Moore's Paradox: New Essays on Belief, Rationality, and the First Person

MOORE’S PARADOX

This page intentionally left blank

Moore’s ParadoxNew Essays on Belief, Rationality,

and the First Person

MITCHELL GREEN

and

JOHN N. WILLIAMS

CLARENDON PRESS · OXFORD

1Great Clarendon Street, Oxford

Oxford University Press is a department of the University of Oxford.It furthers the University’s objective of excellence in research, scholarship,

and education by publishing worldwide in

Oxford New York

Auckland Cape Town Dar es Salaam Hong Kong KarachiKuala Lumpur Madrid Melbourne Mexico City Nairobi

New Delhi Shanghai Taipei Toronto

With offices in

Argentina Austria Brazil Chile Czech Republic France GreeceGuatemala Hungary Italy Japan Poland Portugal SingaporeSouth Korea Switzerland Thailand Turkey Ukraine Vietnam

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

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

the several contributors 2007

The moral rights of the authors have been assertedDatabase right Oxford University Press (maker)

First published 2007

All rights reserved. No part of this publication may be reproduced,stored in a retrieval system, or transmitted, in any form or by any means,

without the prior permission in writing of Oxford University Press,or as expressly permitted by law, or under terms agreed with the appropriate

reprographics rights organization. Enquiries concerning reproductionoutside the scope of the above should be sent to the Rights Department,

Oxford University Press, at the address above

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

British Library Cataloguing in Publication Data

Data available

Library of Congress Cataloging in Publication Data

Data available

Typeset by Laserwords Private Limited, Chennai, IndiaPrinted in Great Britain

on acid-free paper byBiddles Ltd, King’s Lynn, Norfolk

ISBN 978–0–19–928279–1

1 3 5 7 9 10 8 6 4 2

John N. Williams would like to dedicatethis book to Vera, Ben, and Josh.

Mitchell Green would like to dedicatethis book to Lori, Noah, and Sofia.

Acknowledgements

Mitchell Green would like to thank Roberto Mariano, Dean of the Schoolof Economics and Social Sciences at the Singapore Management Universityfor inviting him to visit that institution for the purpose of collaborating withJohn N. Williams on this volume. Green’s research was also supported by aSesquicentennial Fellowship from the University of Virginia.

John N. Williams would like to thank Roberto Mariano, Dean of the Schoolof Economics and Social Sciences at the Singapore Management University forfacilitating Professor Green’s visit, and the Wharton-SMU Research Centre forfunding projects closely related to this book.

The Editors join in thanking Peter Momtchiloff of Oxford University Pressfor his judicious oversight of this project.

Contents

Contributors xi

I . INTRODUCTION AND HISTORICAL CONTEXT

1. Introduction 3Mitchell Green and John N. Williams

2. The All-Seeing Eye: A Blind Spot in the History of Ideas 37Roy Sorensen

I I . MOORE’S PARADOX AND KNOWLEDGE

3. Moorean Absurdity: An Epistemological Analysis 53Claudio de Almeida

4. The Normative Character of Belief 76Thomas Baldwin

5. Moore’s Paradox, Evans’s Principle, and Iterated Beliefs 90John N. Williams

I I I . MOORE’S PARADOX, BELIEF, AND ASSERTION

6. What Reflexive Pronouns Tell Us about Belief: A New Moore’sParadox De Se, Rationality, and Privileged Access 117Jay David Atlas

7. Moore’s Paradox and the Transparency of Belief 146Jonathan E. Adler and Bradley Armour-Garb

IV. MOORE’S PARADOX AND CONSCIOUSNESS

8. Consciousness, Reasons, and Moore’s Paradox 165Andre Gallois

9. Moorean Absurdity and Showing What’s Within 189Mitchell Green

x Contents

V. ARGUMENTS FROM MOORE’S PARADOX

10. My Philosophical Position Says �p� and I Don’t Believe �p� 217Alan Hajek

11. Moorean Pretense 232Robert M. Gordon

Index of Names 243Subject Index 245

Contributors

Jonathan E. Adler (CUNY, Graduate Center)

Bradley Armour-Garb (University at Albany, SUNY)

Jay David Atlas (Pomona College)

Thomas Baldwin (University of York)

Claudio de Almeida (PUCRS, Brazil)

Andre Gallois (Syracuse University)

Robert M. Gordon (University of Missouri, St. Louis)

Mitchell Green (University of Virginia)

Alan Hajek (The Australian National University)

Roy Sorensen (Dartmouth College)

John N. Williams (Singapore Management University)

PART I

INTRODUCTIONAND HISTORICAL CONTEXT

1

Introduction

Mitchell Green and John N. Williams

A. INTRODUCING MOORE’S PARADOX

G. E. Moore observed that to say,‘I went to the pictures last Tuesday but I don’tbelieve that I did’ would be ‘absurd’ (1942: 543). Over half a century later, suchsayings continue to perplex philosophers and other students of language, logic,and cognition. On the one hand, such sayings seem distinct from semanticallyodd Liar-type sayings such as ‘What I’m now saying is not true’. Unlike Liar-typesentences, what Moore said might be true: One can readily imagine a situation inwhich Moore went to the pictures last Tuesday but does not believe that he didso. On the other hand, it does seem absurd to assert a proposition while, with noapparent change of mind, or aside to a different audience, going on to deny thatone believes it. It seems no less absurd to judge true the following proposition: pand I do not believe that p.

Ludwig Wittgenstein was fascinated by Moore’s example, and the absurdityof Moore’s saying was intensively discussed in the mid-twentieth century. Yetthe source of the absurdity has remained elusive, and its recalcitrance hasled researchers in recent decades to address it with greater care. Questions ofthe relation of Moore’s paradox to consciousness, self-knowledge, justification,self-expression, conversation, decision theory, belief, and other topics haveaccordingly come under increasing scrutiny. In addition, recent research hasseen a number of ‘arguments from Moore’s Paradox’, aiming to establish alarge philosophical thesis on the basis of this phenomenon. Such argumentshave been directed toward functionalism in the philosophy of mind (Heal1994; Collins 1987; Milgram 1994), self-knowledge (Shoemaker 1988, 1995;Gallois 1996), the existence of ‘blindspots’ or states that are ‘counterprivate’¹(Sorensen 1988; Gombay 1988), evidentialism in epistemology (Adler 1999,

¹ A person’s state is counterprivate just in case she alone, when she is in that state, cannot judgethat she is.

4 Introduction

2002), consciousness (Rosenthal 1995a, 1995b), skepticism (DeRose 1991,1996, 2002), fictionalism (Szabo-Gendler 2001), and simulation vs. theorytheories of knowledge of other minds (Gordon 2000). These developments haveboth shown the centrality of the topic and the depth of treatment needed for itselucidation. As one author has put it,

Moore’s Paradox can be seen as an emblem for peculiarities in the first-person pointof view, specifically how the possibilities for thinking and talking about oneself aresystematically different from the possibilities of thinking and talking about other people.(Moran 1997)

In the remainder of this Introduction we shall briefly discuss (B) Moore’s viewson the subject, (C) Wittgenstein’s remarks on the matter in response to Moore,(D) constraints on any adequate account of Moore’s brand of absurdity, (E) someavailable approaches to explaining this brand of absurdity, and, finally (F) thecontributions to this volume and their relation to currently open questions onthe topic.

B. MOORE ON PARADOX AND ABSURDITY

Moore’s remarks about the cases that interested him were relatively consistentacross the occasions in which he wrote about the topic. In an untitled andincomplete manuscript of a paper that he gave to the Moral Sciences Club inCambridge, Moore observes that the words, ‘Though I don’t believe it’s raining,yet as a matter of fact it really is raining’ by themselves are not nonsensical.²Moore notes that he could use these words in giving a philosophical examplewithout saying anything nonsensical. He also credits Wittgenstein with theobservation that one could sensibly suppose a situation in which those wordsare true of him. He further observes that no absurdity would arise from a thirdperson or past tense utterance of the sentences in question, such as, ‘Moore doesnot think it is raining, yet as a matter of fact it is,’ or ‘I thought it was not raining,but as a matter of fact it was.’

Moore elucidates this last thought with a semantic consideration, remarkingthat the meaning of the sentence ‘I don’t believe that it is raining’ is to be givenin terms of the speaker’s state of mind:

The words, ‘I don’t believe it’s raining’ when said by a particular person have a definitemeaning in English: we can say that what they mean is something about his state ofmind—what they mean can’t be true unless his state of mind is one which can beproperly described by saying he doesn’t believe that; and so with ‘as a matter of fact it israining’. (Moore 1993: 210)

² Moore (1993; 207). Baldwin dates this paper from 1944. This utterance has the same omissiveform as Morre’s example in his 1942, ‘I went to the pictures last Tuesday but I don’t believe that Idid’, namely, ‘p & I don’t believe that p’.

Mitchell Green and John N. Williams 5

Moore’s point is that what is said in an utterance of ‘I believe it’s raining’ isdifferent from what is said in an utterance of ‘It’s raining’, because only the truthof the former depends upon the state of mind of the speaker. Although this pointmay sound like a platitude, we shall see that it has been denied.

Moore is also careful to distinguish absurdity from paradox. What is absurd,he suggests, is to utter the just-quoted sentence assertively. What is paradoxicalis that there should be such an absurdity that cannot be fully explained in termsof a semantic contradiction generated by the words themselves: ‘It is a paradoxthat it should be perfectly absurd to utter assertively words of which the meaningis something which may quite well be true—is not a contradiction’ (Moore1993: 209). We shall follow Moore in likewise distinguishing between Mooreanabsurdity and Moore’s paradox.

Moore doubted that he had a complete explanation of the nature of theabsurdity (Moore 1993: 211) but he made some suggestions for arriving at one.He tells us that I believe that p follows neither from p nor from I assert that p.Nevertheless, Moore contends, in ‘assertively uttering’ an indicative sentence,one implies, in an everyday sense of ‘imply’, that one believes it. Again in his(1944), Moore remarks that, ‘[t]here seems to be nothing mysterious about thissense of ‘‘imply’’ ’ (p. 542). However, if one goes on to assert that one does notbelieve it, what one says contradicts what one implies (Moore 1993: 210). So hisaccount of the Moorean omissive assertion, ‘p & I don’t believe that p’ requiresthe principle that if I assert that p then I imply that I believe that p (‘Omissive’because the assertion reports the specific omission of true belief). Suppose Iassert that p. Then I imply that I believe that p. But suppose I then assert that Idon’t believe that p. What I assert (that I don’t believe that p) contradicts whatI just implied (that I believe that p). So one part of my conjunctive assertioncontradicts what my assertion of another part implies.

Moore also observes that to say, ‘I believe that he has gone out, but he hasnot’ would also be ‘absurd’ (1944: 204). Unlike his first example, this has thecommissive form, ‘p & I believe that not-p’ (‘Commissive’ because the assertionreports the commission of a specific mistake in belief). Here Moore uses a secondprinciple, that if I assert that p then I imply that I don’t believe that not-p. Soby asserting that p I imply that I don’t believe that not-p, which contradicts thecontent of the second conjunct of my assertion. Since this proposal fails to explainthe omissive case, Moore himself may not have recognized that his two examplesare different forms. For on the second principle, if I assert that (p & I don’t believethat p) then I imply-and-then-assert that I neither believe that not-p nor believethat p, which is neither a self-contradiction nor a contradictory set of beliefs.³

To repair this problem Moore could either apply his first principle to thecommissive case as well or, less economically, apply the first principle to the

³ We assume here assertion distribution: If S asserts that (p & q), then S asserts that p, and Sasserts that q.

6 Introduction

omissive case and the second to the commissive case. On the first option, if Iassert that p, then I imply that I believe that p. But suppose I then assert thatI believe that not-p. This does not contradict what I have just implied. InsteadI have implied-and-then-asserted that I have contradictory beliefs about whetheror not p. So the two absurdities are conceptually distinct. On the second option, Ihave implied-and-then-asserted a contradiction in either case (that I do and don’tbelieve that p, in the omissive case and that I do and don’t believe that not-p,in the commissive case), so now the two absurdities come out as conceptuallyidentical. This means that a reconstruction of Moore’s account must now choosebetween economy of explanandum and economy of explanans. One considerationin favor of the first option is that since the omissive form, which reports a specificinstance of my ignorance, is semantically distinct from the commissive, whichreports my specific mistake, we might expect a resulting structural difference inthe contradiction-like phenomena that constitute the resulting absurdity.

Finally, Moore appears to consider only the absurdity of assertoric utterancesand so nowhere considers the absurdity as it occurs in thought, despite the factthat omissive or commissive propositions appear absurd if I do not assert, butmerely judge them true. It is only relatively recently (due apparently to Sorensen1988) that attention in the literature has turned to the absurdity as one injudgment.

C. WITTGENSTEIN ON MOORE’S PARADOX

Wittgenstein attended Moore’s paper discussing this set of problems at theMoral Sciences Club in 1944. His letters to Moore and notebook entries showthat he was intensely interested in these problems.⁴ Indeed, Malcolm reportsWittgenstein as having ‘once remarked that the only work of Moore’s that greatlyimpressed him was his discovery of the peculiar kind of nonsense involved insuch a sentence as ‘‘It’s raining but I don’t believe it’’ ’ (1984: 66). In section X ofthe second part of the Investigations (which, according to Malcolm, was writtenin 1949), Wittgenstein coins the term ‘Moore’s paradox’ (p. 190) and devotesthe rest of that section to it. He also discusses it in his Remarks on the Foundationsof Psychology (vol. i, §§478–90; vol. ii, §§280–90).

According to his letter to Moore, Wittgenstein thought that the absurdity isimportant because it is ‘something similar to a contradiction, thought it isn’tone’. He also thought that although the explanation of the absurdity will say

⁴ Malcolm (1995) reports counting 130 remarks on the verb ‘to believe’ in Remarks on thePhilosophy of Psychology and in the first volume of Last Writings on the Philosophy of Psychology. Thisconcentration, Malcolm suggests, was provoked in large part by Wittgenstein’s interest in Moore’sparadox. Wittgenstein elsewhere writes, ‘Moore stirred up a philosophical wasps’ nest with hisparadox; and the only the wasps did not fly out was that they were too listless’ (1980: 76). His letterto Moore is in his reason (1974: 177).


‘something about the logic of assertion’ it will also show that ‘logic isn’t as simpleas logicians think it is’ (1974: 177). There, Wittgenstein also points out thatthere is no absurdity in supposing that Moore’s saying is true. In the same letter,Wittgenstein comments:

If someone asks me ‘Is there a fire in the next room?’ and he answers, ‘I believe there is,’ Ican’t say: ‘Don’t be irrelevant. I asked you about the fire, not about your state of mind!’

thus displaying an interest in the relation of a self-report of belief that p to anassertion that p. In the Investigations Wittgenstein formulates ‘Moore’s paradox’as the fact that

‘I believe that this is the case’ is used like the assertion ‘this is the case’; and yet thehypothesis that I believe that this is the case is not used like the hypothesis that this is thecase.

Consistently with this first claim, Wittgenstein further holds that ‘the statement‘‘I believe it’s going to rain’’ has a meaning like, that is to say a use like, ‘‘It’sgoing to rain . . .’’ ’ (1997: pt. II, sect. X). He further claims that if there werea verb meaning ‘to believe falsely’, it would not have any significant first personpresent indicative and adds that because ‘I believe it is so’, ‘throws light’ on mystate of mind, so does ‘It is so’.

In the first volume of his Remarks on the Foundations of Psychology, Wittgensteinreturns to a formulation of the paradox in terms of supposition (§478):

Moore’s paradox may be expressed like this: ‘I believe p’ says roughly the same as ‘ p’; but‘Suppose I believe that p . . .’ does not say roughly the same as ‘Suppose p’.

and goes on to state (§490):

The paradox is this: the supposition may be expressed as follows: ‘Suppose this went oninside me and that outside’—but the assertion that this is going on inside me asserts thisis going on outside me. As suppositions the two propositions about the inside and theoutside are quite independent, but not as assertions.

One position discernible in the above remarks is that if I assert ‘I believe that p’then I assert that p. More generally, both my self-report of belief, ‘I believe thatp’ and my plain assertion, ‘p’ have roughly⁵ similar uses and so, for Wittgenstein,roughly similar meanings. So as Wittgenstein adds,

One might also put it like this: ‘I believe p’ means roughly the same as ‘p’. (§472)

Both tell a hearer something about my own attitude to the world as well assomething about the world itself (as I take it to be). On this view, the absurdityof my assertion that (p & I believe that not-p) lies in the fact that my assertion of

⁵ Roughly: See section E.I.3 below for a view according to which, in asserting p I representmyself as knowing that p, while this does not hold for cases in which I assert that I believe that p.(We thank an anonymous referee for drawing our attention to this point.)

8 Introduction

the latter conjunct is also an assertion of not-p. Thus in asserting (p & I believethat not-p), I have contradicted myself on the matter of whether p.⁶ Wittgensteinseems to hint at a difficulty in this approach in Remarks, vol. ii, §420, where heobserves that ‘if A says ‘‘I believe it’s raining’’ and B says ‘‘I don’t believe so’’,they are not contradicting each other.’ Presumably Wittgenstein would make thesame point in a case in which B says, ‘I believe not’.

Moreover, Wittgenstein seems to suggest a different relation of assertion tobelief in the following passage:

I want to say first of all with the assertion ‘it’s going to rain’ one expresses belief in thatjust as one expresses the wish to have wine with the words ‘Wine over here!’ (1980a:§472)

This passage claims that if I assert that p then I express a belief that p. So inasserting that (p & I don’t believe that p), I express-and-assert a belief and thelack of it, namely a self-contradiction, whereas in asserting that (p & I believethat not-p), I express-and-assert contradictory beliefs about whether p. We willpostpone discussion of this account until Section F below.

Finally, Wittgenstein observes (1980b: §290) that ‘under unusual circum-stances [the] sentence [‘‘It’s raining but I don’t believe it’’] could be given aclear sense’. In (1980a: §§485–7), he gives two examples of non-absurd usesof omissive sentences. The first is when I exclaim in amazement, ‘He’s comingbut I still can’t believe it’ (1980a: §485). Such a case is not absurd since thelocution ‘I can’t believe it’, and its cognates have a conventional use as expressionsof surprise rather than literal disavowals of belief. The second example is of arailway announcer who is convinced that the train whose scheduled arrival heis obliged to report won’t arrive. He announces its impending arrival and adds,‘Personally I don’t believe it’ (1980a: §§486–7). Wittgenstein then gives anexample of a non-absurd use of a commissive sentence, that of a soldier whoproduces military communiques but adds that he believes they are incorrect.Wittgenstein’s point is that the absurdity in speech is not guaranteed by a mereutterance of a sentence of one of Moore’s omissive or commissive forms; rather itrequires their assertion.

⁶ Moran denies that Wittgenstein should be thus construed, which construal he calls thePresentational view: ‘that in the first person present-tense the verb-phrase ‘‘I believe’’ is not in factpsychological, but rather represents a mode of presenting the relevant proposition which followsit’ (1997: 144). He goes on to show that the Presentational view is both implausible and does notcomport with Wittgenstein’s texts, and infers from this that it is not Wittgenstein’s position thatone who asserts a Moorean sentence contradicts herself. However, we do not impute the first ofthe two conjuncts of the Presentational view to Wittgenstein, but rather only the second, or moreprecisely only the doctrine that one who asserts ‘I believe that p’ also asserts p (whatever else shemay be doing).


D. CONSTRAINTS ON AN ACCOUNT OF MOOREAN

ABSURDITY

Our question is the source of Moorean absurdity: What makes utterances suchas ‘It’s raining, but I don’t believe it’, as well as thoughts expressible with suchwords, absurd? Moore traces the absurdity of the utterance of the sentence,‘It’s raining but I don’t believe it’ to the speaker’s being thereby committedto a self-contradiction. On this account, the absurdity is explained as a severefailure of theoretical rationality, a form of rationality concerned with tracking thetruth. Norms constitutive of theoretical rationality include prohibitions againstforming beliefs on insufficient evidence and against drawing inferences that areeither deductively invalid or inductively weak. Extreme failures of theoreticalrationality may be absurd. For instance, it seems absurd for a person to commitherself to a self-contradiction that, with a minimum of reflection, she would beable to see cannot be true, whether in virtue of syntax as in, ‘It is raining and notraining’ or in virtue of semantics, as in, ‘Women aren’t females’. It also seemsabsurd for her to commit herself to a pair of propositions that contradict eachother, such as ‘I believe that p’ and ‘I don’t believe that p’ that, with a minimumof reflection, she would be able to see cannot both be true.

Having discerned one source of absurdity in a severe failure of theoreticalrationality, the question arises whether a different form of rationality couldprovide another. Practical rationality is a form of rationality concerned withprudent action. Among norms thought by many to be constitutive of practicalrationality is the prescription to pursue that action most likely to achieve one’sdesires, given how one believes the world to be. It would be a violation of thisprescription to go to a petrol station to buy petrol in the conviction that it is shut.A related prescription is to choose that action most likely to maximize subjectiveexpected utility, namely the weighted average of the desirabilities of each possibleoutcome of that action, the weights being the probabilities of those outcomes.Thus, unless Jeb places value on risk per se, and so long as he places some valueon monetary gain, his choice of what he deems to be a 50 per cent chance ofa $100 gain over what he deems a sure thing of that gain, would violate thisprescription to maximize one’s subjective expected utility. Might it be possibleto explain the absurdity to be found in Moore’s examples in terms of a failureof practical rationality so extreme that the agent, with little reflection, would seethat her enterprise is an inept way, from among her options, of achieving heraims? We shall find below that some accounts of Moorean absurdity are couchedin terms of practical rather than theoretical rationality.

An account of the source of Moorean absurdity must, it seems, elucidate thatsource in terms of practical rationality, theoretical rationality, or some othersystem of norms. Moreover, many accounts of Moorean absurdity appeal to a

10 Introduction

notion of implication by claiming, for example, that in asserting a proposition p,a speaker implies that she believes that p; or that in asserting that she believes thatp, she implies that p. To be assessed adequately such claims require elucidationin terms of one or another established notion of implication. Among suchnotions of implication are logical ( p implies q just in case there is no way forit to be true that p but false that q); evidential (p is adequate evidence for thetruth of q); probabilistic (it is highly likely that p, given that q); or pragmaticimplication, where this notion of pragmatic implication might in turn take anyof the following forms: semantic presupposition (q’s being true is a necessarycondition for p having truth value at all); pragmatic presupposition (q must beaccepted as part of the conversational ‘common ground’ in order for assertion ofp to be conversationally appropriate); conversational implicature (given generalnorms of conversation, we may infer that one who asserts that p speaker-meansthat q); or conventional implicature (given the meaning of ‘p’, one who uses‘p’ in a speech act must also speaker-mean that q). Any account appealing to anotion of implication is obliged to tell us what kind of implication is at issue:Further, any such account must accommodate the following three data.

1. In addition to the better known omissive, ‘p but I don’t believe it’, ananalogous commissive paradox is to be found in ‘p but I believe thatnot-p’.

2. Moorean absurdity arises when a person does not assert an omissive orcommissive Moorean proposition but rather judges that it is true.

Here is one account of the source of the absurdity of Moorean judgment. Supposethat I judge that (p & I don’t believe that p). On the assumption that judgment,construed as an episodic instantiation of belief, distributes over conjunction, Ijudge, and so believe, that p. This belief in turn falsifies the second conjunctof what I judge and so falsifies the whole conjunction. Although an omissiveMoorean proposition can be true and can be judged true, it cannot be true if it isjudged true. Moreover, discerning this fact about your own judgment of (p & Idon’t believe that p) requires, as we just saw, a minimum of reflection. It is thusnot difficult to see why one who judges true an omissive Moorean propositionis guilty of a severe failure of theoretical rationality, and thus why it is that hisbelief is absurd. Although an omissive Moorean belief is self-falsifying as opposedto a belief in a necessary falsehood (such as someone’s belief that it is raining andnot raining) it is nonetheless like such a belief in the respect that both are severefailures of theoretical rationality.

Consider now the commissive case. Suppose that I judge that (p & I believethat not-p). Then since judgment distributes over conjunction, I judge andthus believe that p. But what I judged is true only if I also believe that not-p.Thus what I judged is true only if I have contradictory beliefs about p, one ofwhich must thus be mistaken. Moreover I should be able to see this fact witha minimum of reflection. This is again a severe failure of theoretical rationality,


and again an explanation of the absurdity of my judgment is ready to hand.This account of the absurdity of the two kinds of Moorean judgment seemssimpler than that for Moorean speech, since the former, unlike the latter, neednot rely upon pragmatic notions such as implicature, assertion, or expression.Instead we have seen two relatively clear senses in which one who believes eithera commissive or omissive Moorean sentence must be in error in a way thatwould manifest itself to minimal reflection. As we will see below, some authorshave fastened upon this asymmetry in the hopes of explaining the absurdity ofMoorean speech in terms of that of Moorean thought. If the explanation of theabsurdity of Moorean assertion can be delivered, with little further explanatorycost, in terms of the absurdity of Moorean thought, then one seems to get bothexplanations parsimoniously.

3. If possible, an account of Moorean absurdity should not appeal tocontroversial principles of epistemic or doxastic logic.

Epistemic (doxastic) logic attempts to codify inferential relations among states ofknowledge and belief as opposed to their contents. Candidates for such logicallaws include the thesis that knowledge distributes over conjunction (one whoknows that p & q knows that p and knows that q), and that belief is deductivelyclosed (one who believes that p, and who believes that p logically entails q,believes that q; or even more strongly that if p logically entails q, and one believesthat p, one believes that q as well.) While knowledge- and belief-distributionseem unexceptionable, other principles are controversial. For instance, it doesnot seem true as a psychological principle that a person’s beliefs are deductivelyclosed: lack of attention or simply an absence of interest in the question mayprevent a person from drawing consequences from her beliefs. Nor does it seemtrue as a psychological principle that a person who believes that p also believesthat he believes that p or conversely. For example, my sincere professions ofopen-mindedness about the status of women or of lack of prejudice againstthem may both be mistaken. So I might believe that women are inferior to menwithout believing that I believe this. Equally, I might mistakenly believe that Ibelieve that women are not inferior.

E. SCHOOLS OF THOUGHT CONCERNING MOOREAN

ABSURDITY

Authors developing accounts of Moorean absurdity may be seen as offeringanswers to the question, ‘In what does this form of absurdity consist?’ Attemptsto answer this question generally bifurcate into two kinds. One explains therelevant sort of absurdity in terms of a severe failure of theoretical rationality, theother as a severe failure of practical rationality. In principle, each kind of approachcan be applied either to Moorean judgment or to Moorean speech. Approaches

12 Introduction

to Moorean absurdity could thus, in principle, fall into four categories: MooreanSpeech as Theoretically Irrational (I), Moorean Judgment as TheoreticallyIrrational (II), Moorean Speech as Practically Irrational (III), and MooreanJudgment as Practically Irrational (IV). Further, for a theory focussing onMoorean judgment (or speech) the question immediately arises whether it canbe applied to Moorean speech (or judgment) as well. Theories of type (II) and(IV) have the advantage that they can make use of what we shall call

Shoemaker’s Principle: What can (coherently) be believed constrains whatcan (coherently) be asserted, whereas the converse is not true.⁷

What can coherently be believed constrains what can coherently be assertedbecause assertion is, inter alia, an ostensible manifestation of belief. If what thatassertion purports to manifest is incoherent, then the assertion will be incoherentas well. For instance, not only is it incoherent to believe, ‘I have no beliefs,’it is also incoherent to assert it—not because the content of that assertion isnecessarily false, but because the assertion purports to manifest a belief which,given the content, cannot be true. On the other hand, belief is not an ostensiblemanifestation of a speech act, assertion or otherwise. This is suggested by the factthat while it is incoherent to assert, ‘I am making no assertions ’ it is coherent tobelieve it.

If Shoemaker’s Principle is correct, then a satisfactory account of Mooreanjudgment as theoretically irrational can be extended, via this principle, to anaccount of Moorean speech, at least for those cases in which the speech act isassertion. Similarly for an account of Moorean judgment as practically irrational.It is less clear whether an account of Moorean speech can be carried over toan account of Moorean judgment. We discuss eighteen approaches to Mooreanabsurdity under the four headings delineated above.

I. Moorean Speech as Theoretically Irrational

1. Self-report-of-belief-as-assertionThis Wittgenstein-inspired approach construes assertion of ‘I believe that p’ asan assertion of p. Thus one who asserts ‘p and I believe that not-p’ has, givenassertion-distribution (if one asserts that p & q, then one asserts that p andone asserts that q), asserted both p and not-p, thereby manifesting an extremefailure of theoretical rationality. Malcolm (1995), Linville and Ring (1991), andJacobson (1996) hold that to assert ‘I believe that not-p’ just is to assert thatnot-p; Heal holds that it is in effect to assert that not-p. So my assertion, ‘p and Ibelieve that not-p’ consists of (for Malcolm, Linville and Ring, and Jacobson) orrequires (for Heal 1994: 296) two contradictory assertions. Evidently the reasonwhy, on this approach, a Moorean utterance is absurd is that one making it

⁷ (1996: 76). This principle is anticipated in Wolgast (1977: 118).


commits herself to these contradictory assertions, and so manifests a severe failureof theoretical rationality.

This approach requires treatment of the omissive case. Assertion of ‘I do notbelieve p’ neither constitutes nor requires the assertion of not-p. Goldstein (1993:94–95) holds that if I assert that I don’t believe that p then I deny that p. Thisis incorrect: an agnostic who truthfully reports, ‘I neither believe that God existsnor believe that He doesn’t’ would, on that principle, be making contradictoryassertions about the existence of God. Surely that is not so, and proponents ofthis approach have not addressed the problem raised by the omissive case.

In a similar vein, Collins (1996) construes all self-ascriptions of belief, ‘Ibelieve that p’, as ways of assigning the truth value T to p. This presumably isnot meant as a semantic thesis, since on that view even the supposition that Ibelieve that p would be an assignment of T to p; rather the intent is evidentlythat one who asserts ‘I believe that p’ assigns T to p. At the same time Collinsprovides a semantic analysis of ‘I believe that p’ as ‘If not-p, then I am mistakenabout p’. On this analysis the sentence, ‘I believe that p, but not-p’ logicallyentails ‘I am mistaken about p’. But, Collins contends, ‘S is mistaken about p’means ‘S assigns a truth value to p, and whatever that truth value is, p has theother truth value.’ As a result, assertion of ‘I am mistaken about p’ commits meto either, ‘I assign T to p, but p is F’; or to ‘I assign F to p, but p is T’. Collinsclaims (1996: 310) without explanation that both of these sentences express aself-contradiction. Yet without such an explanation a sentence such as ’I assignT to p, but p is F’ is at least as mysterious as our original Moorean sentences.Collins also does not address the omissive case.

2. Speaker’s implicatureAccording to this approach, the speaker is said to contradict herself in herassertion of a Moorean sentence, but not by virtue of asserting an explicitcontradiction. Rather she conversationally implicates a content that explicitlycontradicts what she asserts (Martinich 1980; Levinson 1983: 105). Imagine aperson who tells an evidently lost tourist, ‘There’s a Tourist Information bootharound the corner.’ Here the speaker seems to suggest that those manning thebooth are likely to help the tourist to find her way. But were she to add, ‘butthey aren’t likely to help you find your way. You would do better to ask that cabdriver over there’, then by explicitly denying the implicatum of her first remarkshe cancels or revokes it. Doing so would not make her sequence of utterancesabsurd. By contrast, if my remark ‘It’s raining’ conversationally implicated that Ibelieve that it’s raining, then my adding, ‘but I don’t believe it’ would cancel thatimplicatum with no resulting absurdity. Yet that prediction is contradicted by thefact that such an utterance is absurd. So the absurdity of Moorean assertion doesnot seem explicable in terms of ‘imply’ when taken as conversational implicature.

By contrast, one who describes someone as ‘poor but honest’ suggests,insinuates, or implies but does not literally say, that there is a tension of some

14 Introduction

kind between poverty and honesty, perhaps that being impoverished tends tomake a person dishonest. This implication is due to the meaning of ‘but’, assuggested by the oddity of someone’s saying, ‘She was poor but honest, whichis not for a moment to suggest that there is any tension between poverty andhonesty.’ Cases of this sort are thus termed conventional implicature. Since anassertion can be made with virtually any words whatsoever, it is not plausiblethat any words or expressions could themselves have as part of their conventionalmeaning that the speaker believes the proposition they are used to assert.Nor could the implicatum in question be due to the indicative grammaticalmood, which could be used by someone uttering a sentence for the purpose ofputting forth a proposition as a supposition or a guess rather than an assertion.Furthermore, a convention is a practice that could have been otherwise, but itis not an optional feature of assertion that it is used for the manifestation orexpression of belief. As Williamson (1996) has pointed out, a speech act notgoverned by the norm that the speaker believe its content to be true, would notbe the speech act of assertion. Moorean absurdity seems explicable neither interms of conversational nor conventional implicature.

3. Speaker’s representationSomeone who asserts that p represents herself as believing, or knowing that p,and the state of affairs thus represented contradicts her assertion that she does notbelieve (or know) that p (Black 1952: Unger 1975: ch. VI; DeRose 1991, 2002;Williamson 1996, 2000: 252–60). Thus for instance, DeRose (1991) endorsesan explanation of the absurdity of asserting ‘It is raining but I don’t know that itis’ given by Unger (1975: 252–65). According to DeRose

If I ‘flat out’ assert that p then I represent it as being the case that I know that p (2002:597–8)

Here a ‘flat out’ assertion is distinguished from assertions like ‘I think that p’or ‘I’m pretty sure that p’ or ‘Maybe p’. So given that assertion distributes overconjunction, if I assert that (p and I do not know that p) then I assert thatp and so I represent it as being the case that I know that p. But I also assertthat I do not know that p. However, those who exemplify Moorean absurditymust be doing more than representing themselves as believing or knowing whatthey assert. Julie the art teacher can represent herself as, say, an astronaut bypainting and displaying an image of herself in a spacesuit. In so doing, she mayalso represent herself as believing or knowing that she is an astronaut (perhapsshe embellishes the drawing with a thought-bubble enclosing the words, ‘I’min space!’). If she does so while remarking, ‘But I am not an astronaut ’, theperformance generates no absurdity recognizably Moorean. Is the problem herethat Julie has used pictorial rather than verbal representation? Surely not: Shemight represent herself as believing that she is an astronaut by supposing for the


sake of argument that she believes that she is one; she will generate no absurdityif she points out, in the process, that she has never left the troposphere.

4. Assertion as expressionOne cannot coherently assert a Moorean sentence ‘p but I don’t believe that p’because asserting the first conjunct would express an intentional state, namelybelief that p, that the second conjunct denies one is in (Rosenthal 1995a: 197,199; 1995b: 317, 319): if I assert that (p and I don’t believe that p) then I assertthat p and thereby express a belief that p. But I also assert that I don’t believethat p, ‘. . . thereby denying that the whole sentence can be used to make anycoherent assertion’.⁸ It is not, however, clear how the second conjunct does infact deny that the whole sentence can be used to make a coherent assertion.Perhaps the failure of coherence results from a contradiction between what isexpressed (the speaker’s belief that p) and what the speaker asserts, that she doesnot believe that p. Yet beliefs, as opposed to their contents, are particulars andthus don’t stand in relations of contradiction, consistency, entailment, etc., toanything else. Short of an elucidation of the notion of expression showing it tobe a mode of commitment but not a species of assertion, this expressivist positionwill not account for the explanandum. Nor does it provide any suggestions as tohow we might approach the commissive case.

II. Moorean Judgment as Theoretically Irrational

5. Moorean judgment is self-falsifying or requires contradictory beliefsThis approach was explained above in Section D.2 (Williams 1996, 1998).⁹

⁸ (1995b: 317). On p. 319 of the same work, Rosenthal writes, ‘I cannot assertively produce thesentence ‘‘It’s raining but I don’t think it is’’ because asserting the first conjunct would express anintentional state that the second conjunct denies that I am in.’ (See also p. 324.) We shall, however,only take Rosenthal to be committed to the view that one cannot coherently assert the sentence inquestion.

⁹ This position is similar to Armstrong (1971), who defends a Lockean account of communica-tion, according to which my act of assertion is my way of signaling to you my objective to get youto think me sincere. Deutscher comes close to the same position, but fails to clearly distinguish theomissive from the commissive paradox. Addressing the commissive case, he says, ‘What is wrongwith ‘‘p but I believe that not-p’’ is this. If the speaker is correct then what he says is false’ (1965:54). By ‘correct’, Deutscher means, ‘correct in believing the conjunctive saying’. But this is trueonly of the omissive case. In a modified analysis, Deutscher says, ‘if the speaker believes all that hesays when he utters ‘‘p but I don’t believe that p’’, then it is logically impossible for him to hold anycorrect beliefs’ (1967: 184, our italics). This can’t be right, however. Although it is impossible thatmy belief that (p and I don’t believe that p) is correct (because it is self-falsifying), this does notmean that I cannot at the same time correctly believe that p. Deutscher could have repaired this flawby substituting, ‘then it is logically impossible for him to hold all correct beliefs’. That would holdtrue for both omissive and commissive Moorean belief, but would then fail to capture the essenceof the absurdity, since my belief that I have at least one false belief fits that diagnosis without beingabsurd at all.

16 Introduction

6. The impossibility of undefeated evidenceAccording to some epistemologists, a proposition p is warranted for S just in casep is in S’s belief system, there is adequate evidence for p, and that evidence is not‘defeated’ by effective counterevidence. Following Klein, de Almeida remarks thatwarrant for a proposition can always be traced back to others of S’s beliefs. Hethus suggests that we can speak of ‘warrant paths extending from the propositionsthat one believes to the propositions that one is entitled to believe given one’spresent stock of beliefs’ (2001: 46). He also propounds a Rule of Revision: Whena proposition p is added to a belief system, any belief that would block thewarrant path to that belief must be removed from the belief system (2001: 47,citing Klein 1986: 266). Thus suppose that I believe that (p and I believe thatnot-p). Then by belief-distribution it follows that I believe p and that I believethat I believe that not-p. This latter belief is, de Almeida contends, a reason torefrain from believing p. Thus if I believe that (p and I believe that not-p), Iviolate the Rule of Revision. Similarly if I believe that (p and I do not believe p),then I believe both p and that I don’t believe that p. The fact that I believe thatp gives me a reason for believing that I believe that p, which is itself reason forrejecting the belief that I believe that not-p. In both the omissive and commissivecases, then, de Almeida contends that Moorean propositions are ones for whichone can have no non-overridden evidence.

7. The impossibility of justifying Moorean beliefWilliams (2004) adopts a principle that seems implicit in Gareth Evans’s Varietiesof Reference (1982: 225–6) that

Whatever justifies me in believing that p also justifies me in believing that Ibelieve that p.

by arguing that

All circumstances that justify me in believing that p are circumstances thattend to make me believe that p.All circumstances that tend to make me believe that p are circumstances inwhich I am justified in believing that I believe that p.

All circumstances that justify me in believing that p are circumstances thatjustify me in believing that I believe that p.

He also assumes the plausible principle that

Whatever justifies me in believing that (p & q) justifies me in believing thatp and justifies me in believing that q.

Now suppose that I enjoy justification for believing that (p & I do not believethat p). Then given the conjunctive principle above, I have justification forbelieving that p. By Evans’s principle, I enjoy the same justification for believing


that I do believe that p. But by the conjunctive principle, I also enjoy preciselythe same justification for believing that I do not believe that p. This is logicallyimpossible, because anything that justifies me in believing that something is thecase renders me unjustified in believing that it is not the case and vice versa.

To explain the absurdity of the commissive belief, Williams needs the differentprinciple that

Whatever justifies me in believing that p also justifies me in believing that Ido not believe that not-p.

Williams claims that this follows from Evans’s principle, together with theassumption that I am minimally rational and reflective: if I am at all rationalthen I will recognize the fact that whatever justification I have for believing thatp renders me unjustified in believing that not-p. By Evans’s principle, whateverjustification I have for believing that p is justification for taking myself to believethat p. But if I take myself to enjoy justification for holding the belief that p andrecognize that this justification renders me unjustified in believing that not-pthen I should take myself as not believing that not-p.

Now suppose that I am justified in believing that (p but I believe that not-p).Then given the conjunctive principle, I again have justification for believingthat p. By the new principle, this means that I enjoy the same justification forbelieving that I do not believe that it is not raining. But by the conjunctiveprinciple, I also enjoy precisely the same justification for believing that I dobelieve that it is not raining. This is logically impossible, as we saw above.

However Vahid (forthcoming) points out that as it stands, Williams’s argumentfor Evans’s principle proves too much. The conclusion of the argument is aninstance of its first premise. Therefore on the basis of that argument we mayconstruct a valid sorites for the highly implausible conclusion that

All circumstances that justify me in believing that p are circumstances thatjustify me in believing that I believe that I believe that . . . I believe that p.

8. Moorean judgment requires contradictory beliefsHeal’s principle: If I believe that I believe that p then I believe that p, explainsthe absurdity of commissive Moorean belief as follows: If I believe that (p but Ibelieve that not-p) then by belief-distribution, I believe that I believe that not-p.So from Heal’s principle, I believe that not-p. But again by virtue of believing theconjunction, I believe that p. But how is the absurdity of omissive Moorean beliefthat (p & I don’t believe that p) to be explained? Heal could try appealing to ananalogous principle: If I believe that I don’t believe that p then I don’t believethat p. The possibility of omissive Moorean belief refutes this new principle, fornow if I believe that (p but I don’t believe that p) then I do and I don’t believethat p. Moreover, the original principle taken together with the new one wouldprohibit a more sophisticated commissive belief. For if I believe that (I believethat q but I believe that I don’t believe that q) then it now follows that I do and I

18 Introduction

don’t believe that q: If I hold this commissive belief then I believe that I believethat q, so by the original principle, I do believe that q. I also believe that I believethat I don’t believe that q, which by applying the original principle and then thenew one, means that I don’t believe that q.¹⁰

9. The irrationality of conscious Moorean beliefBaldwin claims that I cannot be rational if I consciously believe a Mooreanproposition since this requires me in the omissive case to ‘believe that I believeand fail to believe the same thing’, and in the commissive case to ‘believe that Ibelieve and disbelieve the same thing’ (1990: 230). Baldwin presumably reasonsthat since my belief in the first conjunct is conscious then I believe that I believethat p, and since I believe the second conjunct then I believe that I don’t believethat p in the omissive case and I believe that I believe that not-p, in the commissivecase.

The first case is not accurately described as one in which I ‘believe that I believeand fail to believe the same thing’, for that would be a case in which I believethat (I believe that p and I don’t believe that p). Rather it is a case in which Ihave contradictory higher-order beliefs about whether I believe that p. Nor isthe second case accurately described as one in which I ‘believe that I believe anddisbelieve the same thing’, for that would be a case in which I believe that (Ibelieve that p and I believe that not-p). Rather it is a case in which both myhigher-order beliefs are correct only if I have contradictory beliefs about whetherp. Deriving the single beliefs supposedly required needs the principle that beliefcollects over conjunction. This principle is disputable: I don’t seem to believe theconjunction of everything I believe. Further, the explanation does not clearlyapply to the commissive case. Although in the omissive case I am irrational tobelieve a self-contradiction (if the principle is true) or to hold contradictorybeliefs (if it isn’t), my belief in the commissive case that I hold them is lessso, since my consciousness of my own irrationality may be my first step towardremoving it.

In a similar spirit to that of Baldwin, Kriegel (2004: 108–9) adopts theBrentano-inspired principle

1) If I consciously believe that p then I believe that (p & I believe that p)¹¹

to

2) I consciously believe that (p & I don’t believe that p)

¹⁰ Heal recognizes that her principle might be occasionally defeated. But Moorean belief isalways absurd. Therefore the circumstances in which such principles are false are those in which theabsurdity of Moorean belief persists. Hence the principles cannot explain the persistent absurdity.

¹¹ More accurately, Kriegel’s principle is: if I consciously believe that p then I believe that (p &I myself believe that p). While we agree that this de se element is needed, ignoring it effects neitherour exposition of Kriegel’s account nor our criticism of it.


to give

3) I believe that [p & I don’t believe that p & I believe that (p & I don’tbelieve that p)].

He then appeals to a second principle that

4) If I believe that (q & I believe that p) then I believe that (I believe thatq & I believe that p)

and claims that 3) and 4) entail

5) I believe that (p & I don’t believe that p & I believe that p & I believethat I don’t believe that p).

Kriegel observes that the second conjunct of what I believe in 5) contradictsthe third. So if I have a conscious omissive Moorean belief, then I have aself-contradictory belief. But 3) has the form I believe that (q & I believe that q)where q is p & I don’t believe that p. So all that 4) seems to yield is

5′) I believe that [I believe that (p & I don’t believe that p) & I believe that(p & I don’t believe that p)].

Perhaps Kriegel’s strategy is to apply 4) to the last conjunct of what I believe in3), namely

6) I believe that (p & I don’t believe that p)

and then replace 6) with what follows. But 4) cannot be applied to 6) because 6)does not have the form / believe that (q & I believe that that p).

Kriegel’s account of the absurdity of the commissive belief (2004: 188 n. 28)is also flawed. He observe that 1) and

7) I consciously believe that (p & I believe that not-p)

entail

8) I believe that [p & I believe that not-p & I believe that (p & I believethat not-p)].

which by 4) supposedly yields

9) I believe that (p & I believe that not-p & I believe that p & I believethat I believe that not-p).

So if I hold the conscious commisive belie then I believe (among other things)that I hold contradictory beliefs. But what follows from 4) and 8) is

9′) I believe that [I believe that (p & I believe that not-p) & I believe that(p & I believe that not-p)].

Suppose instead that we apply 4) to the last conjunct of what I believe in 8),namely

10) I believe that (p & I believe that not-p)

and replace 10) with what follows, namely

20 Introduction

11) I believe that p & I believe that not-p

to yield

12) I believe that [p & I believe that not-p & I believe that p & I believethat not-p].

This is slightly different from 9), but still gives the result Kreigel requires. But nowhe needs the principle that if the content of what I believe entails so-and-so, thenso-and-so is part of the content of what I believe. Without further qualification,this is false. I may believe that a triangle is equilateral without believing that it isequiangular.

Kriegel can repair his account of the conscious omissive belief by using 1) plusthe fact that belief distributes over conjunction: if the second conjunct of what Ibelieve in 3) is true then I do not believe that p. But if the last conjunct of what Ibelieve in 3) is also true then (since belief distributes over conjunction) I do believethat p. So 3) describes a self-contradictory belief after all. We get a different resultfor the commisive belief. If the second conjunct of what I believe in 8) is truethen I believe that not-p. If the last conjunct of what I believe is also true then(since belief distributes over conjunction) I believe that p. But as kreigel himselfnotes (2004: 118 n. 28) a better result would be that if I hold the commissivebelief consciously, then I believe that I hold a pair of contradictory beliefs. Thisresult follows from 4) alone. If I believe that (p & I believe that not-p) thenby 4) I believe that (I believe that p & I believe that not-p). But if I hold thecommissive belief unconsciously, do I really think I hold contradictory beliefs?

10. Moorean belief entails contradictory beliefs if that belief is true and onebelieves the consequences of one’s beliefsSorensen contends that

(om1) p & I don’t believe that I believe that p

and

(com1) p & I believe that I believe that not-p

seem less absurd to believe than their original counterparts and that as iteration ofthe belief operator increases, only omissive absurdity appears to decrease. Usingthe notation ‘Bna∼p’ where the superscript denotes the number of iterations ofthe belief-operator (so that ‘B3ap’ means that a believes that he believes that hebelieves that p) Sorensen writes (2000: 42):

My solution endorses the intuition that ‘p & Bn ∼p’ is a Moorean sentence for all n, but‘p & ∼Bnp’ need not be a Moorean sentence when n is a large number. ‘p & Bna∼p’does not entail that a has a specifiable directly opposed belief. But ‘Ba(p & Bna ∼p)’entails that a has directly opposed beliefs about p, under the assumption that a believesthe consequences of his beliefs and that ‘p & Bna∼p’ is true. This entailment followsdirectly for n = 1. When n >1, the entailment is secured by a necessary condition forself-attributing higher-order beliefs.


The condition in question is that if I believem+n that p then I believen that p(2000: 39–42) namely a recursive application of the principle of belief elimination:

If I believe that I believe that p then I believe that p.

Sorensen appeals to this principle together with the principle that belief is closedunder logical consequence:

If a logical consequence of p is q and I believe that p, then I believe that q.

It follows that I cannot hold a true belief in com1 unless I hold contradictory,or ‘directly opposed’ beliefs about whether p (compare Sorensen 1988: 40–2).For if I believe that (p & I believe that I believe that not-p), then a logicalconsequence of what I believe is that p, so I believe that p. But if my belief incom1 is true then I believe that I believe that not-p, in which case the principleof belief elimination ensures that I believe that not-p. Since that principle may beapplied recursively, the same diagnosis of the absurdity will hold for any orderof iteration of the belief-operator, as, say, in com4. It also applies to the originalcommissive belief, in which case the principle is not needed. Sorensen’s accountdiagnoses no such absurdity in om1. If I believe that (p & I don’t believe that Ibelieve that p) then a logical consequence of what I believe is that p, so I believethat p. But if my belief in om1 is true then I don’t believe that I believe that p,in which case the principle of belief elimination fails to apply.

Sorensen must explain the absurdity of the original omissive case as follows: IfI believe that (p & I don’t believe that p) then a logical consequence of what Ibelieve is that p, so I believe that p. But if my omissive belief is true then I don’tbelieve that p. This is not, as Sorensen supposes, a case of contradictory beliefsbut rather a flat contradiction. Secondly, I do seem to be absurd in some way if Ibelieve om1. For example, it would be absurd of me to believe that

God exists but I think I am not a believer.

Finally, Sorensen’s appeal to the truth of the principle that belief is closed underlogical consequence is problematic. It is clear that it fails as a psychologicalprinciple. I may believe that a triangle is equilateral without believing that it isequiangular. Nor can it be true of me as a principle of ideal rationality, givenSearle’s Principle (1992: 155–62):

If I believe that p then I have the ability to think the occurrent thoughtthat p.

This principle explains why although we may intuitively suppose that a doghas rudimentary beliefs about the food in its bowl (which helps us explain itsbehavior as it strains at its leash) we hesitate to attribute to it the belief that it willbe beaten in Lent. Clearly, it does not have the concept of Lent and so lacks theability to think thoughts of Lent. The requirement also explains our difficultyin characterizing the beliefs of other species in any fine-grained way, since it isdifficult to specify, using the linguistic expressions of our thoughts, exactly what

22 Introduction

concepts (or derivatively, thoughts) are available to those with radically differentlinguistic capacities and ways of behaving.

Now suppose that I believe that Singapore is a democracy but have no ideawhat a plutocracy is. Since I cannot think the occurrent thought that Singaporeis a plutocracy, I cannot think the occurrent thought that it is either a democracyor a plutocracy. So by Searle’s principle, I cannot believe that Singapore is eithera democracy or a plutocracy. True, this will not stop me believing that thesentence, ‘Singapore is either a democracy or a plutocracy,’ states some truthor other. But that would be a different belief altogether. For one thing, it isnot necessary that the sentence in question state what it does, because we couldhave used ‘plutocracy’ the way we now use ‘workers’ state’. Although I cannotthink thoughts of plutocracy, I may still think the thought that a sentence thatI don’t understand, states some truth or other. My ignorance of what counts asa plutocracy is an indictment of my knowledge, but hardly counts as a failure ofideal rationality. So if I am a maximally rational thinker, but I have no idea whata plutocracy is, I will see that my belief in the democracy of Singapore entailssome truth or other that is stated by the sentence ‘Singapore is either a democracyor a plutocracy’. Moreover, I will believe that this sentence states some truth orother. But I will not know which truth it is. Nor will I believe that truth. Thisfalsifies Sorensen’s principle, even as a principle of ideal rationality.

A defender of Sorensen might endorse Robert Audi’s (1994) distinctionbetween a dispositional belief and a disposition to believe. Then she couldclaim that Sorensen’s principle should be understood as saying that a maximallyrational person is disposed to believe the logical consequences of her originalbeliefs. This move will not, however, help Sorensen. Since I am unable to thinkthoughts about plutocracies, I cannot form any beliefs about them (as opposedto forming beliefs about sentences). But then I cannot be disposed to form suchbeliefs either, simply because I cannot be disposed to form what I am unable toform. Moreover, since an infinite series of similar disjunctions are entailed bya first disjunct, the principle compels us to say that an ideally rational believerwould believe (or be disposed to believe) each of an infinite series of similardisjunctions. Searle’s principle falsifies this prediction in the case of an ideallyrational thinker who lacks the concepts needed to think the second disjunct.

Sorensen tries to circumvent this difficulty by making my ‘thorough obedience’to the principle a test of my degree of ideal rationality (1988: 37). But althoughwe might admit that degrees of rationality are vague, surely there is a differencebetween total obedience and none. So what is missing from Sorensen’s accountis a principled degree of disobedience to the principle that is distinctive of thedegree of Moorean irrationality. This means that Moorean absurdity cannot beexplained in terms of falsehood of the principle either. Is my failure to believe thelogical consequences of all of my beliefs, a form of irrationality? If so, it is a verymild form. My failure to be disposed to form such beliefs seems even milder. Bycontrast, a Moorean believer is guilty of a severe irrationality.


III. Moorean Speech as Practically Irrational

11. The impossibility of Moorean assertion (because of misuse of language)

When a person utters the words, ‘p but I don’t believe that p’, a presupposition ofhonest assertion, namely that one knows or believes or at least does not disbelievewhat one is saying, is not met. Black thinks it follows that it would be neithertrue nor not true that the speaker has made an honest assertion. And from this, heinfers that she has made no assertion at all. She has instead attempted to make anassertion but failed (Black 1952). But suppose I say, in exasperation to an obtusepsychiatrist who keeps on reassuring me that people aren’t persecuting me, ‘Lookhere, I bloody well know that they aren’t, but I can’t help believing they are!’That would evince a rational recognition of my own irrationality. Couldn’t heunderstand what I was trying to say? Further, as we have seen, it is not clear howone might argue that it is impossible to believe ‘p but I don’t believe that p,’ asopposed to arguing that such a belief must put one in error.

12. The impossibility of Moorean assertion (because of assertabilityconditions)The speech acts of asserting p and of asserting that I think that p have, ‘. . . roughlythe same conditions of assertibility. Any conditions in which I could say that pare conditions in which I could say I think that p,’ and the converse holds as well(Rosenthal 1995b: 320). Thus in response to the question whether it is raining,one could just as appropriately reply, ‘Yes’ as say, ‘I think so.’ Assume now thatif p has a set p(AC) of assertibility conditions, then not-p has a disjoint set ofassertibility conditions, that is p(AC) ∩ not-p(AC)={}; likewise that p(AC)=Ibelieve p(AC). Thus, if ‘p’ is assertible for speaker S, then ‘I believe that p’ is aswell; whence ‘it is not the case that I believe that p’ is not assertible for S. And if ‘Ibelieve that p’ is assertible for S, then ‘p’ is as well; whence ‘not-p’ is not. Hence,assuming that assertion distributes over conjunction, any attempt to assert ‘p butI don’t believe it’ must fail, and one attempting to assert this Moorean sentenceshows a severe failure of practical rationality. On the assumption that ‘not-p’ and‘I believe that not-p’ also have roughly the same assertibility conditions, a similaraccount can apply to the commissive case.¹²

Why, however, should it be granted that the assertibility (as opposed to thetruth-) conditions for p and not-p are disjoint? If this is intended to flow fromShoemaker’s Principle, we shall need an argument for why it is not possible to

¹² Shoemaker (1996) offers a similar account in terms of ‘assent conditions’, that is conditionsunder which it would be appropriate to assent to a proposition, leaving conversational factorsaside. He restricts discussion to what he calls ‘mental assent’, which is an episodic instantiationof belief—what others might call judgment. Shoemaker also espouses only one direction of theconnection, namely that any condition under which it is appropriate mentally to assent to p is acondition under which it is appropriate mentally to assent to ‘I believe that p.’

24 Introduction

believe both p and not-p, or some other incursion into doxastic logic. To thisend Shoemaker (1996) considers ‘assent conditions’, that is conditions underwhich one assents to a proposition. He restricts discussion to what he calls‘mental assent’, which is an episodic instantiation of belief—what others mightcall judgment. He holds as well that one cannot mentally assent to p withoutmentally assenting to I believe that p, at least if both contents present themselvesfor consideration. Hence, if one mentally assents to p and I don’t believe that p,one mentally assents to p and to I don’t believe that p (by belief-distribution); butone’s mental assent to p entails that one also mentally assents to I believe that p ifone entertains the question whether one believes that p. Thus, if one entertainsthis question, and mentally assents to the omissive Moorean proposition, onementally assents both to a proposition q and its negation, thus showing a severefailure of theoretical rationality. Whether an analogous line of thought appliesto the commissive case is unclear. It would require the further premise that ifone mentally assents to the proposition I believe that p, then one mentally assentsto the proposition that p if one considers whether p. We know of no soundargument on behalf of this conclusion.

13. The inexpressibility of Moorean beliefSuppose that one could assert ‘p but I don’t believe it’. In that case the firstconjunct would express belief that p, yet the second conjunct would express beliefthat not-p (Wolgast 1977: 119). But, Wolgast contends, one cannot believe boththat p and not-p (1977: 118). For this reason one cannot use ‘p but I don’tbelieve it’ to express beliefs. Since, further, an assertion is an expression of belief,one cannot, contrary to our supposition, assert the Moorean sentence. Henceany attempt to assert this sentence must show a failure of practical rationality.For this account to be persuasive, we should need reason to deny that one canbelieve contradictory propositions. We would also need reason to believe thata disavowal of belief, ‘I don’t believe that p’, is also an expression of disbelief,namely that I believe that not-p. After all, someone who asserts ‘I don’t believethat God exists nor do I believe that He does not’ has not expressed the beliefthat God does not exist.

14. Moorean assertion as a self-defeating speech actEach speech act requires that certain conditions be met for it to be performed non-defectively. One who utters a Moorean sentence cannot use it to perform a non-defective speech act, and thus cannot use it to perform a speech act (Vanderveken1980; Searle & Vanderveken 1985). For Vanderveken, this is due simply to hisPostulate VIII that, ‘. . . no speaker can succeed in simultaneously committinghimself to having a psychological state and asserting that he does not have thatpsychological state’ (1980: 264). The commitment to the psychological state inquestion is, of course, carried by the assertion of p: The speaker is committed to


believing that p. Postulate VIII seems too strong, however. I could believe a setof propositions that, unbeknownst to me, imply that I believe that p. If in sucha situation I were to say, ‘I do not believe that p’, I would seem to betray myinconsistency by means of a speech act rather than failing to perform a speechact at all. We note also that the present approach focuses on omissive Mooreanpropositions, and leaves it unclear how it is to be applied to the commissive case.

15. The speaker’s self-defeating intentionsSomeone who makes a Moorean assertion that p has self-defeating intentionsdue to her primary intention of imparting knowledge or instilling belief that p inher audience (Baldwin 1990, following a suggestion by Burnyeat 1968). Baldwinholds that if I assert that p to you then I have the primary intention that you willcome to believe that p by recognition of this intention. In so intending I have thesecondary intention to make you believe that p, as well as a secondary intentionto make you believe that I believe that p. So when I assert that p, the second ofmy secondary intentions is to make you believe that I believe that p. But when Igo on to assert, in the omissive case, that I don’t believe that p, the first of mysecondary intentions is to make you believe that I don’t believe that p. So I intendto make you form contradictory beliefs about what I myself believe. And when Igo on to assert, in the commissive case, that I believe that not-p, the first of mysecondary intentions is to make you believe that I believe that not-p. So I intendto make you think that I have contradictory beliefs about whether p.

This latter intention can only be strictly described as my single intention tomake you believe both that I believe that p and believe that not-p. To derivethis from my intention to make you believe that I believe that p, together withmy intention to make you believe that I believe that not-p, would require otherprinciples. Such principles would likewise be needed in the omissive case, inorder to derive my single intention to make you believe both that I do and Idon’t believe that p, from my intention to make you believe that I believe that p,together with my intention to make you believe that I don’t believe that p. Evenif such principles could be provided and defended, it remains unclear why theseintentions are supposed to be absurd. One way an intention can be absurd is bybeing an intention to bring about an impossibility. But given the possibility ofcontradictory beliefs I intend an impossibility in neither case.

Perhaps the thought is instead that I will charitably assume, in the omissivecase, that you will not be so irrational as to form contradictory beliefs. Thus Ishould be able to work out that my primary intention will be frustrated. And inthe commissive case, I will see that you will not be so uncharitable as to think thatI have contradictory beliefs, so I should again realize that my primary intentionwill be frustrated. But in the omissive case, suppose that I do intend to makeyou form contradictory beliefs (for example, in the fashion of 1984’s thoughtpolice), or that, in the commissive case, I do intend to make you think (or evenlet you know) that I am irrational in holding contradictory beliefs (for example

26 Introduction

as in O’Brien’s apology to Smith in 1984). Suppose further that I succeed. Nowthe absurdity of my assertion does not seem explicable merely in terms of thefrustration of these intentions.

By contrast, Jones argues that if I assert that p to you then I have the primaryintention that you will come to know that p by means of your recognition thatI intend this (1991: 185). He argues as well that this intention succeeds only ifyou believe that I know the truth of what I assert. Since assertion distributes overconjunction, if I assert that (p and I believe that not-p) then I assert that I believethat not-p. But Jones thinks that you will then ‘inevitably suppose that there issome reason’ why I believe that not-p (1991: 185) and thus you will not believethat I know that p. So you will not come to know that p, thus frustrating myprimary intention.

The inevitability of this supposition can be challenged (see Welbourne 1992:238). Moreover, Jones does not consider the omissive Moorean assertion, ‘p butI do not believe that p’. However, Jones could repair his account as follows: ifyou believe that I know the truth of what I assert then you will believe that whatI assert is true. So when I assert that (p and I believe that not-p) you will thinkthat I believe that not-p. So you will not think that I know that p, since you willknow that I cannot know what I believe to be false. Similarly, when I assert that(p and I don’t believe that p) you will think that I don’t believe that p and againyou will not think that I know that p, since you will know that I cannot knowwhat I don’t believe. So in either case, my primary intention is frustrated.

Welbourne agrees with Jones that if I make an assertion to you then I havethe primary intention that you will come to know what I assert because yourecognize that I intend this (1992: 237), but thinks that this intention succeedsonly if you come to believe that I know what I assert. But when I assert to youthat (p and I don’t believe that p) I assert that I don’t believe that p. If youbelieve that I know that I don’t believe that p then you must believe that I don’tbelieve that p. So you must accept that a condition of my knowing that p (namelymy belief that p) is not satisfied. Accordingly, you won’t believe that I know thatp, so you won’t believe that I know that (p and I don’t believe that p) either. SoI have frustrated my intention in asserting this, because I have told you in effectthat I am insincere. Jones accounts for the commissive Moorean assertion in thesame way on the assumption that a condition of my knowing something is that Idon’t believe that it is false. We note that it seems plausible to say that I can onlyimpart knowledge to you if I make you think I have the knowledge in the firstplace. (We further note that Jones needs the principle that if you believe that Iknow a conjunction then you believe that I know its conjuncts.)

All three of these accounts assume that Moorean assertions only exemplifyone intentional type. Jones and Welbourne think of asserting as letting know,where the assertor intends to impart knowledge, while Baldwin thinks of it as acase in which the assertor intends to instil belief. So lies fit neither Jones’s norWelbourne’s account, since a liar does not intend to impart knowledge. One’s


assertion may also be described as a provocative contention when one intendsto ‘wind up’ her hearer. One’s intention in this case is to remain divided inbelief and so it fails to fit Baldwin’s account. But the supposition that a Mooreanassertor is lying or ‘winding her hearer up’ does nothing to remove the absurdity,especially since the hearer is rarely in a position to know the correct descriptionof the assertion.

16. Social pragmatismThe utterer of a Moorean sentence fails to meet the conditions given by Brandom’ssocial pragmatism (1983, 1994) for the making of an assertion. According tothis approach, one who asserts that p undertakes justificatory responsibility forthat proposition in the following sense: To an interlocutor’s challenge to thatassertion with such words as, ‘How do you know?’, the assertor is obliged toreply with a justification (be it an argument, an appeal to sensory experience,or deferral to another’s authority). If such a challenge is made and the speakergives an inadequate reply or none at all, she is obliged to retract that assertion.Doran holds that one who asserts both that p and that she doesn’t believe thatp, cannot possibly fulfil both sets of justificatory obligations that this pair ofassertions creates: Justifying p will prevent her from being able to justify herassertion that she does not believe that p, and vice versa. For this reason shecannot use the Moorean sentence to make an assertion (Doran 1995). Hence herutterance is severely practically irrational. Doran does not discuss the commissivecase; evidently she would deny that a speaker could justify both an assertionthat p and that she believes not-p. However, it is unclear what would preventa speaker from justifying the claim that most spiders are harmless (by appeal toarachnidan evidence) and a further claim that she believes otherwise (by appeal toevidence concerning her own phobic behavior). Furthermore, from the premisethat a speaker ought to retract a challenged assertion that she has been unableto justify, it does not follow that she has made no assertion at all; stubbornallegiance to untenable views is unfortunately all too quotidian. Accordingly, weneed a fuller account of normative constraints on assertion to infer that assertionof a Moorean sentence is impossible rather than merely inappropriate.

17. The incredibility of the speakerWilliams holds that an assertor offers his hearer the prima facie justification tobelieve him, in other words, to believe that he is sincerely telling the truth (1996:136). But when the assertion is Moorean, the offer turns out to be worthless. Forif you believe me when I tell you that (p and I don’t believe that p) then youmust think that I believe that p (in virtue of believing me sincere) and also believethat I don’t believe that p (in virtue of believing that I tell the truth). So you havecontradictory beliefs if you believe me. And in the commissive case, if you believeme when I tell you that (p and I believe that not-p) then you must think that I

28 Introduction

believe that p (in virtue of believing me sincere) and also believe that I believethat not-p (in virtue of believing that I tell the truth). So this time you mustthink that I have contradictory beliefs if you believe me. You thus cannot believeme in either case without being theoretically irrational or thinking that I amtheoretically irrational. Williams contends that since an assertor should assumethat both he and his hearer will charitably avoid ascriptions of irrationality ifpossible, the assertor is thus in a position to see with minimal reflection that hewill not be believed in either case. Since being believed is the point of assertion,the assertor should see (again with minimal reflection) that his plan (to bebelieved) is bound to fail. So such an assertor is practically irrational.

It may be doubted, however, whether an assertor who does not plan to bebelieved is thereby severely practically irrational. After all, one might feel sure,under interrogation, that the authorities think her guilty, and yet maintain herinnocence with no intention of being believed. Further, it is not clear that theassertor of a Moorean sentence as Williams conceives her is always in a positionto see that her plan to be believed is bound to fail. Perhaps she thinks it areasonable bet that her interlocutor will simply take her at her word and find herinconsistent. In that case it is not clear that her Moorean utterance is severelypractically irrational.

IV. Moorean Judgment as Practically Irrational

18. The incredibility of Moorean sentencesHintikka holds that ‘the gist of Moore’s paradox may be said to lie in the fact that’the omissive proposition p and I don’t believe that p ‘is necessarily unbelievableby the speaker’ (1962: 67). He espouses the principle that

‘in the case of an individual’ a failure to obey the principle: if I believe thatp then I believe that I believe that p, may be taken as impossible (1962: 67)

He also holds that if I believe that (p and I don’t believe that p) then I believe thatp (since belief distributes over conjunction) and so (from Hintikka’s principle) Ibelieve that I believe that p. But since belief distributes over conjunction, I believethat I don’t believe that p. Thus I have contradictory higher-order beliefs aboutwhether I believe that p, and Hintikka seems to take this to be impossible.¹³

Hintikka’s principle does not seem true as a universal law of psychology, giventhat my belief that p may be a prejudice that I fail to recognize within myself.Nor is its failure that of inconsistent belief. Hintikka could stipulate that myobedience to it, like my avoidance of contradictory beliefs, is a condition of

¹³ Shoemaker at one point also claims that Moorean belief is impossible (1996: 85–6).Presumably he would not on this basis hold that Moorean belief involves a severe failure of theoreticalrationality. Likewise for Goldstein (2000: 86): ‘. . . we not only cannot assert Mooronically but alsocannot believe Mooronically; why we cannot have a Mooronic attitude’.


theoretical rationality. Then my Moorean belief would be impossible, given thatI’m perfectly rational. Then, however, the account faces the problem, discussedabove, of explaining the absurdity of my Moorean beliefs in circumstances suchas self-deception that defeat my obedience to it.

F. OPEN QUESTIONS AND CONTRIBUTIONS TO THIS

VOLUME

We have canvassed eighteen different attempts to explain Moorean absurdity.Not all these attempts appear equally viable, yet even after rejecting a substantialnumber of these we have a surfeit of candidate explanations. On the one handthis suggests the fecundity of Moorean absurdity as a source of philosophicalspeculation; on the other it raises the question whether such speculation mightbe all too ad hoc. If a candidate explanation is to be acceptable it must respect thedesiderata formulated in Section D. In addition, however, a candidate explanationgains plausibility as it provides answers to one or more of the following, currentlyunresolved questions. We raise some of these questions below, indicating whena contribution to this volume offers an answer to it.

Does Moorean absurdity find precursors in philosophers before Moore?

Sorensen in ‘The All-Seeing Eye: A Blind Spot in the History of Ideas’ (Chapter 2)finds isolated approximations to Moorean absurdity in Jean Buridan, Parmenides,Plato, Sextus Empiricus, Augustine, Descartes, Spinoza, and others. He arguesthat Moore’s paradox is the culmination of a critique of idealism within theanalytic tradition and considers the history of theories of vision to provide abackdrop for the notion of a blindspot. In ‘Moorean Pretense’ (Chapter 11),Robert Gordon argues that Descartes’s Meditations can be seen as a form ofMoorean pretence, for example Descartes’s pretence that that he falsely believesthat he has a body.

Can Moorean Absurdity be given a grammatical characterization?

In an effort rigorously to delineate our topic, some have sought a grammaticalcharacterization of Moorean sentences. Jay Atlas, in his ‘What Reflexive PronounsTell Us about Belief: A New Moore’s Paradox De Se, Rationality, and PrivilegedAccess’ (Chapter 6), argues that Moorean absurdity need not arise in an utteranceof the form, ‘p but I do not believe it’. Rather, what is needed is an indirectreflexive pronoun such as ‘I myself ’ whose proper use guarantees that the speakerknows that she is speaking of herself. On this basis, Atlas challenges claimsby Moran (2001) that Moorean absurdity exemplifies a dichotomy between

30 Introduction

theoretical and practical approaches to the self. Atlas also challenges an argumentoffered by Shoemaker (1996) that if one believes that p, then one believes thatone believes that p. He concludes more generally that the ordinary concept ofbelief does not bring with it a commitment to privileged access.

What sort of infraction generates the absurdity?

We said in Section E that extant explanations of Moorean absurdity accountfor that phenomenon either as a severe violation of theoretical rationality, or asa severe violation of practical rationality. Is one of these the correct system ofnorms in terms of which to account for Moorean absurdity, or are there othersystems whose violation might instead account for that phenomenon? Claudiode Almeida, in ‘Moorean Absurdity: An Epistemological Analysis’ (Chapter 3)offers an account of Moorean belief as a form of theoretical irrationality. Hecharacterizes Williams’s (1994) analysis as: You rationally believe that p only ifthere is no simple and compelling argument that we can reasonably expect youto be aware of to the effect that your believing that p is either self-falsifying orensures the presence of contradictory beliefs in your doxastic system. De Almeidaargues that this account is both unnecessary and incomplete as an explanationof the absurdity of Moorean belief, and offers a rival account from which hedraws general conclusions about the nature of justification. Thomas Baldwin in‘The Normative Character of Belief ’ (Chapter 4) construes Moorean absurdityas consisting in a set of inconsistent commitments whether that absurdity beproduced by an assertion or a belief. Baldwin motivates his approach with ananalogy between judgments and performative utterances, and is careful to eschewwhat we described above (Section D.3) as principles of epistemic or doxastic logicin the course of his explanation. Finally, Mitchell Green (‘Moorean Absurdityand Showing What’s Within’, Chapter 9) offers an account of Moorean absurdityas violating either norms of theoretical rationality, or norms internal to speechacts such as assertion, while leaving it open in any given case whether it is theone or the other.

