech act fundamentalists take the production of a declarative sentence to be, in part, constitutive of assertion, speech-act traditionalists place much greater emphasis on attitude expression. It should therefore be no surprise that the most worked-out account of mood among the traditionalists treats mood as a means of expressing a psychological attitude, this expression licensing an inference that the speaker has illocutionary intentions associated with that attitude. However, this approach is not limited to speech-act traditionalists: it is found also in Relevance Theory. Both Bach and Harnish (1979) and Sperber and Wilson (1986/1995; Wilson and Sperber 1988a; 1988b) see the relationship between mood and force as being mediated by the expression of propositional attitudes. Both accounts employ non-illocutionary notions of saying that, telling to and asking whether to characterise the information conveyed by main-clause mood. For Bach and Harnish, the major moods are associated with the locutionary acts of this nature, while for Sperber and Wilson, mood communicates
Assertion and Mood 125
higher-order explicatures that describe the speaker’s verbal behaviour in these non-illocutionary terms. Thus, on both accounts, an utterance of a main-clause declarative sentence such as (29)a licences the inference to (29)b, while an utterance of an imperative such as (30)a licenses the inference to (30)b: (29) a. b. (30) a. b.
The roses are red The speaker is saying that the roses are red Sit down, Peter The speaker is telling Peter to sit down
It is important to stress though, that on neither account is the speechact description a description of an illocutionary act. Rather, it is a first step to inferring the speaker’s illocutionary intention. For Bach and Harnish, (29)b and (30)b represent locutionary acts, and, on their account, different locutionary acts are compatible with distinct propositional attitudes: ‘saying that’ is compatible with belief, so that if a speaker says that P, then, as long as he is speaking literally, he is expressing the belief that P; ‘telling to’ is compatible with desire, so that if a speaker tells the hearer to P, as long as he is speaking literally, he wants the speaker to P. The inference from locutionary act to expressed attitude then licences a further inference from expressed attitude to an illocutionary act that requires, for its analysis, the expressed attitude as a necessary condition. So, as constatives are analysed by Bach and Harnish as expressions of belief, the hearer of (29)a is entitled to infer that the speaker is performing a constative, such as asserting. As directives are analysed by Bach and Harnish as expressions of desire, the hearer of (30)a is entitled to infer that the speaker intends to perform a directive, such as an order.12 For Sperber and Wilson, by contrast, the move is not from speechact description to expressed attitude, but from expressed attitude to speech-act description. For them, what mood encodes is how the proposition expressed is entertained. The declarative is distinguished from the imperative on their account by the claim that the former presents the proposition expressed as being entertained as a description of an actual or possible state of affairs, while the latter indicates the proposition expressed is being entertained as a description of a potential and desirable state of affairs. Saying that P, in the non-illocutionary sense, is then defined as communicating that P is entertained as a description of the actual world, and telling someone to P as communicating that P is entertained as a description of a potential and desirable state of affairs. What is not specified by mood, however, is who is doing the
126
Assertion
entertaining. Thus in saying that P, the speaker merely indicates that P is entertained as a description of the actual world, by somebody, possibly the speaker, but possibly ‘by the person or type of person whose thought is being interpreted in the second degree’ (Sperber and Wilson 1986/1995: 247). Only if the person presented as entertaining this representation as a representation of the actual world is the speaker herself is she asserting that proposition. In telling the hearer to P, the speaker is presenting that state of affairs as potential and desirable to someone, possibly the speaker, but perhaps the hearer. Again, only if the state of affairs represented is presented as desirable to the speaker herself does she issue a directive. The lack of an encoded link between mood and speaker attitude has a number of advantages. It allows, for example, a straightforward explanation of the imperative used to give advice or instructions, as in (31). In such cases, the speaker may be indifferent to whether the state of affairs described is brought about or not. Sperber and Wilson’s account handles these by analysing them as cases of the proposition expressed being entertained as desirable to the hearer.13 (31) A: Excuse me, I want to get to the station B: Take a number 3 bus Despite their appeal, however, attitude-indicating accounts of declarative mood are unattractive from the perspective of the views developed in this book in that they treat interpretation as necessarily a matter of belief ascription, rather than allowing this to be a possible strategy available to sophisticated communicators, as suggested in Chapter 3. It was noted there that an account of assertion in terms of belief attribution requires that this ability be developmentally prior to assertion. As was also noted in that Chapter, the developmental evidence does not support this picture. There are also specific objections that can be levelled the particular proposals. One problem Bach and Harnish’s account faces that it treats as an indirect speech act any act which is not correctly analysed as having the expressed attitude as a necessary condition. This means that a promise made by uttering an explicit performative, such as (32), must be analysed as being performed by means of a direct assertion, for commissives are not defined in terms of an expression of belief that the proposition expressed by the utterance obtains. It cannot be shown, though, that a direct assertion is being performed in this case (see Jary 2007).
Assertion and Mood 127
(32) I promise never to drink again Directives performed by uttering a declarative, such as (28), repeated below, must also be analysed as indirect acts, as belief expression is not constitutive of this genus of acts. But, whereas one can usually respond to an indirect speech act by addressing both the direct and indirect act, one cannot respond to (36) by saying ‘That’s true. I will’. As we saw in section 4.4.1, accounts that treat the declarative as a sign of assertion – and assertion in terms of commitment undertaking – can treat these as direct speech acts. (28) You’re going to leave on that boat tomorrow Sperber and Wilson see their account as offering a unified account of both embedded and main-clause mood. However, they run into trouble if we consider their treatment of the distinction between declarative and infinitive complements. The distinction they draw between declaratives and infinitives is that the former, according to them, present the proposition expressed as a representation of the actual or a possible state of affairs, while the latter presents the proposition expressed as a representation of merely a possible state of affairs. The problem is that their notion of a possible state of affairs cannot be stipulated to exclude the actual, as this would rule out the occurrence of the infinitive complement in (33): the speaker clearly intends the proposition expressed by the infinitive clause to be a representation of the actual world. (33) I believe you to be the best person for the job However, if the notion of a possible state of affairs that they employ does not exclude the actual world, then they have no obvious explanation for the distinction between (17) and (18), repeated below:14 (17) Peter insists that they are here (18) Peter insists that they be here We must conclude, then, that the standard Relevance Theory account of mood does not manage to distinguish declarative mood adequately, and that attitude-indicating accounts of mood are, in general, problematic when applied to the analysis of the declarative.
128
Assertion
6.4 Conclusion Of the approaches to mood surveyed in this chapter, the most successful is that which treats the declarative as an act of proto-assertion. This is able to deal with main-clause and embedded uses of the declarative, as well as with its non-assertoric uses. In doing so, it allows that the declarative is a sign of assertion in that its deployment constitutes behaviour aimed at indicating commitment and putting that commitment forward for acceptance, rejection or confirmation by the addressee. This is its Dummettian side. Its Gricean side, however, allows that the intentions advertised by the use of the declarative need not be present, requiring that the intention to assert by means of the declarative mood must be inferred. In other words, the declarative is an assertion sign in that it constitutes behaviour of one who intends to assert, but this intention must be present if a particular occasion of use is to be correctly treated as an assertion. It might be objected that Barker’s account, relying as it does on attitude ascription to distinguish mere proto-assertions from assertions proper, is open to the same objections that I have levelled against attitude-indicating accounts of mood. But this in not the case. This is because, on Barker’s account, the production of a declarative sentence is the behaviour typical of one who has assertoric intentions, understood in terms of commitment undertaking, not of those who wish to express beliefs. This allows commitment undertaking to be prior to belief in the order of explanation, and this in turn allows commitment undertaking to underlie the development of the notion of belief in children, as suggested might well be the case in Chapter 3, section 3. But, despite its strengths as a conceptual analysis, Barker’s account of mood offers no direct insight into the contribution of the declarative to the process of utterance interpretation. The challenge, then, is to come up with an account of the mechanics of the interpretation of declarative clauses that respects and complements the insights of conceptual analyses such as Barker’s. This challenge is taken up in Chapter 8, once the relationship between assertion and information structure has been discussed in Chapter 7.
7 Assertion and Main Point
7.1 Introduction In previous Chapters, assertion has been discussed mainly in contrast to other illocutionary acts, such as commanding and questioning. However, the term ‘assertion’ is used not only in opposition to other speech acts, but also to contrast what is intuitively felt to be the main point of an utterance with other propositions that are conveyed by that utterance. This relationship between assertion and main-point status derives from the fact that, as we have seen, assertion is widely held to involve an element of putting forward a proposition for acceptance by the audience. If this is accepted as a core function of assertion, then it is natural to view the proposition that is put forward for acceptance as being the main point of the utterance. This has led some to equate assertion and main-point status. Thus Abbott conceives of the asserted element of an utterance being ‘what is presented as the main point of the utterance – what the speaker is going on record as contributing to the discourse’ (Abbott 2000: 1431). This Chapter begins by showing that the tight relationship between assertion and main-point status that Abbott holds cannot be maintained: there are straightforward cases where the main point of an utterance is not asserted by the speaker. Instead, the claim is made that while main-clause declarative syntax can serve as an indicator of main point status, this can be overridden by contextual considerations. That said, it appears that some explicitly expressed propositions cannot felicitously have main-point status, regardless of the context. These ‘conventional implicatures’, to use the term Potts takes over from Grice, are discussed in section 7.3, where it is suggested that these constructions have the effect of marking a proposition as part of the speakers 129
130
Assertion
overall point, and as part of what she asserts, while also blocking these from having main-point status. They have pointhood, the claim goes, but cannot be main-points. This raises the question of just what pointhood is, and this issue is addressed in section 7.4. Sections 7.3 and 7.4 also return to the issue of presupposition.
7.2 Non-asserted main point A straightforward means of establishing main-point status is to embed an example in a question-response pair. This technique is used by Simons (2007), the examples in (1) being adapted from her (2): (1) A: Who was Louise with last night? B: a. She was with Bill b. Henry thinks she was with Bill c. Henry hinted that she was with Bill d. Henry is convinced that she was with Bill The context established by A’s question requires a response which tells A who Louise was with the previous evening. Response (a) is an assertion which gives the required information. In this case, the proposition expressed by (a) is both asserted and has main-point status. In cases (b) to (d), however, the main point of the utterance is expressed by the embedded clause. This main-point proposition is not asserted in these cases, for the speaker does not take responsibility for its truth. Rather, the speaker asserts the proposition expressed by the main clause: it is the proposition that Henry hinted that Louise was with Bill, say, that the speaker can be held to account for if it should turn out to be false. In cases (b) to (d), she takes no responsibility for the truth of the proposition that Louse was with Bill – in any of these cases she could deny this proposition without contradiction. As Simons points out, the main-point status of embedded clauses is highly context-dependent. In the question-response pair in (2) (Simons’s (3)), it is the asserted proposition expressed by the mainclause that has main-point status, not the proposition expressed by the embedded clause. (2) A: What’s bothering Henry? B: He thinks Louse was with Bill last night
Assertion and Main Point 131
The reason for expressing the main point in an embedded proposition in cases such as those in (1)b, c or d is twofold: on the one hand it enables the speaker to put forward the proposition without taking responsibility for its truth; on the other, it allows her to indicate the source of the proposition she declines to assert. In other words, in contexts such as we find in (1), these constructions serve a hearsay, or evidential, function.1 The point of interest here though is that these examples show that asserted content and main-point status do not necessarily go hand in hand. The examples in (1) are cases where the main point of the utterance is not asserted by the hearer, and where what is asserted by the speaker is not the main point of the utterance. Nevertheless, given the context dependency of main-point status, a defender of the view that asserted content has main point status might reasonably argue that, while the asserted content of (1)b, c or d is not the main point of the utterance, it is presented as such by having been expressed in a main clause. Its actual lack of main-point status would be, on this view, pragmatically coerced by considerations of relevance. This view finds some support in the fact that A’s question in (1) could have been appropriately responded to by (1)b , c or d . In these parenthetical formulations,2 the main-point proposition, although still not asserted, is expressed by the main clause. (1) b . She was, Henry thinks, with Bill c . She was, Henry hinted, with Bill d . She was, Henry is convinced, with Bill This means of reformulation is not so felicitous, however, as an alternative response for (2): (2) B : ?Louise was, he thinks, with Bill last night Thus expressed, B’s response to A’s question in (2) appears to give mainpoint status to the proposition that Louise was with Bill, and to limit the role of the proposition that Henry thinks that she was with Bill to an indication about the source of this information, rather than presenting it as the cause of his botherment. These considerations suggest that, while being asserted is not sufficient grounds for counting as main-point, a proposition’s being expressed by a main-clause is closely associated with its having mainpoint status, as it is with assertion. The link between main-point status and main clause can be overridden by contextual considerations, as is
132
Assertion
the case in B’s responses to A’s questions in (1)b, c and d, but in the absence of such contextual factors, the main-clause will be taken to express the main-point proposition. But will the main-clause proposition necessarily be asserted? As we have seen, it seems reasonable to say that, in the responses in (1), the speaker does assert the proposition expressed by the main clause: it is an explicitly expressed proposition that she undertakes a commitment to and to which, if he does not explicitly reject it, the hearer will also become committed. But what of the proposition expressed by the main clause in the parenthetical reformulations (1)b , c or d ? In these cases, the speaker does not assert the proposition expressed by the main clause: she is not committed to the truth of the proposition that Louise was with Bill that night. Rather, she is committed to the proposition associated with the parenthetical element: in the case of (1)c , for example, she undertakes commitment to the proposition that Henry is convinced that Louise was with Bill than night. How can this observation be squared with the view, defended in Chapter 4, that main-clause declarative syntax is an indicator of assertoric force? Recall that, in that chapter, it was shown how contextual features could serve to limit – or ‘cordon off’ – the consequences of expressing a proposition by means of main-clause declarative syntax. These contextual features might be linguistic or non-linguistic: it might be for example, that the speaker is acting in stage play, or her utterance might be governed by a previous utterance such as ‘Let’s suppose the following for the sake of argument’. Utterances such as the responses (1)b , c or d can be dealt with in a parallel manner. Indeed, a common means of analysing utterances containing parenthetical elements such as these is to treat them as double utterances, with one utterance commenting on or ‘fine tuning’ the interpretation of the other (see Ifantidou 2001: 132–138 and references therein). On this view, these examples are best analysed at consisting of two utterances, the utterance corresponding to the parenthetical element commenting on the utterance of the main clause. We can represent this as follows: (1) b . Henry thinks the following: Louise was with Bill c . Henry hinted the following: Louise was with Bill d . Henry is convinced of the following: Louise was with Bill Represented thus, the use of the declarative in the second utterance is governed by the assertion made in the first utterance. This restricts the
Assertion and Main Point 133
consequences of the second utterance, so that the commitments that would follow from asserting the proposition expressed by that sentence are attributed to subject of the first utterance. What the representation in (1)b , c and d lacks, though, is any indication of which utterance carries main-point status. This is just the representation that the parenthetical construction in (1)b , c & d provides. This discussion suggests that main-clause declarative syntax serves as indication of both main-point status and asserted content: in the absence of indications to the contrary, a proposition expressed by a main clause will be taken to have both assertoric force and main-point status. That assertion and main-point status are not to be identified, though, is shown by the fact that either element can be disassociated from the main clause: main-clause propositions may go unasserted and may fail to express the main point of the utterance. This is not to say, though, that main-point status can be held by any propositional element of a sentence. As Potts (2005) shows, syntactically parenthetical elements (‘supplements’), such as those we find in (1)b , c & d , cannot express the main point of the utterance (in his terminology, cannot be ‘at issue’): (3), would be totally inappropriate as a response to the question ‘Who won the European Championship?’: (3) Spain, who won the European Championship, played the best football. In sum, then, it appears that main-point propositions must be expressed by either a main clause or a sentential element of a main clause. Mainclause declarative syntax has a role in marking main-point status, but this can be overridden by contextual considerations, such that the main point is conveyed by an embedded clause. It may also be that the speaker asserts that main point, but this needn’t be the case: both main-clause and subordinate-clause main points can go unasserted by the speaker. This suggests that while main point-status may be a common element of assertion, it isn’t a necessary one. Sometimes speakers want to make a point without undertaking the commitments that assertion results in. Embedding constructions such as those in (1)b , c & d and parenthetical elements such as those found in (1)b , c & d allow them to do this. This raises the question of how non-asserted propositions associated with, and perhaps even communicated by, an utterance should be characterised. This issue is the topic of the next section.
134
Assertion
7.3 Assertion, presupposition and conventional implicature Related to the notion of assertion as main point is the contrast between what is asserted and what is presupposed. Presuppositionhood, as we saw in Chapter 5, is standardly tested for by taking a sentence S, identifying an entailment E of S, negating S, and then enquiring whether E is entailed by ¬S. If E is felt to be strongly implied by ¬S, then E will often be characterised as a presupposition of S, rather than as a straightforward entailment. Further support for classifying E as a presupposition is obtained if the inference to E is also licensed when S is embedded under a modal operator, transformed into a question or embedded in the antecedent of a conditional. Thus, given (4)a as our S, it would commonly be claimed that (4)b is not a presupposition of (4)a as it is not inferable from (4)d to g. (4)c, by contrast, is often taken to be inferable from (4)d to g, and thus counts as a presupposition of (4)a: (4) a. b. c. d. e. f. g.
S: John kicked his brother E1 : John touched his brother E2: John has a brother ¬S: John didn’t kick his brother ?S: Did John kick his brother? ♦S: Perhaps John kicked his brother If S, Q: If John kicked his brother, he should be punished
For some, such as Levinson (1983), survival of an entailment in these environments is a sufficient condition for the status of presupposition. However, for other authors, the mere fact that an entailment survives negation and other embeddings is only a necessary condition. Kadmon (2000) follows Chierchia and McConnell-Ginet (2000) who require that, in addition to passing the above tests, the proposition in question be presented as taken for granted by the speaker, thereby doing justice to the everyday usage of the term ‘presupposition’. On this view, the non-restrictive relative clause (NRRC) in (3) would not be counted as a presupposition, even though it is inferable from (5)a to d. This is in contrast to Levinson, who includes NRRCs in his list of presupposition ‘triggers’. (5) a. Spain, who won the European Championship, didn’t play the best football
Assertion and Main Point 135
b. Did Spain, who won the European Championship, play the best football? c. Perhaps Spain, who won the European Championship, played the best football d. If Spain, who won the European Championship, played the best football, then they deserved to win Partly in response to observations such as these, Potts (2005) draws a distinction between presuppositions and what he terms ‘conventional implicatures’3 . Like presuppositions, he argues, conventional implicatures survive embedding under negation, modal operators etc.4 However, they differ from presuppositions in a number of respects. For example, while presuppositions are widely held to be backgrounded or have common-ground status (recall the discussion of Stalnaker in Chapter 5), conventional implicatures such as NRRCs, ‘as’-parentheticals and nominal appositives (members of Potts’s class of supplements), do not have this status, according to Potts. He argues for this using the following example (his 4.46): (6) Lance Armstrong survived cancer. a. #When reporters interview Lance, a cancer survivor, he often talks about his disease b. And most riders know that Lance Armstrong is a cancer survivor Potts’s reason for the discourse-infelicity of (6)a is redundancy: the nominal appositive ‘a cancer survivor’ introduces information that is already in the common ground as a result of the prior utterance. The acceptability of (6)b, by contrast, is due, according to Potts, to the first sentence providing an antecedent for the presupposition that is triggered by the factive ‘know’ (i.e. that Lance Armstrong is a cancer survivor). An alternative explanation for this data that does not rely on presuppositions requiring antecedents is proposed at the end of this section. Potts’s second distinction between presuppositions and conventional implicatures is that while the former can be cancelled (Levinson 1983; Kadmon 2000), the latter cannot: the discourse (7) is self-contradictory, while the discourse in (8), the first sentence of which would generally be taken to presuppose that the speaker smokes, is perfectly coherent. (7) Spain, who won the European Championship, didn’t play the best football. # Spain didn’t win the European Championship.
136
Assertion
(8) I haven’t stopped smoking. I’ve never smoked. Potts’s third distinction is that while embedding under speech-act descriptions and belief attributions can ‘plug’ (in the sense of Karttunen 1973) a presupposition, conventional implicatures cannot be so plugged. (9) Jane said/thought that Peter noticed that it was raining. But it wasn’t raining. (10) Jane said that Peter, a convicted terrorist, should be hanged. #But Peter isn’t a convicted terrorist In (9), the presupposition associated with the factive ‘notice’ (i.e. that it was raining) is not entailed by the matrix sentence in which ‘notice’ occurs. Hence the denial of that presupposition does not result in an incoherent discourse. In contrast, denying the conventional implicature associated with the nominal appositive in (10) does lead to incoherence, despite the fact that the sentence to which the appositive is attached is embedded in a presupposition ‘plug’.5 On this view, then, the propositions associated with an utterance of a declarative sentence can be divided into three groups: at-issue entailments (Potts’s term for main-point entailments), conventional implicatures and presuppositions.6 For Potts, both at-issue entailments and conventional implicatures are asserted, while both conventional implicatures and presuppositions survive negation. If Potts is correct, then this is further evidence that what is asserted need not have mainpoint status, for, as we have seen, conventional implicatures (which, on his account, are asserted) cannot serve to convey the main point of the utterance. What his work also shows is that what does not have mainpoint status need not be taken for granted. Rather, there are linguistic means of putting forward information that is new to the hearer but at the same time indicating that this information is not the main point of the speaker’s utterance. Accepting Potts’s picture in its entirety, though, requires accepting that presuppositions can be distinguished from conventional implicatures in terms of taken-for-grantedness. Indeed, intuitions about taken-for-grantedness are what Kadmon (2000: 14) suggests we should employ to distinguish presuppositions from other negation-surviving implications. This entails accepting something like Stalnaker’s commonground view of presupposition, on which presuppositions are propositions that speakers treat as having been accepted by all interlocutors. There is, however, considerable evidence that many presuppositions are
Assertion and Main Point 137
incorrectly characterised in these terms. Abbott (2000), for example, details a range of cases where presupposition triggers such as definite descriptions and cleft sentences are employed to communicate information that is new to the interlocutors, and Simons (2007) gives cases of factive predicates used in the same way. Of course, the notion of presupposition accommodation can be called on to account for these phenomena, but, as we saw in our discussion of Stalnaker in Chapter 5 this is highly problematic. What I want to propose, therefore, is that we take Potts’s tests for distinguishing conventional implicatures from presuppositions as tests for pointhood. The idea is that, as Kadmon (2000: 13) suggests, we take survival under negation and other embeddings as a test for foregrounding, or main-point hood: if an implication fails to survive negation, then this indicates that it is presented by the sentence as part of the main point of the utterance (even though it may not actually be the speaker’s main point). For the moment, those implications that survive negation can then be sorted, using Potts’s tests, into presuppositions and conventional implicatures. Conventional implicatures, while they do not have main-point status, are presented by linguistic form as part of the point of the utterance: they have pointhood. Pointhood is what, to use Potts’s terminology, at-issue entailments and conventional implicatures have in common. They differ in that only the former are presented as having main-point status. What are generally termed the ‘presuppositions’ of a positive assertion are entailments that are not presented as part of the utterance’s point: they lack pointhood. But an absence of pointhood need not entail being taken for granted, even though takenfor-granted propositions will often lack pointhood. And, moreover, the fact that an entailment is not presented by linguistic form as part of the speaker’s point, need not mean that it does not in fact contribute towards that point. On this view, (6)a is anomalous not because a presupposition requires an antecedent, as Pott’s suggests, but because the speaker makes the same point twice for no discernible effect.7 (6)b is acceptable, by contrast, because the proposition that Lance Armstrong is a cancer survivor is not presented as part of the speaker’s point in the second utterance. The view just outlined has the advantage of not requiring that presuppositions have common-ground status and hence avoids having to invoke the problematic notion of accommodation to cope with counterexamples. It does, however, rely on the vague, pre-theoretic notion of pointhood. In the next section, we will see how this notion can be rendered more concrete by employing Sperber and Wilson’s notion of relevance.
