[Andrew Brook, Richard C. Devidi] Self-Reference and Self Awarness

Self-reference and Self-awareness

Advances in Consciousness Research

Advances in Consciousness Research provides a forum for scholars fromdifferent scientific disciplines and fields of knowledge who study consciousnessin its multifaceted aspects. Thus the Series will include (but not be limited to)the various areas of cognitive science, including cognitive psychology, linguis-tics, brain science and philosophy. The orientation of the Series is towarddeveloping new interdisciplinary and integrative approaches for the investiga-tion, description and theory of consciousness, as well as the practical conse-quences of this research for the individual and society.

Series A: Theory and Method. Contributions to the development of theory andmethod in the study of consciousness.

Editor

Maxim I. StamenovBulgarian Academy of Sciences

Editorial Board

David Chalmers, University of ArizonaGordon G. Globus, University of California at IrvineRay Jackendoff, Brandeis UniversityChristof Koch, California Institute of TechnologyStephen Kosslyn, Harvard UniversityEarl Mac Cormac, Duke UniversityGeorge Mandler, University of California at San DiegoJohn R. Searle, University of California at BerkeleyPetra Stoerig, Universität DüsseldorfFrancisco Varela, C.R.E.A., Ecole Polytechnique, Paris

Volume 30

Self-reference and Self-awarenessEdited by Andrew Brook and Richard C. DeVidi

Self-reference and Self-awareness

Edited by

Andrew Brook

Richard C. DeVidiCarleton University, Ottawa

John Benjamins Publishing CompanyAmsterdam�/�Philadelphia

The paper used in this publication meets the minimum requirements of American8 TM

National Standard for Information Sciences – Permanence of Paper for PrintedLibrary Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data

Self-reference and Self-awareness / edited by Andrew Brook and Richard C. DeVidi.p. cm. (Advances in Consciousness Research, issn 1381–589X ; v. 30)

Includes bibliographical references and index.1. Self-knowledge, Theory of. 2. Reference (Philosophy) I. Brook, Andrew. II.

DeVidi, Richard C. III. Series.

BD438.S.S45 2001126--dc21 2001043495isbn 90 272 5150�9 (Eur.) / 1 58811 048�6 (US) (Pb; alk. paper)

© 2001 – John Benjamins B.V.No part of this book may be reproduced in any form, by print, photoprint, microfilm, or anyother means, without written permission from the publisher.

John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The NetherlandsJohn Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

Table of contents

Acknowledgments vii

1. Introduction 1

Section IGreat Precursors

2. Kant, self-awareness and self-reference 9

Andrew Brook

3. Frege and the first person 31

Richard DeVidi

Section IIClassical Contributions

4. ‘He’: A study in the logic of self-consciousness 51

Hector-Neri Castañeda

5. Self-reference and self-awareness 81

Sydney S. Shoemaker

6. Self-identification 95

Gareth Evans

7. The problem of the essential indexical 143

John Perry

Section IIIRecent Work

8. The myth of mental indexicals 163

Ruth Garrett Millikan

vi Table of contents

9. Thinking about myself 179

Maite Ezcurdia

10. Introspective misidentification: An I for an I 205

Melinda Hogan and Raymond Martin

11. First-person reference, representational independence, andself-knowledge 215

Christopher Peacocke

12. The constructed and the secret self 247

William Seager

Bibliography 269

Index of names 275

Acknowledgments

We would like to thank Maxim Stamenov, Series Editor, who invited us toprepare this volume and Bertie Kaal, Editor at John Benjamins Books, who hasdone a wonderful job of supporting us in the preparation of this volume. Wethank the authors and publishers of previously published materials who gaveus permission to reprint these works and the authors of the essays commis-sioned for this volume for entrusting us with their work. The volume has beensubstantially delayed. This could not be avoided and we apologize to thecontributors.

Castañeda, H.-N. 1966. ‘He’: A study in the logic of self-consciousness. Ratio8:130–57 is reprinted with permission.

Evans, G. 1982. Self-identity, is Chapter 7 (except for the appendices) of his TheVarieties of Reference. J. McDowell, ed., and is reprinted with the kindpermission of Oxford University Press.

Perry, J. 1979. The problem of the essential indexical.Noûs 13:3–21 is reprintedwith the kind permission of Professor John Perry.

Shoemaker, S. 1968, Self-reference and self-awareness. Journal of Philosophy65:555–67 is reprinted with the kind permission of Professor SydneyShoemaker.

Peacocke C. First-person reference, representational independence, and self-knowledge also appears in Wolfgang Kunne, Albert Newen, and MartinAnduschus, eds. Direct Reference, Indexicality, and Propositional AttitudesStanford, CA: CSLI Publications, 1997.

Millikan, Ruth. 1990. The myth of mental indexicals uses material from her‘The myth of the essential indexical’. Noûs 24: 723–734 with the kindpermission of the editors.

Chapter 1

Introduction

Some intricate, very subtle differences between awareness of self and awarenessof other things and the relationship of these differences to some equallyintricate features of reference to self using first person pronouns (‘I’, ‘my’, ‘me’and ‘mine’) and their third person equivalents (for example, ‘he himself ’) havebeen the subject of a clearly delineated and largely self-contained dialoguewithin analytical philosophy for about thirty-five years now. The investigationbegan with seminal papers by Hector-Neri Castañeda (1966, this volume; 1967;1968), took on important new dimensions in the work of Sidney Shoemaker(1968, this volume;1970) and led to further classical contributions by Perry(1977; 1979, this volume) and Evans (1981; 1982, this volume). It continues toinspire important new work to this day. Shoemaker called his paper ‘Self-reference and self-awareness’. It would be hard to think of a more perfect titlefor the topics under discussion in this volume, so as well as republishing hisseminal paper, we have also gratefully borrowed his title.

Three ideas have been central to this dialogue. They are that:

1. Certain uses of ‘I’ and cognates are ineliminable (the idea of the essentialindexical)1

2. In certain situations at least, we are immune to error throughmisidentifica-tion of another as oneself (the idea of immunity to error through misidenti-fication with respect to the first person)

3. When we refer to ourselves using ‘I’ and cognates, often or always it takesplace without one identifying anything as oneself via properties that one hasascribed to the thing (the idea of self-reference without identification).

1. Note that we do not refer to ‘I’, etc., as indicators. This may seem a bit odd but whether thesepronouns and their cognates are indicators is a matter of debate. For example, one of Castañeda’s threeseminal papers is called ‘Indicators and Quasi-Indicators’. He distinguishes the third person equivalentsof first person pronouns such as ‘I’ from indicators. Because they depend on other indicators and/orreferring expressions, he says that they are really quasi-indicators. It is an indication of the intricatenessand complexity of the recent work on self-reference that even something as basic as this could be amatter for debate.

2 Self-awareness and Self-reference

These three ideas in various configurations have been central to the wholedialogue, from its creation (or recreation) in the 1960s to the present.

Creation or recreation? At the beginning of his 1966 paper, Castañeda saysthat his topic is “almost brand new”. Though his paper is without questionhighly original and of first rate importance, Castañeda’s claim is an exaggera-tion. No idea in philosophy (and precious few anywhere else) is ever construct-ed entirely from whole cloth. With respect to both self-awareness and self-reference in particular and indicators in general, that Frege was an importantprecursor is well-known, especially his (1918/19). TheWittgenstein of The Blueand Brown Books was another, as has been clear at least since Shoemaker’spaper. It is less well-known that Kant also discussed some of the peculiarities ofself-reference and self-awareness, indeed anticipated claims about them thathad to wait ‘til the work we’ve just introduced to get a full articulation.

In this collection, we start by exploring two of the great precursors of therecent dialogue on self-reference and self-awareness just mentioned, Kant andFrege. (There is no separate paper on Wittgenstein, the third, for reasons wewill give in a moment.) Next, four of the classic contributions just mentionedare reprinted. The final section offers five examples of more recent work andwere written for this volume. We will try to say something by way of introduc-ing these papers but there are limitations on how much one can do. Becausethey are dense and intricately argued, there is no substitute for reading thepapers themselves.

The papers on Kant and Frege that make up Section I are by the two editors.Andrew Brook argues that Kant not only anticipated but also clearly articulatedthe phenomenon that Shoemaker calls self-reference without identification, thatis to say, that one can refer to oneself, and to refer to oneself as oneself, without(otherwise) identifying oneself. There are also indications that he was aware ofsomething very much like the contemporary notion of the essential indexical.Brook argues, moreover, that, unlike contemporary theorists at least up toPeacocke (this volume), Kant had at least the makings of a theory to explainthese peculiar features of self-awareness. His theory also yields insight intowhich references to oneself are and which are not immune to the error ofidentifying someone else as oneself (though Kant seems not to have thought ofanything like this notion of immunity to error itself). Peacocke also discussesaspects of Kant’s work on awareness of self.

Frege was the first to take the ‘linguistic turn’ so characteristic of twentiethcentury analytical philosophy; Richard DeVidi shows how this new approach tophilosophical problems invests the connection between the linguistic act of

Introduction 3

referring to oneself and awareness of self with considerable significance. Frege’sfirst faltering steps toward an account of first person reference are generallyconceded to be unsuccessful; but the terms in which he sets the problem havebeen enormously influential on almost all subsequent work in the area. DeVidiargues, against Frege’s critics and his revisionist friends, that Frege’s approachto the theory of meaning requires only a very minor extension to deal with thepeculiar features of first person reference quite successfully.

We did not commission a separate treatment of Wittgenstein’s contribu-tions to our topics for the following reasons. Wittgenstein’s remarks fall mainlyinto two groups, the middle period remarks of the Blue and Brown Books(1933–4) and remarks from his final period in works such PhilosophicalInvestigations (1953), Zettel, andRemarks on the Philosophy of Psychology, thoughthere also some relevant but highly compressed remarks in theTractatus (1921).Shoemaker’s and Evans’ works, both of which had to be part of the volume inany case, discuss the most interesting of Wittgenstein’s middle period remarksso it was not necessary to commission a separate study of them. Wittgenstein’sfinal period remarks on the use of ‘I’ and related matters, many suggest, denythat when we use ‘I’, etc., in the relevant ways, we are using them referentially atall. Peacocke discusses these notions near the end of his paper. The absence of aseparate study ofWittgenstein’s final thoughts on self-awareness may come as asurprise. It reflects the following considerations. First, it is singularly difficult toget a clear, uncontroversial account of whatWittgenstein was trying to show usin his late remarks on first person pronouns, etc. If anything, it is even harder tosee what the considerations were that, in his view, supported his remarks.Second, his remarks have persuaded almost no one who is not already stronglydrawn to his general approach. For example, not a single contributor to thisvolume holds that the relevant uses of ‘I’, etc., are nonreferential and a numberof them explicitly deny the claim. In short, the ‘non-cognitivist’ approach hashad little influence on the work collected in this volume.