Does Moorean absurdity also come in forms involving other epistemicoperators besides belief?

For instance, ‘It’s raining, but I don’t know that it is’ sounds paradoxical tosome. Similarly, ‘It’s raining but I’m not certain that it is’ sounds paradoxical tosome. Other candidates include ‘I believe that Moore has two hands but I don’tbelieve that I know that he has two hands’ and ‘It’s raining but I know that I’mconvinced that it’s not raining’ and even ‘It’s raining but I have no justification atall for believing that it’s raining’. Adler and Armour-Garb, in ‘Moore’s Paradoxand the Transparency of Belief ’ (Chapter 7) investigate the question with somecare, concluding that only cases involving belief generate Moorean absurdity.


Their reason is that only belief has the property of transparency: An attitude Ais transparent to its content C just in case one who bears A to C is such thatC holds (from their point of view). Belief is transparent, given this definition,whereas, for instance, desire is not.

May Moorean absurdity be exemplified in other speech acts besideassertion?

Some authors claim that questions and imperatives can generate Mooreanabsurdity, as in, ‘What time is it? but I don’t want to know what time it is,’ or‘Shut the door but I don’t want you to shut the door’ (Searle and Vanderveken1985; but for dissent see Heal 1977). Others have argued that Moorean absurdityis found even when an indicative sentence is put forth with other than assertoricforce, as in the case of supposition (Green 2000). Adler and Armour-Garbinvestigate the question, as does Green.

Might an account of Moorean absurdity provide a way of identifyingfurther examples of this phenomenon?

For example, are ‘I have no beliefs now’, ‘All my beliefs are false’, or ‘I don’t exist(as a believer)’, cases of Moorean absurdity? If ‘It’s raining but I’m convincedthat I believe it isn’t’ is genuinely Moorean, is ‘It’s raining but I believe I’mconvinced it isn’t’? If Superman informs me that I’m acquainted with him whenhe is disguised as another person, whom I think idiotic, is my remark, ‘I falselybelieve that you are an idiot’, an instance of Moorean absurdity? (Crimmins1992, discussed by Rosenthal 2002, Hajek and Stoljar 2001). Was Luis Bunuel’sremark, ‘I’m still an atheist, thank God,’ as he was banished from Spain forattacking religion, a case of Moorean absurdity? Acknowledging that this abilityto identify further examples of Moorean absurdity is consistent with the inabilityto enumerate all such instances (Sorensen 2000), Williams (‘Moore’s Paradoxes,Evans’s Principle, and Iterated Belief ’, Chapter 5) notes Sorensen’s commentthat

om1) p & I don’t believe that I believe that p

and

com1)p & I believe that I believe that not-p

seem less absurd to believe or assert than

om) p & I don’t believe that p

or

com) p & I believe that not-p.

Moreover, as iteration increases, omissive absurdity appears to decrease, whilecommissive absurdity does not. Thus, with four iterations,

32 Introduction

om4) p & I don’t believe that I believe that I believe that I believe that Ibelieve that p

seems less absurd to believe or assert than om1, whereas the absurdity of believingor asserting

com4) p & I believe that I believe that I believe that I believe that I believethat not-p

seems undiminished. Williams argues that despite appearances, any belief in anyiteration of om or com is equally absurd. However, he agrees that there is adecrease in the absurdity of an assertion of om) as iterations increase. He alsodefends a definition of Moorean belief: that any belief is Moorean just in case itscontent is (i) a possible truth that (ii) reports no irrationality in the believer, but(iii) it is impossible to have justification for that belief.

Is Moorean belief, insofar as it makes its holder absurd, conscious?

Andre Gallois in ‘Consciousness, Reasons, and Moore’s Paradox’ (Chapter 8)argues that only consciously held attitudes can provide one with reasons foraction or for the holding of an attitude. Gallois concludes that only consciouslyheld beliefs are integrated with the rest of the self, and he accounts for theabsurdity of a new species of Moorean absurdity. In the process he relatesthese issues to integration in fiction, and to different ways of attributingreasons.

Does having a Moorean ‘ring’ refute a philosophical position?

Some philosophical positions seem to imply that it would be appropriate to saysomething having the form of a Moorean absurdity. For instance, an eliminativistin the philosophy of mind denies that there are any beliefs. She is thus committedto the following conjunction: ‘There are no beliefs, though I don’t believe it.’Does this commitment refute her position? Alan Hajek’s ‘My PhilosophicalPosition Says ‘‘p’’ and I Don’t Believe ‘‘p’’,’ (Chapter 10), investigates thisquestion, attending to eliminativism about belief, the view that belief requiresassignment of high or maximal probability, the denial of higher-order beliefsor higher-order probabilities, eliminativism about truth, pragmatism, relativismabout truth, truth-value gluts or gaps, and the philosophical ‘meta-induction’that all philosophical theories are false.

Is the study of Moorean absurdity crucially an investigation into thepathology of belief or other attitudes?

It is natural to hold, given the absurdity of such thoughts or utterances as ‘Pbut I don’t believe it’, that Moorean absurdity is primarily of interest for failures


or rationality. However, in his paper ‘Moorean Pretense’ (Chapter 11), RobertGordon argues that the ability to pretend-true a situation such as is expressedby the just-quoted sentence may be a cognitive achievement. On his account,Moorean pretense is not absurd. It underlies Cartesian doubt, is rampant inmodern science, and is a defining feature of dramatic irony. It may underlie theability to understand that people’s behavior is sometimes based on false beliefsrather than on the actual facts. Most children gain this ability at about the age of4. However, children with autism usually do not acquire it at all, and Gordonspeculates that this failure may be tied to their well-established failure to engagein spontaneous pretense.

REFERENCES

Adler, J. (1999), ‘The Ethics of Belief: Off the Wrong Track’, Midwest Studies inPhilosophy, 23: 267–85.

(2002), Belief ’s Own Ethics (Cambridge, Mass.: MIT Press).Albritton, R. (1995), ‘Comments on ‘‘Moore’s Paradox and Self-Knowledge’’ ’, Philo-

sophical Studies, 77: 229–35.Armstrong, D. (1971), ‘Meaning and Communication’, Philosophical Review, 80:

427–47.Audi, R. (1994), ‘Dispositional Beliefs and Dispositions to Believe’, Nous, 28: 419–34.Baldwin, T. (1990), G. E. Moore (Routledge).Black, M. (1952), ‘Saying and Disbelieving’, Analysis, 13: 25–33.Bovens, L. (1995), ‘P and I Will Believe that not-P: Diachronic Constraints on Rational

Belief ’, Mind, 104: 737–60.Brandom, R. (1983), ‘Asserting’, Nous, 17: 637–50.

(1994), Making It Explicit (Cambridge, Mass.: Harvard University Press).Brueckner, A. (1998), ‘Moore Inferences’, Philosophical Quarterly, 48: 366–9.Burnyeat, M. (1968), ‘Belief in Speech’, Proceedings of the Aristotelian Society, 68: 227–48.Castell, P. (1994), ‘Moore’s Paradox and Partial Belief ’, European Review of Philosophy,

1: 45–53.Clark, R. (1994), ‘Pragmatic Paradox and Rationality’, Canadian Journal of Philosophy,

24: 229–42.Cohen, L. (1950), ‘Mr. O’Connor’s ‘‘Pragmatic Paradoxes’’ ’, Mind, 59: 85–7.Collins, A. (1996), ‘Moore’s Paradox and Epistemic Risk’, Philosophical Quarterly, 46:

308–19.(1987), The Nature of Mental Things (South Bend Ind.: University of Notre Dame

Press).Crimmins, M. (1992), ‘I Falsely Believe That P’, Analysis, 52: 191.Davis, W. (2002), Meaning Expression and Thought (Cambridge: Cambridge University

Press).De Almeida, C. (2001), ‘What Moore’s Paradox is About’, Philosophy and Phenomenolo-

gical Research, 62: 33–58.DeRose, K. (1991), ‘Epistemic Possibilities’, Philosophical Review, 100: 581–605.

(1996), ‘Knowledge, Assertion, and Lotteries’, Australasian Journal of Philosophy,74: 568–80.

34 Introduction

DeRose, K. (2002), ‘Assertion, Knowledge, and Context’, Philosophical Review, 111:167–203.

Deutscher, M. (1965), ‘A Note on Saying and Disbelieving’, Analysis, 25: 53–7.(1967), ‘Bonney on Saying and Disbelieving’, Analysis, 27: 184–6.

Doran, K. (1995), ‘Moore’s Paradox, Asserting, and Skepticism’, Southwest PhilosophyReview, 11: 41–8.

Evans, G. (1982), The Varieties of Reference (Oxford: Oxford University Press).Forguson, L. (1969), ‘On ‘‘It’s raining but I don’t believe it’’ ’, Theoria, 34: 89–101.Gallois, A. (1996), The World Without, the Mind Within (Cambridge: Cambridge

University Press).Gillies, A. (2001), ‘A New Solution to Moore’s Paradox’, Philosophical Studies, 105:

237–50.Goldstein, L. (1993), ‘Inescapable Surprises and Acquirable Intentions’, Analysis, 53:

93–9.(1988), ‘Wittgenstein’s Late Views on Belief, Paradox, and Contradiction’, Philo-

sophical Investigations, 11: 49–73.(2000), ‘Moore’s Paradox’, in P. Engel (ed.), Believing and Accepting (Dordrecht:

Kluwer Academic Publishers), 65–92.Gombay, A. (1988), ‘Some Paradoxes of Counterprivacy’, Philosophy, 63: 191–210.Gordon, R. (2000), ‘Sellars’ Ryleans Revisited’, Protosociology, 14: 102–14.Green, M. (1999), ‘Moore’s Many Paradoxes’, Philosophical Papers, 27: 97–109.

(2000), ‘The Status of Supposition’, Nous, 34: 376–99.Hajek, A., and Stoljar, D. (2001), ‘Crimmins, Gonzales, and Moore’, Analysis, 61:

208–13.Hambourger, R. (1984), ‘Moore’s Paradox and Epistemic Justification’, Philosophy

Research Archives, 10: 1–12.Harnish, R. (1980), ‘Searle and the Logic of Moore’s Paradox’, International Logic Review,

11: 72–7.Heal, J. (1977), ‘Insincerity and Commands’, Proceedings of the Aristotelian Society, 77:

183–202.(1994), ‘Moore’s Paradox: A Wittgensteinian Approach’, Mind, 103: 5–24.(2003), Mind, Reason, and Imagination (Cambridge: Cambridge University Press).

Hintikka, J. (1962), Knowledge and Belief (Ithaca, NY: Cornell University Press).Jacobson, R. (1996), ‘Wittgenstein on Self-Knowledge and Self-Expression’, Philosophical

Quarterly, 46: 12–30.Jacquette, D. (2000), ‘Identity, Indiscernibility, and Moore’s Paradox’, Synthese, 123:

279–92.Jones, O. R. (1991), ‘Moore’s Paradox, Assertion and Knowledge’, Analysis, 51: 183–6.Kind, A. (1993), ‘Shoemaker, Self-Blindness and Moore’s Paradox’, Philosophical

Quarterly, 53: 39–48.Klein, P. (1986), ‘Immune Belief Systems’, Philosophical Topics, 14: 259–80.Kobes, B. (1995), ‘Telic Higher-Order Thoughts and Moore’s Paradox’, Philosophical

Perspectives, 9: 291–312.Kriegel, U. (2004), ‘Moore’s Paradox and the Structure of Conscious Belief ’, Erkenntnis,

61: 99–121.


Larkin, P. (1999), ‘Shoemaker on Moore’s Paradox and Self-Knowledge’, PhilosophicalStudies, 96: 239–52.

Lee, B. (2001), ‘Moore’s Paradox and Self-Ascribed Belief ’, Erkenntnis, 55: 359–70.Levinson, S. (1983), Pragmatics (Cambridge: Cambridge University Press).Lycan, W. (1970), ‘Hintikka and Moore’s Paradox’, Philosophical Studies, 21: 9–14.Linville, K., and Ring, M. (1991), ‘Moore’s Paradox Revisited’, Synthese, 87: 295–309.Malcolm, N. (1995), ‘Disentangling Moore’s Paradox’, in R. Egidi (ed.), Wittgenstein:

Mind and Language (Dordrecht: Kluwer Academic Publishers), 195–205.(1984), Ludwig Wittgenstein: A Memoir, 2nd edn. (Oxford: Oxford University

Press).Martinich, A. (1980), ‘Conversational Maxims and Philosophical Problems’, Philosophical

Quarterly, 30: 215–28.Milgram, E. (1994), ‘An Apprentice Argument’, Philosophy and Phenomenological Research,

54: 913–16.Moore, G. E. (1993), Selected Writings, ed. T. Baldwin (London: Routledge).

(1944), ‘Russell’s Theory of Descriptions’, in P. Schilpp (ed.), The Philosophy ofBertrand Russell (La Salle, Ill.: Open Court) 175–225.

(1942), ‘A Reply to My Critics’, in P. Schilpp (ed.), The Philosophy of G. E. Moore(La Salle, Ill.: Open Court), 535–677.

(1912), Ethics (New York: H. Holt and Company).Moran, R. (1997), ‘Self-Knowledge, Discovery, Resolution, and Undoing’, European

Journal of Philosophy, 5: 141–61.Olin, D. (1987), ‘On an Epistemic Paradox’, Analysis, 47: 216–17.Rosenthal, D. (2002), ‘Moore’s Paradox and Crimmins’ Case’, Analysis, 62: 167–71.

(1995a), ‘Self-knowledge and Moore’s Paradox’, Philosophical Studies, 77: 196–209.(1995b), ‘Moore’s Paradox and Consciousness’, Philosophical Perspectives.(1998), ‘Thinking That One Thinks’, in A. Burri (ed.), Language and Thought (De

Gruyter), 259–87.Searle, J. (1992), The Rediscovery of the Mind (Cambridge, Mass.: MIT Press).

and Vanderveken, D. (eds.) (1985), Foundations of Illocutionary Logic (Cambridge:Cambridge University Press).

Shoemaker, S. (1988), ‘On Knowing One’s Own Mind’, Philosophical Perspectives, 2:183–209.

(1995), ‘Moore’s Paradox and Self-Knowledge’, Philosophical Studies, 77: 211–28.Sorensen, R. (1982), ‘Recalcitrant Variations of the Prediction Paradox’, Australasian

Journal of Philosophy, 69: 355–62.(1984), ‘Conditional Blindspots and the Knowledge Squeeze: A Solution to the

Prediction Paradox’, Australasian Journal of Philosophy, 6: 126–35.(1986), ‘A Strengthened Prediction Paradox’, Philosophical Quarterly, 36: 504–13.(1988), Blindspots (Oxford: Oxford University Press).(2000), ‘Moore’s Problem with Iterated Belief ’, Philosophical Quarterly, 50: 28–43.

Stalnaker, R. (2000), ‘On ‘‘Moore’s Paradox’’ ’, in P. Engel (ed.), Believing and Accepting(Dordrecht: Kluwer), 93–100.

Szabo-Gendler, Z. (2001), ‘Fictionalism and Moore’s Paradox’, Canadian Journal ofPhilosophy, 31/3: 293–308.

36 Introduction

Unger, P. (1975), Ignorance: A Case for Scepticism (Oxford: Oxford University Press).Vahid, H. (2005), ‘Moore’s Paradox and Evans’s Principle: A Reply to Williams’, Analysis,

65: 337–41.Vanderveken, D. (1980), ‘Illocutionary Logic and Self-Defeating Speech Acts’, in

J. Searle, F. Kiefer, and M. Bierwisch (eds.), Speech Act Theory and Pragmatics(Dordrecht: Reidel), 247–72.

(1990), Foundations of Illocutionary Logic, i and ii, (Cambridge: CambridgeUniversity Press).

Welbourne, M. (1992), ‘More on Moore’, Analysis, 52: 237–41.Williams, J. N. (1979), ‘Moore’s Paradox—One or Two?’, Analysis, 39: 121–40.

(1982), ‘The Absurdities of Moore’s Paradoxes’, Theoria, 48: 38–46.(1994), ‘Moorean Absurdity and the Intentional ‘‘Structure’’ of Assertion’, Analysis,

54: 160–6.(1996), ‘Moorean Absurdities and the Nature of Assertion’, Australasian Journal of

Philosophy, 74: 135–49.(1998), ‘Wittgensteinian Accounts of Moorean Absurdity’, Philosophical Studies,

92: 283–306.(2004), ‘Moore’s Paradoxes, Evans’s Principle and Self-Knowledge’, Analysis, 64:

348–53.Williamson, T. (1996), ‘Knowing and Asserting’, Philosophical Review, 105: 489–523.

(2000), Knowledge and Its Limits, (Oxford: Oxford University Press).(2001), ‘Comments on Michael Williams’ ‘‘Contextualism, Externalism and

Epistemic Standards’’ ’, Philosophical Studies, 103: 25–33.Willis, R. (1953), ‘Professor Black on Saying and Disbelieving’, Analysis, 14: 24–5.Winch, P. (2001), ‘The Expression of Belief ’, in T. McCarthy and S. Stidd (eds.),

Wittgenstein in America (Oxford: Oxford University Press), 195–214.Wittgenstein, L. (1980a), Remarks on the Philosophy of Psychology, i, ed. G. Anscombe

and G. von Wright (Chicago: University of Chicago Press).(1980b), Remarks on the Philosophy of Psychology, ii, ed. G. von Wright and

H. Hyman (Chicago: University of Chicago Press, 1980).(1980c), Culture and Value, 2nd edn., ed. G. H. Von Wright, trans. P. Winch

(Oxford: Blackwell).(1974), Letters to Russell, Keynes and Moore, ed. G. H. von Wright (Oxford:

Blackwell).(1997), Philosophical Investigations, 2nd edn., trans. G. Anscombe (Oxford:

Blackwell).Wolgast, E. (1977), Paradoxes of Knowledge, (Ithaca, NY: Cornell University Press).

2

The All-Seeing Eye: A Blind Spot in theHistory of Ideas

Roy Sorensen

In a letter to child-friend Mary MacDonald, in 1864, Lewis Carroll warned:

Don’t be in such a hurry to believe next time—I’ll tell you why—If you set to work tobelieve everything, you will tire out the muscles of your mind, and then you’ll be so weakyou won’t able to believe the simplest true things. Only last week a friend of mine set towork to believe Jack-the-giant-killer. He managed to do it, but he was so exhausted by itthat when I told him it was raining (which was true) he couldn’t believe it, but rushedout into the street without his hat or umbrella, the consequence of which was his hair gotseriously damp, and one curl didn’t recover its right shape for nearly two days.

In this passage, Lewis Carroll comes within inches of formulating the sentenceG. E. Moore made famous: ‘It is raining but I do not believe it’.

Why didn’t Lewis Carroll anticipate Moore? Carroll was a capable logicianwith an ear for peculiar sentences. And what about the capable philosophers whopreceded Carroll?

This question arose when I was writing a book on the history of para-doxes (Sorensen 2003). Study of Moore’s paradox had been a formative experiencefor me in graduate school. Finding a precursor of Moore became a pet project.My confidence was partly on general grounds. I was charmed by Stigler’s Lawof Eponymy: ‘No scientific discovery is named after its original discoverer’:Russell’s paradox was discovered by Cantor, Jourdain’s Visiting Card paradoxwas discovered by G. G. Berry, and Moore’s paradox of analysis was formulatedby C. H. Langford as an objection to Moore. (Yes, Stigler did not discover Stigler’sLaw of Eponymy; Robert Merton has priority.)

Stigler’s Law of Eponymy also works for relatives of paradoxes such as cognitiveillusions and long overlooked counterexamples. A particularly germane exampleis Fitch’s ‘knowability paradox’—which Timothy Williamson (2000: 271)classifies as an ‘embarrassment’ rather than a paradox. Idealism, verificationism,and semantic anti-realism each imply that all truths are knowable. Yet FredericFitch (1963) proves, with surprisingly little logic, that there are unknowabletruths of the form ‘p but it is not known that p’. Fitch forthrightly attributes

38 The All-Seeing Eye

the key insight to a referee report composed in 1945 on a manuscript henever published. Recent archival research has revealed the referee to be AlonzoChurch (Salerno 2006). Since G. E. Moore published his discovery of thepeculiarities of ‘p but I do not believe it’ in 1943, the timing and structuralsimilarity of the sentences invites the conjecture that the ‘knowability paradox’was inspired by Moore’s paradox. This legacy would be particularly fitting. The‘knowability paradox’ points the arrow of logic precisely at Moore’s targets.

Is ‘Moore’s paradox’ another tardy counterexample? I felt a precursor waslikely. Sentences of the form ‘p but I do not believe it’ are simple to formulate.Indeed, they have common currency as expressions of surprise. Their oddity issalient. And there is a long history of commentary on similar sentences. Forinstance, analysis of the liar paradox ‘This statement is false’ goes back 2,500years. The fourteenth-century logician Jean Buridan analyzes ‘I do not believethis sentence’ in his Sophismata. All philosophers know that the self-defeatingnature of ‘I do not exist’ became a major preoccupation of philosophers afterDescartes’s cogito ‘I think, therefore, I exist’.

I did find a sequence of approximations to Moore’s paradox that initiallyseemed close enough. But on reflection, none of these stood up as genuineanticipations. To my chagrin, I had better luck finding anticipation of some ofmy own ideas in Blindspots.

I am now inclined to think that there was a blind spot in the history of ideas.Access to Moore’s paradox required a configuration of background beliefs thatonly came into alignment in the twentieth century (and may fall out of alignmentin the future). This consensus was a legacy of the revolution Moore led againstthe Hegelian idealists. Moore’s paradox is a child of analytic philosophy. And itis a counterexample to Stigler’s Law.

I still hold out hope that Carroll’s humor contributed to G. E. Moore’sdiscovery. In his Commonplace book Moore (1962: 123) relates Descartes’s cogitoto the dream passages from Carroll’s Through the Looking Glass. G. E. Moore,J. M. E. McTaggart, and Bertrand Russell, who dined together at Trinity HighTable, were known as the Mad Tea Party of Trinity (Wiener 1953: 194–5). Thiswas partly on the basis of their resemblance to, respectively, the March Hare, theDormouse, and the Mad Hatter (see Figure 1).

The direct connection between Moore’s paradox and the Mad Tea Party is withMoore himself. But one can also see indirect connections with Russell throughhis protege Ludwig Wittgenstein and with McTaggart through his disseminationof Hegelian idealism. My thesis is that Moore’s paradox is the culmination of acritique of idealism within the analytic tradition. Moore instigated this rebellion.Russell soon joined him. However, the roots of Moore’s paradox extend deepinto the historical strata.

Roy Sorensen 39

Figure 1. Alice at the Mad Hatter’s tea party, by John Tenniel

PRAGMATIC PARADOX IN ANTIQUITY

The phrase ‘pragmatic paradox’ only came into philosophical currency withD. J. O’Connor’s article ‘Pragmatic Paradoxes’ in 1948. But the phenomenongoes back to pre-literate folklore.

The walls of Pompeii are festooned with graffiti (preserved by the volcanic ashof Vesuvius in 79 ). Some of the inscriptions lament the defacement: ‘I wonder,O wall, that you have not fallen in ruins from supporting the stupidities of somany scribblers’. Some of the stupidity is an affectation: ‘Everyone writes on thewall except me.’ This inscription shows that Pompeians on the street were familiarwith pragmatic paradoxes (statements that are supported or undermined by thecircumstances of their utterance). Like an analytic statement, there is internalevidence of the pragmatic paradox’s truth or falsity. The analogy with analyticstatements leads post-Kantian philosophers to speak of ‘pragmatic contradictions’and ‘pragmatic tautologies’.

Evidence that ordinary people could appreciate pragmatic paradoxes can bedated much earlier. For pragmatic paradoxes were used by Greek playwrights.

Some of the playwrights may have been influenced by philosophical discussionof the problem of false belief. Parmenides denied that we can think about what isnot. It is not there to be thought about. This puzzle is reinforced by our tendencyto model words as names. The meaning of a name is its bearer. So a namewithout a bearer is meaningless. If a sentence means what caused it to be uttered,


then the meaningfulness of the sentence is sufficient for its truth. Belief is thenjust an effect. Alternatively, if beliefs represent what they most resemble, then itsmeaning is what most closely corresponds to the belief. But such correspondenceis truth. Therefore, to interpret a belief as false is to misinterpret it.

Sophists presented these arguments as support for Protagoras’ maxim thatman is the measure: whatever you think is true for you and so is true (Theaetetus161c–179c). Under the name model of meaning, error is hard to model. In theSophist, Plato responded by distinguishing between believing (the propositionalattitude) and what is believed (a proposition). When a student misspells ‘wine’as ‘win’, he spells something but not what his instructor assigned. The mere factthat the misspelling best resembles an existing word does not make it a spelling ofthat word. Just as a letter can contribute to a spelling without itself be a spelling,a word can contribute to a statement without stating anything. ‘Theaetetus’ in‘Theaetetus is flying’ is not a statement even though it does name an individual.Since Theaetetus is actually sitting, the sentence states something different thanwhat is the case. That is how a sentence can be false. Whereas a false eye is notan eye, a false belief is a genuine belief in a false sentence.

Plato’s strategy is to give partial credit to falsehoods. Falsehoods succeedenough to differ from gibberish but not well enough to qualify as a truth. Hehopes to disqualify meaningless statements with the principle ‘Whenever there isa statement, it must be about something’ (Sophist, 262). However, this rendersmeaningless ‘There are flying horses’. Perhaps Plato could extend the spellinganalogy by considering misspellings that get all the letters wrong. However, therewould remain the problem of true negative existentials such as ‘There are noflying horses’. They do not correspond to any existing thing.

Plato is aware that the very act of asserting a proposition sometimes providesevidence against that proposition. But he does not make more than rhetorical useof this phenomenon. In contrast, twentieth-century commentators try to refuteancient Greek philosophers by pointing out that their doctrines are pragmaticallyself-defeating. Speaking of when a Cyrenaic espouses his philosophy of life, C. I.Lewis writes:

There would be no logical inconsistency in his hortation, ‘Have no concern for thefuture,’ if it should be found engraved by lightning on a rock. But for us to take seriouslyone who puts it forward, or for anyone to take himself seriously in accepting it, wouldimply exactly such concern as this injunction advises that we repudiate. (Lewis 1946: 481)

Greek philosophers attached little weight to this genre of criticism. Theyassimilated accusations of self-defeat to argumentum ad hominem and tu quoque.Since there is no genuine inconsistency in a pragmatic contradiction, no dialecticalvictory could be secured by pointing out a pragmatic contradiction.

The Greeks never explicitly draw the distinction between pragmatic incon-sistency and logical inconsistency. But the distinction is implicit in the rulesgoverning dialectical games. The goal was to make your adversary contradict one

Roy Sorensen 41

of his earlier remarks. The speaker was free to say anything (however implausible)as long as it was consistent with what he previously asserted. For instance, hecould assert ‘I am silent’ as long as he did not admit that he also spoke something.The speaker could also assert ‘I do not believe that I am silent’ because the tworemarks are logically compatible.

This logical purity is surprising because pragmatic inconsistency is importantto lawyers. Law connects with philosophy because it is frequently studied as apreparation for a legal career. Juries are impressed by conflicts between word anddeed. The British barrister Frederick Edwin Smith once cross-examined a youngman who alleged that a negligent bus driver damaged his arm. ‘Will you pleaseshow us how high you can lift your arm now?’ asked the barrister. Grimacing,the man gingerly raised his arm to shoulder level. ‘Thank you,’ said Smith. ‘Andnow, please will you show us how you could lift it before the accident?’ Theyoung man shot his arm above his head.

The young man lost the suit. In litigation, self-refutation is potent stuff. Inthe classical dialectical games used to train lawyers, however, self-refutation wasconfined to logical contradiction. The content of one assertion had to conflictwith the content of another assertion. As in mathematics, the preconditions of anassertion were not counted as part of what was asserted.

The intent was to provide a neutral forum for ideas. In this setting, the dialec-ticians were open to whatever is rationally defensible. Only formal contradictioncounted as a fair refutation. Skeptics entered the debate under these generousterms. For instance, Sextus Empiricus notes that some say it is self-defeating todeny the existence of the soul because the soul is needed to issue denials. ButSextus sees little merit to this question-begging maneuver. Even if the soul werenecessary to make assertions, one would first have to prove that souls are required.The skeptic about perceptual knowledge will be unimpressed by testimony fromthe witnesses to the dialogue.

AUGUSTINE’S INTROSPECTIVE TURN

The first philosopher who makes serious use of pragmatic paradoxes is Augustine.Herenovates theoldpragmaticparadox ‘Idonot exist’ intoa fourth-centurycogito:

I am not at all afraid of the arguments of the Academicians, who say, What if you aredeceived? For if I am deceived I am. For he who is not, cannot be deceived; and if I amdeceived, by this same token I am. And since I am if I am deceived, how am I deceivedin believing that I am? For it is certain that I am if I am deceived. Since, therefore, I, theperson deceived, should be, even if I were deceived, certainly I am not deceived in thisknowledge that I am. And, consequently, neither am I deceived in knowing that I know.For, as I know that I am, so I know this also, that I know. (1872: xi. 26)

Augustine, by training, was a rhetorician rather than a philosopher. This madehim less dismissive of pragmatic paradoxes. His conversion to Christianity also


made him take ad hominem considerations seriously. Christianity originated asa persecuted religion of the poor and oppressed. So the emphasis was on whatcould be privately maintained however humble one’s circumstances. Christ’swarning of an imminent apocalypse reduced the attraction of projects thatcompeted with the cultivation of one’s soul. This introspective turn madeAugustine unprecedentedly concerned with mental states. I agree with MylesBurnyeat’s (1982) judgment that Augustine was the first philosopher to draw thesubjective–objective distinction.

Subsequent Christians elaborated this inward perspective. On the ethical front,there was a growth in intention-based ethics running from Abelard to Ockham.On the metaphysical front, there was an elaboration of a subjective style ofmetaphysics in which Plato’s forms are analyzed as the ideas of God.

Our ideas seem to be based on perception—especially on our eyes. Visionis the master sense for human beings. If reality is fundamentally abstract, asPlato contends, perception should not be a source of knowledge. But the eye isso important to human beings that philosophers commonly use visual modelsof knowledge. This includes philosophers who are officially skeptical aboutperceptual knowledge:

And the soul is like the eye: when resting upon that on which truth and being shine, thesoul perceives and understands and is radiant with intelligence; but when turned towardsthe twilight of becoming and perishing, then she has opinion only, and goes blinkingabout, and is first of one opinion and then of another, and seems to have no intelligence.(Republic, V. 88)

Plato associates darkness with ignorance and light with knowledge even thoughhe elsewhere disparages vision as a source of knowledge. Augustine inherits thistendency from Plato and magnifies the light–dark contrast (perhaps as residueof his Manichean phase). So how can one represent the eye in a ‘God’s-eye view’of the world?

One answer is through the icon of the all-seeing eye. This symbol dates backto the ancient Babylonians. It represents the watchful and protective power ofthe Supreme Being. On the American one-dollar bill the all-seeing eye appearsto the left of the inscription in ‘In God we trust’ (as part of the Great Seal of theUnited States).

If God exists, then his perspective converges with reality. There is an intensionalequivalence between what God perceives and what there is. Thus the all-seeingeye foreshadows idealism: To be is to be perceived (from the perfect perspective).

By the nineteenth century, idealist poet-philosophers used visual imagery toexpress pantheistic rapture:

In the woods we return to reason and faith. There I feel that nothing can befall mein life,—no disgrace, no calamity (leaving me my eyes), which nature cannot repair.Standing on the bare ground,—my head bathed by the blithe air, and uplifted intoinfinite space,—all mean egotism vanishes. I become a transparent eye-ball; I am nothing;

Roy Sorensen 43

I see all; the currents of the Universal Being circulate through me; I am part and particleof God. (Emerson 1836: ch. 1)

Among the utilitarians, the all-seeing-eye influenced prison design. JeremyBentham believed that having prisoners under constant surveillance would reducemisbehavior. Accordingly, he proposed that the guard tower be a cylindrical struc-ture going up the middle of the building. Two ‘panopticons’ were actually built.

DESCARTES’S INHOSPITABLE BACKGROUND

Rene Descartes presents himself as a revolutionary, discontinuous with scholasticphilosophy. But he really extends the introspective tradition pioneered byAugustine. Descartes was immersed in Augustine’s writings (which mentions thecogito seven times) as part of his Catholic education at La Fleche.

Descartes does diverge from the scholastics in the way he presses the skepticinto the service of epistemology. The test for knowledge becomes its defensibilityin the face of skeptical attack. The cogito is the centerpiece of Descartes’sphilosophy, not just an interesting limited case. Self-defeating doubts became ageneral philosophical resource.

Philosophers in the wake of Descartes became sensitized to pragmatic incon-sistency. Spinoza (1632–77) makes a side-remark that constitutes a surfaceanticipation of Moore’s paradox:

If anyone says, then, that he has a clear and distinct, that is, a true idea of substance andnevertheless doubts whether such substance exists, he is like one who says he has a trueidea and yet doubts whether it may not be false. (Ethics 1 p8 s2)

Spinoza is picking up on the problem of ascribing a specific error to oneself. Butthe insight is lost in the sauce. Spinoza is an irrealist about error. His argumentsecho those addressed by Plato: An idea of something must correspond to it. Anidea that succeeds in corresponding is a true idea. So all ideas are true ideas.Spinoza also appeals to Parmenides’ theme that error is about what is not. Whatis not is not there to be erred over. In a similar spirit, Spinoza treats error as aprivation and privations as unreal.

In addition to supporting old obstacles to an understanding of Moore’sparadox, Spinoza’s theory of judgment creates some new ones. According toSpinoza, thinking is believing. We never neutrally entertain a proposition. Theappearance of neutrality is due to a tie between belief and disbelief. We assertthe stronger of our beliefs. If the stronger belief is forgotten then the weaker onewill reassert itself. Therefore, any time we think about both sides of an issue,we are latently inconsistent. Since neutrality is impossible ‘p but I am neutralabout p’ has a contradictory conjunct and so is itself a contradiction. ThusSpinoza’s view of thinkers as hyper-opinionated elbows out the omissive versionof Moore’s problem.


No philosopher before Frank Ramsey (1903–30) has a decent theory of belief.So the infrastructure for a nuanced understanding of Moore’s absurdity wasnot present in seventeenth-century philosophy. Nevertheless, Spinoza’s theoryof judgment did provide the service of resisting some Cartesian doctrines thatharm understanding of Moore’s paradox. One of these Cartesian doctrines isvoluntarism. Descartes maintained that you choose what to believe (B. Williams1978: 163–83). God always gives us a fair opportunity to believe just what is true.But because we are free agents, we often form beliefs without sufficient evidence.

Voluntarism is incompatible with the lessons taught by Moore’s sentence. Ifone can decide to believe then one could, through an act of will, believe that ‘It israining but I do not believe it’. Credence would be like jam that can be spread onanything at will. If there were a limit to your credence, you would have to be carefulnot to run out—lest you suffer the fate of Lewis Carroll’s incredulous friend.

Descartes also maintained that contradictions are only contingently false.God, in his Omnipotence, could make ‘It is raining and it is not raining’ true.This undermines a key contrast favored by commentators on Moore’s paradox:contradictions are necessary falsehoods while Moore’s sentences are contingent.

Moore’s early essay ‘Necessity’ shows that he also lacked the contrast prior to1900. He argues that there are no analytic truths (1900: 296). Contradictionsare not even judgments. The necessary contrast only emerges after Wittgenstein’sTractatus. In addition to containing provocative ideas about the nature of neces-sity, it also develops Arthur Schopenhauer’s key analogy between the I and the eye.

THE DISCOVERY OF THE BLIND SPOT

Johannes Kepler’s work with telescopes led him to suggest that the eye workslike a camera obscura, producing an upside-down image by means of a lens.In Dioptrique, Descartes reports an experiment that confirms this hypothesis.Descartes looked through the eyeball of an ox. Descartes further held thatperception is nothing but a movement transmitted through the nerves. EdmeMariotte reasoned that vision should therefore be strongest where the nerve wasdirectly hit by moving particles carrying the optical information. Leonardo daVinci had made the same conjecture. But Mariotte tested it. He projected animage exactly on to that spot. You can do the same with the spots below:

◦ +Cover your right eye and gaze at the plus sign. Move the page back and forth.

The dot will disappear when it falls on your blind spot.This was completely unexpected because camerae obscurae do not have blind

spots. The camera obscura does have a single hole in front where the light entersand produces an upside-down image on the background. Mariotte wonderedwhy he did not have a big black hole in his visual field. Mariotte also wonderedby no one had previously noticed this blindness in the center of the visual field.

Roy Sorensen 45

Edme Mariotte’s results were published and his experiment became popularin Paris, especially in the Academie des Sciences that he helped to found in 1666.Mariotte demonstrated his discovery in the King’s Library (the later BibliothequeNationale). Even the King of England invited him to perform his experiment.Astronomers were intrigued because the blind spot suggested a physiologicalexplanation of why they often lost sight of a star while watching another star.

There was unresolved controversy about the significance of the blind spot.Mariotte thought the blind spot showed that the retina was not the receptiveorgan of sight. Jean Pecquet maintained that the blind spot was caused by theshadow of a retinal blood vessel.

The blind spot only began to be understood with the assimilation of Kant’stheme that the mind constructs the visual scene. The blind spot is not simplyan area of non-representation. The brain actively fills in the blind spot—usuallygetting it right. If the background is red, then the spot will be filled in as red. Ifthe background is striped, then the spot will be filled in to preserve this pattern.

This filling in should be contrasted with visual effects that involve passivefailures of representation. When you look at a star, you see dim stars off to theside. But when you focus on these dim stars, they seem to disappear. The reasonis that the center of your visual field is packed with cones while rods dominatethe periphery. Rods require less light. So shifting your focus (which works wellin daylight conditions) has the effect of making the dim stars invisible.

In Color and Vision, Arthur Schopenhauer made a Kantian study of Goethe’stheory of vision. He opens ‘On seeing and color’ (1814) with the assertion thatevery perception is the product of reason. Perception is not passive absorption ofdata from the external world. Instead, the mind creates phenomena by imposing aspatial-temporal framework and dictating laws to which all objects must conformto count as objects. The mind is in charge, restlessly interpolating to get acoherent picture of reality. Instead of perceiving a black hole where there isno retinal stimulation, the active brain ‘fills in’ the gap in accordance with itssurroundings.

Schopenhauer used the blind spot to impart an ironic twist to the classicalconnection between knowledge, sight, and the self: ‘. . . the ‘‘I’’ is the dark pointin consciousness, as on the retina the exact point at which the nerve of sightenters is blind, as the brain itself is entirely without sensation, the body of thesun is dark, and the eye sees all except itself ’ (1883: iii. 282). Like the eye, the I isthe ‘center of all existence’ and yet is not present to consciousness or experience.

In the subjective sense, the ‘blind spot’ is an unreceptive portion of one’srepresentation. In the objective sense, the blind spot is the part of reality thatis not covered by this unreceptive portion. Schopenhauer takes the ambiguityfull circle by characterizing the eye itself as a missing objective element of one’srepresentation.

Schopenhauer’s self is an inaccessible entity. It is piece of metaphysics thatcaptivated the young Ludwig Wittgenstein. By the time Wittgenstein wrote the


Tractatus, he considered all metaphysics to be a linguistic illusion. However, hecontinued to be mesmerized by the analogy. To incorporate the I-eye analogy intohis logical framework, Wittgenstein interpreted the eye geometrically rather thanorganically. This is the beginning of a formal accessibility relation (so pivotal inthe development of modern modal logic). Wittgenstein’s semantic views forbidhim from agreeing with Schopenhauer that idealism is false. Instead, Wittgensteinmust conclude that idealism is nonsensical. In the Tractatus, Wittgenstein claimsthat solipsism converges with realism:

5.633 Where in the world is a metaphysical subject to be found? You will say that this isexactly like the case of the eye and the visual field. But really you do not see the eye. Andnothing in the visual field allows you to infer that it is seen by an eye.

If the self is a boundary of reality rather than an element within reality, there is aconvergence between solipsism and realism.

Wittgenstein closely associates solipsism and idealism. After all, if one drivesidealism to its logical conclusion, then to be is to be perceived by me. Theexperiences of others are only available to me as I experience them.

This perspectivalism is frequently overlooked. Peter Geach incorrectly usesWittgenstein’s model to defend the claim that a non-spatial perceiver can seeeverything:

It is perfectly easy to conceive of an observer who contemplates the whole of a spacewithout himself being located in that space; as Wittgenstein said in the Tractatus, the eyeisn’t anywhere in the visual field. We can thus imagine, so to say, an eye of God whichlooks upon all space impartially. (Geach 1979: 101)

Wittgenstein’s geometrical eye involves a point of view. It has a location. It hasan idiosyncratic range of access and a special kind of blindness. This was salientto G. E. Moore because he took notes of Wittgenstein’s 1930–33 lectures,

And he seemed to be quite definite on a point which seems to me certainly true, viz.that I might see without physical eyes, and even without having a body at all; that theconnection between seeing and physical eyes is merely a fact learnt by experience, not anecessity at all; though he also said that ‘the visual field’ has certain internal properties,such that you can describe the motion of certain things in it as motions towards oraway from ‘your eye’; but that here ‘your eye’ does not mean your physical eye, not yetanything, whatever which is in the visual field. He calls ‘your eye’, in this sense, ‘the eye ofthe visual field’, and said that the distinction between motion towards it and away fromit was ‘on the same level’ as ‘the distinction between ‘‘curved’’ and ‘‘straight’’ ’. (Moore1955: 11)

As stressed by David Sanford (1983), Wittgenstein’s point about there beingmotion toward and away from the geometrical eye puts the eye in space andconfers indexical properties.

So what stops Wittgenstein from discovering Moore’s paradox? One problemis that Wittgenstein shares his teachers’ (Russell and Moore) view that contra-dictions are not really judgments. Tautologies and contradictions are ‘senseless’.

Roy Sorensen 47

Wittgenstein also denies that contradictions can be believed. He converts theCartesian picture of what thinking ought to be into a picture of what thinkingmust be. Thinking must be clear and distinct. Thinking is a kind of picturing.Therefore, ‘It is as impossible to represent in language anything that ‘‘contradictslogic’’ as it is in geometry to represent by its coordinates a figure that contradictsthe laws of space, or to give the coordinates of a point that does not exist’(Tractatus, 3.032).

This resonates with many contemporary philosophers because they accepttruth conditional semantics. They reason: To understand a statement is tounderstand its truth-conditions. To understand a statement’s truth-conditionsis to know when it is true and when it is false. So if you understand acontradiction, you know that it is false because you realize that it is false underall conditions. Therefore, you cannot believe a contradiction. Moorean sentenceswould be nonsense insofar as they are inconsistent. They would not be consistentdescriptions that cannot be consistently believed.

Belief does not require clear understanding. Obscurity sometimes enhancescredibility. The early feminist Rachel Strachey had memorable conversationswith her uncle Bertrand Russell: ‘Uncle Bertie has been here for two nights—wehave had long and very interesting discussions on politics and Suffrage, and Ithink I agree with his views—one can’t be sure for he puts them too clearly’(Halpern 1998: 79).

Wittgenstein’s objection to the credibility of contradictions is not simply thatone would recognize their falsehood. For Wittgenstein also denied that anyonecan believe a tautology. A picture that is compatible with everything is not apicture at all. Only contingent propositions are objects of belief. Wittgensteininitially assumed that all contingent propositions are objects of belief. Later hebecame more interested in apparent exceptions.

MOORE’S LINGUISTIC TURN

Moore’s opposition to idealism led him to distinctions that were later used toexploit the Schopenhauer-Wittgenstein analogy. These are linguistic and logicaldistinctions. So the geometrical and ophthalmological character of the analogydisappear.

In ‘The Refutation of Idealism’, G. E. Moore concedes that it is difficult todistinguish between things and things as we are conscious of them. Nevertheless,he believes that the method of introspection can separate consciousness from theobject of consciousness.

. . . the moment we try to fix our attention upon consciousness and to see what, distinctly,it is, it seems to vanish: it seems as if we had before us a mere emptiness. When we tryto introspect the sensation of blue, all we can see is the blue; the other element is as if


it were diaphanous. Yet it can be distinguished if we look attentively enough, and knowthat there is something to look for. (1903: 450)

In Principia Ethica, this psychological method of drawing the distinction issupplemented with an embryonic linguistic method:

It is often pointed out that I cannot at any given moment distinguish what is true fromwhat I think so: and this is true. But though I cannot distinguish what is true from what Ithink so, I always can distinguish what I mean by saying that it is true from what I meanby saying that I think so. For I understand the meaning of the supposition that which Ithink true may nevertheless be false. (1902: 132)

Moore later incorporated this distinction between p and I think that p intohis distinction between asserting and implying. He deploys the assert-implydistinction against subjective analyses of concepts.

The maiden voyage of this distinction was in a criticism of Charles Stevenson’s(1942: 80) thesis that ‘It was right of Brutus to stab Caesar’ means ‘I now approveof Brutus’ stabbing of Caesar’. Moore agrees that the speaker implies that heapproves of the stabbing but he denies that this is what the assertion means.When Moore says ‘I went to the pictures last Tuesday’, Moore implies his beliefthat he went to the pictures last Tuesday but that is not part of the content ofthe assertion. True, there is something odd about ‘The stabbing was right but Ido not approve of it’. But this is only odd in the way the following statement isodd: ‘I went to pictures last Tuesday but I do not believe it’. The speaker maybe pragmatically contradicting himself but his assertion is not a contradiction. Ifthe speaker’s approval were part of the meaning of ‘The stabbing was right’, thenthere should be a semantic contradiction in ‘The stabbing was right but I do notbelieve it’.

Moore explains the absurdity of ‘I went to the picture show last Tuesday butI do not believe it’ as a conflict between what the speaker asserts and what heimplies. The speaker asserts that he does not believe that he went to the pictureslast Tuesday but implies that he believes that he went to the pictures last Tuesday.

Moore went on to offer a solution to the paradox. He maintains that wealmost always believe what we say—so it is statistically surprising that we arefound saying what we do not believe. This solution is generally regarded as apoor one—even as a naïve underestimate of the frequency of lying. However,one need not solve a paradox to be its discoverer. One need only provide themeans for recognizing the paradox as a paradox.

REFERENCES

Augustine (1872), City of God, trans. M. Dods (Edinburgh: T. & T. Clark).Burnyeat, M. F. (1982), ‘Idealism and Greek Philosophy: What Descartes Saw and

Berkeley Missed’, Philosophical Review, 91/1: 3–40.Emerson, Ralph Waldo (1836), Nature (New York: Scholar’s facsimiles & reprints, 1940).

Roy Sorensen 49

Fitch, Frederic (1963), ‘A Logical Analysis of Some Value Concepts’, Journal of SymbolicLogic, 28/2: 135–42.

Geach, Peter T. (1979), Truth, Love and Immortality, (Berkeley and Los Angeles:University of California Press).

Gregory, Richard (1997), Eye and Brain (5 edn., Princeton: Princeton University Press).Halpern, Barbara Strachey (1998), ‘Ray Strachey—A Memoir’, in Wayne K. Chapman

and Janet M. Manson (eds.), Women in the Milieu of Leonard and Virginia Woolf:Peace, Politics, and Education (New York: Pace University Press).

Lewis, C. I. (1946), Analysis, Knowledge, and Valuation (La Salle, Ill.: Open Court).Moore, G. E. (1900), ‘Necessity’, Mind, 9: 289–304.

(1902), Principia Ethica (New York: Prometheus, 1988).(1903), ‘The Refutation of Idealism’, Mind, 12/48 (Oct. 1903), 433–530.(1942), ‘A Reply to My Critics’, in Paul Arthur Schlipp (ed.), The Philosophy of

G. E. Moore (La Salle, Ill.: Open Court).(1955), ‘Wittgenstein’s Lectures in 1930–1933’, Mind, 64/253: 1–27.(1962), The Commonplace Book 1919–1953 (London: Allen & Unwin).

O’Connor, D. J. (1948), ‘Pragmatic Paradoxes’, Mind, 79/314: 358–9.Salerno, Joseph (2006), ‘Who Discovered Fitch’s Paradox and Why Won’t it Go Away?’,

Pacific Division of the APA.Sanford, David (1983), ‘Impartial Perception’, Philosophy, 58: 392–5.Schopenhauer, Arthur (1883), The World as Will and Idea, trans. R. B. Haldane and

J. Kemp (Routledge and Kegan Paul; 9th impression 1950).Sorensen, Roy (1988), Blindspots (Oxford: Clarendon Press).

(2003), A Brief History of the Paradox (New York: Oxford University Press).Stevenson, C. L. (1942), ‘Moore’s Arguments against Certain Forms of Ethical Natural-

ism’, in Paul Arthur Schlipp (ed.), The Philosophy of G. E. Moore (La Salle, Ill.: OpenCourt).

Stigler, Stephen (1999), Statistics on the Table (Cambridge, Mass.: Harvard UniversityPress).

Wiener, Norbert (1953), Ex-prodigy (New York: Simon and Schuster).Williams, Bernard (1978), Descartes: The Project of Pure Enquiry (New York: Penguin).Williamson, Timothy (2000), Knowledge and its Limits (Oxford: Oxford University

Press).

PART II

MOORE’S PARADOXAND KNOWLEDGE

3

Moorean Absurdity: An EpistemologicalAnalysis

Claudio de Almeida

In view of what the philosophical imagination has done with Moore’s paradox,¹there should be no doubt that Moorean Absurdity may be symptomatic of anumber of philosophical ills (that is, the kind of flaw of intellectual or moralcharacter that philosophers seek to understand). A main concern for those whohave been influential in the debate about the problem is explaining why aMoore-absurd belief is a case of irrationality. Indeed, whatever else may truly besaid about it, we will not have found a satisfactory account of Moorean Absurdityuntil we have an unobjectionable explanation of this peculiar form of doxasticirrationality. But, even though much exciting work has been done along theselines, most, if not all, of the available explanations are based on what many ofus see as unreasonable constraints on the concept of doxastic rationality. It is,however, hard to believe that anyone would object to the view that, if a Moore-absurd belief makes you incoherent, it is irrational for you to hold that belief. Inwhat follows, I both argue for the antecedent of this conditional—thus puttingforward what seems to me a satisfactory explanation of Moorean Absurdity—and try to show that it has far-reaching consequences for epistemology.

WILLIAMS’S WAY

John N. Williams (1994) made a definitive contribution to the study of MooreanAbsurdity when he noticed that those contingent propositions the assertion ofwhich G. E. Moore thought ‘absurd’ present us with two distinct threats todoxastic rationality, not just one. Consider the propositional form

I’m very grateful to Robert Audi, Rich Feldman, Nicholas Griffin, Ernie Sosa, and John Williamsfor their generous and insightful comments. I also thank the editors of this volume, Mitch Greenand John Williams, for prodding me into taking another look at Moore’s paradox.

¹ See Chapter 1, ‘Introduction’, in this volume.

54 An Epistemological Analysis

(i) p, but I don’t believe that p.

He noticed that, if you believe an instance of (i), then, on the benign assumptionthat belief distributes over conjunctions (i.e., that, if you believe that p&q, thenyou believe that p and you also believe that q), you acquire a false belief. Yourtype (i) belief cannot be true, because it will necessarily be accompanied by thebelief that p, the belief whose object is the left-hand side of your conjunctive type(i) proposition, which, in turn, falsifies the right-hand side of the conjunction,thus rendering your instance of (i) false. So, the proposition expressed by aninstance of (i) can only be the object of a false belief.

Now, as Williams notes, a belief the object of which is an instance of

(ii) p, but I believe that not-p

will ensure that you hold a false belief in an entirely different way. In this case,belief distribution assures us that you believe that p, and the truth of your type(ii) belief assures us that you believe that not-p (since the right-hand side of theconjunction will have to be true). Thus, when you believe an instance of (ii),either your conjunctive belief is turned false by the falsity of the conjunct on itsright-hand side or it is true and you are inconsistent (for believing both p andnot-p). (Of course, you may still be inconsistent when it’s false—if what makesit false is the falsity of its left-hand side; but this is theoretically irrelevant.) So,whenever you believe a true instance of (ii), you are inconsistent.

This is a potentially important difference for the explanation of MooreanAbsurdity: Self-refutation—in the case of (i)—and inconsistency—in the caseof (ii)—may involve significantly different failures of doxastic rationality, and itdoes seem that any account of the problem which ignores such specificity mustbe doomed.

But Williams hasn’t just claimed that any acceptable analysis of MooreanAbsurdity must be sensitive to that seemingly important difference between (i) and(ii). He has gone as far as to suggest that the mere recognition of the phenomenaexemplified by (i) and (ii) will yield a fully satisfactory resolution of Moore’sparadox—i.e., a satisfactory explanation of Moorean Absurdity—if we only doso much as to associate the above arguments about (i) and (ii) to the followingclaim (which is very clearly implied by what he writes but not his verbatim):

(W) You rationally believe that p only if there is no simple and compellingargument that we can reasonably expect you to be aware of to the effectthat your believing that p is either self-falsifying or ensures the presence ofcontradictory beliefs in your doxastic system.²

² (W) is implied by the characterization of Williams’s proposal, ibid. I thank Williams forshowing me that I unfairly claimed, in my (2001), that he was committed to the view that aMoore-absurd belief ensures the inclusion of a false belief in the believer’s doxastic system in his(1994), even if the misrepresentation was indeed induced by an incautious passage of his (1994). Inhis (1996), however, Williams clearly identifies the belief giving rise to the Preface paradox as one

Claudio de Almeida 55

According to Williams’s original proposal (1994), ‘a Moorean believer is irrationaleither because his belief cannot be correct [as in (i)], or because its correctnessentails contradictory beliefs [as in (ii)]’. I then (in my 2001) took him to taskfor apparently assuming that self-refutation of which the believer is not awarewould suffice to account for the irrationality of a type (i) belief. And he now (inChapter 1 this volume) sees fit to make the following important addition to hisaccount of the irrationality of both a type (i) and a type (ii) belief: ‘Moreover,I should be able to see this fact [i.e., self-refutation if I have a type (i) belief, orinconsistency if my type (ii) belief is true] with a minimum of reflection.’ Bewareof misinterpreting ‘should be able’ here. Surely, the mere ability of the believer topresent himself with his anti-(i) and anti-(ii) arguments is not what Williams’srevised account requires: The original proposal was obviously not aimed atexplaining the irrationality of Moore-absurd beliefs only when these are heldby those whose intellectual resources make them incapable of constructing therelevant arguments. It obviously was intended as an explanation of the irrationalityof every Moore-absurd belief, including that which is held by the believer whois very capable of putting those arguments to himself in soliloquy. (Otherwise, ifWilliams, himself, were to hold a Moore-absurd belief and not notice it, he wouldnot, by his own account, be irrational in so doing!) So, the revised proposal mustrequire that those anti-(i) and anti-(ii) arguments actually be evident (on ‘minimalreflection’, whatever that means) to the rational believer. Whence (W)—in whichthe simplicity requirement does duty for the new minimal-reflection desideratum.

I think it should be granted that an explanation of Moorean Absurdity basedon both (W) and those anti-(i) and anti-(ii) arguments does have its charms:Self-refutation and inconsistency are obviously anathema to the epistemic goalof believing truths exclusively—the goal to which doxastic rationality is, bymost accounts, the (admittedly fallible) means. And we shouldn’t in any waybelittle that crucial step forward in the revised account: If I assert—as an honestexpression of belief—‘I never use a preposition to end a sentence with,’³ I may bestupid, all right—or an airhead. But irrational (for holding an irrational belief)?Consider (perhaps more to the point) a belief of ‘I never have any introspectivebeliefs’—which is self-falsifying when resulting from introspection (rather than,say, from the expert testimony of your psychoanalyst). Again, is the charge ofirrationality warranted? It’s hard to see that it is, even if we are unable to tell,in every case, whether a particular instance of stupidity is a clear-cut case of

that may be rationally believed. My (W) partly results from his complaint. A critical remark fromRobert Audi has also contributed to my formulation of (W).

³ This is my version of an example from Raymond Smullyan (1983: 20). Smullyan’s own versionis ‘A preposition must never be used to end a sentence with’, which is not really ‘self-annihilating’,as he calls it, if ‘self-annihilation’ somehow implies irrationality, since conscious wrong-doing isnot necessarily irrational. At best, his example is a case of ‘do as I say, not as I do’, which doesnot necessarily involve doxastic irrationality. At worst, the speaker does not have the concept ofa preposition, and so doesn’t even understand what he himself asserts, which is philosophicallyuninteresting.


irrationality. Now, if I were to see that I refute myself but were still somehowunable to drop the offending belief, I would unquestionably exemplify what wecall ‘irrationally held belief ’. So, (W) is definitely a step forward for Williams.

There are, however, what I consider good enough reasons to reject his proposedexplanation of Moorean Absurdity.

First, it is unrealistic to assume that those anti-(i or ii) arguments are anywhereclose to being simple enough for even the most demanding, but still fairly plausible,internalistic requirements for doxastic rationality. Notice that, according to (W),the believer is not just required to repudiate a Moore-absurd proposition oninstinct (out of epistemic virtue, one might say); he would have to put togetheran argument to the effect that he ought not to hold a Moore-absurd belief—anargument that will have to include premises about belief distribution andtruth-conditions (no matter how ordinary, how non-philosophical the requisiteconcepts might be allowed to be)! It should, again, be uncontroversial that youravailing yourself of those anti-(i or ii) arguments is a sufficient condition of yourbeing irrational, if you, at the same time, persist in holding a Moore-absurdbelief. But it should be clear that that cannot be a necessary condition of doxasticrationality, as required by (W).⁴

Second, Williams’s explanation is clearly not complete. Granted, I am not ina position to promise a completeness argument here. Moorean Absurdity hasdefied both syntactic and semantic criteria. And the epistemological analysis putforward in Section 3 below will not include a completeness argument. Still, it is,by now, unacceptable for us to rest content with confining our attention to thehistorical forms of Moorean Absurdity, (i) and (ii).

Consider

(iii) p, but it’s not rational for me to believe that p

and

(iv) p, but it’s rational for me to believe that not-p.

Notice that both (iii) and (iv) intuitively seem to be no less cases of MooreanAbsurdity than (i) or (ii). By any standards, if you believe any instance of either(iii) or (iv), you very plainly seem to be doing what, by your own lights, youshouldn’t be doing.⁵ But this is not just any old conscious wrong-doing: Giventhat what a person believes to be rationally incredible is normally not believed by

⁴ I’ve seen a number of very intelligent people react to the well-known Moore-absurd propositionswith a shrug of their shoulders. Catherine Elgin publicly stated, at the 2003 edition of the RutgersEpistemology Conference, that she doesn’t see what all the fuss concerning Moore’s paradox isabout. And I bet it is safe to infer that she hasn’t put those anti-(i or ii) arguments to herself, eventhough it’s quite obvious that she is capable of doing so. And anyone who does so should have nodifficulty feeling the sting of the paradox. That’s Williams’s permanent stamp on the problem.

⁵ In saying that you’re ‘doing’ something by holding a belief, I’m definitely not committed toany controversial views of ‘doxastic agency’. I might as well have said that something untoward ishappening to you.


her, we get the feeling that both a type (iii) and a type (iv) belief are cases ofsignificant cognitive disorder.

Now, notice how non-conjunctively embedded occurrences of both (iii) and(iv) (for instance, ‘If it rained yesterday but it’s not rational for me to believethat it did, I’m being misled by that generally reliable weather forecast’), orhow their instances with a past-tense (or future-tense) main verb in the secondconjunct (‘It rained yesterday, but it was not rational for me to believe that it did,because I had reliable indication to the contrary’), manage to sidestep any senseof absurdity just like (i) and (ii) do when subjected to the embedding/past-tensetest (or ‘EP test’, for short). But the (W) combo is powerless to account for theseseemingly obvious cases of Moorean Absurdity: Neither a type (iii) nor a type(iv) belief will ensure the presence of a falsehood in your belief system (by eitherself-refutation or possession of contradictory beliefs).

If our EP test is a sure-fire indication of Moorean Absurdity, we can safely gofurther in pursuit of completeness. Notice how the stultifying

(v) I have no beliefs

and

(vi) I don’t exist,

two very clear cases of self-refutation (and thus well within the scope of Williams’sdiagnosis), come out Moore-absurd by our seemingly infallible test: While it isobvious that there would be something very wrong with believing either (v) or(vi), nothing seems prima facie problematic with any of the following: ‘If I haveno beliefs, I’m either a vegetable or a radical skeptic, according to our philosophyteacher’, ‘I had no beliefs while they meddled with my brain at the neurosciencelab’, ‘If I don’t exist, you can’t expect me to help you, dear’, ‘I didn’t exist whenWWII ended’.

Our EP test does seem to be a sure-fire way of unmasking Moorean Absurdity:If a proposition is somehow absurd, odd, paradoxical, our test seems to revealwhether it is Moore-absurd. A liar-type proposition will, by our test, fail toexhibit Moorean Absurdity: Notice, for instance, how odd ‘This sentence wasnot true yesterday’ is. It may not even be meaningful! Likewise, a sentence abouta Russellian class such as ‘The class of all classes which are not members ofthemselves both was and was not a member of itself in 1901’ is, if anything, nowdoubly odd!⁶

Whether or not the EP test proves to be the key to a completeness argumentregarding Moorean Absurdity, I’m here making only the more limited claim that

⁶ In latest correspondence, Williams lets me know that he thinks ‘p but I irrationally believe thatp’ is a counterexample to the EP test, since he sees it as absurd but not Moore-absurd, and yet itclearly passes the EP test. We disagree about this: I think it is Moore-absurd. But a discussion ofour disagreement in this regard—and of why the Moorean Absurdity of that proposition can beexplained along the lines of the analysis put forward in Section 3 below—would far exceed theavailable space here.


it is not open to Williams to deny that (iii) and (iv) are clear cases of MooreanAbsurdity. Thus, the upshot of the foregoing is that we have more than one goodreason to think that Williams has not given us a satisfactory explanation of thephenomenon.⁷

THE WAYS OF INCOHERENCE

The key to understanding the nature of Moorean Absurdity is incoherence, Iwill be arguing. And I intend to exploit the fact that we seem to have a veryrobust intuitive grasp of the kind of cognitive malfunction epistemologists call‘incoherence’ in what follows, with technicalities kept to a bare minimum.

Thus, consider this scenario: Suppose you believe that (p) the butler is themurderer. And suppose your belief that p is based (both causally and evidentially)on your beliefs according to which (q) the best criminal science shows that, in98.7 per cent of the murder cases in which the butler has both motivation andopportunity, he is the murderer, and (r), in the case at hand, the butler did havethe motivation and the opportunity to do the deed. But further assume that youalso believe that (s) there is an eyewitness who places the butler several hundredmiles away from the scene of the crime on that fateful night. If p, q, r, and sare all of your crime-related beliefs, epistemologists will want to say that youare incoherent for harboring beliefs which are counterevidence for one another,namely p and s.⁸ We intuitively censure you—or pity you—for failing to noticethat you are not entitled to believe both that p and that s. And we may want tomake the point by saying that you do not rationally believe that p if you, at thesame time, also believe that s (and vice versa).

So, incoherence is what results from the inclusion of effective counterevidencefor a given proposition in your belief system. In our test case, s is effectivecounterevidence for believing that p just because, by hypothesis, there is nothingin your belief system that neutralizes the overriding effect of s (with a view tobelieving that p rationally).⁹ We should then say that, while holding one of thosebeliefs, the other is incoherent with your belief system. (Clearly, if, in additionto believing p, q, r, and s, you were also to hold the belief that t, say, that theeyewitness is married to the butler, incoherence would, again, be eliminated,because s would cease to be effective as counterevidence for believing that p.)

⁷ And I still hold, as I did in my (2001), that Williams’s is the analysis to beat when it comes toexplaining Moorean Absurdity.

⁸ Important discussions of the concept of coherence can be found in Laurence BonJour (1985)and Keith Lehrer (2000). I discuss the concept in my (2003). Unless you’re a coherentist, you’llprobably agree that incoherence is not the complement of coherence. A belief can be neithercoherent nor incoherent with a given belief system. See Robert Audi (2003: ch. 7).

⁹ Like Peter Klein in his (1981), I wouldn’t require that only justified (or ‘rationally held’) beliefsbe allowed to have an effective overriding role.


For our purposes here, it doesn’t really matter how one may want to account forwhat makes a proposition counterevidence for some other proposition for a givenagent. Such an account may require a full-fledged explanation of what makes aproposition good evidence for having another proposition as the content of a belief(for a given agent), and we certainly don’t need any such detailed explanation tounderstand what makes it irrational for you to hold the conflicting beliefs in ourbutler case. For our purposes, I will just note that it seems entirely satisfactory,on purely intuitive grounds, to say that what accounts for the way we react to thebutler case is that we take this world of ours to be such that, when p is true, s tendsto be false, and vice versa. We may want to say that p probabilifies not-s and that sprobabilifies not-p. And that is just about all we need in order to account for theway we intuitively react to the butler case—except, perhaps, that we may think itimportant to add that the irrational believer shares in our belief that p probabilifiesnot-s and that s probabilifies not-p. (It should suffice for him to be disposed tohold a belief the content of which is logically equivalent to our way of puttingit, thus relieving us of any de dicto commitments here.) It seems natural to thinkthat it is epistemic probability that matters to our understanding of incoherence.(Surely, if the would-be irrational believer were at least disposed to believe that theexistence of an eyewitness for the defense had no bearing whatsoever on whetherthe butler did the deed, a judgment of irrationality would seem ill-advised:Strange as it would seem to us, the belief system in question would not, however,display the kind of disarray that characterizes incoherence as we intuitively thinkof it. It may be plagued by ignorance or some other form of intellectual disorder,but no conflict among beliefs of the kind we call ‘incoherence’.) But we certainlyneed go no farther in accounting for our intuitive grasp of incoherence.

The butler case is a paradigmatic case of incoherence, and the way we intuitivelyreact to it makes it clear that, for us, the belief that p is rationally held by anagent S at a given moment t only if p is not incoherent with S’s belief system att—a claim that is honored by every epistemology with a fighting chance.¹⁰ Butincoherence has at least two other philosophically important manifestations—orso I will maintain here.

It is, of course, platitudinous to hold that, if you believe a self-contradictoryproposition, you are incoherent, since you are inconsistent, and it is customarilythought that inconsistency implies incoherence. But it gets surprisingly difficultto find a satisfactory explanation of why you are supposed to be incoherent

¹⁰ See Michael Bergmann (1997). I trust that, when fully analyzed, what he simply calls a ‘no-defeater condition’ would be seen to be a fairly complex anti-incoherence conceptual mechanism apartial analysis of which is given by claims (a), (c), and (e) in Section 3 below. On the need for a ‘no-defeater condition’ and on how it affects our understanding of the internalism/externalism debate,also see Thomas Senor (1996). As a general label, Alvin Goldman’s (1998) ‘non-underminingcondition’ is a better terminological choice. A full-fledged defeasibility theory such as Klein’s (1981)shows that we need a distinction between ‘defeaters’ and ‘overriders’ (or, at any rate, some moreterminology for an important conceptual distinction), rather than just the generic term ‘defeater’.


whenever you are inconsistent. After all, what is inconsistency? At a minimum, aset of propositions is inconsistent if it is logically impossible for every member ofthe set to be true. But it is patently clear that you may rationally believe certainnecessary falsehoods. Suppose that very reliable logician claims to have derived pfrom certain set-theoretic axioms. But, uncharacteristically, he made a mistake,and p is actually a non-obvious necessary falsehood. If you take his word for itand believe that p, we should all agree that you may rationally do so. Thus, wehave to make finer distinctions when considering the notion of inconsistency andthe way it relates to incoherence. It is just not reasonable to hold that every caseof inconsistency is a case of incoherence—for not every case of inconsistency isa case of irrationality, and yet, as universally granted, every case of incoherencemust be a case of irrationality.

The need for a distinction between weak and strong forms of inconsistencyhas been felt elsewhere in the epistemological literature.¹¹ And I, too, will availmyself of such a useful distinction here. However, for reasons that I don’t havethe space to discuss in this paper, I had better steer clear of the distinction as putforward elsewhere and draw it anew for our purposes.¹²

A non-ad hoc distinction of strong and weak inconsistency naturally arises, Isubmit, from attending to the intuitive appeal of a crucial difference: We shouldnotice how much worse off you are, from an epistemic point of view, whenyou have a self-contradictory belief, i.e., when you believe a truth-functionalfalsehood, as opposed to believing necessary falsehoods of other kinds. As claimedabove, there are countless occasions on which you seem to be perfectly rationalin believing certain necessary falsehoods. On all those occasions, not only do younot seem to be plagued by the kind of ‘conflict’ among your beliefs that is essentialto incoherence, the only apparent damage to your epistemic health is that youhold a false belief. That’s a very bad thing, all right. But it is no worse than youreveryday failure to connect with the empirical world by holding a thoroughlyaccurate representation of that slice of it that you hope to have hooked up with(if you have indeed so failed). On the other hand, if you believe a contradiction,even (non-ad hoc) relevance restrictions on logical implication can’t possiblykeep you from holding a belief that gives you an objectively good, entailing groundfor disbelieving that very same contradiction that you believe (as we shall shortlysee). Which is why I think we should all agree that one may rationally believenecessary falsehoods only if those rationally believed necessary falsehoods do notinclude truth-functional falsehoods.¹³ Call this italicized claim the ‘No-RAtional-Belief-In-Explicit-Self-contradictions’ claim—or ‘No-RABIES’, for short. Andthe believer who, by means of Reprehensible Acquisition of Beliefs Instantiating

¹¹ See Klein (1985) and Roy Sorensen (1988). A motivation for the distinction is alreadyrecognized in Richard Foley (1979).

¹² I discuss the distinction as put forward in Klein (1985) in a forthcoming version of my (2003).¹³ 13 If there is a good argument in the epistemological literature to the effect that one may

never rationally believe a non-truth-functional necessary falsehood, I have failed to find it.


Dysfunction as characterized by No-RABIES, believes contradictions shouldproperly be called ‘the RABID believer’. So, you’re RABID if and only if youbelieve a truth-functional falsehood.

There are two routes we can follow in order to be entitled to No-RABIES—thatis to say, two routes to concluding that a self-contradictory belief, unlike othertypes of necessary falsehood, promotes the kind of disarray in your doxastic systemintuitively associated with the term ‘incoherence’: First, having accepted thatbelief distributes over conjunctions, we should notice that every truth-functionalfalsehood is logically equivalent to—and so logically implies—a proposition ofthe form p&∼p. And every belief of such a form is accompanied by contradictorybeliefs. Now, assume that, if you are rational in believing a given truth-functionalfalsehood and rationality is closed under logical implication, then you’re rationalin believing that p&∼p—which you can’t be, since that would saddle you withboth the belief that p and the belief that not-p and would make you obviously‘incoherent’. From this point on, if you stick with the closure principle, ModusTollens, De Morgan, Double Negation, and Disjunctive Syllogism give us oneway of being assured that you’re irrational in believing a proposition which islogically equivalent to p&∼p.¹⁴ Second, a contradiction logically implies its ownnegation. Thus, if, again, you stick with the closure principle, and if you can’trationally believe both p and not-p, because that offends against your most deeplyheld intuitions about what it is to be ‘incoherent’, you’ll find yourself led back tothe conclusion that you can’t rationally believe a contradiction.

Now, what would it take for you to be in a position to reasonably denythat a contradiction implies its own negation, or that every truth-functionalfalsehood logically implies a proposition of the form p&∼p? Certainly, somereasonably held objection that would block the inferential pathways leading tothose conclusions. And that means that, unless you are armed with the kind ofobjection that, say, relevant logicians or anti-realists would try to use to blockthose inferential pathways (in which case you know that you have your work cutout for you), you will lack the motivation to complain about the inference rulesunderpinning those pathways, you will be, that is, like the vast majority of thefolk. And the vast majority of the folk would surely follow us as we derived thatconclusion about the irrationality of RABID believing.¹⁵

By contrast, no such problem—no such commitment to both p and not-p—can reasonably be derived from a belief in a non-truth-functional necessaryfalsehood. Granted, the classical theory of validity will have us believe that any

¹⁴ That is, take the following as your premises: ‘If I’m rational in believing that p [where ‘‘p’’stands for any truth-functional falsehood] and p implies p&∼p, then I’m rational in believing thatp&∼p’, ‘p does imply p&∼p’, and ‘I’m not rational in believing that p&∼p’. Now, apply thoserules to derive ‘I’m not rational in believing that p’.