138
Assertion
7.4 Ordered entailments For Sperber and Wilson (1986/1995), pointhood is key to understanding the process of linguistic communication. According to their theory, every utterance conveys a presumption of its own relevance. Interpretation is a process of inferring how that presumption is to be fulfilled, i.e. of working out the point of the utterance. Relevance, on Sperber and Wilson’s model, is a matter of maximising effect and minimising effort. Effects can be thought of, for present purposes, as implications: the more (non-trivial) implications an utterance has, the greater its effect. For any given level of processing effort, the greater the effect, the greater the relevance. For any given level of effect, the less effort is expended in deriving that effect, the greater the relevance. Relevance can thus be increased by decreasing effort, increasing effect, or both.8 Utterances convey a number of ‘assumptions’ (Sperber and Wilson’s term for a mentally represented propositional form, see section 4.2 for discussion). In any given context, some of these will have more implications – and hence greater effect – than others. Assuming, for the moment, that each of these assumptions can be derived with the same degree of effort, the most relevant in that context will be those that have the largest number of implications. On the Relevance-Theory view, the context is not given prior to the utterance but is constructed online, as comprehension proceeds Therefore, a means is needed of indicating both in which context an utterance should be processed and which assumptions should be added to that context so that implications can be derived. Only if both of these necessary elements of interpretation are successfully identified by the hearer will he ascertain how the utterance is intended to achieve relevance, or – in theory-neutral terms – what the point of the utterance is. Consideration of the means by which intended context and intended effects are indicated can shed light on the phenomena that we are considering in this chapter. The basic idea is that the entailments of an utterance can be divided into those that are grammatically specified and those that are not, and that the set of grammatically specified entailments can then be classified according to whether their role in the interpretation process is to indicate in which context implications should be derived, or whether they should be used as premises in the inferential process of deriving those effects. For an assumption to play a context-indicating role is for it to play an activation role, making information stored in memory more accessible and thereby reducing processing effort. For an assumption to play a role in the derivation of
Assertion and Main Point 139
effects is for it to serve as a premise in the inferential process of implicature derivation. These roles are not exclusive: some assumptions will play both roles. An assumption that serves as premise in the derivation of implicatures is relevant in its own right. An assumption that serves only to activate contextual assumptions contributes to the relevance of an utterance only indirectly. Such an assumption is not relevant in its own right: it is not part of the point of the utterance. (Sperber and Wilson 1986/1995: 209) The set of grammatically specified entailments associated with a sentence can be identified by replacing the syntactic constituents of that sentence with ‘some’-forms. So, the grammatically specified entailments of (11)a will include (11)b to f.9 (11) a. b. c. d. e. f.
John broke the window John did something Someone did something Someone broke the window Someone did something Something is the case
Sperber and Wilson argue that the ordering of these entailments mirrors the syntactic structure of the sentence uttered, with a variable corresponding to each node on the sentence’s tree, as illustrated in 7.1. Linguistic structure gives an indication of whether a proposition is presented as relevant in its own right. As (11)a is processed (assuming, for the moment, unmarked intonation and stress placement), the entailments will be derived in the order specified by the tree. Thus the
S/something is the case
NP/someone
John
VP/did something
broke
NP/something
the window Figure 7.1 Ordered entailments
140
Assertion
order in which entailments will be derived is as in (12). Note that each entailment on the list entails those above it. (12) a. b. c. d.
Something is the case John did something John broke something John broke the window
Generally, the later an entailment is derived, the more likely it is to be relevant in its own right. This is because, given the view of communication outlined above (in particular, the need to construct the context as interpretation takes place), it makes sense to activate the context in which implicatures are to be derived before introducing new premises into that context. Of course, this echoes the oft-noted observation that ‘old’ information tends to precede ‘new’ information in discourse. Sperber and Wilson call a grammatically specified entailment that is relevant in its own right (i.e. one that constitutes part of the point of an utterance) a ‘foreground implication’. Others are ‘background implications’. If stress is unmarked, so that it falls at the end of the sentence (i.e. ‘window’ in (11)), then no indication is given by the speaker of where the line between background and foreground implications should be drawn. This is not to say, though, that no indication can be given. Contrastive stress, Sperber and Wilson argue, is a means by which speaker’s indicate the foreground/background divide. Thus if the speaker were to utter (11) with stress on the first NP, this would indicate that the entailment derived by replacing that constituent with a variable (i.e. (13)b) is one which would activate a context in which the utterance would achieve relevance. This is illustrated by the fact that the stress pattern in (13)a would be an appropriate response to (13)c, which takes for granted that someone broke the window (Note that contrastive stress in listed by Levinson 1983: 183 as a presupposition trigger). (13) a. JOHN broke the window b. Someone broke the window c. Who broke the window? Note, though, that this position does not require that the proposition indicated as the background to the utterance is common ground, nor that the speaker assumes it to be or acts as if she does. As just noted, (13)a would be relevant as an answer to (13)c. It might, therefore, be the case that what the speaker is indicating is that her utterance should be
Assertion and Main Point 141
interpreted an answer to a particular question. Consider, for example, the dialogue (14): (14) A: B: (15) a. b. c.
Who should go last in the relay? JOHN did a good job last time Someone did a good job last time A person who did well once is likely to do so again John should go last again
B’s response in (14) would be quite acceptable even if (15)a, its indicated background implication, were not common ground. (15)a is likely to activate the assumption (15)b, which, in combination with the proposition expressed by B’s reply in (14), would make it possible to derive the conclusion (15)c. In this case, though it is indicated as a background implication, (15)a also functions as a foreground implication. Its background role is to indicate how information should be used to activate encyclopaedic information in order to facilitate the interpretation process. However, there is no requirement that the proposition that activates that information should itself be ‘old’, or ‘given’. As Sperber and Wilson (1986/1995: 210) note, their foreground/background distinction is a functional one, related only indirectly to the linguistic marking of new and old information. One question Sperber and Wilson’s analysis raises is what the role of declarative mood is and how it relates to main-pointhood. They claim that the proposition explicitly expressed by the main clause of a declarative sentence is ‘necessarily a foreground implication’ (1986/1995: 210, when discussing their example (64c)). However, this cannot be the case. Whether an assumption counts as a foreground implication or not depends on whether it is relevant in its own right (i.e. whether it serves as a premise in the derivation of implicatures). This will of course depend on the context. It may be, for example, that the speaker is mistaken and thinks that the assumption that she explicitly expresses will be relevant to a speaker, but that this is not in fact the case because the speaker has already derived the implications that she intended to communicate. This would be a case of miscommunication. Alternatively, it may be the case that the speaker does not intend the proposition explicitly expressed by the utterance to achieve relevance in its own right. This would be the case, for example, if (16)a were uttered in order to communicate (16)b: the speaker would not generally expect the hearer not to know that he had had his hair cut.
142
Assertion
(16) a. You’ve had your hair cut b. I notice that you’ve had your hair cut Thus the fact that a proposition is explicitly expressed cannot guarantee that it is a foreground implication. However, just as we noted earlier that the observation that a main-clause proposition is not in fact the main point of an utterance does not preclude the claim that it is presented as such by the linguistic resources employed, we should also note here that the observation that a proposition does not achieve relevance in its own right does not preclude the claim that it is presented as such by being explicitly expressed by the main clause of the utterance. Thus what Sperber and Wilson mean, I think, when they say that the proposition explicitly expressed is necessarily a foreground implication is that it is necessarily presented as such. The notions of foreground and background are interdependent and this requires that there be at least one foreground implication. Background entailments must be entailed by the foreground, so the only entailment that is necessarily foregrounded is the maximally entailing entailment, i.e. the proposition explicitly expressed. However, we should be careful not to stipulate that a proposition’s being explicitly expressed by a main clause equates with it being presented as relevant in its own right. As Carston (2002: 120) has noted, the propositions explicitly expressed by non-declarative sentences such as (17)a and (18)a are not communicated by utterances of those sentences. Rather, these propositions contribute to the relevance of an utterance only indirectly, by serving as constituents of ‘higher-order explicatures’ such as (17)b and (18)b. (17) a. b. (18) a. b.
Stand up! The speaker is telling the hearer to stand up Do you want a drink? The speaker is asking the hearer whether he wants a drink
It seems, then, that it is being explicitly expressed by a declarative sentence that presents a proposition as relevant in its own right. This idea is developed in the next chapter, where it is shown that thinking of declarative mood in these terms provides a straightforward account both of its main-point status and its role as an assertion sign. Before progressing to this, however, we need first to consider the status of the entailments of an utterance that are grammatically unspecified. This is important as it will relate what has been said in this section
Assertion and Main Point 143
to the discussion of conventional implicature and presupposition in section 7.3. Grammatically unspecified entailments can be divided into three types: those that entail the highest background entailment, those that are entailed by the highest background entailment and those that are neither (Wilson and Sperber 1979: 318). For example, let us assume that (19)a is uttered with unmarked stress, and that the highest background entailment is (19)b. This background entailment is also entailed by (19)c, which is a non-grammatically-specified entailment of (19)a. (19)d, by contrast, is entailed by (19)b, while (19)e is neither entailed by nor entails (19)b. (19) a. b. c. d. e. f.
My daughter is eating an apple My daughter is eating something My daughter is eating a piece of fruit I have a daughter Someone is eating an apple My daughter is not eating an apple
Non-specified entailments that entail the background, such as (19)c are examples of what Potts calls ‘at-issue entailments’. Like the proposition explicitly expressed, they can fall under the scope of negation: the speaker of (19)f would not normally be taken to be committed to (19)c (though, importantly, nor would he necessarily be taken to be denying (19)c either). The terms ‘at-issue entailment’, though, is somewhat misleading as it suggests that entailments such as (19)c must be part of the point of the utterance. However, this is not the case. Unlike the proposition explicitly expressed by the utterance, which is linguistically presented as contributing to the point of the utterance, there is no guarantee that other ‘at-issue entailments’ are part of the speaker’s point. To see this, compare (20)a spoken to an observant Jew and to a vegetarian. In the first case, the entailment (20)b would most likely not contribute to the point of the utterance, while in the second case it most certainly would (cf. Wilson and Sperber 1979: 318).10 (20) a. You’re eating pork b. You’re eating meat Non-specified entailments that are themselves entailed by the background, such as (19)d, potentially play a context-activation role and would not normally be relevant in their own right. In a case where (19)d
144
Assertion
was actually more relevant to the hearer than the assumption presented as such by the utterance of (19)a, there would be strong intuitions of infelicity, for it has not been presented as relevant in its own right. This would be the case if (21)a were uttered to the speaker’s parents, who were completely ignorant of (21)b (this example is adapted from von Fintel 2000: 16): (21) a. My fiancé and I are moving to Seattle next week b. The speaker is engaged to be married Note, though, that spoken to someone less intimately related to the speaker who was also ignorant of (21)b, an of utterance (21)a could well be perfectly felicitous. Imagine, for example, the speaker was explaining to the new tenants of a flat she was giving up her reason for moving. In such a case, (21)b would make a minor contribution to the relevance of the utterance in its own right, but not be more relevant to those hearers than the proposition presented as relevant in its own right, i.e. that expressed by the main clause. This nicely illustrates the fact that, on Sperber and Wilson’s view, any entailment of the utterance can potentially contribute to the relevance of the utterance in its own right, i.e. be part of the point of the utterance. The constraint is that a relevant entailment must not be more relevant than the maximal entailment explicitly expressed by the main clause (unless, like (20)b it entails the potential background entailments itself). Is the speaker of (21)a taking (21)b for granted? It is certainly the case that she is not presenting it as her main point. But it does not follow from this that she is taking it for granted. Moreover, whether she is doing so or not is not really a question for linguistics. This issue is, rather, whether she is presenting, by linguistic means, (21)b as being taken for granted. If failure to survive negation is taken merely as a test for being linguistically presented as main point, as Kadmon suggests, then surviving negation does not show that a proposition is indicated as being taken for granted. All it shows is that it is not presented as main point. The contrast between (21)a spoken to an intimate and to a non-intimate shows that the acceptability of communicating (21)b by means of (21)a depends on how relevant (21)b will be to one’s audience: if it will prove more relevant than the explicitly expressed proposition, which is the only proposition that is presented as relevant in its own right, then the utterance will be defective. Otherwise, it will not. Moreover, on this view, there is no need to posit a process of
Assertion and Main Point 145
presupposition accommodation to account for cases such as the use of (21)a to communicate (21)b. The final set of non-specified entailments are those that neither entail nor are entailed by the highest background entailment. In their earliest paper on this topic, Wilson and Sperber (1979: 319) claim that such entailments ‘should normally play no part in the interpretation of the utterance’. Unqualified, however, this statement is too strong (as the authors later tacitly concede). It predicts, for example, that (22)a should be a much more marked way of communicating (22)e than (22)c, because (22)e does not entail, nor is entailed by, the background (assuming (22)b to be the highest background entailment of (22)a), whereas (22)e is at least entailed by (22)d, which we can assume to be the highest background entailment of (22)c. My intuition, though, is that neither (22)a nor (22)c is a more or less marked way of conveying (22)e, though in neither case, of course, could it be said that (22)d is presented as the point of the utterance. (22) a. b. c. d. e.
I need to phone my husband The speaker needs to phone someone My husband works in banking The speaker’s husband works in something The speaker is married
Discussing a similar case, Wilson and Sperber (1979: 320) argue that what makes (22)a an acceptable means of conveying (22)e is that it constitutes part of the increment of information that must be added to the background (22)b in order to arrive at (22)a and might thus play some part in establishing the relevance of the utterance. This allows, then, that information that is neither entailed by nor entails the background can contribute to the relevance of the utterance. Wilson and Sperber offer an analysis of factive verbs which parallels their explanation of cases such as that just discussed. Factive predicates have long been felt by many to take for granted, rather than put forwards for acceptance, the proposition expressed by their complements. (23) a. b. c. d. e.
Joan regrets that she lost her hat Joan is sad that she lost her hat. Joan realized that she lost her hat Joan knows that she lost her hat Joan lost her hat
146
Assertion
According to Wilson and Sperber, because (23)e is a grammatically unspecified entailment of (23)a to d that neither entails nor is entailed by ‘Joan regrets/is sad that/realized/knows something’, it should play no part in the normal interpretation of (23)a to d. However, it has been argued by Simons (2007) that the complement clauses of factive constructions can carry the main point of an utterance. She gives the following examples (her (21) and (22)): (24) Where did Louse go last week? a. Henry discovered that she had a job interview in Princeton b. Henry learned that she had a job interview in Princeton c. Henry found out that she had a job interview in Princeton (25) How will Louise get to the picnic? a. Henry realized that she can take a bus b. Henry figured out that she can take a bus These examples use the question-response technique to establish mainpointhood and demonstrate that the main point of an utterance can be conveyed by a clause embedded in a factive construction. The following attested example, familiar to those who have bought books in the UK through Amazon, is another case where a clause embedded in a factive clearly conveys the main point of the utterance, i.e. that the items ordered have been sent. (26) We thought you would like to know that the following items have been sent to . . . There seems, then, no block on embedded clauses such as these conveying the main point of the utterance (which, I stress, is not to say that these sentences present the proposition expressed by the embedded clause as the main point). Moreover, they can do so without the ‘slight figurative overtones’ that Wilson and Sperber (1979: 321) suggest should result in such cases. It is not the case, though, that all factives can serve this function, though: (27) could not be used to respond felicitously to (24) (27) Henry is sad that she had a job interview in Princeton
Assertion and Main Point 147
In Chapter 8, I provide an explanation of why this is the case. The point for now is that not entailing nor being entailed by the background does not preclude a proposition from playing a role in the interpretation of an utterance. Such an entailment, by definition, must be entailed by the foreground and can therefore potentially play a role in contributing to the relevance of an utterance. Moreover, while not linguistically marked as carrying the main point of an utterance, entailments explicitly expressed by embedded clauses can achieve main-point status, just as non-entailed but explicitly expressed propositions can, as we saw in section 7.2.11 Potts’s supplements are like the proposition expressed by the main clause of an utterance in that they are marked as contributing to the relevance of the utterance in their own right. However, they are syntactically marked as not having main-point status. This is why, unlike the proposition expressed by an embedded clause, an interpretation on which they the proposition expressed by a supplement such as an NRRC cannot be coerced by pragmatic considerations into having main point-status. Where does this leave the distinction between assertion and presupposition? On the account presented here, the notion of presupposition does not pick out a linguistic category. Rather, it has been used to capture two related but independent phenomena, one linguistic, the other discourse-based. On the linguistic side, there are entailments that survive negation. On the discourse side, there are assumptions we intuitively feel to have been taken for granted by speakers. However, surviving negation has been shown not always to coincide with intuitions about taken-for-grantedness. Furthermore, whether an entailment survives negation is a context-dependent matter, for, as has been noted, it has long been established that putative presuppositions are cancellable without contradiction in negative sentences. It should also be further stressed that it is a matter of context whether a supposed ‘at-issue entailment’ that is not explicitly expressed by the main clause falls under the scope of negation: there is no linguistic guarantee that in uttering (28)a, the speaker intends to deny (28)b, just as there is no commitment to (28)c that cannot be cancelled without contradiction: (28)c can be denied by the speaker of (28)a, just as (28)b can be asserted. (28) a. b. c. d. e.
He didn’t give his wife ROSES He gave his wife flowers He has a wife He gave his wife roses He gave his wife something
148
Assertion
On the account put forward in this section, the reason that we have distinct intuitions about the respective statuses of (28)b and (28)c in relation to (28)a is that (28)b entails a background entailment of the assertion that (28)a denies (i.e. (28)d, of which a likely background assumption is (28)e). (28)c, by contrast, is entailed by that background assumption, and so (28)a is not an appropriate means to deny that the individual in question has a wife. In other words, a denial rejects an element of the point of the associated assertion, not its background. So, while one possible interpretation for (28)a is that the speaker is denying (28)d because (28)b does not hold, the interpretation that she is uttering (28)a to deny (28)c, another entailment of (28)d, is far less accessible, even though the denial of (28)c is truth-conditionally compatible with the utterance of (28)a. As noted above, though, there is no need to invoke common-ground arguments here: all that is needed is a functional distinction between propositions that serve to activate contents and those that serve as premises in the derivation of contextual effects, that is, a distinction between foreground and background assumptions, understood in the sense employed in Relevance Theory.
7.5 Conclusion This chapter started by highlighting evidence that undermines a simple equation of assertion and main-pointhood. The claim made was that while main-clause declarative syntax presents the proposition expressed as the main point of the utterance, in some cases this can be overridden by contextual factors. It was then noted that certain explicitly expressed propositions – members of Potts’s class of supplements – cannot be contextually coerced into taking main-point status. What these elements have in common with propositions expressed by a main clause is that they are presented as contributing to the point of the utterance; how they differ is that they are marked as not having main-point status. This raised the question of just what pointhood is, and this was elaborated on in section 7.4, where it was argued that those propositions that contribute to the point of an utterance are those that are relevant in their own right. In the case of affirmations, presuppositions are entailments that are not presented as fulfilling this role. The linguistically indicated point of denials and questions is to deny or question the linguistically indicated point of the affirmative sentence to which they are related. This explains the intuition that foreground entailments do not survive embedding in such environments.
Assertion and Main Point 149
Thus, on the account presented in this chapter, rather than a distinction between assertion and presupposition, there are distinctions between what is explicitly conveyed by an utterance and what is implicitly conveyed, what is linguistically marked as contributing to the point of the utterance and what is not, what is marked as having main-point status and what is not, and what actually contributes directly to the point of an utterance and what does not. As has been indicated in this chapter and will be demonstrated in the next, these distinctions can be articulated clearly within the Relevance-Theory framework. There it will also be shown how declarative syntax can be seen as an indicator of both assertoric force and main-pointhood without equating the two.12
8 Assertion and Relevance
8.1 Introduction In previous chapters, I have surveyed a range of literature relating to assertion. One of the motivations for this has been to lay the ground for an account of the pragmatics of assertion, focussing on the role of the declarative mood in conveying assertoric force. Providing such an account is the aim of this chapter. My goal is to provide an explanation in cognitive terms of the conditions under which the use of the declarative results in assertoric force, and those in which it does not. The account is designed to cover both main-clause and embedded uses of the declarative. The chapter is organised as follows. I begin by outlining the features of Relevance Theory that are crucial to my account. The first is a distinction between relevance in a context and relevance to an individual. This is crucial because I will claim that what gives declarative mood its assertoric potential is a capacity for sentences of this type to present the proposition expressed as relevant in its own right in a context. Assertoric force is then explained in terms of a speaker using such a sentence to present the proposition expressed as relevant in its own right to an individual (i.e. the hearer). For this to happen, a certain type of context must be accessible. I therefore go on to distinguish two types of context – embedded and factual – and argue that assertion is a means of changing the latter. Next follows a detailed account of the interpretation of mainclause and embedded declaratives. This leads to a discussion of the role of belief-attribution in assertion, which in turn raises the question of whether it is necessary to posit that utterance interpretation is domain of a dedicated cognitive module. I end by relating this cognitive account of assertion to the philosophical treatments of assertion discussed earlier in the book. 150
Assertion and Relevance 151
8.2 Relevance in a context vs. relevance to an individual According to Sperber and Wilson (1986/1995), the notion of relevance is central to an understanding of linguistic communication. An individual attends to an utterance, the theory has it, because all utterances come with a presumption that the information it conveys will be relevant to that individual. This cognitive notion of relevance to an individual is grounded, however, in a more fundamental notion of relevance in a context. Consequently, we will first look at this purely formal notion of relevance. 8.2.1 Relevance in a context A context is defined here as a set of propositions linked to a deductive device. The deductive device derives all the non-trivial implication of the propositions in the context, adding these to that context. A proposition P is relevant in a context C iff the addition of P to C results in the derivation of non-trivial implications. That is the classificatory notion of relevance in a context. Relevance can also be defined in a comparative sense, so that the relevance of two propositions P and Q in C can be compared. Whether P is more relevant than Q in C, however, does not depend only on the number of implications that P and Q have in C. Because the formal notion of relevance in a context is to serve as a basis for a cognitive notion of relevance to an individual, the cost of deriving implication needs to be taken into account, for, once this notion of relevance in a context is instantiated in a physical system, working out these implications is not a cost-free activity: effort is involved. Consequently, when comparing the relevance of propositions in a given context, we take into account the effort involved in deriving these. We therefore say that, for any given level of effort, a proposition P is relevant in C to the extent that it results in non-trivial implications.1 If, for the same degree of effort, P has more contextual implications in C than Q, then P is more relevant in C than Q. Alternately, if P has the same number of implications in C as Q, but the effort involved in deriving these implications is less in the case of P, then P is more relevant than Q in that context. 8.2.2 Relevance to an individual The formal notion of relevance in a context provides the basis for the notion of relevance to an individual. As in the contextual notion, relevance to an individual can be defined in terms of contextual effects, but now these effects must be of a certain type: they must lead to an overall improvement in that individual’s representation of the world. In other
152
Assertion
words, they must lead to an increase in his stock of true beliefs if they are to be relevant to that individual. Contextual effects of this kind – that take place in the cognitive system of an individual and result in an improvement in the functioning of that system – are termed by Sperber and Wilson ‘positive cognitive effects’. Relevance to an individual is thus defined not in terms of mere contextual effects but in terms of positive cognitive effects (the following definitions from Sperber and Wilson 1986/1995: 265–6): Relevance to an individual (classificatory) An assumption is relevant to an individual at given time if and only if it has some positive cognitive effect in one or more of the contexts accessible to him at that time. Relevance to an individual (comparative): Extent condition 1: An assumption A is relevant to an individual to the extent that the positive cognitive effects achieved when it is optimally processed are large. Extent condition 2: An assumption A is relevant to an individual to the extent that the effort required to achieve these positive cognitive effects is small. The distinction between relevance in a context and relevance to an individual is very important to the account put forward in this chapter. The claim that I want to make is that declarative mood, whether main-clause or embedded, is unique among the moods in presenting the proposition expressed as relevant in a context. This enables the declarative to be used to present the proposition expressed as relevant to an individual, the result of which is often, but not necessarily, to imbue the utterance with assertoric force. In order to develop this proposal, I need now to look at the notion of context, as used in Relevance Theory, more closely.
8.3 Context types As far as the formal notion of relevance in a context goes, the context can be thought of as being made up of propositions. When we consider the cognitive notion of relevance to an individual, though, a more concrete notion is needed. Sperber and Wilson, as we have noted, call mentally represented propositional forms ‘assumptions’, and the contexts in which the propositions communicated by such forms
Assertion and Relevance 153
are processed are made up of assumptions. These assumptions come in a number of types. The most basic type is a ‘factual assumption’. This has what Geach (1965) calls ‘assertoric character’, in that to represent a thought in this format is necessarily to treat it as true, one’s behaviour thereby being guided by that thought. Because they constitute the individual’s basic, unreflective, representation of the world, factual assumptions are truth-aiming. The means by which the cognitive system seeks to ensure that its factual assumptions are true is by aiming for consistency. Importantly, this means that the format in which these assumptions are represented must be such that they are able to serve as premises in deductive processes designed to ensure consistency. Factual assumptions are thus consistency aiming. According to Sperber and Wilson, the human cognitive system is also able to represent assumptions in a more reflective manner, as constituents of factual assumptions. Call these ‘embedded assumptions’. Thus the proposition P might be entertained by an individual in a number of ways, as illustrated by (1): (1) a. b. c. d. e.