Section II contains four papers that have achieved the status of classics inwork on our topics. The paper that launched the whole recent dialogue wasHector-Neri Castañeda’s famous work, ‘“He”: A study in the logic of self-consciousness’. In this and other papers on the same topics, Castañeda arguesthat certain uses of ‘I’ and cognates are ineliminable. In particular, no descrip-tion, not even one containing (other) indexicals, can be substituted for theseuses. This is one of the two great claims that launched the recent discussion.

The secondwas first articulated by Sydney Shoemaker in ‘Self-reference andself-awareness’, the idea of immunity to error through misidentification of


another as oneself. This immunity goes with another central peculiarity of self-awareness, what Shoemaker calls self-reference without identification. Weintroduced both ideas earlier.

In ‘The essential indexical’, John Perry picks up Castañeda’s claim thatnothing can be substituted for certain uses of ‘I’ and cognates and articulates itas the idea of the essential indexical. Building his account on what he needs toknow in order to find out that a person making a mess in a store is him, Perryargues that without knowing that some part of a description applies to himself,without therefore knowledge of himself (and as himself), even a completedescription of the person who is in fact making the mess need not be enoughfor him to find out that it is himself.

The chapter from Gareth Evans’ 1982 book, ‘Self-identification’, is amassive, sprawling meditation on many aspects of self-reference and self-awareness. One suspects that Evans might have done somemore work on it hadhe lived long enough. At any rate, it resists summarization. Shoemaker’s twothemes, self-reference without identification and immunity to error throughmisidentification, and their relationship to one another play a central role in thechapter. So does an idea that one must have more indexicals than ‘I’ and itscognates available. In particular, one must be able to locate oneself in theobjective world (and so be able to use ‘here’) and know when a given time hasarrived (and so be able to use ‘now’).

In Section III, we turn to more recent work on self-reference and self-awareness. All the papers were prepared for this volume, though one of them,Peacocke’s, was co-published elsewhere (Peacocke 1997) and Millikan’s paperis a revision of her well-known 1990 critique of Perry. One of the first things tonotice about these papers is how pervasively they have been influenced by thefour classical works of Section II. Every paper of Section III criticizes papers inSection II. Indeed, every one of the former papers makes reference to at leasttwo of the latter papers. One Section III paper, Ezcurdia’s, also criticizes aSection III paper, namely, Millikan’s. Some might suspect that the originalauthors could handle many of the criticisms laid out in Section III but all fivepapers raise significant issues and back their claims with considerable force ofargument. In fact, it would extraordinarily interesting to set the Section IIIcriticisms against the Section II claims that they aim to criticize and adjudicatethe disagreements.

RuthMillikan’s criticism in ‘The myth of mental indexicals’ can be summa-rized quite succinctly: there is something in reference to self that is essential butit is not anything indexical. Why? The relevant uses of ‘I’ do not have the

Introduction 5

features central to indexicals. For example, indexicals have no constant refer-ence but ‘I’ as used by a given person always does. Moreover, any element thatis indexical in uses of ‘I’ does not yield the information crucial to predictingwhat the user of ‘I’ will go on to do when, a central element in reference to self.And so on. In short, the element in reference to self that is essential is not anindexical.

In ‘Thinking about myself ’, Maite Ezcurdia criticizes two authors who, shesays, argue that there need not be any presentation of self when one refers tooneself using ‘I’ and cognates. Mellor is one. Millikan, theMillikan of 1990 andthis volume, is the other. Mellor holds that the job that a representation of selfis supposed to do is in fact done by context. If that were so, Ezcurdia argues,we would have no way to discriminate between an indexical referring to meand an indexical referring to the place I’m at. Since we can discriminate thesethings, Mellor must be wrong. In Ezcurdia’s view, Millikan’s emphasis on therelationship between an indexical and a referent specified by the semantic rulesfor the kind of indexical in question goes with a ‘Millian’ view of indexicalreference, in which the semantics of an indexical are exhausted by its namingfunction and mode of presentation can play no role. Ezcurdia argues that thisview leaves Millikan with no way to distinguish between, for example, a use of‘I’ presenting an object that happens to be myself and a use of ‘I’ presentingwhat I know to be myself.

In ‘Introspective misidentification: an I for an I’, Melinda Hogan andRaymondMartin go after the notion of immunity to error through misidentifi-cation. They do so primarily by presenting four putative counter-examples tothe view. The counter-examples are all of the form: I am aware of someonehaving a feeling in the way that I am aware of feelings that I am having but itmay in fact not be me who is having them. They consider objections to theircounter-examples and reply to the objections. Two of the counter-examples areenmeshed in current controversies about identity of persons over time.

In ‘First-person reference, representational independence, and self-knowl-edge’, Christopher Peacocke isolates a special kind of awareness of self in whichone knows something about oneself, e.g., that one is seeing something, whenthis could not be inferred from anything about how the object of this act isrepresented. Peacocke calls these representationally independent uses of the firstperson. (His distinction is subtle and much more interesting than this gloss.)What interests him is that often representationally independent uses of ‘I’ andcognates give one knowledge, knowledge, for example, that it is oneself who isseeing something. He offers what he calls a delta account to explain this,


according to which a subject having a property is sufficient in the relevant casesfor a subject to refer to itself, and specifically, to ascribe the property to itself. Hethen applies the account to the illusion of a “transcendent subject” that heclaims to detect in Kant, Schopenhauer and the early Wittgenstein, perhapswithout recognizing that his delta account itself has strongly Kantian elements(see Brook, this volume). He concludes with a critical look at ‘no-ownership’theories and theories that the relevant uses of ‘I’ and cognates are not referential.

Seager’s ‘The constructed and the secret self ’ can be seen as a step towardmeeting a need that Evans articulates (1982, 205; this volume, 95). Says Evans,we cannot give a complete account of self-reference and self-awareness withouta complete theory of mind, though we can make some progress while we waitfor one. Seager’s paper takes some steps toward developing such as generaltheory of mind. He urges that from the perspective of a representational theoryof mind, it is natural to arrive at a dual view of selfhood, i.e., of whatever it isthat realizes personhood and carries sameness of personhood across time. Hethen applies the resulting duality to a wide range of phenomena, all the wayfrom Williams’ puzzle cases about personal identity to such mundane mattersas the relative cognitive impenetrability of illusions such as the Müller-Lyerarrowhead illusion and the surprisingly tenacious conflicts between, e.g., ourconsidered values and what we desire that we find in ourselves from time totime, most of us at any rate.

Rich in precursors and stimulated by Castañeda’s and Shoemaker’s seminalpapers, the work of the past thirty-five years on self-reference and self-aware-ness has generated a wealth of deep, sophisticated philosophy.We hope that thisvolume conveys something of that richness.

Andrew Brook and Richard DeVidiCarleton University, Ottawa, Canada

Section I

Great Precursors

Chapter 2

Kant, self-awareness and self-reference

Andrew BrookCarleton University, Ottawa

Introduction

As is well-known, Castañeda (1966, this volume; 1967), Shoemaker (1968, thisvolume), Perry (1979, this volume), Evans (1982, this volume) and others urgethat awareness of self has peculiar features. It is lesswell-known that someof thesepeculiarities were discovered as early as 1781 and the Critique of Pure Reason.1

Two of the key peculiarities are that,

1. In certain kinds of awareness of self, first-person indexicals (I, me, my,mine) cannot be analyzed out in favor of anything else, in particular anythingdescriptionlike,

and that,

2. In such cases, awareness of self is via what Shoemaker calls self-referencewithout identification.One canbe aware of something as oneself without identify-ing it (or anything) as oneself via properties that one has ascribed to the thing.

(2) is often taken to be closely related to another putative peculiarity of aware-ness of self that Shoemaker calls immunity to error through misidentification(Shoemaker 1970, who claims to have found the core of the idea in Wittgen-stein 1933–4: 66–70). This is the idea that in some situations, we cannot becomeaware of a person by being aware of certain experiences, take that person to beoneself, and be wrong. Certainly this latter idea has become amajor preoccupa-

1. I explore the issues in this paper at greater length in Brook 1994: Ch’s 4, 7 and 8. Unless otherwisenoted, references to Kant are to the Critique of Pure Reason, in the Akademie pagination using thestandard ‘A’ and ‘B’ notation for the first two editions (the only two that Kant prepared himself). Areference to one edition only means that the passage in question appeared only in that edition.Translations start fromNormanKemp Smiths’ 1927 translation andGuyer andWoods’ 1998 translationand have been checked against the original text.

10 Andrew Brook

tion in recent literature on self-reference and self-awareness but whether thereis any such immunity and, if there is, how closely it is linked to self-referencewithout identification are other questions — questions about which we willhave something to say, but not until near the end of this paper.

One standard argument for (1), that certain indexicals are essential, goes asfollows.2 To know that I wrote a certain book a few years ago, it is not enoughto know that someone over six feet tall wrote that book, or that someone whoteaches philosophy at a particular university wrote that book, or… or…or…,for I could know all these things without knowing that it was me who has theseproperties (and I could know that it was me who wrote that book and not knowthat any of these things are properties of me). Nor would it help to add detailsof a more identifying kind— the person whose office number is 123 in buildingABC, the person whose office phone number is … . If I don’t know that thatoffice is my office, that that phone number is my phone number, I could knowall these things and still not know that it was me who wrote the book. And vice-versa— through bizarre selective amnesia, I could cease to know all such thingsabout myself and yet continue to know that it was me who wrote the book. AsShoemaker puts it,

… no matter how detailed a token-reflexive-free description of a person is, …it cannot possibly entail that I am that person [1968:560; this volume, 86].

What a curious piece of knowledge — if it even is knowledge!The standard argument for (2), that certain references to self do not require

descriptive identification, goes as follows:

My use of the word ‘I’ as the subject of [statements such as ‘I feel pain’ or ‘I seea canary’] is not due to my having identified as myself something [otherwiserecognized] of which I know, or believe, or wish to say, that the predicate ofmy statement applies to it [1968:558; this volume, 84].

Whether Kant was aware of (1) is an intriguing question, as we will see. He wasclearly aware of (2). Consider this passage:

In attaching ‘I’ to our thoughts, we designate the subject … without noting init any quality whatsoever — in fact, without knowing anything of it eitherdirectly or by inference [1781: A355].