¹⁵ There is, of course, the heroic alternative of rejecting every known version of the closureprinciple for doxastic rationality (or ‘epistemic justification’). See the discussion of claim (b) inSection 3, below.


necessary falsehood logically implies anything whatever, and may thus seemto offer a basis for attacking No-RABIES. But that bizarre consequence is awell-known problem for the classical theory, not for us: It is notoriously absurdto hold that, for instance, 2+2=5 logically implies that Mars is a planet justbecause it is impossible for the former to be true and logical implication ensures thatit is impossible for the implied proposition to be false if the implying propositionis true. (And, of course, the absurdity does not hinge on the classical disregardfor relevance, as it is equally absurd to hold that 2+2=5 implies that 2+2 doesnot, after all, equal 5 just because it is impossible for the former to be true, orjust because it is impossible for the latter to be false—take your pick!—and logicalimplication ensures . . .)¹⁶ So, invoking the classical theory of validity (or of logicalimplication) is a bad move here: Holding on to a notoriously oversimplifiedexplanation of why certain rules of inference appeal to us (or should appeal tous, if we prize what is undeniably good about validity as classically understood,namely, truth-preservation) would hardly give you a good reason for attacking adistinction that is based on the very simple fact that, for the vast majority of us,there are apparently unimpeachable arguments leading from a contradiction toits negation, just as there are equally unimpeachable arguments showing that anytruth-functional falsehood is logically equivalent to a proposition of the formp&∼p. And No-RABIES depends only on the very evident fact that no such claimcan reasonably be made for non-truth-functional necessary falsehoods—that is,no such claim can be made for non-truth-functional necessary falsehoods withoutsqueezing a funny, notoriously perverse argument out of the classical explanationof what it is for an argument to be valid (or for one proposition to be logicallyimplied by another).

If we’re out of the woods with No-RABIES, we have it as a conceptual basisfor drawing a useful distinction of weak and strong inconsistency. For we cannow define ‘strong inconsistency’ as follows:

An agent S is strongly inconsistent at a moment t iff, at t, either there’sa non-empty set of contingent propositions in S’s doxastic system whichlogically implies the negation of some proposition in the system or S isRABID.¹⁷

¹⁶ For a brief but particularly useful discussion of the problem, see Stephen Read (1995: ch. 2,esp. p. 57–9), and notice that his rejoinder on behalf of the classicist won’t work for non-truth-functional necessary falsehoods. See also the discussion in David Sanford (1989: ch. 15), where, asI understand it, it is suggested that one may object to the classical explanation of validity withoutobjecting to any classically valid rule (though Sanford himself does object to certain classically validrules)—because you may not be able to accept certain judgments of invalidity. This is, of course, ahuge topic, but I trust we’ve seen enough for our purposes here.

¹⁷ The reader with the Preface on his mind may have noticed that my definition of ‘stronginconsistency’ implies that the prefatory belief makes the believer strongly inconsistent. Indeed, I’mprepared to argue that that’s the case, and that the weak/strong inconsistency is no help in solvingthat paradox. See my (forthcoming), where a proposed solution of the Preface is put forward.


Note that, by the above definition, you’ll be strongly inconsistent if yourdoxastic system contains either contradictories or logical contraries (when anexistential assumption is in effect) or truth-functional falsehoods, as expected.And we can now use that definition to define ‘weak inconsistency’ as follows:

S is weakly inconsistent at t iff, at t, it is logically impossible for everyproposition in S’s doxastic system to be true, but S is not stronglyinconsistent.

I submit that we can now produce a perfectly general explanation of the kindof ‘conflict’ among beliefs that intuitively characterizes incoherence, one thatmay be absent when you believe a non-truth-functional necessary falsehood butis clearly present when you believe a contradiction of the form p&∼p, just asit is when you believe contradictories, or when you believe logical contraries, orwhen you are the obviously irrational believer in our butler case:

You are incoherent at t iff, at t, your belief system probabilifies the negationof some proposition in the system, and, if you are inconsistent, you’restrongly inconsistent.¹⁸

We assume, of course, that logical implication is a limiting case of probabilific-ation. Thus, the above definition of ‘incoherence’ seems to cover every intuitivelyidentifiable case of ‘conflict ’-irrationality, as we may put it.¹⁹ And it does sofrom a thoroughly objectivist point of view, a point of view according to whichone may be incoherent without knowing, by a combination of introspectionand reasoning alone, that one is, a point of view, that is, according to whichincoherence need not be transparent to the reflective mind just by introspectionand the exercise of reasoning.

We have, therefore, reached an explanation of incoherence according towhich, when you’re incoherent, either you are like the irrational believer in ourbutler case or you’re a RABID believer. Now, this is all leading to the claimthat the Moore-absurd believer belongs in that very same class of incoherentbelievers, which would explain why a Moore-absurd believer is irrational. If we’resuccessful in explaining Moorean Absurdity along those lines, our explanation

¹⁸ It should be noted that, like most people who have written about ‘coherence’ and ‘incoherence’,I take those terms to be defined for propositions (and, derivatively, for beliefs) exclusively. Thus, ifyour favorite epistemology tells you that some non-propositional element is a justifier of p for you,it would be inappropriate to say that such an element is in some coherence-making relation withp for you. So, for the non-coherentist, it is entirely possible for p to be coherent with your beliefsystem while it is not justified for you—because your justification for believing that p is defeated bysome non-propositional element which probabilifies not-p for you. Correspondence with RichardFeldman has shown the need for the clarification.

¹⁹ Well, not quite every intuitively identifiable case; just every case that is relevant for ourpurposes here. If we were building a complete catalog of intuitive cases of incoherence, we’d haveto take into account some cases that do not depend only on what goes on in your belief system ata given moment. Suppose you suspend judgment as to whether p, and suppose that p probabilifiesnot-q. If, while the withholding is in effect, you believe that q, you clearly are incoherent, even ifnothing in your belief system probabilifies not-q.


will satisfy what has seemed to many a necessary condition for a tenableexplanation of the problem: an account of the contradictory-like quality ofMoorean Absurdity. (Whether or not you agree that explaining why the Moore-absurd believer seems to contradict himself in holding a Moore-absurd beliefwithout actually contradicting himself should be a desideratum of a satisfactoryaccount of Moorean Absurdity, it should be obvious that its invulnerabilityto attacks on that front will be a welcome property of my analysis of theproblem.)

MOOREAN ABSURDITY EXPLAINED

So, here’s my main claim about Moorean Absurdity: A Moore-absurd belief makesyou incoherent. My argument for that claim will require the acceptance of thefollowing six theses (a–f ):

(a) If it’s rational for me to believe that p at a moment t, then, at t, it’s notrational for me to hold any belief which is effective (undefeated) counterevidencefor the belief that p (in light of all else that I believe).

That’s the theoretical claim underlying any minimally plausible explanationof how we intuitively react to the butler case: If both p and q are in yourbelief system and q probabilifies not-p—however slightly—without having itsadverse effect on the epistemic status of your belief that p itself defeated byany other belief in your doxastic system, there should be no doubt that you areirrational (since you are incoherent). It is crucial to note that even a very slightprobabilification of not-p by your belief system would make you incoherent inbelieving that p. (Again, it may be entirely rational for you to believe that p evenif your belief system includes counterevidence for believing that p, but only if itis rational for you to believe that the preponderance of the evidence tips the scalein favor of believing that p, in which case that particular piece of counterevidencewill have been rendered ineffective.)

There should be no doubt that we have the relevant intuitions on our sidein this regard. Thus, suppose you are the detective who’s mounted what lookslike an ‘air-tight’, but probabilistic, case against the butler. (You may want toadd that you have evidence of your reliability, independent confirmation fromsources that you take to be reliable, etc.) Now, consider what you should dowhen you receive an anonymous letter according to which, if you call a certainnumber, the stranger taking the call just may decide to give you an intriguingpiece of evidence which just may lead to the exoneration of the butler. By anystandards, you have now acquired very little evidence that the butler is innocentafter all. And yet, shouldn’t you make that call? Would it still be rational for youto keep the belief that the butler is the murderer before making the call (if youhave those beliefs about the possible outcome of the call)? Intuition clearly seems


to call for the silencing of the ‘objection’ (as Keith Lehrer would call it²⁰) posedby that letter. No matter what the moral implications of your actions would be,you just wouldn’t seem to be an epistemically virtuous agent (whatever exactly thatmay mean) if you simply ignored that anonymous letter. The rationality of yourbelieving that the butler did it would clearly be at stake. Probabilification andrationality come in degrees; incoherence (‘conflict’-irrationality) doesn’t: Just anydegree of (undefeated) probabilification of not-p will make it irrational for youto believe that p. It doesn’t take much at all to poison the epistemic well.²¹

(b) If it’s rational for me to believe that p, and p logically implies q, then it isrational for me to believe that q.

Even though a version of (b)—a version of the well-known ‘closure prin-ciple’—was received as news when Edmund Gettier’s seminal paper waspublished, something like (b) has been tacitly assumed ever since the concept oflogical implication seemed important, at the dawn of logic. The above version ofthe principle (with ‘justification’ replacing ‘rationality’) is competently defendedagainst influential criticism in Peter Klein (1995 and 2002).²²

(c) It’s rational for me to believe that p only if I don’t either believe that it’sirrational for me to believe that p or believe that it’s rational for me to believethat not-p.

Something like (c) has been defended by some of our most influentialepistemologists, an assortment of internalists and externalists: All of Keith Lehrer(2000), Richard Foley (1993, 2001), and Ernest Sosa have been explicit aboutthe need to purge one’s mental life of elements that may be used in an exerciseof self-criticism about the epistemic status of one’s beliefs. (Sosa calls it the needfor ‘broad coherence’ in his 1997.)²³

It is noteworthy that, if rationality distributes over conjunctions, an obviousconsequence of (b), one may not consistently both hold that (iii) and (iv) areMoore-absurd and deny (c), or suspend judgment about it—for, if you believe

²⁰ See his (2000: ch. 6).²¹ In saying that incoherence is not a gradable state, I make a fairly unusual claim—but not an

entirely polemical one. Having noted that coherence comes in degrees, Feldman (2003: 64), forinstance, typically says that, if you include both p and not-p in your belief system, you make thesystem ‘less coherent’. And he goes on to add that ‘inconsistency detracts from coherence’. I wouldhave said that inconsistency precludes it altogether, even if it is a fact that, from an epistemic pointof view, having, say, two pairs of contradictories in your belief system is worse than having justone, and having three pairs is worse than having two, and so on. But the discrepancy seems largelyverbal to me. The important point, which is highlighted by my way of speaking about the matter,is that any degree of incoherence is an epistemic disadvantage, and, while no tenable epistemologywould require entailing evidence—or perfect reliability—every tenable epistemology (every tenablenormative epistemology, that is) must condemn incoherence, no matter how minimal.

²² But see my exchange with Jonathan Kvanvig under his post ‘Klein on Useful Falsehoods’ (filedunder ‘knowledge’) at the web forum Certain Doubts (http://fleetwood.baylor.edu/certain doubts/index.php?p=161) for criticism of Klein’s restrictions on the closure principle.

²³ I argue for the recognition of the epistemic importance of immunity to self-criticism along thelines of (c) in my (2000).

http://fleetwood.baylor.edu/certain_doubts/index.php?p=161

http://fleetwood.baylor.edu/certain_doubts/index.php?p=161


that it may be rational for you to believe that p even when you believe that it’sirrational for you to do so, or when you believe that it’s rational for you to believethat not-p, you clearly would have no motivation to regard (iii) and (iv) as Moore-absurd. This is not, however, my explanation of why (iii) and (iv) are Moore-absurd. You may reasonably accept both (b) and (c) but fail to see that (iii) or (iv)is ‘absurd’. But I work here on the assumption that a good enough explanationof Moorean Absurdity will show you that each of (i)–(vi) is Moore-absurd (i.e.,essentially, it is such that, despite being contingent, it can never be rationallybelieved) and does not, therefore, depend on the prior perception that each ofthose is somehow ‘absurd’ (widespread and expected as that prior perception maybe). So, I still owe you an explanation of the Moorean Absurdity of (iii) and (iv).

(d) If I believe that p at t, then, at t, it is (prima facie) rational for me to believethat I believe that p.

I wouldn’t dare pretend that (d) is irresistible. It has been vigorously resisted bythe enemies of foundationalism, as we know.²⁴ (d) is a foundationalist claim, butit is not all there is to a foundationalist theory of doxastic rationality, of course. Itmay be embraced by both internalist and externalist varieties of foundationalism,and by the gamut of competing factions in each camp. It is, therefore, a mainstayof contemporary epistemology.²⁵

(e) If I believe that I believe that p at t, then it’s not rational for me to believethat not-p at t (and, if I believe that I believe that not-p at t, then it’s notrational for me to believe that p at t).

I don’t expect (e) to be self-evident. So, here’s my argument for it:

(e1) If the conditional (e) is false, then you do believe that you believe that pand it may be both rational for you to retain the belief that you believe thatp and also (simultaneously) rational for you to add the belief that not-p inyour belief system.

(e2) If the second conjunct of the consequent of (e1) is true, you may rationallybelieve that p is in your belief system while rationally adding not-p to thesystem.

(e3) Our most basic assumption about rationality here is that you can’t rationallybelieve that p if p is incoherent with your belief system—that is, if your

²⁴ I’m afraid I don’t quite understand what Ernest Sosa (2003: 111, 116) calls ‘propositionally-contentful experiences’, something that, supposedly, a coherentist might be able to appeal to inorder to assimilate a foundationalist element into his epistemology. In any case, it is suggestedby Sosa that some such form of a hybrid coherentism (or a coherentist foundationalism) is notinconceivable. Pending clarification of Sosa’s suggestion, I take (d) to be a hopelessly foundationalistclaim in the exclusive, anti-coherentist sense of ‘foundationalist’. For helpful discussion, see Byrne(2005).

²⁵ In correspondence, an influential contemporary foundationalist, Audi, suggests that (d) wouldnaturally be embraced by foundationalists. But he also believes that a form of foundationalism thatis not committed to (d) is conceivable.


belief system contains effective evidence for not-p (i.e., anything whichprobabilifies not-p, however slightly, without that evidential support beingoverridden).²⁶ And q can be evidence for p only if the presumed truth of qprobabilifies p (on any conception of probability).

(e4) In view of (e3), if the consequent of (e2) is true, you may truly believe thatp is in your belief system while rationally adding not-p to the system (i.e.,the truth of what you already believe, ‘I believe that p’, does not probabilifythe negation of what you add to the system, not-p).

(e5) If the consequent of (e4) is true, you may rationally add not-p to yoursystem while p is in the system.

(e6) But the consequent of (e5) is clearly false, since, if it were true, you’drationally believe that not-p while becoming strongly inconsistent in sobelieving, which we have assumed is impossible.

(e7) So, by successive applications of Modus Tollens, we see that the antecedentof (e1) is false.

If you agree that strong inconsistency is incompatible with rational believing,(e) should now seem entirely unproblematic to you. Any tenable theory ofdoxastic rationality must imply (e).²⁷

(f ) If I believe that p&q, then I believe that p and I believe that q.

The unproblematic (f) completes our set of premises. We now need nomore than classical propositional logic to see that, even though a Moore-absurdproposition is contingent, Moore-absurd believing is like RABID believing inthat it is a non-obvious case of conflict-irrationality, the kind of irrationality weshould identify with incoherence.

In what follows, ‘Bp’ abbreviates ‘I believe that p’, and ‘Rp’ abbreviates ‘It is(prima facie) rational for me to believe that p’. Thus, ‘RBp’ should be read as ‘Itis (prima facie) rational for me to believe that I believe that p’ and ‘Bp&Rp’ willstand for ‘I (prima facie) rationally believe that p’.²⁸ Given that a belief is ultimafacie rational if and only if it is prima facie rational and there is no effectivecounterevidence for it in the agent’s mental life, the following arguments showthat Moore-absurd believing is a case of irrationality without qualification.

²⁶ Here, we disregard the question of whether the beliefs you already have are justified and askourselves what constraints the rational acquisition of new beliefs is under once you believe what youalready do (perhaps unjustifiedly). We want to know what kind of problem incoherence is, how,just by itself, it makes rational believing impossible.

²⁷ Persistent pounding at (e) by Rich Feldman made me see that the argument I had for (e) ina draft of this paper was unacceptably compressed. Similar objections were made by John Williamsand Robert Audi.

²⁸ ‘I rationally believe that p’ is just idiomatic for ‘I believe that p and it is rational for me todo so’. The usual distinction between ‘propositional justification’ (or ‘rationality’, or ‘warrant’, or‘entitlement’) and ‘doxastic justification’ is not pertinent here, since we’re exclusively concernedwith whether you can be ‘propositionally justified’ in having a Moore-absurd belief. For thedoxastic/propositional distinction, see Roderick Firth (1978).


On the Moorean-Absurdity of (i):

1. B(i) ⊃ Bp Assumption from (f )2. Bp ⊃ RBp Assumption from (d)3. R(i) ⊃ R ∼ Bp Assumption from (b)4. R ∼ Bp ⊃ ∼ RBp Assumption from (a)5. B(i)&R(i) Assumption for Reductio Ad Absurdum6. B(i) 5 by Simplification7. R(i) 5 by Sim.8. B(i) ⊃ RBp 1, 2 by Hypothetical Syllogism9. R(i) ⊃ ∼ RBp 3, 4 by HS

10. RBp 6, 8 by Modus Ponens11. ∼RBp 7, 9 by MP12. ∼(B(i)&R(i) ) 5–11 by RAA

On (ii):

1. B(ii) ⊃ BB ∼ p Assumption from (f )2. BB ∼ p ⊃ ∼ Rp Assumption from (e)3. R(ii) ⊃ Rp Assumption from (b)4. B(ii)&R(ii) Assumption for RAA5. B(ii) ⊃ ∼ Rp 1,2 by HS6. B(ii) 4 by Sim.7. ∼Rp 5,6 by MP8. R(ii) 4 by Sim.9. Rp 3, 8 by MP

10. ∼(B(ii)&R(ii) ) 4–9 by RAA

On (iii):

1. Rp ⊃ ∼ B ∼ Rp Assumption from (c)2. B(iii) ⊃ B ∼ Rp Assumption from (f )3. R(iii) ⊃ Rp Assumption from (b)4. B(iii)&R(iii) Assumption for RAA5. B(iii) 4 by Sim.6. R(iii) 4 by Sim.7. Rp 3,6 by MP8. B∼Rp 2,5 by MP9. ∼B∼Rp 1, 7 by MP

10. ∼(B(iii)&R(iii) ) 4–9 by RAA

On (iv):

1. B(iv) ⊃ BR ∼ p Assumption from (f )2. Rp ⊃ ∼ BR ∼ p Assumption from (c)3. R(iv) ⊃ Rp Assumption from (b)4. B(iv)&R(iv) Assumption for RAA


5. B(iv) 4 by Sim.6. R(iv) 4 by Sim.7. BR∼p 1, 5 by MP8. Rp 3, 6 by MP9. ∼BR∼p 2, 8 by MP

10. ∼(B(iv)&R(iv) ) 4–9 by RAA

On (v):

1. B(v) ⊃ RB(v) Assumption from (d)2. RB(v) ⊃ R ∼ (v) Self-evident²⁹3. R ∼ (v) ⊃ ∼ R(v) Assumption from (a)4. B(v)&R(v) Assumption for RAA5. B(v) ⊃ R ∼ (v) 1,2 by HS6. R(v) ⊃∼ R ∼ (v) 3 by Contraposition7. B(v) 4 by Sim.8. R(v) 4 by Sim.9. ∼R∼(v) 6, 8 by MP

10. R∼(v) 5,7 by MP11. ∼(B(v)&R(v) ) 4–10 by RAA

On (vi):

1. R(vi) ⊃ R(I have no beliefs [(v)]) Common sense2. RB(vi) ⊃ R ∼ (v) Self-evident3. B(vi) ⊃ RB(vi) Assumption from (d)4. R ∼ (v) ⊃∼ R(v) Assumption from (a)5. B(vi)&R(vi) Assumption for RAA6. B(vi) 5 by Sim.7. B(vi) ⊃ R ∼ (v) 2, 3 by HS8. R∼(v) 6, 7 by MP9. R(vi) 5 by Sim.

10. R(v) 1,9 by MP11. ∼R(v) 4,8 by MP12. ∼(B(vi)&R(vi) ) 5–11 by RAA³⁰

I take it that we have shown that—the closure principle, belief distribution,and our foundationalist claim aside—from a set of premises whose mainepistemic content is the prevention of incoherence, it follows that Moore-absurdbelieving promotes that kind of irrationality, conflict-irrationality, which we haveidentified with incoherence, QED.

²⁹ Premise 2 is an instance of this self-evident principle: If it is (prima facie) rational for me tobelieve that I believe that p, then it’s (prima facie) rational for me to believe that I have at least onebelief.

³⁰ This argument is in response to an objection that Williams made against my previous argumentfor the irrationality of (vi). I’m grateful to him for identifying a pocket of obscurity in that previousargument.


CONCLUDING REMARKS: MOOREAN ABSURDITY,

KNOWLEDGE, AND JUSTIFICATIONISM

There are at least three interesting (and intimately related) consequences of theforegoing for contemporary epistemology. First, I have given you a foundationalistexplanation of Moorean Absurdity. This is, of course, bad news for the traditionalrivals of foundationalism, which, to my knowledge, haven’t produced any suchexplanation. Nor do we seem to have an inkling of what form either a coherentistor an infinitist resolution of Moore’s paradox might take.³¹

Second, unlike my previous offering on this matter (in my 2001), accordingto which a Moore-absurd belief is one for which there can be no non-overridenevidence, the present explanation is non-evidentialist. It is, therefore, one whichis available to any form of foundationalist epistemology that is compatible withour claim (d). Whereas my previous proposal placed a resolution of Moore’sparadox out of reach for all externalists, the present proposal levels the playingfield (as far as the paradox is concerned) for both internalist and externalist formsof foundationalism. Evidentialism—particularly, the internalist foundationalismfavored by my previous analysis of the matter—loses its foothold here, becauseour analysis of Moorean Absurdity has given us no basis to disqualify an externalistconception of foundational justification (or ‘rationality’).

Third, because we have relied on a premise (our (d)) according to whichthe occurrence of a lower-level belief state produces epistemic entitlement for ahigher-level belief state which describes the occurrence of that lower-level beliefstate, our explanation of Moorean Absurdity reminds us that non-justificationistepistemologies—those for which doxastic rationality (or ‘epistemic justification’,or ‘epistemic entitlement’), no matter how it is conceived, is not a necessarycondition of knowing—have been on a collision course with Moore’s paradox.If you are a non-justificationist, what will you say about Moore-absurd belief ?Since every belief of type (ii) or (iii) or (iv) can be true, can it also be a case ofknowledge? If not, why not? It seems clear that, in response to these questions,the non-justificationist will either remain silent or else will try to bite a very fastbullet and hold that there is, after all, no problem in one’s having a Moore-absurdproposition as an object of knowledge. Truth-tracking theories are, of course,in this category of our victims—most prominently, Nozickian and Dretskeanepistemologies.³² (I ignore, as I think we should, the possibility that such theories

³¹ In a series of recent papers, Peter Klein has single-handedly turned infinitism into a seriousalternative to coherentism, foundationalism, and skepticism. One of the most compelling of thosepapers is his (1999).

³² See Fred Dretske (1991) and Robert Nozick (1981). For an important discussion of alternativeformulations of an interesting truth-tracking principle, see Sosa (1999). According to Fred Adamsand Murray Clarke (2005), the truth-tracking theories are alive and well—or, at any rate, making


might be patched up with justificationist conditions. Their claim to fame residesprecisely in their being based on a ‘brutely external causal requirement’ forknowledge, as Sosa (2003) would put it, thus doing away with the need for ajustification condition.³³ Brutishness implies clarity. The watering-down of thoseepistemologies shouldn’t be a starting point. But I am here suggesting that it istheir inevitable fate when Moore’s paradox is factored into the discussion.)

In his Moore-absurd predicament, the truth-tracking epistemologist is joinedby all those who have endorsed some form of non-justificationism as a viableaccount of propositional knowledge. These include what we might want tocall ‘apartheid epistemologists’: a growing number of philosophers for whomthe theory of doxastic rationality had better be kept apart from the theoryof knowledge, which, for them, may be based exclusively on truth-trackingconsiderations. Richard Foley (2001, 1998) has been an eloquent spokespersonfor apartheid in epistemology. Sosa’s (1991: ch. 13) distinction of truth-tracking‘animal knowledge’ and rationality-based ‘reflective knowledge’ may assign him tothe apartheid group. (Can a Moore-absurd belief be a case of animal knowledge?If not, why not?) The William Alston for whom epistemic justification is animportant notion but epistemological justificationism is a form of ‘professionalparochialism’ (Alston 1988: 181) should be regarded as a charter member of theapartheid club; as apparently also should the Laurence BonJour for whom ‘perhapsexternalism and internalism are best regarded as complementary approaches tolargely different issues, rather than as competitors’ (BonJour 2003: 36). Thoughincreasingly popular—because it promises fast conciliation between warringfactions—segregationism is a deeply tortured view. For hasn’t the concern withhaving good evidence—particularly, the concern with having good reasons—forbelieving always been the very same as the concern with being ‘non-accidentallyin touch with the truth’ (as Sosa puts it)? So, how can we make sense of the viewaccording to which an account of how one succeeds at being non-accidentally intouch with the truth (like Sosa’s virtue epistemology) is ‘simply irrelevant’ to adiscussion of ‘the main internalist account of justification’ (BonJour 2003: 189)?Segregationism has major problems of its own even before colliding with MooreanAbsurdity. Our new problem for apartheid epistemologists is very much the sameas our challenge for non-justificationists in general. The paradox puts them inthe very uncomfortable position of owing us two different explanations: one forwhy a Moore-absurd belief is irrational and the other for why a Moore-absurdbelief can’t be knowledge. And the latter had better not be based on the former!

In light of such somber prospects, it is easy to see how the non-justificationistmay be reminded of that time-honored strategy in philosophy: Try shrugging

a comeback into the mainstream. See Fred Adams (2005) for his defense of truth-tracking theoriesagainst the most popular objections.

³³ In correspondence, Sosa lets me know that his manuscript reads ‘brutely external causalrequirement’, not ‘brutally . . .’, as printed in the published version.


your shoulders and dismissing this rationale for justificationism as arising fromthat funny bit of philosophical extravagance, Moore’s paradox. But I wouldurge the non-justificationist to think again. Moorean Absurdity is an ordinaryenough manifestation of irrationality. And, if you keep it in mind, you’ll see theMoore-absurd wolf in sheep’s clothing often enough.³⁴ For instance, considerthe following scenario: You’re on a tour of that high-tech electronics plant,where your guide is a renowned scientist. You stop in front of what looks likea window, see what looks like the bright blue sky outside, and remark that theweather turned out good after all, in spite of those dark morning clouds. Towhich the scientist replies: ‘It’s actually raining, but you believe it isn’t, becausewhat you’re looking at is not a window; it’s our prototype of an amazinglyrealistic plasma screen, and we’ve invited you here to put it to the test!’ If hisclaim is true and you’re tracking the truth—because your reliable and stabletruth-tracking abilities make you responsive to his authority, etc.—you willbelieve the Moore-absurd ‘It’s actually raining, but I believe it isn’t . . .’ and holdan obviously irrational belief. So, that very mundane contingent propositioncan’t both be true and rationally believed by you.³⁵ Now, would anyone reallywant to maintain that such an obvious case of irrationality is a case of knowledge?Human knowledge it definitely doesn’t seem to be. And I confess to being one ofthose who find it hard to see that we have any other concept of knowledge—nomatter how metaphorical, ambiguous or sloppy we can get in ordinary parlancewhen the word ‘knows’ is thrown around. (For instance, while it may notseem inappropriate to identify knowledge with true belief when asserting that

³⁴ Sorensen (1988: ch. 1) carries a number of interesting non-obvious cases of MooreanAbsurdity.

³⁵ Moore’s paradox poses this wicked problem: If other people can know truths about you thatyou can’t know yourself, because you can’t rationally believe the propositions that convey thosetruths, how can they help you? In correspondence, Williams suggests that, in the plasma screen case,we obviously think that the scientist’s assertion will somehow trigger a revision in the hearer’s beliefsystem. And I think he is right: That assertion must be useful somehow. (At a minimum, we shouldexpect the rational hearer to be driven into suspension of judgment about whether what’s in frontof him is a window. By hypothesis, the exceptionally realistic screen does give him evidence notto believe the scientist. Whether the preponderance of the evidence favors believing the scientist isirrelevant here.) But it’s hard to explain how one goes about rationally making the expected revision.You can’t just pick and choose those parts of what the scientist asserted that you can rationallybelieve. He asserted an instance of ‘p but you believe that not-p’. From what he asserted, you can’tinfer that p and believe it rationally (thus having it as a candidate for knowledge) unless you firstbelieve ‘p but I believe that not-p’, and do so rationally, which, by hypothesis, you can’t do. Mytentative solution: You can rationally believe ‘The scientist has just asserted that (p & I believe thatnot-p)’. Since assertion distributes over conjunctions, you can now infer that ‘The scientist has justasserted that p’. And from this, you finally infer that p. That way, you can come to believe that prationally without having ever believed the Moore-absurd conjunction. Maybe this vindicates ourperception that you can take that treacherous assertion as the proper stimulus for revising your beliefsystem without being irrational in so doing. But my point remains: I have claimed no more thanthat you can’t rationally believe the scientist. The scientist can’t help you revise your belief systemby being believed by you. There is, however, the alternative route from the scientist’s assertion toyour rational belief revision.


a 3 year old ‘knows’ that such-and-such, common sense will readily reject theidentification when one is said to ‘know’ that such-and-such but the person’strue belief obviously results from both cognitive malfunction and luck.)

The deepest moral of my story is a justificationist one for epistemology: Sometrue beliefs can’t be knowledge just because they can’t be rationally held. Surely,this will sound like a truism to some. Interestingly, however, it should come asnews to some of our most venerable epistemologists. It is a safe bet that the reasonfor surprise is that they have tacitly assumed that every belief the content ofwhich is a contingent proposition may be a case of knowledge—provided it is nota self-refuting belief, and provided the would-be knower is not like the typicallyincoherent believer in our butler case.³⁶ But we’ve seen that those two familiarprovisos are not sufficient to root out Moorean Absurdity. In fact, I believe we’veseen quite a bit more than that: We’ve seen that a ‘non-undermining condition’which is rich enough to root out every form of incoherence, including MooreanAbsurdity, is one that includes our foundationalist premise (d) and so is, forthat reason alone, one that non-justificationists cannot afford. One may reallyhave to go all the way and embrace our foundationalist premise (d), thus lettingnon-justificationism—and, with it, the promise of instant conflict-resolution inepistemology—go.

Now, shall we get back to work on our hard-earned conflict?

REFERENCES

Adams, Fred (2005), ‘Tracking Theories of Knowledge’, in Claudio de Almeida (ed.),Perspectives in Contemporary Epistemology, Veritas, 50/4 (Porto Alegre, Brazil).

and Clarke, Murray (2005), ‘Resurrecting the Tracking Theories’, AustralasianJournal of Philosophy.

Alston, William (1988), ‘Justification and Knowledge’, in his Epistemic Justification: Essaysin the Theory of Knowledge (Ithaca: Cornell University Press).

Audi, Robert (2003), Epistemology: A Contemporary Introduction to the Theory of Knowledge(2nd edn., New York and London: Routledge).

Bergmann, Michael (1997), ‘Internalism, Externalism and the No-Defeater Condition’,Synthese, 110: 399–417.

BonJour, Laurence (2003), (co-authored with Ernest Sosa), Epistemic Justification: Inter-nalism vs. Externalism, Foundations vs. Virtues (Oxford: Blackwell).

(1985), The Structure of Empirical Knowledge (Cambridge, Mass.: Harvard Univer-sity Press).

³⁶ Still, BonJour’s (2003: 32) question is a disturbing one: ‘Is there any intelligible rationalefor the requirement of negative internal justification [the juxtaposition of some ‘‘no-defeatercondition’’], which the modified version of externalism accepts, that does not also support therequirement of positive internal justification, which it rejects?’ Can the question, mutatis mutandis,be posed to the internalist who adopts a defeasibility condition to deal with the Gettier problem? Isthat also a move that lacks a compelling, independent rationale? I think the defeasibilist internalistwill ultimately have the upper hand in this regard, but I don’t have the space to justify my answerhere.


Byrne, Alex (2005), ‘Perception and Conceptual Content’, in M. Steup and E. Sosa(eds.), Contemporary Debates in Epistemology (Oxford: Blackwell).

De Almeida, Claudio (forthcoming), ‘Coherence, Defeasibility and the Preface Paradox’,unpublished, presented at the Third International Principia Conference, Florianopolis,Brazil, 10 September 2003.

(2001), ‘What Moore’s Paradox is about’, Philosophy and Phenomenological Research,62/1 33–58.

(2000), ‘Epistemic Deontologism Set Aright’, unpublished, presented at the 2000meeting of the Central States Philosophical Association, Lincoln, Nebraska.

(1999), ‘Moore’s Paradox’, in Robert Audi (ed.), The Cambridge Dictionary ofPhilosophy (2nd edn., Cambridge: Cambridge University Press).

Dretske, Fred (1991), ‘Two Conceptions of Knowledge: Rational Belief vs. ReliableBelief ’, Grazer Philosophische Studien, 40: 15–30.

Feldman, Richard (2003), Epistemology (Upper Saddle River, NJ: Prentice Hall).Firth, Roderick (1978), ‘Are Epistemic Concepts Reducible to Ethical Concepts?’, in

A. I. Goldman and J. Kim (eds.), Values and Morals (Dordrecht: Reidel).Foley, Richard (2001), Intellectual Trust in Oneself and Others (Cambridge: Cambridge

University Press).(1998), ‘Justification, Epistemic’, in Edward Craig (ed.), The Routledge Encyclopedia

of Philosophy (New York and London: Routledge).(1993), Working without a Net: A Study of Egocentric Epistemology (New York and

Oxford: Oxford University Press).(1979), ‘Justified Inconsistent Beliefs’, American Philosophical Quarterly, 16:

247–58.Goldman, Alvin (1998), ‘Reliabilism’, in Edward Craig (ed.), The Routledge Encyclopedia

of Philosophy (New York and London: Routledge).Klein, Peter (2002), ‘Skepticism’, in Paul Moser (ed.), The Oxford Handbook of Epistem-

ology (Oxford and New York: Oxford University Press), 336–61.(1999), ‘Human Knowledge and the Infinite Regress of Reasons’, in James E.

Tomberlin (ed.), Philosophical Perspectives 13, Epistemology, 1999 (Cambridge, Mass.and Oxford: Blackwell), 297–325.

(1985), ‘The Virtues of Inconsistency’, Monist, 68: 105–35.(1981), Certainty: A Refutation of Scepticism (Minneapolis: University of Minnesota

Press).Lehrer, Keith (2000), Theory of Knowledge (2nd edn., Boulder, Colo.: Westview Press).Nozick, Robert (1981), Philosophical Explanations (Cambridge, Mass.: The Belknap

Press).Read, Stephen (1995), Thinking about Logic: An Introduction to the Philosophy of Logic

(Oxford and New York: Oxford University Press).Sanford, David (1989), If p, then q: Conditionals and the Foundations of Reasoning

(London and New York: Routledge).Senor, Thomas (1996), ‘The Prima/Ultima Facie Justification Distinction in Epistemo-

logy’, Philosophy and Phenomenological Research, 56/3: 551–66.Smullyan, Raymond (1983), 5000 B.C. and Other Philosophical Fantasies (New York:

St Martin’s Press).Sorensen, Roy (1988), Blindspots (Oxford and New York: Oxford University Press).


Sosa, Ernest (2003) (co-authored with Laurence BonJour), Epistemic Justification: Intern-alism vs. Externalism, Foundations vs. Virtues (Oxford: Blackwell).

(1999), ‘How to defeat Opposition to Moore’, in James E. Tomberlin (ed.), Philo-sophical Perspectives 13, Epistemology, 1999 (Cambridge, Mass. and Oxford: Blackwell),141–53.

(1997), ‘Reflective Knowledge in the Best Circles’, Journal of Philosophy, 94:410–30.

(1991), Knowledge in Perspective: Selected Essays in Epistemology (Cambridge:Cambridge University Press).

Williams, John N. (1996), ‘Moorean Absurdities and the Nature of Assertion’, Australasi-an Journal of Philosophy, 74: 135–49.

(1994), ‘Moorean Absurdity and the Intentional ‘‘Structure’’ of Assertion’, Analysis,54: 160–6.

4

The Normative Character of Belief

Thomas Baldwin

1

Moore points to the absurdity of saying such things as ‘I don’t believe that it’sraining, though as a matter of fact it is’ and ‘I believe he has gone out, but hehas not’ (Moore 1942: 543; 1959: 175; 1993: 207). He goes on to argue thatthis kind of absurdity is paradoxical because it disappears once these statementsare conjugated into the past tense, as in ‘I did not then believe it was raining,though as a matter of fact it was’, or expressed in the third-person as ‘Mooredoes not believe that it’s raining, though as a matter of fact it is’ (Moore 1993:208–9). For these transformations can be understood to be ways of expressingthe very same proposition as was expressed by the initial statement; and it is,he suggests, paradoxical that for each of us there is a type of statement aboutourselves which we cannot make without absurdity even though the propositionthereby expressed, so far from being self-contradictory, may well be true and canbe expressed in other ways without absurdity.

Moore goes on to suggest that the resolution of this paradox lies in takingaccount of what one ‘implies’ by making a statement in addition to what one‘means’ by it (Moore 1993: 210). In saying ‘I don’t believe that it’s raining,though as a matter of fact it is’, I imply by the second clause that I believethat it is raining; and since this implication conflicts with what I mean by thefirst clause, my statement is absurd: what I imply contradicts what I mean orassert. Furthermore, the fact that the statement’s absurdity is not preserved bytransformations of it into the past tense or the third-person is explained by thefact that although these transformations preserve the proposition expressed, theyalter the implications inherent in expressing the proposition in one way ratherthan another. In using the past tense (‘I did not then believe it was raining,

I am much indebted to John Williams and to an anonymous referee for comments on earlierversions of this paper.

Thomas Baldwin 77

though as a matter of fact it was’), I imply only that I now believe that it was thenraining, not that I did then believe this; and if I use the third-person to describeMoore’s belief (‘Moore does not believe that it’s raining, though as a matter offact it is’), I imply that I believe that it is raining, not that Moore believes this.So in neither case does the implication of my saying what I do contradict what Ithen mean.

Where the issue is presented as Moore presented it, as a paradox about theabsurdity or not of different ways in which the same thing can be said, hisresolution of the paradox must be along the right lines. As was observed by JohnWilliams some years ago (Williams 1979), however, the absurdity inherent insaying ‘I believe he has gone out, but he has not’ needs some further attention.For on Moore’s proposal what is implied by the second clause here is only thatI believe that he has not gone out, and this implication does not contradict mystatement that I believe that he has gone out. This belief is certainly inconsistentwith the belief that I say that I have; but since we can and do have inconsistentbeliefs, the ground for the absurdity of the initial statement cannot be that ofthe other type of case. Nonetheless, because the inconsistency is wholly explicitin this case, the explanation for the absurdity of this second type of Mooreanutterance is that the speaker is presenting himself as someone with beliefs whichare so overtly inconsistent that no rational person should seriously entertain themtogether.

A different issue arises from Moore’s account of the way in which assertionimplies belief. According to Moore this implication rests on an inductiveassociation: it is simply because the association between assertion and belief isso very common that we experience statements in which this implication iscontradicted as absurd (Moore 1993: 210). This cannot be right. It entails thatwhen we are dealing with a known habitual liar who frequently says what weknow he does not believe we should not find any absurdity in an utteranceby him of a Moorean type, such as ‘I don’t believe I returned your book,though as a matter of fact I did’. But the absurdity is just as manifest in thisstatement as in others, and this fact suggests that the absurdity of a Mooreanutterance arises from the fact that it violates some general principles inherentin the practice of assertion. In a previous discussion of this matter I advanceda proposal of this kind, drawing on some of Grice’s suggestions about meaning(Baldwin 1990: 228–9; Grice 1989: 219). My proposal (slightly modified) wasthat since assertion is a speech act in which one tells one’s audience somethingwith the intention of providing them with a reason for believing it through theirrecognition that one intends them to believe that one believes it oneself withgood reason, this intention excludes deliberately representing oneself, within thevery same speech act, as not believing what one asserts, since one would therebyundermine one’s intention of providing them with a reason for believing itthrough their recognition that one intends them to believe that one believes itoneself. So on this account, the absurdity of a Moorean utterance such as ‘I don’t

78 The Normative Character of Belief

believe that it’s raining, though as a matter of fact it is’ is that someone hearingit cannot make sense of the speaker’s intentions; by disavowing his own beliefin that which he is asserting, namely that it is raining, the speaker undermineshis own apparent communicative intention and his audience is left with nostraightforward way of making sense of the speech act (though of course it can bereinterpreted as an expression of astonishment or something similar). As notedbefore, the account of the absurdity of the other type of Moorean utterance, ‘Ibelieve that Jones has gone out, but he has not’ has to be slightly different, forhere the speaker does not disavow belief concerning the matter which he presentshimself as intending to convey to his audience, namely that Jones has not goneout. But since he explicitly affirms the contradictory belief, that Jones has goneout, the speaker still leaves the audience bewildered. He still undermines hisapparent intention to give them a reason to believe that Jones has not gone outthrough their recognition that he believes this for good reason by telling themthat he himself has exactly the opposite belief.

2

This account of the absurdity inherent in putative Moorean assertions still seemsto me to have some merit. But it relies on assumptions about the intentionalstructure of assertion that are disputable. On the one hand, there appear to becases in which a speaker makes an assertion without the intention of therebyproviding his audience with a reason for believing what he asserts, as when heis being tested by someone whom he takes to know the answers already; onthe other hand, there are cases of conversational implicature in which one doeshave the appropriate communicative intentions, but does not assert what onedeliberately implicates. Furthermore, the account is open to the complaint thatit seems misplaced to locate the ground for Moore’s absurdities in the way inwhich they undermine their own communicative potential, as opposed to findingit within the meaning of the assertions themselves. For both reasons, therefore,it would be preferable if one could locate the ground for these absurdities in amore straightforward account of the relationship between assertion and beliefwhich does not rely on assumptions about the communicative intentions ofspeakers.

I will come back later to a suggestion of this kind. For the moment, however,I turn to a different issue. Just as a Moorean assertion strikes us as absurd,it is similarly absurd to suppose someone having the thought: I don’t believethat it’s raining, though as a matter of fact it is; and since the absurdity hereseems to be intrinsic and not dependent on the fact that such a thought couldnot be communicated to others without incoherence, some further accountof this absurdity is needed. In my previous discussion of this issue (Baldwin1990: 230), I suggested that if one thinks of such a thought as a conscious

Thomas Baldwin 79

belief, and thus such that the thinker is aware of believing it when he hasit, a potential contradiction within the thinker’s mind similar to that impliedfor the audience of a Moorean assertion can be generated. For, on the onehand, the Moorean thinker is aware of having the belief: I don’t believe thatit’s raining; but, on the other hand, in being aware that he also has the beliefthat it is raining, he believes that he believes that it is raining. By a reflectivemove this higher-order belief is itself one of which he can become aware; so ifhe is reflective he will become aware that he believes that he believes that it israining, and there is no reason why this awareness should veil his prior awarenessthat he also believes that he does not believe that it is raining. Hence, he willbecome aware both that he does not believe that it is raining and that he doesbelieve this, which is manifestly incoherent. And, I suggested, it is a sense of thispotential incoherence which motivates the impression that the Moorean thoughtis absurd.

A similar argument will yield the conclusion concerning conscious Mooreanbeliefs of the type ‘I believe that Jones has gone out, but he has not’ that areflective thinker can become aware of believing both that he believes that Joneshas gone out and that he believes that he has not gone out. Because thesehigher-order beliefs are not contradictory the incoherence here is not as extremeas in the other case; but since the inconsistency is still overt, no rational thinkercould be content to be aware of believing that he has both these beliefs. So mysuggestion was, again, that it is implicit recognition of this result which explainsthe sense that the Moorean thought itself would be absurd.

This account of the absurdity of Moorean thoughts no longer seems to me tobe satisfactory (but for a recent defence of a version of it, see Williams 2006).It depends on a capacity for reflective self-awareness which is questionable andit applies only to the conscious beliefs of a suitably reflective thinker. But theabsurdity of Moorean thoughts seems to lie deeper than this—in merely havingsuch a belief rather than in the prospect of being aware that one has such abelief. One way to bring this point out is to compare beliefs with assumptions.Like belief and other propositional attitudes, whatever we assume we assume tobe true; in this sense assumption ‘aims at the truth’. Nonetheless, there is noabsurdity in my thinking: it is raining, but I will not assume that it is. For itis precisely the point of attitudes such as assumption that we can employ themin order to distance ourselves in thought from what we take to be the truth.Moore’s paradox arises from the fact that belief is not like this: we cannot in thesame way distance what we take our beliefs to be from the truth as we see it.This feature of belief seems fundamental to it and needs to be explicated; butin looking for an explication it does not seem sufficient to indicate the potentialfor incoherence in the beliefs of a reflective Moorean thinker who is aware of hisbeliefs.


3

In addressing this issue, I want to start from Jane Heal’s suggestion that thereis ‘a sort of performative character’ in declarations that one believes something(Heal 1994: 21). The purpose of this suggestion is to support the claim thatsincere first-person declarations of belief are authoritative because there is a‘constitutive link’ between one’s judgment that one believes something and thebelief itself: in such cases ‘the occurrence of a representation of a state of affairsis itself what constitutes that state of affairs’ (Heal 1994: 21). This is the reason,Heal suggests, why first-person belief claims have the special status reflected byMoore’s absurdities. In making her suggestion Heal acknowledges that believingis not just like promising, and in thinking about her position it is important toconsider the differences between them. The most significant is that one makesa promise simply by deliberately saying ‘I promise’ or something similar, sothat as long as the speech act was intentional, one has thereby promised. Inthis case, the intentional act of producing a verbal ‘representation’ (‘I promise’)does constitute that which is represented, a promise. But in the case of belief,nothing so straightforward obtains: my solemn and intentional declaration ofthe Apostles’ Creed (‘I believe . . .’) does not make it the case that I believe thesethings. Nor does Heal maintain that it does; instead she holds that only mysincere declaration or judgment that I believe that p ‘constitutes in me a beliefthat p’ (Heal 1994: 22).

Sincerity, however, is not enough. Sincerity is a matter of declaring what Ibelieve that I believe and in cases of self-deception such sincerity co-exists withthe absence of first-order belief. Religious belief is a case in point: someonewho has lost their religious belief but does not admit this to herself may makea sincere declaration of faith without thereby resurrecting it. So, contrary toHeal’s proposal, in the case of belief even a sincere occurrence of a first-personrepresentation of a belief (‘I believe . . .’) does not constitute that state of affairs,the belief in question. Furthermore, the role of sincerity in her account isunsatisfactory; for since sincerity involves a first-person higher-order belief, theneed to include it as a condition under which a representation of a first-personbelief constitutes that belief threatens a regress: for what constitutes the higher-order first-person belief, if not a further declaration of this belief that must itselfbe sincere?

These points suffice, I think, to undermine Heal’s proposal. But it is nonethelesshelpful to consider a related line of thought concerning acceptance. I takeacceptance to be a public act whereby one commits oneself to the truth of someproposition which has been put before one for an opinion; it can be expressed bysaying ‘I accept’, though it is normally expressed just by saying ‘Yes’ in responseto a question, just as one expresses rejection by saying ‘No’. These speech acts do

Thomas Baldwin 81

indeed have ‘a sort of performative character’; for by performing them, I makeit true that I accept, or reject, the proposition in question, and unlike the casewith expressions of belief, sincerity is not a precondition of acceptance. WhateverGalileo may have said under his breath about the movement of the earth, byanswering ‘No’ to the Inquisition’s question as to whether the earth moves, hethen accepted that the earth does not move. Yet there is a connection with belief:for even if Galileo did not believe what he then accepted, by accepting it herepresented himself as believing it. In giving a public commitment to the truthof a proposition one shows oneself as thus committed to its truth, and thus asbelieving it.

This performative aspect of acceptance is, not surprisingly, linked to the factthat it should be voluntary. For it is only where a speech act is voluntary that itcan give rise to a commitment of any kind, including a commitment to the truthof the proposition accepted. Where torture or other forms of coercion have beenused to secure acceptance, acceptance normally becomes void in just the way thatcontracts secured under duress are void; an utterance of ‘I accept’ under duressis not a way of entering into a commitment. Nonetheless, I myself may regardmy coerced acceptance as a shameful betrayal; and although others may urge meto forgive myself, the fact that we understand this kind of self-reproach bearswitness to the performative aspect of the speech act of acceptance. Despite theunbearable coercion, it may still appear to me that by uttering the words ‘I accept’I did accept a doctrine that I actually hold to be unacceptable or worse (theseunpleasant questions about coercion and acceptance were famously explored byArthur Koestler in Darkness at Noon and George Orwell in 1984).

Because acceptance is an overt commitment to the truth of a proposition, itwould seem that reasons for accepting a proposition should be grounds for itstruth. Yet one can have reasons for accepting a proposition which are not onlynot grounds for its truth but which outweigh doubts about its truth—as whenone is motivated by ties of personal loyalty or by considerations of personalgain. These are of course cases in which acceptance is not sincere; but whatthey show is that it is precisely the performative character of acceptance thatpulls it away from being a straightforward expression of belief. Hence, althoughthe first-person dimension of Moore’s paradox appears at first to fit well withHeal’s suggestion that belief has ‘a sort of performative character’, on furtherreflection it becomes clear that the necessarily overt and conventional characterof a performative speech act such as acceptance precludes its being a basis for anaccount of belief which explicates Moore’s paradox.

4

Nonetheless, there is something important to be carried forward from thediscussion of acceptance, namely the conception of an act whereby one commits


oneself to the truth of a proposition. For this seems an essential element foran account of belief which, as Moore’s absurdities indicate, deals with the factthat we do not detach ourselves from our beliefs in the way in which we dodetach ourselves from our assumptions. Yet since belief is not itself an act, anycommitment of this kind inherent in belief must accrue to it from some otheract, an act similar to acceptance but tied more closely to belief than acceptanceis. I suggest that if we look to judgment, we will find what we are looking for.

Judgment, unlike acceptance, is not a performative act of any kind. None-theless, like acceptance, it is often prompted by a doubt or a question and itis sometimes marked by an explicitly first-person thought—‘I think so-and-so’.More often judgment is just a straightforward mental affirmation, that so-and-so is the case, and this connection invites a reconsideration of the nature ofordinary linguistic assertion. In my initial discussion (§1) of the relation betweenbelief and assertion I treated assertion as a speech act with a complex structureof interlocking communicative intentions and beliefs which make Mooreanassertions self-defeating. But, as I went on to indicate (§2), it is doubtful that it isa structure of this kind which makes assertion the act it is, and in the light of theprevious account of acceptance, it is natural to consider instead the hypothesisthat assertion is an act in which a speaker commits himself to the truth of whathe says. This hypothesis does not entail that one always asserts what one says; butwe normally do, and where a doubt is raised as to whether the speaker is indeedasserting what he says, rather than merely guessing, being ironic, or somethingsimilar, the matter can be resolved by an overt act of acceptance from which thespeaker can no more subsequently dissociate himself than he can from an explicitpromise. This account of assertion as a way of committing oneself to the truthof what one asserts is not perhaps incompatible with one which concentrates onthe speaker’s communicative intentions; but once assertion is conceived in thisway the relationship between assertion and communication is better thought ofas indirect and not necessarily universal. One can suppose, for example, that thepractice of assertion developed as a way of meeting the communicative needsof speakers; but that, once established, the practice does not require that in allcases speakers should have the communicative intentions it developed to fulfil.However that may be, the important point here is that this account of assertionoffers the prospect of providing a basis for the immediate link between assertionand belief that Moore’s absurdities suggest without the need to bring in thecommunicative intentions of speakers. But to vindicate this suggestion moreneeds to be said about belief and, as a preliminary, about judgment, to which Inow return.

The discussion of acceptance proper showed that because it is a conventionalact, it need not be sincere and may have motivations that are alien to its officialtruth-oriented appearance. Judgment, by contrast, involves no conventionalact which enables it to become detached from the thinker’s deliberations;from the thinker’s perspective reasons for judging that p are simply evidence

Thomas Baldwin 83

for p. Notoriously here too appearances can deceive: wishful thinking and self-deception are always possible, and where they are operative judgment is in factbeing influenced by the thinker’s wishes and fears, and not simply by his appraisalof evidence. But even for a reflective subject who recognizes the inevitability ofsome such wishful thinking on his part, it remains the case that the only reasonsfor him to judge that p are the considerations which he takes to be evidence forp in the first place. The issue that remains delicate here, however, is whetherjudgment, like acceptance, is to be thought of as voluntary. The reason forrejecting that characterization is that the attitudes in which we detach ourselvesfrom truth are precisely voluntary attitudes such as assumption, supposition, andimagination; because we can assume whatever we choose to assume, we do nottake it that that which we do assume is in fact true, even though whatever weassume we assume to be true. And yet the case of perception shows the mistakeof moving to the opposite position, of taking it that our judgments are simplythe involuntary products of processes over which we have no control. For it isbecause we have no control over appearances that we learn that in abnormalcircumstances our perceptions are not to be relied upon; in such circumstances,as familiar skeptical arguments remind us, we learn to doubt the senses, andthus to detach judgment from ‘mere’ appearance. Where then is judgment to besituated?

It seems to me that the point which I have carried over from the discussionof acceptance, namely that in judgment we make a commitment to the truthof a proposition, is intelligible only if judgment is conceived as a voluntaryact. No outcome of an involuntary process could bring with it a commitmentof this kind. Hence what is needed is a way of differentiating judgment fromvoluntary attitudes such as assumption. My suggestion is that whereas in thecase of assumption and similar attitudes we decide what to assume, in the caseof judgment our decision is limited to deciding whether to believe something isthe case, whether to commit ourselves to its truth. So judgment is not decidingwhat to believe, nor indeed what is the case: we have no such cognitive powers.Even wishful thinking is not a matter of deciding to believe what it best suitsone to believe but only of being subconsciously motivated, in deciding whetherto believe something, by one’s wishes—which are not of course reasons for thejudgment at all. Nonetheless, judgment is a voluntary act: thus when doubtsarise we can, if we so choose, suspend belief and initiate further investigations.The outcome of these investigations is, necessarily, not decided by us; but whenthey are over our responsibility returns as we make up our mind whether or notto commit ourselves to the truth of the proposition which they have put beforeus as worthy of commitment.

Implicit in this account of judgment is a normative conception of belief itself,as a state whereby a thinker maintains his commitment to the truth of theproposition judged. This commitment is manifested in the thinker’s use of theproposition in question as a reason, especially as a reason for action. For by this


use of his belief the thinker puts it to work; he relies on it to fulfil his intentionsin a way in which he could not similarly rely on what he merely assumes tobe the case. ‘Actions speak louder than words’ in that they reveal our beliefs byshowing the commitments which actually inform the practical reasonings thatlead us to our actions. Since these practical reasons are causes of action, it followsthat this normative conception of belief sustains our common sense conviction ofthe causal role of beliefs. However, this conception of belief is different from thefamiliar functionalist conception of belief. For that is the conception of a stateidentified by its causal role in mediating interactions between memory, sensoryinputs, motivational drives (desires), and behavioural outputs. So on this viewbeliefs are thought of as states whose representational content has been producedby a combination of sensation, memory, and inference and which is revealed bythe distinctive causal contribution of these states to the causation of behaviourdirected at the satisfaction of desires. On the face of it, this conception of beliefdoes not account for the absurdity of Moorean beliefs. For I can readily have thethought: my physical state is prompting me, or my body, to behave as thoughit is raining, but it isn’t raining; and even if the functionalist position is mademore sophisticated by introducing a normative element so that, say, my beliefsare taken to include a state whose ‘proper’ function is to prompt behaviourwhich will satisfy my desires only if it is raining, it remains the case that I cansensibly hold that my mind/brain includes a state of this kind even though it isnot raining. So Moore’s paradox shows that functionalist accounts of belief areessentially incomplete (see Heal 1994: 12–20 for an extended discussion of thispoint). Before one can infer that Moore’s paradox thereby refutes functionalisttheories of mind, however, there are further moves to be taken into account.In particular, there is a neo-Humean strategy which, starting from a theoryof mind based upon a functionalist account of sub-personal states which donot give rise to Moore’s paradox, holds that our ordinary conception of thesubject of thought (‘I’) as a locus for normative commitments such as thoseexhibited by Moore’s absurdities is a socially constructed artifice which expressespersonal responsibilities but does not have a role in explanatory psychologicallaws (see Dennett 1991: ch. 13 for a position of this kind). I myself do nottake this view; but I do not wish to argue the point here (see Stroud 1977:127–40, 236–7 for a critical discussion of Hume’s account of the self whichis relevant to neo-Humean proposals). For a successful reconciling strategy ofthis kind, so far from undermining the normative, conception of belief advancedhere, will just show how it is possible to employ this conception within ametaphysical perspective founded upon the explanatory paradigms of the naturalsciences.

What does still need some attention is the relationship between perceptionand belief. For on the one hand, the fact that perception does not give riseto further versions of Moore’s paradox, that there is no absurdity in either ‘Itappears to me to be raining, though it isn’t’ or ‘It is raining, though it does not

Thomas Baldwin 85

appear to me to be’, shows that there is a crucial distinction between perceptionand belief. Yet on the other hand, the intimate relationship between them isshown by the way in which our perceptions are constantly furnishing us withperceptual beliefs without any explicit judgment on our part: we have a standingdisposition to trust our perceptions, to believe what we see, unless there aregrounds for doubt. This latter point is central to the common sense strandof British empiricism that runs from Locke to Moore, and is especially wellexpressed by Reid: ‘Perception commands our belief upon its own authority,and disdains to rest its authority upon any reasoning whatsoever’ (Reid 1783: II,ch. 5, pp. 109–10). Yet we also recognize that this authority is fallible—whichis why we recognize that there are no perceptual versions of Moore’s paradox.Hence, despite Reid’s affirmation that perception’s authority does not rest on ‘anyreasoning whatsoever’, our unreflective disposition to trust our perceptions standsin need of some reflective justification, or at least explanation. Empiricists willoffer a justification which invokes beliefs about the reliability of our perceptualsystems. Since these beliefs are themselves perceptually grounded, this approachraises issues concerning circularity that are familiar from discussions of theCartesian circle and induction. This paper is not the place to address these issues,but my own judgment is that externalist conceptions of justification provide away of defusing the charge of circularity while also accounting for the fact thatfor the most part we believe what we perceive without thinking about the matter,and that it could not be otherwise (for a much fuller discussion of this matter, seeBurge 2003).

5

I have been suggesting that belief has a normative character, that it is the state ofbeing committed to the truth of a proposition. I made this suggestion followingmy discussion of Heal’s proposal that Moore’s paradox derives from the quasi-performative character of first-person declarations of belief. Having argued thatHeal’s proposal was not satisfactory, however, it remains for me to show that myown suggestion fares any better as an account of Moore’s paradox.

Consider first the statement:

(1) It’s raining but I don’t believe that it is.

As Moore says, the sense of the absurd arises here from the fact that by assertingthe first clause, the speaker implies that he believes that it is raining, whereas inthe second clause he explicitly disavows this belief. The tricky issue concerns therationale for this first implication. On the account I am proposing, in assertionwe commit ourselves to the truth of the proposition we assert and thereby implythat we believe it since belief just is the state of being thus committed. So thespeaker of (1) first commits himself to the truth of the proposition that it israining and then denies that he has any such commitment. The absurdity is


therefore inherent in the assertion of (1), without the need to bring in any explicitreference to the confusing indications thereby given to an audience as to whatthe speaker intends them to believe. Moore’s second type of absurd assertion isexemplified by the statement:

(2) It’s raining but I believe that it isn’t.

In this case, the belief implied by the assertion of the first clause, that it israining, is obviously inconsistent with the belief which the speaker goes on tosay that he has. So on my account of assertion and belief the situation is one inwhich the speaker begins by committing himself to the truth of the propositionthat it is raining but then affirms that he is also committed to the truth of theproposition that it is not raining. What is absurd here is not the existence ofinconsistent commitments, but the deliberate espousal of a commitment which isobviously inconsistent with another commitment to whose existence the speakeralso commits himself.

I argued in §2 that a full account of Moore’s paradox needs to handle theabsurdity of Moorean thoughts as well as that of Moorean assertions, and thatmy own (1990) account of the former was not satisfactory. So this new approachalso needs to meet this requirement. Consider therefore a type-(1) thought, suchas someone, Jones, judging that it is raining and that he does not believe thatit is. According to my account of what is involved here, by his judgment Jonescommits himself both to the truth of the proposition that it is raining and tothe truth of proposition that he has no commitment to the truth of the sameproposition. But this is an absurd act: for by committing himself to the truthof the proposition that it is raining Jones gives himself a commitment of thekind whose non-existence he also commits himself to in the very same act. Soconsidered as a putative act of commitment the judgment is absurd: it both makesa commitment and equally disavows the existence of any such commitment. Inthe case of type-(2) thoughts the reasoning is similar. Here Jones judges that it israining and that he does not believe that it is; on my account Jones here commitshimself both to the truth of the proposition that it is raining and to his beingcommitted to the falsehood of this proposition. Again this act is absurd, forthe initial commitment obviously conflicts with the further commitment whichJones also affirms in the very same act. We do of course find ourselves from timeto time with inconsistent commitments, but it is absurd to make commitmentswhose inconsistency is obvious in the very judgment itself.

6

This new approach, centred on the normative character of belief, judgment,and assertion, does therefore explicate the absurdity of Moorean judgments inmuch the same way as it deals with that of Moorean assertions. This seems to

Thomas Baldwin 87

me a significant merit. The approach works by identifying conflicts among thecommitments inherent in Moorean assertions and thoughts; as such it is ratherdifferent from approaches which invoke principles of epistemic logic concerningbelief in order to exhibit the absurdity of Moore’s assertions, and in this finalsection I shall briefly explore this difference. To present an account of this latterkind, we should go back to (1). Since this is of the form ‘p but I don’t believethat p’ it is obviously a counterexample to the principle:

(3) p ⇒ I believe p.

But since (3) is clearly untenable, the absurdity of (1) cannot derive from itsbeing a counterexample to (3). (3), however, fails to capture the fact that theabsurdity of (1) depends on the fact that the speaker of (1) is the person who thendisavows the belief which he has implied that he has. So supposing (1) to expressA’s judgment that p but he does not believe that p, we should perhaps take theabsurdity of the assertion of (1) to arise from its implications when combinedwith the principle:

(4) A judges that p ⇒ A judges that he believes that p.

For, given (4), A’s judgment entails both that he judges that he believes that pand that he judges that he does not believe that p; hence, the suggestion is, oneshould explicate the absurdity of asserting (1) by reference to the fact that, given(4), it entails that A makes contradictory judgments. But (4) is clearly false too:judgment does not always bring with it the reflective judgment that one believes(or judges) what one judges. One can avoid this problem by moving to

(5) A judges that p ⇒ if the question arises for A whether he believes p, Ajudges that he believes p

and offer (5) instead as the basis for the absurdity of asserting (1). But thisremains unpersuasive since (5) is untenable for two reasons. First, once thequestion whether he believes p is explicitly raised, A may change his mindconcerning p. Second, A may have prejudices which he does not recognize; forexample A may in fact believe that his wife should do the housework and yet, ifasked about the matter, disavow any such belief.

Many of the same points can be made in connection with (2). Here, because(2) is of the form ‘p but I believe not-p’ a similar path of reasoning leads to thesuggestion that the absurdity of (2) arises from the application of the principle:

(6) A judges that he believes p ⇒ A believes p

to A’s judgment that p but he believes not-p. For, given (6), A’s judgmententails both that A believes not-p and that A believes p, and thus that he hasobviously inconsistent beliefs. (6) does look plausible at first; but as I observedin connection with Heal’s position it is vulnerable to counterexamples involvingself-deception. A may judge that he believes that his work is going well; but itmay also be the case that sub-consciously he does not believe this at all, indeed


that he knows only too well that it is not going well, and that the anxiety thisprovokes in him leads him to form the deceptive belief that it is going well.

What is going wrong in these attempts to exhibit Moore’s absurdities as theviolation of general principles concerning belief is that because of the manyways in which we can mistake ourselves it is doubtful that there are any suitableprinciples of this kind which merit acceptance. One might then wonder whethermy own approach is not equally vulnerable to difficulties of this kind; to, forexample, vindicating (6). I think not: for me, A’s judgment that he believesthat p involves A’s committing himself to the truth of the proposition that he iscommitted to the truth of p; but this act of commitment does not entail that Ais actually committed to the truth of p. Nor does my account of the absurdityof (2) presuppose any such entailment. Instead it points to the absurdity ofembracing, within a single assertion or judgment, a commitment both to thetruth of p and to one’s being committed to the falsehood of p; for even thoughthis act does not entail that one is committed to the falsehood of p, it is absurdto commit oneself to having such a commitment just when one also commitsoneself to the truth of p. Thus, the difference between my approach and thatwhich seeks to exhibit the absurdity of Moore’s assertions in the light of generalepistemic principles is that the absurdity of the commitments I take to beinherent in Moorean assertions and judgments does not require the entailmentsbetween judgments and higher-order attitudes that are characteristic of epistemiclogic. So, although my approach does of course make general claims concerningthe normative character of assertion, judgment and belief, these claims are notcaptured by general principles of this kind. They are instead best exhibitedthrough the absurdity of the first-person assertions about belief by means ofwhich Moore first brought this whole topic to attention.

REFERENCES

Baldwin, T. (1990), G. E. Moore (London: Routledge).Burge, T. (2003), ‘Perceptual Entitlement’, Philosophy and Phenomenological Research,

67: 503–49.Dennett, D. (1991), Consciousness Explained (London: Penguin).Heal, J. (1994), ‘Moore’s Paradox: A Wittgensteinian Approach’, Mind, 103: 5–24.Grice, H. P. (1989), ‘Meaning’, in H. P. Grice, Studies in the Way of Words (Cambridge,

Mass.: Harvard University Press).Moore, G. E. (1942), ‘A Reply to my Critics’, in P. A. Schilpp (ed.), The Philosophy of

G. E. Moore (La Salle, Ill.: Open Court).(1959), ‘Russell’s Theory of Descriptions’, in G. E. Moore, Philosophical Papers

(London: Allen & Unwin).(1993), ‘Moore’s Paradox’, in G. E. Moore, Selected Writings, ed. T. Baldwin

(London: Routledge).

Thomas Baldwin 89

Reid, T. (1783), Essays on the Intellectual Powers of Man (Edinburgh), repr. (2002), ed.Derek R. Brookes (Pennsylvania: Pennsylvania State Press).

Stroud, B. (1977), Hume (London: Routledge).Williams, J. (1979), ‘Moore’s Paradox: One or Two?’, Analysis, 39: 141–2.

(2006), ‘Moore’s Paradoxes and Conscious Belief’, Philosophical Studies, 127:383–414.

5

Moore’s Paradox, Evans’s Principle,and Iterated Beliefs

John N. Williams

1. INTRODUCTION

G. E. Moore famously observed (1942: 543) that to assert

I went to the pictures last Tuesday but I don’t believe that I did

would be ‘absurd’. Moore calls it a ‘paradox’ that this absurdity persists despitethe fact that what I say about myself might be true. For you may consistentlysuppose that I went to the pictures last Tuesday but fail to believe that I did.Moreover, if you contradict my assertion then your words, ‘If you went to thepictures last Tuesday then you believe that you did’ need not be true.¹ Mostpeople who are confronted with Moore’s example would say that in some senseI have contradicted myself, even after admitting that no contradiction lies in mywords. It is no less absurd of me to silently believe such a possible truth. So a

I am grateful for discussion with Tan Yoo Guan, Mark Nowacki, Mitchell Green, and AlanHajek. Two anonymous referees for this volume provided very useful comments.

¹ In parsing your ‘Either you did not go to the pictures last Tuesday or you believe that you did’as ‘If you went to the pictures last Tuesday then you believe that you did’, I take ‘if ’ as implication.Although such an inference is generally invalid, most would allow it here. For example, Stalnaker(1975 and 1984) would allow it on pragmatic grounds because you don’t know whether it is truethat I went to the pictures or whether it is true that I believe I did. If we symbolize ‘You believe thatyou went to the pictures last Tuesday’ as ‘Bp’, we have the following proof:

1. ∼(p & ∼Bp) Suppose the falsehood of omissive Mooreanassertion

2. ∼p v ∼∼Bp De Morgan’s Law

3. ∼p v Bp ∼ ∼ elimination

4. p → Bp → equivalence

So your denial amounts to a possibly true assertion of my omniscience about my own recentmovements.

John N. Williams 91

natural way of solving the paradox is to explain the absurdity of such a belief, aswell as the corresponding assertion, in a way that identifies a contradiction-likephenomenon with something other than its content.

It is important to note that Moore also observes that for me to assert,

I believe that he has gone out, but he has not

would be likewise ‘absurd’ (1944: 204) despite the fact my words might be true.²Unlike his first example, which has the omissive form

om) p & I don’t believe that p,

this has the commissive³ form,

com) p & I believe that not-p.⁴

This difference between the omission of specific true belief and the commissionof a specific mistake in belief is inherited from the genuine difference betweenatheists and agnostics.

So any adequate account of Moorean absurdity must explain the absurdity ofcommissive assertions and beliefs as well.

Such an explanation would also have to identify other examples that intuitivelyseem to share the paradigmatic absurdity of Moore’s own. These include RoySorensen’s examples

All my present beliefs are mistaken⁵

I have no beliefsGod knows that I am not a theist

and

God knows that I am an atheist

none of which is a conjunction. If these examples are essentially like Moore’sthen an adequate account of Moorean absurdity should explain their absurdityin the same way.

² Your denial of my assertion amounts to ‘If you believe that he has gone out then he has goneout’. For if we symbolize ‘You believe that he has gone out’ as ‘Bp’ we have the following proof:

1. ∼(Bp & ∼p) Suppose the falsehood of commissiveMoorean assertion

2. ∼Bp v ∼∼p De Morgan’s Law

3. ∼Bp v p ∼ ∼ elimination

4. Bp → p →equivalence

This time, your denial amounts to a possibly true assertion of the infallibility ofmy specific belief.

³ Sorensen coins these useful terms in (1988: 16).⁴ This formal difference is disguised by Moore’s examples. If we formalize ‘I went to the pictures

last Tuesday but I don’t believe that I did’ as ‘p & ∼Bp’ then ‘I believe that he has gone out, but hehas not’ becomes ‘Bp & ∼p’. This commutes to ‘∼p & Bp’. This may be represented as ‘p & B∼p’.

⁵ See Sorensen (1988: ch. 1). One of Sorensen’s examples is actually ‘Although you think all myopinions are mistaken, you are always right’. To simplify discussion, I take this as ‘All my presentbeliefs are mistaken’.

92 Moore’s Paradox and Evans’s Principle

Sorensen (2000) also gives a second set of examples in which belief operatorsare iterated, such as the omissive

God exists but I don’t believe that I’m a theist

and the commissive

God exists but I believe that I’m an atheist.

These have the forms

om1) p & I don’t believe that I believe that p

and

com1) p & I believe that I believe that not-p

where the superscript denotes the order of iteration of the belief-operator. Let uscall these, Sorensen’s ‘iterated cases’.