P Peter desires that P Jane believes that P In the story, P I conjecture that P
Being unembedded, in (1)a, P is a factual assumption. In (1)b to e, by contrast, P is an embedded assumption. Thus represented, it loses its assertoric character and no longer necessarily has an effect on the behaviour of the individual in its own right. There are, however, important differences in how P is represented in each of these embedded cases. (1)b differs from (1)c to e in that P in (1)b is not consistency aiming. As previously noted, desires – unlike beliefs, stories and conjectures – do not aim at consistency. The assumptions embedded in (1)c to e are thus consistency aiming, in contrast to the assumption embedded in (1)b. Of those that are consistency aiming, though, only in (1)e is the embedded assumption also truth aiming. Belief attributions and thoughts about fiction are not necessarily faulty if the embedded propositions that represent the contents of the attributed belief or fiction are false. Conjectures, though, must be abandoned if shown to be false. This taxonomy of propositional mental representations is illustrated in Figure 8.1 (bear in mind that factual assumptions necessarily aim at both consistency and truth):
154
Assertion
Assumptions
Factual
Embedded
Aim at consistency
Aim at truth Figure 8.1
Don’t aim at consistency
Don’t aim at truth
Types of assumption
Non-consistency-aiming mental representations are excluded from serving as contextual assumptions on the grounds that they are not of the correct format to serve as premises in inference, the means by which consistency is established. In Hare’s (1950/1971; 1970/1971; 1989) terms, they to not have the correct ‘tropic’ (see Chapter 4). This means that there can be three types of context in which an assumption might be relevant: a context formed of factual assumptions, a context formed of truth-aiming embedded assumptions, and a context formed of non-truth-aiming, embedded assumptions. But, as the distinction between the latter two will rarely be pertinent to the discussion that follows, we can reduce the number of context types that we will consider to two: factual and embedded contexts. Embedded contexts come into play in the representation of, for example, the beliefs of others or the contents of a novel. Such contexts mimic factual contexts in that, like a factual context, they aim at accurate representation of a world. While a factual context aims at an accurate representation of the actual world, an embedded context aims at an accurate representation of the actual world according to someone, in the case of belief reports, or of another world altogether, in the case of fiction. In both factual and embedded cases, this aim is to be achieved by seeking consistency in the assumptions that constitute that representation. Of course, in reasoning about the content of a novel or the beliefs of others, we often make use of factual assumptions combined with embedded assumptions. Thus our beliefs about how people normally
Assertion and Relevance 155
behave in certain situations will certainly play a role in our interpretations of a fiction, for example. This requires that embedded and factual assumptions can form part of the same context. This can be allowed for in the current model by stipulating that any conclusions drawn from such reasoning must be embedded in the most embedded assumptiontype from which the context has been formed. This ensures that factual assumptions are not polluted by the contents of a fiction, say, while still allowing that factual assumptions can play a role in the interpretation of fiction. So, a factual context is a context formed of only factual assumptions. An embedded context is a context in which at least one contextual assumption is an embedded assumption. Any change to a factual context directly modifies the individual’s basic, non-reflective representation of the world. Changes to embedded contexts, by contrast, change his fundamental assumptions about the actual world only indirectly, in that they modify a set of assumptions embedded in his assumptions about the world. Embedded assumptions cannot serve as premises in inferences aimed at improving his basic, non-reflective, representation of the world.
8.4 Mood and force Given the distinction that has just been drawn between embedded and factual contexts, we can now consider the role of the declarative mood in the interpretation of utterances and then relate this to assertoric force. The claim I make about the declarative mood, understood as applying to both main-clause and embedded instances of this form, is that it presents the proposition explicitly expressed as relevant in its own right in an accessible context. Importantly, it is the form, not the speaker, that presents the proposition as relevant in this way. Use of the form is thus no guarantee that the proposition expressed is actually relevant in its own right, nor that the speaker intends it to be so. It is, however, an indication of where the relevance of the utterance lies, one that needs to be taken into consideration in the interpretation process. In section 7.4, a proposition was said to be relevant in its own right if it serves as a premise in the derivation of implicatures. This characterisation of a proposition/assumption having relevance in its own right was sufficient for our purposes at that point, but it needs to be revised now in the light of the distinction drawn above between relevance in a context and relevance to an individual. Consequently, we can now say that, for any given level of effort, an assumption is relevant in its own right in
156
Assertion
a context iff it serves as a premise in the derivation of contextual implications; and we can say that an assumption is relevant to an individual in its own right iff it serves as a premise in the derivation of positive cognitive effects. Declarative mood indicates an assumption’s relevance in the former sense, i.e. in its own right in a context. Assertoric force, we will soon see, is a possible consequence of the speaker exploiting this feature of declarative sentences to present the proposition expressed as relevant in its own right in a factual context, and thus to the hearer. So, on the theory presented here, assertion is a possible result of a speaker presenting a proposition as relevant in its own right to an individual. In order to clarify what it means for a proposition to be relevant in its own right, it is worth spelling out what it means for a proposition to contribute to relevance by other means. As has just been said, a proposition is relevant in its own right in a context iff it serves as a premise in the derivation of contextual implications. A proposition contributes indirectly to relevance, by contrast, if it is embedded in proposition that contributes directly (i.e. that is relevant in its own right), or if it serves to make the necessary contextual assumptions accessible to the hearer. An example of the first type of indirect contribution to relevance is to be found in the utterance of imperatives and interrogatives to perform directives and ask questions. As was noted in Chapter 7, Carston (2002: 120) has pointed out that in cases such as (2) and (3), the proposition explicitly expressed by the sentence is not communicated by the utterance. Rather, what are communicated are the higher-order propositions (2)b and (3)b, the proposition expressed by the sentences uttered being embedded in these. It is from the higher-order representations that contextual implications are derived. In these cases, then, the proposition expressed by the utterance does not contribute to relevance in its own right, but only indirectly, as a constituent of a higher-order representation. (2) a. b. (3) a. b.
Stand up! The speaker is telling the hearer to stand up Do you want a drink? The speaker is asking the hearer whether he wants a drink
The second way in which a proposition can contribute to relevance indirectly (i.e. not in its own right) is by serving to reduce effort by making contextual assumptions accessible. As was explained in Chapter 7, the Relevance-Theory analysis of contrastive stress sees this as marking a distinction between the background and foreground entailments of an
Assertion and Relevance 157
assertion. On this view, the intonation pattern in (4)a indicates that (4)b is a background entailment of the utterance, defined in terms of its function of making accessible contextual assumptions in which the foreground entailments of the utterance can be processed. Let us assume that the assumption (5) is mutually manifest to speaker and hearer. If this is the case, then the function of the background entailment (4)b is to make (5) more manifest so that it may be used as a contextual assumption for the processing of (4)a, allowing the hearer to indentify (6) as the key implicature of the utterance. Note that, while (4)b plays a crucial role, it does not serve as a premise in the derivation of the implicature (6): there are only two premises required: (4)a and (5). (4)b contributes to the relevance of the utterance by making a contextual assumption more manifest, not by serving as a premise. It thus contributes to relevance, but not in its own right. (4) a. JOHN gave Mary a lift home b. Someone gave Mary a lift home (5) The person who gave Mary a lift home murdered her (6) John murdered Mary It should not be thought, though, that the proposition explicitly expressed by an assertion must be relevant in its own right. As was also noted in Chapter 7, the speaker may not intend the proposition explicitly expressed by the utterance to achieve relevance in its own right. The example given there was of (7)a being uttered in order to communicate (7)b. The speaker would not generally expect the hearer to be unaware that he had had his hair cut, and thus the relevance of the utterance would lie in a higher-order representation of its content rather than in its content itself. In such a case, the proposition explicitly expressed by the declarative sentence would not be relevant in its own right. (7) a. You’ve had your hair cut b. I notice that you’ve had your hair cut On the current analysis, (7) is a case in which, while the form chosen marks the proposition expressed as relevant in its own right in a context, considerations of the relevance of the utterance to the individual lead to the proposition not achieving relevance in this way. This distinction will be important in the remainder of this section, as will the distinctions between factual and embedded contexts, and between relevance to an individual and relevance in a context. Employing these contrasts,
158
Assertion
it will be possible to give an account of the main-clause and embedded uses of the declarative, and to indentify the circumstance of use under which assertoric force results. In order to do this, we will first look at simple declarative sentences that explicitly express just one proposition, then extend the analysis to complex sentences containing embedded declaratives.
8.4.1 Interpreting simple sentences According to Relevance Theory, all utterances, being instances of ostensive stimuli, convey a presumption of their own optimal relevance (Sperber and Wilson 1986/1995: 164). This warrants the assumption by the hearer that he can identify the speaker’s intended meaning as the first interpretation he arrives at that justifies the effort expended in arriving at that interpretation.2 One way the speaker can assist the hearer in identifying her intended meaning is by indicating how he should employ the proposition explicitly expressed by her utterance. As was first noted in Chapter 7, the declarative mood is distinguished from the imperative and interrogatives in that the proposition it explicitly expresses can form part of the speaker’s message. This observation can be captured in Relevance Theory by saying that while the proposition expressed by a non-declarative can contribute to the relevance of an utterance only indirectly, as a constituent of a higher-order representation (i.e. higher-order explicature), the proposition expressed by a declarative can, uniquely, make a direct contribution to the relevance of the utterance in its own right. It is just this observation that the current account of the declarative as encoding an assumption’s relevance in its own right in a context seeks to exploit. According to the current theory, on encountering the declarative mood in the process of interpreting an utterance, the hearer is entitled to assume that the proposition explicitly expressed by the utterance is relevant in its own right in an accessible context. His next move should therefore be to access the most accessible context and derive the implications that result from adding the proposition expressed to that context. If the effort he expends results in an interpretation that warrants that effort, than he should assume that that is the interpretation intended by the speaker. If not, the he should process the proposition expressed in another way, perhaps by deriving its implications in a less accessible context, or perhaps by processing it as a constituent of a higher-order representation, until he arrives at an interpretation that justifies the effort expended. Of course, the greater the effort expended, the greater
Assertion and Relevance 159
the effect will need to be to justify that effort. At some point, if no such effect is achieved, the interpretation process will be abandoned and communication will have failed. Whether the utterance of a simple declarative sentence conveys assertoric force depends, in part, on whether the proposition expressed is presented by the speaker as relevant in its own right and, if so, on the type of context in which it is presented as relevant. An utterance may have assertoric force, on this account, if the proposition explicitly expressed by the utterance is presented by the speaker as relevant to the individual in its own right. To be presented, by linguistic means, by the speaker as relevant to the hearer in this way, a proposition must (a) be expressed by a form that presents the proposition as relevant in its own right in a context; and (b) be presented by the speaker as relevant in its own right in a factual context.3 Whether the first condition holds depends on the linguistic form used. Whether the second holds depends on processing considerations, specifically on whether the most accessible context available to the hearer to process the proposition expressed by an utterance of a simple declarative sentence is a factual one. Why might an utterance that meets these conditions have assertoric force? From the perspective taken in this Chapter, the primary characteristics of assertion are best seen in terms of safeguards aimed at ensuring that the human ability to represent states of affairs in language and present these as relevant to an individual does not leave that individual open to being misled, and perhaps even exploited, by others. To present a proposition as relevant in its own right in a factual context, and hence as relevant to the individual in its own right, is to present that proposition as liable to lead to an improvements in the hearer’s representation of the world. This cannot be guaranteed, though: the proposition expressed may be false in the actual world, and adopting a factual assumption that expresses a false proposition is, of course, likely to lead to a deterioration of the individual’s representation of the world, rather than an improvement. Allowing others to add to your stock of factual assumptions is, therefore, a risky business. Rather than relying in your own perceptual and inferential abilities as sources of information, you have to rely on theirs. Moreover, they may have motives in presenting proposition as relevant to you in its own right that are not compatible with your own goal of achieving an accurate representation of the world. In other words, they may wish to deceive you. That said, where others are reliable and trustworthy, having them supply you with factual assumptions can be a very efficient means of acquiring information beyond your perceptual and inferential capacities. Thus,
160
Assertion
while a risky business, allowing others to add to your stock of factual assumptions is potentially also a highly profitable one.4 Given these considerations, it is to be expected that certain safeguards have been built into this form of communication. These come in two types: cognitive and social. On the cognitive side, there is the ability to quarantine the proposition presented as relevant in its own right to the individual by embedding it in a factual assumption of the type ‘The speaker is saying that P’ or ‘The speaker wants me to believe that P’. To do this is to adopt the meta-communicative stance discussed in Chapter 3. As noted there, though, there is no need to posit that this stance is always taken: it may be called upon when felt by the hearer to be necessary, perhaps because the speaker is untrustworthy, perhaps because of the nature of the discourse, or perhaps because processing the proposition expressed as relevant in its own right in factual context would lead to a number of strongly held factual assumptions being cancelled. Indeed, in section 8.6 of this chapter it is argued that it is only when the speaker is judged by the hearer to be incompetent or untrustworthy that this meta-communicative stance need be called upon. If the speaker is competent and trustworthy, linguistic communication can succeed without such higher-order representations being employed in the interpretation process. On the social side, there are the sanctions that can be brought to bear on those who falsely present a proposition as cognitively beneficial to a hearer. These vary according to the arena of language use and the intentions the speaker is judged to have. If the speaker is thought to have wilfully misled, then the sanctions are higher than if she has spoken in good faith. This is true in informal settings and in institutional discourse. In the former, mistakes may lead to a loss of status but are unlikely to lead to ostracism, whereas wilful deception might well do so. In the latter, penalties are generally higher for deception than for poor judgement. These social and cognitive safeguards reflect the social and cognitive characteristics of assertion that distinguish it from mere information transfer. On the one hand, as we have seen, full competence in assertion relies on the ability to recognise assertions as sources not of information, but representation and hence belief. This entails recognising that they may be false, and might therefore mislead. Full competence in assertion therefore relies on the ability to metarepresent the content of assertions as expressions of, and attempts to influence, belief. As was noted in Chapter 3, this aspect of assertion is stressed by Grice (1989) and those who follow him in viewing assertion in terms of belief attribution, such
Assertion and Relevance 161
as Bach and Harnish (1979) and Sperber and Wilson (1986/1995). These authors see the cognitive side of assertion as central to its nature. Those who see the social side of assertion as central, such as Brandom (1994) and Williamson (1996), stress the fact that individuals take responsibility for the propositions they assert, and are open to rebuke if they fail to give adequate reason for the content they seek to convey. In doing so, these authors stress the social safeguards on this type of communication. The reality is that both the social and the cognitive aspects of assertion are central to a full appreciation of this activity. So far, I have been careful to say that the speaker’s presenting a proposition as relevant in its own right to an individual may result in assertoric force. It is not the case that it necessarily does so. This is because, when a simple declarative sentence is used to present the proposition as relevant in its own right to the hearer but there is no chance of mistake or deception, then the safeguards we have been discussing are irrelevant and the utterance does not have assertoric force. To see this, consider the following cases: (8) a. I (hereby) offer my resignation b. On my word, I’ll never speak to Peter again c. You will clean the latrines The utterances in (8) are, respectively, an offer of resignation made using an explicit performative, a promise, and a directive. What they have in common is that in each case a simple declarative sentence is used to present the proposition expressed as relevant in its own right to the hearer: in each case the proposition expressed is intended to become a belief of the hearer’s. In the performative case (8)a, the hearer is intended to believe that the speaker offers her resignation; in the promise (8)b, he is intended to believe that the speaker will never speak to Peter again; in the directive case (8)c, the speaker is intended to believe that he will clean the latrines (see Chapter 4 for an analysis of this form of directive). In none of these cases, though, does the utterance have assertoric force. The reason for this is that in each case, witnessing the act itself either is, or is presented as, sufficient grounds for accepting the proposition expressed. The speaker is not presenting herself as a source of knowledge – she is not offering to ‘extend the senses’ of the hearer. Rather, the hearer can rely on his own senses to confirm that the state of affairs described obtains or will obtain. In the case of the explicit performative, the utterance constitutes the state of affairs described, so if the hearer has witnessed the utterance then he has witnessed the state
162
Assertion
of affairs described. In the case of the command, as long as this has been properly given, the hearer, as the agent of the act, must believe that he will bring about the state of affairs described. In the case of the promise, the speaker presents her having spoken as sufficient grounds to believe the proposition expressed. In choosing whether or not to adopt that proposition as a belief, the issue is not whether the speaker is a reliable and trustworthy source of information, but whether she can be relied on to keep her word. There is no issue over whether the speaker knows that the proposition expressed by her utterance obtains, the only issue is whether she will bring about the state of affairs her utterance describes. The issue is thus not over the quality of information on offer, but on the reliability of the speaker to do what she says she will do. Consequently, the safeguards that characterise assertion do not apply to promises, or to explicit performatives and directives. Explicit performatives and directives cannot deceive, so the safeguards have no role to play. And while promises do not offer the same certainty that the proposition expressed obtains, they are presented as offering that certainty: it is kernel to the nature of promising that, having witnessed a promise, one is entitled to act as if the proposition expressed obtains. (Of course, whether one is wise to take up that entitlement on a particular occasion is another matter.) Utterances such as those in (8) thus have in common with other non-assertives such as (2) and (3) (repeated below) the characteristic of themselves counting as sufficient evidence for the state of affairs that they represent. Imperatives used to perform directives and interrogatives used to ask questions represent the speaker as being in particular desire state. So a directive utterance such as (2)a, in the appropriate context, is sufficient reason to believe (2)b, while a question such as (3)a, again in the appropriate context, is sufficient reason to believe (3)b (recall that in Chapter 3 it was noted that in the interpretation of non-assertives, comprehension and acceptance coincide). As in (8), the issue of sincerity and reliability does not arise in these cases because the utterance itself is sufficient evidence for the state of affairs represented: the difference between (8) on the one hand, and (2) and (3) on the other, is that in (8) the proposition explicitly expressed describes the state of affairs represented, while this is not the case in (2) and (3). (2) a. b. (3) a. b.