2. Though most often associated with Castañeda and Perry, a version of it can be found even earlier inNagel (1965).

Kant, self-awareness and self-reference 11

This “attaching ‘I’ to our thoughts” business is interesting. In the kind of self-awareness in question, one is aware not just of oneself. One is aware of oneselfas oneself. One is aware that it is oneself of which one is aware. Put in Fregeanterms, one is not just aware of the being who happens to be oneself. One ispresented to oneself in a certain way, namely, as oneself (see Ezcurdia, thisvolume). By “attaching ‘I’ to our thoughts”, Kant seems to have had somethinglike this in mind, something like “using ‘I’ to refer to myself as the subject of mythoughts”. Since, on Kant’s view, it is not just identifying properties but anyproperties whatsoever that I need not know in order to refer to myself as my-self,3 ‘non-ascriptive reference to self ’ might capture what is special about thisform of awareness of self better than Shoemaker’s ‘self-reference withoutidentification’ (Brook 1975:188).

But (2) is logically linked to (1). If I am aware of myself as myself withoutinferring this from anything else that I know about myself, my knowledge thatit is myself of whom I am aware has to be independent, at least in some respects,of knowing anything else about myself. I can be aware of myself as myselfwithout being aware ofmyself as anything except—myself. This shows that (2),of which Kant was clearly aware, requires something like (1). The existence ofsuch a link between (1) and (2) does not establish that Kant was aware of (1), ofcourse, but is still interesting. As we will see, there is some evidence that he was.

Moreover, Kant went further with these issues than any contemporarytheorist. Unlike theorists of the past few decades, Kant had the makings of atheory to explain (1) and (2). Peacocke (this volume) is the first contemporarytheorist to have anything like such a theory.

Kant called themode of reference to self that gives rise to awareness withoutidentification transcendental designation. In Section 1 we will explore this notionto see what exactly it amounts to and how much Kant knew of the kind ofreference that we use to achieve it. Section 2 lays out Kant’s understanding ofhow awareness of oneself as oneself is different from awareness of one’spsychological states (the latter is what Kant usually meant by ‘inner sense’) andoneself as an object among other objects. Section 3 lays out a further keyelement of Kant’s theory, the notion of what I call the global representation.Kant held that one appears to oneself as a single common subject of a largenumber of psychological states (1781: A350). The notion of the global represen-tation is closely related to that idea. Section 4 draws the elements of the theory

3. Being myself is not a property of me, i.e., something that I and other things could have in common.

12 Andrew Brook

together and in Section 5 we show how it explains (1) and (2). The theory turnsout to have implications for some other things, too, immunity to error throughmisidentification in particular.

1. Transcendental designation: The referential base of self-awareness

Kant’s most common way of describing the awareness of self yielded by tran-scendental designation is this: “… through the ‘I’, as simple representation,nothing manifold is given” (B135). If the kind of reference that yields aware-ness of oneself as oneself, awareness of oneself as subject, is non-ascriptive,then the resulting awareness will not, or certainly need not, present anyproperties of oneself.

If awareness of self as subject is non-ascriptive, the reason, of course, is notthat the self is some strange, indefinable being; as Kant brilliantly discerned, itis because of the nature of the acts of reference used to gain this awareness. Kantspoke of this kind of referring only a few times but when he did, he achievedinsights into it that have only been rediscovered in the past thirty-five years.4

For Kant’s discoveries about reference to self, the crucial text is the passageon A355 of the first-edition attack on the second Paralogism quoted earlier:

in attaching ‘I’ to our thoughts, we designate the subject… only transcenden-tally, without noting in it any quality whatsoever — in fact, without knowinganything of it either directly or by inference.

This doctrine of transcendental designation also appears at B155 of the second-edition Deduction, though not thus labeled.

the I that I think is distinct from the I that it… intuits…; I am given to myselfbeyond that which is given in intuition [B155]

4. Transcendental designation, it is worth pointing out, is purely an epistemological phenomenon.Whether Kant also held that something about the mind itself is ‘transcendental’ is an ontologicalquestion and another matter altogether. Peacocke (this volume) urges that he held to the ontologicalthesis, too, but this is none too clear. Indeed, the nature of the thesis itself is none too clear. PresumablyPeacocke has inmind something that Kant would have called ‘transcendent’, i.e., beyond experience andthe necessary conditions of experience, not what he called ‘transcendental’, i.e. to do with the necessaryconditions of experience. Did Kant think that, in addition to themind of whichwe are aware in, e.g., actsof transcendental designation, there is another, transcendentmindbeyond such awareness? There is littlereason to think that he did (Brook 1994,Ch. 4:5). To be sure, he thought that there is a lot about themind that we cannot know—but here hemeant the ordinarymind of self-awareness and introspection,not some second, transcendent entity.


The central notion is the idea that I can refer to myself (presumably, as myself)using ‘I’ without ‘noting… any quality’ in myself. One can refer to oneself in avariety of ways, of course: as the person in the mirror, as the person born onsuch and such a date in such and such a place, as the first person to do X, and soon, but one way of referring to oneself is special: it does not require identifyingor indeed any ascription to oneself. So Kant tells us. Kant does not specificallysay that transcendental designation makes me aware of myself as myself but heseems to have had it in mind. Indeed, as we will see a bit later, Kant argued thatI must presuppose that it is myself of which I am aware to know some otherthings about myself.5

Did Kant really discover central features of self-reference without identifica-tion or did he merely happen onto something— something that we can identifyin retrospect as self-reference without identification — with no more than avague idea of what he had stumbled upon? Two things make me think that hehad some definite idea of what he had uncovered.

First, he did not refer to the notion only once or in one way. Hemarked thebasic claim in a number of ways and a number of places. He spoke, as we saw,of uses of ‘I’ that designate “only transcendentally”, “without knowing anythingof [the subject]” (A355). But he also said that such uses “denote” but do not“represent” (A382). And, he tells us, even though the referent of ‘I think’ is asubiquitous in experience as the referent of any categorical concept, it is not likeother concepts because “it can have no special designation” (A341=B399). Thatis to say, presumably, it does not pick out its referent as one kind of objectrather than another or as one object rather than another at all. (This last ideawill be central in the fourth section.)6

Second, what if Kant’s work also has the basics of the idea that certainindexicals are essential? That would be a good indication that he understood at

5. One can compare what Kant says about reference to self to his doctrine that existence is not apredicate (A598=B626). In the same way that being aware of something’s existence is not to be aware ofany quality of it, being aware of oneself as oneself is something over and above being aware of qualitiesof oneself. In his criticism of Leibniz’s Amphiboly, Kant says much the same thing about space and time— to be aware of space and time is to be aware of something over and above the qualities of space andtime (A276=B332; see A281=B337).

6. Note that all the phrases just cited are from no earlier than the Paralogism chapter of the first edition.Kant seems not to have developed his theory of reference to self until he needed it to attack rationalpsychology. Note, too, that they are all from the first edition. In the second edition, Kant moved hisdiscussion of awareness of oneself as subject to the Transcendental Deduction — and, regrettably,deleted most of the interesting details. A different but equally stripped-down version can be found in“The Psychological Idea,” §46 of the Prolegomena (1783; Ak. IV:333–4).

14 Andrew Brook

least some aspects of reference to self without identification fairly well, becausethe latter phenomenon requires the former: indexical reference to self could notbe essential unless there is a way of doing such acts of reference that is indepen-dent of (non-indexical) identification.

Whether Kant was aware that, for certain purposes, use of indexicals like ‘I’is essential is not easy to settle with perfect confidence. Such evidence as thereis, however, points to the conclusion that he was. Specifically, Kant argues thatawareness of certain things presupposes awareness of oneself as “subject of thecategories” in a way quite reminiscent of Shoemaker’s claim that:

no matter how detailed a token-reflexive-free description of a person is, … itcannot possibly entail that I am that person [1968:560; this volume, 86].

Here is what Kant says:

The subject of the categories cannot by thinking the categories [i.e. applyingthem to objects] acquire a concept of itself as an object of the categories. For inorder to think them, its pure self-consciousness, which is what was to beexplained, must itself be presupposed [B422].

This passage is from the extremely obscure second edition version of theParalogisms chapter but the phrase ‘its pure self-consciousness’ seems to referto awareness of oneself as oneself, awareness of oneself as subject. If so, what itseems to be saying is something like this: Judgments about oneself, i.e., ascrip-tions of properties to oneself, ‘presuppose … pure self-consciousness’, i.e.,awareness of oneself via an act of ascription-free transcendental designation.We find what may be the same claim in the first edition: “it is … very evidentthat I cannot know as an object that which I must presuppose to know anyobject … .” (A402).

A second passage, this time from Kant’s introduction to the chapter on theParalogisms, offers similar suggestions. The passage begins with a famousvariant on the ‘nomanifold’ theme that we saw at the beginning of this section:

Through this I or he or it (the thing) which thinks, nothing further is repre-sented than a transcendental subject of the thoughts = X.

and then goes on:

It is known only through the thoughts which are its predicates, and of it, apartfrom them, we cannot have any concept whatsoever, but can only revolve in aperpetual circle, since any judgment upon it has always alreadymade use of itsrepresentation [A346=B404].


The last clause is what interests us: “any judgment upon it has always alreadymade use of its representation”. Kant seems to be saying two things. On the onehand, to know that anything is true of me, I must first know that it is me ofwhom it is true. That would seem to require some awareness of myself thatcould not be derived from any knowledge of (what are in fact my) properties.Furthermore (a point that will become vital later), this awareness of myself asmyself is not via some independent representation of myself. The only represen-tations involved are representations of the objects of which I am aware. Aware-ness of self is somehow part of representations of objects. We do not have ‘Ich-Vorstellungen’ of the sort envisioned by Frege or Husserl.

In short, these two passages seem to be saying something like this. In orderto apply the categories to oneself, i.e., in order to make ‘any judgment upon’oneself or know oneself as an object, one must already and independently beaware of oneself as subject, i.e., as oneself. But this is nothing less than the coreof the idea of the essential indexical.

To summarize our results so far. Kant seems to have been aware of twofeatures of reference to self that Shoemaker views as distinctive:

1. Kant was clearly aware of what Shoemaker calls reference to self withoutidentification; in his jargon, we designate the subject “transcendentally, withoutnoting in it any properties whatsoever” (A355); and,

2. There are indications that Kant was also aware of the idea of the essentialindexical. In his terms, awareness of properties as properties of oneself presup-poses awareness of oneself as subject, as oneself.

If so, there is reason to think that Kant did know what he had found when hehit on transcendental designation.