Sorensen comments (2000: 29) that om1 or com1 seem less absurd to believe orassert than om or com. Moreover as iteration increases, omissive absurdity appearsto decrease, while commissive absurdity does not. Thus with four iterations

om4) p & I don’t believe that I believe that I believe that I believe that Ibelieve that p

seems less absurd to believe or assert than om1, whereas the absurdity of believingor asserting

com4) p & I believe that I believe that I believe that I believe that I believethat not-p

seems undiminished.Here is how I will proceed. In §2 I will present the problem of giving

a non-mysterious explanation of what justifies my second-order beliefs withbest authority. In §3 I will supply a tentative argument for ‘Evans’s principle’:if I am justified in believing that p, then I am justified in believing that Ibelieve that p. This delivers the non-mysterious explanation. In §4 I will replyto objections to this argument. In §5 I will use Evans’s principle to explainthe absurdity of believing om or com. In §6 I will extract a definition ofMoorean belief from Moore’s two examples. This enables me in §7 to explainthe absurdity of other beliefs that are essentially akin to Moore’s examples.In §8 I will demonstrate difficulties in Sorensen’s own account of his iteratedcases. In §9 I will argue that despite appearances, any belief in any iterationof om or com is equally absurd, as explained by Evans’s principle. In §10, Iwill explain the absurdity of Moorean assertion in terms of the fact that theintention of such an assertor to be thought a sincere truth-teller is necessarilyfrustrated. Finally in §11, I will use the whole account to explain why onlyomissive assertions, as opposed to beliefs, decrease in absurdity as iterationsincrease.

John N. Williams 93

2. JUSTIFIABLY BELIEVING WHAT MY OWN BELIEFS

ARE: THE APPEAL TO INTROSPECTION

Any account of self-knowledge must explain two facts. First, what justifies me inascribing to myself the belief, say, that it is raining, is unlike what justifies you inascribing that belief to me. Second, I am normally the best authority on my ownbeliefs.

An appeal to introspection would claim that your justification for thinking thatI hold the belief that it is raining consists in your observation of my behaviour,from which you inductively infer the best explanation that I hold that belief.But my justification for ascribing that belief to myself is neither based uponobservation of my own behaviour, nor upon any inference, but is based upon anattention to my own mental states that is normally immune from error.

The problem with this claim is that it is utterly mysterious. What is ‘attentionto my own mental states?’ And why is it normally immune from error? Thinkingof introspection as a form of perception or ‘inward looking’ deepens the mystery.⁶If introspection were a faculty of perception, it would have no physical organ.And the phenomenology of self-knowledge is not that of perception. What itfeels like to ascertain whether I hold a certain belief, if indeed it feels like anythingat all, is not what it feels like to see, touch, or smell something.

Moreover, perceptual mistakes are honest mistakes, but mistakes I make aboutmy own beliefs typically reveal a way in which I am at fault. In a first case ofperceptual mistake, suppose that I come to believe that it is raining because I seewater that looks like rain, although in fact it comes from a concealed sprinkler. Ina second case I acquire the belief because a trick of the light generates the illusionof falling water. My perceptual mechanism may function perfectly in either case.In the first case but not the second, I really do see what I seem to see, namelyfalling water, but mistake it for rain. In a third case, having my drink spiked witha drug causes me to experience the hallucination of rain. In no case am I at faultin mistakenly believing that it is raining. In the first case, my mistake stems frommy ignorance of the source of the falling water. In the second, it stems frommy ignorance that conditions of light make my apparent perceptions unreliable.In the third, it stems from my ignorance that my perceptual mechanism ismalfunctioning. Such ignorance constitutes no fault. My mistakes are honest,not epistemically blameworthy. By contrast, a typical case in which I make amistake about my own beliefs arises when I tell you that I do not believe thatwomen are inferior to men. My assertion may be sincere yet false because I amblind to the way I treat women. You may be in a better position to recognize thatmy boorish behaviour manifests the repressed belief that I sincerely deny having.

⁶ The past participle of introspicere means ‘to look inside’.


This would not be an honest mistake but would reveal a degree to which I amepistemically blameworthy.

We now have reasons to look for an alternative account of what justifies mybeliefs about my own beliefs. I will supply it in §4, via Evans’s principle, to whichI now turn.

3 . EVANS’S PRINCIPLE

In The Varieties of Reference, Gareth Evans observes (1982: 225–6):

If someone asks me ‘Do you think there is going to be a third world war?’ I must attend,in answering him, to precisely the same outward phenomena as I would attend to if Iwere answering the question ‘Will there be a third world war?’ I get myself in a position toanswer the question whether I believe that p by putting into operation whatever procedureI have for answering the question whether p . . . We can encapsulate this procedure foranswering questions about what one believes in the following simple rule: whenever youare in a position to assert that p, you are ipso facto in a position to assert ‘I believe that p’

Now what we might call:

Evans’s Rule: If I am in a position to assert that p then I am in a position toassert that I believe that p

is probably not meant to encompass insincere assertions such as lies. So it isbest read as the rule that whenever I am in a position to sincerely assert that p(in other words to assert that p in the belief that p) then I am in a position tosincerely assert that I believe that p (in other words to assert that I believe that pin the belief that I believe that p). We cannot take being ‘in a position’ to makesincere assertions as having justification for making them. I may have justification,in the form of forensic clues I have yet to analyse, for believing that Smith is themurderer. But since I don’t yet realize that what I have is justification, I am notin a position to sincerely assert that he is the murderer. So it is better to takebeing ‘in a position’ to make sincere assertions as being justified in making them,in the sense that the justification is available to me in a way that the justificationencapsulated in unanalysed clues is not. The justification is available to me if Icould produce it if I were sensibly asked to provide justification. I am justifiedin believing that p just in case I would have a justified belief that p, if I were toform the belief that p. I will not attempt to produce a full analysis of justifiedbelief. But what I have in mind is roughly captured by the proposal that if Ihave a justified belief then I have formed and sustained it in a reliable way and Ihave no reason to doubt the reliability of the way I have formed or sustained it.Moreover, if I were sensibly asked to provide justification for it, I could do so.

This reading coheres with Evans’s remark:

If a judging subject applies this procedure, then necessarily he will gain knowledge ofone of his own mental states: even the most determined sceptic cannot find here a gap inwhich to insert his knife. (1982: 225)

John N. Williams 95

What follows now is what I call

Evans’s Principle: If I am justified in believing that p, then I am justified inbelieving that I believe that p.

Although Evans provides no argument for his rule, it is highly plausible. For ifyou ask me whether I think that it is raining, I will normally treat this questionas the question, ‘Is it raining?’ So any justification I have for a sincere answerto the question of whether p will be precisely the same justification for a sincereanswer to the question of whether I believe that p.

It does seem true that in most pragmatic contexts the two questions are treatedinterchangeably. But not in all. For example, if you ask me whether the pubs arestill open, I may reply, ‘I think so, but don’t quote me’. Since you are interestedin getting to the pubs rather than the state of my beliefs, you would not take thisanswer as a yes to the question, ‘Are the pubs still open?’

Since Evans’s principle is credible, a failure to supply a reason for it would notbe fatal to what follows. However, here is a simple argument for it:⁷

(P1) All circumstances in which I am justified in believing that p arecircumstances that tend to make me believe that p.

(P2) All circumstances that tend to make me believe that p are circumstancesin which I am justified in believing that I believe that p.

(C) All circumstances in which I am justified in believing that p arecircumstances in which I am justified in believing that I believe that p.

The first premise is plausible. For example, given that my apparent perceptionsof rain are generally reliable, they justify me in believing that it is raining. If Iam asked why I believe that it is raining, I may reply, ‘Because it looks like rain’.Such apparent perceptions of rain also tend to make me believe that it is raining.This is the sense in which ‘seeing is believing’. The premise is also supported bythe pragmatic success of justification; if circumstances in which we are justifiedin our beliefs were not those that tend to make us form them, then it would behard to explain the practical value of rationality.

The second premise is plausible as well. Circumstances that tend to put meinto a certain state ipso facto count as inductive justification for thinking that Iam in that state. For example, if drinking brandy tends to make me go red inthe face, then the fact that I am drinking brandy provides me with inductivejustification for thinking that I am red-faced. Likewise, circumstances that tendto make me believe that it raining, such as seeming to see rain, give me goodinductive reason to think that I have come to believe that it is raining. Of course,I don’t have to consciously think, ‘When it seems to be raining, I usually believe

⁷ See Williams (2004) for a brief sketch of this argument. In what follows I develop it and defendit against objections. Two further objections to it are made by Hamid Vahid (2005), to which Ireply in Williams (2006).


that it is raining. It seems to be raining again, so I probably believe that it israining’. Nonetheless the inductive justification is available to me. If someoneasks, ‘Why do you think you believe that it is raining?’ I may reply, ‘Becausethat’s what I always believe when it looks like rain’.

So my apparent perceptions of rain justify both my belief that it is rainingand my belief that I believe that it is raining in virtue of different sets of reliableconnections. I am justified in believing that it is raining on the basis of seemingto see rain, because seeming to see rain is reliably connected with the occurrenceof rain. And the very same basis of seeming to see rain justifies me in thinkingthat I believe that it is raining, because seeming to see rain is reliably connectedwith my coming to believe that it is raining.

My argument for Evans’s principle would hold for any form of justification.For example, if seeming to remember that it rained yesterday justifies me inthinking that it indeed rained yesterday, that is because my apparent memoriesare generally reliable. And the very same basis of seeming to remember that itrained yesterday justifies me in thinking that I believe that it rained yesterday,because seeming to remember that it rained yesterday is reliably connected withmy coming to believe that it rained yesterday.

The principle also explains why I am normally the best authority on my ownbeliefs. Suppose that you and I are both standing at the window looking atthe weather. To ascertain whether I believe that it is raining, I simply ascertainwhether it is raining. I may justifiably ascertain this on the basis of my apparentperception of rain. But your apparent perception of rain will not justify you inthinking that I believe that it raining. Instead, you need to observe my verbaland non-verbal behaviour (as when I unfurl my umbrella or say ‘It’s raining’)and then make an inference to my belief that counts as the best explanation ofmy behaviour. Alternatively, you may observe me observing the rain and theninfer by analogy that since you have come to believe that it is raining then sohave I. In either case you are liable to mistakes in observation. In the first caseyou may mistake my tendency to play with my umbrella as a sign of interest inthe weather or mishear my question ‘It’s raining?’ as an assertion of rain. In thesecond case you may mistake falling water from a hidden sprinkler for rain ormistake my daydreaming at the window for my observation of weather.

By contrast, I need only observe the weather. Nor do I need to make anyinference at all. So you are liable to error in ways that I am not. Of course I maymistake falling water for rain. But in the second case, that mistake would notundermine my justification for correctly ascribing to myself the mistaken beliefthat it is raining.

We now have a non-mysterious alternative to the introspective account ofwhat justifies my beliefs about my own beliefs, namely that it is precisely whateverjustifies that first-order belief itself, whether it is perception, memory, deductiveinference, or even inductive inference from observation of behaviour. Theaccount also explains why I am normally the best authority on my own beliefs.

John N. Williams 97

It is also is consistent with the fact that I may sometimes adopt your methodsof justifiably ascribing beliefs to me. To return to a previous example, afteryou say ‘But look at the way you treat women!’, I might start to observe mynon-verbal behaviour towards women and so discover my repressed belief thatwomen are inferior. There would be nothing irrational in adopting this method,since the discovery it succeeds in unearthing would represent a useful insight intomy lack of self-knowledge.

But that insight could only be useful if it led me to revise my beliefs in a waythat is an overall improvement in my rationality and consequently, my publicbehaviour. Before the insight, I believed that women were inferior but did notbelieve that I held this belief. The most likely explanation of my self-blindness isthat there was no justification for my prejudice against women. For by Evans’sprinciple, had justification been available to me for believing that women areinferior, then that same justification would have been available to me for thinkingthat I hold the belief that women are inferior. But now that I have recognizedthat I do hold that belief, I should look for justification for it. Finding noneavailable, I should abandon my prejudice.

Of course if I were always incapable of acting on Evans’s rule then I wouldbe deeply irrational. For then I would have to constantly employ third-persontherapeutic methods. This would mean forgoing a first-person method ofobtaining a justifiable view of my own cognition that enjoys best authoritybecause it is immune to the mistakes in observation that another person mightmake.

4 . OBJECTIONS TO MY ARGUMENT FOR EVANS’S

PRINCIPLE — AND REPLIES

I anticipate two objections to

(P1) All circumstances that justify me in believing that p are circumstancesthat tend to make me believe that p.

Suppose there is a drug that sharpens my senses, while also making me moresceptical of everything. I take the drug, and I see that it is raining (very distinctly,thanks to the drug). I am justified in believing that it is raining—very muchso, since my sensory apparatus is working even better than usual. But thiscircumstance does not tend to make me believe that it’s raining, since I’vebecome so sceptical.⁸

I reply that in circumstances in which I’m sceptical of everything, I’m notjustified in forming any beliefs. My senses are heightened, but since I don’t trustmy senses, I’m not justified in forming any beliefs on the basis of them.

⁸ I owe this objection to Alan Hajek.


Here is the second objection. Suppose that my neighbour has planted a treein his garden. I look out of my window and see the tree. Although my sensoryapparatus is working perfectly, I misclassify what I see as a bush, so I do not tendto believe that my neighbour has planted a tree. Nonetheless, my clearly seeingthe tree justifies me in thinking that she has planted a tree, in the sense that wereI to form this belief, it would be justified.⁹

My reply is to simply deny this last claim. From my point of view, what I seeis a bush, not a tree. The fact that I seem to see a bush does not justify my beliefthat my neighbour has planted a tree. For suppose that I do form the belief thatmy neighbour has planted a tree. Then if I were asked to justify this belief, theonly ‘justification’ I could produce would be the claim that she has planted abush. That certainly does not count as justification.

Two sorts of examples might be given against

(P2) All circumstances that tend to make me believe that p are circumstancesin which I am justified in believing that I believe that p.

Suppose that my powers of perception are only highly reliable, whereas I am nearlyinfallible in introspecting my own beliefs. I seem to see rain, but for some reason,I do not form the belief that it is raining (it is, after all, only a tendency that I doso). By introspection, it is clear to me that I have not formed the belief that it israining. My introspective justification that I have not formed the belief that it israining, trumps any perceptual justification that I have formed this belief. So onbalance, I am not justified in thinking that I believe that it is raining.

I have two replies. First, the example is effective only if it describes circum-stances in which I tend to believe that it is raining. True, circumstances in whichI seem to see rain are such circumstances. But this is not the fullest, nor thereforethe fairest, description of the circumstances depicted. A fuller description wouldinclude the fact that introspection makes it clear to me that I have not formedthe belief that it is raining, presumably because I think that appearances aredeceptive. In such circumstances I do not tend to believe that it is raining.

Second, by my lights, I have no reason to think that the situation depictedcan arise. The example assumes that introspection is a source of justificationfor my beliefs about what my first-order beliefs are, that is separate from whatjustifies my first-order beliefs. But if my account in the last section is correct,then there is no reason to think that there is any such faculty of introspection,because whatever justifies my second-order beliefs is precisely whatever justifiesmy first-order beliefs.

Now consider a different sort of putative counterexample. Suppose that I takea drug that tends to make me paranoid while at the same time causing suchextensive brain damage that I am incapable of forming higher-order beliefs. Inthis circumstance, I might tend to believe that people are persecuting me, but Iam not justified in believing that I believe that people are persecuting me. In the

⁹ An anonymous referee for this volume made this useful objection.

John N. Williams 99

same spirit, suppose that a small child knows what rain is but lacks the conceptof belief. When it stands in the rain, the child might tend to believe that it israining, but it is not justified in thinking that it believes that it is raining, becauseit is incapable of forming beliefs about beliefs.¹⁰

In reply, recall that in §3, we noted that to say that I am justified in believingthat p is neutral on the question of whether I actually believe that p. So the factthat I do not have the second-order belief does not show that I am not justifiedin holding it. Second, it seems reasonable to think that you cannot sensibly askwhether I would justifiably believe that I believe that people are persecuting me,were I to form this second-order belief. To ask this question already supposesthat I suffer no drug-induced brain damage and so violates the example. LikewiseI can attach no sense to the question of whether the small child standing in therain would have a justified second-order belief, were he to form it. To supposethat he does form it is ruled out by the example. So neither objection is coherent.

Finally, a counterexample might be given to the conclusion, in other words,Evans’s principle:

(C) All circumstances in which I am justified in believing that p arecircumstances in which I am justified in believing that I believe that p.

Suppose that compelling scientific evidence justifies me in believing thatspiders are harmless, yet psychological evidence, such as films of myself avoidingspiders, justifies me in believing that I do not believe that spiders are harmless.Since I cannot be justified in holding contradictory beliefs, I cannot be justifiedin believing that I do believe that spiders are harmless, although I am justified inbelieving that spiders are harmless.¹¹

I reply that the plausibility of this example depends upon separating thecircumstances in which the scientific evidence is available to me from those inwhich the psychological evidence is available to me. Circumstances in which onlythe psychological evidence is available to me will not support the counterexample,because in such circumstances I am not justified in believing that spiders areharmless. But once both sets of evidence are available to me, then since thescientific data is compelling, I will acquire the belief that spiders are harmless. Itis plausible to think that in this situation I would cease to be justified in thinkingthat I don’t hold this belief.

5 . THE SOLUTION TO MOORE’S PARADOX IN BELIEF

One highly plausible principle is

Justification-distribution: If I am justified in believing that (p & q) then Iam justified in believing that p and I am justified in believing that q.

¹⁰ I owe this objection to the same anonymous referee for this volume.¹¹ I thank Mitchell Green for this objection.


For illustration, if my apparent perception of hot and humid weather justifies mein believing that it is hot and humid then that same apparent perception justifiesme in believing that it is hot and also justifies me in believing that it is humid.Of course, examples can’t prove a principle, but the fact that the principle resistsour best efforts to falsify it by counterexample, highly confirms it. That in turnmakes it plausible that the principle is not just a psychological generalization butspecifies part of the nature of justification and belief.

Now suppose that I believe that (it is raining but I don’t believe that itis raining). Suppose too, for reductio, that I am justified in believing this. Byjustification-distribution, I am justified in believing that I don’t believe that it israining. But by the same principle, I am equally justified in believing that it israining. So by Evans’s principle, I am justified in believing that I do believe thatit is raining. This is impossible, because anything that justifies me in believingthat something is the case (in this case that I believe that it is raining) renders meunjustified in believing that it is not the case (in this case that I don’t believe thatit is raining) and vice versa. So it is impossible for me to be justified in believinganything of the form p & I don’t believe that p.¹²

The impossibility is conceptual because it follows from three principles eachof which partly constitutes the nature of justification or belief. This explains thedistinctive absurdity of Moorean belief; it is absurd to hold a belief that I cannot,in principle, justify, despite the fact that it might be true. Beliefs in necessaryfalsehoods are differently absurd. So are beliefs for which any justification is veryunlikely to ever materialize. My belief that it is both raining and not rainingwould be absurd because I cannot, in principle, justify it. But unlike a Mooreanbelief, it cannot be true. My belief that the number of stars is even would beabsurd because in fact I have no reasonable expectation of ever finding anyjustification for it. Unlike a Moorean belief, the far-fetched supposition that

¹² Using ‘BJp’ to denote ‘I am justified in believing that p’, given

JD) BJ(p & q) → (BJp & BJq) Justification-distribution.EP) BJp → BJBp Evans’s principle.

Rat) BJp → ∼BJ∼p Impossibility of being justified in

contradictory beliefs.1. BJ(p & ∼Bp) Suppose for reductio.

2. BJp & BJ∼Bp 1, JD.

3. BJ∼Bp 2, &-elimination.

4. BJp 2, &-elimination.

5. BJBp 4, EP.

6. ∼BJ∼Bp 5, Rat.

7. BJ∼Bp & ∼BJ∼Bp 3, 6, &-introduction. Contradiction.

8. ∼ ♦BJ(p & ∼Bp) 1,7, Reductio ad absurdum.

John N. Williams 101

there is such justification is coherent. There is no contradiction in supposingthe unlikely vindication of a theory of the origin of the universe that accuratelypredicts exactly 800 sextillion extant stars.

So I am guilty of a severe contradiction-like irrationality in holding a Mooreanbelief, despite the fact that what I believe might be true of me. Moreover althoughthe bare truth of what I believe (that I have failed to believe a specific truthabout rain) constitutes no irrationality in me, to believe that I am guilty of thatspecific failure is indeed irrational unless it leads me to revise my beliefs. Thenatural way to do so would be to give up my belief that I don’t believe that it israining.

But how will such an account explain my absurdity holding the commissivebelief that (it is raining but I believe that it is not raining)? Suppose again thatI am justified in believing this. By justification-distribution, I am justified inbelieving that it is raining. I am also justified in believing that I believe that it isnot raining. But Evans’s principle will not turn this into an absurd case in whichI am justified in holding each of a pair of contradictory second-order beliefs.What is needed is the

Analogue of Evans’s principle: If I am justified in believing that p then I amjustified in believing that I do not believe that not-p.

This follows from Evans’s principle, if I am minimally rational and reflective.Suppose that I take myself as believing that p. Then if I recognize that I am at allrational, I must also take myself as not believing that not-p. By Evans’s principle,if I am justified in believing that p then I should take myself to believe that p. SoI should take myself as not believing that not-p.

There is also an independent argument for the analogous principle:

(P1) All circumstances in which I am justified in believing that p arecircumstances in which I tend to not believe that not-p.

(P2) All circumstances in which I tend to not believe that not-p arecircumstances in which I am justified in believing that I do not believethat not-p.

(C) All circumstances in which I am justified in believing that p arecircumstances in which I am justified in believing that I do not believethat not-p.

If the way that I form beliefs is not absurd then I do not tend to form contradictorybeliefs. In that case, these premises are as plausible as their original counterparts.Assuming that my apparent perceptions of rain are generally reliable, theyjustify me in thinking that it is raining. They also tend to make me refrainfrom believing that it is not raining. We might say, ‘seeing is not disbelieving’.The second premise is also plausible. My apparent perceptions of rain are alsoreliably connected with my refraining from believing that it is not raining. So


they afford me inductive reason to think that I do not believe that it is notraining.

Now suppose for reductio that I am justified in believing that (it is rainingbut I believe that it is not raining). By justification-distribution, I am justified inbelieving that I do believe that it is not raining. But by the same principle, I amalso justified in believing that it is raining. By the analogue of Evans’s principle,this means that I am justified in believing that I don’t believe that it is not raining.This is impossible. So it is impossible for me to be justified in the commissivebelief.¹³ Although the bare truth of what I believe (namely that I have made aspecific mistake about rain) constitutes no irrationality in me, to believe that Iam guilty of that specific failure is indeed irrational unless it leads me to revisemy beliefs for the better. The natural way to do so this time is to give up mybelief that it is not raining.

6 . DEFINING MOOREAN BELIEF

We are now in position to extract a definition of Moorean belief from thecommonalities of Moore’s two examples. First, both are examples of possibletruths. Just as I may have forgotten my visit to the cinema, so I may hold themistaken belief that my friend has gone out. Second, if these possible truths areactually true then what follows is that I am not omniscient or that I am fallible.This itself amounts to no irrationality on my part. My forgetfulness no moreimpugns my rationality than the fact that good evidence leads me to mistakenlybelieve that my friend has gone out. Third, as we just saw, it is impossible tobe justified in believing either possible truth. It seems plausible to think thatcombining these three essential features of Moore’s examples is sufficient for anyother belief to be relevantly similar. This gives us the proposal that:

¹³ GivenJD) BJ(p & q) → (BJp & BJq) Justification-distribution.AEP) BJp → BJ∼B∼p Analogue of Evans’s principle.

Rat) BJp → ∼BJ∼p Impossibility of being justified in contradictorybeliefs.

1. BJ(p & B∼p) Suppose for reductio.

2. BJp & BJB∼p 1, JD.

3. BJB∼p 2, &-elimination.


5. BJ∼B∼p 4, AEP.

6. ∼BJ∼B∼p 3, Rat.

7. BJ∼B∼p & ∼BJ∼B∼p 5, 6, &-introduction. Contradiction.

8. ∼ ♦BJ(p & B∼p) 1,7, Reductio ad absurdum.


Any belief is Moorean just in case its content is

(i) a possible truth that

(ii) constitutes no irrationality in the believer

but

(iii) it is impossible to be justified in that belief.

Support for this proposed definition comes from the fact that it makes intuitivelycorrect exclusions. Condition (i) correctly excludes beliefs in self-contradictionssuch as

It is raining and not raining

from being Moorean. Condition (ii) excludes beliefs the truth of which constitutesirrationality in me, such as

It is raining but I believe that it is raining without the least justification.

Condition (iii) excludes beliefs that I might justifiably hold, such as

I am asserting nothing now.

After all, I might be justified in quietly believing in my continuing obedience toa vow of silence. It also excludes

I have at least one false belief.

My believing this guarantees that it is true. For if that belief is false, then noneof my beliefs is false. So all my beliefs are true, including my belief that I have atleast one false belief. This means I have inconsistent beliefs, namely a set of beliefsthat cannot all be true. But it also means that I cannot be mistaken in believingthat I have at least one false belief. Since I almost certainly have some false beliefsanyway, my belief that this is so represents a rational motive for finding outwhich beliefs they are, notably by looking again at the quality of evidence.

Inconsistency in my beliefs need not undermine my justification in the waythat contradictory beliefs would. Justification for my belief that I have at leastone false belief, such as the fact that I have held false beliefs in the past, need notcount against any particular one of the vast number of beliefs I now hold. Forthe truth of my belief that at least one of my beliefs is false, does not entail that Ihave beliefs that contradict each other. Clearly, I need not believe that all of mybeliefs are true, for I can see that this would count as hubris.

7 . APPLYING THE SOLUTION TO OTHER MOOREAN

BELIEFS

We may now apply this account to other Moorean beliefs. But first we shoulduse the definition above to decide if the absurdity of a given belief is reallyMoorean. Consider


All my present beliefs are mistaken.

This can be read in two ways. If we read it as

All my present beliefs (excluding this) are mistaken

then this might be true of me, but then I am not irrational in believing it. For itmight report the fact that I have just learned that I am the victim of systematicbut convincing delusion. Moreover, I might be justified in believing that this isso. Such knowledge might well improve my rationality. So on this reading it isnot Moorean, because it is excluded by condition (iii). On the other hand if weread it as

All my present beliefs (including this) are mistaken

then it is not a possible truth. For if it is true then my belief of it is mistaken, soit is false. So it is excluded by condition (i). So on no reading is this Moorean.By contrast, my belief that

I have no beliefs

intuitively shares the paradigmatic absurdity, as predicted by the definition,despite the fact that it is not a belief in a conjunction.

We may now explain its absurdity. Suppose for reductio that I am justifiedin believing that I have no beliefs. Then by Evans’s principle, I am justified inbelieving that I have at least one belief, namely the belief that I have no beliefs.This is impossible, because I cannot be justified in holding each of a pair ofcontradictory beliefs (that I have no beliefs and that I have at least one belief ).

Or suppose for reductio that I am justified in believing that God knows thatI am not a theist. Then I would be justified in believing that I do not believethat God exists. But I would also be justified in believing that God exists. Soby Evans’s principle, I would be justified in believing that I do believe that Godexists. But it is impossible to enjoy the same justification for each of a pair ofcontradictory beliefs. Parallel reasoning from the analogue of Evans’s principle,shows that it is impossible for me to be justified in believing that God knows thatI am an atheist, for then I would be both justified in believing that I do believethat God does not exist and justified in believing that I do not believe that Goddoes not exist.

8 . SORENSEN’S ACCOUNT OF HIS ITERATED CASES

We are now well placed to consider Sorensen’s comment that om1 or com1 seemless absurd to believe or assert than om or com and that as iteration increases,only omissive absurdity appears to decrease. Using the notation ‘Bna∼p’ wherethe superscript denotes the number of iterations of the belief-operator (so that


‘B3ap’ means that ‘A believes that he believes that he believes that p’) Sorensenwrites (2000: 42):

My solution endorses the intuition that ‘p & Bn ∼ p’ is a Moorean sentence for all n, but‘p & ∼Bnp’ need not be a Moorean sentence when n is a large number. ‘p & Bna∼ p’does not entail that a has a specifiable directly opposed belief. But ‘Ba(p & Bna ∼ p)’entails that a has directly opposed beliefs about p, under the assumption that a believesthe consequences of his beliefs and that ‘p & Bna ∼ p’ is true. This entailment followsdirectly for n = 1. When n >1, the entailment is secured by a necessary condition forself-attributing higher-order beliefs.

The condition in question is that if I believem+n that p then I believen that p(2000: 39–42) namely a recursive application of the principle of

Belief elimination: If I believe that I believe that p then I believe that p.

Sorensen appeals to this principle together with the principle that

Belief is closed under logical consequence: If a logical consequence of thefact that p is that q, and I believe that p, then I believe that q.

It follows that I cannot hold a true belief in com1 unless I hold contradictory, or‘directly opposed’ beliefs about whether p. For if I believe that (p & I believe thatI believe that not-p), then a logical consequence of what I believe is that p, so Ibelieve that p. But if my belief in com1 is true then I believe that I believe thatnot-p, in which case the principle of belief elimination ensures that I believe thatnot-p. Since that principle may be applied recursively, the same diagnosis of theabsurdity will hold for any order of iteration of the belief-operator, as in com4. Italso applies to com, in which case the principle is not needed. Sorensen’s accountdiagnoses no such absurdity in om1. If I believe that (p & I don’t believe that Ibelieve that p), then a logical consequence of what I believe is that p, so I believethat p. But if my belief in om1 is true then I don’t believe that I believe that p, inwhich case the principle of belief elimination fails to apply.

But there are a number of problems with this account. First, Sorensen mustexplain the absurdity of om as follows: If I believe that (p & I don’t believe thatp), then a logical consequence of what I believe is that p, so I believe that p. Butif my belief in om is true then I don’t believe that p. But this is not, as Sorensenseems to think, a case of contradictory beliefs but rather a flat contradiction.Second, I do seem to be absurd in some way if I believe om1. For example, itwould be absurd of me to believe that

God exists but I think I am not a believer.

Finally, Sorensen’s appeal to the truth of the principle that belief is closed underlogical consequence is problematic. It is clear that it fails as a psychologicalprinciple. I may believe that a triangle is equilateral without believing that it isequiangular. Nor can it be true of me as a principle of ideal rationality, given


Searle’s Principle (1992: 155–62): If I believe that p then I have the abilityto think the occurrent thought that p.

This principle explains why although we may intuitively suppose that a doghas rudimentary beliefs about the food in its bowl (which helps us explain itsbehaviour as it strains at its leash) we hesitate to attribute to it the belief that itwill be beaten in Lent. Clearly, it does not have the concept of Lent and so lacksthe ability to think thoughts of Lent. The requirement also explains our difficultyin characterizing the beliefs of other species in any fine-grained way, since it isdifficult to specify, using the linguistic expressions of our thoughts, exactly whatconcepts (or derivatively, thoughts) are available to those with radically differentlinguistic capacities and ways of behaving.

Now suppose that I believe that Singapore is a democracy but have no ideawhat a plutocracy is. Since I cannot think the occurrent thought that Singaporeis a plutocracy, I cannot think the occurrent thought that it is either a democracyor a plutocracy. So by Searle’s principle, I cannot believe that Singapore is eithera democracy or a plutocracy. True, this will not stop me believing that thesentence ‘Singapore is either a democracy or a plutocracy’ states some truth orother. But that would be a different belief altogether. For one thing, it is notnecessary that the sentence in question state what it does, because we couldhave used ‘plutocracy’ the way we now use ‘workers’ state’. Although I cannotthink thoughts of plutocracy, I may still think the thought that a sentence thatI don’t understand, states some truth or other. My ignorance of what counts asa plutocracy is an indictment of my knowledge, but hardly counts as a failureof ideal rationality. So if I am a maximally rational thinker, but I have no ideawhat a plutocracy is, I will see that my belief in the democracy of Singaporeentails some truth or other that is stated by the sentence ‘Singapore is eithera democracy or a plutocracy’. Moreover, I will believe that this sentence statessome truth or other. But I will not know which truth is stated. Nor will Ibelieve that truth. This falsifies Sorensen’s principle, even as a principle of idealrationality.

A defender of Sorensen might endorse Robert Audi’s (1994) distinctionbetween a dispositional belief and a disposition to believe. Then she couldclaim that Sorensen’s principle should be understood as saying that a maximallyrational person is disposed to believe the logical consequences of her originalbeliefs.¹⁴

But this move won’t help Sorensen. Since I am unable to think thoughts aboutplutocracies, I cannot form any beliefs about them (as opposed to forming beliefsabout sentences). But then I cannot be disposed to form such beliefs either,because surely I cannot be disposed to form what I am unable to form. Moreover,since an infinite series of similar disjunctions are entailed by a first disjunct, the

¹⁴ As suggested by a second anonymous referee for this volume.


principle compels us to say that an ideally rational believer would believe (orbe disposed to believe) each of an infinite series of similar disjunctions. Searle’sprinciple falsifies this prediction in the case of an ideally rational thinker thatlacks the concepts needed to think the second disjunct.

Sorensen tries to circumvent this difficulty by making my ‘thorough obedience’to the principle a test of my degree of ideal rationality (1988: 37). But althoughwe might admit that degrees of rationality are vague, surely there is a differencebetween total obedience and none. So what is missing from Sorensen’s accountis a principled degree of disobedience to the principle that is distinctive of thedegree of Moorean irrationality. This means that Moorean absurdity cannot beexplained in terms of the falsehood of the principle either. Is my failure to believethe logical consequences of all of my beliefs, a form of irrationality? If so, it is avery mild form. My failure to be disposed to form such beliefs seems even milder.By contrast, a Moorean believer is guilty of a severe irrationality.

9 . THE MOOREAN ABSURDITY OF BELIEF IN THE

ITERATED CASES

Is a belief in om1 or com1 Moorean as defined above? As an instance ofom1 consider:

It is raining but I don’t believe that I believe that it is raining.

This is a possible truth that constitutes no irrationality in me. Suppose that I haveno way of discovering the truth that it is raining because I have been incarceratedin a sealed room. Then I might reasonably withhold the belief that it is rainingby suspending judgement either way, without mistakenly thinking that I believethat it is raining. As an instance of com1, consider:

It is raining but I believe that I believe that it is not raining.

Again this is a possible truth that constitutes no irrationality in me. Suppose thatwhile incarcerated, my captors fool me with the utterly convincing illusion of dryweather. Then I might be justified, not only in believing that it is not raining,but also in thinking that I hold this second-order belief.

So whether belief in om1 and com1 is Moorean depends upon whether itis possible to be justified in holding such a belief. The Evans-inspired accountvindicates the intuition that this is impossible. Take om1 first. Suppose forreductio that I am justified in believing that (it is raining but I do not believethat I believe that it is raining). By justification-distribution, I am justified inbelieving that I do not believe that I believe that it is raining. By the sameprinciple, I am also justified in believing that it is raining. By a double applicationof Evans’s principle, it follows that I am justified in believing that I do believethat I believe that it is raining. Since this is impossible, I cannot be justified


in believing that (it is raining but I don’t believe that I believe that it israining).¹⁵

Now deal with com1: Suppose for reductio that I am justified in believing that(it is raining but I believe that I believe that it is not raining). By justification-distribution, I am justified in believing that I do believe that I believe that it isnot raining. By the same principle, I am also justified in believing that it raining.By a double application of the analogue of Evans’s principle, it follows that I amjustified in believing that I do not believe that I believe that I believe that it isnot raining. Since this is impossible, I cannot be justified in believing that (it israining but I believe that I believe that it is not raining).¹⁶

¹⁵ GivenJD) BJ(p & q) → (BJp & BJq) Justification-distribution.EP) BJp → BJBp Evans’s principle.

Rat) BJp → ∼ BJ∼p Impossibility of being justified in contradictory beliefs.

1. BJ(p & ∼BBp) Suppose for reductio.

2. BJp & BJ∼BBp 1, JD.


4. BJ∼BBp 2, &-elimination.

5. BJBp 3, EP.

6. BJBBp 5, EP.

7. ∼BJBBp 4, Rat.

8. BJBBp & ∼BJBBp 6, 7, &-introduction. Contradiction.

9. ∼ ♦BJ (p & ∼BBp) 1, 8, Reductio ad absurdum.

¹⁶ GivenJD) BJ(p & q) → (BJp & BJq) Justification-distribution.AEP) BJp → BJ∼B∼p Analogue of Evans’s principle.

Rat) BJp →∼ BJ∼p Impossibility of being justified in contradictorybeliefs.

DN) Double negation under the scope of belief operators may be eliminated.

1. BJ(p & BB∼p) Suppose for reductio.2. BJp & BJBB∼p 1, JD.


4. BJBB∼p 2, &-elimination.

5. BJ∼B∼p 3, AEP.

6. BJ∼B∼∼B∼p 5, AEP.

7. BJ∼BB∼p 6, DN.

8. ∼BJ∼BB∼p 4, Rat.

9. BJ∼BB∼p & ∼BJ∼BB∼p 7, 8, &-introduction. Contradiction.

10. ∼ ♦BJ(p & BB∼p) 1, 9, Reductio ad absurdum.


Since both Evans’s principle and its analogue may be applied recursively, itfollows that it is impossible to be justified in believing om4 or com4. Indeed thisholds for any iteration of belief operators. So a belief in any iteration of om orcom is equally Moorean.

But doesn’t this result contradict Sorensen’s observation that as iterationincreases, omissive absurdity appears to decrease, while commissive type absurditydoes not? In fact the contradiction is only apparent. To appreciate this, recall thepoint that I might believe that Singapore is democracy but fail to believe that itis either a democracy or a plutocracy because I have no idea what a plutocracy is.

Having granted Searle’s principle, now suppose that my belief that I believethat it is raining is iterated in the following series:

I believe that I believe that I believe that it is rainingI believe that I believe that I believe that I believe that it is raining. . . and so on.

Although I do not lose the concepts of rain, belief, or of myself as the seriesprogresses, eventually the sheer complexity of the iteration will prevent anyhuman being from thinking thoughts of it. The same complexity will eventuallyrender me incapable of distinguishing any would-be nth-order beliefs from anywould-be n−1th order belief. Moreover, it could be argued that this renders meunable to understand the content of either would-be belief.

Now suppose, for the sake of argument, that the point at which I am firstunable to think the relevant thought in the series is marked by the millionthiteration of the belief. Suppose also that I am justified in believing that it israining. Consistently with Evans’s principle, I am still justified in holding aone-millionth-iterated belief that it is raining, but only in the counterfactualsense that if I could form such a belief then I would be justified in holding it. Inthat case it would be absurd of me to believe that

It is raining but I don’t believe that I believe that . . . I believe that it israining

where the iteration is a million because I could not possibly enjoy justificationfor holding it.

What explains the apparent absence of absurdity is an easily made confusionbetween the supposition that I hold such a belief and the supposition that I holdthe different but non-absurd belief that

It is raining but I don’t hold a one-millionth-iterated belief that it is raining.

By contrast, I would be justified in holding such a belief, since I may sensiblyrecognize the fact that despite the rain, only God could form beliefs of suchcomplexity. The abbreviated thought of a ‘one-millionth-iterated belief ’ that weformed in considering the series above is not particularly complex, as opposedto the thought that we would have formed in actually holding a one-millionthiterated belief. So my inability to think one-millionth-iterated thoughts does not


prevent me from thinking the thought of myself holding a belief that is iterateda million times.

10. THE ABSURDITY OF MOOREAN ASSERTION

When I make a Moorean assertion to you, no absurdity resides in your beliefthat my words are true. For you can quite sensibly believe that I am ignorantin a specific way or that I hold a specific mistaken belief. But my full intentionin making an assertion to you is not simply to get you to believe my words. Inattempting to inform or let you know that p, I intend to get you to know thatp. When I lie to you that p, I intend to get you to mistakenly believe that p. Inany such case I intend to get you to believe my words. But I cannot succeedin this attempt unless I also get you to think that I am sincere in making theassertion. For if you think that I’m play-acting or recognize that I’m lying, thenyou have no reason to accept my words, so my attempt to impart knowledgeor lie to you will fail. Since I should see with minimal reflection that this is so,my full intention must be to get you to believe my words by getting you to thinkme sincere in uttering them. It follows that I must intend to get you to believethat I am sincerely telling the truth. We may call this the intention to ‘believeme’.¹⁷ When my assertion is Moorean, this aim is necessarily frustrated. It seemsuncontroversial that

Assertion distributes over conjunction: If I assert that (p and q) then I assertthat p and I assert that q.

For illustration, if I tell you that it is hot and humid then surely I tell you thatit is hot and tell you that it is humid. So if I tell you that (p and I don’t believethat p) then I tell you that p. So in virtue of believing me sincere in asserting thefirst conjunct, you must think that I do believe that p. But by the same principle,I also tell you that I don’t believe that p. So in virtue of believing that I tell thetruth in asserting the second conjunct, you must think that I do not believe thatp. So you must have contradictory beliefs if you believe me.

Now suppose that I make the commissive assertion to you that (p and I believethat not-p). Since assertion distributes over conjunction, I tell you that p. So invirtue of believing me sincere in asserting the first conjunct, you must again thinkthat I believe that p. But I also tell you that I believe that not-p. So in virtue ofbelieving that I tell the truth in asserting the second conjunct, you must think thatI believe that not-p. So this time you must think that I have contradictory beliefs.

¹⁷ Although intuitions about what counts as ‘believing me’ are not robust, there is reason tothink that taking it to constitute believing that I am a sincere truth-teller is not just a convenientstipulation. For if you don’t believe what I say then clearly you won’t believe me. Nor will youbelieve me, as opposed to merely believing my words, if you accept my words but know that I ammerely parroting information or inadvertently telling the truth in an attempt to deceive you thathas failed because I have got my facts wrong.


This itself is no obstacle to your believing me. Perhaps you are prepared toacquire contradictory beliefs or ascribe them to me. But when I attempt tocommunicate with you by making an assertion, I should assume that we wouldboth charitably avoid such ascriptions if possible. On this assumption I am ina position to see with minimal reflection that my plan to be believed, in otherwords to be thought a sincere truth-teller, is bound to fail. So it is practicallyirrational of me to go ahead and make the assertion.

I admit that there are two types of assertion that do not fit the central accountof assertion I have just given. For if I make either type of assertion to you, I donot aim to make you think I am sincerely telling the truth. The first case is whenI say something to you just to ‘wind you up’. For example, suppose that I knowthat you think that North Korea has nuclear weapons, an opinion I in fact share.Nonetheless I insincerely state that it has no such weapons just to ‘rattle yourcage’. Here my intention is to get you to believe that I am sincerely asserting afalsehood in order to keep you verbally opposed to my words. The second casearises when on learning that you have just discovered that I am a habitual liar, Idecide to tell the truth for once. So when you ask me if the pubs are still open, Itell you the truth that they are, in order to deceive you into mistakenly thinkingthat they are not. Here my intention is to get you to believe that I am insincerelyasserting a falsehood.

Since I have explained the absurdity of Moorean assertion in terms of thecentral account, the objection arises that we may coin Moorean assertions of thetwo non-central types. Then the central account will not be able to explain theabsurdity in terms of the assertor’s intention to be thought a sincere truth-teller.

This is perfectly true. But my account of Moorean assertion still has theresources to explain the absurdity of ‘winding-up’ or ‘double-bluff ’ Mooreanassertions. In the winding-up case, I can hardly hope to prolong verbal disagree-ment with you unless you think (mistakenly) that I’m sincere. But when mywinding-up assertion is Moorean, I am in a position to see that you couldn’t takeme to hold a Moorean belief unless you thought I was irrational. So although Icould still irritate you by pretending to be mad, I could not sensibly try to annoyyou by making you think that we are divided in opinion.

In the second case, my intention in asserting that p is to get you to mistakenlybelieve that not-p. This means that I myself believe that p. But when mydouble-bluff assertion is Moorean, I cannot rationally believe what I assert.

Moreover, my attempt to make you think me insincere is parasitic upon myexpectation that you will normally think me sincere. This is precisely why it is adouble-bluff. So the full description of such an assertion includes the fact thatwhen I assert to you that p, I intend to get you to mistakenly believe that I’minsincere because I know that normally I will get you to think I am sincere. Butwhen my double-bluff assertion is Moorean, I know no such thing, because thereis no normal case in which I can sensibly try to make you think I hold a Mooreanbelief.


My central analysis easily explains the absurdity of the Moorean assertionsthat we considered at the start. Suppose that you believe me when I tell you thatI have no beliefs. In virtue of accepting my sincerity, you must believe that Ihave at least one belief, namely that in my own words. But in virtue of acceptingthe truth of what I say, you must also believe that I have no beliefs. So if youare to believe me then you must hold contradictory beliefs about my beliefs. Orsuppose that you believe me when I tell you that God knows that I am not atheist. In virtue of accepting my sincerity, you must believe that I do believe thatGod exists. But in virtue of accepting the truth of what I say, you must alsobelieve that I do not believe that God exists. So if you are to believe me then youmust hold contradictory beliefs about my religious convictions. Finally, supposethat you believe me when I tell you that God knows that I am an atheist. Invirtue of accepting my sincerity, you must believe that I believe that God exists.But in virtue of accepting the truth of what I say, you must also believe that Ibelieve that God does not exist. So if you believe me this time then you mustthink that I hold contradictory beliefs about the existence of God.

11. THE DIFFERENTIAL ABSURDITY OF ASSERTING

SORENSEN’S ITERATED CASES

I have argued that believing om1 or com1 is equally absurd as believing om orcom. But asserting om1 or com1 does seem less absurd than asserting om or com.Moreover, as iteration increases, the absurdity of the omissive assertion doesappear to decrease, unlike that of the commissive assertion. My account explainswhy this is so. As we saw, when I assert om to you, you can only believe me bysacrificing your own rationality in acquiring contradictory beliefs. You need notmake this sacrifice in order to believe me when I assert

om1) p & I don’t believe that I believe that p.

Suppose that you do believe me. Since you think me sincere in asserting thefirst conjunct, you think that I believe that p. And since you believe what I sayin the second conjunct, you think that I don’t believe that I believe that p. Inother words, you must think that I have a belief that I don’t think I have. Sincethis involves no sacrifice of rationality on your part, I am correspondingly morecredible.

I also become more credible when I assert com1 instead of com. As we saw,when I assert com to you, you can only believe me by thinking that I holdcontradictory beliefs. This is a harsher judgment than the one you must pass onme if you believe me when I assert

com1) p & I believe that I believe that not-p.

Suppose that you do believe me. Then since you think me sincere in asserting thefirst conjunct, you think that I believe that p. And since you believe what I say in


the second conjunct, you think that I believe that I believe that not-p. So if youare to believe me, you must judge that I really hold a belief that contradicts whatI think I believe. Given your charity in withholding the judgement that I havecontradictory beliefs, you may only make the criticism that I hold a mistakenbelief about what I believe. This seems a milder criticism than the charge that Ihold contradictory beliefs. Given that I am more credible when charity demandsa lesser criticism, I am more credible in asserting com1 than in asserting com.

The absurdity of omissive assertion decreases as iteration increases. That I failto have a one-millionth-iterated belief in what I really do believe is perfectlycredible, since neither you nor I can humanly hold such beliefs. My assertionbecomes a truthful report of psychological limits. By contrast, my credibilityin making commissive assertions is not strengthened by further iteration. Yourjudgement that I have a belief that contradicts what I take myself (over a millioniterations) to believe should be that I am still at fault to the same degree. Myiterated belief still commits me to a belief that is iterated one order less, and soon back down the series until I am committed to contradictory beliefs. If youbelieve me when I assert (com1,00,000), you are still in a position to see that I canavoid contradictory beliefs only if, somewhere in the series, I take myself to havea belief that I don’t in fact have.

REFERENCES

Audi, R. (1994), ‘Dispositional Beliefs and Dispositions to Believe’, Nous, 28/4: 419–434.Evans, G. (1982), The Varieties of Reference (New York: Oxford University Press).Moore, G. E. (1944), ‘Russell’s Theory of Descriptions’, in P. Schlipp, The Philosophy of

Bertrand Russell (Evanston, Ill.: Tudor), 175–225.(1942), ‘A Reply to My Critics’, in P. Schlipp (ed.), The Philosophy of G. E. Moore

(Evanston, Ill.: Tudor), 535–677.Searle, J. (1992), The Rediscovery of the Mind (Cambridge, Mass.: MIT Press).Sorensen, R. A. (1988), Blindspots (Oxford: Clarendon Press).

(2000), ‘Moore’s Problem with Iterated Belief ’, Philosophical Quarterly,50/198: 28–43.

Stalnaker, R. (1975), ‘Indicative Conditionals’, Philosophia, 75/5: 269–86.(1984), Inquiry (Cambridge, Mass.: MIT Press).

Vahid, H. (2005), ‘Moore’s Paradox and Evans’s Principle: A Reply to Williams’, Analysis,65/4: 337–41.

Williams, J. N. (2004), ‘Moore’s Paradoxes, Evans’s Principle and Self-Knowledge’,Analysis, 64/4: 348–53.

(2006), ‘In Defence of an Argument for Evans’s Principle: A Rejoinder to Vahid’,Analysis, 66/2: 167–70.

PART III

MOORE’S PARADOX, BELIEF, ANDASSERTION

6

What Reflexive Pronouns Tell Us aboutBelief: A New Moore’s Paradox De Se,

Rationality, and Privileged Access

Jay David Atlas

I . SOME THIRD-PERSON, LOGICO-LINGUISTIC

OBSERVATIONS: QUINE, GEACH, ET AL.

Quine (1960) raised questions in the 1940s about the referential occurrences ofsingular terms in lower clauses in third-person, present tense, indicative sentencesthat contain intentional verbs, e.g. ‘believes’. In sentence

(1) John believes that Gray Davis was re-elected governor of California.

Quine distinguished between a ‘‘notional sense’’ of ‘believes’ and a ‘‘relationalsense’’ of ‘believes’. For the notional sense he considered several paraphrases: apropositional attitude, understood first as a two-place relation between John andan intension, (2a), a relation between John and a paraphrase of the lower clauseby an eternal sentence, (2b), a three-place relation between John and an eternalsentence in the belief-ascriber’s idiolect, (2c), and, finally, a predicate (a general

The first version of this paper was presented to the Colloquium, Department of Philosophy,California State University, Northridge, April 30, 2003, under the title ‘‘The Language of First-Person Belief: Grammar and Privileged Access.’’ A later, shortened version was presented to theInstitute for Cognitive Science, State University of New York at Buffalo, November 17, 2004. I amgrateful to Takashi Yagisawa, Gregory Trianosky, and to their colleagues at CSUN for commentsand stimulating discussion, and to the audience at the University of Buffalo for the same. Inparticular I am indebted to Elisabeth Camp, Martin Hackl, John Kearns, Jean-Pierre Koenig,Gail Mauner, William J. Rapaport, David Rosenthal, Peter Ross, Erwin Segal, Stuart Shapiro,John Williams, and Richard Zubin for comments, and to Norton Starr, Dan Velleman, AlexanderGeorge, Jonathan Vogel, William B. Starr, Robert and Phyllis Sleigh, and Edmund Gettier ofthe Department of Mathematics and the Department of Philosophy, Amherst College, and of theDepartment of Philosophy, University of Massachusetts, Amherst, for their hospitality—intellectualand institutional—during the Fall Semester, 2004–5.

118 What Reflexive Pronouns Tell Us

term) true of John, (2d). Quine expressed the relational sense of ‘believes’ as athree-place relation among John, the attribute of being re-elected governor ofCalifornia, and Gray Davis in (3). The relational senses allow quantification onthe occurrence of ‘Gray Davis’ in the lower clause of (1), as in (4), while (2a),(2b), (2c), and (2d) do not.

(2) a. John believes [Gray Davis was re-elected governor of California].b. John believes-true ‘Gray Davis is re-elected governor of

California at time to’c. John believes-true ‘Gray Davis . . .’ in z’s idiolect.

(z is the asserter of the sentence)d. John believes-that-Gray-Davis-was-elected-governor-of-

California.(3) John believes x[x was re-elected governor of California] of Gray

Davis.

(4) ∃y (John believes x[x was re-elected governor of California] of y).

These observations generated and continue to generate a vast and sophisticatedphilosophical enterprise—the semantic analysis of belief attributions, or the‘‘logic’’ of ‘believes’ sentences.

Beginning in the late 1950s Peter Geach (1957, 1972), Hector-Neri Castaneda(1966), and Arthur Prior (1967), followed by David K. Lewis (1979a), andJohn Perry (1977), made linguistic observations that raised quite differentissues, semantically ill-understood but having more direct connections withtraditional problems in metaphysics and the philosophy of mind. Supposewe alter example (1) so that the singular term is the same in subject pos-ition in the higher as in the lower clause, e.g. (5a) for co-referring propernames, or so that the singular terms are co-referring, e.g. the anaphoric ‘he’in (5b) and the definite description ‘the man whom every California statelegislator dislikes’ in (5c). What Geach and Castaneda noticed was that anindirect

(5) a. Gray Davis believes that Gray Davis was re-elected governor ofCalifornia.

b. Gray Davis believes that he was re-elected governor ofCalifornia.

c. Gray Davis believes that the man whom every California statelegislator dislikes was re-elected governor of California.

reflexive pronoun ‘he himself ’ could be inserted in subject position in the lowerclause of (5b) to produce (6a), with a constraining effect on interpretation. Theindirect reflexive pronoun creates a predicate the natural interpretation of whichis a one-place predicate of x in (6b), allowing binding by existential and universalquantifiers in (6c) and (6d). The force of the indirect reflexive pronoun can beapproximately paraphrased in direct discourse in (6e).

Jay David Atlas 119

(6) a. Gray Davis believes that he himself was re-elected governor ofCalifornia.

b. x believes that he himself was re-elected governor of California.c. Someone believes that he himself was re-elected governor of

California.d. Everyone believes that he himself was re-elected governor of

California.e. Gray Davis believes, ‘‘I was re-elected governor of California.’’

What was not discussed by these philosophers, unsurprisingly, was the reflexivepronoun/proper name case, produced by inserting ‘himself ’ in the subject of thelower clause of (5a):

f. Gray Davis believes that Gray Davis himself was re-elected governor ofCalifornia,

for there cannot be a proposition believed by Gray Davis that has the contentliterally expressed by:

g. <<Gray Davis himself was re-elected governor of California>>.

The sentence-string ‘*Gray Davis himself was re-elected governor of California’is ungrammatical; so it cannot literally express a propositional content.¹

The sentences in (5) possess, of course, their usual, relational, co-referentialinterpretations in (7a,b,c).

(7) a. Gray Davis believes, of Gray Davisi, that hei was re-electedgovernor of California.

b. Gray Davis believes, of Gray Davis, i.e. of himself, that he wasre-elected governor of California.

c. Gray Davis believes, of the man that every California statelegislator dislikes, that he was re-elected governor ofCalifornia.

d. Gray Davis believes [Gray Davis was re-elected governor ofCalifornia].

e. Gray Davis believes [he was re-elected governor of California].

What is logically interesting is that the de se ascription (6a) Gray Davis believesthat he himself was re-elected governor of California entails the co-referential de reascription (7a), but not conversely. Moreover, the de se ascription (6a) GrayDavis believes that he himself was re-elected governor of California neither entails

¹ Even worse grammaticality problems will afflict the corresponding sentence for (5c), since theresulting belief-sentences are themselves ungrammatical: *Gray Davis believes that the man himselfwhom every California state legislator dislikes was re-elected governor of California, or: *Gray Davisbelieves that the man whom every California state legislator dislikes himself was re-elected governorof California. Similar constraints on the grammaticality of first-person sentences will be discussedbelow in (11), where I discuss cases of stress on ‘myself ’. Stress can make the utterance-type ‘‘GrayDavis HIMSELF was re-elected governor of California’’ seem linguistically acceptable.


nor is entailed by the de dicto interpretations of the ascriptions (5a) Gray Davisbelieves that Gray Davis was re-elected governor of California, (5c) Gray Davisbelieves that the man whom every California state legislator dislikes was re-electedgovernor of California. This is clear when Gray Davis does not believe that‘Gray Davis’ denotes him, e.g. he has forgotten his name, or when he does notbelieve that ‘the man whom every California state legislator dislikes’ denoteshim. Thus the de dicto and de se ascriptions are logically independent, and thede se ascriptions entail the de re ascriptions but are not logically equivalent tode re or to de dicto, non-pronominal counterparts. (I am indebted here to theilluminating work of Boer and Lycan (1986), though they fail to consider thepronoun case (5b).)

The case of the pronoun ‘he’, the case ignored by Boer and Lycan, is moreinteresting. The pronoun ‘he’ may be anaphoric, syntactically co-indexed withits antecedent Noun Phrase, or it may not be anaphoric. Correspondingly, of(5b) we can get two de re interpretations, one of which is the non-coreferentialGray Davisi believes, of himj , that hej was re-elected governor of California and theother of which is the coreferential (7b) Gray Davisi believes, of himi —that is, ofhimself—that hei was re-elected governor of California; the de se (6a) Gray Davisbelieves that he himself was re-elected governor of California entails the de re (7b)Gray Davis believes, of himself, that he was re-elected governor of California.

How shall we construe a de dicto interpretation of (5b) Gray Davis believes thathe was re-elected governor of California? There are two possible dicta; the first isthe non-anaphoric, non-coreferential [hej was re-elected governor of California],and the second is the anaphoric, coreferential [hei was re-elected governor ofCalifornia]. When Russellian propositional attitudes were thought of as relationsbetween persons and the intensions of Quine’s eternal sentences, there was alwaysthe lurking problem of ‘‘which eternal sentence?’’ But one presupposed that whatwas meant, by the sentence or by the speaker, was one proposition or eternal-sentence-meaning. One simply would have concluded that the interpretationappropriate to (5b) depended on the analysis of the pronoun ‘he’ in the lowerclause of the sentence uttered by the speaker.

How shall we represent the dictum in (5b) that Gray Davis would be saidto believe? Here are some choices (8b-e), depending on how one construes thepronoun ‘he’ in (5b), repeated in (8a).

(8) a. He was re-elected governor of California.bi. He who is identical to Gray Davis was re-elected governor of

California.bj. He who is not identical to Gray Davis was re-elected governor

of California.c. The male person was re-elected governor of California.

Jay David Atlas 121

ci. The male person who is identical to Gray Davis was re-electedgovernor . . .

cj. The male person who is not identical to Gray Davis wasre-elected governor . . .

d. The unique male person was re-elected governor of California.²di. The unique male person who is identical to Gray Davis was

re-elected governor . . .

dj. The unique male person who is not identical to Gray Davis wasre-elected . . .

e. Gray Davis was re-elected governor of California.

We eliminate (8d) if truth-values are to be assigned correctly. One may simplifythe problem by removing the non-coreferential cases (8bj, cj, dj). That leavesthe remaining (8b’s), (8c’s), and (8e). Still, if a philosopher or a linguist asksme ‘‘what Gray Davis is said to believe’’ by one who asserts sentence (5b) GrayDavis believes that he was re-elected governor of California, my choice seems nota matter of clear linguistic intuition; it seems as much a matter of one’s theoryof pronouns. Not even the appealingly simple (8e) Gray Davis was re-electedgovernor of California is necessarily the correct answer to what is intended by ade dicto ascription of (5b). Interpretation (8e) would result from treating ‘he’ asa ‘‘pronoun of laziness,’’ one that merely goes proxy for an occurrence of theearlier Noun ‘Gray Davis’, and not as an anaphoric pronoun; in that case thedictum of (5b) would be the same as the dictum of (5a) Gray Davis believesthat Gray Davis was re-elected governor of California. But it does not seem rightto ignore the difference between anaphoric pronouns and pronouns of lazinessin the expressions of objects of belief, unless, of course, one has a theoreticalreason to do so.³ After all these complexities, if the question is what did thespeaker—the asserter—of (8a)/(5b) mean in his ascription of a belief-contentto Gray Davis, a unique, determinate, distinctive answer (i.e. an answer differentfrom (5a)) seems even less obvious.

² I do not assume that the determiner ‘the’ semantically requires uniqueness. So I distinguishbetween ‘the N’ and ‘the only N’ in (8c) and (8d).

³ Those who take the dicta to be propositions, and singular propositions in particular, would,of course, say that it does seem right that (5a) and (5b), if ‘he’ is an anaphoric pronoun, express abelief -relation between Gray Davis and the same proposition. On that interpretation (5a) and (5b)ascribe the same belief-content to Gray Davis. The singular proposition theorist asks: Why shouldnot pronouns of laziness and anaphoric pronouns coincidentally yield the same proposition as theobject of belief ? One tends to answer, rather simple-mindedly one fears: the propositional conceptsexpressed by the pronoun and by the proper name sentences would seem to be different. Whyshould not these different linguistic entities express different ways of believing or different objectsof belief in the mind of Gray Davis?


I I . SOME FIRST-PERSON, LOGICO-LINGUISTIC

OBSERVATIONS

The permutations of first-person and third-person in (9) are all grammatical, butthe permutations in the de se (10) are not. By contrast with the grammatical (10a)I believe that I myself was re-elected governor of California, in sentence (10c) *Johnbelieves that I myself was re-elected governor of California the indirect, first-personreflexive pronoun in the lower clause cannot grammatically co-occur with anon-first-person higher subject Noun Phrase, even if there are speakers named‘John’ who might utter the ungrammatical (10c) with a semantic interpretationidentical to that of an utterance of the grammatical (10a). The grammaticalityof the English sentence cannot depend on the existence of speakers named‘John’—a point reminiscent of a view of Wittgenstein’s in the Tractatus Logico-Philosophicus. This way of contrasting (10a) and (10c) treats them respectively asgrammatical and ungrammatical sentence-strings.

(9) a. I believe that I was re-elected governor ofCalifornia.

b. John believes that John was re-elected governor ofCalifornia.

c. John believes that I was re-elected governor ofCalifornia.

d. I believe that John was re-elected governor ofCalifornia.

(10) a. I believe that I myself was re-elected governor of California.b. John believes that he himself was re-elected governor of

JohnCalifornia.

c. *John believes that I myself was re-elected governor ofCalifornia.

d. *I believe that he himself was re-elected governor of California.e. *I believe that John himself was re-elected governor of

California.

Sentence (10d) is manifestly ungrammatical. The Proper Name cases (10c)and (10e) are more interesting. A more tolerant theory of the set of grammaticalsentence-strings might un-star (10c) on the grounds that there is some acceptableinterpretation of some utterance-type—not just of a nonce utterance-token—ofthe sentence-string, a case of which I shall discuss below. Such tolerant criteriawould admit as grammatical various ungrammatical or semantically anomalousMalapropisms, e.g. ‘It’s a ten-hour receptacle’ (from ‘It’s a ten-hour spectacle’),and ungrammatical Spoonerisms, e.g. ‘Cicero denounced the sat old fenator’(from ‘Cicero denounced the fat old senator’). The unexpected, verbal pleasure of

Jay David Atlas 123

many Spoonerisms depends on its being only an amusing accident that the resultis grammatical and meaningful: Prince Albert kissed the dear old Queen becomesPrince Albert kissed the queer old Dean. Narrower, less tolerant criteria force usto accept the possibility that ungrammatical sentences like (10c) could have anacceptable interpretation of its utterance-type. I shall return to this possibilitybelow.

Sometimes speakers do not notice the difference between the stressed andunstressed statements without having the stress brought to their attention,but the indirect, first-person reflexive pronoun cannot occur acceptably in anindependent clause unless the reflexive pronoun is contrastively stressed; compare(11a) with (11b).

(11) a. *I myself was re-elected governor of California.⁴b. I MYSELF was re-elected governor of California.c. Myself, I was re-elected governor of California.

If identity criteria for sentences include phonological information on stress andintonation, then (11a) and (11b) are distinct sentences. We can then mark (11a)as an ungrammatical sentence and (11b) as a grammatical sentence—as I havedone above.

If identity criteria for sentences are not so formulated, then we should treat(11a) and (11b) as distinct utterance-types of one sentence-type generatedby the grammar (a ‘‘system-sentence’’), and the strings in (11a,b) shouldbe double-quoted as unacceptable and acceptable utterance-types: ‘‘?I myselfwas re-elected governor of California’’, ‘‘I MYSELF was re-elected governor

⁴ Castaneda (1966: 151–4) has a splendid discussion of the ungrammaticality of ‘He himself isF’ and of ‘He himself is a’, the conclusion of which is:

(H*1) The pronoun ‘he*’ [‘he himself ’] is strictly a subordinate pronoun: it is by itself an incomplete,or syncategorematic, symbol, and every sentence or clause containing a tokenw [i.e. token in thewide sense, allowing related grammatical forms, e.g. ‘him’] of ‘he*’ which is not in oratio obliqua, isalso an incomplete or syncategorematic sentence or clause. (1966: 154)

One charming part of this discussion I cannot resist quoting, with some change in notation,though I shall resist the temptation to give a critical gloss of it here:

Doubtless, in the sentence ‘a believes that he himself is F’ the token of ‘he himself ’ does refer to theperson a. This suggests both that we could have statements of the form ‘‘he himself is a’’ and thatwhoever believed or asserted ‘‘a believes that he himself is F’’ should certainly know the truth ofthe corresponding statement ‘‘he himself is a.’’ It would seem, then, that tokens of ‘he himself ’ canappear in oratio recta. However, if there were complete statements of the form ‘‘he himself is a’’,these statements could be known to be true by the person a himself. But suppose that a, who doesnot know that he himself is a, does assert (or just thinks, for that matter) ‘‘a believes that he himselfis F.’’ That is, a could come to know that he himself is a simply by thinking that the man a has someproperty or other. For instance, the heaviest man in Europe could come to know that he himselfweighs more than anybody else without resorting at all to scales and comparison of weights!Thisabsurdity arises simply from allowing the tokensw of ‘he*’ to function as independent symbols, i.e.,as referring devices in their own right, without the need of a grammatical and logical antecedent.Hence, we must conclude that there are no complete statements of the form ‘‘he* is a.’’ That is, asentence containing a tokenw of ‘he*’ can, given ordinary meanings, formulate a statement only ifthe tokenw in question has an antecedent in the same sentence. (1966: 151–2)


of California’’.⁵ The question is now, is the one sentence-string in (11a,b)grammatical or ungrammatical? One answer is that, since the string has oneacceptable, phonological realization in (11b), the sentence-string I myself wasre-elected governor of California is grammatical. Another answer is that, since thestring has one unacceptable, phonological realization in (11a), the sentence-stringI myself was re-elected governor of California is ungrammatical.

If the sentence-string is ungrammatical in (11a,b), by contrasting it withthe grammatical sentence in (10a) I believe that I myself was re-elected gov-ernor of California, we can formulate the question: why is the complementclause part of a grammatical sentence in (10a), but if it stands alone as asentence in (11a), it is ungrammatical? First-person reflexives in lower clausesoccur grammatically, but first-person reflexives in highest clauses do not occurgrammatically.

If the sentence-string is grammatical in (11a,b), no such syntactical questionis formulable. The data would have to be described as follows: a grammaticalstring (11a,b) embeds grammatically in a higher ‘I believe . . .’; an utterance-type ‘‘I MYSELF . . .’’ of the same string embeds acceptably in an utterance‘‘I believe . . .’’ of a higher ‘I believe . . .’ sentence: ‘‘I believe that I MYSELFwas re-elected governor of California’’, and an unacceptable utterance-type‘‘?I myself was re-elected governor of California’’ of a grammatical string ‘Imyself was re-elected governor of California’ embeds in an acceptable utterance‘‘I believe that I myself was re-elected governor of California’’ of a higher,grammatical ‘I believe . . .’ sentence. This description of the data has the odditythat a grammatical string, when uttered naturally (the ‘‘unmarked’’ case), is anunacceptable utterance but when uttered with ‘‘unnatural’’ stress (the ‘‘marked’’case) is an acceptable utterance. One also would wonder why it is that prefacingan unacceptable utterance with an acceptable utterance of ‘‘I believe . . .’’ shouldmake the unacceptable utterance-type acceptable.

Contrasting the preceding two paragraphs, with their two distinct formulationsof the data, one understands why one might prefer, for clarity and coherence,the first of the two. Moreover, as (11c) shows, a first-person sentence with aleft-dislocated first-person, reflexive pronoun is grammatical; (11c) exhibits asyntactical means of accomplishing in a sentence what stress accomplishes inthe utterance-type (11b). (Utterances of the sentence (11c), with normal stressand intonation, are perfectly acceptable, of course.) But then one must note adistinction between the forms of the sentences in (11a,b) and (11c). The first isungrammatical; the second is grammatical.

In Quine’s relational sense of (1) John believes that Gray Davis was re-elected governor of California, viz. (3) John believes x[x was re-elected governorof California] of Gray Davis, ‘John’ and ‘Gray Davis’ occur transparently in a

⁵ For more on the difference that contrastive stress can make, see Atlas and Levinson’s (1981:16–18) discussion of the difference in Noun Phrase focus of simple declarative and cleft statements.

Jay David Atlas 125

three-place ‘believes’ predicate that is true or false of <John, Gray Davis, x[x wasre-elected governor of California]>.

(12) a. I believe of myself that I was re-elected governor of California.b. *I believe of myself that he was re-elected governor of

California.c. *I believe of myself that John was re-elected governor of

California.d. I believe of John that he was re-elected governor of California.e. *I believe of John that I was re-elected governor of California.

A first-person, subject Noun Phrase case would be (12a), where ‘I’ and‘myself ’ occur transparently in a three-place ‘believes’ predicate true or falseof <I, myself, x[x was re-elected governor of California]>. There are somesentences that are manifestly ungrammatical, e.g. (12b), even if a speak-er named ‘John’ who asserts (12c) or (12e) would assert an ungrammaticalsentence-token whose assertion nonetheless would generate an accommodatedinference that I (the speaker) am identical to John.⁶ For the reasons discussedearlier in this section, the Proper Name sentences (12c) and (12e) are bestregarded as ungrammatical even though they possess accommodated utterance-interpretations for tokens of them uttered by a speaker named ‘John’. Thealternative is to think of them as grammatical sentence-types because there existsuch interpretable utterance-tokens; I have rejected that option earlier in thissection.⁷

⁶ I use ‘accommodate’ in the technical sense of ‘presuppose in order to make possible aninterpretation of an infelicitous or ungrammatical assertion’. See Lewis (1979b); Atlas (2004, 2005).

⁷ Castaneda (1966: 146–7) mentions cases of non-anaphoric, referential occurrences of ‘me’with a non-first-person subject: Johni believes of mej that I j am F, where he takes ‘I’ to be merely asyntactic artifact, dependent for its interpretation on ‘me’, and cases of non-anaphoric uses of ‘I’ inthe lower ‘believes’ clause (where there cannot be co-reference with the subject NP): Johni believesthat I j am F, the latter of which he calls ‘directly self-referring uses of ‘I’ ’ that refer to the speakerof the sentence, as in (9c). Oddly enough, he does not consider the first-person subject cases. Healso never mentions the ‘I myself ’ case of (10a), whether the subject NP is first-person or not. Hedoes say of the two sentences that he mentions that ‘‘ . . . the difference . . . is very intriguing. Butwe do not have to dwell upon it here. We need note only that it lies in the claim’’ made by thelatter sentence, but not the former, that John has a way of referring to the speaker of the sentencethat enables John to ‘‘pick out’’ or ‘‘identify’’ the speaker as the individual whom John believes tobe F (Castaneda 1966: 150). Or to put it slightly differently, Johni believes that I j am F ‘‘makes theclaim’’ that John knows who the speaker is, while the de re, ‘of me’ version does not. Castaneda’sintuition about the use and implications of the former sentence makes the sentence with the directlyself-referring use of ‘I’ a semantical hybrid: in some sense it is both de dicto and de re. Castanedainterprets the sentence to express a relation between John and a content expressible by <<�isF>>, for some �, where ‘�’ ranges over identifying referring expressions by which John picksout the speaker, who in uttering the sentence refers to himself by ‘I’. This was an insight into thesemantics of the indexical ‘I’. By contrast with the intended de re sentence Johni believes of mejthat I j am F, there is surely something to Castaneda’s intuition that the speaker conversationallyimplicates that John knows who he or she is in asserting Johni believes that I j am F. The question iswhether it is more than a conversational implicature, whether, as Castaneda seems to have intended,it is entailed. I believe that it is a conversational implicature, but I shall not argue the point here.