Stand up! The speaker wants the hearer to stand up Do you want a drink? The speaker is asking the hearer whether he wants a drink
Assertion and Relevance 163
There are also cases of the non-assertoric use of the declarative that do not aim at adding to the individual’s stock of factual assumptions. These include supposition for the sake of argument and fiction, which are cases in which the proposition expressed is presented by the speaker, in part by virtue of her use of the declarative, as relevant in an embedded context. Embedded contexts, as was noted above, mimic factual contexts in aiming at consistency. For this reason, they are able to serve as representations of particular worlds (see Chapter 5). A hearer will process the proposition expressed by a declarative in an embedded context if this is more accessible than a factual context, for this will reduce processing effort and hence help to optimise relevance. An embedded context can be made accessible by the speaker saying something like ‘Let’s suppose the following for the sake of argument’ or ‘Once upon a time . . .’, though this may not be necessary if other contextual features, or the content of the utterance, makes an embedded context the most accessible.5 So, canonically, assertoric force is the result of the speaker using a form that presents the proposition expressed as relevant in its own right in a context in order to present that proposition as relevant in its own right to an individual, viz. the hearer. This entails presenting that proposition as relevant in its own right in a factual context, and hence as contributing towards an improvement in the hearer’s representation of the world, or, in theory-specific terms, as a source of positive cognitive effects. Main-clause declaratives are the form of choice for performing assertions because, unlike embedded declaratives, the context in which the proposition expressed is presented as relevant is not specified by that utterance. Thus, given the presumption of relevance that comes with all utterances, and given that the declarative marks the proposition expressed as relevant in its own right in a context, if it is mutually manifest that the most accessible context is a factual one, then the hearer will be justified in assuming that the speaker intends to present the proposition expressed as relevant in a factual context in its own right. In other words, he can assume that she intends the proposition expressed to be relevant to him in its own right. To reiterate, then: canonically, assertion is a case of a speaker presenting, by linguistic means, an explicitly expressed proposition as relevant to the hearer (i.e. as relevant to an individual) in its own right. Assertion cannot be defined thus, though. In order for an utterance to have assertoric force, it must also be subject to the cognitive and social safeguards that distinguish assertion. These may be triggered/applicable if the mere fact of having witnessed the utterance does not constitute sufficient reason for adopting the proposition expressed as a factual
164
Assertion
assumption. The reason assertion is such a valuable activity in human affairs is that it allows us to gather information far beyond our individual perceptual and inferential capabilities. This also presents a risk, though, and hence safeguards, both cognitive and social, have developed in order to minimise these. It is the applicability of these safeguards that distinguishes assertion both from other illocutionary acts and from other forms of information transfer. One advantage of the account developed here is that, because it is grounded in a notion of relevance to an individual, it offers an explanation of why assertions are of interest to hearers. On this view, an assertion comes with a presumption that the proposition explicitly expressed will be of relevance to the individual. That is to say, it comes with the promise of positive cognitive effects derivable from the proposition expressed. The association of assertion with truth follows from this: only true assumptions are likely to lead consistently to positive cognitive effects. A problem with accounts of assertion not couched in terms of relevance is that they give no explanation of why the activity is of such central importance in human affairs. There is an indefinite amount of true information, but most of it is of no use or interest to most individuals. Accounts of assertion merely in terms of commitment to truth thus miss out on the point of assertion. To be sure, we tend to punish liars more heavily than bores, suggesting perhaps that truth is valued more highly than relevance. However, in intending to deceive, liars present false assumptions as relevant. Someone who merely spouted irrelevant falsehoods would not be said to be lying. Rather, her behaviour would most likely be deemed irrational. It is falsehoods which appear relevant that we view most negatively, for these are liable to lead to negative cognitive effects.6
8.4.2 Interpreting complex sentences So far we have been considering the interpretation of simple declarative sentences. However, the account of the declarative presented here is intended to apply to both main-clause and embedded instances of declarative mood. In this sub-section, therefore, we will focus on complex sentences which explicitly express more than one proposition. The issues that we will consider are the parenthetical use of embedded declaratives, the embedding of declaratives in factive environments, and the contrast between declarative and non-declarative embedded clauses. A sentence in which a declarative clause is embedded as the object of a speech-act or propositional-attitude verb presents the proposition
Assertion and Relevance 165
expressed by the embedded clause as relevant in its own right in an embedded context. This is, in effect, to present the complex (i.e. embedding) proposition as relevant to the individual in its own right, for any modification of an embedded context results in an indirect change to the speaker’s stock of factual assumptions. To illustrate this point, consider the following examples: (9) I believe that Peter is guilty (10) John says that Peter is guilty Each of the sentences (9) and (10) presents the proposition that Peter is guilty as relevant in an embedded context: in (9), in a context representing the speaker’s representation of the world; in (10), in a context representing John’s view of the world. Because each of these contexts is embedded in the hearer’s set of factual assumptions, a change to the embedded assumption will result in a change to the hearer’s basic representation of the world, i.e. to his factual assumptions. This change, though, will be indirect, brought about by modifying a set of assumptions embedded in this. An assumption that is relevant in an embedded context thus contributes to relevance to an individual only in an indirect manner: by being relevant in its own right in an embedded context, it is indirectly relevant to the individual. But the net effect is that the proposition expressed by the main clause is presented by the sentence as relevant in its own right to an individual (i.e. the hearer). One way the embedding clause in (9) and (10) can be viewed, then, is as a means of specifying the type of context in which the proposition expressed by the embedded clause is to achieve relevance. This, though, is what is specified by the sentence, not the utterance. While sentence meaning and utterance meaning may converge in this respect, they need not. As was noted in Chapter 7, sentences such as (9) and (10) can be used in contexts in which the proposition expressed by main clause has main-point status. This was illustrated with the following examples (taken from Simons 2007): (11) A: Who was Louise with last night? B: a. She was with Bill b. Henry thinks she was with Bill c. Henry hinted that she was with Bill d. Henry is convinced that she was with Bill
166
Assertion
In (11), A’s question makes it clear that details of Louise’s whereabouts the previous the previous evening would be relevant to him. It is therefore manifest to B that (11)a would be relevant to A in its own right in a factual context (i.e. relevant to A as an individual). However, the sentences employed in B’s responses (11)b to d do not present the proposition expressed by (11)a as relevant in this way. Rather, (11)b to d present that proposition as relevant in an embedded context representing Henry’s epistemic state. In these cases, then, the means by which relevance is to be achieved that is indicated by the sentence is at odds with the requirements of the context: A’s question in (11) does not enquire after Henry’s epistemic state. This divergence between the requirements of the conversation and what is indicated by the sentence employed can be explained in the following terms. While the sentence presents the proposition expressed by the embedded clause as relevant in its own right in an embedded context, the speaker is well aware that it will in fact be most relevant to the hearer in a factual context. However, because she only has hearsay evidence for the proposition that Louise was with Bill, she cannot present this as relevant in its own right in a factual context by simply uttering (11)a, for to do so would make her responsible for the truth of that proposition and, if that proposition turned out to be false, subject her to the social sanctions, discussed earlier, that act as safeguards in this type of communication. Consequently, she uses a sentence that presents the relevant proposition as relevant in an embedded context, and passes to the hearer the responsibility for processing it in a factual context, thereby reducing the chance of social sanction. In doing so, she is exploiting the relevance-maximising tendency of human cognition: she is aware that the hearer’s interests will ensure that the embedded proposition will achieve relevance in a factual context, even though she does not, in her choice of linguistic form, present it as relevant in this way. By contrast, in using a parenthetical construction, as in (11)b to d , the speaker, by granting it main-clause status, acknowledges that the proposition that Louise was with Bill is relevant to the hearer in its own right, but again avoids explicitly presenting it as such by the use of the parenthetical element. Once more, the possibility of social sanction is mitigated. What distinguishes (11)b to d from (11)b to d is that the proposition expressed by the parenthetical element in the latter is linguistically precluded, due to its supplement status, from being the main source of relevance of the utterance, as we saw in the previous Chapter.7
Assertion and Relevance 167
(11) b . She was, Henry thinks, with Bill c . She was, Henry hinted, with Bill d . She was, Henry is convinced, with Bill When the embedding construction is a factual one, though, as in the responses in (12) and (13), the speaker cannot avoid taking responsibility for the truth of the proposition expressed by the embedded clause, for the proposition expressed by the embedded clause is entailed by that expressed by the main-clause. In such cases, then, the explanation for the embedding of the most relevant clause must be that the additional information conveyed by the embedding clause justifies the extra processing effort involved, perhaps because it gives the source of the information conveyed by the embedded (main-point) clause. (12) A: Where did Louse go last week? B: a. Henry discovered that she had a job interview in Princeton b. Henry learned that she had a job interview in Princeton c. Henry found out that she had a job interview in Princeton (13) A: How will Louise get to the picnic? B: a. Henry realized that she can take a bus b. Henry figured out that she can take a bus In the quest for an account of assertion, factive predicates (of the sort typified by (12) and (13)) complicate the picture somewhat because, on the one hand, they present the proposition expressed by the complement clause as relevant in an embedded context but, on the other, they also commit the speaker to its truth. The linguistically indicated point of the utterance does not seem to be, then, to present the proposition expressed by the complement clause as relevant in a factual context, but the speaker is nevertheless committed to that proposition. A similar situation arises with first-person belief attributions. Although they are not factive, in these cases the speaker does commit herself to the truth of the proposition expressed.8 As in the factive case, she thus undertakes assertoric commitment, though without – linguistically, at least – indicating that her aim is to put that proposition forward for acceptance, rejection or confirmation by the hearer. Of course, in practice it is often mutually manifest that in reporting a first-person belief, the speaker is putting that object forward for acceptance. The point is, nevertheless, that in such cases we explicitly mark that proposition
168
Assertion
as relevant in another way (i.e. in an embedded context representing our own representation of the world). So, what factives, such as (12) and (13), and first-person belief reports show is that there are linguistic means of presenting a proposition as relevant in an embedded context whose use also results in the speaker undertaking assertoric commitment to that proposition. However, as illustrated by the dialogues (12) and (13), these can be exploited by speakers to communicate the content to which assertoric commitment is undertaken by virtue of the fact that the declarative form employed to express that content presents that proposition as relevant in a context. As was noted in Chapter 7, though, not all factive predicates can be used to present the proposition expressed as relevant in its own right: (14) could not be used as a response to A’s question in (12). (14) Henry is sad that she had a job interview in Princeton That the presence of an embedded declarative in (14) that does not present the proposition expressed as relevant in its own right in any context is a challenge for the account being put forward here. My response is to claim that while the choice of mood in the embedded clause might appear to present the proposition expressed as relevant in its own right, the semantic relationship between the embedded and the embedding proposition precludes this. Constructions such as (14) have been analysed by a number of authors (Bosque 1990; Heim 1992; Quer 1998) as having causal meaning: Louise’s having a job interview in Princeton is presented in (14) as the cause of Henry’s sadness. In the responses in (12) and (13), by contrast, no causal link is claimed between the state of affairs described by the embedded proposition and the state of knowing attributed to the hearer. The causal link in (14) means that what is presented as relevant are not the independent propositions, albeit in distinct contexts (factual and embedded), as in (12) and (13), but a single complex proposition of the form ‘P caused Q’. This analysis is supported by the fact that in some languages, notably Spanish, such constructions take a subjunctive complement, as illustrated by (15). This is in contrast to cases such as those exemplified by (12) and (13), where, in Spanish, a declarative complement is found in positive sentences (for extensive discussion of the Spanish subjunctive from the perspective of the framework presented in this chapter, see Jary 2009). (15) Henry se lamenta de que Louise tenga una entrevista en Princeton ‘Henry is sad that Louise has- SUBJ an interview in Princeton’
Assertion and Relevance 169
The idea, then, is that the difference between factive environments such as (12) and (13) and those of the type illustrated by (14) can be accounted for by the claim that only the former present the embedded proposition as relevant in its own right. Cases such as (14) present a complex proposition, in which two propositions are related causally, as relevant in its own right. This is why (14) cannot be employed as a means of answering A’s question in (12): this question requires that a proposition regarding Louise’s whereabouts be presented as relevant in its own right. The distinction between an embedded proposition being presented as relevant in its own right and a complex proposition being presented as relevant in this way also underlies the current framework’s analysis of mood-distinct minimal pairs, such as we find in (16): (16) a. John insists that the children are there b. John insists that the children be there On the reading intended here, (16)a makes a statement about John’s expressed beliefs, while (16)b reports an expressed desire. The difference can be accounted for as follows: (16)a, having a declarative embedded clause, presents the proposition expressed by that clause as relevant in its own right in an embedded context, i.e. one that is formed of assumptions that constitute the hearer’s representation of John’s representation of the world; (16)b, by contrast, has a non-declarative complement and hence does not present the embedded proposition as relevant in any context. Rather, (16)b presents the complex proposition expressed by the sentence as a whole as the source of contextual implications. In terms of encoded conceptual meaning, (16)a and b are identical. We can represent what they encode as in (17)a. As it stands, (17)a is ambiguous between (17)b and (17)c. However, what the representation in (17)a does not indicate is whether the embedded proposition is represented as relevant in its own right: this is just what mood-choice indicates. A declarative indicates that the proposition is presented thus, indicating that a reading as in (17)b is intended; an infinitive indicates that the proposition expressed is not presented as relevant in its own right and hence that an interpretation as in (17)c is intended. (17) a. > b. > c. <wants-strongly, John, >
170
Assertion
The reason that an infinitive complement is related with a desire-report is that desires, as was noted earlier, are not consistency aiming. Consequently, in representing the desires of others we do not aim at achieving a consistent set of propositions: a list of inconsistent propositions is potentially an accurate representation of another’s desires (and will necessarily be so if that individual is in a dilemma). Because they do not have to be tested for consistency, propositional forms constituting the representations of the desires of another do not have to be of an appropriate format to serve as premises in inference, and hence do not need to be of a format that can be relevant in its own right. On the account presented here, then, it is to be expected that desire attributions do not take declarative complements, and this expectation can be used to explain the difference in meaning between (16)a and b. However, not all declarative/infinitive complement alternations can be explained in this way. Such an account will not help with the distinction between (18)a and b, for example. In these cases, there appears to be no truth-conditional difference between the two. Nevertheless, an account can be given in terms of the current account that makes novel, accurate predictions and reveals new patterns in the data. It is presented in detail in Jary (2008). Here, I give only a brief outline. (18) a. We believe that you are the best person for the job b. We believe you to be the best person for the job Utterances reporting the beliefs of others can be of interest to hearers in two distinct but related ways: either for the content of the belief attributed or for the help it gives us in explaining behaviour. In the first case, we may be interested in adopting the content of that belief as one of our own, if the attribuee is an expert or reliable source of information. In such cases, the proposition expressed by the embedded clause is relevant in its own right: we are interested in the implications of the embedded proposition. This is not the case, though, in behaviourexplaining cases. The standard folk-psychological schema for explaining intentions is, ‘If someone wants that Q and believes that P will lead to Q, then, all things being equal, that person will try to bring about P’. In such a schema, P and Q are not serving as premises in their own right; rather, the forms in which they are embedded (i.e. ‘X wants that Q’, ‘X believes that P’) are. In other words, the embedded proposition is not relevant in its own right. The prediction that the current analysis makes, then, is that there should be a preference for using infinitive complements in behaviour-explaining uses of ‘believe’. And indeed, the
Assertion and Relevance 171
corpus study in Jary (2008) found that uses of the infinitive complement were more likely to occur in contexts in which behaviour explanation was the goal than in those where the content of the belief had been presented as relevant in its own right.9 That infinitival complements are, on some occasions, used to present the proposition expressed as relevant in its own right, as was also evidenced by the above corpus study, might be seen as a problem for the current account. But I do not believe this to be the case. Use of this structure is restricted to formal, prepared discourse, and some of the effort that, on the current account, would be involved in interpreting the proposition expressed by the infinitival complement as relevant in its own right, can be explained in terms of the stylistic effects, perhaps of indirectness, that characterise this mode of discourse. As noted earlier, how the sentence presents information and how the utterance does need not coincide: while a sentence may fail to present an embedded proposition as relevant in its own right, a hearer may be coerced to process it in that way in order to derive sufficient cognitive effects in order to fulfil the presumption of relevance. In such cases, though, there should be a payoff in terms of additional effects that justify the effort expended that would not have been needed had the relevant proposition been presented as such by the speaker’s choice of linguistic form. In summary, the current framework assumes that, on coming across a proposition expressed by a declarative clause, embedded or not, the interpretation process blindly seeks to derive implications from that proposition in the most accessible context. This context may be factual or embedded, independent of whether the proposition is expressed by a main clause or a subordinate clause. Whether this utterance counts as an assertion will depend on both whether the most accessible context is a factual one, and whether the speaker’s choice of linguistic form presents the proposition expressed as relevant in an embedded context. If the form does present the proposition as relevant in this way (e.g. in presenting the content of that proposition as the object of a third person’s belief), then the speaker cannot be held responsible for the truth of the embedded proposition, even if the conversational context makes it mutually manifest that the utterance can only achieve relevance if the hearer processes the proposition expressed by the embedded declarative in a factual context. By contrast, a main clause-declarative offers no indication that the proposition expressed is relevant in a non-factual context. Consequently, in uttering a main-clause declarative when it is mutually manifest that the most accessible context is a factual one, the
172
Assertion
speaker will be taken to have asserted that proposition and thereby taken on the associated commitments. This will be the case unless merely having witnessed the utterance is sufficient grounds for coming to believe the proposition expressed. Only if further reason to accept that proposition might be sought will it be correct to say that the speaker has asserted the proposition expressed. On the view presented in this section, main-clause declarative mood is an assertion sign. When used by a competent asserter under certain conditions, the speaker thereby commits herself to the truth of the proposition expressed by virtue of having used that form. The conditions in question are that it is mutually manifest that the optimally relevant interpretation of the utterance will result from the speaker processing the proposition explicitly expressed for relevance in its own right in a factual context. This is not to say, however, that main-clause declarative mood encodes assertoric force. What the declarative does, rather, is present the proposition expressed as relevant in its own right in a context, without specifying which context. This enables it to be employed to present a proposition as relevant in its own right to an individual, and it is for this reason that the declarative is so tightly linked with assertoric force. The most crucial respect in which assertions differ from the other main illocutionary-act types (directives, questions) is that only assertion presents the proposition expressed as relevant to the hearer. Directives and questions, by contrast, present as relevant an internal state of the speaker that constitutes an attitude towards the proposition expressed. While it is correct to say that assertions can be used this way (i.e. to express beliefs), that this – in some important sense – is not their primary function is evidenced by the fact that they are evaluable in terms of truth and falsity. Were the function of assertions primarily to express beliefs, the issue of truth would not arise, just as it does not in the case of directives and questions.10 The notion of relevance in a context is thus fundamental to the analysis of the declarative given in this chapter. It is relevance in a context that is encoded by the declarative mood. This feature allows the speaker to present as relevant to the hearer not only her action (i.e. the production of a propositional form), but the very proposition her utterance expresses. She does this by presenting the proposition expressed as relevant to the hearer in a factual context (i.e. in a context made up assumptions drawn from his basic, unreflective stock of assumptions about the world). Non-declaratives, by contrast, do not afford this opportunity. An utterance of an imperative or an interrogative sentence presents as relevant to the hearer a higher-order representation of which
Assertion and Relevance 173
the proposition expressed by the sentence is a constituent. It is for this reason that they do not have assertoric potential.
8.5 Assertion and main point Analysing assertion in terms of relevance opens up the possibility of shedding light on the relationship between assertion and main-point status, for, as we saw in Chapter 7, main-pointhood can be straightforwardly explained in terms of relevance. Here, we return to this issue, and consider the conditions under which main-pointhood and assertion may coincide, those in which they may diverge, and the reasons for this. As we saw in Chapter 7, the proposition that is the major source of the contextual implications of an utterance is its main point. By virtue of the fact that it is a source of contextual implications, such a proposition is relevant in its own right, and it is this feature that links main point status and assertion, for, as we have seen in this chapter, assertion is a case of a proposition expressed by a declarative clause being presented as relevant in its own right in a factual context. Assertion and main-pointhood thus coincide when the asserted proposition is the main source of the relevance of the utterance. But, while the notion of a proposition being relevant in its own right is kernel to both assertion and main pointhood, there are two dimensions along which they may diverge: on the one hand, what is asserted may not be the main source of the utterance’s relevance, whereas this is part of the definition of mainpointhood; on the other hand, the main point might not be presented as relevant (in its own right) in a factual context, while this is essential to assertion. This is summed up in Table 8.1: Table 8.1
Assertion and main point compared
Asserted proposition Main-point proposition
Presented as a source of relevance in its own right
Main source of the relevance of the utterance
Presented as relevant in a factual context
Always
Sometimes
Always
Always
Always
Sometimes
174
Assertion
Typical cases where main-point status and assertion do not coincide occur when the main point of an utterance is expressed by the complement of a propositional-attitude or speech-act predicate, as illustrated by (11)b to d from Simons (2007), already discussed in Chapter 7: (11) A: Who was Louise with last night? B: a. She was with Bill b. Henry thinks she was with Bill c. Henry hinted that she was with Bill d. Henry is convinced that she was with Bill In these cases, it is mutually manifest that the proposition expressed by (11)a will be relevant in a factual context. Nevertheless, the form chosen by the speaker of (11)b to d presents that proposition as relevant in an embedded context. In presenting the most relevant proposition thus, the speaker avoids asserting the main point of her utterance. For an assertion to take place, the communicated proposition must be presented as relevant in its own right in a factual context, not merely processed in this way. In (11)b to d, the responsibility for processing the proposition that Louise was with Bill in a factual context is thus passed to the hearer, allowing the speaker to avoid taking on the responsibilities that would result from asserting that proposition. In the parenthetical equivalents to (11)b to d, the speaker presents the proposition expressed by (11)a as the main source of relevance, but also asserts, by means of the parenthetical utterance, the proposition expressed by (11)b to d. This second utterance is marked as not being a potential candidate for the main source of the utterance. Rather, its relevance lies in the indication it gives that the speaker is unwilling to take responsibility for the truth of the main-clause proposition. It does this by making more accessible an embedded context in which she could be taken to be indicating that the proposition expressed by the main clause should be processed. Thus while the formulations (11)b to d present the proposition expressed by (11)a as the main point of the utterance, they also indicate that its relevance lies in an embedded context, not a factual one, and hence the speaker is able to avoid assertoric commitment to the main-clause proposition. (11) b . She was, Henry thinks, with Bill c . She was, Henry hinted, with Bill d . She was, Henry is convinced, with Bill
Assertion and Relevance 175
So, the difference between the formulations in (11)b to d and (11)b to d is that there are two candidates for main-point status in the former cases, but only one in the latter. The reason that main-clause declarative syntax is an indicator of main-pointhood is that the proposition explicitly expressed by the main clause is the maximal entailment of an utterance that is linguistically marked as relevant in its own right. The distinction between background and foreground implications discussed in Chapter 7 requires that there be a foreground/background distinction between the grammatically specified entailments of an utterance of a declarative sentence. This, in turn, requires that the utterance have at least one foreground implication. If there are no indications to the contrary, then the hearer is therefore justified in assuming that the point of the utterance lies in the proposition explicitly expressed, for this, at least, must be a source of contextual implications. This assumption, though, can be overridden. In the case of (11)b to d, it is overridden by the fact that another proposition expressed by a declarative clause is manifestly of greater relevance in the most accessible context. It might also be overridden by the assumption that the speaker could not have intended the proposition explicitly expressed by the main clause to be the main point of the utterance. This would be the case in (18)a: the speaker would not normally expect the hearer to be ignorant of his having had his own hair cut; hence the hearer cannot reasonably assume that the relevance of the utterance lies in this proposition. Rather, he should assume that the speaker’s point is that she has noticed the haircut, and intends her utterance to be evidence of this. This is an example of – in Sperber and Wilson’s terms – a non-ordinary assertion, in which the speaker undertakes assertoric commitment to the proposition expressed by the main clause of a declarative sentence, but manifestly does not intend it to become a belief of the hearer’s as a result of her utterance. (19) a. You’ve had your hair cut b. The speaker notices that the hearer has had his hair cut We saw in Chapter 7 that there are some restrictions on which declarative clauses can convey a proposition with main-point status. Nonrestrictive relevant clauses and the objects of so-called ‘factive emotives’ (e.g. ‘be sad that’), for example, cannot felicitously be used to convey the main point of an utterance. The ability to distinguish between main and subsidiary points is crucial to written discourse, where the writer
176
Assertion
must take all the responsibility for guiding the interpretation process, the possibility of negotiation and repair being absent. In such discourse, the implications of the main point will form highly accessible contextual assumptions for the processing of subsequent utterances, and hence indicating which point is central is vital if overall text coherence is to be achieved. To see this, consider the following example: (20) C, who had been granted refugee status in 2000, was arrested in December 2001, just after the Anti-Terrorism, Crime and Security Act-which sanctioned detention without charge or trial and was introduced after the September 2001 terror attacks on New York and Washington-gained royal assent.11 This sentence explicitly expresses a number of propositions, listed in (21): (21) a. C was arrested in December 2001 just after the Anti-Terrorism, Crime and Security Act gained royal assent b. C had been granted refugee status in 2000 c. The Anti-Terrorism, Crime and Security Act gained royal assent shortly before December 2001 d. The Anti-Terrorism, Crime and Security Act sanctioned detention without charge or trial e. The Anti-Terrorism, Crime and Security Act was introduced after the September 2001 terror attacks on New York and Washington Only one of these can convey the main point of the utterance: (21)a. This is the only proposition expressed by an independent clause, the remainder being expressed by supplements and adjuncts. By presenting (21)a as the main point of (20), the writer indicates that the discourse is about the individual referred to as C, rather than about The AntiTerrorism, Crime and Security Act. In other words, the addressee is justified, on the basis of the fact that the communicator has presented (22)a as the main point, in expecting the next sentence to be about C. And indeed this is the case. (22) is the next sentence in the discourse: (21) He had been sentenced in absentia to 15 years imprisonment in Egypt for trying to recruit army officers to a terrorist group In terms of the current proposal, by expressing the propositions (21)b to e in adjuncts and suppositions, the writer indicates that the effects
Assertion and Relevance 177
derived from processing these propositions are not intended to play a significant role in the processing of subsequent utterances. In indicating thus their comparative lack of relevance, the writer ensures that, even though these propositions are relevant in their own right, the reader will devote less effort to processing them. The speaker thus asserts – i.e. presents as relevant in its own right in a factual context – each of the propositions in (21). Only (21)a, however, is a candidate for the main point of the utterance.12 From the preceding discussion we must conclude that while there are linguistic means of precluding an explicitly expressed proposition form being a candidate for the main point of an utterance, the relationship between main-clause declarative syntax and main point-status is less straightforward. In order to explain this relationship, two factors must be considered. First, there is the lack of any linguistically encoded feature (such as we find in supplements and adjuncts) that prevents the proposition explicitly expressed by a main-clause declarative from achieving main-point status. Then, there is the expectation that at least one of the grammatically specified implications of an utterance of a declarative will be a foreground implication. As the maximal entailment is the proposition explicitly expressed, in the absence of a competitor, this should be assumed to be the utterance’s main point (i.e. its main source of contextual effects). However, if there is a competing proposition expressed by a declarative clause that is not linguistically precluded from having main point-status, as in examples (11)b to d, and this is manifestly more relevant in the most accessible context, then this will be the main point of the utterance. Alternatively, if there is no competitor, but the proposition explicitly expressed by the main-clause declarative cannot achieve relevance in its own right in an accessible context, then the hearer is justified in embedding that proposition as an element of a higher-order explicature and looking for the relevance of the utterance in the implications of that in the most accessible context, as is the case in (18).
8.6 Relevance maximising vs. relevance optimising In section 8.4.1, I claimed that a distinguishing feature of assertion is that it is subject to certain cognitive safeguards, made necessary by the need to avoid polluting an individual’s set of factual assumptions with false assumptions liable to lead to a degradation in his overall representation of the world. In this section, I want to look more closely at this aspect of the theory, and relate it to the view of assertion outlined in
178
Assertion
Chapter 3. There, recall, it was claimed that the comprehension of assertions by fledgling communicators does not involve belief attribution. The ability to attribute beliefs, it was claimed, develops from a protocompetence in assertion. Once developed, this ability can be employed in the interpretation of assertions (and indeed, full competence in assertion is possible only if this capacity is present), but it needn’t be. Under the right circumstances, fully competent asserters can use the same processes that they employed before this ability developed, keeping their belief-attributing abilities in reserve, to be called on if needed. What I am going to argue is that much pragmatic inference relies not on the ability to attribute belief, but on the ability to attribute knowledge. To attribute knowledge, in the sense intended here, is the ability to attribute what, from the perspective of the attributor, is shared information. From the point of view of someone able to attribute knowledge, but not belief, there is no such thing as false information: there is just knowledge and ignorance. Such an individual has no grasp of the intentionality of thought: for her, the perceived world drives behaviour directly, unmediated by mental representation. To show how knowledge attribution account for much of the mentalstate attribution required for pragmatic inference, I will consider the interpretation strategies identified by Sperber (1994; see also Wilson 2000; 2003). These are ‘naïve optimism’, ‘cautious optimism’ and ‘sophisticated understanding’. According to Sperber, these strategies vary in terms of their metarepresentational requirements. However, I will argue that, on one reading of the term, naïve optimism and a variety of cautious optimism require no metarepresentation at all. To say that an individual has the ability to metarepresent the mental states of others can be understood in at least two ways. It can mean that that individual has the ability to track the information state of others, though without any grasp of the notion of representation that is required in order to attribute false-beliefs. On this view, a creature has metarepresentational ability if it can behave in a manner that can be explained by attributing to it a thought of the type shown in (23), where X is a conspecific, ‘:’ a dummy term for a mental-state attributing predicate such as ‘know’ or ‘believe’, and ‘that P’ a proposition. On this view, the ability to evaluate the truth value of P is not a requirement for having metarepresentational ability. Indeed, the notions of truth and falsity may not be part of the creature’s conceptual repertoire. On this understanding of metarepresentation, then, the creature claimed to have metarepresentational ability is not claimed to have any notion of representation as a driver of behaviour. For it, information drives behaviour
Assertion and Relevance 179
directly. Consequently, we can replace the dummy term with ‘knows’, understood in the sense outlined above. This gives us (24). Creatures with this ability can entertain thoughts of the type exemplified by (24) and (25), demonstrating this by correctly predicting X’s behaviour on the basis of that mental-state attribution. Again, the possible falsity of P does not arise for these creatures. (23) X: that P (24) X knows that P (25) X doesn’t know that P But one does not understand the representational nature of thought unless one has the notions of truth and falsity. Consequently, another notion of metarepresentation can be identified according to which to metarepresent is to entertain a propositional form as a representation, and assign it a truth-value. To say that a creature has the ability to entertain metarepresentations of this type is to say that that creature has a conception of truth and falsity. Such a creature is able to attribute thoughts of the type exemplified by (26), and to demonstrate this by making a correct prediction about X’s behaviour when P is false. Indeed, only this ability demonstrates metarepresentational competence of this type: if P is true, then the ability to entertain thoughts of the type exemplified by (24) cannot be distinguished from the ability to entertain thoughts of type (26).13 (26) X believes that P This latter sense of ‘metarepresent’ is what is in play when a creature is said to have the ability to consider the implications of the embedded assumption without thereby adopting them as beliefs. It is thus the notion that is in play, for example, in Cosmides and Tooby (2000), where it is argued that the human metarepresentational ability serves the function of allowing thoughts to be entertained, and their consequences established, without behaviour necessarily being triggered. What I want to argue is that if this second use of the term ‘metarepresentation’ is applied, then much pragmatic inference does not require metarepresentation. I will illustrate this by looking at the cognitive demands made by Sperber’s three interpretation strategies. First, though, I want to comment on which of these notions of metarepresentation is employed by Sperber.