Few of Kant’s students have paidmuch attention to transcendental designa-tion. Doubtless there are many reasons for this. Kant managed to give a clearstatement of what he thought about the topic only once in the whole Critique,namely, on A355. Not only is this description exceedingly brief but he droppedboth it and the term in the second edition. It is also buried in the middle of anobscure andwhatmany, I thinkwrongly, take to be a parochial discussion of thesimplicity of the soul. Perhaps most importantly, the remarkable insights intothemechanics of reference to and awareness of self that Kant sketched there werelost again with his death, to reappear at the earliest with Wittgenstein (1933–4:66–70) in his notion of the use of ‘I’ as subject and probably not until Castañeda

16 Andrew Brook

and Shoemaker.7 Indeed, it would have been difficult for anyone to haverecognized what Kant had spotted prior to Castañeda’s and Shoemaker’s work.Kant himself lacked the apparatus needed to describe his discoveries adequately.

2. The sources of self-awareness

Kant was not merely aware of some of the distinctive features of non-ascriptiveself-awareness. Unlike recent theorists prior to Peacocke, he had themakings ofa theory to explain them. Kant held that one gains awareness of oneself asoneself, as subject, in a way very different from the way in which one gainsawareness not just of external objects but also of one’s own psychological states(and even of oneself when one is an object of one’s own mental states — forexample, when one sees oneself in a mirror). He also held that the awareness ofself that results is different from awareness of anything else in certain respects,respects that explain the resulting peculiarities. Let us start with how onebecomes aware of oneself and what the resulting awareness is like.

Kant’s theory of self-awareness compares and contrasts in interesting wayswith some contemporary views. We have already seen a parallel between Kant’sviews and Peacocke’s. We will return to it. Kant’s views also contrast with somecurrent theories, Rosenthal’s higher-order thought theory and Dretske’sdisplaced perception theory in particular. Since these contrasts bring out someof the originality and power of Kant’s theory, we will look at them, too.

Since the distinctions just canvassed are part of Kant’s doctrine of innersense, we should start with an account of is doctrine of inner sense but wewon’t. His doctrine of inner sense is a mess. Here are just a few of the problems.Kant insists that all representational states are in inner sense, including thoserepresenting the objects of outer sense (i.e., spatially located objects),8 but healso says that the object of inner sense is the soul, the object of outer sense the

7. Though Shoemaker attributes the core of his treatment to Wittgenstein, one wonders. In his laterwritings Wittgenstein seems to have maintained that apparently self-referential uses of ‘I’ and cognatesin fact are not referential at all. It is hard to tell whether he held the same view in his middle period.Perhaps we could put it this way: Kant anticipated what may be one thread in Wittgenstein’s middleperiod work; Shoemaker developed the idea that Kant had anticipated, probably without knowing thatKant had worked on it.

8. “Whatever the origins of our representations, whether they are due to the influence of outer things,or are produced through inner causes, whether they arise a priori, or being appearances have anempirical origin, they must all, as modifications of the mind, belong to inner sense.” (A98–9).


body (including one’s own). He comes close to denying that we can be aware ofthe denizens of inner sense — they do not represent inner objects and have nomanifold of their own. Yet he also says that we can be aware of them —representations can themselves be objects of representations — and thatrepresentations can make us aware of themselves. In its role as a form of ormeans to awareness of self, apperception ought to be part of inner sense. YetKant regularly contrasted apperception, a means to awareness of oneself andone’s acts of thinking, with inner sense as a means to awareness of — what?Presumably, particular representations: perceptions, imaginings, memories, etc.Here is a passage from the Anthropology:

§24. Inner sense is not pure apperception, consciousness of what we are doing;for this belongs to the power of thinking. It is, rather, consciousness of what weundergo as we are affected by the play of our own thoughts. This consciousnessrests on inner intuition, and so on the relation of ideas (as they are eithersimultaneous or successive). [1798, Ak. VII:161]

And these are just the most obvious problems. As I said, Kant’s doctrine ofinner sense is a mess.

Nevertheless, the quotation just introduced contains three interesting ideas.The first starts from the distinction between awareness of oneself and awarenessin ‘inner intuition’ of ‘what we undergo’, i.e., awareness of our representationalstates. One of the things that we undergo in inner intuition is representation ofoneself. If so, there is an important distinction between “apperceptive” aware-ness of self and awareness of self in intuitions. As Kant puts it in the second-edition Deduction,

… the I that I think is distinct from the I that it, itself, intuits …; I am given tomyself beyond that which is given in intuition, and yet knowmyself, like otherphenomena, only as I appear to myself, not as I am… [B155].9

This distinction between the ‘I that I think’ and the I that this I ‘intuits’ is thedistinction between being aware of oneself as subject of all of one’s representa-tional states and being aware of oneself as the object of some of them.

The second idea is the suggestion that apperceptive awareness of self is“consciousness of what we are doing” — doing. For Kant, we become aware ofobjects of representation via apperceptive acts of synthesis tying a manifold of

9. In German, the first part is, “Wie aber das Ich, der Ich denke, von dem Ich, das sich selbst anschauet,unterschieden … .” Kemp Smith translates this, “How the ‘I’ that thinks can be distinct from the ‘I’ thatintuits itself …” and inserts some unhelpful emendations.

18 Andrew Brook

intuition into coherent, recognizable, reidentifiable particulars. However, thestandard way of becoming aware of an act of representing is quite different fromthis.We become aware of acts of representing not by receiving intuitions but bydoing them: “… synthesis …, as an act, … is conscious to itself, even withoutsensibility” (B153); “… this representation is an act of spontaneity, that is, itcannot be regarded as belonging to sensibility” (B132). Kant tells us that herewe do not represent by forming an object. That is not to say that acts ofrepresenting are not themselves represented in inner sense; they certainly are.It is just that they are not the object of a representation.

We can be aware of acts of representing via intuition, too, of course, butwhat is special about them is that we can be aware of them just by doing them,just by representing something. Another Kantian way of capturing what he hadin mind here would be to distinguish between ‘awareness by doing’ and‘awareness by having an image’ (for Kant, we are aware of all intentional objectsin images [A120]— he even thought that all intentional objects are representedspatially [B154–5, B156, B158–9]). The point he is making is that we do notneed to represent acts of representing in images. The passage from B153 citedabove makes that clear.

Equally, we can be aware of ourselves as subject just by doing acts ofrepresenting.When I am aware of myself as the subject of a representation, I amaware of myself as doing the act of representing. One can of course be aware ofoneself via intuition, too — by seeing oneself, for example, in a mirror. This isthe way in which one becomes aware of one’s size, shape, colour, etc. But whenone is aware of oneself as the subject of one’s own representations or agent ofone’s acts, it is by being aware of acts of representing by doing them. Here is apassage from later in the Critique where Kant says this very clearly:

Man,…who knows the rest of nature solely through the senses, knows himselfalso through pure apperception; and this, indeed, in acts and inner determina-tions which he cannot regard as impressions of the senses [A546=B574].

How does awareness of our own acts of representing and of ourselves as theirsubject work? The act of representing makes us aware of three things. Considerthe sentence:

1. I am looking at the words on the screen in front of me.

Kant’s claim seems to be that the representation of the words on the screen is allthe experience I need to be aware not just of the words and the screen but alsoof the act of seeing them and of who is seeing them, namely, me. A single


representation can do all three jobs. In Kant’s words, the awareness of the lattertwo items is given “not indeed in, but with … intuitions” (B161).

Let us introduce a term for this function. Let us call an act of representingthat can make me aware of its object, itself and myself the representational baseof my becoming aware of these items. Almost any representation will do.Imagining Pegasus will do just as well as perceiving external objects such ascomputer screens. Indeed, representational states which have no apparentobject such as pains or feelings of hunger will do just as well. Nor does arepresentation itself have to be recognized to provide a representational base forself-awareness. Just recognizing the object of a representation is enough for meto be aware that it is me who is aware of it. Having the representational base forrecognition of a state is not actually recognizing it, nor indeed myself, but it isto have all the representation I need. This is why, to return to a point madeearlier, the basis of awareness of oneself is not some separate Ich-Vorstellung.

This theory of self-awareness is remarkably powerful. It rests on an idea thatnext saw the light of day in any clear form in Peacocke’s delta theory (thisvolume). In the delta theory, having a property is sufficient in relevant cases forits subject to refer to and ascribe the property to itself — just Kant’s most basicidea. Kant’s theory also neatly avoids some of the problems that afflict otherleading current theories of self-awareness.

Consider, for example, Rosenthal’s (1991) higher-order thought theory.According to this theory, to be conscious of some representational state, A, onemust have another representation (a thought) of which A is the object. ForKant, all we need is A itself. A representation itself has the power to make usaware of it. Rosenthal’s view runs foul of objections such as that lots of crea-tures conscious of their own states don’t seem to have such complicatedthoughts, and that the model readily leads to a regress of thoughts aboutthoughts about thoughts, and there is little independent reason to postulatesuch a hierarchy of thoughts. Now of course Rosenthal has answers to suchobjects — but for what we might call Kant’s same-order model, no suchobjections arise. Advantage Kant.

Or consider Dretske’s (1995) displaced perception view of self-awareness.Here when we are aware of our own representations (and, by extension,ourselves as their subject?), we infer this somehow fromwhatwe are representing,that this item is being represented. Hence ‘displaced perception’. This theory runsfoul of the objection that we seem to know a lot about our representations thatgoes well beyond what is being represented in them, e.g., whether they arestriking or faint, in one sensible modality or another, perception, memory or

20 Andrew Brook

imagination, and so on and so forth. Despite a superficial resemblance engen-dered by the idea that representation of an object is all we need for awareness ofthe representation itself, Kant’s theory is not a displaced perception theory.Indeed, nothing is displaced from the object being represented. Representationof the act of representing and representation of self are distinct activities. It isjust that a single representations provide for all three; we do not need a distinctrepresentations for each act. Once again, Kant is safe from objections that faceothers. Once again, advantage Kant.

Now the third idea introduced in the passage quoted earlier. It is this. Whenone is aware of oneself as subject by doing acts of representing, one is not awareof oneself as an object of representations of any kind. (By ‘object’ I meanintentional object.) Here are some passages: “it is … very evident that I cannotknow as an object that which Imust presuppose to knowany object… .” (A402).“[The representation] ‘I’ is… as little an intuition as it is a concept of any object”(A382). “The proposition, ‘I think’, in so far as it amounts to the assertion, ‘I existthinking’ … determines the subject (which is then at the same time object) inrespect of existence” (B429; second emphasismine). To be aware ofmyself as anobject requires not only “spontaneity of thought”, that is, acts of transcendentalapperception, but also “receptivity of intuition”; that is, it requires “the thoughtof myself applied to the empirical intuition of myself” (B430–31).