I I I . SENTENCE-TYPES VERSUS UTTERANCE-TYPES

The distinction between sentence-types and utterance-types (see Atlas 1989;Bach 1987; Levinson 1995) has not been given sufficient attention in thesecases, since philosophers of language and linguists have not examined carefullythe mixed cases of third-person and first-person, where the distinction becomesparticularly useful. Like the grammatical sentences in (9), the utterances in(13a) are acceptable. Like the ungrammatical sentence (10c), the sentence in(13b) is ungrammatical, but the utterance of (13b) can receive an acceptable,‘‘accommodated’’ interpretation. A hearer can put an acceptable interpretationon the utterance if he ‘‘accommodates’’ the co-extensiveness of ‘I myself ’and ‘Gray Davis’ in the speaker’s utterance of (13b); he infers what wouldmake the utterance acceptable, viz. that the speaker must be Gray Davis. LikeRichard Nixon, during his Presidency, referring to himself by tokens of ‘thePresident’, Gray Davis can utter tokens of (13b) in which the tokens of ‘GrayDavis’ and ‘I myself ’ are co-extensive. Likewise, the tokens of ‘myself ’ and‘Gray Davis’ are accommodated by the hearer to be co-extensive in tokens of(13c), if the utterance of the ungrammatical sentence is to have an acceptableinterpretation.

(13) a. ‘‘Gray Davis believes that Gray Davis was re-electedgovernor . . .’’

‘‘Gray Davis believes that I was re-electedgovernor . . .’’

‘‘I believe that Gray Davis was re-electedgovernor . . .’’

b. ‘‘*Gray Davis believes that I myself was re-electedgovernor ofCalifornia.’’

c. ‘‘*I believe ofmyself that

Gray Davis was re-electedgovernor ofCalifornia.’’

Recall that the logical relationships of the third-person cases were as follows:the de se ascription (6a) Gray Davis believes that he himself was re-elected governorof California entails the de re ascription (7a) Gray Davis believes of Gray Davisithat hei was re-elected governor of California, but not conversely.⁸ The de dictoascription (5a) Gray Davis believes that Gray Davis was re-elected governor ofCalifornia and de se ascription (6a) Gray Davis believes that he himself wasre-elected governor of California are logically independent.⁹

⁸ The latter claim was also made by Castaneda (1966: 139–40).⁹ This claim was also made by Castaneda (1966: 135).

Jay David Atlas 127

What about the first-person cases in (9a, 10a, 12a)? If the logical relationshipswere analogous to the third-person cases, the de se self-ascription (10a) I believethat I myself was re-elected governor of California would entail the de re self-ascription (12a) I believe of myself i that I i was re-elected governor of California,but not conversely, and the seemingly de dicto self-ascription (9a) I believe that Iwas re-elected governor of California would be logically independent of the de seself-ascription (10a) I believe that I myself was re-elected governor of California.

In fact, like the third-person case, the de se (10a) entails the de re (12a),and not conversely. But, as a matter of linguistic fact, the utterance-type ofsentence (9a) I believe that I was re-elected governor of California itself hastwo natural interpretations, a de se interpretation in which the self-ascribedbelief-content must be expressed in the first-person, viz. <>¹⁰ and a de re interpretation in which the self-ascribedbelief-content need not be expressed in the first-person, e.g. <<Gray Davis wasre-elected governor of California>>.¹¹ What is peculiar about the first-personutterance of sentence (9a) is that there is no de dicto interpretation; the naturalinterpretation is the de se interpretation. What is also peculiar to the first-personutterance is that there is a de re interpretation I believe of myself that I wasre-elected governor of California the content of which could be expressed ormentally represented objectively rather than subjectively, using a first-person,singular pronoun ‘I’ that is referentially dependent on ‘myself ’, in <>.¹²

What is notable about the first-person utterance of sentence (9a), then, is this:unlike the third-person case there is no independent de dicto interpretation of thefirst-person utterance of (9a). I summarize these observations in (14a) and (14b).

¹⁰ For example, a question to Gray Davis: ‘‘What do you believe?’’; answer: ‘‘That I was re-electedgovernor of California’’.

¹¹ How one should analyze this de re interpretation, whether it should be an expression in aFodorian Language of Thought, or whether it is a Russellian singular proposition <Gray Davis, xwas re-elected governor of California> containing in the ordered-pair Gray Davis and the predicate,or perhaps the property expressed by the predicate, is an interesting question. The simple, unadornedRussellian singular proposition cannot distinguish between these two de re and de se interpretations,so it is not a useful semantic entity for this analysis. One would need at least what I shall call athree-place Braunian singular proposition expressed by the lower clause of a ‘believes’ utterance Uin a context K(U), <Gray Davis, x was re-elected governor of California, Denotes (‘I’, Gray Davis,K(U) )>, where ‘K(U)’ denotes a conversational context in which the sentence-token U is uttered.By the way, this is not Braun’s formalism; it is just my shorthand way of making use of his ‘‘waysof believing’’ a singular proposition. See David Braun (1998, 2002). Braun does claim that propernames are substitutable salva significatione in belief attributions, but that definite descriptions arenot.

¹² What I mean by an objective use of ‘I’ is a referential use dependent on ‘myself ’ in phraseslike ‘of myself ’, ‘I hurt myself ’, ‘I kicked myself for forgetting his birthday’. Some philosophers,linguists, as well as ordinary speakers, have trouble with the notion that a de re interpretation couldbe expressed using ‘I’. I believe that their difficulty is an effect of failing to notice the objective useof ‘I’. A conclusion of this paper concerning privileged access—my answer to the question whether<Bp → BBp> is correct—does not depend solely on the availability of this interpretation; theconclusion will still follow from my observations on the de se interpretation. See n. 31.

(14) a. The utterance-type ‘‘I believe that I was re-elected governor ofCalifornia’’ has two interpretations, (i) the de se and (ii) thede re:

i. I believe that I myself was re-elected governor of California.ii. I believe of myselfi that Ii was re-elected governor of

California.(That is: I believe, of x, that x was re-elected governor of

California, wherex = the speaker of the utterance-token, the extension of the

utterance’s token ‘I’.)b. The de se interpretation of (9a) entails the de re interpretation

of (9a);there is no de dicto interpretation of the utterance-type (9a).

So, in particular, uttered by Gray Davis, an utterance of (9a) I believe that I wasre-elected governor of California actually expresses the essentially first-person de se(15(i) ) or a de re (15(ii) ), where I have refined in (15(ii) ) the de re paraphrasegiven in (14a(ii) ).

(15) Gray Davis asserts a token of I believe that I was re-elected governorof California.

Gray Davis makes a de se assertion, or a de re assertion, reported by:i. G.D. believes <>.ii. G.D. believes of G.D., i.e. of himself, the property x[x was

re-elected governor of California].

It is a linguistically natural assumption to make in interpreting the utterance-typethat there exists a genuine, de se interpretation in which the belief-contentmust be expressed using the first-person, singular pronoun. This intuition leadsimmediately to the question: what does it mean to attribute a belief-content thatis essentially expressed in the first-person?

IV. MOORE’S PARADOX, PERSPECTIVE ON THE SELF,

AND RICHARD MORAN

In order to shed some light on those questions, let us consider the conjunction(16) that contains (9a):

(16) I believe that I was re-elected governor of California, but I was not.

It will be granted that (16) resembles a peculiar utterance-type, of a sort madefamous by Cambridge University philosopher G. E. Moore in the 1940s.¹³ But

¹³ Two of the forms of Moore sentences are the familiar examples in (a,b), schematized in (c,d).

a. It’s raining, but I do not believe it.b. I believe that it’s raining, but it is not.

Jay David Atlas 129

first, let us note that (16) will have the de se and de re interpretations in (17a)and (17b). Approximate logical forms for (17) are given in (18).

(17) a. I believe that I myself was re-elected governor of California,but I was not.

b. I believe of myself that I was re-elected governor of California,but I was not.¹⁴

(18) a. B(I, <<RI>>) ∧ ¬ RI

b. BEL (I, myself, x [x was re-elected governor of California] ) ∧¬x [x was re-elected governor of California] (Imyself )

Let’s consider an English paraphrase of the de re (17b) in (19). The typicallinguistic oddity of the Moore Paradox sentence does not appear!Compare (19)with a traditional G. E. Moore utterance (20), which certainly seems linguisticallyodd. By contrast, the de re utterance (19) just sounds like a psychiatric patientreporting a persistent but acknowledged delusion, and it is not linguisticallyodd.¹⁵ The de se interpretation (21b) of (16)—repeated as (21a)—on the otherhand, sounds distinctly odd, after the Moore’s Paradox fashion, but it doesn’t

c. P ∧ ¬BI[P]d. BI[P]∧¬ P

How exactly to describe what is paradoxical in asserting sentences of these forms is up for discussion,but one observes that the assertion of a sentence of these forms certainly seems linguistically peculiar.Moore (1993: 207) himself thought that ‘‘it’s perfectly absurd or nonsensical to say such things’’.By ‘say’ he meant ‘utter a sentence assertively’. The logician C. H. Langford (1942/1968: 333) saidthat the assertion ‘‘sounds self-contradictory.’’ The linguist George Lakoff (1975: 264–5) thinksthe assertion ‘‘sounds odd’’, as does John Searle (1983: 9). Linguists and philosophers have offeredvarious explanations for this oddity. Wittgenstein (1974: 177) wrote to Moore in October 1944,just after hearing a discussion by Moore of the paradox at a meeting of the Moral Sciences Club inCambridge, ‘‘You have said something about the logic of assertion . . . [that] it makes no sense toassert ‘p is the case and I don’t believe that p is the case’. This assertion has to be ruled out . . . justas a contradiction is.’’ Wittgenstein thinks that the Moore statements are an ‘‘ ‘absurdity’ which isin fact something similar to a contradiction, though it isn’t one . . .’’. Some speakers believe thatthe past-tense form of the content-clause in (16), as contrasted with the future or past-tense formsof the ‘believe’ verb, which admittedly block the assertoric oddity of the utterance of the Mooresentence, mislead the ear and make my examples sound more acceptable than they should sound.So, in addition to (16), we shall look at sentences like (16′) I believe that I am the governor-elect ofCalifornia, but I am not. Nonetheless, I am on the side of historical tradition in my example (16);see my discussion of C. H. Langford and A. M. MacIver in Sec. VII, below.

¹⁴ (17′) a. I believe that I myself am governor-elect of California, but I am not.b. I believe of myself that I am governor-elect of California, but I am not.

¹⁵ Philosopher Greg Trianosky has confirmed my interpretation of this linguistic datum; hereports that, during a Medical Ethics Workshop at the University of Texas, Houston MedicalCenter, he observed psychiatric patients saying things of just this sort. Psycho-linguist Gail Maunerhas also pointed out to me that John Nash reports his deluded mental states in the same way; seeSylvia Nasar, A Beautiful Mind (New York: Simon and Schuster, 1998). The first-personness of thelower clause in (19) makes it much easier, for me, linguistically to accept (19) than to accept (20),which does not have a first-person lower clause. Psychologist Richard Zubin has commented to methat if (19) were uttered acceptably by a deluded patient, so would (20) be. Zubin took this claimto be a reductio argument against my view of (19). But Zubin is ignoring the linguistic differencesbetween the pronominal (19) and the non-pronominal (20).

have the form of a Moore’s Paradox sentence <BI[P] ∧ ¬ P>, which would bethe ungrammatical (21c), ungrammatical because of the ungrammatical, secondconjunct.

(19) I believe myself to have been re-elected governor of California,but I was not.¹⁶

(20) ?I believe it’s raining, but it isn’t.

(21) a. I believe that I was re-elected governor of California, but I wasnot. [16]

b. ?I believe that I myself was re-elected governor of California,but I was not.

c. *I believe that I myself was re-elected governor of California,but I myself was not.¹⁷

Thus one form of a standard Moore’s Paradox sentence, where the content-clauseis cast in the first-person, has a natural de re interpretation (19) but an odd dese interpretation (21b). The problem is that the odd de se interpretation is notan interpretation with the grammatical form of a Moore’s Paradox sentence, andthe de re interpretations of a sentence that has the grammatical form of a Moore’sParadox are not odd.

An alternate, and the original, form of Moore’s Paradox is (22). The corres-ponding interpretations for the first-person example (23a) would be the de se(23b) and the de re (23c) interpretations.

(22) ?It’s raining, but I do not believe it.

(23) a. I was re-elected governor of California, but I do not believe it.b. I was re-elected governor of California, but I do not believe

that I myself was re-elected governor of California.c. I was re-elected governor of California, but I do not believe

myself to have been re-elected governor of California.¹⁸

But the de se (23b) is not an utterance that has the form of a Moore’s Paradoxsentence. And it does not have the assertoric oddity of (22). The utterance-type(23b) could be the statement of an amnesiac Gray Davis who has not rememberedhaving been re-elected governor of California, has been told that he is Gray Davis

¹⁶ (19′) a. I believe of myself that I am governor-elect of California, but I am not.b. I believe myself to be the governor-elect of California, but I am not.

¹⁷ (21′) a. I believe that I am governor-elect of California, but I am not.b. ?I believe that I myself am governor-elect of California, but I am not.c. *I believe that I myself am governor-elect of California, but I myself am not.

¹⁸ (23′) a. I am governor-elect of California, but I do not believe it.b. I am governor-elect of California, but I do not believe that I myself am

governor-elect of California.c. I am governor-elect of California, but I do not believe myself to be

governor-elect of California.

Jay David Atlas 131

and that Gray Davis was re-elected governor of California, but being an amnesiacand not necessarily believing what he was told, would still disavow the first-personthought <>.

The case I have in mind is a case where the amnesiac patient Gray Davis is told,‘‘You are Gray Davis, and Gray Davis was re-elected governor of California.’’Here are some possibilities in this context: Gray Davis can now say, takingthe latter conjunct of the informant’s word for it, which it would be easy foran amnesiac to do, as it does not involve his own identity, ‘‘Gray Davis wasre-elected governor of California, but I do not believe that I myself was re-electedgovernor of California’’—not understanding or, if understanding, not taking inor, if taking in, not accepting from the first conjunct that he is Gray Davis—heis still an amnesiac after all and does not know who he is—and not rememberingand so rejecting in the second conjunct a thought of the form <> even when ‘I’ and ‘Gray Davis’ are co-referential inthe utterance.¹⁹

Furthermore, he can acceptably utter assertively ‘‘I was re-elected governorof California, but I do not believe that I myself was re-elected governor ofCalifornia’’ (as Moore (1993: 207) would have put it, distinguishing the casefrom a case of assertion—see n. 13), since the semantic value of the first ‘I’ is GrayDavis in this example, and Gray Davis was re-elected governor of California,even though he doubts the word of the informant that he is Gray Davis. Hecould warrantedly assert the sentence if he were willing to accept the testimony ofthe informant that he is Gray Davis and that Gray Davis was re-elected governor,even if he does not remember being re-elected governor of California.

He can felicitously assert the sentence ‘Even if I was re-elected governor ofCalifornia, I do not believe that I myself was re-elected governor of California’,which ought to be just as linguistically ‘‘odd’’ an utterance as the allegedly ‘‘odd’’Moore Paradox ‘even’ sentence ‘Even if it is raining, I do not believe that it israining’. Finally, he can felicitously assert the sentence ‘I was re-elected governorof California, but I do not believe that I myself was re-elected governor ofCalifornia’ if he is willing to humor the informant and echo him in the firstconjunct of his utterance by using the pronoun ‘I’, in what I shall call ‘anindirectly referring, echoic use of ‘I’ ’, a speaker’s use parasitic on the informant’sbeliefs about the referent of the informant’s use of ‘you’. When Gray Davis speaks,the informant’s ‘you’ shifts to the speaker’s ‘I’. ‘I’ will be used by the speakerin place of the informant’s ‘you’ to refer to the person whom the informanthas implied, falsely in the view of the utterer of ‘I’, satisfies the description ‘theperson re-elected governor of California’, which, as it happens, contrary to thebeliefs of the utterer, is himself. The speaker has a meta-linguistic belief aboutthe informant’s use of ‘you’. In light of this belief, his own use of ‘I’ will be

¹⁹ The importance of conjunctive sentences in examples of these kinds has been clear from theearliest discussions in Geach (1957, 1972) and in Castaneda (1966: 130, n.1; 145).

acceptable in a conversation with the informant, whom the patient believes tobelieve that the individual re-elected governor of California is he the patient,believing as he does that the informant believes that he the patient is Gray Davisand Gray Davis was re-elected governor of California. The linguistic complexityhere results from the utterer of ‘I’ also simultaneously using ‘I’ non-parasiticallyto refer to himself, which I shall call ‘a directly referring use of ‘I’ ’.²⁰ The twouses converge on one referent.²¹

My case of an indirect referring use is a variant of the Donnellan referring useof a definite description, applied to ‘I’. In this kind of case the speaker wouldbelieve that the hearer believes falsely that the man is drinking a martini, and sothe speaker wants to cooperate with the hearer in the conversation by maximizingagreement in their usage of referring terms, getting the reference right (from thepoint of view of the hearer) through the vehicle of the hearer’s false belief. But thespeaker’s belief is itself false, and the man he describes, incorrectly he thinks, by‘the man drinking a martini’ is actually drinking a martini. The case is differentfrom Donnellan’s, since the speaker intends to make a ‘‘referring use’’ butunintentionally succeeds in making an ‘‘attributive use’’ of a definite description.

In my amnesiac example, the speaker would be prepared to assert the firstconjunct ‘‘I was re-elected governor of California’’ (and the assertion turns outto be true), but, in the case as I described it, he would not himself rememberthat he was re-elected, i.e. not himself accepting any thought of the form <<Iwas re-elected governor of California>>, and as a skeptical amnesiac he doesnot accept what he was told by the informant—that he is Gray Davis. So,again, since he disavows a belief whose content is <>, he would have to say ‘‘but I do not believe that I myselfwas re-elected governor of California’’. Such is the significance of the reflexivepronoun ‘myself ’ in the second conjunct of (23b).²²

I do not think that in the case that I just elaborated the utterance of (23b)sounds odd. I think the utterance is felicitous in the context that I described. ButI do not deny that it is an unusual context. For the example to work, one cannotassume that ‘‘told by an informant’’ entails ‘‘accepted by the hearer.’’ This is not

²⁰ Castaneda (1966: 144) uses similar language, viz. ‘a directly self-referring use of ‘I’ ’.²¹ It is also possible for a parasitic use to diverge from the direct referring use of ‘I’. Castaneda

(1966: 146) wants to hold the simple principle that ‘‘nobody can refer to another person by meansof ‘I’ ’’. But then he gives an example of Norman Kretzmann’s to the contrary (Castaneda (1966:147, n. 12)):

I am indebted to Norman Kretzmann for the following teasing use of ‘I’ which looks like acounter-example to this claim. Suppose that there is a play about Privatus and that Privatus is inthe audience. Suppose further that the actor representing Privatus is losing his moustache and thatPrivatus referring to the actor says ‘‘I am losing my moustache.’’

Castaneda calls this a teasing use, implying that the counter-example is spurious, but I think that ateasing, and so parasitic, use is a use rather than a non-use and so not spurious.

²² I am indebted to John Williams and Elisabeth Camp for comments that provoked me toclarify this example. The argument proceeds in the same way for (23′ b).

Jay David Atlas 133

an entailment, even if it is often a reliable inference, anymore than ‘the mancalled ‘the greatest showman on earth’ ’ necessarily denotes the greatest showmanon earth. (An observation made by Kripke in Naming and Necessity.)²³

By contrast, the de re utterance-type (23c) ‘‘I was re-elected governor ofCalifornia, but I do not believe myself to have been re-elected governor ofCalifornia’’ is an interpretation of an utterance of sentence (23a) I was re-electedgovernor of California, but I do not believe it that has the form of a Moore’sParadox sentence, but when asserted by Gray Davis, recovering consciousnessafter plastic surgery but otherwise unaddled, it can be felicitously and trulyasserted when Gray Davis does not yet know that he is the same person as theperson in the post-operative photograph, of whom he is asked, ‘‘Was this manre-elected governor of California?’’. Davis thinks that the anaesthesiologist is justchecking his recovery from the anaesthetic by testing his knowledge of history.So the de re interpretation (23c) of a sentence that does have the form of a MooreParadox sentence is not a linguistically paradoxical utterance-interpretation.²⁴

Compared to the previous standard form of a Moore’s Paradox sentence (20),where the problem with the case of the first-person content-clause is that theodd de se interpretation no longer has the form of a Moore’s Paradox sentence,and the de re interpretation of a sentence that has the form of a Moore’sParadox is not odd, in this second standard form of a Moore’s Paradox sentencein the case of a first-person content-clause, again the de re interpretation ofa sentence that has the form of a Moore’s paradox is not odd, but neitheris the de se interpretation, which again does not have the form of a Moore’sParadox sentence. Thus explanations that hinge essentially on the standard formsof Moore’s Paradox sentences cannot account for these content-clauses in thefirst-person. The interpretations of forms of Moore’s Paradox sentences are notodd, and the interpretation that is odd of the one version (20) of a standardMoore’s Paradox sentence does not have the Moore’s Paradox form.

The discussion of this section has implications for the views elegantly sketchedby Richard Moran (2001). In defending the cogency of the distinction between‘‘subjective’’ and ‘‘objective’’ perspectives, or as Moran (2001: 68) more eleg-antly—and Sellarsianly—puts it, between ‘‘deliberative’’ and ‘‘theoretical’’perspectives, Moran writes that an ‘‘examination of the first-person shouldaccount for why someone’s need to rely on behavioral evidence to report on hismental states would suggest something wrong with him, some state of disso-ciation, and would raise doubts about the rationality of those attitudes of hiswhich are not accessible to him in the normal ‘immediate’ way’’. The linguisticexamples of this section, of a psychiatric patient recovering from delusional statesand of a surgical patient recovering from plastic surgery who does not yet know

²³ School-boy joke. A: If you call the tail of a dog a leg, how many legs does the dog have? B:Five. A: No, four; calling a tail a leg does not make it one.

²⁴ I would make the same claim for (23′ a,c).


what he looks like, are felicitous, Moore’s Paradox utterances. It is precisely inthese cases that the speaker relies on more than introspectible information inassessing his own mental states, and he or she is right to do so. In such speakersit is a sign of psychological harmony between their ‘‘inner’’ and ‘‘outer’’ ratherthan, as Moran suggests, evidence of dissociation.

The use of Moore sentences indicates, Moran (2001: 68–9) suggests in thecase of his example ‘‘I know this must be envy that I feel toward him (I believeit’s raining out), although there is nothing about him to be envied (but it’s notraining out),’’ that ‘‘whatever thought may be seeking expression here couldonly [my emphasis] be entertained under great tension’’. Like Sartre, and onhis reading (though not on mine) Wittgenstein too, Moran wishes to criticize‘‘theoretical certainty as our model for the achievement of self-knowledge’’. Yetmy linguistic examples above, instances of the form of Moore’s paradox sentences,are not cases of entertaining a thought under great tension; they are an exampleof a linguistic report of the overcoming of a psychotic state and an example ofunderstanding the contrast between facts and one’s limited knowledge of them.There is every reason to believe that a psychiatric patient’s ‘‘objective’’ empiricalknowledge of the self, as in my first example, can contribute to a lessening ofpsychotic delusion and to a reducing of tension between the ‘‘inner’’ and the‘‘outer,’’ a reduction that the utterance of the Moore sentence is designed toexpress. To turn Moran’s (2001: 68) point upside down, I would rephrase hisremark that in ‘‘its most radical form, the sort of alienation [my emphasis] we’reconsidering would place the person in a situation he could only describe in anutterance that was a version of Moore’s paradox’’ by saying that in its mostradical form the sort of psychological integration we’re considering would placethe person in a situation he could only describe in an utterance that was a versionof Moore’s paradox. The saying, or the entertaining of, such a thought is notparadoxical; it is not even an infelicitous utterance or verbal thought.

Moran (2001: 84–5) is aware of such cases. He (2001: 84) even says that‘‘were it to make sense for one to take such a purely theoretical view of oneself,then the thought expressed in a Moore-type sentence would describe a perfectlycoherent empirical possibility on which one could sensibly report’’. This seemsconsistent with my examples, except that it is hard to understand the examplesas ones in which a ‘‘purely theoretical view of oneself ’’ is being taken. Surelythe utterances of my medical patients do not constitute thoughts in which theytake a purely theoretical, scientific view of themselves. Their ordinary utterancesseem a mix of the theoretically pure and the subjectively impure. Certainlythere is no reason to think that their talk of their beliefs is solely dependenton ‘‘objective’’ empirical evidence. In example (19) I believe myself to have beenre-elected governor of California, but I was not²⁵ one would have thought it quitethe contrary. Looking at these uses of ‘believe’, one suspects that there is no more

²⁵ I believe myself to be governor-elect of California, but I am not.

Jay David Atlas 135

to the inner/outer distinction in the sense of ‘believe’ than Iris Murdoch (1985)or Hilary Putnam (2002) thought there was a fact/value distinction in the senseof ‘jealous’ or ‘bumptious’ (Murdoch 1985: 13, 23) or in the sense of ‘cruel’(Putnam 2002: 34–5).

The dichotomies that interest Moran, ‘‘inner’’/‘‘outer’’, ‘‘subjective’’/‘‘objective’’, and ‘‘rational’’/‘‘theoretical’’, are represented by him (2001:84)—and this is supposed to be a good thing—as implying that the first‘‘transcends any description of my psychological state’’ reported by the second.Thus it should be possible to contrast a ‘‘purely’’ theoretical/empirical belief-attribution, dependent on objective, psychological evidence, with a ‘‘rationalagent’’, first-person conception of oneself as having beliefs about the world,beliefs to which one has access without ‘‘objective’’ evidence. But there is justno linguistic reason to think that people talk or think dichotomously in thisway, or that in thinking of themselves as having beliefs about the world, theythink of the world, and their ordinary beliefs about it, as transcending everydescription of their psychological states. The descriptions ‘jealous’ and ‘cruel’ areworld-involving as much as mind-involving. First-person, present-tense, indicat-ive mood uses of ‘believe’ are impure; they are mixtures of reporting and avowing,to use Moran’s (2001: 86) Wittgensteinian distinction. Reporting a belief isgrounded in evidence; Moran claims that reporting a belief, even reporting one’sown belief, does not entail a commitment to the truth of the lower constitu-ent ‘‘. . .’’ in ‘‘I believe . . .’’. By contrast, avowing is the expression of one’scommitment to the truth of the lower constituent.

When one examines the utterances of the medical cases in (19), one realizesthat, in Moran’s terms, the patient is both avowing and disavowing, hencepresumably reporting his or her belief-state without commitment to the truthof the lower constituent. But the theory of Indirect Speech Acts permits justsuch complexity in utterances (Searle 1975; Davison 1975; Lycan 1984). Theindicator of illocutionary force in the first conjunct is what, following J. F. M.Hunter (1990), I have called a ‘‘quasi-performative’’ use of ‘I believe’. Its useby a speaker in the first conjunct of (19) purports to establish the existenceof a belief-state committed to the truth of the content. But a denial of thatcontent by the second conjunct of (19) creates for an utterance of the wholeconjunction an indirect speech act of reporting a belief-state with a false content,with the peculiarity that the belief-state is the speaker’s own. (The possibilityof this peculiarity is noted, happily, by Moran 2001: 89 n. 11.) Thus in ourlinguistic understanding of the utterance, we focus primarily on the indirectillocutionary force—reporting—of the conjunction, while also understandingthat secondarily the speaker is also committed by the grammatical form ofthe utterance to the truth of his or her belief-content. However one hopes toexplain psychologically the mental or brain states of such individuals, or evento describe commonsensically what is going on ‘‘inside’’ the patient, there isnothing linguistically anomalous in the language used. The patient is revealing

as much as his or her competence in the language permits. Nothing in whatis said indicates how much of what is uttered is about the inner or the outer.So, contrary to Moran’s (2001: 83–4) hopes, Moore Paradox sentences are nottouchstones ‘‘for difficulties arising from the introduction of an empirical ortheoretical point of view whose deliverances may clash with what we want to sayfrom the point of view of a rational deliberator, a point of view ‘transcendent’with respect to the psychological facts as currently constituted’’. The fact thata Moore sentence conjunction can have two, independently true conjuncts ‘Ibelieve it is raining’ and ‘It is not raining’ does not suggest by being used tomake both a true and an infelicitous assertion that there is a deep clash betweena third-person, objective perspective and a first-person subjective perspective.Moran is guilty of a metaphysical subliming of language that ordinary speakersof English avoid. Many Moore Paradox sentences when uttered are felicitous,indirect speech acts of reporting while simultaneously remaining grammaticallydirect but secondary speech acts of avowing. In indirect speech acts like (19),‘I believe’ expresses a ‘‘thick’’ concept, one that is not purely theoretical andscientific, or independent of ‘‘expressing’’ a commitment to the truth of thecontent. That is why Moran (2001: 85) is mistaken to say in general that whena ‘‘belief is described, it is kept within the brackets of the psychological operator,‘believe’,’’ without a commitment to the truth of the content.

Moore sentences with first-person content-clauses are not paradoxical. Neverthe-less the linguistic evidence that I have considered above does suggest a new formof a Moore’s Paradox in sentences of belief de se. I summarize the conclusion inthe next section.

V. MOORE’S PARADOX DE SE

From the evidence in Section IV, I conclude that (a) if QI is a first-person subjectNoun Phrase sentence, there is no linguistic oddity to the form <QI ∧ ¬ BI

[QI]> of a Moore Paradox sentence and that (b) there is a linguistic oddity to

the form <BI [QI∗] ∧ ¬ Q>I , but only if the first-person subject Noun Phrase in

QI∗ is an indirect reflexive ‘I myself ’, i.e. is a Castaneda (1966) ‘‘quasi-indicator’’‘I*’. The form must be <BI[QI∗] ∧ ¬ Q>

I , as in (21b), but this sentence is notof a Moore Paradox form <BI[QI] ∧ ¬ Q>

I .The standard forms of Moore’s paradox belief utterances do not, in fact,

seem to me logically or linguistically paradoxical in the first-person content-clause cases, since they would be uttered felicitously by recovering, previouslydelusional psychiatric patients. The truly problematic forms of Moore’s Paradoxare quite linguistically different. The New First-Person Problem of Moore’sParadox De Se is this: we need an explanation of the oddity of the de se utterance(21b), and, more generally, of the oddity of the de se form (24).

(24) ?I believe that I myself F, but I do not F.

Jay David Atlas 137

I shall not, in this essay, attempt to provide an analysis of the New Problem ofMoore’s Paradox De Se. Instead I would like to show what consequences for ‘‘priv-ileged access’’ to beliefs follow if one takes indirect reflexive pronouns seriously.

VI . SHOEMAKER’S FIRST EXPLANATION OF MOORE’S

PARADOX

David Rosenthal (1986, 1993) and Sydney Shoemaker (1996) have offeredaccounts of conscious intentional states and of first-person privileged access tomental states that were, in part, motivated by consequences they drew from thelinguistic absurdity of asserting standard forms of Moore’s Paradox sentences.David Rosenthal (1986) has used Moore’s Paradox to show that ‘‘expressing’’a mental state is distinct from ‘‘reporting’’ that one is in it, that the oddity ofasserting a Moore Paradox sentence consists in one conjunct of the sentencereporting the absence of a mental state that the other conjunct purports to express[It’s raining but I do not believe it] and that there is a causal connection betweenspeech acts and the mental states that they express. In a later essay Rosenthal(1993) goes much further, using the distinction between expressing a mental stateand reporting it to argue for his famous conclusion, that a mental state’s beingconscious consists simply in its being accompanied by a higher-order thought.Sydney Shoemaker (1996) uses Moore’s Paradox to a different end, offering adifferent view of the significance of Moore’s Paradox, one that leads him to anargument for first-person access to our beliefs: a defense of the principle that‘‘If I believe that P, then I believe that I believe that P.’’ Here my task is toexamine Shoemaker’s arguments for ‘‘privileged access’’, as they are suggested byhis analysis and explanation of Moore’s Paradox first-person sentences.

In his influential essay ‘‘Moore’s Paradox and Self-Knowledge’’ Sydney Shoe-maker (1996) has two explanations of the oddity of Moore’s Paradox. He firstproposes that instead of considering the relation between assertion and beliefin explaining the paradoxicality of Moore sentences, he will focus instead onthe peculiarity of believing what one has said or thought when one has said orthought . On this view, while the thought of the form can be true, the conjunction (C) <(P ∧ ¬ BI[P]) ∧ BI (P ∧ ¬ BI[P])>

is alleged by Shoemaker to be inconsistent. On the assumption that believing aconjunction entails believing each conjunct, an explicit contradiction <¬BI[P]∧ BI[P]> is derivable. But what should one conclude from this? Shoemaker(1996: 76) concludes that if the sentence of the Moore Paradox form ‘‘can be believed at all, the subject of such a belief could not, logically,believe that she had it’’. But that is not what follows from the negation of (C).Rather, what follows, among other consequences, is , i.e. if P then the person who does not (first-personally) believe Pcannot (first-personally) believe both P and that he/she does not (first-personally)

believe it. And, of course, if one does not believe something, even if it were true,why should one have the belief that one does not believe it? There are manypropositions that I fail to believe, among them many propositions that I havenever considered and that are true; I do not expect myself to believe that I do notbelieve a particular, true, never-thought-of proposition. Shoemaker’s argumentmerely demonstrates a truism about belief.

But Shoemaker thinks that he has demonstrated something more than a meretruism. He thinks that he has shown that it is logically impossible for one tobelieve a Moore sentence of the form . Of course, if Shoemakerhad shown <¬♦BI (P ∧ ¬ BI [P])>, he would have shown that it is necessarythat is compatible (consistent) with what I believe, i.e. he wouldhave shown <

�CI (P → BI [P])>. So, for example, he would have shown thatit is a necessary truth that it is compatible with everything I believe that if thesmallest integer that can be expressed by the sum of two cubes in two ways is1729, then I believe that the smallest integer that can be expressed by the sumof two cubes in two ways is 1729, and the same goes for any P whatsoever.But I am not omniscient, and I believe that I am not. That is, I believe thatthere is some true proposition that I do not believe; so an axiom of my beliefsystem is <BI (∃p(Tr(p) ∧ ¬ BI(p) )>. Hence, it is compatible with everything Ibelieve that for some proposition po, <if po, then I believe p>

o is false. But then,what Shoemaker has claimed to have shown is inconsistent with the denial ofmy (and your) omniscience. And I really don’t think one wants a philosophicalexploration of the consequences of Moore Paradox belief sentences that beginswith the claim that Sydney, or any of us believers, is omniscient.²⁶

VII . SHOEMAKER’S SECOND EXPLANATION

OF MOORE’S PARADOX

Shoemaker then offers a second explanation of Moore’s Paradox by supporting aclaim of ‘‘self-knowledge’’, more precisely a claim of first-person access to one’sbeliefs. I shall summarize his account. Suppose I have a belief with content<>, i.e. I believe P. In the stereotypical cases of belief, e.g. not self-deception, repressed belief, or forgetfulness (Shoemaker 1996: 80), because itis believed, the content <> is ‘‘available’’—I would say ‘accessible’—as

²⁶ So beliefs for the non-megalomaniacs among us are ‘‘incomplete’’, in a sense analogous tothe incompleteness of a formally axiomatized system rich enough to contain formalized arithmetic.In earlier work I explored the parallel between the report of the existence of a true but formallyunprovable Godel sentence and a Moore paradox sentence like (20); see Jay David Atlas, ‘‘Whatis Paradoxical about G. E. Moore’s Paradox?—Rationality, Sincerity, Implicature, and the Self-limitations of Belief ’’, November 1995, MS, Pomona College, Claremont, California. Needless tosay, I do not think that a Godel sentence is assertorically anomalous, nor do I think that the sentencethat reports its existence is asserted anomalously. The relationship between a Moore sentence andthe report of the existence of a Godel sentence is of great interest, logically and linguistically.

Jay David Atlas 139

a premise in my theoretical and practical reasoning. When so, I consciouslythink it and, furthermore, ‘‘mentally assent’’ to it. (Mentally assenting is likeverbally asserting it or answering a query about its truth by replying ‘yes’—onlyquieter.) But it is analytic (for Shoemaker) that mentally assenting to <>

is a conscious process; Shoemaker (1996: 79) actually writes ‘necessary’ insteadof ‘analytic’, but I take the liberty. When doing so, one ‘‘manifests’’—welanguage mavens would have said ‘expresses’, though that is not without itsproblems—that one believes P.

This is a causal model: one asserts P because one believes P (and isbeing entertained), just as one has spots because one has measles. Having spots‘‘expresses’’/‘‘manifests’’ measles (Sellars 1969/1996).)

The next step in Shoemaker’s argument is one for which I can find noargument in his essay. He wants to go from ‘‘manifesting’’ that one believes Pto ‘‘manifestly assenting’’ to one’s believing P (Shoemaker 1996: 82–3). Forif this inferential move were acceptable, then one would be assenting to one’sbelieving P. The argument would now show that if one assents to , oneassents to one’s believing P. Since assenting is conscious (an analytic claim forShoemaker), one is aware of assenting to believing P. Since assent is a kind ofbelieving—a stipulation (Shoemaker 1996: 78), one is consciously believing thatone believes P.

By this argument, Shoemaker gets two large philosophical payoffs. First, hegets the claim that if one assents to , one assents to . Second,he gets the claim that we have first-person access to our beliefs: (a) if one believesP, one will be able to access the content <>; (b) if one is able to access thebelief-content <>, one will both believe (at least tacitly) that one is ableto access the belief-content <> and consciously believe that one is able toaccess the belief-content <>.

The question is how, in the crucial step, Shoemaker goes from manifestingthat one believes P to ‘‘manifesting assent’’ that one believes P. There is a hintin the text (Shoemaker 1996: 82). According to Shoemaker, saying (sincerely) can manifest that one believes P. If one sincerely asserts , and one sincerely asserts it because (and only because?) one mentally assentsto it, then one believes that one believes P, because it is analytic for Shoemakerthat one mentally assents to believing P only if one believes that one believesP. And the deed is done; if one manifests believing P, one manifestly assents tobelieving P. So if one believes P, one believes that one believes P—privilegedaccess (Shoemaker’s ‘‘self-intimation’’).

The crux of the matter is Shoemaker’s claim that asserting (sincerely) manifests that one believes P. If ‘manifest’ is paraphrasable by (one sense of )‘express’, one has the claim <asserting (sincerely) expresses thatone believes P>. But this claim is only plausible if ‘express’ is not taken in itscausal, manifesting sense, e.g. the sense in which one’s asserting P expresses one’sbelieving P. It is plausible only in the semantic sense of ‘express’, that in which

my assertion expresses the thought that I believe P. (See Sellars1969/1996.)²⁷

In short, the apparent success of Shoemaker’s argument relies essentially onan equivocation between two senses of ‘express’, the causal (manifesting) senseand the semantic sense.

Had Shoemaker’s argument succeeded, he would have had the followingexplanation of the oddity of the Moore sentence schema . OnShoemaker’s view we may derive from <BI (P ∧ ¬ BI[P])>, on the assumptionthat believing a conjunction entails believing each conjunct, <BI[P] ∧ BI [¬BI[P]]>. But if it is a theorem of the logic of belief that if I believe P, then Ibelieve that I believe P, one then infers <BI [BI[P]] ∧ BI [¬ BI[P]]>. By theconverse conjunction principle to the previous one, one may infer <BI [BI[P]∧ ¬ BI[P]]>. So I, the believer of the Moore sentence ‘It’s raining but I don’tbelieve it’, if my believing P entails my believing that I believe P, and if beliefdistributes over conjunction, believe an overt contradiction.

What is essential to this explanation of the oddity of the original form of aMoore’s Paradox sentence by appeal to the existence of a contradictory belief-content is Shoemaker’s elaborate argument in support of the view that if I believeP, then I believe that I believe P. But that is just the argument that fails; itseems to succeed only because of an equivocation, a slide between two sensesthat motivates an illicit move from ‘‘manifesting’’ that I believe P to ‘‘manifestlyassenting’’ to .²⁸

Shoemaker has assumed that by deducing a contradictory belief-content frombelieving the Moore sentence, he has shown why the uttering of the sentenceassertively is ‘‘odd’’. This is a traditional line to take. To my knowledge itwas first employed by the American logician C. H. Langford in 1942, in hiscontribution to the Schilpp volume on Moore ‘‘The Notion of Analysis inMoore’s Philosophy’’ (Langford 1942/1968: 333), where he writes:

I want to cite an example which is due to A.M. MacIver and which is worth repeatingon its own account.∗ Suppose someone to remark: ‘‘He thinks that he has been toGrantchester but he has not.’’ The person referred to may entertain this proposition asan hypothesis. But suppose he actually asserts the proposition: ‘‘I think that I have beento Grantchester but I have not.’’ This sounds self-contradictory, and the reason is thathe will actually be saying that he thinks that he has been to Grantchester, whereas thebut-clause in the indicative mood will signify or mean pragmatically# that he does notthink so.∗∗

*See Analysis, Vol. 5 (1937–8), 43–50, and The Journal of Symbolic Logic, Vol. 3(1938), 158.

²⁷ ‘Assertion’ has the traditional act/object ambiguity in English: assertion as an act of asserting,assertion as what is asserted.

²⁸ Hintikka’s (1962) seminal discussion of the logic of belief takes the Moore sentence tobe ‘‘doxastically indefensible,’’ i.e. that one’s believing one’s Moore sentence is inconsistent withHintikka’s modal S4 axioms for the belief-operator. Of course, the characteristic axiom for S4 is, inthe case of the belief operator, just <Bp → BBp>.

Jay David Atlas 141

#(My note: Langford (1942/1968: 332) wrote: We may call what a man does notstate, but intends his linguistic behavior to signify, his pragmatical meaning, and we maydistinguish this from the sense of his words, which is the proposition expressed by them.$

[Langford’s note:$ This term has been used by Charles Morris in the same or a similarsense.])

**The course of Moore’s argument in ‘‘A Defence of Common Sense’’ will be clearerif this distinction between formal and pragmatical contradiction is carefully observed.For in saying that certain philosophers contradict themselves when they assert, in effect,‘‘There have been many other human beings beside myself, and none of them, includingmyself, has ever known of the existence of other human beings,’’ Moore is not holdingthat a formal contradiction can be derived from the sense of those words, but only thatthe pragmatical meaning of such an assertion is incompatible with its literal meaning.

The peculiarity of Langford’s analysis is that he is diagnosing the ‘‘apparentself-contradiction’’ of a statement of the form <BI[P]&¬P> by the hypothesisthat in saying <¬P> the speaker ‘‘pragmatically means’’ <¬B[P]>. There is norecognition from Langford of the need to discuss an apparent scope distinctionbetween the wide-scope negative belief-proposition that Langford takes to bethe speaker’s pragmatical meaning and the narrow-scope negative propositionB[¬P], which might be an equally good candidate for the pragmatical meaningof uttering <¬P>. Without defense—and I find the claim dubious—Langfordchooses the wide-scope negative formula as the speaker’s meaning in order toexplain the apparent self-contradiction by an actual logical contradiction betweenthe sense of ‘I believe I have been to Grantchester’ and the speaker’s allegedpragmatical meaning ‘‘I don’t believe I have’’ when asserting ‘I have not’.²⁹ Inaddition he thinks it is self-evident that such a logical inconsistency betweenthese different sorts of ‘‘meaning’’ suffices to explain the apparent ‘‘oddness’’of asserting the Moore sentence. But no theory providing the explanatoryconnection is offered at all.

In a similar vein, George Lakoff (1975: 264–5) discusses Moore’s paradoxin the form <P&B[¬P]>. Lakoff ’s explanation of the paradoxicality of theutterance is a version of a solution that appeals essentially to a speech act sinceritycondition, to an incorrigibility thesis, and to rationality assumptions. In additionto the sincerity condition ‘‘if x sincerely asserts P, x believes P’’, Lakoff makesthree assumptions: (i) that if one believes a conjunctive statement, one believes

²⁹ There is an appealing, but a mistaken, argument to support Langford’s reasoning. Supposea la Moore in asserting <¬P> the speaker pragmatically means <BI[¬P]>. The consistency ofbelief requires the axiom <B[¬P] → ¬B[¬¬P]>, while Belief Double Negation yields <B[P] →B[¬¬P]>, from which there follows <B[¬P] → ¬B[P]>. If pragmatic meaning is preserved underlogical consequence (modulo the axioms of rational belief ), then if the speaker pragmatically means<B[¬P]>, he also means <¬B[P]>. Such consequences might be thought to be plausible if aspeaker intends his utterings to signify the logical consequences—or perhaps more restrictedlythe direct logical consequences (Atlas 1991: 137), modulo the axioms of rational belief—of whathe intends his utterings to signify. The problem is, the preservation of pragmatic meaning underlogical consequence is really quite implausible. So the easy argument for Langford’s position doesnot succeed.

the individual conjuncts (belief conjunction-elimination <B[P&Q] → (B[P]&B[Q]) >), an assumption also made by Shoemaker in his first argument, (ii)the incorrigibility of beliefs about mental states: that if one believes that onebelieves P, then one believes P (i.e. <BB[P] → B[P]>), and the consistencyassumption (iii) <B[P] → ¬B[¬P]>. He then can deduce the contradiction<B[¬P]&¬B[¬P]> from the assumption of a sincere, rational assertion of hisMoore sentence. By explaining rather than Moore’s form , Lakoff avoids assuming Shoemaker’s self-intimation of belief (i.e. <B[P]→ BB[P]>, the converse of the incorrigibility thesis). Lakoff (1975: 265) goeson to claim that the same principles will allow a deduction of contradiction fromthe assumption of a sincere assertion of the original Moore sentence form in order to explain the latter’s assertoric oddness, but so far as I can seethis claim is incorrect. In order to deduce a contradiction, he needs the S4-like,self-intimation principle that, anti-Cartesianly, Lakoff thinks is questionable,though he does not explain why he thinks so (Lakoff 1975: 264).

Langford and Lakoff explain the peculiarity of the Moore assertion bydeductions of a logical contradiction of two sorts. In Langford’s case, preciselybecause the original assertion sounds self-contradictory to him, he claims that aninconsistency can be derived from the conjunction of the speaker’s unassertedpragmatical meaning and the asserted literal sense. In Lakoff ’s case, becausethe assertion of the Moore sentence sounds odd to him, he claims that asincere, rational assertion of the Moore sentence is logically inconsistent with theprinciples of rational belief and of speech-act theory.

Yet it seems to me an open question whether the assertion of the Mooresentence does sound self-contradictory, and a closed question that even if assert-ing it seemed odd, the alleged reductio of the rational, sincere assertion of theMoore sentence could not explain the assertoric oddness. A difficulty with Lang-ford’s explanation is that there are a denumerable number of self-contradictorystatements that are perfectly felicitous, so that the logical inconsistency of thetotal signification of a statement cannot be sufficient for its assertoric oddness.³⁰Half of the needed contradiction is not even asserted. In the case of Lakoff ’sexplanation, since neither a speaker’s insincerity nor the failure of the speaker’sbeliefs to conform to Lakoff ’s two axioms of rational belief and the incorrigibilitythesis could be explanatory of the Moore statement being linguistically odd, itis bizarre to think that the impossibility of the rational, sincere assertion of aMoore sentence could explain the oddity of uttering a Moore sentence assertively

³⁰ For example, ‘0 = 1’, or even better ‘147/356 = 0.41292’, though contradictory, is notuttered infelicitously, especially not in a reductio ad absurdum proof, nor is the sentence ‘ 1

3

√π =

−∞∫ ∞

e−x2dx’, an equation whose right-hand side is a definite Riemann integral, which is alsoa contradictory sentence but felicitously utterable, nor is the following, necessary falsehood, anon-theorem of the calculus, believed by European mathematicians for two hundred years to be atheorem and asserted as such, ‘Every real-valued function f: R → R that is continuous on the closedinterval [a, b] is differentiable on the interval.’

Jay David Atlas 143

(as Moore (1993: 207) would put it), even if the impossibility of a sincere,‘‘rational’’ assertion explains why either the utterance is insincere, or expresses a‘‘non-rational’’ belief or a corrigible belief, or is not an assertion—oddities all,but none of insincerity, non-rationality, corrigibility, and an assertoric speech-act‘‘misfire’’ is the oddity of uttering the Moore sentence assertively. None ofthose consequences would seem to explain the particular oddity of uttering theMoore sentence assertively. (As Paul Benacerraf once remarked, in reaction tothis claim, an adequate account of the oddity must appeal to principles that everyspeaker/hearer can be expected to ‘‘know’’, in the sense in which we know ourlanguage and its uses. Not every speaker, surely, is natively a Port Royal logi-cian.) The oddity remains even if the Moore utterance merely purports to be anassertion. Furthermore, ‘‘nonrationality’’ and insincerity of belief are propertiesof the asserter not of the utterance, so it is a category error to think that they canexplain the linguistic oddity of the sentence uttered assertively.

Arguments that appeal to self-contradiction or inconsistency with supposedrational principles of belief just do not explain the linguistic phenomena. Whatwould explain it is a tale for another occasion. But if Lakoff finds privileged accessdubious, Shoemaker and Hintikka find it self-evidently or necessarily true. I havealready shown that Shoemaker’s arguments fail to support the privileged accessthesis. Can the approach that I have been taking in this paper offer a definitiveanswer to the question whether the privileged access thesis is correct?

VIII . A LOGICO-LINGUISTIC ARGUMENT AGAINST

PRIVILEGED ACCESS (“SELF-INTIMATION”)

Why, apart from Cartesian prejudices about the transparency of the mind, shouldone have expected the privileged access thesis <BI[P] → BI[BI[P]]> to be correct?

(25) If I believe it’s raining, then I believe that I believe it’s raining.

The sentence in (25) has only two interpretations, the de se in the consequentof the conditional, in (26a), the de re in the consequent of the conditional, in(26b), and the de dicto in the antecedents.

(26) a. I believe it’s raining ⇒ I believe that I myself believe it’s raining.b. I believe it’s raining ⇒ I believe myself to believe it’s raining.

Neither interpretation makes the conditional true, either as an analytic entailmentor a theorem of a logic of belief. The de re interpretation goes from thehypothesized truth of a non-first-person-content de dicto antecedent to anobjectively first-person content, de re consequent. The de se interpretation goesfrom a non-first-person content de dicto antecedent to an essentially-first-person-content de se consequent. Neither implication is semantically justified.³¹ The

³¹ As I mentioned in n. 12, those who cannot get the de re interpretation of the consequent in(26b) will at least get the de se interpretation in (26a), which suffices for my argument.


English grammar of ‘believes’ sentences carries no commitment to first-personbeliefs being ‘‘self-intimating’’ in Shoemaker’s sense and so to no distinctivefirst-person access to our beliefs. In fact, the falsity of (26) shows that theordinary concept of belief, the one we express in our everyday talk, the one meantby the lexical item ‘believes’, explicitly eschews such a commitment to privilegedaccess!This is what indirect reflexive pronouns tell us about belief: the ordinary,linguistically expressed notion of belief cannot be Cartesian.

At his commentary on an early version of Shoemaker’s essay, at the PacificDivision of the American Philosophical Association in 1994, Rogers Albritton,reported by Shoemaker (1996: 81), in reaction to the idea that if I believe Pthen I believe that I believe P, besides pointing out the obvious infinite regressof first-person higher-order beliefs, remarked, of someone who says ‘I believeit’s raining’, ‘‘Where is it written that it takes two beliefs, or even one with twocontents [‘‘P’’ and ‘‘I believe P’’], to confess that one believes it’s raining. Whycan’t I just bare my soul in the matter?’’ Such an Albrittonian comment!Thelinguistic argument of this essay shows that Albritton’s intuition was correct.

REFERENCES

Atlas, J. D. (1989), Philosophy without Ambiguity (Oxford: Clarendon Press).(1991), ‘‘Topic/Comment, Presupposition, Logical Form, and Focus Stress

Implicatures: The Case of Focal Particles ‘Only’ and ‘Also’’ ’, Journal of Semantics, 8:127–47.

(2004), ‘‘Presupposition’’, in G. Ward and L. Horn (eds.), Encyclopedia of Pragmatics(Oxford: Blackwell), pp. 29–52.

(2005), Logic, Meaning, and Conversation: Semantical Underdeterminacy,Implicature, and Their Interface (New York: Oxford University Press).

and Levinson, S. (1981), ‘It-clefts, Informativeness, and Logical Form: RadicalPragmatics (Revised Standard Version)’, in P. Cole (ed.), Radical Pragmatics (NewYork: Academic Press), pp. 1–61.

Bach, K. (1987), Thought and Reference (Oxford: Clarendon Press).Boer, S., and Lycan, W. (1986), Knowing Who (Cambridge, Mass.: MIT Press).Braun, David (1998), ‘‘Understanding Belief Reports’’, Philosophical Review, 107:

555–95.(2002), ‘‘Cognitive Significance, Attitude Ascriptions, Ways of Believing’’, Philo-

sophical Studies, 108: 65–81.Castaneda, H.-N. (1966), ‘ ‘‘He’’: A Study in the Logic of Self-Consciousness’, Ratio, 8:

130–57.(1989), Thinking, Language, and Experience (Minneapolis: University of Minnesota

Press).Davison, A. (1975), ‘‘Indirect Speech Acts and What to Do With Them’’, in P. Cole and

J. Morgan (eds.), Syntax and Semantics 3: Speech Acts (New York: Academic Press),pp. 143–85.

Jay David Atlas 145

Geach, Peter (1957), ‘‘On Beliefs about Oneself ’’, Analysis, 18: 23–4; repr. in P. Geach(1972), Logic Matters (Berkeley and Los Angeles: University of California Press),pp. 128–9.

Hintikka, J. (1962), Knowledge and Belief (Ithaca, NY: Cornell University Press).Hunter, J. F. M. (1990), Wittgenstein on Words as Instruments: Lessons in Philosophical

Psychology (Edinburgh: Edinburgh University Press).Kripke, S. (1980), Naming and Necessity (Cambridge, Mass.: Harvard University Press).Lakoff, G. (1975), ‘‘Pragmatics and Natural Logic’’, in E. L. Keenan (ed.), Formal

Semantics of Natural Language (Cambridge: Cambridge University Press), pp. 253–86.Langford, C. H. (1942/1968), ‘‘The Notion of Analysis in Moore’s Philosophy’’, in P.

A. Schlipp (ed.), The Philosophy of G.E. Moore, vol. i (LaSalle, Ill.: Open Court),pp. 321–42.

Levinson, S. C. (1995), ‘‘Three Levels of Meaning’’, in F. R. Palmer (ed.), Grammar andMeaning (Cambridge: Cambridge University Press), pp. 90–115.

Lewis, D. K. (1979a), ‘‘Attitudes De Dicto and De Se’’, Philosophical Review, 88/513–43;repr. in Lewis, Philosophical Papers, vol. i (New York: Oxford University Press),pp. 133–56.

(1979b), ‘‘Scorekeeping in a Language Game’’, Journal of Philosophical Logic, 8:339–59; repr. in Lewis, Philosophical Papers, vol. i (New York: Oxford UniversityPress), pp. 233–49.

Lycan, W. (1984), Logical Form of Natural Language (Cambridge, Mass.: MIT Press).Moore, G. E. (1993), Selected Writings (London: Routledge).Moran, R. (2001), Authority and Estrangement: An Essay on Self-Knowledge (Princeton:

Princeton University Press).Murdoch, I. (1985), The Sovereignty of Good (London: ARK).Perry, J. (1977), ‘‘Frege on Demonstratives’’, Philosophical Review, 86: 474–97.Prior, A. N. (1967), ‘‘On Spurious Egocentricity’’, Philosophy, 42: 326–35.Putnam, H. (2002), The Collapse of the Fact/Value Dichotomy and Other Essays (Cambridge,

Mass.: Harvard University Press).Quine, W. V. O. (1960), Word and Object (Cambridge, Mass.: MIT Press).Rosenthal, D. (1986), ‘‘Intentionality’’, in P. French, T. Uehling, and H. Wettstein

(eds.), Midwest Studies in Philosophy, 10 (Minneapolis: University of Minnesota Press),pp. 151–84.

(1993), ‘‘Thinking that One Thinks’’, in M. Davies and G. Humphreys (eds.),Consciousness (Oxford: Blackwell), pp. 197–223.

Searle, J. (1975), ‘‘Indirect Speech Acts’’, in P. Cole and J. Morgan (eds.), Syntax andSemantics 3: Speech Acts (New York: Academic Press), pp. 59–82.

(1983), Intentionality (Cambridge: Cambridge University Press).Sellars, W. (1969), ‘‘Language as Thought and as Communication’’, Philosophy and

Phenomenological Research, 29; repr. in H. Geirsson and M. Losonsky (eds.), Readingsin Language and Mind (Oxford: Blackwell, 1996), pp. 214–32.

Shoemaker, S. (1996), ‘‘Moore’s Paradox and Self-knowledge’’, in S. Shoemaker, TheFirst Person Perspective and Other Essays (Cambridge: Cambridge University Press),pp. 74–93.

Wittgenstein, L. (1974), Letters to Russell, Keynes, and Moore (Oxford: Blackwell).

7

Moore’s Paradox and the Transparencyof Belief

Jonathan E. Adler and Bradley Armour-Garb

A number of philosophers have thought to extend Moore’s Paradox to conceptsother than belief, e.g., I want Tina to go out with me, but I’m indifferent towhether she does. The scope of Moore’s Paradox is the problem of what conceptsbesides belief will generate a Moore’s (or Moorean) Paradox from the followingschema:

(MP@) I @ that (p, but it’s not the case that I @ that p),

where @ is a schematic variable ranging over elements of a delimited classof propositional attitude verbs (e.g., desire, belief). In particular, we want toinvestigate the question, ‘Is the Moore’s Paradoxical nature of the paradigm,

(MPi) p, but I don’t believe that p,

inherited from being instantiated in (MP@)?’ If so, what does this tell us aboutthe nature of belief ? If we compare (MPi) with another putative instance of(MP@),

(MPii) I desire that (p, but I don’t desire that p),

an immediate difference is evident—indeed, one that we take to hold generally:Only with (MPi) do we eliminate an explicit place for an instance of ‘@’. Soan affirmative answer to whether (MPi) inherits its Moore’s Paradox naturefrom (MP@) depends upon—but not only upon—whether assertion expressesor implies belief. We examine this matter by reference to a speech-act accountof illocutionary force and of the Unger-Williamson thesis¹—that, roughly, thenorm of assertion is that one properly asserts that p only if one knows it. As wewill show, reflection on this norm serves to elucidate the relevant notion of belief.

Thanks to John N. Williams and an anonymous referee for comments.

¹ Peter Unger, Ignorance (Oxford: Oxford University Press, 1975). Timothy Williamson,‘Knowing and Asserting’, Philosophical Review, 105/4 (1996), 489–523, repr. with minor additionsin his Knowledge and Its Limits (Oxford: Oxford University Press, 2000), ch. 11.

Jonathan E. Adler and Bradley Armour-Garb 147

The question of whether (MPi) fits under schema (MP@) leads us to criticalreflection on the property we regard as central to resolving Moore’s Paradox—thatbelief is transparent to its content: To believe that p is for it to be the case (fromone’s point of view) that: p.²

With others, we construe Moore’s Paradox as essentially first-personal.³ Doesthe first-personal transparency of belief fail for other instances of @ that mightbe held to generate variants of Moore’s Paradox? It does not appear to hold fordegrees of belief as with forms like:

(MPiii) I’m pretty sure that (p, but I am not inclined to believe that p).

In order to address these questions, we return to the beginning: Is Moore’sParadox paradoxical? That is, does it affirm a set of statements, each of whichis highly credible, and yet which are incompatible? If it is paradoxical, whatyields the paradox? If not, what does it tell us about the putative culprits that areimplicated, belief and assertion?

1. PARADOXICALITY AND MOORE’S PARADOX

Moore’s Paradox is perplexing or puzzling, and the task is to explain how itmight arise and how it is avoided. The puzzle is that a sentence that could betrue is ‘heard’ as inconsistent, when asserted. Since the apparent inconsistencyis among statements each of which has a plausible grounding, Moore’s Paradoxcan be presented in a paradox-like form. Colloquially, an instance of Moore’sParadox, which takes off from a sentence of the form

(1) p but I do not believe that p,

appears to affirm one thing only to deny it, or to take it back. The suggestionof a contradiction when asserted (or even believed) does not lead us to rejecta plausible assumption of the apparent paradox set, as is standard in diagnosingand treating a paradox; rather, it demands an explanation as to why it arises andwhy it is neither asserted nor believed.⁴ More specifically, the target questionbehind Moore’s Paradox is this: Why does assertion (or belief) transform aconsistent, assertible sentence into a recognized (and unacceptable) inconsistency?

² Our account draws upon that in Jonathan E. Adler, Belief ’s Own Ethics (Cambridge, Mass.:MIT Press, 2002), ch. 7.

³ For insightful reflections on Moore’s Paradox from the first-person point of view see SydneyShoemaker, The First-Person Perspective and Other Essays (Cambridge: Cambridge University Press,1996), ‘Moore’s Paradox and Self-Knowledge’, ibid. 359–70, and Richard Moran, Authority andEstrangement: An Essay on Self-Knowledge (Princeton: Princeton University Press, 2001).

⁴ It bears noting, for what follows, that one author (Armour-Garb) of this paper allows for theacceptance of some contradictions. While he finds ‘dialetheism’ an attractive and plausible view, herecognizes the virtue of consistency and, thus, aims to investigate (and resolve) Moore’s Paradoxwith that norm in mind. For more on dialetheism, see Armour-Garb, Dividing the (Semantic)Paradoxes (Stanford: CSLI, forthcoming).

148 Moore’s Paradox: Transparency of Belief

In order to answer this target question, we should note the two forms ofMoore’s Paradox:

A. p but I do not believe that p

and

B. p but I believe that not p.⁵

For what follows, we focus on the A form (herein, ‘MP’), as we take it to bemore basic than the B form, which form, as we establish below (§2), whilecontradictory, is not paradoxical. The task, then, is to investigate its maximallygeneralized form, in order to determine the scope of Moore’s Paradox.

There is, we maintain, a general form of MP, viz.

(2) p, but I do not @ that p

for ‘@’ an attitude verb.The ‘I’ in (2) is necessary, for only then are we ensured the self-recognition that

is needed for the proposition in question to be attached to the thinker/speaker insuch a way as to generate the symmetric, but negated, content with the secondconjunct. As already noted, the paradox is, essentially, first-personal.⁶

We maintain that no other attitude save belief will yield an MP as it stands(leaving aside for now ‘as if ’ belief attitudes, such as supposing, assuming, etc.),as no other attitude both links that attitude with assertion and has the uniqueproperty of transparency, which links a belief to what is believed: To believe thatp is to represent p as being the case. In order to make the case, we return, briefly,to the aforementioned link (we will develop the point further, in §3).

Sincerely to assert that p is both to put that p forward as true and to expressyour belief that p. This is no accident: By sincerely asserting something, youexpress that you believe it; to believe it is for it to be represented to you, thebeliever, as true; accordingly, to assert something is to put it forward as believedand, thus, as believed to be true. This explains why an instance of the A form isnot (and, indeed, ought not ever to be) asserted:⁷ In asserting a sentence of theA [MP] form, e.g.,

⁵ Anthony S. Gillies, ‘A New Solution to Moore’s Paradox’, Philosophical Studies, 105/3 (2001),237–50, discusses the B form, though his analysis does not seem to leave room for the contrastbetween it and the A form.

⁶ What is crucial is self-recognition and self-ascription. One can, of course, construct a versionof MP that relies on the first-personal report without the personal pronoun, as one can refer tooneself by name. Of course, such cases can be problematic (think of Kripke’s Pierre), so we ignorethem here (although, for more on such cases, see Dale Jacquette, ‘Identity, Intensionality, andMoore’s Paradox’, Synthese, 123/2 (2000), 279–92). Since both self-recognition and self-ascriptionare guaranteed by ascriptions (or thoughts) that make use of the first-personal pronoun, we willemploy it, for what follows.

⁷ In so claiming, we accept Rosenthal’s (‘Moore’s Paradox and Crimmins’s Case’, Analysis, 62/2(2002), 167–71) and Hajek and Stoljar’s (‘Crimmins, Gonzales and Moore’, Analysis, 61/3 (2001),208–13) treatment of Mark Crimmins’, case (‘I falsely believe that p’, Analysis, 52/3 (1192), 191),but maintain that the foregoing diagnoses it in a slightly different way.


Jones is rich but I do not believe that Jones is rich,

you express, in a single thought, that

I believe that (Jones is rich, but I do not believe that Jones is rich).

Thus, given that belief distributes over conjunction, and that all that is takingplace is within a single, overt thought, we get

I believe that Jones is rich and I believe that I do not believe that Jones isrich.

This thought is the same as (i.e., is logically equivalent to)

I believe that Jones is rich,I believe that I do not believe that Jones is rich.

Recall that the transparency of belief holds that to believe that p is for it to bethe case, for one’s point of view, that: p. From the first-person point of view,belief is factive, a key property of knowledge (if S knows that p then p).⁸ Fromthe first-person point of view, then, the last statement implies

I do not believe that Jones is rich.

There is now an explicit contradiction with the previous statement, ‘I believe thatJones is rich.’ In simple and obvious steps, we then have derived a contradictionfrom the nature of assertion (and belief) alone, which is what is necessary foran explanation of why the assertion of Moore’s paradoxical sentence is heard ascontradictory.

Our analysis assumes a strong parallel between belief and assertion. Assertionexpresses, not only implies, one’s belief; but it also the case that belief, likeassertion, claims the truth of its content, even though that claim only becomesapparent, upon attending to one’s belief. Thus, one stands behind it as true, asthe speaker stands behind his assertion as true.⁹

As no other attitude claims the truth of p, no other attitude is expressed byassertion, in which case no instance of (2), with ‘@’ a propositional attitudedistinct from belief, will yield a contradiction with the first conjunct. With theother attitudes, we need to add the @ explicitly, as in

(3) I desire [hope, guess, imagine] p, but I do not desire [hope, guess,imagine] p.

But there is nothing puzzling here, since the starting assumption of the consistencyof the sentence affirmed fails. (3) is, effectively, an instance of the explicitlycontradictory form:

C. @Ip & ∼@Ip.

⁸ Of course, we do not mean to imply that belief actually is factive. Rather, we mean that, inbelieving that p, the believer represents p as actually being the case.

⁹ We can then say that a speaker is entitled to assert that p (viz., that that p is assertible (for thespeaker)) only if she is in a position that entitles her to claim that what she asserts is true.


(3) and its ilk (for other attitudes) are contradictory sentences. So there is nodiscrepancy between the consistency of the sentence and the inconsistency ‘heard’upon assertion.

What of the B form,

(4) I desire [hope, guess, imagine] p, but I desire [hope, guess, imagine]not p?

The more general form here is

D. @Ip & @I ∼ p,

with the ascribed propositions overtly contradictory. Sentences of the D formindicate the same attitude towards contradictory propositions, rather thanan ascription of the same attitude toward a contradiction, and in the Dform, we surrender the crucial implication of (4) that the agent grasps ina single first-personal thought his same attitude toward the contradictorypropositions.

But what of the attitude itself ? Can we desire that p while simultaneouslydesiring its negation (in a single consciousness, as it were)? Assuming a preferencefor consistency, we can have no plan to implement, to act on, or to realize,both, as the result would be overtly self-defeating.¹⁰ But what of cases like this:I want Mary to come to the party (so I can meet her) but I don’t want [want notfor] Mary to come to the party (because she’ll bring her unpleasant pal David)?Once the tacit restriction to a certain respect is factored in the contradictiondissipates.

Or consider the following:

(5) I desire to eat ice cream (because of the taste) but do not desire [or:want] to eat ice cream (because it is not good for me).

This yields an inconsistency, if treated as of the C form. But, as is more plausible,it is believed as consistent—by speaker and hearer alike. That is, the sentence,when asserted, is ‘heard’ as inconsistent but is ‘treated’ (once interpreted) asconsistent. So interpreted, it expresses a first-order desire to eat ice cream, and,as well, a second-order desire not to have that desire, a form of conflict that wediscuss further below.

Whether these are actually inconsistent or not, they are not, and could notbe, cases of MP. For, even if contradictory, the contradiction arises from thesentence, rather than from the assertion of the sentence.¹¹ By contrast, in the

¹⁰ Cf. Michael Bratman, Intentions, Plans, and Practical Reason (Cambridge, Mass.: HarvardUniversity Press, 1987).

¹¹ It might seem that the analogy with (5) (and others like it) could be sustained by simplyprefacing assertions of the (MPii) form with the operator, ‘I believe that’. In fact, however, to assert‘I believe that p’ is to express a hesitancy to assert that p as all-out believed. It is, rather, to putforward a qualified, or a guarded, assertion that p, which is to convey a conversationally inducedweakening of the assertion of that p, rather than an unqualified expression of belief (consider, forexample, ‘I believe that George is in the office but do not hold me to it’).


case of belief, the contradictory nature of MP follows only from the MP sentence(in the A form) as asserted.¹² This enables us to distinguish the first form of MPfrom the second: An instance of the B form implies the inconsistency of what isbelieved, rather than inconsistencies of whether something is believed, as occursin an instance of the A form.

2. MOORE’S PARADOX AND VARIETIES OF BELIEF

Thus far, we have marked a difference between belief and other attitudes, insupport of our claim that no attitude but belief can (or: will) yield a MP, andby identifying both the features and the links that give rise to it. But ‘belief ’ is ablanket term that covers divergent attitudes of different degrees of strength. Aswe will discuss below (§3), we treat belief as full belief, as only full belief possessesthe needed transparency. Now, we consider a sharp contrast of belief with theother attitudes, which contrast emerges from a consideration of iterations ofsecond-order attitudes.

Consider the notorious datum that you can desire to smoke a cigarette butnot desire that you desire to do so, i.e., that attending to your desire does notinvariably yield a (second-order) desire, and so it goes for other attitudes (hoping,liking, etc.).¹³ Here we have a credible instantiation of (MP@), with ‘desire’ for@, i.e.,

(6) I desire that (I smoke this cigarette, but also that I do not desire that Ismoke this cigarette).

There is nothing contradictory about (6), whether asserted or thought, evenif it is awkwardly expressed.¹⁴ Also, and corroboratively, there is no differencebetween (6) and its second-person attribution:

(7) He desires that (he smoke that cigarette, but also that he does not havethat desire).

¹² Here we have a clear difference of the (B) form. In the (B) form an asserted instance wouldbe, effectively, to assert a contradiction, e.g.,

Jones is rich but I believe that Jones is not rich.

Treating ‘I believe’ here not as a qualifier (a weakener) of the assertion, but (in its standard way) astransparent, the assertion reduces to

Jones is rich and Jones is not rich.

Of course, to infer such a contradiction would be a confusion, given the pragmatically impliedweakening in the assertion of a self ascription of belief (cf. n. 11, above).

¹³ For a discussion of related issues, see John N. Williams, ‘Wittgenstein, Moorean Absurdityand Its Disappearance from Speech’, Synthese, 149/1 (2006), 225–54, §11.

¹⁴ In fact, it is not at all unusual for one to desire to have the opposed first-order desire, which is(roughly) an analogue of the B form, i.e.,

(6*) I desire that (I smoke this cigarette and that I have the desire not to smoke it).


What is, trivially, not possible is only for the second-order desire to succeed, andfor the first-order desire to remain.

But if you believe that the #3 train stops at Franklin Avenue, and if you attendto that belief, then (unless you revise your original belief) you believe that youbelieve it, which can be captured by the following abstract form,

(BI) B(p) (and attending to your belief that p) B(B(p)).¹⁵,¹⁶

What is involved is not a further endorsement of p, but is, rather, a non-optionaljudgment that one does so believe. That is, the recognition that it is true thatone has the belief that p is thereby to believe that one believes that p.

What then of degrees of belief—of uncertain, non-full, or partial belief ? Thereare a number of ways to think of these. We begin with

(8) Probably p, but it’s not probable that p.