180
Assertion
The term Sperber uses in much of his writing about metarepresentation is ‘believe’. Thus he illustrates the levels of metarepresentation that are required, according to him, in ostensive-inferential communication as follows (2000: 125): (27)
Mary intends that Peter should believe that she intends that Peter should believe that these berries are edible
This use of ‘believe’ might lead us to conclude that Sperber uses ‘metarepresentation’ in the second sense, on which the individual said to have the metarepresentational ability to attribute thought is aware of the representational nature of thought. That is to say, she can assign a positive or negative truth value to the embedded proposition. However, on other occasions Sperber’s writing suggests that the ability to assign a truth value to the embedded proposition that is the object of the metarepresentation is not required in communication. Consider the following quotation: The attribution of a meaning to a speaker, and the prediction that a person with a false belief will act on this belief, though both involving mindreading, are two very different performances. The formal resources involved in the two cases are not the same. In the case of speaker’s meaning, what is needed is the ability to represent an intention of someone else about a representation of one’s own – a second-order metarepresentation of a quite specific form. [. . .] In the case of false beliefs, a first-order metarepresentation of a belief of someone else is sufficient, but what is needed is the ability to evaluate the truth-value of the metarepresented belief and to predict behaviour on the basis of false belief. We are not aware of any argument to the effect that the ability needed to pass the false-belief task is a precondition for the ability needed to attribute speaker’s meaning. There is nothing inconsistent or paradoxical therefore in the idea of an individual capable of attributing speaker’s meaning and incapable of attributing false beliefs (and conversely). (Origgi and Sperber 2000: 163)
Assertion and Relevance 181
This passage is a strong indication that a grasp of the representational nature of the object of the metarepresentation is not, for Sperber, a requirement for a creature to be felicitously said to have metarepresentational capacities. It suggests, in particular, that the ability to assign a truth value to the embedded proposition, while required for false-belief attribution, is not required for attributing speaker’s meaning. Thus it would seem that, despite his frequent use of ‘believe’, Sperber in fact uses metarepresentation in the first, and weaker, of the two senses discussed above. In what follows, my use of ‘metarepresent’ will be restricted to the second sense, because I don’t think the first merits the use of the term. If one does not have a notion of truth and falsity, then one does not have a notion of representation in the sense applied to assertions and thoughts. Without this ability, one cannot metarepresent in a sense that makes the use of this term felicitous. Having made this point, I will now go on to look in detail at Sperber’s three interpretation strategies in terms of their metarepresentational requirements. As has already been explained, the Relevance-Theory view of communication is that utterances are ostensive stimuli which presuppose their own relevance. They thus warrant the speaker following a path of least effort until the goal of sufficient effect/benefit is reached in interpreting an utterance. This is based on the assumption that human cognition has evolved so as generally to minimise cost and maximise benefit, in other words, to maximise relevance. However, the presumption which utterances are claimed to convey is not one of maximal relevance, but optimal relevance, ‘optimal’ because the speaker will not generally be expected to go against her own preferences, and will never be able to go beyond her abilities, in crafting her utterance. This means that the interpretation that would be most relevant to the hearer may be beyond the speaker’s abilities and/or incompatible with her desires. Consequently, a sophisticated hearer must take the speaker’s preferences and competence into account when interpreting her utterance and so seek not the interpretation that will give him the greatest effects (for any given level of effort), but the interpretation that will give him the greatest effects given the speaker’s preferences and competence. This is summed up in the presumption of optimal relevance: Presumption of optimal relevance (revised 1995): a) The ostensive stimulus is relevant enough for it to be worth the addressee’s effort to process it.
182
Assertion
b) The ostensive stimulus is the most relevant one compatible with the communicator’s abilities and preferences (Sperber and Wilson 1986/1995: 270). What I want to argue is that one of Sperber’s interpretation strategies – ‘naïve optimism’ – never requires taking into account the speaker’s competences and preferences. It thus not a relevance optimising procedure but one of maximising relevance, requiring no metarepresentation. Moreover, certain cases of cautious optimism, while being relevance optimising procedures, also require no metarepresentational abilities in the strong sense being employed here. Consider first naïve optimism, the most basic of the three strategies. A hearer following this strategy simply assumes that the first relevant interpretation of the speaker’s utterance that occurs to him is the one that she intended. This strategy is successful in arriving at the speaker’s meaning only if the speaker is both benevolent and competent: she must be competent to such a degree that she is able to choose a linguistic form such that the first interpretation the speaker arrives at will be her intended one, and benevolent so that she does not choose a form which, thought it may lead to sufficient effects to warrant the effort expended, is likely to deceive the speaker. An individual employing a strategy of naïve optimism, however, does not consider the speaker’s competences and preferences, rather he implicitly assumes that she is competent and benevolent. He thus seeks to maximise relevance, rather than to optimise it. This can be seen by considering how (28), the example discussed by Sperber (1994), would be interpreted by an individual following this strategy: (28) It’s late In interpreting (28), a naïve optimist will simply take the most accessible referent for ‘it’ (or a non-referring interpretation if none is easily accessible) and the most accessible interpretation of ‘late’ (as meaning, say, that it is time for he and the speaker to leave) and, as long as he judges it to be relevant enough to have warranted paying attention to, assume this interpretation is the one intended by the speaker. Metarepresentation enters this story, as Sperber tells it, in two ways: first, in recognising the utterance as an ostensive stimulus and thereby taking it as a premise from which her message will be inferred; and, second, in attributing to her the intended message. Thus, if, as in Sperber’s example, Mary utters (28) in order to inform Peter that it is time to go home,
Assertion and Relevance 183
the initial premise will be (29) and the interpretation (30) (with square brackets indicating levels of representation): (29) Mary says: ‘It’s late’ (30) Mary intends [me to believe [that it is time to go home]] However, neither of these metarepresentations is needed if a strategy of naïve optimism is applied. Consider first (29). Part of the reason given for requiring an initial representation of this type is that it justifies following the interpretation procedure warranted by ostensive stimuli, i.e. the search for optimal relevance by following a path of least effort, taking into account the speaker’s preferences and competences. However, it has already been shown that the naïve-optimism strategy involves not optimising but maximising relevance. This is the same strategy as is applied to the processing of any stimulus which is attended to, for the tendency to maximise relevance is a generalisation about human cognition as a whole, not merely about utterance interpretation. What distinguishes ostensive stimuli is that they come with a presumption of relevance that is in itself justification for attending to them. Now, one way a theorist might conceive of this is in metarepresentational terms: in recognising a stimulus as ostensive, the system represents it in a certain fashion (under ‘says that’, for example) and this triggers the interpretation process. Alternatively, though, it might be conceived of in non-metarepresentational terms: the recognition of an ostensive stimulus could simply interrupt processing such that resources are directed at processing that stimulus above competing stimuli, without that stimulus being metarepresented. The advantage of this second construal is that it does not require the hearer to have metalinguistic abilities such that it can represent the linguistic form of the utterance. Rather, this can be seen as input that feeds directly into the linguistic decoder, the output being inferentially developed to maximise relevance. As for (30), given that the naïve optimist assumes both competence and benevolence, ‘believe’ can here be replaced by ‘know’ for the issue of falsity does not arise. This gives us (31), which, on the current construal, is not a metarepresentation, as it requires no awareness of the representational nature of the most embedded proposition. The best way to view (31), I think, is as a way of capturing the hearer’s understanding of the speaker’s behaviour as goal directed, the goal being to inform him that it is time to go home. (31) Mary intends [me to know [that it is time to go home]]
184
Assertion
The archetypal naïve optimists are very young children engaging in basic communication, but according to Sperber, when ‘everyday communication takes place between people who are benevolent to one another and who know one another well enough’ (1994: 197–8) then the strategy can be employed by adults. The difference is that adults, even when employing naïve optimism, have manifest to them more sophisticated assumptions about the nature of communication that can be employed if necessary. Let us now consider the strategies that result from bringing these mechanisms to bear on the process of utterance interpretation. A hearer following Sperber’s second strategy of cautious optimism assumes that the speaker is benevolent, though not necessarily competent. He will therefore realise that the stimulus the speaker employs may appear to be the most relevant to the speaker but may in fact not be. For example, the speaker of (28) may mean that it is time to leave, while for the hearer the most relevant interpretation may be that a package he has been waiting for, but of which the speaker is unaware, will arrive late. Following a strategy of naive optimism would result in the hearer accepting the most relevant interpretation to him (i.e. that the package is late) as the one intended by the speaker. A cautious optimist, by contrast, would realise that such an interpretation was beyond the speaker’s competence, and hence ignore, or pass over, this possibility in interpreting her utterance (Sperber 1994: 191–4). Is metarepresentation required for this interpretation strategy? The belief-free basic ontology of mental states proposed above as a possible alternative to, or precursor of, full-blown belief attribution would certainly allow a form of cautious optimism: a hearer would be justified in ignoring interpretations which relied on information unknown to the speaker. Thus, the interpretation of (28) as referring to a package could be avoided by a hearer equipped only with this basic repertoire: just as the naïve-optimist follows a path of least effort, taking linguistic senses, assigning referents and considering hypotheses according to accessibility (Sperber 1994: 190), the cautious optimist can do the same, the only difference being that accessibilities will be determined to some degree by information the hearer assumes that the speaker has available to her. On this view, the maximally relevant interpretation would not be arrived at and discarded in favour of the optimally relevant one. Rather, the most accessible interpretation would be the optimally relevant one. This is at odds with Sperber’s conception of cautious optimism, on which a hearer takes a cautiously optimistic approach only when the interpretation arrived at by following a path of least effort
Assertion and Relevance 185
results in an interpretation that, while it would be relevant to the hearer, is incompatible with his assumptions about the speaker’s competences (Sperber 1994: 192). What the lack of a full-fledged concept of belief would do is prevent the hearer from identifying the speaker’s intended meaning when this relies on him attributing a false belief to the speaker. For example, suppose Mary utters (28) intending to inform Peter that a train she wrongly assumes him to be waiting for is delayed. Unless Peter is able to attribute to her the false belief that he is waiting for the delayed train, he will be unable to identify her intended meaning. Cautious optimism can thus be divided into two sub-strategies according to whether it relies on the attribution of an absence of ignorance, or the attribution of a false belief. Note that the latter case is clearly a case of explaining behaviour in terms of false-belief. Indeed, it would be strange if the processes underlying the interpretation of an utterance motivated by a false belief were distinct from explanations of other behaviours motivated by false belief. In at least one type of cautious optimism, then, false-belief attribution plays a role. Behaviour-explanations in terms of false belief have much in common with certain cases of utterance interpretation, then, pace the claim to the contrary made by Origgi and Sperber above. That said, he use of metarepresentations as premises is a defining feature only of the third of Sperber’s strategies: sophisticated understanding. Employing this strategy, the speaker does not assume that the speaker is either necessarily competent or benevolent, but only that she intends to appear so. Given this assumption, hearers may still identify the speaker’s intended meaning, even if the informative intention is not fulfilled. That is to say, a hearer may realise what a speaker intended him to believe, even though he may not go on to believe it as a result. Like cautious optimism when it involves false belief attribution, the process that allows such interpretations is clearly both metarepresentational and reliant on full-blown belief attribution: given what the speaker has done (i.e. said that ‘P’), the hearer must work out what effects the speaker could have intended her utterance to have, and for this he must attribute to her certain beliefs about what effects her utterance could have had if he had accepted it. Such inferences cannot be the result of processing a stimulus simply by following a path of least effort without giving any consideration to the nature of the stimulus or to the speaker’s intentions: both must be explicitly represented and must play a direct causal role (i.e. serve as premises) in the interpretation process. Table 8.2 summarises the above analysis of Sperber’s interpretation strategies. It shows that as long as the recognition of ostension is
186
Assertion
Table 8.2 Summary of relationship between interpretation strategies and metarepresentational requirements Metarepresentation of utterance necessary as initial premise?
Belief attribution necessary?
Metarepresentation of intended meaning necessary?
Naïve optimism
No
No
No
Cautious optimism (not involving false belief attribution)
No
No
Cautious optimism (involving false belief attribution) Sophisticated understanding
Yes
Not necessarily full-blown belief: knowledge attribution may be sufficient Yes
Yes
Yes:
Yes
Yes
characterised in non-representational terms, then the metarepresentation of the utterance is not required as an initial premise in the interpretation process for either naïve optimism, or cautious optimism that is not reliant on the attribution of false belief. Similarly, neither requires that the speaker’s intended meaning be metarepresented as such: a child, for example, could respond appropriately to an utterance by following either of these strategies, without explicitly representing the effect of the utterance on him as the speaker’s intended meaning. (However, as was noted above, the disadvantage of this position is that it does not characterise the child as recognising the speaker’s behaviour as goal-directed.) While belief attribution is not necessary for naïve optimism, some form of either belief or knowledge attribution is required for both forms of cautious optimism. The important point is that this does not have to serve as a premise in the interpretation process unless this process rests on a false belief. Sophisticated understanding does require a full-fledged concept of belief, as well as the explicit representation of the speaker’s utterance as an initial premise, and of the speaker’s intended meaning as an output. The preceding discussion has a number of consequences. First, it shows that a form of ostensive-inferential communication which does not rely on metarepresentational abilities can be conceived of. In circumstances in which the hearer can identify the speaker’s intended message either by maximising relevance, or by optimising relevance
Assertion and Relevance 187
in the manner of the more basic form of cautious optimism, then no metarepresentational abilities need be posited. It is important to keep apart notions of metarepresentation and inferential communication: under the right circumstances, you can have the latter without the former. Second, it shows that it is only necessary for the hearer to represent the utterance as an initial premise if it is also necessary to attribute to the hearer a false belief in order to arrive at the intended interpretation, that is to say, in the third and fourth of the strategies in Table 8.2.14 It needs to be stressed, though, that what is being claimed is that the hearer does not need to metarepresent the utterance as an initial premise if he employs a strategy of naïve optimism or the first type of cautious optimism. In both these cases, however, it may be necessary for a theorist to employ such a premise in providing a rational reconstruction of the interpretation process. The claim is thus that the inferential process of interpretation does not necessarily follow the same steps as an argument aimed at showing that an interpretation is rational. The difference between a theorist’s perspective and a hearer’s cognitive processes also underlies the third point that follows from the discussion in this section. Both hearer and theorist are interested in speaker-meaning, but it does not follow from this that the hearer necessarily represents the speaker’s meaning qua speaker meaning. That is to say, he may not necessarily partition all of those implications he derives from interpreting the speaker’s utterance into those that are intended by the speaker (i.e. the implicatures of the utterance) and those that are not. This may be because he is incapable of doing so, or because he has no interest in doing so. Consider (32) spoken to a child in order to prevent him from eating a sweet he has picked up off the floor. This could have the desired effect without the child representing the speaker’s meaning as in (33). Rather, he could simply follow a path of least effort and derive (34), which, combined with his assumption that dirty objects should not be eaten, would direct him not to eat the sweet (and perhaps allow him to derive further cognitive effects).15 (32) It’s dirty (33) The sweet is dirty (34) She intends [me to believe/know [that this sweet is dirty]] The final point that results from the preceding discussion relates to claims that have been made within Relevance Theory for pragmatics module. As this is rather a lengthy issue, it gets its own subsection:
188
Assertion
8.6.1 A pragmatics module? Because the most basic, and developmentally earliest (according to Sperber 1994) interpretation process is a relevance maximising procedure, rather than one of relevance optimisation, it can be explained in terms of general cognitive tendencies, avoiding the need to posit a specific interpretation module. This is at odds with the position taken by Serber and Wilson, who claim that utterance interpretation relies on procedures specific to that process, and that this warrants positing a dedicated module. The arguments for such a module have been presented in a number of places (Sperber 2000; Sperber and Wilson 2002), and are summarised by Wilson (2003). They can be termed ‘the underdeterminacy argument’, ‘the overtness argument’, and ‘the effect argument’. I consider each in turn below. The underdeterminacy argument rests on the observation that, where non-ostensive intentional behaviour is concerned, the effects that a speaker can hope to achieve in a certain situation are quite limited, and that this greatly facilitates the task of inferring the speaker’s intention. Someone jumping up and down under an apple tree, the argument goes, can rationally only expect to have a limited number of effects, making it easy to form reliable hypotheses about his intentions (i.e. that he intends to grasp an apple or shake one from the tree). Sentence meaning, on the other hand, greatly underdetermines the speaker’s intended meaning, making the inferring of the speaker’s intention a far more challenging task, one which requires a dedicated procedure. This argument fails to acknowledge, however, that although linguistic meaning is indeed under-determinate as regards speaker meaning, language does allow much richer and more fine-grained evidence to be presented than do other forms of behaviour. Moreover, as the previous discussion has shown, even given its underdeterminacy, a speaker’s meaning can be identified by following the relevance-maximising procedure that is used when processing non-ostensive stimuli. Thus, the underdeterminacy of linguistic meaning is not a strong argument for a pragmatics sub-module. The overtness argument points out that Grice’s conception of ‘meaningnn (1989: 213–23), which distinguishes linguistically communicated meaning from natural forms of meaning, requires that the speaker must not only have an informative intention but intend that this intention be recognised by the audience. Interpreting such behaviour thus requires the attribution of a multi-level intention of the form ‘the speaker intends me to know that she intends me to believe that P’. Given that two-year olds, on the one hand, engage in basic
Assertion and Relevance 189
communication but, on the other, fail to demonstrate any ability to form such higher-order representations in false-belief tasks, a possible solution to this problem is to posit a comprehension module which is available to such children and is able to process such higher-order representations, but whose representations are unavailable to other modules of the mind, such as the general mind-reading module. Consequently, although able to represent the content of beliefs attributed to others as representations, the child is unable to use these representations in inferences aimed at explaining or predicting behaviour, the argument would go. As Wilson acknowledges (2003: 313, fn. 8), one response to this argument is to ask whether the sort of communication that children are engaging in is indeed full-blown ostensive communication. It could well be that they are merely maximising relevance, the higher-order intentions of the speaker not manifest to them. The reason that they may seem to be engaged in ostensive communication is that adults assume that children are characterising adult ostensive behaviour as an expression of the sort of higher-order intentions that characterises Grice’s meaningnn , when in fact they treat it as merely attention-directing. At the same time, adults may also mistakenly characterise a child’s attention-directing behaviour as being driven by a full-blown Gricean communicative intention (i.e. one reliant on the notion of belief). An important question raised by the claim that two-year olds have the ability to attribute a communicative intention of the Gricean-type is what use this would be to them if they were also unable to attribute belief more generally. The crucial effect of the recognition of the communicative intention is to make it mutually manifest that the speaker has a certain informative intention and thereby encourage the hearer to infer what the content of this might be on the basis of assumptions about what the speaker might believe to be relevant to the hearer. Without the ability to represent such beliefs, though, it is not clear what use the attribution of a higher-order intention might be. In a shared cognitive environment, the same stimulus will be relevant in the same way to all those who share it. This is just the type of communication predicted above to be successful in symmetrical communication between non-belief-attributors. Mutuality is needed when a stimulus is employed that requires that the audience adjust their representation of the shared environment in a manner indicated by the speaker. By making her informative intention overt, the speaker invites the addressee to attune his cognitive environment to hers on the basis of her stimulus, what is manifest to him about her beliefs and desires, what is manifest to them
190
Assertion
about her assumptions about their assumptions and so on.16 If children under four are unable to mentally represent assumptions as attributed to their speaker, it is not clear what would be the use of attributing a communicative – rather than merely informative – intention. Thus the overtness argument wrongly assumes that complex Grice-style reasoning must underlie the interpretation of ostensive stimuli. But it has been shown above that this need not be the case (and is not required to be so by the more basic of Sperber’s interpretation strategies), and so the overtness argument does not make a strong case for a pragmatics module. The effect argument goes as follows: general intention attribution works by first identifying an effect that the agent could have both predicted and desired and then assuming that this was the intended effect; this procedure is not capable of accounting for utterance interpretation because the desired effect just is the recognition of the speaker’s intention. However, this again assumes that all utterance comprehension is aimed at arriving at a complex representation of the speaker’s communicative intention. Once this assumption is dropped, though, utterance comprehension and general intention attribution can be seen to follow the same pattern: when following strategy of naive optimism or the first form of cautions optimism, the hearer – by seeking to maximise relevance – identifies the effect that the speaker could have predicted and then assumes that this is what she intended. In other words, naïve optimism and the first type of cautions optimism are effectdriven: the communicator produces a stimulus which the addressee processes by seeking to maximise relevance. If the effect achieved is the one intended, then communication succeeds; if it is not, then it fails. The effect is not calculated on the basis of a recognition of the speaker’s intentions; rather, the effect achieved is the only one considered, any interpretations requiring the attribution of a false belief not being considered. In sum, neither the underdeterminacy, overtness nor the effect argument provides a compelling case for a pragmatics module. Furthermore, it is not clear what sort of evolutionary account one could give for a pragmatics module. If this has ostensive acts as its domain, then these must have existed prior to the evolution of the pragmatics module, otherwise there would be nothing for it to evolve to track. However, if a pragmatics module is necessary for the interpretation of ostensive acts, then it is not clear why such acts would exist prior to the development of a the module: they are only useful if they can be recognised as such, and if they can be recognised as such without domain-specific
Assertion and Relevance 191
cognitive abilities, then why should these develop? Sperber’s answer to this question is that early, simple ostensive-inferential communication was interpreted by employing general mind-reading abilities, but that once this was underway ‘a more specialised adaptation aimed at comprehension evolved’ (2000: 130). However, this position concedes that a pragmatics module is not a prerequisite for successful ostensive-inferential communication. But even though there is no clear case for a pragmatic module, it is worthwhile considering why one might feel the need to postulate one (cf. Carston 2002: 132). When Relevance Theory was first developed, part of the aim was to respond to Fodor’s (1983) claim that there could be no theory of central cognitive systems because of what has become known as ‘the frame problem’ (the problem of limiting the premises that can be employed in a process of nondemonstrative inference such that the problems it sets out to solve become tractable). By arguing that human cognition tends towards maximising relevance, Sperber and Wilson put forward a robust challenge to Fodor’s pessimism. Since Fodor (1983), however, the original distinction he drew between central and modular systems has been questioned by many, with a number of theorists now arguing that the mind is modular through and through (Sperber 1994; Pinker 1998; Carruthers 2005). In such a revised picture, the question of which module is responsible for utterance interpretation naturally arises. Given that linguistically communicated meaning, on the Gricean conception, is intentional, it is natural to assume that the putative theory-of-mind module is a likely candidate. However, as this appears to come online (in the crucial respects) after communicative abilities develop, the idea that there is a sub-module responsible for utterance interpretation becomes attractive. Indeed, the arguments presented for a pragmatics module are essentially arguments against the view that utterance interpretation falls within the domain of a theory-of-mind module, rather than being motivated by any explanatory need that the nonmodular central-systems story was unable to fulfil (cf. Bloom 2002, who also argues against a pragmatics module). The picture developed here, by contrast, is neutral concerning the modularity of central systems. It simply follows the original relevance-theoretic line that central systems, be they modular or not, seek to maximise relevance. Mindreading abilities have a role to play, but this is not to say that utterance interpretation is part of the domain of a mind-reading module. If this position is adopted, then the need to argue for a pragmatics module disappears.
192
Assertion
8.6.2 Summary of section 8.6 This section has considered the metarepresentational requirements of pragmatic inference. The conclusion has been that the ability to metarepresent is only a necessity for sophisticated interpretation strategies relying on the attribution of a false belief. If false-belief attribution is not required (and the speaker is benevolent), then simply by following a path of least effort until that effort is justified, the hearer will indentify the speaker’s intended message. This view of the relevance-driven interpretation procedure undermines the case for a pragmatics module.