Let us spell this idea that awareness of oneself as subject as not awareness ofoneself as an object of a representation out a bit. Kant tended to tie awarenessof objects very closely to sensibility, to appearances and intuitions (see A104);all awareness of objects seems to be via sensibility. But, as Kant put it, “…synthesis…, as an act,… is conscious to itself, even without sensibility” (B153);being aware of an act of representing by doing it is not being aware of it byreceiving intuitions of it. In addition, he says that the “unity of the synthesis ofthe manifold” (i.e., that we are representing the manifold in a single representa-tion) is given “not indeed in, but with … intuitions”. Since one is aware ofoneself as subject by being aware of acts of representing by doing them, not viaintuitions in sensibility (B430), it would follow that one is not aware of oneselfas a represented object. (On the other hand, Kant does say that the subject is anobject, the ‘transcendental’ object of inner sense, at A341 and A361. Because heuses Gegenstand, not Objekt on those occasions, I take him to be talking aboutobjects in a loose way here that does not negate the distinction so carefullydrawn in the passages quoted above.)

For Kant, this distinction between awareness of self by doing acts ofsynthesis and awareness of things as objects is of fundamental importance.


When one is aware of oneself by doing cognitive and perceptual acts, one isaware of oneself as spontaneous, rational, self-legislating, free — as the doer ofdeeds, not just as a passive receptacle for representations: “I exist as an intelli-gence which is conscious solely of its power of combination” (B158–159), of“the activity of the self” (B68) (Sellars, 1970–1; Pippin, 1987).

To close this section, let me situate Kant vis-à-vis one more strand in recentthought about self-awareness, the so-called noncognitivist approach associatedwith the later Wittgenstein. Strangely enough, some recent commentators, forexample Powell (1990), have taken Kant’s assertion that awareness of self assubject is not intuitional awareness of objects to be an assertion that what seemsto be reference to self as subject is not a referential act at all and even that thereis nothing there to refer to. The idea is this. If all (noninferential) awareness ofthings is intuition of objects, then, if uses of ‘I’ are nonintuitional, they areeither nonreferential or do not give us awareness of anything or both. I know ofnothing in Kant that could support such an interpretation and it is not plausiblein its own right. As Shoemaker says, “in making a judgment like ‘I feel pain’ oneis aware of [no]thing less than the fact that one does, oneself, feel pain.”(1968:563; this volume, 90). I think these exotic readings of Kant rest on twomistakes. One is the mistake of thinking that, for Kant, all noninferentialawareness is intuitional. The other is the mistake of confusing non-ascriptivereference with absence of reference.

To summarize this section: According to Kant, we have two ways ofbecoming aware of ourselves — by working intuitions up into intentionalobjects and by being aware of acts of representing and of oneself as their subjectby doing them. The former makes us aware of represented objects, the lattermakes us aware of ourselves as the subject and agent of the act of representing.This distinction anchorsKant’s theory of the peculiarities of awareness of oneselfas subject. Beforewe can lay out that theory, we need to add one further element.

3. The global representation

So far we have focused on individual representations. For Kant, however, therepresentations that serve as the representational base of awareness of oneself assubject are usually much ‘bigger’ than that, i.e., contain multiple objects andoften multiple representations of them tied together into what Kant called‘general experience’.

22 Andrew Brook

When we speak of different experiences, we can refer only to the variousperceptions, all of which, as such, belong to one and the same general experi-ence. This thoroughgoing synthetic unity of perceptions is indeed the form ofexperience; it is nothing else than the synthetic unity of appearances inaccordance with concepts [A110].

Call this general experience a global representation. When I am aware of manyobjects and/or representations of them as the single object of a global represen-tation, the latter representation is all the representation I need to be aware notjust of the global object but also of the global representation itself (the ‘oneconsciousness’ and the various individual representations that I have of theseobjects), and of myself as not just the subject of individual representations butas the common subject of all the constituent representations.

One can be aware of more than one act and object of representing. Howev-er, to become aware of any such acts and/or objects is to integrate themwith theacts and objects of representation of which one is already aware. At any one time,there will be one largest act of representing. That is one’s current global repre-sentation. In short, the global representation is the home of the unity of con-sciousness, a unity that has both a synchronic and a diachronic dimension.10

According to the last paragraph, each of us could have only one globalrepresentation at one time. If so, to use a phrase of Wittgenstein’s, a globalrepresentation has no neighbour; there will not be other, simultaneous globalrepresentations of which one is aware by having them fromwhich to distinguishany one of them. This uniqueness will become important when we draw outKant’s theory of the peculiarities of awareness of self in Section 4.

Howdoglobal representations serve as the representational base of awarenessof self as subject? Here, stated with less than pellucid clarity, is Kant’s account:

the mind could never think its identity in the manifoldness of its representa-tions … if it did not have before its eyes the identity of its act, whereby itsubordinates all [the manifold] … to a transcendental unity … [A108].

Kant seems to mean that what allows me to become aware of my identity as thecommon subject of my various representations is that I can be aware of thesingle, unified and unifying acts of representing by which I combine the objectsof these representations, and sometimes also the representations themselves,

10. It is notoriously difficult to elucidate the kind of unity involved (Brook 1994, Ch. II:5). However,noticing that it has both synchronic and diachronic dimensions can help us sort out where immunity toerror through misidentification could occur and where it could not, as we will see later.


into a single object of a global representation. Kant called these acts transcen-dental apperception. I think he was expressing the same idea when he said in thesecond edition that I am aware of myself as the single common subject of acertain group of experiences by being aware of “the identity of the conscious-ness in … conjoined … representations” (B133).

Put differently, Kant thought that synthesis into global objects is a necessarycondition of awareness of self as common subject.Without objects of represen-tation being tied together as a single complex object of a single representation,one might be aware of the subject of an individual representation but one couldnot be aware of the subject of one such representation as the subject of othersuch representations. Rather, I should have “as many-coloured and diverse aself as I have representations of which I am conscious …” (B134) — as are infact had by me, for I would not, of course, be aware that it was me. It takes aglobal representation to serve as the representational base of awareness of selfas common subject.11

Is a global act of unifying representation also sufficient for being aware ofoneself as subject? No; despite many claims to the contrary, notably by Strawson(1966), Kant was clear that one could represent objects without being aware ofoneself (A113; A117fn.; B132). If one can have representations of which one isnot aware, as was suggested earlier, one could have global representations ofwhich one was not aware — for example, if one’s attention was totally focusedon the complex scene being represented. Even if each global representation isthe full representational base of self-awareness, a direction of attention or somecognitive apparatus necessary for taking advantage of the available representa-tional opportunities might be missing.12 (Something like this might explainwhy nonhuman animals are not aware of themselves as subjects.)

To summarize. One’s global representation is the representational base ofbeing aware of objects, of the representation itself, and of oneself as the com-mon subject of one’s representations. We turn now to the theory of the pecu-liarities of self-awareness that can be built on this base. All the pieces of it are tobe found in Kant.

11. Discussions with Richard DeVidi saved me from some errors here.

12. Kant was aware of attention but discussed it only once in the first Critique (B156 fn.).

24 Andrew Brook

4. Why awareness of self is as it is

The basic idea behind the theory whose pieces can be found in Kant is fairlysimple. When the medium of awareness of oneself is the doing of acts ofrepresenting, the medium imposes sharp constraints on what the resultingawareness can be like. These constraints account for the peculiarities of aware-ness of self as subject identified at the beginning of this paper. That Kantspotted distinctive features of “reference to self without identification” isremarkable enough. That he had the makings of a theory to explain some ofthem is even more remarkable. As we saw in Section 2, the special awarenessthat we have of ourselves as subject is not the only form of awareness of self thatwe have but it is different from intuitional awareness of oneself as an object andthe differences can explain some of the peculiarities of self-awareness.

This theory is neither easy to spot nor simple to unravel. Kant never laid itout completely, indeed he only hints at some of its most important features.Thus any reconstruction of it is bound to be speculative. The key component ofthe theory, not surprisingly, is the global representation. If A355 is the crucialtext for Kant’s view of awareness of self as subject and the kind of reference thatyields it, A108 is the crucial text for the theory of why those things are as theyare. Between them these two pages contain the core of Kant’s whole picture ofself-awareness. Regrettably, they could not be more obscure.

One can be aware of oneself by seeing oneself in a mirror, by acts and statesof inner sense such as feelings, thoughts, etc., being objects of representations,by inferring things about oneself from other things of which one is aware, bybeing aware of oneself through perceptions of one’s body, behaviour, etc. inouter sense (A347), and perhaps in other ways. In all these cases, one would beaware of oneself as the object of a representation.

To be aware of myself as an object, I need a representation devoted to that(intentional) object, a representation of which I rather than something else amthe object. Now return to the earlier claim that any representation that I amhaving can be the representational base of awareness of self as subject, nomatter what its object is.13 If that is right, then all representations of which Iam aware by having them present the same subject to me, namely, me, and in thesameway, asme.A fortiori, the single common subject of a global representation

13. Even some representations that might be somebody else’s by some other criterion seem to providethat possibility. This will complicate the business of immunity to error through misidentification whenwe get to it.


is presented to itself as the same subject throughout this global representation.If so, when one is aware of oneself as oneself, as subject, this awareness is notexperience-dividing, to use a term of Bennett’s — “i.e., [statements expressingit have] no direct implications of the form ‘I shall experience C rather than D’”(Bennett, 1974:80). In a statement such as

1. I am looking at the words on the screen in front of me,the verb expression or the object expression may divide experience but thesubject expression does not. In this, awareness of self as subject is unlike allother awareness.

Why is this kind of awareness of self not experience-dividing? As Kant putsit, “if I want to observe the mere ‘I’ in the change of all representations, I haveno other correlatum to use in my comparisons except again myself” (A366; cf.A346=B404, B422). That is to say, I could not compare this ‘I’ to or contrast itwith what is presented in any representation I am having in which it does notappear. To useWittgenstein’s phrase oncemore, awareness of oneself as subjecthas no neighbour. In no representation of which I am aware by having it doesthe subject appear differently from how it appears in any other.14 If so, I couldnot distinguish the self presented in one such representation from the selfpresented in any other such representation.

We can now explain why, when one appears to oneself as oneself, one is notappearing as the object of a representation. A representation is individuated,differentiated from other representations, by its object. But no representationof mine is made different from any other representation of mine by the fact thatit makes me aware of myself as its subject. To represent something as an objectis to place it vis-à-vis other objects, and usually to ascribe properties to it. If so,to appear in a representation as subject is not to appear as an object of any kind,just as Kant said (A342=B400). (Interesting enough, Shoemaker says somethingsimilar [1968:564; this volume, 90]).