This is just an instance of the explicitly contradictory D. If we move the negationin, i.e., to

(9) Probably p, but probably not p,

then we seem to have another instance of (8), given the negation law forprobability: For the probability of p to be less than .5 is for the probability of∼p to be greater than. 5. However, ‘not probable’ can be used as a really fullnegation, where one has no degree of belief in a proposition at all (Of course,this is forbidden in standard treatments).¹⁷

The following seem MPs as well, and so degrees of belief stand with belief,rather than with the other attitudes in this respect:

(10) p, but it’s not probable that p,

understood as something like

(11) p, but the probability of p is less than half,

which may (but need not) be heard as something like

(12) p turned out to be true, despite its being improbable,

¹⁵ The same goes for

(BE) B(B(p)) (and attending to your believing it) B(p)

and

(B∼E) B(∼B(p)) (and attending to your believing it) ∼B(p),

as we hope is obvious. We should note that we employ the turnstile because, as is standardly used,it captures both truth preservation and inferential entitlement, both of which we apply to thepresent cases. Because constitutive of belief is only the judgment or claim that p, we think that theseiterative belief principles do not run afoul of Williamson’s powerful criticisms of the assumption ofluminosity (Knowledge and Its Limits, ch. 4). But we cannot address the issue here.

¹⁶ Given (BI), considerations from n. 4, and (BE), and assuming that we are attending to ourbeliefs (NB, even believing that we believe), we get, and so are entitled to infer (from the attending)

(BE/O) B(p) B(B(p)).

¹⁷ So, for example, we have no degree of belief that Hamlet blinked an even number of times onthe day of his sixteenth birthday, nor do we have a degree of belief in its negation.


and which, if so heard, is not only consistent (as asserted), but is, in fact, familiar.What then of epistemically qualified assertion, as in (e.g.)

(13) I’m pretty sure that (p, but I don’t believe that p [highly])

or

(14) I’m pretty sure that (p, but it’s doubtful that p)?

(13) either itself, or as asserted, is not just consistent, but possibly an implic-ation like the following: ‘I’m [just] pretty sure that Jones is in Alaska, so,of course, I do not [fully] believe it.’ Likewise, (14) seems intelligibly under-stood as e.g. ‘I’m [just] pretty sure that Jones is in Alaska, though I haveacquired reasons to doubt that he is.’ Or, better, with ‘because’ substituted for‘though’.

However, assertions as expressions of all-out belief (or acceptance as true)must represent themselves as satisfying a total relevant available evidence con-dition. The body of evidence grounding each conjunct must be the samebody of evidence that purports to be the total relevant evidence. Not onlyis this evidence all that the agent has available, but in accepting that pro-position the agent implies this is all the evidence necessary to establish it astrue.

Qualified assertions like (13) or (14) need not have the latter implication. Theyremain open to inquiry, to the discovery of further relevant evidence, and so toan alteration in the degree of qualification. They are failed assertions because thefundamental goal of assertion is to transmit truths about the world, rather thanabout one’s attitude toward it (except, of course, in the odd cases in which theseare of interest to the hearer). So, for example, in asking after the whereabouts ofJones, what you want to know is where Jones is, not where the speaker thinksthat he is. Accordingly, it does not satisfy your interest in being so informed, tolearn only that the speaker is pretty sure.

Also, notice that these examples fall in place with the other propositionalattitudes, in having to explicitly represent the attitude in the sentence, and unlikeeither (full) belief or the basic case, A.¹⁸ But why is this so? We maintain that thereason turns on certain features of full belief, which have ramifications for other(familiar and neighboring) paradoxes.

3 . BELIEF AND TRANSPARENCY

A central difference between full belief and partial, or qualified, belief (in additionto other attitudes) is that only the former is transparent to its content. Recall:For you fully to believe that p is for it to be true that p (from your first-personal

¹⁸ As a consequence, the logic of the statement is the same as the logic of the assertion—markedlyunlike cases under (A).


point of view).¹⁹ Thus, full belief is not essentially any strength of attitude—you‘disquote’ from your attitude.

Now consider any strength of belief-like attitude in an MP version:

E. p, but I’m not very sure that p.²⁰

Although, in this stark form, the E form assertion seems puzzling, and so calls forthe speaker to enter some explanation, such assertions (or their kin) are assertible,when construed literally. In order to see this, consider

(15) Suzy loves me, but I’m having doubts about it.

Nervous Ned explains: ‘All the signs are that Suzy loves me, and we getalong great. But that hunk Tony just walked by, and I thought he mightbe attracted to Suzy.’ Assuming that Nervous Ned remains competent tojudge his love-relations, his expressions of doubt allow for the assertibil-ity of his (psuedo-) MP, and its literal acceptance: Both conjuncts of (15)can be true (i.e., true-from-Ned’s-point-of-view), and assertible, without para-dox.

We think that this result holds quite generally. It undermines any attempt toread full or all-out belief as too weak to imply the truth of what is believed (fromthe first-person point of view).²¹ For if belief is any strength of attitude, it shouldexclude doubt, as in the standard Peircean view.²²

4. THE NORM OF ASSERTION AND BELIEVING WHAT

YOU KNOW

Recall that the Unger-Williamson thesis maintains that, in asserting that p, onerepresents oneself as knowing that p. Along a different vein, Moore claims that‘by asserting p positively, you imply, though you don’t assert, that you know thatp.’²³ We are sympathetic to the claim that (sincerely) asserting that p representsoneself as knowing that p and, following Williamson, will assume ‘the KnowledgeRule’ that tracks it, i.e., the norm according to which you must assert that p onlyif you know it. Rather than arguing for the Knowledge Rule, we shall employ it,together with reflection on full belief and assertion.

¹⁹ By contrast, partial belief ‘marks’ a qualified, or guarded, assertion, rather than a bald assertionof that p itself. The uptake of partial belief is that it transforms putatively paradoxical utterances intonon-paradoxical, even familiar, ones. Thus, consider a paradoxical (or incoherent) assertion of theform, ‘p but it is not very probable that p’.

²⁰ For a related variant of MP and a different diagnosis, see Byeong Lee, ‘Moore’s Paradox andSelf-Ascribed Belief ’, Erkenntnis, 55/3 (2001), 359–70.

²¹ Here, and in what follows, we use ‘full’ and ‘all-out’ interchangeably.²² Cf. Charles, Peirce ‘The Fixation of Belief ’, in his Essays in the Philosophy of Science, ed. V.

Thomas (New York: The Bobbs-Merrill Co., 1957), 3–30.²³ The Commonplace Book of G. E. Moore 1919–1953, ed. Casimir Lewy (London: George Allen

& Unwin Ltd; New York: Macmillan Co., 1962), 125.


The Knowledge Rule states:

(KR) One must: Assert that p only if one knows that p,

where ‘must’ expresses a kind of obligation and where the rule thus yields thatonly knowledge warrants assertion. As Moore, Unger, and Williamson note,²⁴we seem to get just a variant of MP, with ‘believe’ replaced by ‘know’. Given(KR), this is easily explained: If only knowledge warrants assertion then to havewarrant to assert the conjunction,

(MP*) p but I do not know that p,

is for the asserter to know that (p and that he does not know that p). Assumingthat one can know a conjunction only if one knows each conjunct then, if one haswarrant to assert an instance of (MP*) then one knows the first conjunct— thatp. But one knows a conjunction only if it is true, which, via disquotation and theaforementioned norm, yields that one does not know that p. Thus, (KR), and theassumption that the speaker has warrant to assert an instance of (MP*), yields acontradiction, thereby supporting the paradoxicality (and, so, the unassertability)of (MP*).²⁵

The knowledge rule forbids the combination: one asserts that p when onedoes not know that p. Thus, if (KR) is in place, we are obliged to assert that ponly if we know it; and, if we do not know it, we are obliged not to assert it.Of course, we violate (KR) all of the time, as Williamson is well aware. But thenorm motivates an observation that will play a role in what follows:

To make an assertion is to confer a responsibility (on oneself) for thetruth of its content; to satisfy the rule of assertion, by having the requisiteknowledge, is to discharge that responsibility, by epistemically ensuring thetruth of the content.²⁶

Granting the obligation (which we do), it does, however, raise a question: Evenif we must assert that p only if we know it, under what conditions do we takeourselves to be entitled to assert it? To be sure, one might reply that we areentitled to assert that p when we have warrant (and, thus, authority) to do so, andthat we have warrant to assert it only when we know it. But an infinite regress

²⁴ Ibid. 277; Unger, Ignorance, 256–60; and Williamson, Knowledge and Its Limits, ch. 11.²⁵ Beyond noting this key argument for (KR), the Unger-Williamson thesis, and a cautionary

word, we shall just assume them.The cautionary word is this. Someone might naturally and sincerely (and perhaps dejectedly)

assert, about the outcome of a fair lottery in which he holds a single ticket,

(15∗) I’ll lose.

In so asserting, the speaker surely does not imply that he knows that he will lose. However, in thiscase, it is presumed as mutually evident that a lottery is a gamble and that its outcome (like manykinds of future events) cannot be known in advance. So there is a mutual accommodation to allowfor the assertion of (15∗), rather than for what is, strictly speaking, correct, viz.,

(15∗∗) I’m almost certain I’ll lose.

²⁶ Williamson ‘Knowing and Asserting’, 521–2.


threatens, if our entitlement to assert depends on a confirmation of its warrant,which notions, of course, are not identical.

This, however, is a mistake, although an important one. The knowledge rule,qua norm, is taken to be constitutive of the speech act of assertion. As such, ittells us the conditions under which we have the authority to perform that act.But it does not follow from (KR) that if we do not know that which we have justpurported to have asserted (because it is false, say) then we have failed to assertit. We have so asserted, but we oughtn’t to have done so: We violated the rules,rather than failing to perform the speech act.

But this does raise a question: How, and under what conditions, do we takeourselves to be entitled to assert? The answer, it seems, is clear, assuming (KR):²⁷As we take ourselves to be entitled to assert that p only when our assertionconforms to (KR), we take ourselves to be entitled to so assert when we takeourselves to know that p, i.e., we take ourselves to be entitled to assert that whichwe believe that we know.²⁸,²⁹

Recall (BI)—that if one attends to her belief that p then she believes that shebelieves it. Now suppose that the thinker attends to her belief and that what shebelieves is presented to her as fully believed and, thus, as transparent. In thatcase, if she attends to her belief, so that its content appears to her transparently,then, we maintain, she believes herself to know that content, in which case shetakes herself to be entitled to assert it, as it is presented to her as known, even if,in fact, it is false, she does not know it, etc. So, what is fully believed is therebytaken by the believer to be known; and, insofar it is so taken, it is taken to beassertible, in accordance with the Knowledge Rule.³⁰

If this is right then we have a ‘test’ for determining both full belief and theappearance (to a speaker) of assertability: When one attends to her belief andwhen the content of that belief is presented, to the believer, as transparent, then

²⁷ Williamson (ibid., passim) makes clear that he takes the knowledge rule to be the actualnorm of assertion—the norm that we actually employ and that guides our practices of asserting.Accordingly, and assuming the norm, we take it to play a role in elucidating the actual concept offull belief.

²⁸ Williamson (ibid. 512–13) argues against a norm that bears a resemblance to our presentproposal. The norm he rejects is the BK rule,

(BKR) One must: assert that p only if one believes that one knows that p.

It is important to note that the present proposal does follow Williamson in rejecting (BKR). Thepresent proposal is to treat an element of (BKR) as related to (KR), by setting out a condition underwhich we take ourselves to be entitled to assert: We take ourselves to be entitled to assert that pwhen we have a full belief that p, and we have a full belief that p when, having attended to ourbelief, we come to believe that we know it.

²⁹ Notice, in addition, that, if correct, this is related support for the claim that full belief is notessentially any strength of attitude.

³⁰ One might object that, while it may well be true that we take ourselves to be entitled to assertsomething, when we take ourselves to know it, it is wildly implausible to assume that, when weattend to a belief, that belief is then represented to us as known, in a non-metaphorical way. Thisobjection seems to be based on a mistake, for, when one attends to, or is guided by, a full belief thatp, what she attends to, or is guided by, is that p, rather than one’s attitude toward it.


she, in effect, believes that she knows it and so takes herself entitled to assert it,as it is what she fully believes.

5 . FULL BELIEF AND FALLIBILITY

Try as we might, and with only the best of intentions, we are liable to assertor believe things that are false and are liable to disbelieve things that are true.That is, just as the preface-assertion (‘at least one statement in this book isfalse’) is likely true, though not assertible, the same is true of (what we call) thefallibilist-assertion viz.,

(FA) I believe at least one false proposition.

(For now, we put aside the pragmatic weakening implied by the ‘I believe’, andtreat this as an accurate report in thought.)³¹

Since we have such an extensive and wide-ranging set of beliefs, it would behubris not to be sure that some of them are false (we are, after all, not omniscient).So (FA) is likely—perhaps, inevitably—true, and, if it is believed, our full set ofbeliefs is inconsistent.

But, if (FA) is asserted, we can run an argument that parallels the PrefaceParadox, with the same disastrous results. We review it briefly, with especialattention to the standard presentation of the Preface Paradox, with which webegin.³²

5.1 The Preface Paradox

In a version of the Preface Paradox applicable to all, Smith believes each of anenormous and diverse set of propositions:

Bp1, Bp2, . . . , Bpn.

Call the set of propositions that Smith believes ‘Smith’s corpus’. As Smith isminimally fallibilist, he thinks that some of the members of the his corpus areerroneous, i.e.,

B∼(p1 & p2 &, . . . & pn).

Call the proposition that he believes, the preface-proposition. What Smith takesto be erroneous—what his fallibilism appears to commit him to—is not oneor another of the members of the corpus; rather, it is that he thinks of thefull range of members of the corpus as likely to contain an error. His fallibilistbasis for his inference to the preface-proposition is second-order—a reflectionon his believing (the members of) the corpus, and not on the content—theparticulars—of what he believes.

³¹ See nn. 11 and 12.³² D. C. Makinson, ‘The Paradox of the Preface’, Analysis, 25/6 (1964), 205–7.


The paradox depends on assuming that the person not only fully believeseach pi, but the fallibility-proposition, as well. If so, the person’s beliefs areinconsistent. However, it is standard to claim that, whereas inconsistent beliefscould be rational to maintain, only an outright belief in a contradiction isforbidden:

B[(p1 & p2,&, . . .& pn) & ∼(p1 & p2 &, . . .& pn)].

We can avoid this last step if we reject a number of assumptions—mostprominently, a conjunction rule (to believe each of a set of propositions is tobelieve their conjunction). However, this route is not open to us because, giventhe transparency of belief, we can ‘detach’ from the ‘B’ operator, and then we doget the out-and-out contradiction:

(p1 & p2,&, . . .& pn) & ∼(p1 & p2 &, . . .& pn)

by logical equivalence (of p1,. . .,pn and p1&. . .&pn). Instead of rejecting theconjunction rule, we reject the fallibility proposition—no one all-out believesthat one of her beliefs is false.³³

In a natural framework of assertion, the person asserts each sentence thatexpresses these propositions—including his denial of their joint truth (viz., thefallibility-proposition). In the context of assertion, the conjunction principleholds:

There is no significant contrast between a conjunction of assertions and anassertion of a conjunction.³⁴

Now while we do find people saying things like ‘Blah-blah-blah, though ofcourse I’m fallible, I could be mistaken,’ if we take this expression in a woodenlyliteral way, rather than as a rhetorical device, we seem to have an overt expressionof belief that is unassertible because blatantly inconsistent: The speaker claims aset of propositions (simultaneously) true, and then denies it.³⁵

Of course, as a practical matter, we are not held to all our assertions over time.We forget, and we change our mind without in each case explicitly having towithdraw a distant assertion.

The connection between the Preface Paradox and Moore’s Paradox is eventighter, given that disbelief implies a lack of belief.³⁶ But, in addition, a problem

³³ Like Moore’s Paradox, as well as (FA), the Preface Paradox is essentially first-personal. To seethis, consider the fact that all or most of us can (and do) all-out believe that at least one of anothercognizer’s beliefs will be false. And this is so even though we may not know which ones are theto-be-rejected culprits.

Something similar applies, when we take a third person attitude towards ourselves qua believers.When we do, we find that we can say, for example, that at least some of our beliefs are likely false.Thanks to an anonymous referee, who pressed us to address this issue.

³⁴ Michael Dummett, Frege: Philosophy of Language (2nd edn. Cambridge, Mass.: HarvardUniversity Press, 1981), 336.

³⁵ We assume, of course, that, in asserting, a speaker represents himself as recognizing explicitrelations like negation.

³⁶ So that the (B) version implies the (A) version.


emerges: Whereas the unassertibility of Moore’s Paradox corresponds to anincoherence of thought, the Preface Paradox develops a line of coherent thought.Thus, many follow Makinson in taking the right response to the Preface Paradoxto be that one can rationally be involved in (recognizable) inconsistency.³⁷Accordingly, in order consistently to resolve the Preface Paradox, one mustresolve it in such a way as to explain the coherence of the preface-assertionwithout its assertability.

5.2 A Fallibilist Paradox³⁸

As fallibilists, we recognize that it is overwhelmingly likely that some of our beliefsare false. Thus, like the preface-assertion we are faced with the fallibilist-assertion(FA), and, with it, the contradiction in belief that appears to ensue. This is theFallibilist Paradox. In both paradoxes, we seem faced with a dilemma: We havederived a contradictory belief from a collection of beliefs each of which we accept.³⁹

5.3 Resolving the Preface and the Fallibilist Paradoxes

Our resolution consists of two related parts, already suggested and implicit inour account of Moore’s Paradox: First, the fallibility-proposition is not all-outbelieved and, second, the justification for the fallibility-proposition, and theproper expression of fallibility, is second-order: It is to cast doubt on ourperfection as believers, not to cast doubt on what we believe (viz., the first-orderpropositions).⁴⁰

There is no inconsistency in (fully) believing each of a set of propositions, andhaving a high degree of belief that not all of them are true—indeed, somethinglike the latter follows from the conjunction principle for probability, under very

³⁷ Cf., e.g. Graham Priest, ‘What is So Bad about Contradictions?’, Journal of Philosophy, 95/8(1998), 410–26. It bears pointing out that dialetheists, while they believe that some contradictionsare both true and assertible, do not accept that all contradictions are true and assertible. Thus,accepting dialetheism by no means entails that dialetheists will (must, should) take the relevantinstance of (e.g.) a Moore-style sentence to be genuinely inconsistent. Indeed, dialetheists, liketheir consistentist kins, subscribe to a consistency-governed norm; they simply maintain that thereare reasons for accepting some contradictions. For the details, see Graham Priest, J. C. Beall,and Bradley Armour-Garb (eds.) The Law of Non-Contradiction: New Philosophical Essays (Oxford:Oxford University Press, 2004).

³⁸ The indefinite article is used in recognition that there is already a‘fallibility paradox’ (cf. KeithLehrer and Kihyeon Kim, ‘The Fallibility Paradox’, Philosophy and Phenomenological Research, 50,suppl. (1990) 99–107). Ours is distinct from theirs, but indebted to their presentation (Oxford:The Clarendon Press, 2004).

³⁹ Although the presentation of the paradox is fairly standard, one might balk at the assumptionsthat drive it. To assuage the reader’s worry, consider the following: If one accepts the preface-assertion and the fallibilist-assertion then a contradiction can be adduced, from the fact that, foreach proposition that we believe, qua belief, we believe it to be true.

⁴⁰ See also Simon Evnine, ‘Learning from One’s Mistakes: Epistemic Modesty and the Natureof Belief ’, Pacific Philosophical Quarterly, 82/2 (2001), 157–77.


weak conditions. Moreover, the denial of all-out belief fits with one’s treatingthe fallibility-proposition as an epistemic possibility that all of one’s beliefs arecorrect.⁴¹ This, in turn, serves to explain why the fallibility-proposition is notall-out believed (though it may be partially believed) and so why it is not judgedto be assertible by the believer, even if, in fact, it is true.

Can one take oneself to know the fallibility-proposition? We maintain that onecannot, since (i) one will take it to be epistemically possible, even if wildly unlikely,that all of her beliefs are true, and (ii) knowledge implies that it is not possible(from one’s point of view) that one is mistaken. We grant that it is overwhelminglylikely (from your point of view) that at least one of your beliefs is false. But wedeny that there is any basis for detaching from this subjective probability to theall-out belief that the fallibility-proposition is true. The situation parallels that ofa fair lottery: There is no ground to detach from the overwhelming probabilitythat one will lose (with a single ticket) to take it as simply true (as if the lotterywinner had already been announced) that one will lose.

That said, there does seem to be something that we believe and, thus,something that we take ourselves entitled to assert. But, as our belief is merelypartial, e.g., of the sort that it is likely that at least one of the sentences of my book isfalse (though I know not which it is), the assertion that expresses such a (partial)belief is guarded (or qualified) as well. The same holds for the fallibilist-assertion:It is guarded or qualified because what it expresses is a partial belief, whichlacks the transparency that is a hallmark of full belief. But the (qualified orguarded) assertion that at least one of the sentences of my book is false is, itself,neither paradoxical nor problematic, even given the full belief in each of thesentences of the book. Thus, the paradox is diagnosed— it arises from the falseassumption that we fully believe the fallibilist-proposition—and treated— wecan fully believe our qualified belief, which belief thus does not (and, indeed,cannot) yield paradox.

We have claimed that a key to resolving these paradoxes is to resist thetemptation to read such beliefs (like the fallibility proposition) as full beliefs.A further reason for not treating the fallibility-proposition as all-out believed isthat, contra the consequences of the Preface Paradox, where full belief in thecontent of each of the sentences of the author’s book is retained, in the case ofthe Fallibilist Paradox, the analogous belief casts doubt back on the individualbeliefs.

So, for example, suppose that, after grading the final exams for your logic class,you discover that no students have scored below 90 per cent, which is far abovethe class average. You consider each student and provide reasons for thinkingthat student to be honest and so conclude, of each of the many students, that that

⁴¹ Indeed, if one believes, of each of his beliefs, that it is true, he will not (and cannot,without contradiction) deny the epistemic possibility that all of what he believes—including theirconjunction—are true.


student did not cheat. But the class average is just too high and so you concludethat at least some of the students must have cheated, in spite of the fact that youdo not have evidence that would enable you to convict any particular student.

It seems then that a dilemma ensues, for it seems that you (fully) believe eachof the students to be honest but you also believe that someone has cheated. As inthe previous cases, the dilemma is resolved by noticing that you would not (or:should not) all-out believe that each of your students have performed the examhonestly.

Accordingly, this case is contrasted with the Preface Paradox, where the warrantfor the preface-assertion is simply the high probability that one of p1, p2, . . . ,pn isfalse, which licenses no inference to the preface-assertion itself. In the case of thestudents, there is a basis for an inductive inference: The best explanation for thecoincidence is that some students cheated (not merely that this is probable).⁴²

In the case of the exam, once you conclude that something has gonewrong—that it is overwhelmingly likely that someone has cheated—you findthat you can no longer fully believe that each of the students are honest testtakers, and this is the case consistent with your fully believing that some of yourstudents have cheated.

6 . CONCLUSION

Our account implies the controversial claim that instances of the following arevariants of MP and, thus, are paradoxical, if asserted or fully believed:

p, but I lack sufficient evidence that p.p, but my reasons do not establish that p.

Any hearer would refuse to accept the former conjunct, given an assertion ofthe latter. (Recall previous discussion that assertion and all-out belief both claimto satisfy a condition of total relevant available evidence as representative of thepossible non-misleading evidence.)

For another example, consider

(16) There are no misspellings in my book, and I have not checked the lastsection.

Reasons that establish the truth of p would normally be taken as reasons sufficientto know that p. Reflection on these examples amounts then to an extension, aswell as to a corroboration, of our basic analysis. More generally, it seems that anystatement of the form

p but I M that p,

⁴² Analogously, while you do not all-out believe (FA) (you partially believe it), your partial beliefdoes not undermine the epistemic status afforded to the content of what you fully believe.


if all-out believed, will be a version of Moore’s Paradox, if M serves to cancel thegrounds or reasons for fully believing that p. But no paradox ensues, or is heard,from an assertion of a qualified form, such as

I am almost certain that p but I do not know that p.

This, we maintain, provides further support for our diagnosis of Moore’s Paradoxand, with it, the factivity of full belief, from the first-person point of view.

PART IV

MOORE’S PARADOXAND CONSCIOUSNESS

8

Consciousness, Reasons, and Moore’s Paradox

André Gallois

On a number of occasions Wittgenstein discusses what he calls Moore’s Paradox.Moore himself discussed the phenomenon Wittgenstein so labels in his paper,‘Russell’s ‘‘Theory of Descriptions’’ ’, and in a reply to Charles Stevenson.¹Moore’s own treatment of the paradox is perfunctory. He clearly thinks of theproblem posed by examples of Moore paradoxicality as having minor significancewith a relatively straightforward solution. Evidently Wittgenstein disagreed.Norman Malcolm reports that Wittgenstein thought of Moore’s Paradox asMoore’s great discovery. Wittgenstein was right to attach the importance he didto the puzzling phenomenon Moore identified.² One reason he was right todo so is, far from being an isolated logical curiosity, it plays a role in a largenumber of central philosophical issues. I propose to investigate the bearing ofMoore’s Paradox on some aspects of the relation between consciousness andhaving reasons.

Here is how I will proceed. In the first section, after, all too briefly, discussingthe distinguishing feature of Moore-paradoxical statements and thoughts, I iden-tify an instance of Moore-paradoxicality that directly links Moore-paradoxicalityto consciousness. In the same section I go on to say why a standard treatmentof Moore-paradoxicality fails to yield a persuasive account of that instance.In the second section I distinguish between two ways of attributing a reasonfor being in a psychological state with propositional content such as believing,imagining, or intending. In the third section I elaborate on the link betweenMoore-paradoxicality and the thesis, which I defend, that only consciously heldbeliefs can supply reasons. The fourth section is devoted to exploring the link

¹ G. E. Moore, ‘Russell’s ‘‘Theory of Descriptions’’ ’ in Philosophical Papers (London: Allen andUnwin 1959), 151–95; Moore’s discussion of the paradox is on pp. 172–7. His later discussion ofthe paradox is on pp. 542–3 of P. A. Schlipp (ed.), The Philosophy of G. E. Moore (La Salle, Ill.:Open Court, 1942).

² To entitle that phenomenon ‘Moore’s Paradox’ is somewhat misleading in something like theway ‘Naturalistic Fallacy’ is misleading. Just as the alleged mistake entitled the Naturalistic Fallacyis no fallacy so ‘Moore’s Paradox’ labels a phenomenon which, however puzzling, is not a paradox.At least it is not a paradox in the sense of presenting something which is evidently true that seemsto imply something evidently false.

166 Consciousness, Reasons, Moore’s Paradox

between Moore-paradoxicality and what I call the deletion and augmentationfunctions of consciously held belief. What emerges is a connection betweenconsciously held belief having those functions, and one being able to form anintegrated view of the world. In the fifth section I attempt to explain whatis meant by an integrated view of the world by invoking a comparison withintegration in fiction. The sixth section provides an account of the distinctionbetween two ways of attributing a reason introduced in the second. Finally, inthe last section I utilize the materials developed in the preceding ones to give anaccount of the target example of Moore paradoxicality relating to consciousness.

1 . MOORE-PARADOXICALITY

Let us use ‘Moore-paradoxicality’ to label a feature that is shared in common byinstances of, among others, the following schemas:

(1) P, but I do not believe that P,

(2) P, but I believe that not-P. ³

What is it about (1) and (2) that would lead us to classify them as Moore-paradoxical?⁴ Here we need to be careful. In one way it is unfortunate that‘Moore-paradoxicality’ has become such an inevitable label for the feature Moorecalled our attention to since it strongly suggests that the thoughts and statementspossessing it, or sentences expressing those thoughts and used to make thosestatements, are, in virtue of doing so, automatically defective. We should resistthat suggestion. Whatever we think of eliminativism about propositional attitudesthere is nothing defective in an avowed eliminativist’s assertion that eliminativismis true, but not believed by her to be so. Likewise there is nothing defective abouta dialethist who believes in true contradictions producing an instance of (2).

So ‘Moore-paradoxicality’ should not be taken to pick out a feature whosepossession automatically indicates its possessor is defective. Since that is so weshould not identify Moore paradoxicality with a certain form of irrationality, orbad type of inconsistency. We need to separate what Moore-paradoxicality isfrom a diagnosis of why it is usually a bad thing to be.

What then is Moore-paradoxicality? We may be tempted to say that to beMoore-paradoxical is just to be an instance of (1) or (2), and leave it at that. Heretoo we should resist temptation. Additional examples of Moore-paradoxicalityare provided by:

³ I am making the assumption that there is some feature, other than a disjunctive one whosedisjuncts are not significantly connected, that (1) or (2) have in common that deserves to be calledMoore-paradoxicality.

⁴ Strictly it is instances of (1) or (2) that are Moore-paradoxical. However, since a statement orthought will be Moore-paradoxical in virtue of instancing a certain schema I will refer to a schemaof the appropriate type as Moore-paradoxical.

Andre Gallois 167

(3) P, but I am not at all justified in believing that P,

(4) P, but I do not know that P,

(5) P, but it might not be the case that P,

and by;

(6) P, but it is not at all likely that P.

Despite this proliferation we may well think that we can discern the nature ofMoore-paradoxicality even if we restrict our attention to (1) and (2). As a numberhave pointed out, Moore-paradoxicality seems to arise in the most basic way atthe level of thought. When we ask what makes thinking any of (1) or (2) Mooreparadoxical many have found the following answer attractive. Consider (1).Belief plausibly distributes over conjunction. That is, believing both P and Qimplies believing P and believing Q. Some hold that, at any rate, conscious beliefiterates. That is, consciously believing P implies believing that one believes P.In addition, suppose that the following rationality condition holds for belief.Believing that P implies not believing that not-P. If distribution, iteration,and the rationality condition hold for conscious belief then it is impossiblefor there to be a belief in an instance of (1) that is consciously held.⁵ What,on this view, makes (1) Moore-paradoxical is that a conscious belief in (1) isruled out.

Advocates of this account of the Moore-paradoxicality of (1) allow that theprinciples dictating the impossibility of consciously believing (1) do not rule outconsciously believing (2). What those principles imply is that a conscious believerof (2) has inconsistent beliefs. By distribution, consciously believing P, but Ibelieve Not-P implies consciously believing P and believing that one believesNot-P. In turn, consciously believing P implies believing that one believes thatP. So, one who consciously believes P has the belief that she believes P as well asthe belief that she does not believe P.

Call an account of Moore-paradoxicality along these lines the propositionalattitude account.⁶ We may reasonably hope to extend the propositional attitudeaccount to cover (3)–(6). Clearly, (3) and (4) implicate belief.⁷ Moreover, on

⁵ The argument for this being so goes as follows. Suppose you consciously believe some instanceof (1). Say you believe both that Moore is a philosopher and that you do not believe that Moore isa philosopher. By distribution you consciously believe that Moore is a philosopher. So by iterationyou believe that you believe that Moore is a philosopher. But by distribution you also believe thatyou do not believe that Moore is a philosopher. So by the rationality condition you do not believethat you believe that Moore is a philosopher.

⁶ The above version of the propositional attitude account is given, without endorsement, byRoy Sorenson in ‘Moore’s Problem with Iterated Belief ’, Philosophical Quarterly, 50/198 ( Jan.1998), 28–43. Somewhat different versions of the same account are defended by Sidney Shoemakerin a number of papers reprinted in his collection The First Person Perspective and Other Essays(Cambridge: Cambridge University Press, 1996), and Uriah Kriegel in ‘Moore’s Paradox and theStructure of Conscious Belief ’, Erkenntnis, 61/1 (2004), 99–121.

⁷ That said, how to extend the propositional attitude account to cover, for example, (4) is byno means obvious. Consciously believing (4) does not imply believing that one does not believe


one natural reading (5) and (6) may be taken to do so as well. After all themodality in (5) is plausibly epistemic, as is arguably the probabilistic estimateimplied by (6).

Despite that, the propositional attitude account should be rejected as acomprehensive account of Moore-paradoxicality. As Moore in effect pointed out,there are examples of Moore paradoxicality that do not, directly or indirectly,involve the denial or attribution of a propositional attitude. For example there isas much reason to hold that:

(7) ‘Snow is white’ does not mean that: snow is white,

is Moore-Paradoxical as there is to so regard any of (1)–(6) as being so.⁸Moreover, if (7) is Moore paradoxical, so is:

(8) Napoleon is not called ‘Napoleon’.

What then make all of (1)–(8) Moore-paradoxical?Suppose we agree that beliefs, sentences, statements, and singular terms are, in

some suitably broad sense, representations, and, in that light ask what, if anything(1) to (8) have in common relevant to characterizing Moore-paradoxicality.Compare the following instances of (1):

(1∗) Syracuse is in New York, but I do not believe that Syracuse is in NewYork

and:

(2∗) I believe that Syracuse is not in New York, but Syracuse is in NewYork,

with the Moore-paradoxical:

(7) ‘Snow is white’ does not mean that: snow is white,

and;

(8) Napoleon is not called ‘Napoleon’.

that P. Tim Williamson in Knowledge and Its Limits (Oxford: Oxford University Press, 2000) hasattempted to account for the Moore paradoxicality of assertions of instances of (4) by appealingto a constitutive rule governing assertion. Since Williamson’s account rests on such a rule it is notobvious how it would apply to the Moore paradoxicality of thinking an instance of (4).

⁸ In ‘Russell’s ‘‘Theory of Descriptions’’ ’, Moore does not initially give instances of any of (1) or(2) to illustrate what has come to be known as Moore paradoxicality. Instead he gives as an openingexample ‘The sentence ‘At least one person is the King of France’ means that at least one personis the king of France.’ He observes that it would be a mistake to think that the last mentionedsentence is a tautology. Later, using ‘Z’ to stand in for the sentence ‘At least one person is the Kingof France’ Moore remarks ‘Hence if we were to assert ‘‘Z does not mean that at least one personis the King of France’’ we should imply that Z can be properly used to mean what, on the secondoccasion on which we are using it, we are using it to mean’. He continues ‘To make our assertion bythe use of this language is consequently absurd for the same reason [my italics] for which it is absurdto say such a thing as ‘‘I believe he has gone out, but he has not’’ ’. Moore, ‘‘Russell’s ‘Theory ofDescriptions’’ ’, 173–5.

Andre Gallois 169

Consider sentence (7). Sentence (7) contains the sentence ‘Snow is white’mentioned on the left, and used on the right. Sentence (7) says somethingtrue only if, in virtue of having a meaning, it represents something being so.Moreover sentence (7) represents something being so only if any, used rather thanmentioned, sentence it contains represents something being so. No sentence thatlacks a meaning represents something being so. Hence, (7) represents somethingto be so which, if it is so, fails to represent anything to be so.

Sentence (8) conforms to the same template. Sentence (8) represents somethingto be so only if the name ‘Napoleon’ represents someone that sentence (8) saysit does not represent.

Now consider (1∗) and (2∗). Taking thoughts to represent things being sowe can say this about (1∗) and (2∗). A thought that (1∗) represents Syracuseas being in New York, but also represents the thinker of that thought as notrepresenting, in thought, Syracuse as being in New York. Likewise, a thoughtthat (2∗) represents Syracuse as being in New York, but also represents its thinkeras having a thought that represents the opposite.

This suggests an account of Moore-paradoxicality which may be called the rep-resentational account. Applied to our initial examples of Moore-paradoxicality,(1) and (2), the representational account tells us that thinking (1) or (2) implieshaving thoughts that represent in conflicting ways. For example thinking:

(1) P, but I do not believe that P,

implies having a thought that represents P being so, and also having the thoughtthat one does not so represent P. Asserting (7) by uttering sentence (7) requiresusing a linguistic representation that, according to (7), fails to represent. Asserting(8) by means of uttering sentence (8) likewise requires using a name to representan individual who, according to (8) is not so represented.

No doubt the representational account could be more precisely elaborated.I shall not attempt to do so since it is, at best, unclear how it would apply tothe following example of Moore-paradoxicality which plays a major role in theremainder of the paper:

(9) P, but I only non-consciously believe that P.

Consider this instance of (9):

(9∗) Syracuse is in New York, but I only non-consciously believe thatSyracuse is in New York.

Compare (9∗) with the following paradigm of Moore paradoxicality:

(1∗) Syracuse is in New York, but I do not believe that Syracuse is in NewYork.

Here is an undefended, but, it seems to me, compelling intuition. Whateverrenders stating or thinking (9∗) unacceptable is the same feature that renders


stating or thinking (1∗) unacceptable. If that is right, we have reason to classify(9∗) with (1∗) as Moore-paradoxical.

The Moore-paradoxicality of (9∗) is not easily accommodated by the repres-entational account. One who thinks (9∗) has a thought that represents Syracuseas being in New York. The problem with applying the representational accountto (9∗) is this. (9∗) does not imply that the same thinker either does not entertaina thought representing Syracuse as being in New York, or entertains a thoughtrepresenting Syracuse as not being in New York. On the contrary, (9∗) onlyimplies that a thinker of (9∗) does entertain a thought, albeit a non-consciousone, representing Syracuse as being in New York.

We might try to extend the representational account to cover (9∗) by invokingthe following argument. One who thinks that (9∗) not only represents Syracuseas being in New York, but also consciously represents Syracuse as being there.Hence, one who thinks that (9∗) consciously represents that Syracuse is in NewYork while denying that she is so representing the location of Syracuse.

This brings us back to the propositional attitude account of Moore-paradoxicality. According to the version of that account we looked at earlier,consciously believing (1∗) is precluded by distribution: belief distributes overconjunction; iteration: believing that P implies believing that one believes P,and the rationality condition: believing that not-P implies not believing that P.Whatever one thinks of the plausibility of the iteration and rationality conditions,they do not exclude believing (9∗). One who believes (9∗) does believe that shebelieves that Syracuse is in New York. So, there is no prospect of deriving fromconsciously believing (9∗) that one who does so fails to believe that she believesSyracuse is in New York.

If we dispense with the rationality condition but retain distribution anditeration, we can at least show that consciously believing (1∗) implies havinginconsistent beliefs⁹. Can we likewise show, by invoking only distribution anditeration, that consciously believing (9∗) implies having inconsistent beliefs? Itall depends on the version of iteration we employ.

There are the following options:

I1: Consciously believing that P implies believing that one believes thatP¹⁰.

I2: Consciously believing that P implies consciously believing that onebelieves that P.

⁹ The argument goes as follows. By distribution consciously believing (1∗) implies believingthat Syracuse is in New York, and believing that one does not believe that Syracuse is in New York.By iteration believing that Syracuse is in New York implies believing that one believes that Syracuseis in New York.

¹⁰ This is the strongest version of iteration that so-called higher order thought (HOT) theoriesof consciousness are typically committed to.

Andre Gallois 171

I3: Consciously believing that P implies believing that one consciouslybelieves that P.

I4: Consciously believing that P implies consciously believing that oneconsciously believes that P.

Combining distribution with I1 fails to result in consciously believing (9∗)implying having inconsistent beliefs. Applying distribution and I1 to (9∗) willonly yield the following. Consciously believing (9∗) implies believing that onebelieves that Syracuse is in New York, and believes that one only non-consciouslybelieves that Syracuse is in New York. Nothing is inconsistent about thatcombination of beliefs.

The same goes for I2. We can, at most, extract from distribution and I2 thatconsciously believing (9∗) implies consciously believing that one believes thatSyracuse is in New York, and consciously believes that one only non-consciouslybelieves that Syracuse is in New York. Again there is no inconsistency of belief.

I3 and I4 can be used to show that consciously believing (9∗) implies havinginconsistent beliefs. The problem with using I3 to explain what is wrong withconsciously believing the Moore-paradoxical (9∗) is more subtle. I3 would servethat purpose if it could be used to show that consciously believing (9∗) implieshaving an irrational belief. It cannot. The problem is this. What we want is anaccount of Moore-paradoxicality that shows why it is problematic to consciouslyendorse (9∗)¹¹. Let us ask: what follows about consciously believing (9∗) fromdistribution and I3? Just this. Consciously believing (9∗) implies having thepossibly non-conscious belief that you consciously believe that Syracuse is inNew York together with the conscious belief that you only non-consciouslybelieve that Syracuse is in New York.

Now let us pose a further question. If distribution and I3 are both true, andyou consciously believe that (9∗), then, if you are fully rational, what else canyou be brought to acknowledge? Since you consciously believe that you non-consciously believe Syracuse is in New York, you can be brought to acknowledgethat you non-consciously believe that Syracuse is in New York. Hence, you canbe brought to acknowledge that you do not consciously believe that Syracuse isin New York.

You consciously believe that Syracuse is in New York. It follows from I3 thatyou, it may be non-consciously, believe that you consciously believe that Syracuse

¹¹ Earlier I distinguished an account of what, for example, it is to have a Moore paradoxicalthought from what, in those cases where it is wrong, is wrong with having a Moore paradoxicalthought. We need to draw such a distinction in the case of even the paradigm of Mooreparadoxicality: P, and I do not believe that P. As we saw there are instances of that schema, such asthe eliminativist’s: there are no beliefs, and I do not believe that there are no beliefs, that do notresult in any defective thought. In contrast, there are, so far as I can see, no non-defective instancesof: P, but I only non-consciously believe that P.


is in New York. Does it follow that, if you are rational, you can be brought toacknowledge that you have that conscious belief ? That would only follow if youconsciously believe that you consciously believe that Syracuse is in New York.Unless, on pain of irrationality, you can be brought to acknowledge that youdo consciously believe that Syracuse is in New York, there is no obvious reasonwhy it would be irrational to have the inconsistent beliefs implicit, according todistribution and I3, in consciously believing (9∗).

What we need for conscious belief in (9∗) to incur not only inconsistentbeliefs, but irrationality is I4. The trouble is that I4 is implausibly strong. Theobvious problem with it is the infinite regress of beliefs it generates.

No doubt a resourceful defender of the propositional attitude account wouldhave responses to the above objections.¹² I will not attempt to anticipate thoseresponses. My disquiet with the propositional attitude account has a differentsource. It is this. Suppose a believer Sam comes to consciously believe:


We who are not prepared to endorse instances of:

(9) P, but I only non-consciously believe that P,

are in a position to recognize that Sam does consciously believe that (9∗). Hence,we are in a position to recognize that Sam does consciously believe that Syracuseis in New York. So, assuming I3, we are in a position to recognize that Sam hasinconsistent beliefs.

How, using the propositional attitude account, might we show Sam thathe has inconsistent beliefs? Obviously we begin by pointing out to Sam that,whether or not he consciously believes that Syracuse is in New York, he at anyrate consciously believes (9∗). Having secured that concession, it should, usingthe propositional attitude account, be an easy matter to persuade Sam that heconsciously believes that Syracuse is in New York.

The difficulty is that if Sam is willing to endorse an instance of (9) such as(9∗), he will have no trouble endorsing the following instance of (9): (9∗), but Ionly non-consciously believe that (9∗).

Perhaps, on pain of irrationality, Sam can be brought to acknowledge thatif he consciously believes (9∗), he has inconsistent beliefs. That is no help inshowing Sam what is wrong with (9∗), unless he can be brought to acknowledgethat he does consciously believe that (9∗). Again, unless we have recourse to I4,it is hard to see how the propositional attitude account can be used to show Samthe error of his ways.

I am suggesting that the propositional attitude account is wanting in thatit cannot be used to show someone like Sam who finds nothing amiss in

¹² For example, we might try weakening I4 to: consciously believing P implies being in a positionto consciously believe that one consciously believes P.

Andre Gallois 173

consciously believing (9∗) that there is something amiss.¹³ In doing so I amimplicitly imposing what may be called a first-person adequacy condition on anaccount of the Moore-paradoxicality of:


According to that condition an account of the Moore-paradoxicality of (9) shouldenable anyone who consciously believes an instance of (9) such as (9∗) to identifywhat has gone wrong.¹⁴ I take it to be a virtue of the account of the Moore-paradoxicality of (9) given in the closing section that it satisfies the first-personadequacy condition. Whether or not that is a virtue, the account will, hopefully,have a further one. Developing it will throw light on the relation between twoof the most significant features of our mental lives: being conscious and havingreasons.

2 . ATTITUDE AND CONTENT-RELATED REASONS

Suppose, you are in the power of a tyrant who controls an efficient team ofmind readers called the thought inspectors. You will be horribly tortured unlessyou come to believe that the tyrant discovered Neptune. If you do not acquirethat manifestly false belief, the thought inspectors will, almost certainly, detectyour failure to do so, and act to your great detriment. Do you have a reason tobelieve the tyrant discovered Neptune? Certainly, in one sense you have everyreason to self-induce that belief. You have every reason to resort to hypnotism,brainwashing, reading bad works, or whatever else it takes to get yourself tobelieve the tyrant discovered Neptune. In another sense you have no reason to

¹³ It is important to distinguish between Sam finding something amiss in consciously believing(9∗), and finding something amiss with consciously believing (9∗). Sam consciously believes (9∗). Asnoted, the propositional attitude account can be used to show Sam what is wrong with consciouslybelieving (9∗). So, we may assume he accepts that there is something wrong with consciouslybelieving (9∗). Since Sam does consciously believe (9∗) he ought, if he is rational, to concedethat there is something amiss. Since it is open to Sam to, at least, remain agnostic about whetherhe consciously believes (9∗), the propositional attitude account is impotent to explain why, inconsciously believing (9∗), Sam ought to find something amiss.

¹⁴ We might ask: gone wrong with what? The difficulty of answering that question is, in my view,symptomatic of something that lies at the heart of Moore-paradoxicality. Someone who consciouslybelieves (9) ought to find something amiss with doing so. Suppose we ask such a believer what isamiss. There seem to be two candidate answers. The first is that (9∗) is true. Clearly, that will notdo. There is nothing, in the relevant sense, amiss with Syracuse being in New York, and the believerfailing to believe that it is. At first sight, more promising is the answer that the believer consciouslybelieves that (9∗). We have already seen, in effect, why that answer will not do. Suppose our believerin (9∗) ought, on pain of irrationality, to find something amiss with consciously believing (9∗), butdoes not. In that case, there should be something we can call Sam’s attention to which will alerther to what has gone wrong. Without strain Sam consciously believes that Syracuse is in New Yorkwithout conceding that she does. So she will find nothing difficult about consciously believing (9∗)without conceding that she does. Hence, we cannot alert Sam to what has gone wrong by pointingout that she consciously believes (9∗)


believe the tyrant discovered Neptune. It is perfectly acceptable for you to reporton your situation by saying something like the following. Even though I have noreason to believe the tyrant discovered Neptune, I have every reason to bring itabout that I hold that belief.¹⁵

In this case there seems to be a straightforward way of marking the relevantdistinction. In the above scenario you have no reason that bears on its truthto acquire the belief that the tyrant discovered Neptune. The only reason youhave for so believing is one that bears on your self-interest. However, that wayof marking the distinction will not generalize to the next case. For a variety ofreasons it is greatly in your interest to join the society of bottle top collectors.Unfortunately, you regard collecting bottle tops as a deeply irrational activity,and the society, which includes among its members some thought inspectorsfrom the previous example, will admit only those who have the desire to performthat activity. You can, it seems, correctly say the following. I have no reason towant to collect bottle tops even though I have every reason to cultivate the desireto do so. As in the previous case, in one sense, you have every reason to have anattitude towards a proposition P which, in another sense, you have no reason tohave towards the same proposition. However, in this case we cannot mark thecontrast as the contrast between having a reason to acquire a desire that bears onself-interest rather than one that bears on the truth of that desire. Desires cannotbe true or false.

Here is a final example, familiar from the literature, illustrating the distinctionbetween attitude and content-related reasons. You have a strong incentive tointend to perform some act even though you have an equally strong incentivenot to carry out your intention. In the best known case, you have every reasonto intend to ingest a toxin that you have every reason not to ingest.¹⁶ Placedin this predicament, you may correctly say the following. I have no reason tointend to take the toxin even though I have every reason to form the intentionto do so.

We have before us a distinction between two senses in which someone canhave reason to adopt a propositional attitude. So far, I have attempted to illustraterather than explain the distinction. When it comes to explaining the distinctionbetween attitude and content-related reasons, there are different ways of doingso. One has already been implicitly introduced. A content-related reason forbelieving, wanting, or intending that P is a reason for believing, wanting, orintending that P is so. An attitude-related reason is a reason for bringing it aboutthat one is in a certain attitudinal state. An attitude-related reason for believing,

¹⁵ Pascal’s Wager arguably illustrates the distinction being drawn here. Pascal’s Wager may give areason to get oneself to believe in the existence of God. It gives no reason to believe in the existenceof God.

¹⁶ This example is due to G.Kavka, ‘The Toxin Puzzle’, Analysis, 43/(1983), 33–6.

Andre Gallois 175

wanting, or intending that P is a reason for bringing it about that one believes,wants, or intends that P.

Here is a second way of articulating the distinction between attitude andcontent-related reasons. Consider belief and desire. A content-related reason forbelieving or desiring that P counts in favor of P being believable or desirable.An attitude-related reason for believing or desiring that P contributes not at allto P being believable or desirable. In the above examples the attitude-relatedreasons for believing the tyrant discovered Neptune and wanting to collect bottletops contribute not at all to the believability of the tyrant having discoveredNeptune, or the desirability of bottle top collecting. Generalizing, in contrastto an attitude-related reason, a content-related reason for believing, desiring, orintending that P indicates that P merits belief, desire, or intention.¹⁷

I have a preference for a third way of drawing the distinction.¹⁸ So far thedistinction between attitude and content-related reasons has been presented as adistinction between two kinds of reason. That is a natural, but not inevitable, wayof taking it. Consider the case of believing that the tyrant discovered Neptune.We could say the individual in that case has one kind of reason, an attitude-relatedone, for believing the tyrant discovered Neptune, but not another kind, a contentrelated one, for holding that belief. Alternatively, rather than acknowledging twokinds of reason, we could say this. She has no reason at all for believing that thetyrant discovered Neptune. Instead she has a reason for wanting to believe thatthe tyrant discovered Neptune.

If we construe the distinction between content and attitude-related reasonsin this way, it transmutes into a distinction between reasons for having higheras opposed to lower-order attitudes. What looks like a different kind of reasonfor having, say, a first-order attitude is really the only kind of reason there is forhaving a different kind of second-order attitude.

When the time comes to make use of the content-attitude related distinctionin an account of Moore-paradoxicality I will presuppose this last way of takingthat distinction.

3 . CONSCIOUSNESS AND REASONS

Consider Fred who only non-consciously believes that he is unmusical. Supposethat, despite having that unconscious belief, Fred should pursue a musical

¹⁷ I owe this way of putting the distinction to Eve Garrard.¹⁸ This way of drawing the distinction is adopted by Derek Parfit in ‘Rationality and Reasons’

in Dan Egonsson, Bjorn Peterson, Jonas Josefsson and Toni Ronnow-Rasmussen (eds.), ExploringPractical Philosophy: From Action to Values (Aldershot: Ashgate, 2001): 17–41, and by Alan GibbardWise Choices, Apt Feelings: A Theory of Normative Judgment (Cambridge, Mass.: Harvard UniversityPress, 1990). 37


career, and consciously believes that he should. Suspecting that there is anunconscious impediment to his effectively pursuing his chosen career, Fred visitsa psychoanalyst. As a result he is persuaded that he does have an unconsciousbelief in his unmusicality. Fred comes to consciously believe that he has theunconscious belief that he is unmusical.

Here is an asymmetry between Fred’s unconscious belief and his consciousbelief that he ought to be a musician. Without it becoming conscious, Fred isable to consciously self-attribute his belief that he is unmusical. Since that is so,Fred can take his unconsciously believing himself to be unmusical as a reason forbelieving, wanting, or acting. For example, Fred may realize that his unconsciousbelief that he is unmusical inhibits him from properly exploiting his musicaltalent. In that case, Fred unconsciously believing that he is unmusical may givehim a reason to persist with the therapy revealing he has that belief.

Now consider Fred’s consciously held belief that he should become a musician.As we have seen, Fred may take the fact that he unconsciously believes himselfto be unmusical to be a reason to attempt to divest himself of that belief. Fredmay, likewise, take the fact that he believes he should become a musician as areason for acting. He may, for example, take his believing that he should becomea musician as a reason for not discussing his future career with his parents whovery much want him to become a lawyer.

So far, no asymmetry has emerged between the way in which a consciously,as opposed to a non-consciously, held belief can feature in a reason for actingor adopting a propositional attitude. An asymmetry emerges when we considerthe following. Fred is able to take the fact that he believes he should become amusician to be a reason for acting. Since he holds that belief consciously, he isalso able to take the fact that he should become a musician as a reason for acting.Fred is also able to take the fact that he believes he is unmusical as a reason foracting. Since he holds that belief only unconsciously, he is not able to take thefact that that he is unmusical as a reason for acting.

Consciously holding the belief that P enables one to treat the fact that P,rather than just the fact that one believes P, as a reason for acting, or adoptingfurther propositional attitudes. Only non-consciously holding the belief that Pdisables one from taking that P as a reason. What explains this asymmetry? Onecandidate explanation can be immediately discounted. Let us say that if onebelieves that P, and is able to take the fact that P as a reason, then the belief thatP is reason-giving. Suppose we grant that a belief is reason-giving only if one isaware of holding it. It may be suggested that a non-conscious belief cannot bereason-giving because the one who holds it is not aware of doing so.

We need to draw a distinction between being conscious of holding a belief,and that belief being consciously held. In the case described above, Fred, as aresult of visiting his psychoanalyst, is fully conscious of holding the belief that heis unmusical. Despite that, the belief he is fully conscious of holding is not one

Andre Gallois 177

that he consciously holds. So the question remains to be answered: why is it thatonly a consciously held belief can be reason giving?

Here is the principle we are currently examining:

C: A believer is in a position to take that P as a reason for acting, orconsciously holding a propositional attitude such as believing, wanting,hoping intending etc, only if she consciously believes that P.

The question before us is this. What explains the truth of C? One answer to thatquestion has proved unsatisfactory. C is not true because only a belief that anindividual is aware of holding can be reason-giving. An initially more promisinganswer goes like this. Take Fred’s unconscious belief that he is unmusical. He has,we are supposing, every reason to believe that he holds that belief. Nevertheless,he has no reason to believe that he is unmusical. Only a belief held for a reasoncan, itself, be reason-giving. Moreover unconscious beliefs are, typically, not heldfor reasons. So, non-conscious beliefs are, typically, not reason-giving. That iswhy C is true.

At most this shows that an unconscious belief, even if consciously self-attributed, is unlikely to be reason-giving. It fails to show that such a beliefcannot be reason-giving.

We are still left with the question: why cannot a non-consciously held beliefbe reason-giving? In brief the answer is that only a consciously held belief canengage in the right way with other consciously held propositional attitudes, suchas desires, to deliver a reason for adopting those attitudes.

To see how only consciously held belief can engage with a consciously helddesire to yield a reason consider a type of case familiar from the literature onbelief and the will. Sam has a very strong desire to believe that God exists. Ashe sees it he has compelling truth-related reasons to believe there is no God. Asa result he becomes severely depressed. So depressed that he has an overridingnon-truth-related reason to reacquire his former consciously held belief in theexistence of God.

Prior to reacquiring that belief, Sam reasons thus. What I most want is thatGod exists. There is nothing I can do to satisfy that desire. In particular, mycoming, once again, to consciously believe in God’s existence will make nodifference to his existing. Still, given that I am unable to modify my desire forGod to exist, if I do bring myself to consciously believe that God exists, I will nolonger be depressed.

Sam has a desire he recognizes he cannot satisfy: the desire that God exists. Hav-ing that desire, together with the depression resulting from its non-satisfaction,leads him to form a desire he can satisfy: the desire that he consciously believe thatGod exists. Suppose that, by adopting the relevant Pascalian procedures, Sambrings himself, for what he takes to be good truth-related reasons, to consciouslybelieve that God exists. At that point Sam is asked whether he feels depressedbecause he lives in a universe without God. From the point of view of alleviating


his depression what matters is that, from his perspective, Sam is in a positionto give the following answer. I don’t feel in the least depressed about living in auniverse without God because I don’t live in such a universe: God exists.

Before he succeeds in repossessing his belief in God’s existence, Sam takes hisdesire for God to exist to be unsatisfiable, but sees that he is able to satisfy adistinct desire to consciously believe that God exists. After repossessing his beliefin God’s existence, Sam not only takes his desire to believe that God exists to besatisfied, but, in addition takes his desire for God to exist to be satisfied. Were henot to take the second desire to be satisfied, he would still feel depressed.

What the case of Sam illustrates is the distinctive impact on one’s psychologyof having a belief consciously. Consciously having the belief that P enables one tointegrate P into one’s view of the world in such a way that it being the case thatP can have a distinctive impact on one’s mental life. We may label that impactthe integrative function of consciously holding a belief. In summary, in virtue ofa belief being consciously held it is possible for what is believed to be integratedinto the believer’s world view in a way that what is believed in the case of annon-consciously held belief cannot be.

In the remainder of the paper I will attempt to do the following. First tospell out in three interconnected ways the integrative function of holding a beliefconsciously. Second, to say how conscious belief having that function helps toexplain the Moore-paradoxicality of: P, but I only non-consciously believe thatP. Third, to relate the foregoing to the distinction drawn earlier between attitudeand content-related reasons.

4 . CONSCIOUS BELIEF: DELETION AND INTEGRATION

Consider the following inferences:

(A) Necessarily: there is only one even prime.

There is only one even prime.

(B) It is true that: snow is white.

Snow is white.

(C) John intends that: he John will leave soon.

John will leave soon.

Each of (A)–(C) have the following in common. Each one is an inferenceconsisting of a premiss which supports the corresponding conclusion. In the caseof (A) and (B) the premiss entails the conclusion. In the case of (C) the premissdoes not entail, but makes probable the conclusion. We may say that in the caseof each of (A)–(C) the truth of its premiss supports the truth of its conclusion.

Here are some further features shared in common by (A)–(C) that I wish tocall to attention. Ceteris paribus, belief in the premiss of each of (A)–(C) makes

Andre Gallois 179

rational belief in its conclusion. We may say that, in the case of each of (A)–(C),the transition from believing its premiss to believing its conclusion is a defeasiblyrational one.

Another feature shared in common by (A)–(C) plays a prominent role inthe following discussion. Consider (A). Suppose an individual makes what Icalled the rational transition from believing (A)’s premiss, necessarily: thereis an only even prime, to believing its conclusion: there is an only evenprime. Making such a transition amounts to deleting the operator expressed by‘necessarily’ from that premiss. An alternative way to put the point is to saythat making the relevant transition amounts to taking a proposition that fallsunder the scope of the necessity operator in the premiss outside the scope of thatoperator.

The same point applies to (B) and (C). Making the defeasibly rationaltransition from a belief in the premiss of each one to a belief in its conclusionresults in deleting the operator featured in that premiss.

For our purposes calling attention to the features shared in common by(A)–(C) has the following relevance. Compare (A)–(C) with:

(D) I consciously believe that: G. E. Moore smoked a pipe.

G. E. Moore smoked a pipe.

(D) shares the following in common with (A)–(C). It is defeasibly rationalfor anyone who consciously endorses (D)’s premiss to make the transition toendorsing its conclusion. If you accept, as you might put it, that you believein full consciousness that G. E. Moore smoked a pipe then, in the absence ofcountervailing considerations, it is rational for you to believe that G. E. Mooredid indeed smoke a pipe.

The next feature (D) shares in common with (A)–(C) is the crucial one.Making the transition from believing (D)’s premise to believing its conclusionamounts to deleting the associated operator. In this case it amounts to deletingthe operator expressed by ‘I consciously believe that’. Again we may put thepoint in terms of scope. Moving in thought from (D)’s premise to its conclusionplaces the proposition that G. E. Moore smoked a pipe outside the scope of theoperator expressed by ‘I consciously believe that’.

Before examining the significance of this point about deletion or scope, weneed to note a difference between (D) and each of (A)–(C). I called the movefrom a belief in any of (A)–(C)’s premisses to a belief in its conclusion adefeasibly rational transition. Why call that move a defeasibly rational transitionrather that something more familiar such as a warranted or justified inference?The reason is this. Belief in the premise of any of (A)–(C) does warrant or justifybelief in its conclusion. One indication of this is the following. It is rational forone who believes (A)’s premiss to believe its conclusion. In the case of (A) it isalso rational to believe the conditional whose antecedent is (A)’s premiss, andwhose consequent is (A)’s conclusion. It is rational to believe:


(A∗) If it is necessary that there is an only even prime then there is an onlyeven prime.

Moreover, unless it is rational to believe (A∗) it is not rational to move from abelief in (A)’s premiss to a belief in its conclusion.

The same goes for (B) and (C). The rationality of moving from a belief in (B)or (C)’s premiss to its conclusion makes rational, in the case of (B), belief in theconditional:

(B∗) If it is true that snow is white then snow is white,

and, in the case of (C), the conditional:

(C∗) If John intends that he will leave soon then John will leave soon.

As with (A) and (A∗) the rationality of belief in (B∗) and (C∗) is required for thecorresponding belief transitions to be rational.

In this respect (D) contrasts with (A)–(C). Here once again is:

(D) I consciously believe that: G.E. Moore smoked a pipe.

G. E. Moore smoked a pipe.

It is rational for me to move from a belief in (D)’s premiss to a belief in itsconclusion even if it is not rational for me to believe:

(D∗) If I consciously believe that: G. E. Moore smoked a pipe then G. E.Moore smoked a pipe.¹⁹

We might say that the move from a belief in (D)’s premiss to a belief in itsconclusion is inference-like. Despite that, and this cannot be overemphasized,that move is not a rational one to make because it is a justified or warrantedinference.

Here is another way to make the same point that, as we will see, links therationality of the belief transition represented by (D) to the Moore paradoxicalityof: P, but I only non-consciously believe that P. Each of (A)–(C) is an inferencethat it is rational to make. As we have observed, each one involves deleting anoperator from the proposition constituting its premiss. Clearly, in the case of(A)–(C), in order for that deletion to be rationally made the relevant premissmust, at least, be believed. For example, the truth of (A)’s premiss is by itselfinsufficient to make the deletion of the operator expressed by ‘it is necessary that’rational. In contrast, it is enough for (D)’s premiss to be true for the deletion ofthe operator expressed by ‘I consciously believe that’ to be rationally made.

What makes the transition from consciously self-attributing a consciously heldbelief to the belief consciously self-attributed a rational one? In brief the answeris this. The deletion function of consciously held belief enables a believer tointegrate reasons and what they are reasons for into a unified view of the world.

¹⁹ Of course, it may be rational for me to believe D∗. It would be so if I had reason to believemyself authoritative about G. E. Moore’s smoking habits. The point is it need not be rational forme to believe D∗ in order for it to be rational for me to move from D’s premiss to its conclusion.

Andre Gallois 181

My next task is to clarify what kind of integration is at issue. I will attempt to doso in two ways. First, by invoking a comparison I have used elsewhere with therole of integration in fiction.²⁰ Second, and more importantly, by reintroducingthe distinction drawn earlier between two ways of attributing a reason.

5 . INTEGRATION IN FICTION

Part of a fictional narrative may consist in some character within the story relatinga fictional narrative: a story within a story. Such a story may unfold like this.The original storyteller, it may be an actual author, tells a fictional tale in whicha group of individuals gather to have a beer together, and one of them tells astory. Call the story told within the original story the embedded story. Supposethe embedded story begins like this. A traveler journeyed to a distant land, andthere met an old man who told her fortune.

So long as it remains within the original story, as they say, just a story, theembedded story fails to be integrated with the original. There are a number ofways to signal this failure of integration. The most obvious is this. Suppose weare asked to specify what would have to be the case for the original story to betrue. We would need to mention that a group of individuals gathered together tohave a beer. We would not need to mention that a traveler journeyed to a distantland. What we would need to mention is that one of those having a beer told astory about such a traveler.

Another indicator of the relevant failure of integration is this. Events in theoriginal and embedded stories do not bear on one another in a way that theydo in a pair of stories that, to anticipate, present an integrated fictional world.One of the protagonists in Charlotte Brontë’s Jane Eyre is Rochester’s wife. JeanRhys’s Wide Sargasso Sea is what these days would be called a prequel to Jane Eyre.In the former novel, Rochester’s wife’s early life in the West Indies is narrated inpart to explain the way she reacts in Bronte’s narrative.

Despite being by different authors, the events in The Wide Sargasso Sea dobear on the events in Jane Eyre in a way that the events in the embedded andoriginal stories do not bear on each other. An example of this is given above.Events narrated in The Wide Sargasso Sea may be invoked to explain eventsnarrated in Jane Eyre, not just by the readers of those works, but also by theirparticipants. For example, though it did not, reflecting on his wife’s earlier lifeas depicted in Jean Rhys’s novel could have given Rochester explanatory insightinto his wife’s behaviour as the events in the embedded story could not provideexplanatory insight into the behaviour of the individuals listening to it. Again,the events in Jean Rhys’s novel can provide the protagonists in Jane Eyre with

²⁰ In Gallois, ‘First-Person Accessibility and Consciousness’, Philosophical Topics, 28 (2000)/2101–24.


reasons for, say, acting or believing. Events in the embedded story cannot providethe protagonists in the original story with like reasons to act or believe.

The next indication of integration failure relates directly to what was said inthe last section about conscious belief. In the narration of the original story useis made of an operator: one, as it may be, expressed by ‘one of the individualsdrinking beer began to tell the following story’. Call an operator of this type thestory operator. Events in the embedded story fail to be integrated with those inthe original because the former events fall within the scope of the story operator.Taking those events outside the scope of that operator has the effect of integratingthem into the original story. For example, if one of those in the original story,without entering any qualification, remarks that the traveler in the embeddedstory has just entered the room, that has the effect of making that traveler part ofthe same fictional world as the world of the original story.

In the case of a fictional narrative we have identified signs or marks ofintegration. A fictional work F1 fails to be integrated with a fictional work F2 ifwhat would have to be the case for F1 to be literally true need not be mentionedin specifying what would have to be the case for F2 to be true. F1 fails to beintegrated with F2 if what is the case according to F1 does not bear, in anexplanatory or reason-giving fashion, on what is the case for the participants inF2. F1 fails to be integrated with F2 if what is the case according to F1 cannot beplaced outside the scope of the story operator in F2. The point of the precedingremarks about integration is to construct a, hopefully illuminating, parallel withthe sense in which consciously held beliefs are integrated. Of the three marks ofintegration the last two apply to conscious belief. Consider the case mentionedat the beginning of Section 3. Fred consciously believes he should become amusician, but only unconsciously believes he is unmusical. From Fred’s point ofview, that he should become a musician bears on what, for example, he has reasonto do or think. In contrast, that he is unmusical can have no bearing on what heshould do or think, unless his belief that he is unmusical becomes conscious.²¹

The third mark of integration concerns scope. If pair of fictional works areintegrated then what is the case according to one will fall outside the scope ofthe story operator in the other. Fred is, likewise, in a position to place that heshould become a musician outside the scope of a belief operator as he is not ableto place his being unmusical outside the scope of that operator.

6 . REASONS, CONSCIOUSNESS, AND INTEGRATION

It is now time to return to the distinction between attitude and content relatedreasons. Consider the case I used to introduce the distinction in Section 2. The

²¹ Again, Fred can take that he believes he is unmusical as a reason for acting, but his belief thathe believes he is unmusical is one he consciously holds.

Andre Gallois 183

tyrant will torture you unless you believe he discovered Neptune. In that case,you have an attitude- but no content-related reason to believe he discoveredNeptune. Now consider a belief which you do have a content-related reason forholding. You have a content-related reason for believing that the tyrant did notdiscover Neptune. One content-related reason you have for holding that beliefis, say, that the tyrant was not born when Neptune was discovered.

Here is a striking feature of a content-related reason which differentiates itfrom its attitude-related counterpart. The belief for which you hold a contentrelated reason is the belief that:

(i) The tyrant did not discover Neptune.

One content related-reason you have for believing (i) is:

(ii) The tyrant was not born when Neptune was discovered.

In addition, you have an attitude-related reason for believing:

(iii) The tyrant did discover Neptune,

which is:

(iv) You will be horribly tortured unless you adopt the belief that (iii).

(ii) specifies the content of a belief that gives you a reason for believing (i). (iv)likewise specifies the content of a belief that gives you a reason for believing(iii). One of these reasons is content-related, and the other only attitude-related.What makes for this difference?

Consider the pair of beliefs at issue in what we may call the tyrantcase:

(Bi) The belief that the tyrant did not discover Neptune.

(Biii) The belief that the tyrant did discover Neptune.

Your reason for adopting (Bi) is that:

(ii) The tyrant was not born when Neptune was discovered.

and for adopting (Biii) that:

(iv) You will be horribly tortured, unless you adopt (Biii).

Intuitively there is a relation between (ii) and (Bi) which does not hold between(iv) and (Biii). Making use of the terminology used in the last section tocharacterize one aspect of integration in fiction, (ii) bears on what (Bi) is aboutin a way that (iv) does not bear on what (Biii) is about. We may say that (ii)bears, in a way (iv) does not, on a question that (i) and (iii) supply competinganswers to. (ii), unlike (iv), bears on whether the tyrant did discover Neptune.

How should the distinctive connection between a content-related reason andthe content of the belief it rationalizes be understood? One answer is clearlyinadequate. It goes like this. (ii) is, we are supposing, your reason for having:

(Bi) The belief that the tyrant did not discover Neptune,


and, we may also suppose, you would have no other reason to hold (Bi) if you hadnot believed (ii). If so, your having a belief with (Bi)’s content is counterfactuallydependent on your believing (ii). We may be inclined to say that the distinctiverelation that holds between a reason Q and the belief that P is content-related ifone would not have had a belief with that content, unless one had believed Q.²²

Of course, the trouble with this suggestion is that you would not have had(Biii) if you had not believed (iv). There is no obvious relation of counterfactualdependence that distinctively holds between a content-related reason and thebelief it is a reason for having.

What then is the best way to understand the link between a content-relatedreason and the belief it rationalizes? The answer I favor is best given against twobackground assumptions whose defence must be left to another time and place.Consider the following pair of claims:

(1) That it has rained recently is a reason for it being true that the groundis wet.

(2) That Jones’s light is on is a (content-related) reason for Sam believingthat Jones is at home.

The first assumption is that the expression ‘is a reason for’ in sentences (1) and(2) attributes the same relation that holds between one thing being the case,or being true, and something else being the case or being true. The secondassumption is the crucial one. In the case of (1) the reason-giving relation holdsbetween it having rained recently and the ground being wet. In the case of (2) thesame reason-giving relation does not hold between Jones’s light being on andSam believing that Jones is at home. Instead, it holds between Jones’s light beingon and Jones being at home.

Here is an all too brief sketch of how one might employ the machinery ofpossible worlds to fill in this basic picture. In the case of (1) the reason-givingrelation holds between something true at the actual world, it raining recently,and something else, the ground being wet, also true at the actual world. Inthe case of (2) the same reason-giving relation holds between something true atthe actual world, Jones’s light being on, and something, Jones being at home,holding true at the members of some set of worlds. But the worlds in this caseare Sam’s belief-worlds. We may put it like this. If (1) is true, it having recentlyrained is a reason for it being true in the actual world that the ground is wet. If(2) is true, Jones’s light being on is a reason for it being true in the world as Sambelieves it to be that Jones is at home.

Content-related reasons for other propositional attitudes behave in the sameway. Sam, let us say, has the following content-related reason for going for a

²² An obvious objection to this proposal arises from the possibility of overdetermination ofcontent-related reasons. You have a sufficient content-related reason, (ii), for (Bi). Suppose, youhave an additional content-related reason which is also sufficient for (Bi). In that case, it may befalse that had you not believed (ii), you would not have had (Bi).

Andre Gallois 185

walk: it is a fine day. In that case, it being a fine day is a reason for it being truein the world as Sam wants it to be that he goes for a walk. Seen in this light inthe tyrant case we say you have the attitude-related reason:

(iv) You will be horribly tortured, unless you adopt (Biii)

for adopting:

(Biii) The belief that the tyrant did discover Neptune

because (iv) being true is a reason for it being true in the world as you want itto be that you believe the tyrant discovered Neptune. (iv) is not a reason for itbeing true in the world as you either want or believe it to be that the tyrant diddiscover Neptune. That is why (iv) provides an attitude- but no content-relatedreason for (Biii).

7 . CONSCIOUSNESS, REASONS, AND MOORE’S

PARADOX

We have covered a good deal of territory since taking note of the Moore-paradoxicality of instances of:


The representational account of Moore-paradoxicality was found wanting in thatit provides no explanation of the Moore-paradoxicality of (9). My aim is toprovide, at least, the beginnings of such an account by utilizing as its componentswhat has been covered in the previous five sections. It is now time to review thosecomponents, and say how they bear on the Moore-paradoxicality of (9).