8.7 Comparison with other frameworks The view of assertion and the role of declarative mood in the performance of this act that has been put forward in this chapter has drawn inspiration from other frameworks. In this section, I highlight these. 8.7.1 Stalnaker, context change and representation Like Huntley (1984) and Farkas (1992) (both discussed in Chapter 6), the present account sees the declarative mood as apt for the representation of particular worlds. However, on the current account, this is a result of this mood presenting proposition as candidates for use as a premise in a deductive process. Because this deductive process aims at consistency, the set of propositions that is its output can serve as a representation of a particular world. In the case of deductions performed over factual assumptions, the aim of the consistency requirement is to preserve truth, and hence the particular world represented is the actual world. In other words, as the declarative can be used to present propositions as relevant in a factual context, it can be used to represent the actual world. If used to present a proposition as relevant in its own right in an embedded context, though, then the consistency-aiming deductive processes need not be geared to deliver true outputs, and the world represented need not be the actual world. It may be, for example, a fictional world. The current account is thus unique in relating particular world representation to the potential to serve as a premise in inference, this potential in turn being due to the proposition being expressed in a particular format. In linking the declarative to world representation in this way, the current account avoids the problems raised by simply postulating that the declarative is related to particular worlds, non-declaratives to sets of worlds. A major problem with this approach, recall from Chapter 7,
Assertion and Relevance 193
is that it fails to provide an account of belief reports. The objects of belief reports, though generally declarative, do not pick out individual worlds but sets of worlds. What they represent is the way the actual world is according to the attribuee. This world, though, could be any of those in which the object of the belief report is true. So, according to the current account, both declaratives and non-declaratives, being propositional, pick out sets of possible worlds. What makes the declarative distinct is that it presents its proposition as a premise. This means that the set of worlds it picks out can be reduced. If that set of worlds is the content of a representation of the actual world, then uncertainty about the nature of that world is reduced by establishing with which other propositions the proposition expressed by the declarative clause is compatible. Representation, in the sense of a guide to how the world is, on this account, is a consequence of aptness for inference, not of mere propositionhood. This view of representation owes a great deal to my reading of Stalnaker (see Chapter 5). For me, the main insight to be drawn from Stalnaker does not relate to the relationship between assertion and common ground, but to the relationship between assertion and representation. For Stalnaker, recall, to represent the world as being a certain way is to locate the actual world in the basic context set. This is a set of worlds picked out by a set of consistency-aiming propositions, and one locates the actual world in this by making an assertion. In other words, to represent the world as being a particular way is to make a consistency-claim. In terms of the present proposal, the speaker manifests her intention to do this by presenting the proposition expressed by her utterance as relevant in a factual context. This intention is manifested in part by her choice of linguistic form, for the declarative mood has the function of presenting the proposition expressed as relevant in its own right in a context.17
8.7.2 Commitment accounts of assertion The first half of this book was concerned primarily with philosophical accounts of assertion. I came down in favour of broadly Dummettian views of assertion, as defended by Brandom and Barker, with Barker offering the most worked-out account of the role of the declarative in the performance of this act. I noted, though, that the present proposals were not in competition with these views, but intended, rather, as complementary. Recalling, from Chapter 2, Brandom’s distinction between explaining what the ‘trick’ of assertion consists in and showing how it
194
Assertion
is done in creatures wired up the way we are, my aim in this chapter has primarily been to contribute to the second project. That said, there are parallels between the philosophical account of assertion favoured in the first half of the book and the cognitive account presented here. These parallels are to be found primarily the way in which nonassertoric uses of the declarative are explained. In the discussion of the declarative as an assertion sign, in Chapter 4, it was argued that nonassertoric uses fall into two groups: those where the consequences of assertion are restricted, such as supposition and fiction, and those where the consequences of witnessing the act are the same as those of accepting an assertion of the same propositional content, as in the case of explicit performative utterances and directives performed through the use of future-tense declaratives. This explanatory strategy is mirrored in this chapter: cases such as supposition and fiction are cases of the proposition expressed by a declarative sentence being presented as relevant in its own right in an embedded – rather than factual – context, thereby limiting the consequences of acceptance; explicit performatives, declarations and declarative directives share with assertion the fact that they present the proposition expressed as relevant in its own right in a factual context, but they differ from assertion in that merely witnessing the act is sufficient grounds for acceptance – no further reason may be sought. They do not therefore have the potential to trigger the cognitive safeguards that, in part, characterise assertion.
8.8 Conclusion The reason that the declarative mood is apt for the making of assertions is that, uniquely, it presents the proposition expressed as relevant in its own right. That is to say, it presents the proposition as a premise to be used in inference. This has a number of consequences. First, it enables the declarative to be used to represent particular worlds, including the actual world. This is so because world representations aim at consistency and consistency is established through inference. One type of world-representing set of assumptions is an individual’s set of factual assumptions. To assert is, in part, to present a proposition as relevant in its own right to an individual. This is done by employing the declarative mood to signal that the proposition expressed is relevant in a context formed of factual assumptions. Allowing one’s thoughts to be influenced in this way is potential advantageous but also highly risky, for the communicator may be incompetent, or have interests that diverge from
Assertion and Relevance 195
one’s own. Consequently, safeguards – both social and cognitive – have emerged, and it is these that distinguish assertion. The cognitive safeguards are the ability to see utterances for what they are: representations, rather than information. This requires an ability to metarepresent, and one cannot be a competent asserter without this ability. That said, however, it does not follow from this that the interpretation of assertions always relies on metarepresentation. Direct information-transfer, with hearers treating utterances of declarative forms as information, can and does occur, as a number of authors have argued. However, it is always open to the competent asserter to step outside, as it were, this information-transfer process and see it for what it is: the production and consumption of representations. In this way, the danger of allowing others to add to one’s stock of beliefs is mitigated.
9 Conclusion
In this book, the phenomenon of assertion has been discussed from a variety of perspectives, and the picture that emerges perhaps seems somewhat contradictory. On the one hand, I have come down in favour of analyses that view assertion as an act which is constituted by the utterance of a declarative sentence and the concomitant undertaking of a particular set of commitments, a view found in the work of Dummett, Brandom and Barker. According to this view, assertion is not to be explained in terms of the attitudes it expresses, but in terms of its practical and inferential consequences. By contrast, the view of assertion presented in the previous chapter could be viewed as a belief-expression account of the broadly Gricean type: the speaker has a thought and chooses a stimulus by which to communicate that thought to the hearer. In this conclusion, I want to suggest that these two positions are not incompatible, but result from distinct perspectives being taken. On more than one occasion in the book, I have noted Brandom’s distinction between explaining what the trick of assertion consists in and how it is performed by ‘creatures wired up like us’. The distinction is between a conceptual analysis of assertion and a cognitive model of utterance interpretation. From the former point of view, assertion is to be analysed in terms of the commitments associated with the employment, under appropriate conditions, of a particular linguistic form, viz. the declarative mood. The notion of belief, following Dummett, can then be explicated in terms of this act. From the perspective of a cognitive model of utterance interpretation, however, there are two main reasons why this will not do. First, what is encoded by an assertion grossly underdetermines its truth conditions, suggesting that there must be another level of representation which does encode the truth conditions of our assertoric utterances, and over which inference can take 196
Conclusion 197
place. Belief is the obvious candidate. Second, humans are not borne asserting, and assertion is not a practice that in our primate cousins engage in, so we need to tell a story about how the practice develops, both in the individual, and in the species. From that perspective, the view that belief is prior to assertion is compelling: there must have been beliefs before there were utterances, else there would have been nothing to say. A cognitive perspective thus gives priority to belief over assertion and views assertion, as Frege did, as a means of expressing thoughts and judgements. These two positions can be reconciled if the human ability to ask questions about meaning and attitudes towards it – an ability which underlies the cognitive science project – is explained in terms of competence in assertion. Assertion, we have seen, requires meta-linguistic awareness: one cannot ask whether what another has said is true or false unless one can represent and reflect upon her utterance (as opposed to merely responding to it with the appropriate behaviour). Thus the cognitive safeguards that developed to protect the individual from misinformation through linguistic communication require that the individual has a degree of metalinguistic awareness. If this is correct, then the notion of assertion is absolutely fundamental both to the human ability to ask questions about meaning and to our possession of a notion of belief: both are by-products of a need to protect against misinformation and/or exploitation. Moreover, there is evidence, as noted in Chapter 3, that this process is replicated at the ontogenetic level: children appear to need a meta-awareness of declarative syntax in order to develop a full-fledged notion of belief. If this is correct, then the conceptual priority of assertion over belief reflects the ontogenetic and phylogenetic needs to develop a notion of imperceptible drivers of action that is grounded in perceptual drivers of action. Once this notion of imperceptible action-drivers is grasped, however, new perspectives become available: armed with the concept of belief, psychological models of communication can be developed in which belief is fundamental, and assertions relegated to the role of expressing those beliefs. The impoverished representations that utterances encode can be viewed as mere indicators of meaning encoded at the level of thought. This is the fully determinant level at which inference takes place, assertions serving as warrants that licence inferences, rather than as premises themselves. Further reflection, however, leads to the realisation that assuming such a level of representation, returns us to assertion, this time viewed through Frege’s lens. The lesson drawn from Frege in Chapter 4, recall,
198
Assertion
was that if one wants to posit a symbolic system aimed at developing an accurate world-representation, then one has also to posit, on pain of regress, a form such that tokening a form of that type results in undertaking a species of commitment, one that restricts which other representations of that format can also be tokened. In other words, one needs an assertion sign, whereby tokening one representation of this type rules out the possibility of tokening certain others, in the way that making one assertion removes the entitlement to make certain others. It is because of these constraints that assertion can function as a means of representing how the world is: if one assertion did not preclude others, then it would be uninformative as to the way the world is. In Stalnaker’s terms, assertions rule out possible worlds, and thereby reduce uncertainty as regard the state of the actual world. Thus it seems that if, in a desire to build a cognitive model of utterance interpretation, we posit forms whose tokening constitutes acceptance, as in Sperber and Wilson’s notion of a factual assumption, then we are still treating the notion of assertoric commitment as fundamental, even if we do not treat the speech act of assertion as basic. From a philosophical point of view, however, this move will not be wholly satisfactory, for we leave the notion of commitment unanalysed. And commitment must, I think, be explained in social terms, which once more gives precedence to the illocutionary act of asserting and its consequences. The extent to which cognitive approach to utterance interpretation squares this circle depends on whether the human notion of belief that results from our competence as asserters cuts nature at its joints, as some have argued, or whether it is merely a useful idiom for explaining behaviour, as others have claimed. On the former view, in coming up with the notion of belief, humans made a discovery; on the latter, we invented a tool. I have found the assumption that beliefs are physically instantiated mental representations with causal powers a fruitful working hypothesis, but I concede that this hypothesis is but a metaphorical extension of the observation that assertion, qua social practice, places constraints on subsequent behaviour, both verbal and otherwise.
Notes
1
Introduction
1. Pagin (2008) provides a very thorough overview of assertion from a purely philosophical perspective.
2 Assertion in Speech-act Theory 1. It is not clear in which group Austin would be best placed. As Hornsby (2006: 906) notes, when Austin was writing, the camps had not yet been marked out. 2. Austin (1950/1979: 120) treats ‘statement’ and ‘assertion’ as denoting essentially the same phenomenon. I will do the same. 3. For criticism of this dimension of the distinction, see Recanati (1987: 161–3). 4. Note that on Searle’s view, absence of hearer ignorance has no effect on felicity if the speaker is unaware of this ignorance. 5. The type of intention is important: it must be what Bach and Harnish term an R-intention, i.e. one that is ‘intended to be recognized as intended to be recognized’ (1979: xiv). That such intentions are central to linguistic communication has its roots, of course, in Grice (1989: ch. 14). The emphasis given to such intentions is a major difference between Bach and Harnish and Searle. 6. Cases such as these are discussed in detail by Alston (2000: 40–50) in his critique of Schiffer’s (1972) intention-based analysis of illocutionary force. 7. Millikan herself stresses the true-belief-forming history of indicative mood rather than the practice of assertion. For her, having this function is constitutive of a linguistic form counting as an instance of indicative mood (Millikan 1984: 53). 8. For detailed discussion of the notion of direction of fit, see Humberstone (1992). 9. For a useful summary of Alston’s proposals, with discussion and criticism, see Barker’s review (Barker 2002). 10. Frege held that only ‘yes/no’-questions and assertions express propositions (‘thoughts’ in his terminology). Orders cannot be said to express propositions, he claimed, because the question of truth does not arise for them (Frege 1918–19/1997: 329). For discussion of this aspect of Frege’s thought, see Dummett (1981: 307–8) 11. This second position is considered, and/or adopted by (Strawson 1964; Stenius 1967; McGinn 1977; Bach and Harnish 1979; Segal 1991; Harnish 1994; Boisvert and Ludwig 2006). 12. We return to this issue in Chapter 6, when we take a detailed look at satisfaction-based accounts of linguistic mood. 199
200
Notes
13. For a discussion of the problem of defining propositions, see Schiffer (2006). 14. Importantly for Brandom, asserters do this by employing linguistic form specified for making such undertakings: the way we recognise a declarative sentence type, on his account, is by identifying a form the utterance of which results in a speaker assuming the authority to pass on entitlements and commitments and undertaking the responsibility to defend these (Brandom 1994: 168). Thus one cannot refute Brandom’s claims by arguing, as Pagin (2004: 840) does, that Brandom’s account predicts that an utterance of ‘I hereby authorize you to claim whatever follows from P and I undertake the responsibility of justifying P’ should count as an assertion that P, when in fact it does not. This is because, for Brandom, the commitments undertaken by assertion are associated with the utterance of a particular linguistic form. It does not therefore follow that an attempt to undertake those commitments by other means should count as a token of the same assertion-type. 15. Other means by which a proposition can be put forward as true, such as intimating that P, giving to understand that P, strongly hinting that P, and conversationally implicating that P could be distinguished, on Brandom’s account, from assertion by the fact that these are means of avoiding undertaking public commitment to P, and/or of having to provide reason for P. 16. The precise sense in which assertions are ‘fodder for inference’ is discussed in Chapter 4, section 3. 17. Brandom uses a score-keeping framework (Lewis 1979) to model how these commitments are tracked by interlocutors. A similar approach is taken by Green (2000). 18. Indeed, content is explained in terms of force on Brandom’s account: conceptual distinctions are defined in terms of the different inferential consequences that would result from asserting a sentence S, if a constituent of S were substituted for by another. In other words, conceptual content is explained in terms of the consequences of assertion: ‘Semantics must answer to pragmatics’ (Brandom 1994: 83). 19. Or at least this is Brandom’s stated view: not all authors agree that his position is as deflationist as he claims. See Bar-on and Simmons (2006). 20. The notion of commitment is not expanded on in Barker (2004), but elsewhere he presents a position similar to Brandom’s in that to defend a commitment state is seen as taking epistemic authority, where this is understood as to ‘represent oneself as able to offer reasons for that state’ (2007: 193). 21. Barker uses the term ‘complex’ where I use ‘state of affairs’. As far as possible, I attempt to present his ideas in theory-neutral terms. 22. Recall Dummett’s distinction between assertions which have verifiable truth condition and those which do not. While it is not clear that this maps directly to Barker’s distinction between reportive and expressive assertions, there is certainly some overlap. 23. ‘Why does a language like ours allow for orders rather than merely reports about audience-directed desire states? The order, unlike the question, preempts dissent.’ (Barker 2004: 44, fn.18). 24. Barker also appeals to cognitive science in his explanation of semantic relationships such as entailment. For him, to assert that an entailment
Notes 201 relationship hold is to assert that tracking state A follows computationally from tracking state B. From this it follows that the rightness of a reasoning procedure depends on how the cognitive system is set up. This commits him to the view that reason is not eternal, but historically contingent (Barker 2004: 215–19). He claims that relativism does not follow from this: ‘Reason is historically conditioned, but at the same time relativism is false. To say that an inference pattern is valid is not to assert that it is accepted by members of my linguistic group or cultural community. It is to assert expressively commitment to a certain commitment law’ (2004: 220). 25. This second stipulation is clearly indebted to Barker’s (2004) account of the major mood/force distinctions, as discussed in section 2.3.2.
3 Assertion, Belief and Knowledge 1. This is not to claim that the monkey is asserting that a snake is present. The point is simply that a communicative act might require recourse to the notion of belief and belief expression for its full explication, yet not require belief attribution by its audience for its correct interpretation by its intended audience. 2. In Meaning, force and convention, Dummett says only that assertion is an exemplification of treating as true, not that it is the best exemplification of this practice (1993: 221). However, his view of the relationship between truth and assertion, as evidenced by his other writings (most notably his 1981), makes it reasonable to attribute to him the view that assertion is the practice that best exemplifies treating as true. 3. The picture painted in Jary (2010) takes its inspiration from Brandom’s (1994; 2000) development of Dummett’s ideas on assertion into an account of the sort of representation involved in belief attribution. In this, Brandom is indebted, as well as to Dummett, to Sellars (1956/1997). In Jary (2010), I also discuss evidence the has been taken as evidence of precocious false-belief attribution in children (Clements and Perner 1994; Onishi and Baillargeon 2005; Southgate et al. 2007; Surian et al. 2007). 4. See Dummett (1993: 149–50) for a discussion of the correct order of explanation in relation to language and thought. He argues there that the conceptual priority of language over thought would not be undermined by evidence that a creature could have ‘full-fledged thought’ prior to having language. 5. For a discussion of testimony from a philosophical perspective, see Coady (1992). Harris and his colleagues have done interesting psychological work on children’s reliance on testimony as a source of information about the world (Harris 2002; Clément et al. 2004; Harris 2004). 6. Green (2007) calls this the ‘extended senses view’. 7. McDowell doesn’t actually mention vervet monkeys. He uses an imagined species of bird to make his point. 8. Green (2007: ch. 3.6, fn. 43) takes issue with this definition of manifestness. He says that Sperber and Wilson’s official definition of manifestness ‘has it that a fact is manifest to an individual just in case that individual is capable of representing it mentally, and accepting it as true or probably true’. He then goes on to complain that ‘this definition counts as manifest
202
Notes
any fact that I am capable of thinking about and believing true or probably true, regardless of my evidence’. This criticism ignores Sperber and Wilson’s cognitivism, however. On their account, to unreflectively treat an assumption as true is to represent it in an unembedded format (i.e. as a factual assumption) (Sperber and Wilson 1986/1995: 73–5). This is not something that one can consciously decide to do. Rather, the cognitive system is set up, on their framework, to allow only assumptions for which one has sufficient evidence (or warrant) to have the status of factual assumptions. Thus one cannot decide that an assumption will be manifest regardless of the evidence: the available evidence (together with the individual’s cognitive capacities) determines what is manifest, and to what degree. Of course, one can reflectively decide that an assumption is true, but doing so will not allow that assumption interact willy-nilly with factual assumptions. Such reflective (i.e. non-factual) beliefs are hived off (by being embedded in factual assumptions), so that their consequence is limited. On this last point, see Sperber’s (1985: ch. 2; 1997) distinction between intuitive and reflective beliefs. 9. Williamson insists that a constitutive rule is not a convention, because it is not contingent (1996: 490). By this he means that there is no other rule that could apply to assertion such that the act in question would still be assertion. For critical discussion of Williamson’s notion of a constitutive rule see García-Carpintero (2004: 161–4) and Hindriks (2007: 394–9). 10. In relation to assertion and knowledge, it is worth recalling the discussion of McDowell in section 3.3. There it was suggested that creatures (such as early humans and young children) without the notion of belief in their conceptual repertoire might be able to engage in a form of proto-assertion, treating such acts as sources of knowledge, rather than expressions of belief. The idea behind such speculations is that these creatures lack the notion of belief. The conception of knowledge attributed to these creatures is a conception of information as something that one can either have or not have. The contrast available to these creatures, then, is not between knowledge and belief, but between knowledge and ignorance. For them, knowledge just is access to information, and is not analysable, from their perspective, as justified true belief, for they have no notion of truth or belief.
4
A Sign of Assertion
1. The term ‘proposition’ as used in this book and Frege’s notion of a ‘thought’ do not coincide. Frege held that thoughts could be expressed only by assertions and ‘yes/no-questions’ (Frege 1918–19/1997: 329). Propositions, in the sense intended in this book, can be explicitly expressed by any sentential speech act. 2. I take the distinction between expressing a premise and being a vehicle of inference from Vicente’s (2010) distinction between expressions and vehicles of thought. 3. While (3) and (4) aim to represent formal derivations, ordinary discourse is abound with enthymatic inferences and these could just as easily be used to test for assertoric force in an inferential integration test.
Notes 203 4. See Jary (2007: 214–17) for arguments that show that explicit performatives cannot serve as premises in inference. In that paper, I argue that explicit performatives are not assertions. Accepting that Peter’s utterance in (5) introduces a premise in this ‘inference’ does not conflict with this view, for what is asserted by Peter’s utterance is not the proposition that is expressed by the utterance (i.e. that Peter thereby asserts that 73 is a prime number), but the proposition expressed by the subordinate clause (i.e. that 73 is a prime number). 5. This means that, by his own admission, Pagin’s case against what he terms ‘social accounts’ of assertions is seriously undermined, for he accepts (2004: 833) that if ‘I hereby assert that 73 is a prime number’ counts as an assertion that 73 is a prime number, then his system of refuting social accounts by reformulating them as explicit performatives over-generates counterexamples. 6. Dummett is not, of course, the only writer who has made this claim. Others who are committed to it include Barker (2004), Brandom (1994), Geach (1965), and Williamson (1996). 7. This isn’t an argument against Green. His aim is not to show that the declarative is not, in any sense, an indicator of assertoric force. Rather, his claim is that the declarative does not function in the same manner in English as Frege’s assertion sign does in the Begriffsschrift. Green (1997; 2000) does show, however, that English has a sign that is a weak indicator of assertoric commitment regardless of its linguistic context. This is a first-person ‘as’parenthetical containing an assertive verb. For example, in uttering ‘If snow is, as I claim, white, then coal is black’, the speaker cannot avoid undertaking assertoric commitment to the proposition that snow is white, even though this is embedded in the antecedent of a conditional. The existence of such a form goes some way to undermining Davidson’s thesis of the autonomy of linguistic meaning. 8. We ignore here problems of reference in fiction. 9. This is not to deny that there are experimental modes of fiction that may allow this type of inconsistency, but the fact that these attempt to subvert standard expectation of how fiction works only serves to underscore the existence of such expectations. 10. This should not be read as a claim that supposition and fiction are identical, merely that they have relevant features in common. 11. Of course, you might want to claim that the conditions of truth are all there is to assertion, and all this talk of practical and inferential consequences is superfluous. What I aim to show in this section, though, is that thinking of assertion in terms of its consequences allows one to maintain the view that the declarative is a marker of assertoric force and allow that it has other uses without thereby claiming that these are indirect speech-acts. If this can be done, then it supports the view that assertion is best understood not only in terms of truth conditions, but in terms of its consequences, as Dummett suggests. 12. It might be argued that the hearer can respond to an utterance such as (12) as he would an assertion, thereby undermining the claim that such utterances are not indirect directives. For example, the hearer might respond to (12) by saying ‘Oh I will, will I? How interesting!’. But in responding thus, he is
204
Notes
refusing to treat the utterance as a directive and is instead treating it as an assertion, using the declarative form as an excuse to do so. This does not show that, when (12) treated as a directive, it is also an assertion. For that to be shown, it would have to be possible to respond in a way that indicated both acceptance of (12) as true and compliance. But one cannot felicitously respond to (12), spoken as a directive, by saying ‘That’s true’. 13. For insightful discussion of the parallels between assertion and promising, see Watson (2004). 14. This is not a mistake that Stainton makes. He correctly notes (2006: 221) that Dummett’s Assertion Thesis (though he doesn’t call it this) is compatible with the existence of sub-sentential assertions.
5
Assertion and Common Ground
1. This brief outline gives a rather simplified picture of Stalnaker’s account of assertion. In particular, it ignores the relationship between the common ground and beliefs about the common ground, considerations of which have been to the fore in Stalnaker’s later work (see, in particular, Stalnaker 2002). For a thorough mapping of developments in Stalnaker’s thinking on assertion, presupposition and common ground, see Simons (2003). 2. For a recent attempt to incorporate imperatives into a Stalnakerian framework, see Portner (2007). This work is discussed in Chapter 6, section 2.2. 3. Stalnaker’s dual use of the notion of representation (as something which both speakers and propositions do) reflects two ways in which possible worlds can be used in semantic theorising: as the content of a representation and as its object. An assertion picks out a set of possible worlds by virtue of its content, i.e. those in which it would be true. However, an assertion is also a representation of a particular possible world, viz. the actual world. As with the two notions of truth at play in Stalnaker’s model, the first use of possible worlds relates to sense, the latter to force. Others have used possible worlds in only one of these senses. For example, Huntley (1984) distinguishes between the declarative and the imperative in the following terms: a declarative, he says, involves deictic reference to a particular world at which it is to be evaluated as true, while an imperative lacks such a linkage to any particular world. This linkage explains, according to Huntley, the use of the declarative for assertion: asserting, on his account involves a deictic link to the actual world. Thus, because Huntley is concerned with issues of mood and force rather than sense and reference, possible worlds are employed to serve as what is represented by an assertion, rather than to constitute the content of an assertion. Because Stalnaker is concerned with both, he makes dual use of possible worlds, as the content and the object of representation. 4. The notion of presupposition goes back, of course, to Frege. For an introduction to – and history of – presupposition studies, see Levinson (1983: ch. 4). For more up-to-date discussion, see Beaver (2001) and Kadmon (2000). 5. That view was prevalent when Stalnaker started publishing on presupposition. Ironically, despite the influence of his work, it is still the dominant view over a quarter of a century later, even among those linguists who see themselves as implementing his framework.