The basis in theory for Kant’s insistence that one’s awareness of oneselfas subject is not via “noting qualities” of oneself can be put this way. We candistinguish the subject from all objects (A342=B400). What we cannot do iscompare it to, contrast it with, one object rather than another. If so, awarenessof self as subject does not distinguish me from or identify me with anything of

14. An obscure remark in the attack on the third Paralogism may be based on the same idea: “in theapperception, time is represented… only in me” (A362). Part of what Kant may have meant here is thatall my representations that locate something in time or are themselves located in time will also representmyself.

26 Andrew Brook

which I am aware as an object, anything in “the world”. Something of greatinterest follows: so far as anything my awareness of myself as subject could tellme, I could be any object or any compilation of objects or any succession of objectswhatsoever. Not by accident are these exactly the topics of the first threeParalogisms. One of the mistakes of rational psychology is precisely to take thesimplicity (lack of manifoldness) in the unified representation of self to be arepresentation of simplicity and unity. Kant, of course, insists that awareness ofself as subject tells us nothing about what the self is like (A355; B156).

5. Putting the theory to work

Let us now put the idea of the representational base and the observations wehave just made about awareness of self as subject to work on some of thepeculiarities of awareness of self as subject discussed at the beginning of thepaper. First, (2), self-reference without identification.

Two questions. First, how is self-reference with identification possible? It ispossible because one always appears to oneself in the same way in awareness ofoneself as the subject of representations of which one is aware by having them.Thus, to recognize oneself as subject does not differentiate the entity thusrecognized from anything else presented as subject. If a representation does notthus differentiate, does not divide experience, then one does not need to ascribeproperties to it in order to achieve reference, indeed unique reference, to it. Ifrepresentation can be without ascription in this way, then it is possible that, asKant put it, “through the ‘I’, as simple representation, nothing manifold isgiven” (B135), that there is a way of referring to oneself that cannot be “accom-panied by any further representation” (B132), in which ‘I’ can only ‘denote’,not ‘represent’ (A382), can designate “only transcendentally” (A355), that can“have no special designation” (A341/2=B399). This awareness is a ‘bareconsciousness’ (A346=B404; B158) that is not knowledge (B157).

Second question. Must self-reference take place without identification? Itmust. If I appear to myself in the same way in every representation of which Iam aware by having it, awareness of self as subject is not just possible withoutascription; it cannot involve ascription. Ascribing properties would produce justthe differentiations that are not there in awareness of self as subject. Now, ifself-awareness requires reference to self, and if self-awareness takes place withno ascription, then reference to self must take place without identification. If so,we have explained (2), self-reference without identification.


And also (1), the essential indexical. The idea of the essential indexical isthat to make references to self via ascribing properties, one must be able tomake references to self that do not ascribe properties. What alternative toascriptive reference is there? So far as I know, only indexical reference. If so,indexical reference is essential to awareness of oneself as subject. Castañeda(1966) seems to hold that this feature of awareness of self as subject is simply abrute fact of self-awareness and cannot be further explained. Had Kant everhave thought about the issue, I conclude, he would have disagreed.

However, this does not explain everything that needs to be explained.Castañeda and Shoemaker argue not only that indexical references to oneselfare essential, but that in them one must refer to oneself as oneself (or cognate).Does Kant have anything to explain this final feature of indexical reference toself? Not that I know of.

The materials that Kant left us can help with some other things, however.First, an element of his account shows that indexical reference using ‘I’ orcognates is essential in a narrower range of cases than Castañeda and Perrythought. They claim that from knowledge of properties by themselves, onecould not know that the properties were one’s own unless one already knew ofoneself and as oneself. Notice that the way in which one is aware of the propertyis not specified here. If we become aware of certain properties in a certain way,the claim does not hold. Consider feeling a pain and becoming aware of it byfeeling it. When I am aware of a pain by feeling it, I know or certainly can knowthat it is a pain of mine. Yet we described the pain non-token-reflexively (‘theperson involved is aware of the pain by feeling it’.) If so, the claim that indexicalreference is essential holds in a narrower range of cases than has been thought.

Indexical reference is essential only in those situations in which how one isaware of a state or event does not settle who has that state or event. Roughly,how one is aware of a state or event settles ownership for psychological stateswhere there are two or more ways of becoming aware of them, namely, byhaving them and by observing and inferring them in various ways and one isaware of the state by having it. The element in Kant’s account that makes thisnarrowing possible is, of course, his distinction between awareness of a state byhaving it and awareness of it as the object of a representation.

This distinction, between awareness by having an experience and beingaware in other ways, can also be put to work to explain— and again to limit—Shoemaker’s notion of immunity to error. This, as we said, is the idea thatcertain judgments are “immune to error throughmisidentification with respectto the first person” (1970:269–70; see 1968:556). What Shoemaker meant, as

28 Andrew Brook

we saw at the beginning of this essay, is that in some situations we cannotbecome aware of a person by being aware of certain experiences, take thatperson to be oneself, and be wrong. To take the most plausible kind of example,I could not be aware of seeing a scene, decide that it is me seeing it, and bewrong. Shoemaker introduced the notion in the context of past-tense memoryjudgments but if it ever obtains, it obtains most clearly in the present. (We willconsider the past in a moment.)

Kant’s account suggests that if I am ever immune to error in identifying aperson as myself, it could only be when I am aware of the person in questionby having the experiences or doing the actions via which I am picking out theperson. Clearly there is no such immunity when I am aware even of myself onany other basis, e.g., by looking in a mirror, seeing a body part, or hearingsomeone discussed.

Indeed, there may be no such immunity even in some cases where I amaware of experiences or actions by having or doing them. Suppose that onecould be aware of a feeling by feeling it when this feeling, as judged by any othercriterion, is someone else’s. Feeling a feeling is one criterion for a feeling beingone’s own but it is not the only one. Perhaps changes in the feeling are causallydependent on another’s body, the feeling has as its object a scene representedfrom the perspective of another person, another person (or at any rate anotherbody) can report on the feeling in exactly the way that I can, and so on. Ifapplication of the various criteria for who is having that feeling produced amixed result of this sort, then Shoemaker would not be obviously right to holdthat we are immune to error through misidentification with respect to the firstperson even about some experiences and actions of which we are aware byhaving or doing them.15 Nor does such a split seem impossible. One couldimagine wiring that let one feel a feeling, perceive a scene, etc., where thesefeelings and perceptions were in every other respect associated with anotherperson. They were seen through the other’s eyes, the other person can reporthaving and feeling them in exactly the same way and on the basis of exactly thesame kind of experience as I can, and so forth. Simply being aware of a percep-tion, feeling, etc., by having it does not render the question of whose perceptionor feeling it is trivial.

15. Powell (1990) raises some doubts about whether such immunity is unique to first-personalreference, referring to Evans’ similar thoughts (1982; this volume). He also challenges Anscombe’sappeal to this immunity as a reason to think that uses of ‘I’, etc., are non-referential. Strangely, in the endhe himself opts for a non-referentialist reading of Kant.


Still further, if immunity to error throughmisidentification with respect tothe first person obtains at all, it obtains only of states, feelings, actions, etc., thatone is currently undergoing. There is no immunity with respect to things thatI remember having or doing. When I remember some thought or feeling oraction from the point of view of thinking or feeling or doing it, I will “automati-cally assume” (Parfit 1970:15) that it is me who thought or felt or did it; I willbe in the same situation as if it were me that I am aware of. Yet, if traces ofautobiographical memories (more exactly, autobiographical q-memories) canbe transferred from person to person in certain ways, this need not be true.Suppose that I (q-)remember seeing the scene from the top of Mt. Robson. Iwill ‘automatically assume’ that it was me who saw that scene. Yet I have neverbeen on top ofMt. Robson. Sally has, however, andmemory-traces set up in herwhen she saw that scene have been transferred to me. If so, it is Sally that I amremembering seeing it, not me. I take it to be me because I remember the scenefrom the point of view from which she saw it but it is still her that I am (q-)remembering. If this analysis is right, there is no immunity to error throughmisidentification in autobiographical (q-)memory.16

The ur-phenomenon for Kant is the manifoldlessness of the self as itappears to itself. Interestingly, this lack of qualitative manifold appears evenphenomenologically. One can easily observe in oneself that having a representa-tion gives one information about that representation (what type of representa-tion it is, how it was formed, what we are representing by it, etc.) but it merelypresents oneself as subject of it. True, the self is presented as me and this mightbe a mode of presentation of a sort (Ezcurdia, this volume) but it is at best anextremely stripped downmode of presentation. Beyond this ‘information’ thatthe thing presented isme, the representation tellsme nothing aboutmyself. Thisbarrenness in one’s awareness of oneself as subjectwas perhaps one of the thingsthat led Hume to think that no subject is to be found in self-awareness at all.

To summarize. Kant seems to have anticipated the idea of the essentialindexical. He unquestionably anticipated the idea of reference to self withoutidentification. And he sketched a most interesting theory of why it must be so.The theory is based on his claim that having a representation, any representa-tion, is the representational base for awareness of self as subject. It seems unlikely

16. Evans (1982:238; this volume, 124–5)disagrees.He thinks that some autobiographicalmemories areimmune to error through misidentification with respect to the first person. It seems to me that he runstwo notions together, namely, awareness of self without identification and immunity to error throughmisidentification. They need to be kept separate.

30 Andrew Brook

that anything like immunity to error throughmisidentificationwould ever haveoccurred to him—but his theory can help us limit and thinkmore clearly aboutthis notion, too. Not a bad record for someone writing over 200 years ago!

Nor have we exhausted Kant’s contribution to our understanding of self-awareness, not by any means. For example, his distinction between awarenessof self as subject via doing acts of apperceptive synthesis and awareness of one’spsychological states via representations in inner sense is very much like thedistinction Evans (1982, 224–35; this volume, 113–22) draws between noniden-tificatory awareness of self and introspection. Again and also like Evans, Kantwas interested in how awareness of oneself as subject and awareness of oneselfas object among other objects connect. For example, how can we know that thething of which we are aware as subject is one of the things of which we areaware as object (B155)? These issues deserves their own unhurried treatment;I have discussed some of them elsewhere (Brook, 1994). I hope that I haveshown here that Kant anticipated some of the most important ideas on self-reference and self-awareness of the past thirty-five years.