In Section 2 we looked at two ways, attitude and content-related, in whicha reason can be a reason for believing. In the last section we noted that, as solabeling them suggests, content-related reasons have a distinctive connection tothe contents of the beliefs they rationalize. In that section an attempt was madeto articulate that connection. The relation of being a reason for holds betweencontents. In the case we are concerned with it holds between belief-contents.If Q is a content-related reason for having the belief that P, that is a matterof Q being true being a reason for P being true. Why not then simply say thatQ being true is a reason for P being true without making reference to belief ?Because the reference to belief indicated the standpoint from which Q being trueis a reason for P being true. Q being true is a reason for P being true from thestandpoint of belief rather than desire, hope, fear, intention, or, for that matter,actuality.

Section 3 explored the relation between consciousness and reasons. Thereit was claimed that only a consciously held belief can supply a reason. Thatobservation is linked to the topic of Section 4: the deletion and integrationfunctions of consciously held belief. One may consciously self-attribute both


consciously and non-consciously held beliefs. A striking difference between aconsciously and non-consciously held belief is this. In the case where the beliefthat P is consciously held one is typically able to delete the qualification thatP is something one believes. Such self-effacement is intimately connected toa function of conscious belief I called its integration function. The contentsof consciously held beliefs can be integrated into a single view of the world.If one consciously believes that P and consciously believes Q then one candelete the qualification that P and Q are believed. As a result the contents ofthose beliefs are integrated into a view of the world in which both P and Qobtain.

Now for the story about the Moore-paradoxicality of (9). To see how thatstory goes, consider the instance of (9) introduced in the first section:


Suppose Sam comes to consciously hold (9∗). The representational account tellsus that the following are consequences of Sam consciously believing (9∗):

(3) Sam consciously believes that: Syracuse is in New York,

and:

(4) Sam consciously believes that: he only non-consciously believes thatSyracuse is in New York.

The task is to say why Sam holding the conscious beliefs attributed by (3) and(4) should be problematic for Sam given that (3) does not imply that Samconsciously believes (3), but, at most, that he believes (3).

According to (4) Sam attributes to himself the belief that Syracuse is in NewYork as one that he non-consciously holds. Since that is so Sam will not beprepared to move from:

(5) I (Sam) believe that: Syracuse is in New York,

to:

(6) Syracuse is in New York.

But, Sam does consciously believe (6). So, from Sam’s point of view, (6) fallsoutside the scope of the belief operator in (5). Despite that, since he is onlyprepared to self-attribute his belief in (6) as a non-conscious one, Sam is notprepared to move from (5) to (6). He is thus not prepared to delete the beliefoperator in (5). Hence, in virtue of consciously believing (5) and consciouslyself-attributing a non-conscious belief in (6), Sam has conflicting dispositions.He is disposed to delete the belief operator in (5). He is also disposed not todelete the belief operator in (5).

Can we leave it at that? Can we explain the Moore-Paradoxicality of (9) bysimply noting that, in virtue of consciously endorsing (9∗), Sam will have theconflicting dispositions in question? To do so would be unsatisfactory since the

Andre Gallois 187

resulting account of (9)’s Moore-paradoxicality would fail to meet the first-personconstraint mentioned at the end of Section 1. Simply observing that consciouslybelieving (9∗) induces conflicting dispositions in the believer leaves out what,from the believer’s first-person perspective, generates that conflict.

We can do better by taking into account the integration function of consciouslyheld belief, and its connection to content-related reasons. In consciously believing(9∗) our believer self-attributes the belief that Syracuse is in New York as oneonly non-consciously held. As so attributed the believer cannot place Syracusebeing in New York outside the scope of the belief operator. Since the believercannot take Syracuse being in New York outside the scope of that operator, shecannot integrate Syracuse being in New York into her antecedent view of theworld. In particular, she cannot take Syracuse being in New York as a contentrelated reason for acting or having further propositional attitudes, or take anycontent-related reason as a reason for Syracuse being in New York from thestandpoint of belief.

Our believer does consciously believe (9∗). So Syracuse being in New York isintegrated into her overall view of the world. She is in a position to take Syracusebeing in New York as a reason, and take content-related reasons as reasons forSyracuse being in New York, again from the standpoint of belief.

In this way, consciously believing (9∗) generates an unstable view of the worldbelieved in. In virtue of consciously believing (9∗), our believer consciouslybelieves that Syracuse is in New York. Since that is so Syracuse being in NewYork is integrated into the world as the believer takes it to be. Our believer alsoconsciously believes that she only non-consciously believes that Syracuse is inNew York. Since that is so Syracuse being in New York fails to be integrated intothe world as the believer believes it to be.

I will conclude by, once again, using the comparison with integration in astory to throw light on what is going on here. In that case the counterpart toconsciously believing (9∗) would be this. Suppose we begin reading the originalstory. In doing so we are presented with a fictional world. That fictional worldincludes a number of individuals having a conversation while drinking beer ina pub. Suppose one of those individuals announces that David Lewis’s realismabout possible worlds has been conclusively established. If so, the existence ofsuch worlds spatio-temporally disconnected from the world of the original storyhas been integrated with that world. So far no failure of integration. Some storiesare inconsistent. Suppose the original story turns out to be inconsistent. Supposethe author of the original story implies that one of the beer drinkers is a teetotaler.Still no failure of integration. The fictional world depicted is one in which somebeer drinker is a teetotaler. At this point, one of the characters begins relatingthe embedded story. Now there is a failure of integration. The embedded storyintroduces into the original story a fictional world that is not integrated into theworld of the original story.


The world-embedded story may in various ways become integrated into theworld of the original story. The counterpart of consciously believing (9∗) wouldarise if a reader of the original story were invited to both integrate and notintegrate the world of the embedded story into the world of the original. Such astory would not, as an inconsistent one is, be a story about the impossible, butan impossible story. In like fashion, consciously believing (9∗) leads a believer totell an impossible story about the actual world.

9

Moorean Absurdity and ShowingWhat’s Within

Mitchell Green

1. MOOREAN ABSURDITY AND MOORE’S PARADOX

Moorean absurdity occurs in any utterance or thought in which an agent overtlyexpresses an intentional state that she also explicitly disavows; it also occurs inany utterance or thought in which an agent overtly expresses an intentional statewhose content is incompatible with that of another intentional state that she alsoexplicitly avows. Her expression-and-disavowal, or expression-and-avowal, mightoccur in her utterance or thinking of a single sentence such as ‘It’s raining but Idon’t believe it.’ On the other hand this expression-and-disavowal, or expression-and-avowal, might occur without the medium of a single sentence. It mightoccur through the utterance or thinking of two consecutive sentences such as

It’s raining. I don’t believe that it is.

Alternatively, for all we know about what a conscious mind can do, it mightoccur through the simultaneous thinking or uttering of the above two displayedsentences.

The above disjunctive characterization of Moorean absurdity (as either overtlyexpressing an intentional state that one also explicitly disavows, or overtlyexpressing an intentional state whose content is incompatible with that ofanother intentional state that one also explicitly avows) is not a grammaticalcharacterization. Someone who utters a sequence of sentences such as those

Earlier versions of this paper were presented at the University of Virginia and at Texas A&MUniversity. I thank audiences at both institutions for their insightful comments. Special thanks toJohn Williams and a referee for Oxford University Press for their illuminating comments on anearlier draft. Research for this paper was supported in part by a Summer Grant from the ViceProvost for Research and Public Service at the University of Virginia. That support is here gratefullyacknowledged.

190 Showing What’s Within

displayed above may simply be using her second sentence to express surprise oramazement. Doing so is not at all absurd. Nor is it absurd if the above sequenceregisters a mid-utterance change of mind. Likewise, as I recover from laryngitisI might test my voice by uttering the above sequence. In this case I don’t meana word of what I say, and nothing in my performance is absurd. An actor whoutters ‘It’s raining but I don’t believe it’ on stage in the course of a play about, say,G. E. Moore, is also not behaving absurdly; nor is he harboring absurd beliefs.This is due to the fact that so long as he is not a method actor, he is not expressingany of his beliefs. Instead, he is portraying a character expressing and avowingbeliefs, and thereby portraying a character who is behaving absurdly. That iswhy the non-absurdity of the actor’s performance is compatible with the abovecharacterization of Moorean absurdity.¹ Similarly, in Wittgenstein’s example thetrain announcer who says over the loudspeaker, ‘Train number 121 arriving ontrack #3 in two minutes (personally I don’t believe it),’ is not behaving absurdly.The reason is that he is only saying his first conjunct ex cathedra in the hopesof keeping his job, rather than putting it forth as an expression of his state ofmind. That is why the non-absurdity of his announcement is compatible withthe above characterization of Moorean absurdity.

My account of Moorean absurdity invoked the notion of an overt expression.It may be possible to express an intentional state without doing so overtly.Perhaps when I involuntarily grimace as the paring knife slices into my finger Iam expressing pain. I am not, however, overtly doing so. Accordingly, if whilethus grimacing as the knife slices flesh I also say, ‘. . . but I’m not in pain,’I am in error but am not absurd. Likewise, my blushing might express myembarrassment even though I cannot help it. If while blushing I also explicitlydeny that I am embarrassed, I am simply wrong rather than absurd. Contrary towhat these examples suggest, it is not enough for overtness that my expression beintentional. We will consider what more is required in Section 3 below.

Whereas Moorean absurdity is a phenomenon for which we theorists need noapology, Moore’s Paradox is a testament to the ignorance of students of languageand thought. For Moore’s Paradox is the apparent tension that arises from thefacts that (1) such utterances and thoughts as ‘It is raining but I don’t believe it,’seem absurd, but (2) their absurdity is not due to a semantic contradiction.

An assertion of an explicit contradiction P&-P is absurd, but with the helpof standard logical techniques we may explain its absurdity without landing inparadox. Likewise, a self-falsifying utterance such as an utterance of ‘I am not nowuttering any words’ is absurd, but the absurdity is not paradoxical because thereis no mystery how the utterance of that sentence falsifies itself: We explain thephenomenon with standard logic plus a gentle reference to pragmatics. A slightly

¹ It also shows why Szabo-Gendler (2001) is mistaken to contend that the actor’s performance isabsurd. Szabo-Gendler is confusing absurdity in the actor with absurdity in the character portrayed.This mistake undermines the gravamen of his criticism of modal fictionalism.

Mitchell Green 191

fuller reference to pragmatic notions such as is available from double-indexingsemantics shows what is absurd about an utterance of ‘I am not here now’.With minimal assumptions about the structure of contexts of utterance plusthe behavior of indexicals ‘I’, ‘here’, and ‘now’, we may infer that this sentenceexpresses a contingent proposition even though in every context in which it is(or could be) thought or uttered it is not true. Once again we account for theabsurdity without landing in paradox.

Matters are not so simple with cases of Moorean absurdity. ‘It is raining butI don’t believe it,’ and ‘It is raining but I believe it is not,’ not only expresspropositions that could be true. They are also sentences that could be trulyuttered or truly thought.² If I utter ‘It is raining but I don’t believe it,’ what Iutter could be true so long as I am not sincere. If I say to myself without utteringany words, ‘It is raining but I don’t believe it,’ what I say to myself could be trueso long as I am not sincere. (Below I show that it is possible to say something tooneself in the privacy of one’s own thoughts that one does not also believe.)³ Welabel rather than explain the phenomenon with such expressions as ‘pragmaticparadox’ or ‘pragmatic absurdity’. In what, then, might we locate the source ofMoorean absurdity?

My aim in what follows will be to resolve Moore’s Paradox by explaining thesource of Moorean absurdity. With apologies to existentialists, the first step shallbe to elaborate on the notion of absurdity.

2 . ABSURDITY AND THE VIOLATION OF NORMS

Absurdity arises from severe violation of a system of norms.⁴ Some well-knownsystems of norms are theoretical rationality and practical rationality. We find asevere violation of theoretical rationality in any agent whose system of beliefs isguaranteed to put her in error no matter how the world happens to be, and ina way that she could in principle discern with no empirical investigation. Jane’sbelief that Hesperus is shining puts her in error if in fact Venus is not shining, butit is not absurd simply to be in error. So too, Jane’s belief that Hesperus is shining

² Cargile (1967) is to my knowledge the first to observe this.³ Kriegel (2004) asserts that although Moorean sentences can be true, they cannot be truly

asserted and they cannot be truly believed. As we shall see below, only the latter claim is correct.Kriegel’s view that a Moorean sentence cannot be truly asserted seems to rest on his view that anutterance is an assertion only if it is an expression of a belief, that is, only if it is sincere. This viewsimply stipulates away the possibility of a lie, and will not be assumed here.

⁴ ‘In ordinary life, a situation is absurd when it includes a conspicuous discrepancy betweenpretension and aspiration or reality: someone gives a complicated speech in support of a motion thathas already been passed; a notorious criminal is made president of a major philanthropic foundation;you declare your love over the telephone to a recorded announcement; as you are being knighted,your pants fall down.’ This characterization from Nagel (1979), is more inclusive than the one givenin the text, but I will only need the less inclusive characterization in what follows.


but Phosphorus is not shining puts her in error no matter how the world turnsout to be compatible with metaphysical possibility. However, Jane’s ignoranceof the identity of Hesperus and Phosphorus may prevent her from being ableto discern her error except by empirical investigation, and if that is so, then herbelief that Hesperus is shining but Phosphorus is not, need not be absurd. On theother hand, if Jane believes that Taylor saw Hunter but that Hunter was not seenby Taylor, then not only is she in error no matter what else is true of the world,she can in principle discern her error with no empirical investigation. That iswhy it is absurd for her to believe that Taylor saw Hunter but that Hunter wasnot seen by Taylor. (Harboring an absurd belief need not make her irrational;perhaps she is irrational only if she violates theoretical rationality in a way thatwould be open to minimal, as opposed to extensive, reflection.)

A belief or system of beliefs that severely violate the norms of theoreticalrationality is absurd even if prudential, moral, or other norms enjoin one tohold it. In that case it would be absurd from the point of view of theoreticalrationality, while permissible, and perhaps even mandatory, from the point ofview of some other system of norms. Thus suppose that the only way to save thelife of a loved one is to take a pill and undergo hypnosis with the result that Icome to believe a contradiction. Suppose further that I agree to induce that beliefin this way. Then I undertake to do something that is absurd from the point ofview of theoretical rationality, but morally mandatory—or at least permissible.Instead of this gaseous ‘point of view’ talk, we could just describe the situationas one in which my coming to believe a contradiction is both absurd (full stop)and morally permissible.

Theoretical rationality is not the only system of norms admitting of severeviolation. We find a severe violation of practical rationality in one whose systemof plans, together with her utilities and subjective probabilities, guarantee thatshe does not maximize subjective expected utility, and in a way that she could inprinciple discern with no empirical investigation. Suppose I accept a sure thingof $1 over a wager in which a fair coin is tossed, such that if it comes up heads Iwin $1,000 and if it comes up tails I gain nothing. This choice guarantees that Ido not maximize subjective expected utility, and I could in principle discern thiswith no empirical investigation. A rival to this classical style of decision theory,prospect theory,⁵ offers a distinct set of norms whose violation takes a differentform from that adduced by subjective expected utility theory. We need not settlethis rivalry here. Suffice it to say that each theory offers an account of practicalrationality, and with it an account of what constitutes a severe violation of thatform of practical rationality. No matter the theory, a severe violation of whatis by its lights practically rational will be deemed absurd. As with the case oftheoretical rationality, we may expect the norms of practical rationality to collide

⁵ See for instance Tversky and Kahneman (1992), and Wakker and Tversky (1993). Green1999a discusses the philosophical significance of these approaches to decision theory.

Mitchell Green 193

with other norms. When that happens, what is absurd from the point of view oftheoretical rationality might be acceptable, and perhaps even mandatory, fromthe perspective of that other system of norms.

My system of beliefs might be absurd without its following that I am beingabsurd in reporting on or otherwise acknowledging them. My therapist showsme films of my behaving in phobic ways around spiders, even the harmless ones.I conclude from these observations of myself that I believe that all spiders aredangerous. However, after learning how few spiders are truly dangerous, I cometo see that on the whole they are not, and that those that are tend to be reclusive.Yet old phobias are hard to shake, and I might acknowledge both this fact andmy new empirical investigations into arachnids with the words, ‘Even thoughI believe that all spiders are dangerous, really they aren’t.’ This is an eminentlyreasonable thing for me to report. After all, each conjunct of that report is bothtrue and justified. Yet if this report is sincere and correct, it follows (as we shallsee below) that I am in error. Indeed I could know that at least one of my beliefsis in error with no further empirical investigation. According to our account ofabsurdity, it follows that my system of beliefs is absurd. That is compatible withthe fact that when I assert, ‘Even though I believe that all spiders are dangerous,really they aren’t,’ this assertion is true, justified, and not the least absurd.

The assertion might be true, justified, and not absurd. It might also be sincere.If it is, then it reports-and-expresses a pair of beliefs that cannot fail to put mein error. It reports my belief that all spiders are dangerous. It expresses my beliefthat not all spiders are dangerous. This pair of beliefs is absurd. But there the pairsits, and I have good evidence for the presence of that pair of beliefs. So it can betheoretically rational for me to report or express a system of beliefs that are sureto violate theoretical rationality, even when those beliefs are my own. If by dintof phobia or obtuseness you’ve violated norms of theoretical rationality to thepoint of absurdity, you’re only being reasonable in acknowledging the fact.

It can be theoretically rational for me to acknowledge someone’s severeviolation of theoretical rationality, even my own. It can also be theoreticallyrational to commit such a violation. Anyone who writes a sizeable book has agood chance of making an error somewhere. If she predicts in her preface thatshe has made an error somewhere in the ensuing five hundred pages, she is onlybeing reasonable. On the other hand her preface together with the ensuing textcannot all be true. In addition, if we assume that she believes everything she saysin her book (including the preface), her belief system is absurd. How then canher prefatory prediction of error be reasonable? The phobic stands back fromhis belief that all spiders are dangerous and acknowledges its falsity while alsoacknowledging that, being bound up with habits of mind, autonomic responses,and the like, it is hard to shake. The modest author stands back from the manybeliefs she expresses in her book to acknowledge that the limitations of herintellect (and those of her research assistants and other authors on whom she hasdrawn) make it likely that she has erred somewhere in those five hundred pages.


Those limitations are epistemic limitations—limitations on one’s ability to liveup to norms of theoretical rationality. It is theoretically rational to acknowledgethose limitations; after all, the author knows she is no less fallible than others.This is to say that it can be theoretically rational to harbor a system of beliefs thatseverely violate norms of theoretical rationality.⁶ Hence, it can be theoreticallyrational to harbor a system of beliefs that is absurd from the point of view oftheoretical rationality. The so-called Preface Paradox is thus a misnomer. Insteadwe should acknowledge the existence of prefatory absurdity while denying thatsuch utterances or beliefs are paradoxical at all.

The phobic puts his finger on his absurdity; he knows which of his beliefsis wrong without being able to shake it. The author draws a circle around herabsurdity; she acknowledges that at least one of her many beliefs is likely to bewrong without being able to pinpoint which ones. You might even know thatI harbor an absurdity without knowing whether the source of that absurdityis theoretical rationality, practical rationality, or some other system of norms.Suppose that a severe violation of one system of norms N1 would result inabsurdity, and a severe violation of another system of norms N2 would do so aswell. Suppose now that you know that my behavior severely violates either N1 orN2 but you do not know which. In such a case, surely, you know my behavior isabsurd in spite of your not knowing the precise source of that absurdity.

Similarly, suppose N1 and N2 are systems of norms, and I perform an actionof which I can know, with no further empirical investigation, that it will violateeither N1 or N2. Here, while we cannot infer, of either N1 or N2, that I haveseverely violated it, we can infer that I have severely violated their conjunction,namely N1 & N2. In that case my behavior is again absurd in spite of our notbeing in a position to locate the source of that absurdity in a violation either ofN1 or of N2.

3. SPEAKER MEANING

In what follows I shall use ‘meaning what s/he says’ and its cognates to refer tocases in which an agent says something in such a way as to take responsibilityfor it. On this usage, one means what one says even if one is a liar, and even ifone is mistaken. On this usage, one who assertively utters an indicative sentenceP means what she says only if she stands to be right or wrong on the issue

⁶ Descartes’s theism and views about evil lead him to conclude that so long as we are using ourminds in the most rigorous and careful way possible, God would not permit us to be in error. Forthis reason, Descartes would hold that so long as the author in question has done her utmost toensure the accuracy of her claims, that is, so long as she clearly and distinctly perceives the truth ofeach of these claims, she could not be justified in making her prefatory prediction of error. In lieuof Descartes’s extraordinary but implausible position, we do well to acknowledge the likelihood ofour error even when, as authors, we have done our best.

Mitchell Green 195

of P depending on whether P is true; she also puts forth P as something shebelieves, whether or not she really does believe it. Ronald Reagan did not, wehope, mean what he said in warming up a microphone before a press conferencewith the words, ‘The bombing begins in five minutes.’ So too one does notgenerally mean what one says in rehearsing lines from a play. By contrast, youmean what you say in promising to educate a dying friend’s children only ifyou commit yourself to educating those children, and only if your sincerity inmaking that promise depends upon your intending to educate them. The notionof meaning under scrutiny here is often referred to by the misleading phrase,‘speaker meaning’.

What is speaker meaning? I shall pursue only as much of an answer to thisquestion as is needed for the account below. Paul Grice is widely believed tohave shown that to achieve a case of speaker meaning, one must make anutterance with the intention of producing an effect on an audience, with thefurther intention that this effect be achieved at least in part by the audience’srecognition of your intention. (Grice 1957, 1969, 1982) This is a so-calledreflexive communicative intention. Unfortunately, Grice’s conditions for speaker-meaning are too restrictive (Davis 2003; Green 2003; forthcoming). Speakermeaning is possible in the absence of a reflexive communicative intention. Forinstance, when Herod presents Salome with St John’s severed head on a charger,he both shows her that St John is no longer and means that St John is no longer.⁷He intends Salome to come to believe that St John is dead, but presumablyintends her to conclude this from the presence of the severed head rather thanfrom any recognition of his intention that she believe anything. After all, thesevered head is there for her to see.

Speaker meaning does not require a reflexive communicative intention. Infact, it does not even require a communicative intention (an intention to producea cognitive effect, such as a belief, on an audience). A framed suspect might meanthat she is innocent in saying, ‘I am innocent!’ Yet she is fully aware that noone will believe her and, being realistic, she does not intend to convince anyone.She might not even intend her interrogators to believe that she believes she isinnocent, since she might know that they’re certain she’s lying. Or, gazing intomy newborn daughter’s eyes I might say, ‘All things valuable are difficult as theyare rare,’ meaning what I say, without having the slightest intention to producebeliefs or other attitudes in her or in anyone else. Again, in the film Sleeper,Woody Allen’s character Miles Monroe discovers in his solitary exploration ofa futuristic world a genetically modified chicken the size of small house. Milesremarks, ‘That’s a big chicken.’ In saying this he does not seem to be intendingto produce an effect on anyone, himself included.

⁷ The fact that he is not telling her that St John is no longer is neither here nor there: We knowindependently that speaker meaning does not require telling. One can, for instance, speaker-mean,‘How many apples are in the bowl?’, without making an assertion.


An intention to communicate, to say nothing of a reflexive intention tocommunicate, is not required for speaker meaning. However, one way ofachieving speaker meaning is to harbor another kind of reflexive intention,specifically, that built into the notion of overtness. Doing something overtlyinvolves as a minimum making some aspect of oneself manifest. I might makemanifest the bruise on my arm by pulling back the sleeve that had been covering it.In so doing I need not cause anyone actually to perceive that bruise. What mattersinstead is that the bruise be open to view by appropriately situated observers. Ican likewise manifest my commitment to some proposition P by making thatcommitment open to inspection to appropriately situated interlocutors. I mightdo that by asserting P in front of an audience, be they distracted or not; Imight do it by asserting P in the privacy of my study. In my locked, unbuggedstudy, the only audience to whom I am manifesting my commitment is myself.I nevertheless mean what I say when I say to myself, ‘All things valuable aredifficult as they are rare.’

I can, however, make some aspect of myself manifest without being overt. Thebulging vein on my forehead makes my anger manifest without itself being, orbeing a part of, any overt behavior on my part. I can even intentionally makesome aspect of myself manifest without behaving overtly. As we approach eachother in a dark alley I cough to keep from startling you when we get closer;but I need not be overtly coughing. Contrast this with a case in which I am aschoolteacher who has come upon a young couple displaying affection, and Istentoriously clear my throat before they see me. Here my throat-clearing willbe overt: Not only do I intentionally manifest my presence, I also manifest myintention to manifest my presence.

Overtly to do something requires doing it with the intention that the actbe manifest, and further with the intention that that very intention itself bemanifest. But it is not true that I can achieve overtness merely by having twointentions, namely (a) an intention that some commitment of mine be manifest,and (b) a further intention that this very intention be manifest. Consider thefollowing Strawson-inspired example (Strawson 1964). You are exploring a housefor possible purchase. I want to manifest to you my belief that it is rat-infested,and so enact the following plan. I will enable you to see me leaving a rat in thehouse, while acting as if I think I am unobserved. I also know that you think mea good and veracious friend, and know that, although you won’t conclude fromthe presence of the rat that it is genuine evidence that the house is rat-infested,you will nevertheless conclude from my odd behavior that I must believe thatthe house is rat-infested.

In acting as I do in the above example, I intend that my belief that the houseis rat-infested be manifest. I thus satisfy condition (a) above. I will also intendin so acting that this very intention (that my belief that the house is rat-infestedbe manifest) itself be manifest. Hence, I satisfy condition (b) as well. However,in this case it does not seem that I overtly display my belief that the house is

Mitchell Green 197

rat-infested. The problem appears to be that, as in the case of my secreting yourmonogrammed handkerchief into the crime scene in order to implicate you,something about my performance is under the table—or at least the floorboards.Contrast this with a case in which you and I are discussing the aforementionedhouse over lunch, and you ask whether I think it infested with anything. I happento have a rat in my briefcase, and wordlessly take it out, dangling it by the tail infront of you. Here, surely, I overtly display my belief that the house has a verminproblem.

From the foregoing it seems that at least one means to achieve overtness in away that will make speaker meaning possible, requires all the relevant intentionsbe out in the open.⁸ One approach to filling this need is to demand that to beovert, the intention that my commitment be manifest self-referentially requiresthis very intention to be manifest as well.⁹ That suggests the following sufficientcondition for speaker meaning:

Sufficient Condition for Propositional Speaker Meaning: S means that P if1. S performs an action A intending that:2. In performing A, S’s commitment to P be manifest, and that it bemanifest that S intends that (2).

P might be the proposition that it is blustery outside, with A being the uttering ofcertain words or a non-conventional action such as the throwing open of curtainsto reveal a looming storm. In the Herod case the P in question is the propositionthat St John is dead, and the action is Herod’s presenting of St John’s head onthe charger. Once again, one’s commitment to P might be manifest withoutanyone being aware of this fact. As a result one can overtly intend that one’scommitment to P be manifest without intending to produce effects on others. Inlight of our Sufficient Condition for Propositional Speaker Meaning, then, onemight mean something without intending to produce effects on any audience,and one can mean something in the course of overtly making one’s commitmentmanifest.¹⁰

⁸ That is compatible with our having ulterior motives in cases of speaker meaning. In remarkingon the weather I might be trying to be sociable, or for that matter be intending to distract you whilepurloining your maraschino. In either case I nevertheless speaker-mean that it’s a nice day when Isay, ‘Nice day.’ For further discussion of the distinction between ulterior and ostensible motives inspeaker meaning see Green (1999b).

⁹ Green (forthcoming: ch. 2), develops this point in further detail. In addition to elucidatingthe sufficient condition given below for propositional speaker meaning, speaker-meaning that P,that work also develops the notion of objectual speaker meaning, in which an agent means α, whereα is an object.

¹⁰ Some authors have been skeptical of the very possibility self-referential intentions. (Suchscepticism is not to be confused with scepticism of the doctrine that all intentions are self-referential.One can accept that self-referential intentions are possible while remaining neutral on the questionwhether all intentions are self-referential.) Thus for instance Seibel (2003) writes, ‘the content of[the self-referential intention] contains an element which refers to the intention itself. But what doesthat element look like? . . . How does it single out the intention and nothing but it? By identifyingfeatures, i.e., properties which are exclusively possessed by the intention? But what could be these


4. SPEAKER MEANING IN YOUR HEART

‘Speaker meaning’ is a misleading label because it applies to cases involving nospeech. Suppose that we share no language and I want to alert you to the danger ofthe quicksand nearby. I mime out before you my being pulled under the surfaceby quicksand. In so doing I might speaker-mean that there is quicksand nearbywithout making a sound—indeed without moving my lips. The label ‘speakermeaning’ is also misleading because one can speaker-mean things without doinganything publicly observable. To see this, consider first of all that the distinctionbetween speaker-meaning and merely mouthing words has an analogue in therealm of thought. I can think through the lines of a poem in my head withoutassenting to what those lines say. On the other hand, I might not only saysomething to myself but also mean it in such a way as to be committed to it. Aswith speech, I might thereby mean some proposition P in such a way as to standto be right or wrong on the issue of P depending on whether P is true. If in thecourse of a vigorous morning run I resolve to make it up to the top of the hillbefore me by saying in my heart the words, ‘I shall conquer that hill!’, I mightmean what I say. If I do, then I undertake a commitment that will be satisfied ifand only if I conquer that hill.¹¹

I take the phenomenon of speaker-meaning something to oneself to be familiarand uncontroversial. In spite of this, talk of so-called ‘inner speech’ tends tomake philosophers nervous because of its association with discredited attemptsto base an account of linguistic meaning upon an incorrigibly private language.¹²However, I here make no claim that linguistic meaning can be explained in

features?’ Intentions admit of the same act/object dichotomy as do many other intentional states,and like other mental events, are spatiotemporally located. On the modest assumption that notwo intentions have identical spatiotemporal coordinates, we may then use such coordinates toindividuate intentions. One might still wonder what the content is of an intention whose contentrefers to that very intention, which itself comprises both an intending (a state or act) and a content.One answer may be given in terms of an analogue of truth conditions applicable to intentions,namely satisfaction conditions. Just as the thought, had as I regain consciousness after a near-fatalaccident, ‘This thinking is miraculous’, will be true just in case that thinking is, indeed, miraculous,so too, the intention, ‘This intention shall be manifest’ will be satisfied just in case that intention is,indeed, manifest.

¹¹ Anselm distinguishes between two things the fool might be thought to be doing when he saysin his heart, ‘There is no God’ (1995: 101). On the one hand the fool might be silently saying thesewords to himself. Anselm thinks this case is possible. What he does not think possible is anothercase in which the fool not only says to himself, ‘There is no God,’ but also understands what he issaying. Anselm’s point is that if the fool were to grasp the concept of God, he would immediatelysee that God could not fail to exist. In what follows I will not lay down any limitations on whatabsurd thoughts a person can entertain even as he understands the words—if such there be—inwhich they are couched.

¹² For a creditable such attempt, however, see Davis (2003).

Mitchell Green 199

terms of mental acts of ascribing meaning to words. For all I say in this essay,acts of inner speech are only possible in the context of a linguistic community.However, one might still challenge the idea that it is possible overtly to dosomething in the privacy of my own thoughts. Does overtness, invoked in mysufficient condition for speaker meaning above, make any sense in this context?I contend that it does. My believing that it is raining outside does not guaranteethat I will tell myself that it is. For instance if I am conscious of the rain butnot attending to it, I will not tell myself that it is raining. Hence I can believethat it is raining outside without speaker-meaning anything. I might also makemy belief that it is raining manifest to myself by, for instance, going to the closetto take out an umbrella. Here too, however, I haven’t speaker-meant anything,but have merely acted on and thereby displayed my belief that it is rainingoutside.

What if I intentionally manifest to myself my belief that it’s raining? Thistoo is not sufficient for speaker-meaning something to myself. I film myselfgoing outside with an overcoat, hat, and umbrella. Unfortunately I also haveanteriograde amnesia, which prevents me from retaining new information formore than a few seconds. (Assume that I know that I suffer from this malady.) Ileave the film of myself for a later stage of myself to watch. I know that that laterperson-stage will watch the film and see an earlier person-stage going out dressedfor bad weather. I also know that the later person-stage will not know who orwhat made the film, and with what intent. Because of this, the earlier person-stagemight intend, in placing the film where he does, to manifest to himself his beliefthat it is raining. However, he will in all likelihood not believe that in so doing heis manifesting to himself this very intention. After all, the later person-stage willhave no idea where the film came from; and the earlier person-stage knows this.Because of this, the earlier person-stage will in all likelihood be unable overtlyto manifest his belief that it is raining. This squares with our intuition that inleaving the film there for the later stage of himself to see, the earlier stage doesn’tspeaker mean that it is raining.

In order to speaker-mean something to myself, what is needed instead is acase in which I not only intentionally manifest to myself my belief that it israining, but also intentionally make my intention to manifest that belief tomyself, itself manifest. If I have anteriograde amnesia, and know it, I mightcreate a movie to be a screen-saver on my computer showing me ostentatiouslygoing to the closet to get a hat and umbrella. In that movie I stare significantlyat the camera while donning gear for the weather. In placing the movie whereI do, I may reasonably intend not only to make my belief that it’s rainingmanifest to my later self, but also intend to make manifest to my later self thisvery intention. (Assume that I have retained enough knowledge from before theaccident producing my amnesia that I know that the only way this movie couldhave ended up as a screen-saver was if I had put it there myself.) As with the


case of the quicksand example above, here I speaker-mean something withoutproducing any words.¹³

The situation is similar when I say something to myself while uttering nowords aloud. If I say in my heart, ‘It’s raining,’ I may reasonably intend to makemy belief that it’s raining manifest to myself. Further, in a standard such case,not only do I intentionally manifest to myself my belief that it’s raining, I alsointentionally make that very intention manifest to myself. For I will normally beaware that the only reason I would ‘hear’ myself saying these words was that Ihad intended to manifest my commitment to myself. In an unusual case, suchas schizophrenia, I might be suffering from delusions of the sort that cause meto hear voices. If I am schizophrenic and know it, then I probably also knowthat I am prone to hear such voices. In that case it is less clear that I canintend, by saying something to myself, to manifest my intention to manifest abelief to myself. One symptom of schizophrenia, then, may be an inability tospeaker-mean things to oneself by saying things in one’s heart.

Speaker-meaning something in the privacy of one’s own thoughts seemspossible, and indeed not at all unusual. This becomes clear in light of oursufficient condition for speaker meaning, which requires no communicativeintentions, much less reflexive communicative intentions. However, it mightseem that a disanalogy between speaker meaning in one’s heart and speakermeaning done in public is that only the latter can fail to be sincere. No mysterysurrounds the idea of lying to others. What about lying to oneself ? Sometimeswe use this expression to refer to people believing things on insufficient evidence.For instance, Hunter’s lying to himself in thinking he can make it in time for hisChicago flight might be due merely to the facts that (a) his chances are so slim,and (b) he should have known better. This need not involve his telling himselfanything that he thinks untrue. On the other hand, everyday experience alsosuggests that I can say something to myself, meaning it, without believing whatI say. I tell myself that this is my last piece of pie when I know perfectly well thatby the time I leave the room, the pie plate will be clean. The phenomenon isalso reflected in literature. For instance, in Ann Packer’s, The Dive from Clausen’sPier (Knopf, 2002), the narrator Carrie is visiting her fiance in the hospital afterhis spinal injury from a dive off a pier. She is gradually losing interest in caringfor him, and wants to move away:

Again he closed his eyes, and now tears seeped out, a single trail moving down eachcheek. I set his hand down and began stroking his forearm again. I wish I could say I feltselfless then, unaware of myself. That I was thinking only of him, or that I wasn’t eventhinking. But I was: This is me doing the right thing. This is me being brave and strong forMike. (p. 102)

¹³ Those sympathetic to a ‘memory links’ account of diachronic personal identity need not bedistracted with the question whether a person could survive an amnesia of this sort. Grice’s (1941)version of such an account can accommodate this kind of case.

Mitchell Green 201

When Carrie says, ‘This is me being brave and strong for Mike,’ she is tellingherself that she is being brave and strong for Mike. She speaker-means it. Yet shedoesn’t really think she is being brave and strong for him. She knows full wellthat she is moving away from him emotionally, that her visits to the hospital arebecoming less frequent and more halfhearted, and that she may not even lovehim any more.

We have intuitive support for the idea that one can speaker-mean somethingto oneself. We now have, in addition, intuitive support for the idea that one canspeaker-mean something to oneself without believing it. Intuitions are fallible,however, and this latter intuition might seem mistaken. A speech act that isnot sincere is typically made with the intent to deceive. Is it plausible thatone can intend to deceive oneself by saying something to oneself? Likewise,speaker-meaning is widely construed as an attempt to produce an effect on anaudience by means, at least in part, of their recognition of the speaker’s intentionto produce that effect. I have rejected that account and suggested that it is at leastsufficient for speaker meaning that one intentionally and overtly manifest one’scommitment to a proposition. Is it possible to do this in the privacy of one’s ownthoughts without being sincere?

I manifest my commitment to P by making that commitment available.I might do that by asserting P in front of a distracted audience; I make mycommitment available to them even if they don’t acknowledge that commitment.If I manifest my commitment to P by asserting it in the privacy of my own study,I make that commitment available to myself and no one else. (Assume that thereare no recording devices or eavesdroppers.) But much the same holds for thingssaid in my heart. There are many things I believe that are not manifest to me.Introspection, psychotherapy, and elenchus are all ways of dredging up beliefsinto consciousness. When I follow one or more of these paths, I might articulatewhat I find by consciously thinking to myself some such thing as: ‘I do seek mycolleagues’ approval!’; or ‘I don’t think that consequences are all that matter formorality!’ As in the case of what I utter in my locked study, when I say one ofthese things ‘in my heart’ I make my belief manifest to myself and to no one else.That is still enough for me to speaker-mean that I reject consequentialism, andit is still enough for me to mean that not enough of my self-worth comes fromwithin.

I cannot make manifest, to myself or anyone else, a belief that I do not have.How then can I fail to be sincere in the confines of my own thoughts? In thecases in which I make a sincere assertion I not only make my belief manifest, Ialso (intentionally and overtly) manifest my commitment to the content of theproposition that I assert. Hence, when I sincerely assert that it is raining, I notonly manifest my belief that it is raining, but also manifest my commitment tothe proposition that it is raining. By contrast, I might manifest commitment toa proposition that I do not believe. Similarly, when Carrie says to herself, ‘Thisis me being brave and strong for Mike,’ she intentionally and overtly manifests


commitment to the proposition that she is being brave and strong for Mike,without actually believing that proposition. She would like to be being brave andstrong for Mike, but she doesn’t think she is (and in fact she isn’t).

5 . SPEECH ACTS AND THEIR NORMS

We now see that speaker meaning is possible in the context of one’s ownthoughts, and that it is even possible to speaker-mean something in such a milieuwithout being sincere. Further, like beliefs (which are governed by the norms oftheoretical rationality), and actions (which are governed by the norms of practicalrationality), speech acts are governed by their own system of norms. That systemoverlaps with but is not coextensive with the norms of theoretical rationality.According to speech act norms, assertions are to be relevant, justified by thespeaker’s evidence, and sincere. Promises are to be sincere and not obviouslyimpossible to fulfil. Questions are not to presuppose anything controversial. Andso forth.

What would be a severe violation of speech act norms? It is absurd to attemptto promise you something that is obviously beyond my control to bring about,e.g., to make it the case that the Napoleonic wars did not occur if it is commonknowledge between us that those events took place in the past. It is absurd toattempt to bequeath something to you that, as you and I both know, is not myown to give, such as the Horsehead Nebula. These cases square with the accountof absurdity given in Section 2 above, for in all of them one can discern with nofurther empirical investigation that one will violate a system of norms.

Unlike some other speech acts, assertion is beholden to a norm of theoreticalrationality, namely that one is to assert only what is true. It also includes a normwhose source is less clear: Assert only what one believes. We don’t need to decidehere whether this sincerity norm governing assertion flows from theoreticalrationality or from some other system of norms. Perhaps it does, or perhaps onthe other hand it is a sui generis norm of assertion. For in either case, an assertionabout which it can be known, just by inspecting its content, that it is false, isabsurd. So are assertions whose content is falsified by particular aspects of theiruse. For instance, an utterance of ‘I am not now uttering any words,’ is falseeven though its content could be true. It is thus absurd knowingly to utter sucha sentence even if it expresses a proposition that could be true. Similarly for aconscious thinking of, ‘I am not now thinking’: It is absurd consciously to thinkthis sentence even if it expresses a proposition that could be true. (I assume thatif one consciously thinks this sentence, one knows that one is doing so.)

Suppose that we know of some promise P that anyone who makes it orpurports to make it, meaning what he says, has either promised to do somethingthat she will not, in fact, do, or has made a lying promise. Suppose, in addition,that we may infer this with no further empirical investigation. Were we able to

Mitchell Green 203

infer with no further empirical investigation that the promise will not be kept,we would be able to conclude that the promise is absurd. Were we able to inferwith no further empirical investigation that the promise is not sincere, we wouldbe able to conclude that it is absurd. Knowing that no matter how the worldis, the promise is either one or the other of these is also enough to justify ourconcluding that it is absurd. It does not follow from this, though it may also betrue, that such a promise is impossible. Fortunately we need not settle that issue.

Similarly, suppose that we know of some sentence or thought S that anyonewho asserts (thinks) it, meaning what they say (think) is either mistaken or isnot sincere. We know this without requiring any empirical investigation, andno matter what else is in the speaker’s (thinker’s) mind. Did we know with nofurther empirical investigation that the speaker’s (thinker’s) utterance is mistaken,we could conclude that the utterance is absurd. Did we know with no furtherempirical investigation that the speaker’s (thinker’s) utterance is not sincere, wecould conclude the same thing. As with the case of the promise just contemplated,our ability in principle to infer with no further empirical investigation that theassertion is either mistaken or not sincere is also enough to justify our conclusionthat it is absurd. Here too it may be going too far to say that such an assertion(thought) is impossible, and here too we are fortunate that we need not settlethat issue.

I will argue that all instances of Moorean absurdity are absurd on accountof severely violating norms for speech acts in this way: We may know with nofurther empirical investigation that either the speaker is in error, or is not sincere.We might not know which one it is, but as we saw in Section 2 above, thisdoes not matter. What matters is that we may know with no further recourse tofacts about the world that the speaker is violating one of the norms internal toassertion (or whatever speech act is the vehicle of her Moorean utterance). Thatis perfectly compatible with the content of what is asserted being true. It is alsoperfectly compatible with the speaker believing what is asserted.

6 . MOOREAN SPEECH AND MOOREAN THOUGHT

My approach assumes that all instances of Moorean absurdity are speech acts,even while some are not acts of speech. This assumption seems to be in conflictwith the widely shared view that Moorean absurdity can be realized merely bybelieving some such thing as ‘P, but I don’t believe it’, rather than saying it.Surely belief is not a form of assertion, not even a form of mental assertion? It isconsiderations like these that make those interested in Moorean absurdity doubtthat a speech-act approach could possibly be broad enough to cover the relevantexplananda.

It is not, however, true that merely believing, ‘P, but I don’t believe it,’ isabsurd, or causes one to be absurd. If either that entire belief or one component


of it is not accessible to conscious introspection, then the agent may be unableto discern with no further empirical investigation that she is in error. Supposethat either the entire belief, ‘Spiders are harmless, but I believe they aren’t,’ orone of its two conjuncts, is locked in my unconscious in such a way that itwould require at least a year of intense psychotherapy including films of my ownbehavior, virtual reality exercises, fMRI information, and so forth, to come tobe aware of it. In believing both conjuncts of ‘Spiders are harmless, but I believethey aren’t,’ I cannot fail to be in error.¹⁴ However, the same goes for my beliefthat Hesperus is shining but Phosphorus is not. In both cases it would takeempirical investigation to determine that I must be in error. In the latter case Imust learn the identity of Hesperus and Phosphorus; in the former case that Ibelieve both conjuncts of, ‘Spiders are harmless, but I believe they aren’t.’ Justas it is difficult to see what would be absurd about believing that Hesperus isshining but Phosphorus is not, so too it is difficult to see what would be absurdin this case of believing both conjuncts of, ‘Spiders are harmless, but I believethey aren’t.’

If some ‘part’ of a Moorean belief is unconscious, then one may be unable tobring it into consciousness by introspection. Rather, one may need to do someempirical investigation to detect its presence. Because of that, an unconsciousbelief of this sort is no more absurd than believing both that Hesperus is shiningand that Phosphorus is not. It may not be intuitively clear what an unconsciousbelief having the form either of a commissive or omissive Moorean sort wouldbe like; yet we need not dwell on this issue. The reason is that approaches takingMoorean belief as the explanans and other cases of Moorean absurdity as theexplananda,¹⁵ predict that believing the following two propositions puts one ina situation of Moorean absurdity: (a) P, (b) I believe that not-P. One need notbelieve their conjunction in order to exemplify Moorean absurdity, accordingto this standard account. After all, the argument used in the footnote to thelast paragraph applies equally well to anyone who merely believes both (a) and(b). Similarly for the commissive case. Accordingly, all we need to imagine is an

¹⁴ We discuss the omissive and commissive cases in order. Assume that belief distributes overconjunction. Then one who believes ‘P, but I don’t believe it’ believes P, and believes that he doesnot believe that P. That latter belief must be in error. For the commissive case suppose that I believethat (p & I believe that not-p). Then assuming that belief distributes over conjunction, I believethat p. But what I believe, that p & I believe that not-p, is true only if I also believe that not-p.Thus what I believe is true only if I have contradictory beliefs about p, one of which must thus bemistaken. A Moorean belief, be it omissive or commissive, conscious or unconscious, cannot fail toput the person harboring it in error.

I should mention as well that we are here assuming that the unconscious belief about spiders canonly be made known to me by empirical investigation: it is ‘subconscious’ rather than ‘preconscious’.According to theories of the ‘adaptive unconscious’ now gaining currency (canvassed in Wilson2002), some unconscious material can only be known in a ‘third personal’ way rather than in a ‘firstpersonal’ way; they can thus only be known by empirical investigation. I take the spider belief to bea case of this kind.

¹⁵ Positions of this kind are discussed in Chapter 1, Section E.2 in this volume.

Mitchell Green 205

agent whose belief (a) is unconscious while his (b) belief is conscious. In thatcase, while his entire set of beliefs put him in such a position that he cannot failto be in error, he is not absurd for holding this set of beliefs. He cannot fail tobe in error, yet he will be unable to discern this fact without further empiricalinvestigation. For this reason, his system of beliefs is no more absurd than hisbelieving that Hesperus is shining and Phosphorus is not shining.

I claim that a necessary condition of an agent’s being absurd is that her severeviolation of norms is one that she can in principle come to be aware of withno further empirical investigation. Accordingly, if some of my Moorean beliefis submerged in my unconscious, then I am not, at least on this basis, absurd.Rather, one needs consciously to think some such sentence as ‘P, but I don’tbelieve it,’ in order to generate absurdity recognizably Moorean. But this byitself is not enough, for there are many propositions we think through withoutcommitting ourselves to them. Instead, one needs not just consciously to thinkit, but more specifically to speaker-mean it in the way described in Section 3above. As we saw in that section, this does not require that one believe what onesays to oneself. It does require that one manifest commitment to what one says tooneself, and in particular a commitment to the truth of what one says. (Strictlyspeaking, I have given only a sufficient condition for speaker meaning. However,I take it as not in need of argument that merely thinking a thought, and merelymanifesting commitment to oneself, are not sufficient for speaker meaning.)

7 . MENTAL ASSENT

The approach adumbrated thus far might also seem to be superseded by the viewthat takes ‘mental assent’ as the core notion, and explains other cases in its terms.However, as this term is normally used, mental assent must be sincere (Shoemaker1995). The reason is that mental assent is construed by authors such as Shoemakeras an episodic instantiation of belief. On this usage, one cannot mentally assentto a proposition that one does not believe. Because of this, Shoemaker’s approachdoes not have sufficiently broad scope. It does not account for cases in which aperson says to himself a Moorean sentence that he does not in fact believe. Asnoted above we have everyday familiarity with the experience of saying things tooneself that one means but does not believe. Doing so with a Moore sentence canstill be absurd. It follows that the mental assent approach lacks adequate scopeto be the source of a general explanation of Moorean absurdity.

8 . SHOWING WHAT ’S WITHIN

Showing comes in at least three forms. First of all, I might show my courageby acting bravely. My brave behavior is good evidence for my courage. A


grammatical tag for this category is showing-that, since the brave behavior alsoshows that I am courageous. Second, recall from our discussion of manifestationabove that I might show something in such a way as to make it perceptible.I show my bruise, and thereby enable others to see that bruise. Let us putthis perception-enabling form of showing under the rubric of showing-α, where‘α’ is a singular term. Finally, I might also show how something looks, feels,sounds, etc. I present your nose with a durian and enable you to know howit smells. Similarly, the trepidation in my voice might enable you to knowhow my anxiety feels if you are sufficiently empathetic. If you are sufficientlyempathetic, then hearing my voice may enable you to imagine feeling as I do.If you can do that, then you know how I feel. These three forms of showingwe may label showing-that, showing-α, and showing-how. The discussion belowdepends primarily on the notion of showing-that; showing-α has a cameo inSection 10.

We are now in a position to tie together various strands to support an accountof what is absurd in the utterances that Moore was the first to discover. Assertionspurport to show beliefs. If they are sincere, they do show those beliefs. This isnot because they make beliefs perceptible (that doesn’t seem to make sense), norbecause they show how a belief feels (beliefs don’t seem to feel like much ofanything). Rather, an assertion is (inter alia) evidence that the speaker believeswhat is asserted. That is why the sincerity of an assertion shows that you believewhat is asserted. Now suppose you show a belief or other attitude that you alsodeny having. Then since ‘show’ is a success verb, that denial must be in error.The showing might be public (in an utterance) or private (in a saying to oneself ).If you show a belief (or other attitude) and then go on to describe yourself asbelieving its contradictory, then whether or not this latter statement is true, youare in error. On the other hand if you purport to show a belief (or other attitude)that you do not in fact have, then you are not sincere. Suppose then that you(perhaps silently) assert

1. P but I don’t believe that P.

Then by the assumption that assertion distributes over conjunction, you haveasserted P and have asserted that you don’t believe that P. The former assertion iseither sincere or not. Suppose it is sincere. In that case, it shows your belief thatP, but then your other assertion, that you don’t believe that P, is in error. In thatcase you’re in violation of the norm that assertions are to track the truth. On theother hand suppose that the assertion of P is not sincere. Once again you are inviolation of a norm of assertion, namely to assert only those things you believe.So either the assertion is sincere or it is not; but in either case we may infer withno further empirical investigation that you are in violation of a norm of assertion.Hence with no further empirical investigation we may conclude of someone whoasserts, ‘P but I don’t believe it’, that she is in violation of a norm of assertion.It is a severe violation of a norm or system of norms to perform an act of which

Mitchell Green 207

it can in principle be inferred with no further empirical investigation, that itwill violate that norm. From our original contention that absurdity consists ina severe violation of a system of norms, we may infer that assertion of ‘P but Idon’t believe it,’ is absurd.

So too, suppose you assert

2. P but I believe that not-P,

With assertion-distribution we may infer that you have asserted P and haveasserted that you believe that not-P. Your first assertion is either sincere or itis not. Suppose it is sincere. Then your second assertion, that you believe thatnot-P, is either correct or incorrect. In the former case, you are in error: for yourfirst assertion, being sincere, shows your belief that P, and so you believe thatP; while the correctness of your second assertion implies that you believe thatnot-P. Anyone who believes both P and not-P is in error. On the other hand,if the second assertion, that you believe that not-P, is incorrect, then you arein violation of a norm of assertion. Likewise and as before, if the first assertion,namely that P, is not sincere, you still violate a norm of assertion. It follows thatif you assert ‘P but I believe that not-P’, then we may conclude with no furtherempirical investigation that you are in violation of some norm of assertion. Itis a severe violation of a system of norms to perform an act of which it can inprinciple be known with no further empirical investigation that it will violatethose norms. From our original contention that absurdity consists in a severeviolation of a system of norms, we may infer that assertion of ‘P but I believethat not-P,’ is absurd.

I cannot be sure that the norm that assertions are to track the truth is a normof assertion. It might instead be a norm of theoretical rationality, applying toassertion, as with any other activity aiming at the truth, simply by universalinstantiation. However, if this is so it will not undermine our explanation ofthe absurdity in cases such as (1) and (2). For if this is a norm of theoreticalrationality only, then one who asserts (1) performs an act that, no matter howthe world turns out to be, either violates a norm of assertion or violates a norm oftheoretical rationality. Likewise, if the norm that assertions are to track the truthis a norm of theoretical rationality only, then one who asserts (2) performs an actthat, no matter how the world turns out to be, either violates a norm of assertionor violates a norm of theoretical rationality. As we saw in Section 2 above, thatstill suffices for absurdity.

9 . MIXED ILLOCUTION CASES

The approach offered here generalizes with little difficulty to cases other thanthose involving only assertion. I shall consider two such cases, one involvingsupposition, and the other involving interrogatives.


Even off stage, one can utter an indicative sentence with other than assertoricforce by, for instance, forwarding a proposition merely as something to beentertained, as a supposition for the sake of argument, or as a conjecture.Taking, for brevity, only the case of supposition for the sake of argument (whichhereafter I shall just refer to as supposition), offhand it seems we can intelligiblyinquire whether a speech act of supposing is sincere no less than we can inquireinto the sincerity of a compliment, an assertion, an apology, or a promise.¹⁶A supposition’s sincerity does not require the speaker to believe what she hasproffered as a supposition. Rather, it is at least sufficient for her sincerity thatthe speaker perform such a mental act of supposing as one might do in one’sunspoken hypothetical deliberations. (I doubt that such a mental act is requiredfor the speaker to be sincere, but that question need not be settled here.) Theresult of such a mental act is that the speaker is in the intentional state ofsupposing, occupancy of which I shall take to be at least a necessary conditionfor her speech act of supposing to be sincere.

That a speech act of supposing can be assessed for sincerity might be obscuredby the fact that it is not pragmatically deviant to assert or believe

3. P, though my state of mind is not one of supposing that P.

One can also suppose this sentence (or its content) by imagining a case in which:P holds but one refrains from supposing P. Further, one can assert or believewithout oddity

4. P, though my state of mind is one of supposing that not-P.

One can also suppose it, thereby supposing both P and the proposition thather state of mind is one of supposing not-P. No paradox need result.¹⁷ On theother hand, as exemplified by the relevance of sentences involving interrogativesto the question what it is for an interrogative to be sincere, we need not restrictour inquiry to examples in which the two conjuncts of either of the sentencesjust displayed are put forth with the same illocutionary force, or are held underthe same propositional attitude. Accordingly, consider a situation in which aspeaker inscribes P under the scope of a supposition sign of the sort used innatural deduction systems. That sign indicates without asserting that P is putforth with the force of supposition, and will indicate that all reasoning carriedout to its right and below P are within P’s scope. Assume further that the speaker

¹⁶ In its use as part of a verb phrase of the form ‘A supposes’ taking complements of the form‘that P’, ‘suppose’ is often used to impute beliefs, sometimes with the suggestion that the believeris in error. I shall nevertheless consider only its use to refer to the acceptance of a premise for thesake of argument. This usage of ‘suppose’ is thus also to be distinguished from uses of ‘assume’ torefer to a person’s commitment, often unacknowledged, to the truth of a proposition. (‘Assume’ ishowever used at other times to refer to the use of a premise for the sake of argument, as are ‘say’,‘pretend’, and ‘imagine’, and what is said below will apply to all these uses.)

¹⁷ Sorensen (1988) argues along similar lines that there are no imagination blindspots, and I takeit that he would say the same for supposition. We are about to see that these points may be grantedwithout its following that there is no analogue for the case of supposition of the Moore paradox.

Mitchell Green 209

is using the sign of supposition to guide her informal reasoning rather than toaid her in the use of a formal system, and that she appends to her inscriptionof P the parenthetical remark that her state of mind is not one of supposing P.Such a parenthetical remark would normally be read not as within the scope ofthe supposition sign but rather as being put forth assertorically, and thus wouldbehave analogously to the parenthetical that occurs in

5. If (as is indeed the case) snow is white, then grass is green.

Here the speaker asserts the conditional, ‘If snow is white, then grass is green’while also putting forth ‘snow is white’ assertorically even though grammaticallyspeaking the parenthetical clause occurs in the antecedent.¹⁸ Likewise, in

6.P (though my state of mind is not one of supposing Pfor the sake of argument)

the content of the parenthetical clause will normally be read as being put forthassertorically rather than as part of what is being supposed. In addition, thecontent P and the content expressed in the parentheses can be conjoined toform a logically consistent proposition. Nevertheless, if someone were to writethe above display on a chalkboard addressing an audience familiar with theconventions of natural deduction, their audience may have no choice but to findthe performance absurd. A first, charitable response to this performance might beto construe the parenthetical remark as retracting the supposition of P. However,just as we may be unable to construe one who says, ‘P but I don’t believe it’ asexpressing a mid-utterance change of mind, so too this interpretation may beunavailable if, for instance, the speaker goes on to infer things from P. A secondcharitable response is that the speaker is dissociating herself from her suppositionof P, perhaps because her commitment to not-P is so deeply entrenched that shecannot bring herself to reason as if P is true. This construal will also be ruledout by the speaker’s going on to reason under P’s scope with adequate facility. Athird charitable response might be to construe the speaker as signaling that herstate of mind is not merely one of supposing P, but is instead one of acceptingP in a way that may seem stronger than supposition (i.e., belief or conjecture).This interpretation, too, may be ruled out by contextual factors, and wouldhave been explicitly ruled out had the speaker instead inscribed ‘P (though mystate of mind is not one of accepting P in any way at all)’, within the scopeof the supposition line. Attempts at charitable interpretation might, in the end,

¹⁸ This perspective on parentheticals is defended in Green (2000b).


meet with failure, with the result that the audience would have no choice butto find the speaker’s performance absurd. Such a case would be an analogue forsupposition of the absurdity to be found in certain utterances of ‘P but I don’tbelieve it’.¹⁹

How shall we explain what is absurd in such a case? Suppose that the speaker’sputting forth P as a supposition for the sake of argument is sincere. Thenher doing so shows her state of mind of supposing that P. Her parentheticalutterance, having the force of an assertion, must therefore be in error, whenceshe is in violation either of theoretical rationality or norms of assertion (or both).On the other hand, if the speaker’s putting forth P as a supposition for the sakeor argument is not sincere, then she is in violation of a norm of the speechact of supposition. (No absurdity arises from a person’s putting forth P as asupposition for the sake of argument while going on to avow that she supposesnot-P.) Accordingly, anyone who utters (6) above in the conditions described(including the condition of being within the scope of the supposition sign) is inviolation of some system of norms, and we may determine this with no furtherempirical investigation. That is what makes her utterance absurd.

Some authors have suggested that the following utterance exhibits Mooreanabsurdity:

7. It’s raining but I don’t know that it is.

The point, however, needs to be handled with care. It is clear that someutterances of this sentence in a speech act generate no absurdity. We knowfrom the discussion of supposition just offered that a speaker can utter (orthink) an indicative sentence in a speech act without making an assertion.Suppose, then, the first conjunct of (7) is uttered as, say, a conjecture. It isperfectly appropriate to put forth a conjecture while making clear that you don’tknow it to be true. This is attested by the fact that while it is appropriate torespond to my assertion of P with the challenge, ‘How do you know?’, it isnot appropriate to challenge my conjecture with that question. It seems, moregenerally, that (7) is a case of Moorean absurdity only if the first conjunct is putforth assertorically.

¹⁹ Some authors, for instance Searle and Vanderveken (1985), and Rosenthal (1998), havesuggested analogues of the Moore paradox involving neither belief nor assertion without developingthe possibility of such a paradox for the case of supposition. However, not all of these authorskeep sight of the fact that a speech act can generate Moorean absurdity only if it characteristicallyexpresses an intentional state. As Heal (1977) argues, although a speech act such as an imperativemight seem to generate Moorean absurdity, as in ‘Shut the door, but I don’t want you to shutthe door’, this appearance is probably deceptive. The reason is that imperatives are not speech actsone of whose roles is the expression of an intentional state, and it is for this reason not the casethat imperatives are speech acts that characteristically express an intentional state. One uttering animperative might provide her addressee with evidence of her intentional state (perhaps a desire), butit does not follow from this that the imperator expresses any such state. Further, one performing animperative might adventitiously express such an intentional state as a desire, but this fact is of littleinterest to the study of Moorean absurdity.

Mitchell Green 211

Let us suppose then that the first conjunct of (7) is an assertion. Then it doesseem plausible that we have a case of Moorean absurdity. If so, then we mayaccount for this fact as follows. An assertion, if sincere, justified, and correct,shows not only one’s belief, but also one’s knowledge. Going on to deny that oneknows what one asserts must put one in error. Similarly, in

8. It’s raining but I know that it isn’t

if one is sincere, justified, and correct, the first conjunct shows one’s knowledge,whence the second conjunct must be in error. (Unlike believing p and believingnot-p, one can’t know this pair of propositions.)²⁰

I close this section with a remark about non-indicative versions of Mooreanabsurdity. It is not clear to me whether non-indicative cases of Moorean absurdityexist. For instance, I do not know whether, ‘Shut the door, but I don’t wantyou to shut the door,’ exemplifies Moorean absurdity. Likewise, I do not knowwhether ‘What time is it, even though I have not the slightest interest in knowingthe time?’, exemplifies Moorean absurdity. Both cases are pragmatically odd, butit would be rash to infer that they exhibit the same sort of oddity that we findin Moore’s cases. However, if either one of these cases does exemplify Mooreanabsurdity, the approach offered in this paper explains why. The explanationwould proceed by observing that if the first conjunct is sincere, then it shows astate of mind that the second conjunct mistakenly disavows; thus whether or notthe first conjunct is sincere, the speaker is in violation of some system of norms,and she is in violation of such norms in a way that is open to inspection with nofurther empirical information.

10. NON-ILLOCUTIONARY CASES

We glossed Moorean absurdity as any utterance or thought in which an agentovertly expresses an intentional state that she also explicitly disavows; or anyutterance or thought in which an agent overtly expresses an intentional statewhose content is incompatible with that of another intentional state that she alsoexplicitly avows. This account does not strictly require that the agent in questionuse words, even in the privacy of her own thoughts. That raises the questionwhether we find Moorean absurdity in cases in which an agent scowlingly deniesthat she is angry, or exuberantly avows her lack of exuberance.

Just as we have found cases of speaker meaning that do not require utteringany words, a speech act does not require any act of speech. For instance, anextended finger at an auction is a promise to purchase the item at the amount bid

²⁰ In (8) of course, we also have a semantic contradiction, but we would need some reason tothink that absurdity is ‘additive’ before predicting that (8) ought to sound ‘more’ absurd than othercases that we have considered.


on the condition that no one else outbids that offer. Pulling down one’s lowereyelid in some Mediterranean cultures is a warning. So too, nodding one’s headis an assertion under the right conditions. That is why it can be the vehicle ofMoorean absurdity. As Williams (forthcoming) observes, if you ask me whetherthe pubs are open and I nod my head in emphatic agreement while saying, ‘Idon’t believe so,’ this seems no less absurd than one of the standard cases 1 or 2above. This and similar cases involving nonverbal speech acts can be explainedin terms of the line of thought we have developed thus far: A nodded head, forinstance, shows one’s acceptance of some salient proposition if in fact that head’sowner is sincere; together with her subsequent utterance we may infer with nofurther empirical investigation that one is either in error or not sincere.

Other cases involving nonverbal behavior are quite different. When an agentscowlingly denies that she is angry, or exuberantly avows her lack of exuberance,she might be in error only. She might simply be unaware of the fact that she isscowling or that she is behaving exuberantly. The scowl might not be one that shehas noticed, and she might not be conscious of the exuberant behavior. In thosecases, her disavowal or avowal is simply a mistake of fact and so is not absurd: Itis not the case that with no further empirical investigation she could concludethat she is in violation of a system of norms. If we find such cases amusing it isbecause it is easy to be amused at people who are blind to their own emotionaldisplays. Likewise, one’s companion on a midnight walk through a cemetery whotremblingly says ‘Not scary at all,’ might be unaware of the tremors in her voice,and might be unable to detect her own fear without empirical investigation.Given these possibilities, we cannot conclude from these performances that theagent is behaving absurdly, or is in some other way absurd.

The emotional expressions in these last cases are not intentional, to say nothingof overt. Might an agent overtly express an intentional state without conventionaldevices such as words or gestures? If so, that would suggest that a case of Mooreanabsurdity might be found in which an agent behaves both expressively and overtly(rather than performing a speech act) while disavowing what she expresses. Tothat end, imagine that I not only scowl, but overtly do so: according to the glossgiven above, it would be sufficient to achieve this result if I not only intentionallydisplay my anger, but also intend to display this very intention. In such a caseit would be natural to describe me as scowling significantly. In fact, in such acase it is also natural to describe me as speaker-meaning that I am angry. Thatdoes not fall under our sufficient condition for propositional speaker meaningas given in Section 3 above. The reason is that condition applies when an agentmanifests her commitment (to a proposition, state of affairs, etc.). However, itis not clear that I am committing myself to anything in scowling, even overtly.Instead, what I am doing in overtly scowling is displaying my anger—makingthat anger perceptible rather than merely giving evidence for the presence of thatanger. (This is suggested by the fact that if I am dissimulating, not showing myanger but merely seeming to do so, I may be misleading or mendacious but no

Mitchell Green 213

liar.) In this case, I am showing my anger in the showing-α way discussed inSection 8, rather than merely showing that I am angry.

Speech acts are not the only vehicles by which we express ourselves. They arealso not the only way in which speaker meaning is achieved. In a case in whichwe intentionally and overtly show our intentional state, we speaker-mean thatstate without performing a speech act, and we express that intentional state aswell. Suppose that I scowl in such a way as not only to express my anger, but todo so overtly. If I could at the same time deny that I am angry, then we wouldhave the makings of Moorean absurdity. However, it is not clear that I could doboth these things at once. For in light of what I literally say when I deny thatI am angry, it is hard to see how an interpreter could sensibly construe me asintending overtly to display my anger. My literal utterance will put pressure onthe interpreter either to construe my facial behavior as inadvertent or covert; orat least as facetious. That is why, when it might seem as if a case of Mooreanabsurdity involving nonverbal, non-conventional expressive behavior is in theoffing, the best we may be able to do is either to describe the agent as protestingtoo much or as hamming it up.

REFERENCES

Anselm (1078 [1995]), Monologion and Proslogion, ed. T. Williams (Indianapolis:Hackett).

Collins, A. (1996), ‘Moore’s Paradox and Epistemic Risk’, Philosophical Quarterly, 46:308–19.

Cargile, J. (1967), ‘On Believing You Believe’, Analysis, 27: 177–83.Davis, W. (2003), Meaning, Expression and Thought (Cambridge: Cambridge University

Press).Green, M., (1999a), ‘Attitude Ascription’s Affinity to Measurement’, International Journal

of Philosophical Studies, 7: 323–48.(1999b), ‘Illocutions, Implicata, and What a Conversation Requires’, Pragmatics

& Cognition, 7: 65–92.(1999c), ‘Moore’s Many Paradoxes’, Philosophical Papers, 28: 97–109.(2000a), ‘The Status of Supposition’, Nous, 34: 376–99.(2000b), ‘Illocutionary Force and Semantic Content’, Linguistics & Philosophy, 23:

435–73.(2003), ‘Grice’s Frown: On Meaning and Expression’, in G. Meggle and C. Plunze

(eds.), Saying, Meaning, Implicating (Leipzig: University of Leipzig Press), 200–19.(forthcoming), Self-Expression (Oxford: Oxford University Press).

Grice, P. (1941), ‘Personal Identity’, Mind, 50: 330–50.(1957), ‘Meaning’, Philosophical Review, 66: 377–88; repr. in Grice (1989).(1969), ‘Utterer’s Meaning and Intentions’, Philosophical Review, 78: 147–77;

repr. in Grice (1989).(1982), ‘Meaning Revisited’, in Mutual Knowledge, ed. by N. Smith (New York:

Academic Press), 223–43; repr. in Grice (1989).(1989), Studies in the Way of Words (Cambridge, Mass.: Harvard University Press).


Heal, J. (1977), ‘Insincerity and Commands’, Proceedings of the Aristotelian Society, 77:183–202.

(1994), ‘Moore’s Paradox: A Wittgensteinian Approach’, Mind, 103: 5–24.(2003), Mind, Reason, and Imagination (Cambridge: Cambridge University Press).

Kriegel, U. (2004), ‘Moore’s Paradox and the Structure of Conscious Belief ’, Erkenntnis,61: 99–121.

Nagel, T. (1979), ‘The Absurd’, in Mortal Questions (Cambridge: Cambridge UniversityPress).

Rosenthal, D. (1998), ‘Thinking That One Thinks’, in A. Burri (ed.), Language andThought (Berlin: De Gruyter), 259–87.

Searle, J., and Vanderveken, D. (1985), Foundations of Illocutionary Logic (Cambridge:Cambridge University Press).

Shoemaker, S. (1988), ‘On Knowing One’s Own Mind’, Philosophical Perspectives, 2:183–209.

(1995), ‘Moore’s Paradox and Self-Knowledge’, Philosophical Studies, (77): 211–28.(1996), The First Person Perspective and Other Essays (Cambridge: Cambridge

University Press).Siebel, M. (2003), ‘Illocutionary Force and Attitude Expression’, Linguishes and Philo-

sophy, 26: 351: 66.Sorensen, R. (1988), Blindspots (Oxford: Clarendon Press).Strawson, P. (1964), ‘Intention and Convention in Speech Acts’, Philosophical Review,

73: 439–60.Szabo-Gendler, Z. (2001), ‘Fictionalism and Moore’s Paradox’, Canadian Journal of

Philosophy, 31: 293–308.Tversky, A. (1975), ‘A Critique of Expected Utility Theory: Descriptive and Normative

Considerations’, Erkenntnis, 9: 163–73.and Kahneman, D. (1992), ‘Advances in Prospect Theory: Cumulative Represent-

ations of Uncertainty’, Journal of Risk and Uncertainty, 5: 297–323.Wakker, P. and Tversky, A. (1993), ‘An Axiomatization of Cumulative Prospect Theory’,

Journal of Risk and Uncertainty, 7: 147–76.Williams, J. (forthcoming), ‘Wittgenstein, Moorean Absurdity and its Disappearance

from Speech’, Synthese.Wilson, T. (2002), Strangers to Ourselves: Understanding the Adaptive Unconscious (Cam-

bridge, Mass.: Harvard University Press).

PART V

ARGUMENTS FROM MOORE’SPARADOX

10

My Philosophical Position Says �p�

and I Don’t Believe �p�

Alan Hájek

INTRODUCTION

There is typically something strange about asserting or believing Moore-paradoxical sentences. In this paper I want to harness this strangeness todo a particular kind of philosophical work. I will argue that various prominentphilosophers are committed to asserting and believing various Moore-paradoxicalsentences in virtue of the very philosophical positions that they hold. Some ofthe philosophers in question may be surprised to learn of their commitment andfind it unwelcome; others may not be troubled by their commitment and claimthat not all Moore sentences are paradoxical after all; still others may positivelycelebrate their commitment and the paradoxicality, perhaps in the name ofnotoriety or boldness of thought. In any case, observing the commitment mayhelp lay bare what we find peculiar in these philosophical positions, wherepreviously we may only have had vague feelings of unease; or we may regard theMoore-paradoxical commitments as further reductios of the positions; or we mayconclude that not all Moore sentences are paradoxical after all. Whichever waythings go, I hope that some philosophical progress will be made. Along the way,I will consider some philosophical positions that may not actually have been heldby any philosopher, prominent or otherwise, but that still have some interestingMoorish consequences.

I thank Jon Kvanvig for helpful discussion at an early stage, and Graham Priest, Roy Sorensen,and especially Andy Egan, Jordi Fernandez, Mitch Green, and John Williams for very astutecomments on earlier drafts.

218 My Philosophical Position

MOORE SENTENCES, AND MORE SENTENCES

(OF THE SAME KIND)

We begin with the canonical versions¹ of Moore sentences—sentences of theform:

(1) p and I believe that not-p.