Notes 205 6. Note that, if Stalnaker wants to distinguish asserting that P from conversationally implicating that P, then he will need to acknowledge that assertion has another essential effect on the context set: for an utterance to count as an assertion, even if it modifies the context set, the proposition that the speaker has said that P (where ‘say that P’ means something like ‘explicitly express P in declarative mood’) must also be added to the context set. In other words, unless some sort of explicitness condition is added to his framework, there is no way, in that framework, to distinguish asserting that P from conversationally implicating that P. 7. Stalnaker considers this objection, I think, in Common Ground (2002: 712–13). His response to it, as Abbott (2006: §4.1) implies, is hard to interpret. 8. This is not to say that all presupposition phenomena could be treated this way. In particular, a different sort of account will have to be given for the implications associated with contrastive stress and cleft sentences. These are discussed in Chapter 7.
6 Assertion and Mood 1. Cases such as (11) and (12) are argued by Sweetser (1990: 118–21) to be examples of speech-act conditionals, i.e. cases where the antecedent places a condition on the performance of the speech act indicated by the consequent. For an alternative view, and further references, see Siegel (2006). 2. The question of whether an account of the major moods should generalise to apparent embedded cases, such as we find in the Pendlebury cases, divides scholars. A major problem that we have already touched on is that most of the main sentence types do not appear to embed directly. The sentence in (i) is represented by a ‘whether’-clause in (ii), and this does not exhibit the subject-verb inversion that marks (i) as interrogative in mood, while we find the imperative in (iii) represented by a infinitival-clause in (iv). (i) (ii) (iii) (iv)
Are you hungry? He asked whether I was hungry You eat up! He told me to eat up
Huntley (1984: 113–17) argues that, semantically, infinitive complements such as we find in (iv) are of the same type as imperatives, and hence that imperatives do embed, differences between the main-clause forms used for issuing directives and their embedded counterparts being only syntactic. Certainly, all other things being equal, an account that provides a unified explanation of the contribution to utterance interpretation of main-clause mood and its (possible) embedded counterparts will be preferred over two separate accounts. Furthermore, there is good syntactic evidence that, whatever the case for imperatives and interrogatives, declaratives – again, our main concern here – do embed: a report of an assertion employs the same syntactic structure to denote the original assertion as is used to perform the original act, as illustrated by (v) and (vi): (v) The Earth is flat (vi) He said that the Earth is flat.
206
Notes
3. Boisvert and Ludwig are actually concerned with ‘asks that’ and ‘asks whether’, but the point generalises. 4. See Harnish (1994: 426–9) for discussion of Pendlebury’s approach. 5. Villalta (2000) also analyses subjunctives as requiring reference to an ordered set of worlds. As she does not consider the declarative, however, her work is not discussed here. 6. Recall Alston’s (2000) point that non-assertives require the speaker taking responsibility for a range of propositions not explicitly expressed by, but nonetheless associated with, the utterance (see Chapter 2, section 2.3). 7. Portner also posits a Question Set, consisting of sets of propositions. Interrogatives are then claimed to denote sets of propositions, following, among others, Karttunen (1977/2002). For criticism of this approach, see Stainton (1999). 8. Given that the ‘by’-adverbial in passive sentences is typically associated with agent role, then this analysis would mean that agent role is assigned to two arguments. This would be odd, but then (23)b is odd. 9. See Chapter 2, section 3.2 for a discussion of Barker’s notion of an expressive assertion. 10. Having this disposition is tantamount to asserting both conjuncts. Barker recognises this and acknowledges an alternative analysis on which ‘and’ connects assertions, rather than mere proto-assertions (Barker 2004: 89, fn. 7). The problem with this, though, is that it requires an alternative analysis of embedded conjunctions, such as ‘If S1 and S2 , then S3 ’, for in such cases the conjuncts are not asserted. 11. Biscuit conditionals, dubbed such after Austin’s (1956/1961: 158) famous example, can be seen as a subset of speech-act conditionals, see fn. 1, this chapter. 12. The original proposals made by Bach and Harnish are developed and refined in Harnish (1994), and subsequent publications. 13. Harnish (1994: 448) questions whether B’s response in (31) is in fact an imperative, asking whether it might be a case of discourse ellipsis, so that it is an infinitival continuation of an implicit ‘to get to the station I advise you to. . .’. That this is not the case can be seen by considering what form would be used if negative advice were given. If this were a case of ellipsis, one would expect the negation to be indicated by ‘not’, rather than ‘don’t’, but as (31)’ shows, this is not the case. (The negation test for imperativehood is found in Clark (1991: ch. 2)). (31)’ B: Don’t take a number 3 bus; that goes all round the houses. Take a 73 instead. 14. This problem become particularly pronounced when Sperber and Wilson’s approach to mood is applied to the analysis of the indicative/subjunctive contrast, as in Rouchota (1994). For discussion, see Jary (2002: 161–164; 2005: 155–158).
7
Assertion and Main Point
1. In languages other than English, the possibility of an embedded clause having main-point status appears to depend, to some extent, on whether it
Notes 207
2. 3.
4. 5.
6.
7.
8.
9.
10.
11.
exhibits the same syntactic features as a main-clause declarative. For example, in Spanish, the possibility of interpreting an embedded clause as the main point of the utterance depends on the verb being in the indicative, rather than the subjunctive, mood (Terrell and Hooper 1974), while German appears to require that the verb appear after the subject (i.e. verbsecond, as in main-clause declaratives), rather than in a clause-final position (Meinunger 2006). Furthermore, a recent survey of Scandinavian languages concludes that for an embedded clause to have main-point potential, the syntactic environment in which it occurs must permit verb-second (i.e. declarative) word order, even if that word order is not realised (Wiklund et al. 2009). In their analyses, both Meinunger and Wiklund et al. employ the taxonomy of predicates proposed by Hooper and Thompson (1973), which has also been influential in the discussion of the indicative/subjunctive contrast in Spanish. For extensive discussion of the indicative/subjunctive contrast in Spanish, and how it relates to assertion, see Jary (2009), and references therein. The literature on parentheticals begins, of course, with Urmson (1952). Potts takes the term ‘conventional implicature’ from Grice, of course, but he does not apply it to the same set of phenomena (e.g. ‘but’) to which the term is commonly applied. For a discussion of his use of the term, see Potts (2005: 5–13). Here I am only considering certain cases of his class of supplements. I have nothing to say about expressives. From now on, I will use ‘survival under negation’ as a cover term for passing all of the tests illustrated by (5). Example (10) illustrates another feature of Potts’s CIs: they are speaker oriented. The view that Peter is a convicted terrorist is not necessarily attributed to Jane in this utterance. Rather, commitment to this proposition is undertaken by the speaker. This ignores, for simplicity, conversational implicatures and the distinction that Potts draws between conventional and context-dependent presuppositions. This is not to say that the same point may not be felicitously presented as the main point a number of times in one stretch of discourse: recall Anthony’s repeated statement that Brutus is an honourable man (Julius Caesar, III, 2). What makes this repetition non-redundant, of course, is the gradual revelation of the attitude which accompanies that statement: at first Anthony appears sincere, later ironic. This very brief overview of Relevance Theory is elaborated on in Chapter 8, section 2. Those unfamiliar with the theory might want to read that section before continuing with this section. A well known-issue here is that this account requires that ‘NOBODY broke the window’ entails ‘Someone broke the window’, which is plainly wrong. However, the explanatory potential of the account justifies putting this concern on hold, pending an adequate theory of negation. It is considerations such as these that lead Carston (2002: 140–1) to the view that entailments that contribute to the relevance of the utterance in their own right should be classified as implicatures. The contribution to utterance interpretation that a proposition (entailed or not) expressed by embedded clause can make in its own right has been
208
Notes
recognised by Sperber and Wilson since at least the second edition of Relevance (1995), where they suggest that their definition of explicit communication needs to be revised in order to allow for explicit communication by means of embedded clauses (see p.182, fn. a). Such consideration led to the revised definition of explicature given by Carston (2002: 124). 12. One topic which has not been touched on in this chapter is Horn’s (2002) notion of ‘assertoric inertia’. According to Horn, entailments of utterances containing certain elements are marked as not contributing to the point of an utterance, and hence said to be ‘assertorically inert’. For example, ‘You almost died’ entails ‘You didn’t die’, yet one cannot say ‘How wonderful that you almost died’ to rejoice the fact that the addressee didn’t die. Assertorically inert entailments, in the terms of this chapter, thus seem to be linguistically precluded from contributing to the relevance of the utterance in their own right.
8
Assertion and Relevance
1. I am simplifying here in that I am ignoring contextual effects that do not result in the addition of propositions to the context. Sperber and Wilson also allow that a proposition is relevant in a context if it results in the elimination of propositions from that context, or in the strengthening or weakening of propositions in that context (Sperber and Wilson 1986/1995: 109–17). These sources of contextual effect play no role in the analysis put forward in this chapter. 2. I ignore for the moment the importance of the speaker’s competence and preferences and hence the difference between maximising and optimising relevance. This is addressed in section 6 of this chapter. 3. Sub-sentential assertions of the type discussed by Stainton (1993; 2005; 2006) are cases where a proposition is presented as relevant in its own right to an individual, but not by linguistic means. 4. The idea of metarepresentational abilities serving as cognitive safeguards is one that Sperber has been propounding for a long time, at least since his (1985). More recently, he has referred in lectures to these abilities a stemming for a need for ‘epistemic vigilance’. My own views have clearly been influenced by his work in this area. 5. One case in which a declarative will be interpreted in an embedded context, but still result in assertoric commitment, is if it is mutually manifest that the hearer intends her utterance to serve only as an expression of her belief. As noted in Chapter 2, section 2.1.1, Sperber and Wilson (1986/1995: 181) call utterances that communicate the proposition expressed ‘ordinary assertions’. ‘Assertions’ made without an intention to inform the hearer of the proposition expressed are, on this view, unordinary. 6. Another author who stresses the importance of relevance (in the nontechnical sense) to the analysis of assertion is Gauker (2007: 132). 7. What about embeddings such as ‘Somebody says that P’ and ‘Nobody says that P’? The latter can be seen as means of denying the relevance of P in any context. This is supported by the fact that, in Spanish, the embedded
Notes 209
8.
9. 10.
11. 12. 13.
14.
15.
clause after ‘Nadie dice que’ (‘Nobody says that’) is commonly subjunctive, which, on the current analysis, would mean that it is not marked as relevant in its own right. By contrast, ‘Somebody says that P’, presents P as relevant in an embedded context that represents a world-view, but does not say to whom this world-view belongs. As in cases where the holder of the world-view is named, such utterances can result in the proposition expressed by the embedded clause achieving main-point status, if the most accessible context is a factual one. Again, this will mean that the speaker is able to avoid assertoric commitment to the proposition expressed by the embedded clause, even though this is manifestly the main point of her utterance. The difference between factive and first-person belief reports is that, while uttering either commits the speaker to the proposition expressed by the complement clause, only by accepting the former does the hearer necessarily also take on that commitment. The results were statistically significant. Hence, I think, the temptation felt by some to view ‘I believe that P’ utterances as similar to explicit performatives (see, e.g. Wierzbicka 2006: 217). Clearly, such utterances are not explicit performatives. However, they share with these both a structural similarity and the feature of being less than amenable to truth judgements. The reason some might feel tempted to say that present-tense, first-person belief reports are not truth apt is because they are hard to view as being put forward for acceptance or rejection. On what grounds might we reject such an utterance? They are surely the best evidence that one get that the proposition expressed obtains, thus the intuition that a truth judgment in such cases applies to the object of the belief report, rather than to the proposition expressed by the utterance as a whole. Taken from the Guardian website 1s February 2005: http://www.guardian.co.uk/terrorism/story/0,12780,1403335,00.html. See Unger (2006) for an analysis of text coherence in terms of Relevance Theory. This is, of course, the central insight behind the classic theory-of-mind tests. Importantly, being able to track ignorance is insufficient for being able to predict behaviour based on a false representation of the world. The fact that John doesn’t know that the cookie is no longer in the jar is no reason for him to look in the jar to find it. What is good reason is his belief that the cookie is still in the jar. Note how this meshes with the observations, made in Chapter 3, concerning the relationship between assertion and belief. There it was suggested that a meta-awareness of declarative mood was required in order for a child to acquire the notion of belief. Here, it is claimed that such an awareness is needed before interpretation strategies relying on false-belief attribution can be employed. Cf. Recanati (2002: 113, original emphasis) ‘we don’t have to think about the speaker’s communicative intentions; we merely have to grasp the right meaning, where the right meaning is the speaker’s meaning – but we don’t have to think of it as the speaker’s meaning.’
210
Notes
16. This goes on ad infinitum, but becoming more weakly manifest at each stage such that after a few embeddings these manifest assumptions are not mentally represented. 17. Farkas (2003) and de Bustos and Aliaga (1996; Aliaga and de Bustos 2002) also link the declarative to context-change potential, but they do not do so in terms of a form’s ability to serve as a premise in inference.
References
Abbott, B. 2000: Presuppositions as nonassertions. Journal of Pragmatics, 32, 1419–37. Aliaga, F. and de Bustos, E. 2002: Mental spaces and epistemic attitudes: On the Spanish subjunctive/indicative alternation. In Gutiérrez-Rexach, J. (ed.). From words to discourse: Trends in Spanish semantics and pragmatics. Amsterdam: Elsevier. Alston, W. P. 2000: Illocutionary acts and sentence meaning. Ithaca, NY: Cornell University Press. Anscombe, G. E. M. 1959/1996: An introduction to Wittgenstein’s Tractatus. Bristol: Thoemmes Press. Astington, J. W. and Jenkins, J. M. 1999: A longitudinal study of the relationship between language and theory-of-mind development. Developmental Psychology, 35, 1311–20. Austin, J. L. 1950/1961: Truth. In Urmson, J. O. and Warnock, G. J. (eds). Philosophical papers Oxford: OUP. Austin, J. L. 1956/1961: Ifs and cans. In Urmson, J. O. and Warnok, G. J. (eds). Philosophical Papers. Oxford: OUP. Austin, J. L. 1962/1975: How to do things with words. Cambridge MA: Harvard University Press. Bach, K. 1994: Conversational impliciture. Mind and Language, 9, 124–62. Bach, K. 2010 Knowledge in and out of context. In Campbell, J. K., O’Rourke, M. and Silverstein, H. S. (eds). Knowledge and skepticism. Cambridge (MA): MIT Press. Bach, K. and Harnish, R. M. 1979: Linguistic communication and speech acts. Cambridge, MA: MIT Press. Bach, K. and Harnish, R. M. 1992: How performatives really work: a reply to Searle. Linguistics and Philosophy, 15, 93–110. Bar-on, D. and Simmons, K. 2006: Deflationism. In Lepore, E. and Smith, B. C. (eds). The Oxford handbook of philosophy of language. Oxford: OUP. Barker, S. 2002: Illocutionary acts and Sentence meaning. Mind, 111, 633–9. Barker, S. 2004: Renewing meaning. Oxford: OUP. Barker, S. 2007: Semantics without the distinction between sense and force. In Tsohatzidis, S. L. (ed.). John Searle’s philosophy of language: Force, meaning and mind Cambridge: CUP. Baron-Cohen, S., Leslie, A. M. and Frith, U. 1985: Does the autistic child have a theory of mind? Cognition, 21, 37–46. Beaver, D. I. 2001: Presupposition and assertion in dynamic semantics. Stanford: CSLI. Bloom, P. 2002: Mindreading, communication and the learning of names for things. Mind and Language, 17, 37–54. 211
212
References
Boisvert, D. and Ludwig, K. 2006: Semantics for nondeclaratives. In Lepore, E. and Smith, B. C. (eds). The Oxford handbook of philosophy of language. Oxford: OUP. Bosque, I. 1990: Las bases gramaticales de la alternancias modal. Repaso y balance. In Bosque, I. (ed.). Indicativo y subjuntivo. Madrid: Taurus Universitaria. Brandom, R. 1994: Making it explicit: Reasoning, representing, and discursive commitment. Cambridge, MA: Harvard University Press. Brandom, R. 2000: Articulating reasons: An introduction to inferentialism. Cambridge, MA: Harvard University Press. Breheny, R. 2006: Communication and folk psychology. Mind and Language, 21, 74–107. Carruthers, P. 2005: The case for massively modular models of mind. In Stainton, R. (ed.). Contemporary debates in cognitive science. Oxford: Blackwell. Carston, R. 2002: Linguistic meaning, communicated meaning and cognitive pragmatics. Mind and Language, 17, 127–48. Carston, R. 2002: Thoughts and utterances: The pragmatics of explicit communication. Oxford: Blackwell. Chierchia, G. and McConnell-Ginet, S. 2000: Meaning and grammar: An introduction to semantics. Cambridge, MA: MIT Press. Clark, B. 1991: Relevance theory and the semantics of non-declaratives. Unpublished thesis. University of London: London. Clément, F., Koenig, M. and Harris, P. L. 2004: The ontognesis of trust. Mind and Language, 19, 360–379. Clements, W. A. and Perner, J. 1994: Implicit understanding of belief. Cognitive Development, 9, 377–395. Coady, C. A. J. 1992: Testimony. Oxford: OUP. Cosmides, L. and Tooby, J. 2000: Consider the source: The evolution of adaptation for decoupling and metarepresentations. In Sperber, D. (ed.). Metarepresentations: A multidisciplinary perspective. Oxford: OUP. Davidson, D. 1967/2001: Truth and meaning. Inquiries into truth and interpretation. Oxford: OUP. Davidson, D. 1979/2001: Moods and performances. Inquiries into truth and interpretation. Oxford: OUP. de Bustos, E. and Aliaga, F. 1996: Indicative, subjunctive and context. In Gutiérerez-Rexach, J. and Silva-Villar, L. (eds). Perspectives on Spanish linguistics. Los Angeles: UCLA. de Villiers, J. G. and de Villiers, P. A. 2000: Linguistic determinism and false belief. In Mitchell, P. and Riggs, K. (eds). Children’s reasoning and the mind. Hove: Psychology Press. de Villiers, J. G. and Pyers, J. E. 2002: Complements to cognition: A longitudinal study of the relationship between complex syntax and false-belief understanding. Cognitive Development, 17, 1037–1060. Dennett, D. C. 1981: Brainstorms: Philosophical essays on mind and psychology. Brighton: Harvester. Dummett, M. 1981: Frege: Philosophy of language. London: Duckworth. Dummett, M. 1989/1993: Language and communication. The seas of language. Oxford: Clarendon. Dummett, M. 1993: Mood, force and convention. The seas of language. Oxford: Clarendon Press. Dummett, M. 1993: Truth and meaning. The seas of language. Oxford: Clarendon.
References 213 Farkas, D. 1992: On the semantic of subjunctive complements. In Hirschbuher, P. (ed.). Romance languages and modern linguistic theory. Amsterdam: John Benjamins. Farkas, D. 2003: Assertion belief and mood choice. Paper presented at ESSLLI, Conditional and Unconditional Modality Workshop, Vienna. Fodor, J. A. 1975: The language of thought. Cambridge, MA: Harvard University Press. Fodor, J. A. 1981: Representations. Hassocks: Harvester Press. Fodor, J. A. 1983: The modularity of mind. Cambridge, MA: MIT Press. Fodor, J. A. 1998: Concepts: Where cognitive science went wrong. Oxford: OUP. Fodor, J. A. 2008: LOT 2: The language of thought revisited. Oxford: OUP. Frege, G. 1897/1997: Logic. In Beaney, M. (ed.). The Frege reader. Oxford: Blackwell. Frege, G. 1918–19/1997: Thought/Der Gedanke. In Beaney, M. (ed.). The Frege reader. Oxford: Blackwell. García-Carpintero, M. 2004: Assertion and the semantics of force-markers. In Bianchi, C. (ed.). The semantics/pragmatics distinction. Stanford: CSLI Publications. Gauker, C. 2007: On the alleged priority of thought over language. In Tsohatzidis, S. (ed.). John Searle’s philosophy of language: Force meaning and mind. Cambridge: CUP. Geach, P. T. 1960: Ascriptivism. The Philosophical Review, 69, 221–5. Geach, P. T. 1965: Assertion. The Philosophical Review, 74, 449–65. Ginet, C. 1979: Performativity. Linguistics and Philosophy, 3, 245–65. Giorgi, A. and Pianesi, F. 1997: Tense and aspect: From semantics to morphosyntax. Oxford: OUP. Green, M. S. 1997: On the autonomy of linguistic meaning. Mind, 106, 217–43. Green, M. S. 1999: Illocutions, implicata, and what a conversation requires. Pragmatics and Cognition, 7, 65–91. Green, M. S. 2000: Illocutionary force and semantic content. Linguistics and Philosophy, 23, 435–73. Green, M. S. 2002: The inferential significance of Frege’s assertion sign. Facta Philosophica, 4, 201–29. Green, M. S. 2007: Self-expression. Oxford: OUP. Grice, P. 1989: Studies in the way of words. Harvard: University Press. Grice, P. 2001: Aspects of Reason: Clarendon Press. Hale, C. M. and Tager-Flusberg, H. 2003: The influence of language on theory of mind: A training study. Developmental Science, 6, 346–59. Hare, R. 1950/1971: Appendix. Practical inferences. London: Macmillan. Hare, R. 1970/1971: Meaning and speech acts. Practical inferences. London: Macmillan. Hare, R. 1989: Some subatomic particles of logic. Mind, 98, 23–37. Harnish, R. M. 1994: Mood, meaning and speech acts. In Tsohatzidis, S. (ed.). Foundations of speech act theory: Philosophical and linguistic perspectives. London: Routledge. Harris, P. L. 2002: Checking our sources: the origins of trust in testimony. Studies in History and Philosophy of Science Part A, 33, 315–33. Harris, P. L. 2004: What do children learn from testimony? In Carruthers, P., Siegal, M. and Stich, S. (eds). Cognitive bases of science. Cambridge: CUP.
214
References
Heim, I. 1983/2002: On the projection problem for presuppostions. In Portner, P. and Partee, B. H. (eds). Formal semantics: the essential readings. Oxford: Blackwell. Heim, I. 1992: Presupposition projection and the semantics of attitude verbs. Journal of Semantics, 9, 183–221. Hindriks, F. 2007: The status of the knowledge account of assertion. Linguistics and Philosophy, 30, 393–406. Hintikka, J. 1962: Knowledge and belief. Ithaca, NY: Cornell University Press. Hooper, J. and Thompson, S. A. 1973: On the applicability of root transformations. Linguistic Inquiry, 4, 465–79. Horn, L. R. 2002: Assertoiric inertia and NPI licensing. CLS: The panels, 38, 55–82. Hornsby, J. 1986: A note on non-indicatives. Mind, 95, 92–99. Hornsby, J. 2006: Speech acts and performatives. In Lepore, E. and Smith, B. C. (eds). The Oxford handbook of philosophy of language. Oxford: OUP. Humberstone, I. L. 1992: Direction of fit. Mind, 101, 59–83. Huntley, M. 1984: The semantics of the English imperative. Linguistics and Philosophy, 7, 103–33. Ifantidou, E. 2001: Evidentials and relevance. Amsterdam: John Benjamins. Jary, M. 2002: Mood in relevance theory: A re-analysis focusing on the Spanish subjunctive. UCL Working Papers in Linguistics, 14. Jary, M. 2005: Assertion and mood: A cognitive account. Unpublished thesis. University College London: Jary, M. 2007: Are explicit performatives assertions? Linguistics and Philosophy, 30, 207–34. Jary, M. 2008: The relevance of complement choice: A corpus study of ‘believe’. Lingua, 118, 1–18. Jary, M. 2009: Relevance, assertion and possible worlds: A cognitive approach to the Spanish subjunctive. In De Brabanter, P. and Kissine, M. (eds). Utterance interpretation and cognitive models. Bingley: Emerald Group Publishing. Jary, M. 2010: Assertion and false-belief attribution. Pragmatics and cognition, 18:1, 17–39. Kadmon, N. 2000: Formal pragmatics. Oxford: Blackwell. Karttunen, L. 1973: Presupposition of compound sentences. Linguistic Inquiry, 4, 169–93. Karttunen, L. 1977/2002: Syntax and semantics of questions. In Portner, P. and Partee, B. H. (eds). Formal semantics: The essential readings. Oxford: Blackwell. Kratzer, A. 2001: The notional category of modality. In Portner, P. and Partee, B. H. (eds). Formal semantics: The essential readings. Oxford: Blackwell. Levinson, S. C. 1983: Pragmatics. Cambridge: CUP. Lewis, D. 1973: Causation. Journal of Philosophy, 70, 556–67. Lewis, D. 1979: Score-keeping in a language game. Journal of Philosophical Logic, 8. Lohmann, H. and Tomasello, M. 2003: The role of language in the development of false belief understanding: A training study. Child Development, 74, 1130–44. McDowell, J. 1980/1998: Meaning, communication and knowledge. Meaning, knowledge and reality. Cambridge, MA: Harvard University Press. McGinn, C. 1977: Semantics for non-indicative sentences. Philosophical Studies, 12, 301–11. Meinunger, A. 2006: On the discourse impact of subordinate clauses. In Molnár, V. and Winkler, S. (eds). The Architecture of Focus. Berlin: Mouton de Gruyter.