Chapter 3

Frege and the first person

Richard DeVidiCarleton University, Ottawa

I

From the outset, the workings of language have been a central concern ofanalytical philosophy. This is not to say that all parts of speech have been judgedequally worthy of attention. One category of expression that suffered relativeneglect, especially before the mid-1960’s, is indexicals — words such as ‘here’,‘there’, ‘this’, ‘that’, ‘now’, ‘then’, ‘you’, and ‘I’ — terms the reference of whichdepends on the context in which they are uttered.1 Perhaps it was tempting tothink of indexicals as Russell once thought of verb tense: as a ‘vulgarity’ ofordinary language we could well do without.2 Like sentences containing tensedverbs, sentences containing indexicals can vary in truth-value depending on thecircumstances in which they are uttered. From Frege to Quine, philosophersinterested in the regimentation of language in the service of science have cast awary eye on the features of natural language that introduce such apparentambiguity. However, developments in the 1960’s made it impossible to contin-ue treating indexicals as an easily dismissed byproduct of the pliancy of naturallanguage. The key moments were the publication of influential papers byHector-Neri Castañeda (this volume) and Sydney Shoemaker (this volume).Quite independently of each other, and approaching the issue from differentdirections, Castañeda and Shoemaker arrived at surprisingly similar insights intothe peculiar features of the first person pronoun (I and its cognates), features

1. This characterization of indexicals is not a definition. A full definition would distinguish indexicalsfrom, for example, anaphoric uses of pronouns: the word ‘she’ in ‘Daria knew she deserved a better fate’is not an indexical. Its reference is determined by the proper name that precedes it. On the other hand,‘she’ accompanied by a pointing gesture in ‘She is the dean of arts’, is an indexical. In short, the contextin question is not the sentence, or even the discourse of which the indexical is a part.

2. The Philosophy of Logical Atomism, David Pears, ed. 1985:117).

32 Richard DeVidi

that set it apart from other singular terms (see the Introduction and Brook, thisvolume). For instance, an utterance of ‘I’ cannot fail to refer. The sentence

I subscribe to Alberta Report

may be false, but not because the word ‘I’ fails to refer. Contrast this with thesentence

Hermes Trismegistus subscribes to Alberta Report,

which is false, not because the legendary Egyptianmagus has failed tomail awayfor a subscription, but because the name ‘Hermes Trismegistus’ fails to refer.

Secondly, Castañeda and Shoemaker pointed out that ‘I’ is immune to errorthrough misidentification of the referent. For example, I might report that

Conrad Black is a sweet-tempered fellow

by becoming confused about exactly who I was speaking with on an occasionwhen I met a number of people. On the other hand, if I say

I am an accomplished pianist

I utter a falsehood, but not because I am mistaken in who I am talking about.Finally, Castañeda and Shoemaker point out that ‘I’ cannot always be

eliminated in favour of a proper name or a description. If I tell you that

Red wine goes to my head

I may well deny that red wine goes to RD’s head, or that red wine goes to thefather of Alex and Natalie’s head, while continuing to allow the truth of myutterance — especially after a few glasses of red wine. This shows that theindexical sentence is not synonymous with the sentences formed by replacing‘my’ with a name or description.

These insights are not without precedent. In the period between theTractatus and the Philosophical Investigations, Wittgenstein seems to have hadat least an inkling of the peculiar properties of the first person pronoun (see TheBlue Book, pp. 66–70). On the strength of the features pointed out above, onemight infer that the first person pronoun has a ‘stickier’ relationship to itsreferent than do other singular terms. What underlies the apparently specialconnection between ‘I’ and its referent? Some have invoked The Blue Book inarguing that the answer to this question is — nothing. In ‘The First Person’, forinstance, Elizabeth Anscombe argues that the special properties of ‘I’ only go toshow that it is not a referring expression at all, but merely a grammatical

Frege and the first person 33

placeholder, like ‘it’ in ‘It is raining’. To suppose otherwise, in her opinion,would be to suppose that there is some special epistemic connection betweenthe person who utters the word ‘I’ and the referent of the word. Anscombe’sposition has attracted few adherents: obviously, the conclusion that the firstperson pronoun is not, despite all appearances, a referring expression is hard toswallow no matter how much one might dislike Cartesianism with respect toself-knowledge. Given that ‘now’ and ‘here’ share many of the peculiarities of‘I’, and that Anscombemakes no similar move to deny that ‘now’ ever picks outa time, nor that ‘here’ ever picks out a place, it is hard to quell the suspicion thather conclusion is ad hoc.

An even earlier precedent for the insights of the 1960’s is to be found inGottlob Frege’s ‘The Thought’ (1918/19), published a little more than a decadebefore Wittgenstein gave the lectures that were the basis for The Blue Book.Wittgenstein knew Frege personally — he visited Frege at his home during theGreat War — and seems to have been well-acquainted with his writings.However, the doctrine set out in ‘The Thought’ could not be further from theview Anscombe teased out of Wittgenstein’s work. Frege takes for granted thatthe first person pronoun is a referring expression, and ascribes to it ‘a specialand primitive sense’ unlike the sense of any other expression.

The view Frege expresses in ‘The Thought’ is an innovation over earlierviews he held on the topic. According to classical views on propositions, it isrecognized that sentences may be context dependent, but this is held to showonly that the propositional content of a sentence may vary from context tocontext. Different utterances of ‘Yesterday was Colville’s birthday’ havedifferent truth values, but this is because the sentence is used to expressdifferent timeless propositions on different occasions, and not because itexpresses a proposition whose truth value may vary from occasion to occasion.Frege apparently wishes to employ his notion of a thought in similar fashion. In‘The Thought’, he says of the words ‘yesterday’ and ‘today’:

If someone wants to say today what he expressed yesterday using the word‘today’, he will replace this word with ‘yesterday’. Although the thought is thesame its verbal expression must be different in order that the change of sensewhich would otherwise be effected by the differing times of utterance may becancelled out. The case is the same with words like ‘here’ and ‘there’ (Frege1918/19: 64).

Frege made a similar point in the manuscript ‘Logik’, written in the 1890’s. Butthere, the suggestion was that the same thought one person could express using

34 Richard DeVidi

‘I’ — for example, ‘I am cold’ — could be expressed by another person ‘usinga name to designate the one who feels cold’ (Hermes et al. 1979:135).

However, with respect to sentences involving ‘I’, Frege appears to havechanged his mind by the time he wrote ‘The Thought’. Immediately afterstating the doctrine for ‘yesterday’ and ‘today’ which corresponds to hisremarks concerning ‘I’ in the earlier work, he says that ‘the occurrence of theword “I” in a sentence gives rise to some further questions’ (Frege 1918/19: 65).Among the considerations he has in mind is the suggestion that certainthoughts involving ‘I’ are incommunicable. Just before this, however, Fregepoints out that a thought expressed with ‘I’ cannot be identical with a thoughtexpressed using a proper name.

The example Frege uses to illustrate this point involves one Dr. GustavLauben; we are to suppose that both Leo Peter and Rudolf Lingens attach thesame sense to the name ‘Dr. Gustav Lauben’, so that for each, the sentence ‘Dr.Gustav Lauben was wounded’ will express the same thought. Moreover, bothPeter and Lingens are present when Lauben says, ‘I have been wounded’; again,both grasp the same thought. However, Lingens does not know that it was Dr.Lauben who said this. So when Peter says at some later time, ‘Dr. GustavLauben was wounded’, Lingens cannot know that the same affair is in question.Therefore, Frege says, ‘the thought which Leo Peter expresses is not the same asthat which Dr. Lauben expressed’ (Frege 1918/19: 65).

Frege does not quite get to, but perhaps deserves some credit for anticipat-ing, Castañeda’s (1966, this volume; 1967) celebrated discovery of the inelimin-ability of indexicals: a sentence containing certain types of indexicals, includingthe first person pronoun, cannot be paraphrased by a sentence with a coreferen-tial proper name (or indexical-free description) replacing the indexical, and yetretain the same content. In any case, by 1918, Frege seems to have realized thathis earlier view is not correct; however, the issue becomes bound up with thenotion that thoughts expressed using ‘I’ are in some sense private, that they canonly properly speaking be grasped by the person who entertains or utters them.

Frege says of Lauben’s utterance of the sentence ‘I have been wounded’ that

… everyone is presented to himself in a special and primitive way, in which heis presented to no-one else. So, when Dr. Lauben has the thought that he waswounded, he will probably be basing it on this primitive way in which he ispresented to himself. And only Dr. Lauben himself can grasp thoughts speci-fied in this way. But now hemay want to communicate with others. He cannotcommunicate a thought he alone can grasp (Frege 1918/19: 66).


This passage stands out as the only place in all of Frege’s writings where hespeaks of incommunicable thoughts and senses.

Frege’s remarks on indexicals in general are few and far between. Theviews expressed in the unpublished logic and in ‘The Thought’ are the onlywritings we have in which Frege directly addresses this category of expressionafter he made the distinction between sense and reference in 1891. ‘TheThought’ is the only place in which Frege demonstrates an awareness that firstperson reference requires special treatment. While the short passage in ‘TheThought’ is arguably quite prescient, subsequent writers have often found itboth odd and striking. Some have taken the passage to be a clear and deeplyrevealing mistake. Others have claimed to find in it the key to developing aworkable version of the theory of sense and reference. Whether as intellectualancestor or as whipping boy, Frege became the interlocutor of choice for manyof the philosophers who pursued the work initiated by Castañeda and Shoe-maker. In short, Frege’s brief remarks on the first person have become grist formore than one philosophical mill.

For instance, John Perry (1977) finds Frege’s remarks in ‘The Thought’ tobe both uncharacteristic and philosophically unattractive. Perry takes them tomark Frege’s realization of fundamental difficulties in his theory of thought andsense. On the other hand, Gareth Evans (1981; 1982, this volume) sees greatmerit in the same passage, and takes it to point the way to a more generalunderstanding of Frege’s theory of thought and sense.

In Perry’s opinion, ‘Frege was led to his doctrine of incommunicable sensesas a result of some appreciation of the difficulties his account of demonstrativesfaces, for these come quickly to the surface when we think about “I”,’ (Perry1977:474).3 On the other hand, Evans takes the emphasis on the ‘the way inwhich [the referent of ‘I’] is presented to himself ’ as providing the prototype fora general theory of sense.

Perry’s criticism of Frege, and the theory he advances to account for thefirst person pronoun and other indexicals, have been influential particularly inAmerica. Since the 1970’s, the judgment of themajority has gone against Frege,and Perry’s work has played an important role in this development. Evans’

3. ‘Indexicals’ is the most commonly used term for the context sensitive expressions under discussion,although other authors have used other terms. Somewriters use the term ‘demonstratives’ to refer to theentire class of expressions. This is a tendency to which, for example, Perry has succumbed. In myopinion, this is poor usage, as the term ‘demonstratives’ is best reserved for the pronouns ‘this’ and ‘that’which differ from other indexicals in generally requiring supplementation by a pointing gesture (ademonstration) if their reference is to be unambiguous.