(2) p and I don’t believe that p.

In very much the same spirit, we have:

(3) p and I believe that ‘p’ is not true.

(4) p and I don’t believe that ‘p’ is true.

In fact there is a spectrum of sentences displaying the same peculiarity to varyingdegrees. For example:

(5) p and I assign ‘p’ low probability.

(6) p and I assign ‘p’ middling probability.

(7) p and I don’t assign ‘p’ high probability.

Along the way we will encounter further sentences that are similarly anomalous.The simplest way to reveal the varying degrees of oddity of such sentences

is to begin with the idea that in asserting p, one represents oneself as knowingthat p. (Cf. Unger 1975; Slote 1979; De Rose 1991; Williamson 1996.) But thesmaller the probability that one assigns to p, the further one is from believing anda fortiori from knowing that p, and thus the greater is the discrepancy betweenhow one represents oneself by an assertion of p, and how one represents oneselfby an assertion of a probability assignment to p.

Even if we do not accept this rather stringent account of assertion, we cansurely agree that in asserting p, one conveys at least reasonable confidence inp—one conveys at least a moderately high subjective probability assignment top. (Cf. Lewis 1976.) But one sends mixed messages by conveying also that thissubjective probability is low, or middling, or not moderately high. After all,probability functions, even subjective probability functions, are functions: theycannot assign two different values to the same proposition. So there is apparently

¹ There are non-canonical instances of the same puzzling phenomenon—e.g. ‘God knows thatI am an atheist’ (Sorensen 1988). Conversely, there are non-puzzling instances of Moorean-lookingsentences: for example, ‘it is raining, and I don’t believe that there is precipitation’, uttered bysomeone who does not know what the word ‘precipitation’ means. More generally, someone maynot know that the sentence denoted by ‘p’ expresses or is implied by the proposition p. Thus,a sentence’s having the canonical Moorean syntax is neither necessary nor sufficient for its beingMoore-paradoxical. I will not attempt to give an analysis of just what Moore-paradoxicality consistsin. Instead, I will play the Justice Stewart defense, resting content with knowing the phenomenonwhen I see it; moreover, I believe that all the examples I adduce clearly display it. I thank JohnWilliams for discussion on this point.

Alan Hajek 219

no way of interpreting someone who utters (5)–(7) as having a single probabilityassignment for p. And the lower the probability stated in the second conjunct,the harder this task of interpretation becomes. (Note that it is not strange in thesame way to say

p, and ‘p’ is probable,

since both conjuncts send the same message of confidence in p. Strangely wordy,perhaps—but not strange in the same way.)

Thus, if you assert any of (1)–(7), you apparently represent yourself to othersas having a certain attitude to the world, which you then undermine or contradictwith the representation of another attitude. But Moore’s paradox is as mucha puzzle for belief as it is for assertion. And it would be equally puzzling torepresent yourself to yourself in any of these ways—by believing any of thesesentences. Moreover, it should seem puzzling for you to believe something thatentails any such sentence—thus implicitly committing yourself to the truth ofthe sentence—if the entailment is easily recognized.

The lore has it that asserting or believing Moore-paradoxical sentences isproblematic. And yet a number of philosophers are implicitly committed todoing both in virtue of the philosophical positions that they espouse: thingsthat they explicitly say entail instances of (1)–(7), where the entailments areeasily recognized. If we alert them to their commitment to Moore-paradoxicalsentences, they should either assert and believe these sentences—which many,following Moore, find ‘absurd’—or rethink their philosophical positions.

SOME PHILOSOPHICAL POSITIONS WITH

MOORE-PARADOXICAL CONSEQUENCES

‘There are no beliefs’

Churchland (1981) and Stich (1983) are skeptical about the very notion of belief.‘Belief ’ is part of a suspect folk psychology, likely to go the way of phlogistonand vital spirits—that is, ultimately to be discarded by science. On this view,neither I, nor you, nor anybody else ever has, ever had, or ever will have beliefs.Rather, we have whatever mental items will be postulated by a fully developedpsychological theory. Thus, Churchland and Stich are committed to utteringsentences such as:

‘It is raining and I don’t believe that it is raining (and neither do you, noranybody else, for ‘‘belief ’’ is part of a suspect theory of the mental)’.

We have an easy way of generating sentences of type (2).Now, Churchland and Stich would surely not be fazed by this gambit—nor

by the putative reductio that ‘they offer a philosophical position that, by theirown lights, they don’t really believe!’ The right thing for them to say, of course, is


that their mental state regarding the rain, or their own position, is whatever thefully developed psychological theory postulates it to be. Still, their commitmentto Moore sentences is genuine, as is the puzzlement that it may induce in manyof the rest of us. And we see our first trick for generating Moore sentences, onethat we will see again: subject belief to different standards from those for (sincere,warranted) assertion. Indeed, according to Churchland’s and Stich’s standards forbelief, nothing meets them: nobody ever had, ever has, or ever will have a belief.Yet presumably their standards for assertion are the usual ones. Thus, on theirview, one may succeed in asserting something (sincerely and with warrant) whilelacking a belief in that thing.²

Now suppose that you are not totally convinced by Churchland and Stich, butyou think that it is at least an open question—something else close to Moore’sheart—whether or not they are right. It seems that you are committed tosentences such as:

‘It is raining, and it is an open question whether I believe it is raining (andwhether you do, and whether anybody else does, for it is an open questionwhether Churchland and Stich are right)’

and perhaps:

‘It is raining, and I am agnostic about whether I believe it is raining . . .’

These, too, sound pretty Moorish. If you want to avoid such a predicament,you should not be even agnostic about whether Churchland and Stich are right.Not that they would welcome this talk of agnosticism either, since presumablyby their lights it is another folk psychological notion to be jettisoned. So if youreally want to keep an open mind regarding their position, perhaps even the lattersentence is too committal. Better to play it safe:

‘It is raining, and it is an open question whether I am agnostic aboutwhether I believe it is raining . . .’

And so on.Churchland’s and Stich’s position is programmatic; they await the details of a

fully developed psychological theory. Let me go out on a limb and suggest oneway that things could conceivably go. Psychology could adopt wholesale the termsof Bayesianism, eschewing talk of beliefs in favor of subjective probabilities. Thereis already this tendency in the work of Jeffrey (e.g. 1968), who seeks to replacethe concept of knowledge with that of subjective probability, thus downplaying

² I say ‘presumably’ because I assume that even qua eliminativists, Churchland and Stich feelfree to assert things very much as the rest of us do—after all, even their philosophical works oneliminativism are full of assertions, any one of which I could use to make my point. To be sure,assertion is usually characterized in terms of its role as the (purported) expression of belief, orin similar mentalistic terms. An eliminativist account of assertion would presumably look ratherdifferent. The mentalistic adjective ‘sincerely’ may similarly require an eliminativist gloss. (Thankshere to an anonymous referee for this volume.)

Alan Hajek 221

skeptical concerns about knowledge. Pushing this further, a Jeffrey-inspiredpsychologist might insist that ‘belief ’ cannot simply be reduced to subjectiveprobability (citing, perhaps, the lottery and preface paradoxes, of which moreshortly), and that subjective probability is the proper doxastic notion: folkpsychology should be eliminated in favor of Bayesianism. Then speaking in hiscapacity as eliminativist, the psychologist may well say:

‘It is raining, and I don’t believe that it is raining (instead, I assign highsubjective probability to it raining).’

‘Beliefs are propositions assigned subjective probability 1’ or ‘. . . veryhigh subjective probability’

On the Churchland/Stich/Jeffrey-inspired psychologist view, beliefs are muchharder to come by than you might think (since earning their keep in a maturepsychological theory is a tough standard to meet)—so much so that there simplyaren’t any. But we need not adopt positions as radical as theirs in order togenerate Moore sentences. It suffices to adopt unusually high standards for beliefwhile keeping normal standards for sincere, warranted assertion. Thus, we mightbe less demanding than Churchland, Stich, and the Jeffrey-inspired psychologistabout the notion of belief, but demanding enough. For example, Hawthorneand Weatherson (2004) argue that ‘S believes that p’ should be analyzed as ‘Sassigns subjective probability 1 to p’. Beliefs do exist on this view, but still theyare harder to come by than you might think. If sincere, warranted assertionsremain as easy to come by as you think, we can find propositions that make thecut for assertion, but that do not make the cut for belief, so understood. For sucha proposition p, presumably S may properly assert p while disavowing belief in it.

We might be moved by the lottery paradox,³ for example, to say that nothreshold of subjective probability below 1 is sufficiently high to count as belief:you don’t really believe that your ticket will lose, you merely assign it probability0.999999 of doing so. As long as a lower threshold suffices for (sincere, warranted)assertion, the conditions for Moore sentences are in place. Thus, it apparentlybecomes reasonable to say:

‘My ticket will lose (I assert this because my probability is above thethreshold for assertion), but I don’t believe that my ticket will lose (sincemy probability falls below the threshold of 1 for belief)’.

We may lower the standard for belief while arguably maintaining somedaylight between the new standard and that for warranted assertion—and thisstill suffices to generate Moore-paradoxical sentences. Suppose that we set thebar for belief not at probability 1, as Hawthorne and Weatherson did, but at

³ Kyburg (1961).


0.999. Plausibly, various sentences will still clear the bar for assertability, but notthe bar for belief. As it might be: ‘the pubs are open (I am entitled to assert thissince my credence in it is sufficiently high), but I don’t believe that the pubs areopen (since that credence does not quite reach 0.999)’.

The preface paradox⁴ furnishes an example of a related phenomenon, this timeproducing commissive Moore sentences of type (1). You preface a long book thatyou have written with the modest words: ‘Despite all my efforts, I am sure thatthere is at least one mistake somewhere in the book’. But the book itself can beregarded as a very long assertion—the conjunction of many individual assertions.Indeed, suppose that we replace all periods but the final one throughout the bookwith ‘and’s.⁵ You should be as committed to this unwieldy sentence as you wereto the original book (even if you have acquired some stylistic qualms about it).Conjoining the unwieldy sentence to your preface, we have an assertion of theform:

[MY BOOK] (the unwieldy sentence), and I believe that �MY BOOK� isfalse (since I believe that at least some conjunct in it is false).

Schematically, we have the dreaded

p, and I believe that �p� is false.

The trouble is that assertion is an on/off, all-or-nothing act, whereas degreesof belief come in degrees. We do not have devices for giving assertion all ofthe nuance that we might want—say, boldness of typeface that varies with thestrength of our convictions. Imagine that with each ‘and’ that we insert betweensuccessive sentences, their print fades accordingly; by the time we have conjoinedthem all, they become invisible!

In a way, all of us—and not just certain idiosyncratic philosophers—findourselves in the same uncomfortable shoes as the modest preface-writer, for onpain of gross immodesty, we all admit that some of our beliefs are false. Each ofus would thus assert something of the form:

[LONG CONJUNCTION OF MY BELIEFS] and I believe that �LONGCONJUNCTION OF MY BELIEFS� is false.

Note that this does not require you to provide a complete enumeration of allyour beliefs—an impossible task, surely. It suffices that you can find some longconjunction of beliefs of yours that you believe is false—an easy task, surely.For example, a few dozen beliefs of yours about the capital cities of countries inthe world, or about phone numbers, or about the names of the children of yourfriends and colleagues, may well do the job.

The conjunction might not even have to be at all long. Consider a recalcitrantparadox that has a grip on you that can be presented as a short list of premises,

⁴ Makinson (1965).⁵ I assume that the book consists solely of sentences that have truth-value—I ignore, for example,

imperatives or questions.

Alan Hajek 223

each of which you believe, but which you recognize to be jointly contradictory.You assert:

[SHORT CONJUNCTION OF THE PREMISES] and I believe that�SHORT CONJUNCTION OF THE PREMISES� is false (since Irecognize them to be jointly contradictory).

Perhaps one of Kant’s antinomies will fit the bill. In fact, soon we will see a singlepremise giving rise to this phenomenon—although only for certain idiosyncraticphilosophers.

‘There are no higher-order beliefs’

Suppose that you are not skeptical about beliefs, the way that Churchland andStich are, but that you are skeptical about higher-order beliefs. So you are happyto speak of beliefs (‘I believe that it is raining’, and so on) but you have no truckwith beliefs about one’s own beliefs (‘I believe that I believe that it is raining’,and so on). Now take a belief of yours—say, that it is raining. Then you shouldbe prepared to assert:

‘I believe that it is raining, and I don’t believe that I believe that it is raining(for that would be a higher-order belief, with which I have no truck)’

—something of the form ‘p, and I don’t believe that p’.Now suppose that you are not totally skeptical about whether there are higher-

order beliefs, but you think that it is at least an open question. Then it seems thatyou are committed to sentences such as:

‘I believe that it is raining, and it is an open question whether I believe thatI believe that it is raining’

—something of the form ‘p, and it is an open question whether I believe that p’.

‘There are only so many higher-orders that beliefs can reach’

Less radically, you might allow second-order, third-order, and perhaps stillhigher-order beliefs, but insist that the hierarchy must stop somewhere. Youmight say, for example, that the finiteness of our heads imposes limits on justhow many times the belief operator can be iterated. (This is a commonplacein the literature on common knowledge in which it is acknowledged that theputative infinite iterations of ‘I know that you know that I know that . . . p’ arean idealization.⁶ See also Sorensen 2000.) Suppose, then, that you think that youhave a highest order of belief: an nth order belief, for some n > 2. Find, if youcan, an nth order belief of yours—say, that it is raining. Then you should beprepared to assert:

⁶ I thank Mitch Green for this point.


‘I nth-order-believe that it is raining, but I don’t believe that I nth-order-believe that it is raining (for that would be an (n+1)th order belief, whichexceeds my limit)’.

And as before we can open the door to an ‘open question’ version of this.

‘There are no higher-order probabilities’; or ‘There are only so manyhigher-orders that probabilities can reach’

De Finetti (1972) and Savage (1954) are champions of subjective probabilities,but they are skeptical about higher-order subjective probabilities, or probabilitiesconcerning one’s own probabilities. For example, by their lights it is nonsenseto say ‘my probability that the coin lands heads is 1/2, and my probability thatthis really is my probability is 0.99’. De Finetti writes: ‘Any assertion concerningprobabilities of events is merely the expression of somebody’s opinion and notitself an event. There is no meaning, therefore, in asking whether such anassertion is true or false or more or less probable.’ (1972:189). One of theirarguments, roughly, is that any putative second-order probabilities would collapseto ordinary first-order probabilities—a reductio of the idea that there really weresecond-order probabilities in the first place. Both of these authors also seem toregard the threat of infinite regress as fatal to higher-order probabilities. Savage:‘once second order probabilities are introduced, the introduction of an endlesshierarchy seems inescapable. Such a hierarchy seems very difficult to interpret,and it seems at best to make the theory less realistic, not more’ (1954:58). DeFinetti: ‘we have events and probabilities of events only; otherwise we wouldhave the beginning of an infinite regression (probability of a probability, and soon)’ (1972:193). Consider, then, some statement of probability that de Finettior Savage is prepared to make—say, ‘The probability that the coin lands headsis 1/2’. Now conjoin to it their skepticism about the notion of this statementin turn having a probability, and thus a fortiori, skepticism about this statementhaving a high probability:

‘The probability that the coin lands heads is 1/2, and I don’t assign thisclaim high probability (for this claim is not the sort of thing that has aprobability at all)’.

That is, we have a sentence of the form (7).Less radically, you might allow second-order, third-order, and perhaps still

higher-order probabilities, but insist that the hierarchy must stop somewhere(again, perhaps because our heads are finite). Suppose, then, that you think thatyou have a highest-order probability assignment: an nth-order assignment, forsome n > 2. Find, if you can, an nth-order probability assignment of yours—say,that it is raining. Then you should be prepared to assert:

‘My nth-order probability that it is raining is x, but I don’t assign highprobability to the claim that my nth-order probability that it is raining is x

Alan Hajek 225

(for that would be an (n+1)th order probability assignment, which exceedsmy limit).’

Again, we have a sentence of the form (7).

‘There is no such thing as truth’

Paralleling our discussion of Churchland and Stich, now consider a philosopherwho is skeptical of the very notion of truth. Let us suppose that he is prepared toassert sentences as usual, but he balks at assertions of the truth of sentences. Thus,a true⁷ Nietzschean ought to be prepared to assert sentences such as:

‘It is raining, but I believe that ‘it is raining’ is not true (for I reject the verynotion of truth)’,

a sentence of form (3). Similarly:

‘It is raining, but I don’t believe that ‘‘it is raining’’ is true (for I reject thevery notion of truth)’,

a sentence of form (4). Certain French philosophers and certain literary criticshave a similar disdain, disrespect, or disregard for truth. They will pay for itsimilarly in Moorean ways (which is not to say that this is the worst of theirproblems).

If someone were happy with the notion of (first-order) truth, but skeptical ofthe notion of higher-order truth, we could set them some Moorean bait:

‘ ‘‘It is raining’’ is true, but I don’t believe that ‘‘ ‘It is raining’ is true’’ istrue (for that would involve higher-order truth, of which I am skeptical).’

And, much as before, we could also shanghai someone who thinks that therecan be only n levels of higher-order truth, for some n > 2, with an nth-ordertruth-assertion of theirs that they do not believe is true.

A skeptic about subjective probability—perhaps along the lines of Harman(1986)—could presumably be prepared to assert:

‘It is raining, and I don’t assign ‘‘it is raining’’ high probability (or indeedany probability, for I am skeptical about subjective probability)’,

a sentence of form (7).

‘Truth is just what is useful for our purposes’ or ‘what is the consensusof our community’, or somehow ‘is relative’

We have just seen how to drive a wedge between an assertion and a belief inthe truth of the assertion, by being unusually demanding about the notion of

⁷ I can’t resist this jab. Was Nietzsche a true Nietzschean in this respect? He seems to be whenhe writes: ‘Truths are illusions about which one has forgotten that this is what they are’ (Nietzsche1994: 47). I thank Harold Langsam for this reference.


truth. Going in the other direction, we could drive the wedge by being unusuallyundemanding about it. A certain kind of naive pragmatist, for example, mighttranslate talk of truth into talk of what is useful for our purposes. Thus, perhaps:

‘God does not exist, but I believe that ‘‘God exists’’ is true (it being usefulfor our purposes generally to act as if God exists)’.

I dare say that this position is too naive to take seriously. A bit more promising,perhaps, is the position of a rabid communitarian about truth: he thinks that thetruth of a proposition p consists in a consensus of his community that p. He mayfind himself with a case of the Moores when he disagrees with such a consensus.For example:

‘God does not exist, but I believe that ‘‘God exists’’ is true (this being theconsensus of my community).’

A certain kind of Peircean might append the words ‘. . . at the end of inquiry’,with similar results.

Anyone who holds a relativist account of truth should strictly speaking makeno sense of locutions of the form ‘p is true’. For by a relativist’s lights, there isno one-place predicate ‘ is true’ at all, but rather a two-place relation of theform ‘ is true relative to ’. The second argument-place will be filled indifferent ways by different relativists. A Protagorean may relativize an attributionof truth to a person; a Foucaultian may relativize it to a discursive formation; aKuhnian may relativize it to a paradigm; a Quinean may relativize it to a theoryor a language; a MacFarlanean may relativize it to a context of utterance or ofassessment⁸ . . . The upshot is that a relativist should hold that an attributionof truth simpliciter is ill-formed, and thus not something that can properly bebelieved. Much as strictly speaking it is nonsensical to believe or to assert ‘Reneeis younger’ or ‘Los Angeles is west’, so it is nonsensical to believe or assert ‘pis true’, according to the relativist. And yet relativists seem to have no troublebelieving or asserting things in the normal way (some of them, anyway). Thus,Moorean sentences should glide off their tongues:

‘It is raining, and I don’t believe that ‘‘it is raining’’ is true (being nonsensical,as it lacks a needed second relatum).’

To be sure, context often makes missing relata clear. In a conversation in which Iam comparing my age to various people, you will have no trouble understandingme when I say ‘Renee is younger’, because it is obviously elliptical for ‘Reneeis younger than me’. Likewise, if we are both standing in New York, I maypermissibly say ‘Los Angeles is west’, leaving tacit the relativization to here,understood by both of us to be New York. That does not show that ‘younger’or ‘west’ may suddenly become one-place predicates. Rather, context simply

⁸ See MacFarlane (2005).

Alan Hajek 227

makes obvious what the second argument-place is. So to be sure, context maymake obvious which person, or discursive formation, or paradigm, or theory orlanguage, or context of utterance or of assessment . . . is salient. Then a relativistmay permissibly say that something is true, it being understood to what thisattribution of truth is being relativized. But this only shows that sometimesit is permissible not to speak strictly. And context does not always make clearthe missing relatum. When it does not, and the empty argument-place remainsunfilled, the result is nonsense. Or so should say the relativist—about ‘younger’,or ‘west’ (as we all are), and about truth (as are the Protagorean, the Foucaultian,the Kuhnian, the Quinean, the MacFarlanean . . .).

‘There are truth gluts’

We already saw with the preface paradox that someone can be prepared to assertsomething while believing that it is false. There, at least, the Moore-paradoxicalitywas in a sense diffused over a long conjunction of beliefs, with no individualconjunct being simultaneously asserted and believed to be false. But there arecases where the Moore-paradoxicality is far more localized—indeed, to a singlesentence.

Faced with stubborn paradoxes like those generated by ‘liar’ sentences (‘thissentence is false’, and its brethren), some philosophers embrace the existence oftruth gluts—sentences that are both true and false. Priest (1987) is perhaps mostfamous for this view. Presumably, then, Priest will not bat an eye at saying:

‘The liar sentence is true, and I believe that the liar sentence is false (andtrue)’.

Now, I suspect that he would hardly be troubled biting the bullet of asserting aMoore sentence, having already bitten the nuclear bomb of contradiction. Andthose of us who reject dialethism probably recoil at the very notion of therebeing true contradictions. Thus, you might call my drawing attention to theseMoore-paradoxical commitments overkill. I prefer to call them icing on the cake.

An approach to vagueness, common to both supervaluating and subvaluating,considers all the various permissible ways of precisifying a vague predicate.(See Hyde 1997.) Subvaluating treats a statement as true simpliciter if it istrue according to any such precisification. (We will consider supervaluatingshortly.) Suppose that Bruce is a borderline case of ‘tall’. Then there are someprecisifications of ‘tall’ according to which he is tall, and others according towhich he is not tall. The subvaluator then adjudicates him to be both tall and nottall. Moore awaits (as well as the nuclear bomb):

‘Bruce is tall (and not tall), and I believe that Bruce is not tall (and tall)’,

or deleting the parenthetical conjuncts,

‘Bruce is tall, and I believe that Bruce is not tall.’


‘There are truth gaps’

Another non-classical reaction to the ‘liar’ sentences has it that they are truthgaps—sentences that lack truth-value altogether. Various philosophers considervarious other sentences to be truth gaps. Expressivists about moral discourse, suchas Ayer (1946), regard moral claims such as ‘murder is wrong’ as expressionsof disapproval, but not truth-apt. Such an expressivist, then, should happily saythings of the form:

‘Murder is wrong (as I am wont to express), and I don’t believe that ‘‘murderis wrong’’ is true (since the sentence is not truth-apt)’.

To be sure, ‘murder is wrong’ may not be an assertion according to expressivists(any more than ‘Boo to murder!’ is). Still, there is no denying that our examplefits the Moorean template, and that expressivists should be prepared to utter it.

Wittgenstein apparently has a similarly expressivist view about certain mentalstate reports, such as being in pain: ‘To say ‘‘I have a pain’’ is no more a statementabout a particular person than moaning is’ (1958:67). Then he should not flinchat saying:

‘I have a pain, and I don’t believe that ‘‘I have a pain’’ is true’.

Adams (1975), Edgington (1995), and Bennett (2003) argue that conditionalslack truth-values although they may well be assertible. They should thus beprepared to say things like:

‘If it’s raining then the ground is wet, and I don’t believe that ‘‘if it’s rainingthen the ground is wet’’ is true (since the sentence lacks a truth-value)’.

Some authors believe that vague predicates generate truth gaps. On this view,if Bruce is a borderline case of ‘tall’, then the sentence ‘Bruce is tall’ lacks atruth-value. In particular, this sentence is not true (as well as being not false).Now suppose that it is permissible, although of course not required, to believe asentence that one takes to be a truth gap. Then this has Moorish consequences.As it might be:

‘I believe (permissibly) that Bruce is tall, and ‘‘Bruce is tall’’ is not true(being a truth gap).’

Consider, for instance, the supervaluational approach to vagueness, whichidentifies truth simpliciter with truth according to all permissible precisifications,so-called ‘super-truth’. Suppose that ‘Bruce is tall’ is true according to somepermissible precisifications (as must be the case, on this approach, if Bruceis a borderline case). Then arguably my belief can permissibly follow theseprecisifications. Thus:

‘I believe that Bruce is tall (permissibly, since he is tall on some admissibleprecisfications), and ‘‘Bruce is tall’’ is not true (since it is not super-true).’

Alan Hajek 229

There are probabilistic analogues of this phenomenon. Suppose that your stateof opinion cannot be represented by a single probability function—e.g., yourprobability for rain is vague over the interval [0.5, 0.75]. Following Levi (1974),Jeffrey (1983), and van Fraassen (1990), we may represent you with a set ofprobability functions, each of which precisifies your probability for rain withsome sharp number in the interval. What is determinately true of your opinionis agreed upon by all the functions in this set. It is determinately true in theexample that your probability for rain is at least 1/2. On the other hand, anystatement about your opinion that is true according to some functions in theset and false according to others is indeterminate, and thus not determinatelytrue—e.g. that your probability for rain is at least 0.6.

Now suppose that there is a proposition p for which your probabilityassignment is vague over a wide interval—wide enough that according to theleft-hand endpoint, p is not probable, while on the basis of the right-handendpoint, it is assertable. For example, suppose that your probability of therebeing life on Mars is vague over the interval [0.4, 1]. According to somepermissible precisifications of your opinion (e.g. 0.4), you do not assign highprobability to ‘there is life on Mars’; thus you do not determinately assign highprobability to ‘there is life on Mars’. According to others (e.g. 1), ‘there is life onMars’ is assertible; thus, it is permissible for you to assert ‘there is life on Mars’.Conjoining these facts, we now get you to assert:

‘There is life on Mars (I permissibly assert), and I don’t (determinately)assign ‘‘there is life on Mars’’ high probability.’

We arrive at a sentence of form (7) once we drop the parenthetical reminders ofhow we got there.

I said at the outset that there is apparently no way of interpreting someonewho utters (7) as having a single probability assignment for p. I am offeringa way of interpreting such a person if we may ascribe to them vague opinion,represented as multiple probability assignments for p.

So far I have tried to impale various specific philosophical positions on Moorishsentences, suggesting that even if the philosophers who propound these positionsdon’t feel any discomfort there, the rest of us may well do so. But let us not gettoo smug. Perhaps I should not single out these philosophers. Perhaps they arein good company.

A PHILOSOPHICAL PESSIMISTIC META-INDUCTION

Philosophy is a strange business. We have strong incentives to assert things,preferably in print (and the higher the profile of that print, the better), whereour commitment to them is made even more manifest. This much is commonto any intellectual discipline. But unlike many other disciplines, there is good


reason to think that much that is asserted in philosophy is simply false. Moreover,we philosophers know this. We need merely remind ourselves how muchphilosophers disagree; at most one party to a disagreement can be right (assumingthat the disagreement is genuine and the parties are not merely talking past eachother, and ignoring dialethism). Even prior to such disagreement, we mustacknowledge that philosophy is a subtle business, and that saying somethingphilosophically interesting and true is no mean feat; yet what is philosophicallyuninteresting is less likely to make it to print (still less high-profile print), so wehave incentives to stick our necks out. And whatever the explanation, there issurely the brute historical fact that many, and perhaps even most, substantivephilosophical positions that have been offered are false. (Indeed, some of thepositions, believe it or not, even have Moore-paradoxical commitments!) Forevery important philosophical position that you claim is true, I will respond withten such positions that we agree are false.⁹ But I’m sure you don’t need me—youcan do it yourself.

Laudan (1981) offers a ‘pessimistic meta-induction’ concerning the truthof scientific theories, on the basis of the historical track record of science.A philosopher like you may well likewise run a pessimistic meta-inductionconcerning the truth of philosophical theories, on the basis of the historical trackrecord of philosophy. And now here you are, advancing your own ambitious, boldphilosophical position: p. You assert it vigorously, you defend it in (high-profile)print, and so on. But do you really believe it? Suppose that you must bet at highstakes on p’s truth, and that God will settle the matter. Still feeling confident init? You may not think, after all, that you are that much more reliable than variousphilosophers who have come before you whose equally vigorous assertions havenot withstood the test of time. And so your standards for assertion and for beliefare sundered. You publicly assert p, maybe even in (high-profile) print, but ifyou are honest with yourself, you admit that you believe that p is probably false,or that you are at best agnostic about p; in any case you don’t believe that p. Atsuch a reflective and reflexive moment, then, you may catch yourself assertingsotto voce: ‘My philosophical position says ‘‘p’’ and I don’t believe ‘‘p’’ ’. Not thatyou will ever say that in print.

REFERENCES

Adams, Ernest (1975), The Logic of Conditionals (Dordrecht: Reidel).Ayer, A. J. (1946), Language, Truth, and Logic (Middlesex: Penguin Books).Bennett, Jonathan (2003), Conditionals (Oxford: Oxford University Press).Churchland, P. M. (1981), ‘Eliminative Materialism and the Propositional Attitudes’,

Journal of Philosophy, 78/2: 67–90.

⁹ Lewis’s list of philosophical positions with dubious credentials (1991:59) is a good start; seealso Stove (1991), throughout.

Alan Hajek 231

Edgington, Dorothy (1995), ‘On Conditionals’, Mind, 104/414:235–329.De Finetti, Bruno (1972), Probability, Induction, and Statistics (New York: John Wiley).De Rose, Keith (1991), ‘Epistemic Possibilities’, Philosophical Review 100/4:581–605.Harman, Gilbert (1986), Change in View (Cambridge, Mass.: MIT Press).Hawthorne, John, and Weatherson, Brian (2004), ‘Beliefs Old and New’, http://brian.

weatherson.net/belief.pdfHyde, Dominic (1997), ‘From Heaps and Gaps to Heaps of Gluts’, Mind 106/424:

641–60.Jeffrey, R. C. (1968), ‘Probable Knowledge’, in I. Lakatos (ed.), The Problem of Inductive

Logic (Amsterdam: North-Holland Publishing Company), 166–80; repr. in Probabilityand the Art of Judgment (Cambridge: Cambridge Studies in Probability, Induction,and Decision Theory, 1992).

(1983), ‘Bayesian with a Human Face’, in J. Earman (ed.), Testing Scientific Theories(Minnesota Studies in the Philosophy of Science, X; Minneapolis: University ofMinnesota Press), 133–56; repr. in Probability and the Art of Judgment (Cambridge:Cambridge Studies in Probability, Induction, and Decision Theory, 1992).

Kyburg, Henry (1961), Probability and the Logic of Rational Belief (Middletown: WesleyanUniversity Press).

Laudan, Larry (1981), ‘A Confutation of Convergent Realism’, Philosophy of Science,48/1: 19–49.

Levi, Isaac (1974), ‘On Indeterminate Probabilities’, Journal of Philosophy 71/15:391–418.

Lewis, David (1976), ‘Probabilities of Conditionals and Conditional Probabilities’,Philosophical Review, 85/3:297–315.

(1991), Parts of Classes (Oxford: Blackwell).MacFarlane, John (2005), ‘Making Sense of Relative Truth’, Proceedings of the Aristotelian

Society, 105/3:321–39.Makinson, D. C. (1965), ‘The Paradox of the Preface’, Analysis, 25/6:205–7.Nietzsche, Friedrich (1994), The Portable Nietzsche, trans. Walter Kaufmann (Har-

mondsworth: Penguin Classics).Priest, Graham (1987), In Contradiction (Dordrecht: Martinus Nijhoff).Savage, Leonard (1954), The Foundations of Statistics (New York: John Wiley).Slote, Michael (1979), ‘Assertion and Belief ’, in J. Dancy (ed.), Papers on Language and

Logic (Keele: Keele University Library).Sorensen, Roy (1988), Blindspots (Oxford: Clarendon Press).

(2000), ‘Moore’s Problem With Iterated Belief ’, Philosophical Quarterly, 50/198:28–43.

Stich, Stephen (1983), From Folk Psychology to Cognitive Science (Cambridge, Mass.: MITPress).

Stove, David (1991), The Plato Cult and Other Philosophical Follies (Oxford: Blackwell).Unger, Peter (1975), Ignorance: A Case for Scepticism (Oxford: Oxford University Press).Van Fraassen, Bas (1990), ‘Figures in a Probability Landscape’, in J. M. Dunn and

A. Gupta (eds.), Truth or Consequences (Dordrecht: Kluwer).Williamson, Timothy (1996), ‘Knowing and Asserting’, Philosophical Review, 105/4:

489–523.Wittgenstein, Ludwig (1958), The Blue and Brown Books (New York: Harper and Row).

http://brian.weatherson.net/belief.pdf

http://brian.weatherson.net/belief.pdf

11

Moorean Pretense

Robert M. Gordon

Wittgenstein noted that the absurdity of assertions of the form,

p but I don’t believe that p,

does not carry over to suppositions. He was right. The absurdity also doesnot carry over to what I take to be the more inclusive category, pretense.¹ Forexample, you can pretend—or suppose or imagine—that a certain species ofmushroom is poisonous but that you are ignorant of this fact. You might evenpicture yourself innocently sauteing your mushrooms, then eating them andgetting very sick. Or you can imagine walking down the crowded aisle of abusy department store, not looking where you are going, and walking into asupporting column in the middle of the aisle. You imagine: I am walking rightinto the column, but I don’t believe I am.

By a Moorean Pretense, I mean a pretense that has at least one premise of eitherthe ‘omissive’ form already mentioned,

p but I don’t believe that p,

or the ‘commissive’ form,

p but I believe that not-p.²

I speak of the premises of a pretense because pretense is plausibly construedas having an inferential structure. For example, when young children pretendthat certain globs of mud are cherry pies, they are typically able to use thatidentification, together with additional information drawn from observation,

The author is much indebted to Mitchell Green and John Williams for comments on an earlierversion of this paper.

¹ The intuition behind calling it more inclusive is roughly this: When S pretends that p, Sintentionally acts in some way as if p. The action may be limited to arguing (e.g., ‘supposing forthe sake of argument’). There is of course also a cognitive or epistemic condition: perhaps, ‘S doesnot believe that p’ or perhaps only, ‘S is not certain that p.’ For example, during an athletic event,athletes may find it useful to pretend they are executing a perfect performance. They may wellbelieve they are doing so—they are just not sure.

² See J. N. Williams, ‘Moore’s Paradox—One or Two ?’, Analysis, 39/3 (1979), 141–2 andChapter 1 in this volume.

Robert M. Gordon 233

to answer such questions as, ‘How many cherry pies are there?’ and, ‘Whichcherry pie is biggest?’³ To answer the ‘How many’ question, for example, therelevant inference might be, ‘Each mudpie is a cherry pie [pretended premise],there are three mudpies [empirical finding], therefore there are three cherry pies[pretense-bound conclusion].’

1 . THE INDEXICAL CHARACTER OF MOOREAN

ABSURDITY

Before discussing pretense, it is important to be clear about the indexical characterof Moorean absurdity. The belief clauses,

I don’t believe that pI believe that not-p

contain the first-person indexical ‘I’ and implicitly the temporal indexical ‘now’,signified by the present tense. For an assertion to be Moorean-absurd, the beliefclause must contain these or relevantly similar indexicals. Consider a contrastingcase: You are walking down a crowded aisle, peering intently at a video monitoras you go. On the monitor you see people walking down a crowded aisle, andyou presume correctly that you are watching a live real-time video of the aisleyou are in. Suddenly you point to the monitor and remark,

That person doesn’t know it, but (s)he is about to walk into a column!

Or, to force the point, suppose you say,

That person is about to walk into a column, but (s)he doesn’t believe(s)he is!

If I happen to know that in fact you are the person in the monitor, I may findyour remark comically absurd in context, but I do not find it Moorean-absurd.There is nothing inconsistent or self-stultifying about asserting,

p but that person doesn’t believe that p.⁴

Although the person you are referring to as not believing what you have juststated is yourself, you are not referring to that person indexically as yourself.

The corresponding point holds for the temporal indexical. Suppose yourecognize yourself in the video but do not appreciate that it is a live video:

I didn’t know it at the time, but I was about to walk into a column.

I was about to walk into a column, but I didn’t believe I was.

³ See K. L. Walton, ‘Pictures and Make-Believe’, Philosophical Review, 82/3 (1973), 283–319.⁴ Even referring to yourself by name would be insufficient for Moorean absurdity, for it would

leave open the possibility that you do not know that the name refers to you.

234 Moorean Pretense

These are not examples of Moorean absurdity either.⁵Similar indexical requirements hold for Moorean Pretense. A premise of the

form,

p but that person doesn’t believe that p,

does not make a pretense Moorean. Nor can one convert it into a MooreanPretense merely by identifying the person who doesn’t believe that p with oneself,the person who is doing the pretending. For example, I mentally picture someonewho looks like me walking unknowingly into a column, and I think, ‘That’sme I’m picturing walking unknowingly into a column.’ This is not a case ofMoorean Pretense, because the indexical ‘I’ is introduced only in my commenton the pretense, not within the pretense itself. A pretense is Moorean only ifthere is an ‘I’ within the scope of the pretense.

It should be noted that an external identification with the ignorant non-believer is not only not sufficient for Moorean Pretense; it is also not necessary.That is, it is not necessary that the ‘I’ be understood to refer to myself, the veryperson who is doing the pretending. I can pretend to be someone else, a real orfictitious individual other than the person who is doing the pretending. I canalso pretend the time to be other than the time of the pretending. The ‘I’ and‘now’ within the pretense need not be co-referring with the ‘I’ and ‘now’ outsidethe pretense. That is, they need not refer, respectively, to the pretender and thetime of the pretending. What matters is not what is referred to but the mode ofreference, the I-now character of the reference.

2 . THE DIVIDED PRETENDER

Psychologically there is a major difference between pretending,

That person doesn’t believe that p,

and pretending,

⁵ A curious problem arises as it dawns on you that the unfortunate person on the monitor isactually you in realtime. Suppose you make this inference:

That person is about to walk into a column, but (s)he doesn’t believe she is!Wait, that’s me!I’m about to walk into a column, but I don’t believe I am. (Or: I’m the one who is about towalk into a column, but doesn’t believe it.)

Would your conclusion be Moorean-absurd? I have some inclination to say no, as long as the beliefdenial is strictly evidence-based:

(Evidently) I don’t believe I am.

(I would be similarly inclined where one’s assertion of the belief conjunct in a ‘Moorean’ sentenceis based solely on neuroscientific evidence, such as results from brain imaging.) This is an unstablesituation, however; once the ‘I and now’ identification is made, people would and should abandonthe belief denial; otherwise, it would lead to absurdity.


I don’t believe that p.

What is the difference? The latter pretense requires that the ignorance or non-belief be represented in a ‘first person’ way, by pretending not to believe that p.Likewise, pretending,

I believe that not-p,

requires pretending to believe that not-p.When I pretend to believe or not to believe something, I thereby modify the

set of ‘facts’ that my decision-making within the pretense will have access to. Forexample, if I pretend to believe that the mushrooms on the plate are poisonous,then within the pretense I might refrain from eating them on the grounds thatthey are poisonous. If I pretend to believe that there is an obstacle in my path,then within the pretense I can plausibly change direction because there is anobstacle in my path. Not only actions, but also emotions, desires, and otherbeliefs will have access to this ‘fact’. If on the other hand I pretend not to believethat p, then I will not do anything, for example, change directions, on the basisof the ‘fact’ that p; nor will I form emotions, desires, or further beliefs on thatbasis. In short, one pretends that

I believe (do not believe) that p

only if one pretends to believe (not believe) that p; and pretending to believe (notbelieve) something constrains the facts on the basis of which one can act, emote,or make inferences within the pretense.

Such constraints pose a psychological problem for Moorean Pretense. If onepretends

p but it I don’t believe that p,

then one pretends that p in a special way: a way that makes the pretend ‘fact’that p unavailable to the ‘ignorant’ agent one plays. The pretend fact that p mustsomehow be cordoned off so that it cannot move one to action or emotion or leadone by rational inference to form new beliefs. Thus Moorean Pretense requires thepretender to represent the world twice over: once as the ‘objective’ or ‘outer’ world,which includes the pretend-fact that p; and once as the ‘subjective’ or ‘inner’world, which excludes the pretend-fact that p. In a commissive Moorean Pretenseone would feed contrary or contradictory premises into the two pretenses. In themushroom example, one feeds into the outer pretense the premise,

F. These mushrooms are false morels.

However, one feeds into the inner pretense, as a possible basis for action, emotion,and other beliefs, the contrary premise,

T. These mushrooms are (true, genuine) morels,


Only the outer pretense has access to F, and only the inner has access to T.There is no leakage of information from one to the other; each is informationallyinsulated from the other.

Compartmentalization has its limits, however. Consider the clown who playsthe part of an innocent person walking, eyes averted, into a column—feigningignorance, yet cognizant that unless he swerves from his path he will indeedwalk into a column. It is sometimes difficult to feign such ignorance. A novicemight telegraph his prescience by visibly decelerating as he nears the column,and by showing anticipation on his face. A seasoned performer, on the otherhand, somehow manages to hide his knowledge from himself, so that it does notshow in any readily observed emotions or actions. His finesse helps his audienceengage in the two tasks as well: to project themselves into the clown’s innocenceand yet at the same time to be fully clued in.

For certain types of action, it appears to be not just difficult but humanlyimpossible to prevent our knowledge from influencing what we do. Discoveringwhat we already know seems to be such a type. Consider a two-player gamein which Player A hides something and Player B tries to discover or at least toguess correctly what it is. The hidden object may simply be what A is ‘thinkingof ’, such as a particular number. Or it may be the particular coordinates ofthe squares that A has filled in on a hidden grid. In the Battleship game, thesquares represent ‘ships’ of various types, each consisting of a certain number ofcontiguous filled-in squares in a straight line. Player B’s goal is to hit (bomb,torpedo) all of the squares comprising each of the opponent’s ships. B calls outthe coordinates of the targeted square, and A replies by indicating whether theshot is a hit or a miss, and if a ship was hit, what type of ship was hit. Obviouslyit is important that A’s placement of the ships be hidden from B. (A similargame is the Windows game Minesweeper, which pits the human player againstthe computer.)

Suppose one were to try to play solitaire Battleship or solitaire Minesweeper.One and the same human being would be both A the scene-setter, who sets upthe ships or the mines, and B the scene-player, who tries to locate A’s ships ormines. The aim of the solitaire game would be to discover (as Player B) somethingyou already consciously know (as Player A). Barring certain pathologies or a timeinterval long enough to allow one to forget what one did in setting the scene,such a task would seem at least psychologically (if not logically) impossible. Fornothing set up by Player A will be hidden from Player B: Knowledge will ‘leak’from scene-setter to player. Even more obviously, there appears to be no possiblesolitaire version of the game of ‘Guess what number I am thinking of.’

Moorean Pretense need not be as difficult as playing a game that requires thatwe discover something we already know. I can play in my mind the innocentmorel-eater, knowing at the same time that I am the unknowing dupe of nature’sdeceit. Nonetheless, pretending that p and simultaneously pretending not tobelieve that p (or to believe that not-p) is pretense of a sophisticated sort.


3. FALSE BELIEF

I defined a Moorean Pretense as having at least one premise of either the‘omissive’ form,

p but I don’t believe that p

or the ‘commissive’ form,

p but I believe that not-p.

Thus far, I have spoken as if the pretender were pretending each of the conjunctsto be true—that is, merely pretending. However, it is often the case that one ofthe conjuncts is imported into the pretend world from the actual world—thatis, the world as the pretender actually believes it to be. Here are three distincttypes of Moorean Pretense:

1. (ascription-preserving) I pretend that p, carrying over into my pretensemy knowledge or belief that I don’t really believe that p (or do believethat not-p). Here my inner pretense, but not my outer, is simply carriedover from the actual world.

2. (world-preserving) I pretend that I don’t believe that p (or do believe thatnot-p), carrying over into my pretense my knowledge or belief that p.Here my outer pretense, but not my inner, is simply carried over fromthe actual world.

3. (non-preserving) I pretend both that p and that I don’t believe that p(or do believe that not-p), carrying neither conjunct over from what Iactually know or believe. Here neither pretense, inner or outer, is carriedover from the actual world.

Type 1 (ascription-preserving) Moorean Pretense is not uncommon. Philo-sophers and scientists pride themselves on asking themselves, ‘What if I am (weare) wrong?’ Sometimes this is a matter simply of pretending something contraryto what they actually believe, where it is not essential that at the same time theycontinue to believe as they do. But in other cases it is essential that they imaginecontinuing to believe as they do—and being wrong. For example, they may wantto assess the cost of acting on a false positive relative to that of acting on a falsenegative. Accordingly, they carry out in their mind an action plan based on theiractual belief and imagine the consequences likely to ensue if the belief is false.

Descartes’s pretenses in the Meditations are also ascription-preserving MooreanPretenses. He tries to persuade himself that nothing he currently believes is so,that the physical world he seems to perceive does not in fact exist, and thatnothing he seems to remember is true. However, along with this pretense, heimagines himself continuing to believe as he actually does:


I will consider myself as without hands, eyes, flesh, blood, or any of the senses, and asfalsely believing that I am possessed of these. (Meditation I, paragraph 12)

Descartes’s beliefs within the pretense do not match the facts that shape thepretend world within which he holds these beliefs. It was this Moorean Pretensethat made it possible to introduce the cogito and, in its wake, three of themajor issues in modern philosophical thinking about the mind: the mind–bodyproblem, the problem of phenomenal consciousness, and the problem of otherminds.⁶

Type 2 (world-preserving) Moorean Pretense may be even more common. Itmay underlie our ability to anticipate and explain the behavior of people whomwe know to be (or perhaps intend to be) mistaken about their environment.Developmental psychologists have devised a number of tests of the ability ofchildren of various ages to recognize false beliefs in others. In the most famousof these, that of Wimmer and Perner, the child subject observes two puppetcharacters, trustful Sally who puts her marble away in her toy box and covetousAnne who transfers it to her own basket while Sally is out of the room.⁷ WhenSally wants to play with her marble again, where will she go to get it? An adultor an older child is likely to say, Sally will go to her toy box, where she left themarble. Asked why she doesn’t go to Anne’s basket, where the marble actuallyis, they might answer, Sally was unaware that the marble had been transferredto Anne’s basket. She was unaware of this because she was out of the room atthe time and therefore didn’t see it being moved. Young children, on the otherhand, don’t seem to get it. Asked where Sally will go to get her marble, they willpoint to Anne’s basket, evidently using their own awareness of where the marbleactually is. Wimmer & Perner and numerous other experimental studies haveshown that nearly all children make incorrect predictions—until about age 4,when something seems to click, and nearly all get it right.

What is it that clicks? According to the ‘simulation’ account I have givenelsewhere⁸, it is the capacity for a kind of pretense—what I would now call Type

⁶ David Rosenthal has drawn a connection between Descartes’s cogito and Moore’s paradox in anumber of his writings. The distinction between types of Moorean pretense is my own.

⁷ H. Wimmer and J. Perner (1983), ‘Beliefs about Beliefs: Representation and ConstrainingFunction of Wrong beliefs in Young Children’s Understanding of Deception’, Cognition, 13/1:103–28.

⁸ I have argued for such an account, although I did not label the kind of pretense involvedas ‘Moorean Pretense’, in several publications, including ‘Folk Psychology as Simulation’, Mindand Language, 1/2 (Summer 1986), 158–71; repr. in W. Lycan (ed.), Mind and Cognition: AnAnthology 2nd edn. (Oxford: Blackwell Publishing, 1998). A similar view has since been adoptedby a number of developmental psychologists and has led to numerous experimental studies. Mostof the early contributions by philosophers and developmental psychologists to the ‘simulation vs.theory’ debate first appeared in Mind and Language, 7/1–2 (Spring–Summer 1992), Special Issueon ‘Mental Simulation: Philosophical and Psychological Essays’); repr. in M. Davies and T. Stone(eds.), Folk Psychology: The Theory of Mind Debate (Oxford: Blackwell, 1995). For some of the laterinterdisciplinary work on the topic, see P. Carruthers and P. Smith (eds.), Theories of Theories ofMind (Cambridge: Cambridge University Press, 1996).


2 Moorean Pretense.⁹ Suppose the child subject decides where Sally will go toget her marble by mentally playing the role of Sally. As Sally, the child wants toget her marble, and the place to get it, as the child knows, is Anne’s basket! Tomove beyond this automatic application of what the child knows, what must bepretended is the following:

M. Sally’s marble has been moved to Anne’s basket, but I, Sally, don’tbelieve it has.

The logical form of M is, of course,

p, but I don’t believe that p.

Again supposing that the child subject answers this question, ‘Where will Sallygo to get the marble?’ by role-playing, it is essential that she be capable ofcompartmentalized reasoning. Otherwise, she would fail to cordon off theinformation she gained as spectator while Sally’s head was turned. In that case,she would make the incorrect prediction that Sally will look for her marble whereit actually is, namely, in Anne’s basket.

Does the empirical evidence support the hypothesis that in human beings thecapacity to ascribe false beliefs depends on the capacity for Type 2 MooreanPretense? Although far from conclusive, the evidence does offer some support.One relevant empirical issue concerns neuropathologies that deprive people ofthe capacity for Type 2 Moorean Pretense. It is well established that autism isassociated with a notable lack of spontaneous pretense, particularly sophisticatedforms such as Type 2 Moorean Pretense. A number of experimental studieshave shown that people with autism generally exhibit a profound deficit in thecapacity for false belief ascription.¹⁰ Nonetheless, some people with autism domanage to ‘pass’ various false belief tests. Can the same individuals also carryout sophisticated pretense? Do they, perhaps, find ways to ‘cheat’ on these tests,providing the right answers without invoking the concept of belief? Further, evenif it is likely that a causal connection explains the coupling of these twin deficitsin most people with autism, it is not yet clear which way the connection flows.Perhaps the lack of Type 2 Moorean Pretense explains the failure to ascribe falsebeliefs; perhaps the failure to ascribe false beliefs explains the lack of Type 2

⁹ I also believe that the capacity for Type 1 Moorean Pretense is needed, too, for ascribing beliefto oneself —not just saying ‘I believe’ as a prefix to an assertion, but understanding one’s utteranceas a genuine ascription of belief. I develop this point, among others related to belief ascription andMoorean Pretense, in a paper to appear in a special issue of Synthese on the topic, ‘Self-Ascriptions ofAttitudes’. For an earlier application of Moore’s Paradox to self-ascriptions of belief, see my paper,‘Sellars’s Ryleans Revisited’, Protosociology: An International Journal of Interdisciplinary Research, 14(2000), 102–14.

¹⁰ For a fuller application of the simulation theory to autism, with some references to theempirical literature, see R. M. Gordon and J. Barker, ‘Autism and the ‘‘Theory of Mind’’ Debate,’in G. Graham and L. Stephens (eds.), Philosophical Psychopathology: A Book of Readings (Cambridge:Mass.: MIT Press, 1994), 163–81.


Moorean Pretense; and perhaps both deficits are connected only by having acommon cause.

Another relevant empirical issue concerns the age of onset of the capacity forType 2 Moorean Pretense in children who are not autistic. Obviously, if thecapacity to ascribe false beliefs depends on the capacity for Type 2 MooreanPretense, it must not emerge earlier. However, methodological issues cloud theissue. Not only is it difficult to establish when such pretense begins; there hasalso been disagreement about the age at which genuine false belief ascriptionemerges.¹¹ Pretense of a less sophisticated sort usually becomes evident beforeage 2, long before the child can master false belief tests such as the Sally–Annetask.¹² However, the crucial and more difficult question concerns the capacity forType 2 Moorean Pretense. Although I am not aware of any study that specificallyasks when children begin to engage in such pretense, my bet is that it is not laterthan the age of false belief ascription.

Finally, in Type 3 (non-preserving) Moorean Pretense, I pretend both that pand that I don’t believe that p (or do believe that not-p), carrying neither conjunctover from what I actually know or believe. The fantasies described in the firstparagraph may be construed as examples of this type. In the inner pretense I amsauteing ordinary non-poisonous morel mushrooms, and in the outer pretenseI am sauteing poisonous false morel mushrooms. Actually, I am doing neither.Or, in the inner pretense I am walking down the aisle with no obstacle in theway, and in the outer pretense I am walking into a supporting column in themiddle of the aisle. Actually, I am merely sitting at home, fantasizing. Neitherthe premise of the outer pretense nor that of the inner pretense is carried overfrom the actual world of the imaginer.¹³

When a Moorean Pretense is acted out, whether in stage or film acting or inchildren’s pretend play, it is typically of Type 3. The actual world does not furnishthe premise of either the inner or the outer pretense. For example, suppose, asbefore, children pretend certain mudpies to be ordinary cherry pies. However,that is just the inner pretense. The children also share an outer pretense: thatone of these pretend-pies is a pie with secret magical powers. Eating such a pie

¹¹ However, the 4-year-old chronology does seem to be holding its own. See H. Wellman, D.Cross, and J. Watson, ‘Meta-Analysis of Theory of Mind Development: The Truth about FalseBelief ’, Child Development, 72/3 (2001), 655–84.

¹² See P. L. Harris and R. D. Kavanaugh, ‘Young Children’s Understanding of Pretense’,Monographs of the Society for Research in Child Development, 58/1 (1993).

¹³ Whether these examples should be classified as Type 3 Moorean Pretenses depends, strictlyspeaking, on how the premise of the inner pretense is stated. If it is, ‘I am walking down the aisleof a department store, and not walking into a column’, then it is of course not true of someonewho is merely sitting at home, fantasizing. However, if the premise is stated merely as, ‘I am notwalking into a column’, then it is, strictly, true of the imaginer. Likewise, ‘I am eating ordinary(non-poisonous) morels’ is not true of the imaginer, but ‘I am not eating poisonous mushrooms’,is. Although this relativity is worth noting, I think it can be fixed with a little more work. (I noticedit just before going to press.) I am sure it raises no important problem for my distinction betweenthree types of Moorean Pretense.


immediately transforms one into a donkey. The children innocently pretend-eatthe other pies, to no bad effect. Then they partake of the magic pie; and in amoment, acting out the sequelae of unseen forces, they begin to walk on all foursand bray. The mudpies are of course neither cherry pies nor magic pies. Neitherof these premises is a holdover from the actual world.

Moorean Pretense, particularly Type 3 Moorean Pretense, appears to be adefining property of dramatic irony. For example, in Oedipus Rex, Sophoclessecures the audience’s complicity in a double pretense. We pretend that the actorwho plays the protagonist is a man who has slain his father and married hismother, two events that mark him for destruction. We also pretend to see theworld through the protagonist’s innocent eyes: The king he slew was not hisfather and the queen he married was not his mother. The audience is at onceclued in and empathetic. It shares in the outer pretense that sets the scene for theaction, and it shares in the inner pretense that allows the actor to get behind hisscripted behavior.

4 . SUMMARY

Moorean Pretense is not absurd. It underlies Cartesian doubt and is rampant inmodern science; it may underlie the ascription of false belief, and it is a definingfeature of dramatic irony. Although not absurd, it is psychologically complex. Itscomplexity hinges on a feature that is also crucial to the Moorean absurdity ofassertions: namely, the I-now character of the belief clause,

I don’t believe that p,

or,

I believe that not-p.

In an omissive Moorean Pretense, one is pretending ignorance of a premise ofone’s own pretense, namely that p. In a commissive Moorean Pretense, oneis pretending that a premise of one’s own pretense is false. What keeps thecommissive pretense from being not only absurd but self-contradictory is thecapacity to compartmentalize. Although one is at the same time pretending thatp and pretending that not-p, one assigns these contradictories to distinct logicalspaces, one space containing the world and the other, the world relativized to aparticular point of view.

Index of Names

Abelard 42Adams, E. 228Adler, J. 3, 30–1Albritton, R. 144Allen, W. 195Alston, W. 71Atlas, J. 29–30, 126Armour-Garb, B. 30–1Audi, R. 22, 106Augustine 29, 41–3Ayer, A.J. 228

Bach, K. 126Baldwin, T. 18, 25, 30 , 76–8Benaccerraf, P. 143Bennett, J. 228Bentham, J. 42Berry, G.G. 37Black, M. 14, 23Boer, S. 120Bonjour, L. 71Brandom, R. 27Brentano, F. 18, 20Bronte, C. 181Buridan, J. 38Burge, T. 85Bunuel, L. 31Burnyeat, M. 42

Cantor, G. 37Carroll, L. 44Castaneda, H. 118, 136Church, A. 38Churchland, P. 219–21, 223, 225Collins, A. 3, 13Crimmins, M. 32

Da Vinci 44Davis, W. 195Davison, A. 135De Finetti, B. 224De Almeida, C. 16, 30De Morgan, A. 61Dennett, D. 84DeRose, K. 4, 14, 218Descartes 29, 33, 38, 43–4, 47, 85, 142–4,

237–8, 241Doran, K. 27Dretske, F. 70

Edgington, D. 228Emerson 42Evans, G. 16–17, 31, 92, 94–7, 100–2, 104,

107–9

Fitch, F. 37Foucault 226–7Foley, R. 65, 71

Galileo 81Gallois, A. 3, 32Geach, P. 46Gettier, E. 65Goethe 45Goldstein, L. 13Gombay, A. 3Gordon, R. 4, 29, 33Green, M. 3, 30, 31, 195Grice, P. 77, 195

Hajek, A. 31–2Halpern, B.S. 47Harman, G. 225Hawthorne, J. 221Heal, J. 3, 13, 17, 31, 80, 84–85Hegel 38Hintikka, J. 28 , 143Hume 84Hunter, J.F.M. 135Hyde, D. 227

Jacobson, R. 12Jeffrey, R.C. 220–1, 229Jones, O.R. 26Jourdain, P. 37

Kant 39, 45, 223Kepler 44Klein, P. 16, 65Koestler, A. 81Kriegel, U. 18–20Kripke, S. 133Kuhn 226, 227

Lakoff, G. 141–3Langford, C.H. 7, 140–2Laudan, L. 229Lehrer, K. 65

244 Index of Names

Levi, I. 229Levinson, S.C. 13, 126Lewis, C.I. 40Lewis, D.K. 118, 187, 218Linville, K. 12Locke, 85Lycan, W. 120, 135

MacDonald, M. 37MacFarlan, J. 226–7Maclver, A.M. 140McTaggart, J.M.E. 38Makinson, D. 159Malcolm, N. 6, 12, 165Mariotte, E. 44–45Martinich, A. 13Merton, R. 37Milgram, E. 3Moore, G.E. 3–6, 9, 29, 37–8, 46–8, 53,

76–7, 85, 88, 90–1, 128–9, 155, 165Moran, R. 4, 29, 128, 133–6Morris, C. 141Murdoch, I. 135

Nietzsche 225Nozick, R. 70

Ockham 42O’Connor, D.J. 39Orwell, G. 81

Packer, A. 200Parmenides 29, 39, 43Pecquet, J. 45Peirce, C. 226–7Perner, J. 238Perry, J. 118Plato 29, 40, 42Priest, G. 227Prior, A.N. 118Protagoras 226–227Putnam, H. 135

Quine, W.V.O. 117–8, 120, 124, 226–7

Ramsey, F. 44Reagan, R. 195Reid, T. 85Rhys, J. 181

Ring, M. 12Rosenthal, D. 4, 15, 23, 31–2, 137Russell, B. 37–38, 46–47, 57

Salerno, J. 38Sanford, D. 46Sartre, J.P. 134Savage, L. 224Schopenhauer, A. 44–46Searle, J. 21–22, 24, 31, 106–7, 109, 135Sellars, W. 139–140Sextus, E. 41Shoemaker, S. 3, 12, 23–4, 28, 30, 137–140,

142–4, 205Slote, M. 218Smith, F.E. 41, 157Sophocles 241Sorensen, R. 3, 6, 21–2, 29, 31, 91–2,

104–7, 109, 111, 223Sosa, E. 65, 71Spinoza 29, 43Stevenson, C.L. 48, 165Stich, S. 219–21, 223, 225Stigler, S. 37–8Stoljar, D. 31Strachey, R. 47Strawson, P. 196Stroud, B. 84Szabo-Gendler, Z. 4

Tversky, A. 14

Unger, P. 14, 146, 154–5, 218

Vahid, H. 17Van Fraassen, B. 229Vanderveken, D. 24, 31

Weatherson, B. 221Welbourne, M. 26Wiener, N. 38Williams, B. 44Williams, J.N. 3, 15–17, 27–8, 30–2, 53–8,

76–7, 92, 212Williamson, T. 14, 32, 155, 218Wimmer, H. 238Wittgenstein, L. 3, 4, 6–8, 12, 44–5, 47, 122,

134, 146, 154, 165, 190, 228, 232Wolgast, E. 24

Subject Index

absurdity 5–8, 10–13, 15, 21, 25–7, 29,62–71, 73, 76, 78, 90, 92, 100, 106, 109,137, 190–3, 192, 194, 202–10, 212,219, 232–3, 241.

commissive 5, 8, 17, 19, 20, 21, 31, 91–2,107–9,112,

decrease of 20, 31–2, 92, 104, 109, 112–13examples of 3, 5, 8, 21, 30–1, 56, 72, 76,

90–1, 104, 147–8, 152, 153–5, 161,166–7, 190, 192–3, 210, 211–12,233.

grammatical characterisation of 29, 56, 189Moorean 4, 9, 10–12, 14, 19, 22, 29–32,

44, 48, 53–8, 77–80, 82, 84–8,90–2, 100, 107, 110–112, 189–91,203–6, 210–13, 233–4, 241

ommissive 6, 8, 17, 20–1, 31, 91–2, 101,104–5, 107–9, 112–13

paradigmatic 91, 104acceptance 153–4, 212agnosticism 13, 81, 173, 220alienation 134all-seeing eye 29, 37, 42amnesia 130–2, 199antinomies 223apartheid epistemologies 71

assent 198, 205conscious 139manifest 139–40

assertability 154–160, 222assertion 10–14, 26–8, 30, 40–1, 55, 82, 96,

111, 118, 125, 128, 132–3, 136–7,146–150, 153–4, 158, 166, 169, 190,193, 196, 201, 207–8, 210–12, 218–22,224–5, 228, 230, 232

absurdity of 3, 5, 7–8, 18, 32, 86, 88,90–1, 110, 207, 190

and belief 1, 3, 7, 12, 30, 41, 43, 77–8,82–3, 85–6, 88, 121, 137, 140, 143,146–7, 148–9, 154, 156, 167, 161,202–3, 206–7

and expression 12, 15, 24, 128–40, 146,148–9, 153, 160

commissive 5, 27, 31–2, 86, 91–2, 110,112–13

context of 26, 27, 158, 131, 226–7, 191its distribution over conjunction 12, 10, 14,

23, 26, 67, 110, 158, 193, 204,206–07

and implication 4–6, 8, 10, 13–14, 76–8,86–7, 90, 146

logic of 7, 8, 143, 118, 153Moorean 4, 5, 11–13, 23–7, 53, 78–9, 82,

86–7, 92, 110–112, 137, 142, 149,154, 155, 210–211, 233, 241

omissive 5, 13, 26–7, 31–2, 43, 86, 91–2,112–113

purported 143, 156, 206rational 142–3, 202, 207, 210sincere 11, 93–4, 112, 139, 141–3, 148,

201–3, 206–8, 211–12uttering assertively 5, 129, 131, 137, 140,

142–3, 194attitude-related reasons 148, 173–5, 178, 183,

185authority 92–3, 96–7, 156–7autism 239

Bayesianism 220–1belief 174, 178, 183–4, 190–3, 195, 197,

208, 219–25, 227, 230, 233, 241and closure 11, 21, 61, 65, 69, 105concept of 99, 106conscious 78–9, 165–7, 171–2, 175–80,

182, 185–8, 201, 204–5commissive 91, 92, 101, 102, 109, 204–5de dicto 59, 120, 126–7, 143degree of 152, 159de re 119–20, 126–30, 133, 136, 143de se 119–20, 126–30, 133, 136, 143dispositional 22, 106its distribution over conjunction 54–6, 58,

61, 140, 142, 149, 167, 170–2, 204,206, 207

and the distribution of justified belief overconjunction 99–102, 107, 109

expression of 55, 81, 128, 135, 139–40,190, 193, 222, 224, 228

elimination 21, 105higher-order 79–80, 144and iterated operators 167, 170, 180, 182,

186, 187manifest 77, 139, 196, 199–201meta-linguistic 131Moorean 53–57, 64–73, 77–9, 84, 86, 91,

93, 96, 97, 101–04, 204–5notional sense of 117and occurrent thought 21, 22omissive 10–11, 16–21, 25, 28, 91–2,

104–5, 107, 109, 112, 204–5

246 Subject Index

belief (cont.)and principle of elimination 21, 100, 102,

105, 108rational 55, 72, 142–3, 171, 179, 191, 194relational sense of 117–18, 124reporting 7, 12, 135, 137, 193and Searle’s principle 106–7, 109second-order 92, 98–9, 101, 107self-ascriptions of 13

blindspot 38, 44–5

Cartesian doctrines 33, 38, 43–4, 47, 85,142–4, 237–8, 241

coherence 12, 15, 45, 53, 58–61, 63–7, 70,73, 78–9, 94, 99, 101, 124, 134, 159

commitment 61, 135–6, 144, 196–202, 205,209, 212, 217, 219–20, 227, 229–30

communitarian 226community 199, 225–6conditionals 143, 209, 228consciousness 150, 165–6, 175, 179, 182,

185, 189, 199, 201–2, 212, 236, 238consensus 225–6content-related reasons 173–5, 178, 183–5,

187contradiction 90, 99–103, 105, 108–9, 190,

192, 227counterevidence 16, 58–9, 64, 67, 70

deletion 166, 178–80, 185deliberative perspectives 133delusion example 200dialectical games 40discourse 118, 226–7dissociation 6, 133, 134

eliminativism 32, 166, 221EP test 57–8epistemic goal 55epistemic probability 59Evans’s principle 15–17, 92, 94, 96–7, 99,

101, 102, 104, 107–8expressivism 228

fact/value distinction 135fallibility 55, 57, 85, 91, 98, 101–2, 157–60,

194, 201,first-person point of view 122, 149, 154, 162folk psychology 219, 221foundationalism 66, 70functionalism 3, 84

God 13, 21, 24, 31, 42–4, 46, 91–2, 104,105, 109, 112, 177–8, 226, 230

grammaticality 14, 29, 119, 122–6, 130,135–6, 189, 206, 209

higher-order thought 137idealism 37–8, 42, 46–7illocutionary force 135, 146, 208inconsistency 25, 28, 30, 40–1, 43, 47, 54–5,

59–60, 62, 63, 77, 79, 86–7, 103, 131,137–8, 141–3, 147, 150–1, 157–9,167–8, 170–2, 187–8, 233

Strong versus weak 63, 67incorrigibility 141–2indeterminacy 229indexicals 191, 233–4infinite regress 80, 144–5, 155, 172, 224Inner and Outer 7, 134–6, 198–9, 235–7,

240–1.Integration 32, 134, 166, 178, 180–3,

185–8intention 25–8, 42, 77–8, 80, 82, 84, 92,

110–11, 132, 174–5, 185, 190, 195–7,199–201, 212

introspection 42–3, 47, 55, 63, 93, 96, 98,134, 201, 204

irrationality 53, 55–6, 59–61, 101–3, 107,166, 172

judgment (judgement) 3, 6, 10, 80, 85–8,107, 113, 152, 154, 160

justification 3, 16–17, 27, 30, 32, 65, 67,70–3, 85, 92–104, 107–9, 143, 159,167, 179, 180, 198, 202–3, 211

knowledge 3–4 ,11, 14, 22–31, 37–8, 41–3,45, 47–8, 63, 70–1, 94, 104, 106, 110,149, 154–6, 160, 199, 202, 211, 220–1,223, 236–7

language 23, 47, 126, 135–6, 143, 190, 198,226–7

malapropism 122meaning 3–5, 7, 10, 14, 39–40, 48, 57,

77–8, 120, 123, 140–2, 169, 199speaker 141, 194–5, 197–9, 200–1, 205,

211–13literal 141–2pragmatic 140–2

meta-induction 32, 229–30moral discourse 228

norms 56, 82–3, 84, 85, 88, 146–7, 154–6,191–4, 202–3, 205, 207, 210–12, 221226

omnipotence 44omniscience 92, 102, 138, 157open question 142, 220, 223–4

Subject Index 247

pain 190, 222, 228paradigm 84, 91, 104, 146, 169, 171, 226–7paradox 222, 227

history of 37Jourdain’s Visting Card 37liar 3, 38, 57, 227–8lottery 160, 221Moore’s 3, 37–8, 43–4, 90–1, 100,

189–91, 203, 219preface 54, 157–61, 193–4, 221–2, 227Russell’s 37

perception 45, 83–6, 93, 95–6, 98, 100–1,206

performatives 30, 80–2, 85, 135phlogiston 219plastic surgery example 133pragmatics 10–11, 27, 32, 39, 90, 95, 141precisification 227–9predicate 117–18, 125–6pretense 232–241

commissive 232, 235, 237, 241omissive 232, 237, 241

privileged access 117, 137, 139, 143–4probability 152, 159, 161, 218–19, 221,

224–5, 229higher-order 224subjective 160, 192, 218, 220–1, 225

pronounsanaphoric 118, 120–1indirect reflexive 118, 136–7, 144of laziness 121

propositional attitudes 79, 117, 120, 146,149, 153, 166–8, 170, 172, 174, 176–7,184, 187, 208

quantification 118quasi-performative 85, 135

rationality 9, 15, 23, 28, 32, 33, 53, 55, 56,59–61, 65–6, 72, 95, 97, 101–4, 107,112, 117, 133, 141, 143, 166–7, 170,172, 180

degrees of 22, 65, 107doxastic 11 53–6, 66–7, 70–1ideal 21–2, 105–7practical 10, 12, 23, 24, 30,191–2, 194,

202theoretical 9–10, 12, 13, 24, 29, 30,

191–4, 202, 207, 210relativism 32, 37, 225–7, 237, 241

scope 57, 141, 146, 148, 179, 182, 186–7,205, 208–10, 234

self-awareness 79

self-knowledge 93, 97, 102, 134, 137, 138self-refutation 41, 53–5sentence

identity criteria of 123Moorean 11, 13, 15, 20, 23–4, 27–9, 47,

105, 134, 136–8, 140–3, 205,217–22, 226–7

string 119, 122–4system 123, 138token 125type 123, 125–6, 219, 222

simulation theory 238sincerity 80–1, 112, 141, 143, 195, 202, 206,

208singular term

referential 117, 127transparent 124–5

skepticism 224social pragmatism 27speaker-meaning 194–202, 205, 211–13speech acts 24, 77, 82, 88, 135–7, 146, 156,

201–3, 208, 210–13spoonerism 122–3Stigler’s Law 37–8sub-valuationism 227super-valuationism 227–8supposition 7, 13–14, 31, 48, 83, 100, 109,

207–10, 232

theoretical perspectives 133thick concepts 136transparency 30–1, 143, 146–9, 151, 153,

158, 160truth 9, 10, 13, 22, 23, 27–8, 37, 39, 40, 42,

44, 47, 55, 56, 60–3, 70–2, 79, 81–2,85–6, 88, 92, 101, 106, 110–112, 121,135–6, 139, 143, 149, 153–5, 158, 161,174, 177–8, 180, 205–7, 219, 226, 230

gaps 32, 228gluts 32, 227higher-order 225possible 32, 90, 102–4, 107

truth-tracking epistemologies 9, 70–2

Unger-Williamson thesis 146, 154utterance

Moorean 77–8, 203token 122, 125–6, 128type 122–24, 126–8, 130, 133

vagueness 227, 228validity 9, 17, 61–2vital spirits 219visual field 44–6

Moore's Paradox: New Essays on Belief, Rationality, and the First Person

Documents

Transcript of Moore's Paradox: New Essays on Belief, Rationality, and the First Person