References 215 Millikan, R. 1984: Language, thought and other biological categories. Cambridge, MA: MIT Press. Millikan, R. 1993: In defense of proper function. White queen psychology and other essays for Alice. Cambridge, MA: MIT. Millikan, R. 2004: The varieties of meaning: The 2002 Jean Nicod lectures. Cambridge, MA: MIT Press. Millikan, R. 2005: Language: A biological model. Oxford: OUP. Onishi, K. H. and Baillargeon, R. 2005: Do 15-month-old infants understand false-beliefs? Science, 308, 255–258. Origgi, G. and Sperber, D. 2000: Evolution, communication, and the proper function of language. In Carruthers, P. and Chamberlain, A. (eds). Evolution and the human mind: Language, modularity and social cognition. Cambridge: CUP. Pagin, P. 2004: Is assertion social? Journal of Pragmatics, 36, 833–59. Pagin, P. 2008: Assertion. In Zalta, E. N. (ed.). The Stanford Encyclopedia of Philosophy. (Fall 2008 Edition). URL=. Pendelbury, M. 1986: Against the power of force: Reflections on the meaning of mood. Mind, 95, 361–72. Pinker, S. 1998: How the mind works. London: Penguin. Portner, P. 2005: The semantics of imperatives within a theory of clause types. In Watanabe, K. and Young, R. B. (eds). Proceedings of Semantics and Linguistic Theory XIV. Ithaca, NY: CLC Publications. Portner, P. 2007: Imperatives and modals. Natural Language Semantics, 15, 351–383. Potts, C. 2005: The logic of conventional implicatures. OUP. Quer, J. 1998: Mood at the interface. The Hague: Holland Academic Graphics. Quer, J. 2001: Interpreting mood. Probus, 13, 81–111. Recanati, F. 1987: Meaning and force: The pragmatics of performative utterances. Cambridge: CUP. Recanati, F. 2000: Oratio obliqua, oratio recta: An essay on metarepresentation. Cambridge, MA: MIT Press. Recanati, F. 2002: Does linguistic communication rest on inference? Mind and Language, 17, 102–26. Recanati, F. 2004: Literal meaning. Cambridge: CUP. Rouchota, V. 1994: The subjunctive in modern Greek: Dividing the labour between semantics and pragmatics. Journal of Modern Greek Studies, 12, 185–201. Ruffman, T., Slade, L., Rowlandson, K., Rumsey, C. and Garnham, A. 2003: How language relates to belief, desire, and emotion understanding. Cognitive Development, 18, 139–58. Russell, B. 1905: On denoting. Mind, 14, 479–93. Schiffer, S. 1972: Meaning. Oxford: Clarendon. Schiffer, S. 2006: Propositional content. In Lepore, E. and Smith, B. C. (eds). The Oxford handbook of philosophy of language. Oxford: OUP. Schlenker, P. 2005: The lazy Frenchman’s approach to the subjunctive (Speculations on reference to worlds and semantic defaults in the analysis of mood). In Geerts, T., van Ginneken, I. and Jacobs, H. (eds). Romance languages and linguistic theory 2003: Selected papers from ‘Going Romance’ 2003. Amsterdam: John Benjamins.
216
References
Searle, J. R. 1969: Speech acts: an essay in the philosophy of language. Cambridge: CUP. Searle, J. R. 1975/1979: Indirect speech acts. Expression and meaning: Studies in the theory of speech acts. Cambride: CUP. Searle, J. R. 1975/1979: A taxonomy of illocutionary acts. Expression and meaning: Studies in the theory of speech acts. Cambridge: CUP. Searle, J. R. 1989: How performatives work. Linguistics and Philosophy, 12, 535–58. Segal, G. 1991: In the mood for a semantic theory. Proceedings of the Aristotelian Society, 91, 103–118. Sellars, W. 1956/1997: Empiricism and the philosophy of mind Cambridge, MA: Harvard University Press. Seyfarth, R. M., Cheney, D. L. and Marler, P. 1980: Vervet monkey alarm calls: Semantic communication in a free-ranging primate. Animal Behaviour, 28, 1070–94. Siegel, M. E. A. 2006: Biscuit conditionals: quantification over potential literal acts. Linguistics and Philosophy, 29, 167–203. Simons, M. 2003: Presupposition and accommodation: Understanding the Stalnakerian picture. Philosophical Studies, 112, 251–78. Simons, M. 2007: Observations on embedding verbs, evidentiality, and presupposition. Lingua, 117, 1034–56. Slaughter, V., Peterson, C. C. and Mackintosh, E. 2007: Mind what mother says: Narrative input and theory of mind in typical children and those on the autism spectrum. Child Development, 78, 839–58. Smith, N. J. 2000: Frege’s judgement stroke. Australasian Journal of Philosophy, 78, 153–75. Sobel, D. and Copp, D. 2001: Against direction of fit accounts of belief and desire. Analysis, 61, 44–53. Southgate, V., Senju, A. and Csibra, G. 2007: Action anticipation through attribution of false belief by 2-year-olds. Psychological Science, 18, 587–92. Sperber, D. 1985: On anthropological knowledge. Cambridge: CUP. Sperber, D. 1994: The modularity of thought and the epidemiology of representations. In Hirschfeld, L. A. and Gelman, S. A. (eds). Mapping the mind: Domain specificity in cognition and culture. Cambridge: CUP. Sperber, D. 1994: Understanding verbal understanding. In Khalfa, J. (ed.). What is intelligence? Cambridge: CUP. Sperber, D. 1996: Explaining culture. Oxford: Blackwell. Sperber, D. 1997: Intuitive and reflective beliefs. Mind and Language, 12, 67–83. Sperber, D. 2000: Metarepresentations in an evolutionary perspective. In Sperber, D. (ed.). Metarepresentations: A multidisciplinary perspective. Oxford: OUP. Sperber, D. and Wilson, D. 1986/1995: Relevance: Communication and cognition. Oxford: Blackwell. Sperber, D. and Wilson, D. 2002: Pragmatics, modularity and mind-reading. Mind and Language, 17, 3–23. Stainton, R. 1993: Non-sentential assertions. Unpublished thesis. MIT: Massachusetts. Stainton, R. 1997: What assertion is not. Philosophical Studies, 85, 57–73. Stainton, R. 2005: In defense of non-sentential assertion. In Szabo, Z. (ed.). Semantics vs. pragmatics. Oxford: OUP.
References 217 Stainton, R. 2006: Words and thoughts: Subsentences, ellipsis, and the philosophy of language Oxford: OUP. Stainton, R. J. 1999: Interrogatives and sets of answers. Crítica 31, 75–90. Stalnaker, R. C. 1973: Presuppositions. Journal of Philosophical Logic, 2, 447–57. Stalnaker, R. C. 1974/1999: Pragmatic presuppositions. Content and context. Oxford: OUP. Stalnaker, R. C. 1975/1999: Indicative conditionals. Content and context. Oxford: OUP. Stalnaker, R. C. 1978/1999: Assertion. Context and content. Oxford: OUP. Stalnaker, R. C. 1984: Inquiry. Cambridge, MA: MIT Press. Stalnaker, R. C. 1988/1999: Belief attribution and context. Context and content. Oxford: OUP. Stalnaker, R. C. 1988/1999: On the representation of context. Content and context. Oxford: OUP. Stalnaker, R. C. 1991/1999: The problem of logical omniscience, I. Context and content. Oxford: OUP. Stalnaker, R. C. 2002: Common ground. Linguistics and Philosophy, 25, 701–21. Stenius, E. 1967: Mood and language game. Synthese, 17, 254–74. Strawson, P. 1964: Intention and convention in speech acts. Philosophical Review, 73, 439–60. Strawson, P. 1971: Meaning and truth. Logico-linguistic papers. London: Methuen. Surian, L., Caldi, S. and Sperber, D. 2007: Attribution of beliefs by 13-month-old infants. Psychological Science, 18. Sweetser, E. E. 1990: From etymology to pragmatics: Metaphorical and cultural aspects of semantic structure. Cambridge: CUP. Terrell, T. D. and Hooper, J. 1974: A semantically based analysis of mood in Spanish. Hispania, 57, 484–94. Unger, C. 2006: Genre, relevance and global coherence: The pragmatics of discourse type. Basingstoke: Palgrave. Urmson, J. 1952: Parenthetical verbs. Mind, 61, 480–96. Vicente, A. 2010: In Recanati, F., Stojanovic, I. and Villanueva, N. Contextdependence, perspective and relativity. New York: Mouton de Gruyter. Villalta, E. 2000: Spanish subjunctive clauses require ordered alternatives. In Jackson, B. and Mathews, T. (eds). Proceedings of SALT X. Ithaca, NY: Cornell University. Watson, G. 2004: Asserting and promising. Philosophical Studies, 117, 57–77. Wierzbicka, A. 2006: English: Meaning and culture. Oxford: OUP. Wiklund, A.-L., Bentzen, K., Hrafnbjargarson, G. H. and Hróarsdóttir, o. 2009: On the distribution and illocution of V2 in Scandinavian that-clauses. Lingua, 119, 1914–38. Williamson, T. 1995: Is knowing a state of mind? Mind, 104, 533–65. Williamson, T. 1996: Knowing and asserting. The Philosophical Review, 105, 489–523. Wilson, D. 2000: Metarepresentation in linguistic communication. In Sperber, D. (ed.). Metarepresentations: A multidisciplinary perspective. Oxford: OUP. Wilson, D. 2003: New directions for research on pragmatics and modularity. UCL Working Papers in Linguistics, 15, 105–27.
218
References
Wilson, D. and Sperber, D. 1979: Ordered entailments: An alternative to presupposition theories. In Oh, C.-K. and Dinneen, D. (eds). Syntax and sematics: presupposition. London: Academic Press. Wilson, D. and Sperber, D. 1988a: Mood and the analysis of non-declarative sentences. In Dancy, J., Moravcsik, J. M. E. and Taylor, C. C. W. (eds). Human agency: Language, duty and value. Stanford: Stanford University Press. Wilson, D. and Sperber, D. 1988b: Representation and relevance. In Kempson, R. (ed.). Mental representation: The interface between language and reality. Cambridge: CUP. Wittgenstein, L. 1958: Philosophical investigations. Oxford: Blackwell.
Index
Abbott, B. 96, 97, 129, 137, 205 Aliaga, F. 210 Alston, W. P. 7, 11, 13–16, 121, 199, 206 Anscombe, G. E. M. 53 assertion (and assertoric) and convention 56, 60–82 and inference 1, 6, 19–21 and justification 19–20, 72–4, 200 as belief attribution 38–40, 126 as belief expression 2–3, 5, 9, 12, 15, 32–8, 80, 128, 196 as information/knowledge transfer/‘perception-by-proxy’ 40–6, 46–50 assertoric character 6, 55–6, 59, 72, 82–3, 153, assertoric commitment 55, 70, 78, 167–8, 174, 198, 203, 208, 209 assertoric inertia 208 conceptual vs. cognitive analyses of 36–8 consequences of 19–20, 31, 61, 69–70, 77, 194, 200 ‘double-barrelled’ 16 knowledge account of 46–50 explicitness of 6, 13–16, 205 proto-, see Barker quasi-, see Dummett sign 52–83, 105, 128, 142, 172, 194, 198, 203 simulated 69 stage 64–7 sub-sentential 77–8, 79, 204 see also main point assertive class of illocutionary acts 11–12, 15, 18, 87, 203 assumption (in Relevance Theory sense) 44–5, 100–2, 202 consistency-aiming 153–4 embedded 153
factual 56, 153, 194, 198 truth-aiming 153 Astington, J. W. 37 Austin, J. L. 7, 11, 74, 199, 206 constative vs. performative 7–8 Bach, K. 32 Bach, K. and Harnish, R. M. 2, 11–12, 39, 50, 74, 161, 199 informative function of assertion 9–11, 32 mood 124–6 background/foreground distinction 140–2, 147–8, 156–7, 175–7 Baillargeon, R. 201 Bar-on, D. 200 Barker, S. 2, 7, 25, 16, 23–30, 87, 88, 110, 128, 193, 196, 199, 200, 201, 203, 206 contrasted with Brandom 28 communicative vs. representative intentions 25 expressive nature of truth judgements 18, 25 expressive vs. reportive assertion 24–6 mood 121–4 proto-assertion 27, 121–4, 202, 206 role of assertion in the analysis of other illocutionary acts 26–7 see also direction of fit Baron-Cohen, S. 37 Beaver, D. I. 204 belief attributions/reports 92, 103, 113–15, 123, 126, 154, 168, 193, 209 relationship to assertion, see assertion Bloom, P. 191
219
220
Index
Boisvert, D. 106, 108–13 Bosque, I. 117, 168 Brandom, R. 1, 2, 9, 16, 19–23, 31, 110, 121, 161, 193, 196, 200, 201, 203 assertion and knowledge 50 assertoric commitments and entitlements 21–2 contrasted with Barker 28 criticism of Stalnaker 88–91 expressive nature of truth judgements 18, 22 see also assertion: and inference Breheny, R. 40 Carruthers, P. 191 Carston, R. 32, 142, 156, 191, 207, 208 cautious optimism 178, 184–5, 186–7, 209 Chierchia, G. 68, 98, 134 Clark, B. 206 Clément, F. 201 Clements, W. A. 201 Coady, C. A. J. 201 cognitive effects 152, 156, 163, 164, 171, 187 core meaning 6, 14, 30 conjecture 67–70, 153 consistency 68–9, 85–8, 90–1, 100, 115, 153–4, 163, 170, 192–4 context activation 143 basic 4, 85–6, 193 embedded 4, 144–55, 157, 163, 165–72, 174, 194, 208, 209 factual 154–5, 156, 157, 159–60, 163, 166–8, 171–2, 173–4, 177, 192–193, 194, 209 Relevance Theory notion of 138, 140 see also relevance: in a context and Stalnaker: context set context-change potential (CCP) 115–17, 210 contextual effects/implications 148, 151–2, 156, 169, 173, 175, 177, 208
Copp, D. 13 commissives 72–3, 126 conventional implicature 27–8, 129, 134–8, 207 Cosmides, L. 179 Davidson, D. 21, 34–5, 64, 78, 86, 89 autonomy of linguistic meaning 61–2, 203 ‘Davidsonian’ accounts of mood 108–13 paratactic account of mood 106–8 see also Dummett: inadequacy of a disquotational theory of meaning de Bustos, E. 210 declarations 59, 75–6, 194 declarative sentences/mood 1, 3, 4, 5, 7, 19, 21, 23–3, 27, 30–1, 37–8, 41, 52–7, 60–83, 87, 95, 97, 105–28, 129, 132–6, 141–2, 148–9, 200, 203, 204, 205, 206, 207, 208, 209, 210 author’s account of 150, 152, 155–77, 192–5, 196–7 see also assertion: sign de Villiers, J. G. 37 de Villiers, P. A. 37 Dennett, D. C. 90 direction of fit 11–13, 199 Barker on 29 Searle on 12 directives 2, 4, 16, 35, 70–2, 73, 76, 82, 87, 108–12, 116–20, 123, 125–7, 156, 161–2, 172, 194, 203, 204, 205 discourse embedding 67–70, 77 Dummett, M. 2, 3, 21, 31, 32, 39, 40, 52, 124, 128, 196, 199, 200, 201, 203, 204 conventional nature of assertion 60–82 consequences of assertion 19 declarative as a conventional sign of assertion 60–82 Assertion Thesis 60, 77–81 Declarative Thesis 60, 61–77 Not-Dummett 79
Index 221 Grice’s account of meaning 46 inadequacy of a disquotational theory of meaning 34–35, 86–7, 88 priority of assertion over belief 33–8 tactical role of mood and force 78–9 quasi-assertion 19 see also assertion: stage and Davidson: paratactic account of mood effort, processing 138, 151–2, 155–6, 158–9, 163, 167, 171, 177, 181–7, 192, 209 entailment(s) 21, 28, 93–4, 134, 175, 200, 207, 208 ‘at issue’ 136–7, 143 grammatically (un)specified 138, 140, 142–7 ordered 138–48 factive predicates 16, 42, 94, 96, 117, 135–7, 145–6, 167–9, 175, 209 Farkas, D. 113, 114, 115 felicity conditions 8–9, 14, 122–3, 199 fiction 4, 62, 67–70, 77, 91, 153–5, 163, 192, 194, 203 force, see sense/force distinction foreground, see background/foreground distinction Fodor, J. A. 30, 101, 191 Frege, G. 1, 16, 24, 32, 197, 199, 202, 203, 204 assertion sign 53–6, 60, 61, 64–6 conception of inference 54–5 fulfilment/satisfaction conditions 13, 17–18, 31, 109–13, 199 García-Carpintero, M. 47, 49, 202 Gauker, C. 208 Geach, P. T. 15–16, 53, 55, 153, 203 Ginet, C. 74 Giorgi, A. 117 Green, M. S. 54, 62–3, 68, 200, 201, 203
Grice, P. (and Gricean) 2, 9, 37, 32, 38, 65, 80, 94, 124, 128, 129, 160, 191, 196, 199, 207 assertion as the communication of attitudes 39–41, 44–6, 188–90 Hale, C. M. 37 Hare, R. 62, 108 neustic, tropic and phrastic 64–6 Harnish, R. M. (see also Bach, K. and Harnish, R. M.) 108, 199, 206 Harris, P. L. 201 Heim, I. 115, 117, 168 Hindriks, F. 50, 202 Hintikka, J. 114 Hooper, J. 207 Horn, L. R. 208 Hornsby, J. 108, 199 Humberstone, I. L. 199 Huntley, M. 113–14, 192, 204, 205 Ifantidou, E. 132 indicative mood see declarative mood inference ‘inferential integration test’ 58–9 vehicle of 57–9, 202 see also assertion: and inference and Frege: conception of inference infinitive complements 127, 169–71, 205 Jary, M. 38, 59, 74, 117, 126, 168, 170–1, 201, 203, 206, 207 Jenkins, J. M. 37 Kadmon, N. 134, 135, 136, 137, 144, 204 Karttunen, L. 136, 206 knowledge attribution of 178–86 contrasted with attribution of belief 43–4 see also assertion: and knowledge Kratzer, A. 119 Levinson, S. C. 134, 135, 140, 204 Lewis, D. 95, 117, 200
222
Index
Lohmann, H. 37 Ludwig, K. 106, 108–13 main point 4, 5, 97–8, 99, 129–48, 173–7 non-asserted main point 130–4 manifestness 44–5, 201–2, 210 mutual 99–103, 157, 163, 167, 171, 189, 208 McConnell-Ginet, S. 68, 98, 134 McDowell, J. 2, 39, 42, 42–3, 201, 202 McGinn, C. 106, 109, 199 Meinunger, A. 207 metarepresentation 160, 178–92, 195, 208 Millikan, R. 2, 39, 40, 45, 57, 199 proper function 10–11 mood attitude-indicating accounts 124–7 ‘Davidsonian’ accounts of 108–13 embedded 109, 121–3, 164–71 force-indicating accounts 121–4 imperative 8, 15, 26–7, 35, 65, 71–2, 87, 106–8, 110–12, 113, 118–20, 125–6, 156, 162, 172, 204, 205, 206 interrogative 5, 8, 27, 107, 108, 110–11, 121, 123, 156, 158, 172, 205, 206 paratactic account 106–8 possible-world accounts 113–20 subjunctive 113, 115, 117, 168, 206, 207, 208–9 see also declarative sentences/mood naïve optimism 182–4, 186–7, 190 negation 24 as a test for presupposition 93, 98, 134–7, 143–4, 147 neustic, see Hare: neustic, tropic and phrastic Onishi, K. H. 201 Origgi, G. 163, 185 Pagin, P. 58, 87, 199, 200, 203 parentheticals 131–5, 164–6, 174, 203, 207
Pendlebury, M. 109, 111–12, 205, 206 performatives 7–8 explicit 58–9, 73–6, 82, 91, 124, 126, 161–2, 194, 203, 209 perception-by-proxy, see assertion: as information/knowledge transfer/‘perception-by-proxy’ Perner, J. 201 phrastic, see Hare: neustic, tropic and phrastic Pianesi, F. 117 Pinker, S. 191 pointhood 130, 137–8 Portner, P. 118–20, 204, 206 possible worlds 30, 85–8, 100 representing particular possible worlds 88–91, 114–15 Potts, C. 129, 133, 135–7, 143, 147, 148, 207 presupposition 49, 92–9, 103, 116, 117, 134–8, 145, 147–9, 204, 205, 207 accommodation 95–9, 137, 145 and taken-for-grantedness 134–6 pragmatics module 188–91 promising 8, 11, 13, 15, 17, 21, 42, 72–3, 74, 95, 126–7, 161–2, 204 proposition 1, 7, 14, 16–18, 21–2, 26–7, 30–1, 32, 36, 57–9, 69, 85, 89–91 complex 168–9 embedded 114, 131, 153, 166–71, 180–3 Pyers, J. E. 37 Quer, J.
117, 168
Recanati, F. 12, 32, 59, 64–6, 69, 70, 72, 74, 199, 209 relevance 4–5, 51, 123, 131, 138–47 in a context 151–8, 163, 168, 172, 193–4, 208 in own right 4, 139–48, 150, 155–77, 193–4, 207, 208, 209 optimal vs. maximal to an individual 4–5, 151–65, 172, 194
Index 223 Relevance Theory 4, 55–6, 99–103, 124–7, 138–48, 151–8, 181, 187–92, 207, 209 Rouchota, V. 206 Ruffman, T. 37 Russell, B. 15, 16, 53, 64 safeguards, cognitive and social 4, 159–66, 177, 194–5, 197, 208 satisfaction conditions, see fulfilment/satisfaction conditions Schiffer, S. 199, 200 Schlenker, P. 115, 117 Searle, J. R. 2, 7, 11, 14, 17, 19, 59, 71, 74, 199 characterisation of assertion 8–9, 11 taxonomy of illocutionary acts 11–12 see also direction of fit Segal, G. 106, 108, 109, 111, 199 Sellars, W. 201 sense/force distinction 17 Barker on 24, 28–30 Brandom on 21 in Stalnaker’s framework 85–8, 89 Seyfarth, R. M. 33 Siegel, M. E. A. 205 Simmons, K. 200 Simons, M. 96, 130, 137, 146, 167, 174, 204 Slaughter, V. 38 Smith, N. J. 54 Sobel, D. 13 sophisticated understanding 185–6 Southgate, V. 201 speech act fundamentalists 7, 16–30, 31, 110, see also mood: force indicating accounts taxonomy of 11–12 theory 2, 6–31 traditionalists 2, 6–16, 24, 27, 30, 31, 85, 87, 110, see also mood: attitude indicating accounts Sperber, D. 90, 163, 185, 202, 208 metarepresentation 179–81 utterance interpretation strategies 178–86, 209
Sperber, D. and Wilson, D. (also Wilson, D. & Sperber, D.) 2, 4, 5, 11, 32, 45, 85, 151–3, 158, 161, 175, 207–8, 208 factual assumptions 55–6, 104, 198 manifestness 44, 99, 201–2 mood 124–8, 206 ordered entailments 138–48 pragmatics module 188–91 Relevance Theory compared with Stalnaker’s common ground famework 99–103 Stainton, R. 57, 206, 208 see also assertions: sub-sentential Stalnaker, R. C. 3, 4, 30, 105, 115–16, 118, 119–20, 135, 192–3, 198, 204, 205 acceptance 69 context set 85–8, 91–3, 95–6, 99–103, 115–17, 205 distinguishing assertion in his framework 91–2 essential effect of assertion 95–6, 99, 205 general framework 84–5 his framework compared with Relevance Theory 99–103 non-informative functions of assertion 102–3 presupposition (and accommodation of) 92–9, 136 representation 88–91 sense and force 87–8 truth 86–7 views on linguistic conception of mental representation 101–2 Stenius, E. 199 Strawson, P. 38, 199 supplement 133, 135, 147–8, 166, 176–7, 207 supposition 4, 67–70, 91–2, 122, 163, 194, 203 Surian, L. 201 Sweetser, E. E. 205 Tager-Flusberg, H. 37 Terrell, T. D. 207 Thompson, S. A. 207
224
Index
Tomasello, M. 37 Tooby, J. 179 tropic, see Hare: neustic, tropic and phrastic truth compared with fulfilment/satisfaction 17–18, 111, 112–13 expressive conception of 18–19, 31 Unger, C. 209 Urmson, J. 207
Vicente, A. 202 Villalta, E. 206 Watson, G. 13, 204 Wierzbicka, A. 209 Wiklund, A.-L. 207 Williamson, T. 68, 161, 202, 203 knowledge account of assertion 46–50, 51 Wilson, D. (see also Sperber, D. and Wilson, D.) 178, 189 Wittgenstein, L. 7, 53, 55