36 Richard DeVidi

work, too, has been influential; by bringing to light new possibilities for neo-Fregean approaches to the philosophy of language, Evans’ work has served as acounterweight to Perry.

Evidently, Perry and Evans disagree over the merits of Frege’s remarks onthe first person pronoun. This disagreement is partly, but only partly, a matterof divergent interpretations of Frege’s philosophy. There are also deep diver-gences over what would count as an adequate account of thought and language,divergences that come out quite clearly in their respective treatments of the firstperson pronoun. My aim in this paper is to show: first, that despite Perry, it ispossible to frame a Fregean account of our use of ‘I’; and second, that a Fregeanaccount of first person reference need not, and should not, follow Evans’ model.Along the way, I shall try to settle the outstanding exegetical issues betweenPerry’s and Evans’ competing interpretations of Frege.

II

Everyone agrees that by uttering the words ‘I am cold’ a speaker may saysomething and that what she says may be true or false. But an account of thisfact that is consistent with Fregean principles must say more: such an accountmust hold that by uttering the words ‘I am cold’, the speaker expresses aFregean thought, and a thought is supposed to have an absolute truth value —that is, its truth is not relative to where, when, or by whom it is uttered. Athought is supposed to be the sense of a sentence, composed of the senses of theconstituent parts of that sentence. The sense of an expression is supposed to beidentified with its meaning. Sense is also supposed to be the ‘mode of presenta-tion’ of the reference, that is, it is supposed to determine what the reference ofan expression is. Finally, sense is supposed to account for the difference incognitive value between coreferential expressions.

What is not objectionable to Perry is Frege’s basic notion of a thought. Perryaccepts the explanatory necessity of abstract complexes which serve as themeanings of sentences and the objects of mental acts; however, Perry generallyprefers the term ‘proposition’ to Frege’s term ‘thought’. He also accepts acompositional view of propositions which is based on a syntactic analysisultimately owing to Frege. Finally, Perry agrees that propositions are truthbearers, and that their truth values do not vary from time to time, person toperson, or place to place.

What is objectionable, according to Perry, is that Frege demands too much


of his notion of sense: Frege’s single notion of sense cannot simultaneously bemode of presentation, cognitive significance, and thought constituent. Themistake involved in assimilating truth conditions to mode of presentation andcognitive significance is said to become apparent when we turn our attention toindexicals (e.g., ‘I’, ‘here’, ‘yesterday’, etc.).

Perry sets out the problem in the following way. According to Frege’scompositional view of thoughts, the sense of the proper name ‘2’ and the senseof the concept expression ‘ξ is a prime number’ combine to form the thoughtthat 2 is a prime number. The sentence ‘2 is a prime number’ is the completeexpression of a thought. Now consider the sentences

(1) Russia and Canada quarrelled when Nemtsanov defected,

and

(2) Russia and Canada quarrelled today.4

On the Fregean analysis, the clause ‘Russia and Canada quarrelled’, which isarrived at by removing the time specification from (1) or (2), is to be viewed asa concept expression. The clause stands in need of an object name (for adefinite time) in order to become the expression of a thought. Similarly, thesense it expresses stands in need of the sense of an object name in order tobecome a complete thought.

By specifying a time, the expression ‘when Nemtsanov defected’ completesthe sense of sentence (1). Presumably, ‘today’ must do the same for sentence(2). But suppose ‘Russia and Canada quarrelled today’ expresses something trueon August 1 but false on August 2. If ‘today’ has the same sense on August 1 andon August 2, then the sense of ‘Russia and Canada quarrelled today’ must be thesame on both days. But then the sense of the sentence must be incomplete,since Frege denies that a complete thought could change in truth value fromone day to the next. Since Frege also seems to believe that we can get from anutterance of a sentence such as ‘Russia and Canada quarrelled today’ to athought (Frege 1918/19: 64), the indexical must provide a completing sense.

The same considerations arise in the case of the first person pronoun.Assume that an utterance of ‘I am cold’ on a particular occasion expresses a

4. The example is Perry’s (1977:477–479), although it appears to be adapted from an example involvingPrussia and Austria in ‘On Sense and Reference’ (Frege 1892:42n 14, 49). In a number of places Fregequite explicitly states his view that certain sentences of natural language which lack, for example, a timeindication, do not express a thought (see Frege 1884: § 46; Hermes et al.1979:135).

38 Richard DeVidi

Fregean thought. Now, identity of truth-value is a necessary condition on theidentity of Fregean thoughts. But when Bob and Marie each utter the sentence‘I am cold’, the thoughts expressed may differ in truth value. Since Frege deniesthat a complete thought could change in truth value depending onwho expressesit, the sense of the sentence must be incomplete. If so, the utterance of the word‘I’ must contribute a sense that completes the concept expression ‘ξ is cold’.

Perry argues that no such completing sense is to be found but that one isnot required either. Perry assumes that the appropriate completing sense‘would have to be intimately related to the sense of a unique description5 of thevalue [i.e., the referent] of the demonstrative in the context of utterance,’ (Perry1977:485). He offers a number of hypothetical cases which, given this assump-tion, provide compelling reason to reject the Fregean theory of indexicals.Consider, for instance, the case of Hume and Heimson (1977:487–488,491–494):6 suppose that on a particular day in 1775, David Hume says, ‘I amDavid Hume.’ Suppose also that Heimson says, ‘I am David Hume.’ Finally, letus suppose that Heimson is Hume’s twin earth counterpart. Now the sentencethat Hume and Heimson both utter, ‘I am David Hume’, has the same Fregeansense on both occasions, where that sense is understood as a purely qualitativedefinite description. But the thoughts expressed must be different, because theyhave different truth values. Hume’s is true, Heimson’s false. The moral we areto draw from this example is that the Fregean sense is not sufficient to deter-mine which thought is expressed. Of course, this objection to descriptive sensesin the context of the first person pronoun does not apply only to ‘I’. The sameargument can be made with respect to proper names.

However, the case of Lingens lost in the library7 does depend on propertiesparticular to indexicals. Perry asks us to imagine Lingens, an amnesiac lost onthe fifth floor of the Stanford library. He has before him the biography of oneRudolf Lingens, a description as complete as you like — e.g., to the point thatLingens is said to have amnesia and is lost on the fifth floor of the Stanfordlibrary. Lingens might read through the entire biography without ever realizing,

5. Perry takes it that the description involved must be purely qualitative; that is, it must not involve anyproper names or indexicals.

6. I adapt Perry’s cases freely.

7. Perry’s case of the aircraft carrier (1977:483) offers a similar objection involving the demonstratives‘this’ and ‘that’. The locus classicus for objections of this kind — that is, cases designed to make itplausible that, for any description ‘the F’, and any general quality ‘G’, I could believe that the F is G butfail to believe that I am G, even if I am the F — is Castañeda’s case of the editor of Soul (1966:134–135;this volume, ).


‘I am Lingens.’ Howevermuch Lingensmay know about Lingens, he can alwaysbe unaware of the fact that he is Lingens: the question ‘Am I Lingens?’ canalways be raised. With ‘I’, a question of this kind can always be raised, howevercomplete the description might be. Hence, an indexical utterance of the word‘I’ is not replaceable with a definite description.8

A further objection, originally raised in Perry’s 1977 paper,9 is developedat length in ‘The Problem of the Essential Indexical’ (Perry 1979; this volume).Here, the objection involves a hypothetical case in which there is an agentwhose behaviour seems impossible to explain or predict in a way consistentwith the Fregean theory of sense. Consider, for example, the belief I might cometo have if I come to believe that I am inadvertently making a mess in a super-market by spilling sugar out of my cart (Perry 1979:3; this volume, xxx). If Icome to believe that I am making a mess, my belief seems to contain anessential indexical element. This is shown by the fact that no paraphrase of mybelief into nonindexical terms will capture exactly the belief that I am makingamess. Imight circle the aisles following the trail of spilled sugar, trying to catchup with the person who is making a mess; but until I realize that I ammaking amess, I will not take appropriate action. Not only is the first person pronounused to express knowledge that is not reducible to non-indexical propositionalknowledge, the knowledge it is used to express is essential to action.

Perry quite rightly takes these cases to constitute a devastating, even adecisive blow against the possibility of a Fregean account (so understood) ofindexicals. So long as sense is understood as something that it must be possibleto specify with a purely qualitative definite description, it cannot carry out all ofthe roles which Frege assigned to it. No such completing sense is to be found.To see why Perry thinks no such completing sense is necessary, let’s consideragain Bob and Marie and their utterances of ‘I am cold’.

The predicate employed by the speaker in each case is evidently the same;hence, the subject term must be responsible for the difference. However, thesubject term Bob uses to express his thought is the same as the subject termMarie uses to express hers. It is not that the words accidentally sound the same;the first word of each utterance has the same linguistic meaning, viz., that thereferent of an utterance of ‘I’ is the author of that utterance. As we have

8. This is not in principle true of names; see Evans’ discussion (this volume) of names introduced witha definition, e.g., “‘Julius’ is the inventor, whoever that might be, of the zipper.”

9. In ‘Frege on Demonstratives’, Perry’s cases of the bear about to attack and of the meeting at noonoffer the same kind of objection (Perry 1977:494).

40 Richard DeVidi

sameness of linguistic meaning in the presence of variation in referent, linguis-tic meaning cannot, at least in the case of ‘I’, be identified with Fregean sense.

Perry’s solution to this problem involves rejecting the idea that a Fregeansense is required at all in the case of indexicals. When Bob says ‘I am cold’, thereseems to be no need to ascribe to the pronoun ‘I’ a sense that can be graspedand that determines its reference on that occasion of its use. It is enough toknow the meaning of the pronoun in the language and to have heard theutterance. Moreover, each utterance of ‘I am cold’ is to be regarded as express-ing, not a thought as Frege conceived it, but a thought in which the componentcorresponding to ‘I’ is its subject, i.e., something like a proposition as Russellthought of it.

Perry suggests that the passage from ‘The Thought’ quoted earlier, in whichFrege departs from his own doctrine that thoughts are essentially public, a‘common store’ which is transmitted from person to person and generation togeneration, is a reaction to problems of the kind raised by his hypothetical cases.These problems, he says,

… turned on the failure to find a suitable description for the value of thedemonstrative, whose sense would complete the sense of the sentence in justthe right way. If the sense we are looking for is private and incommunicable, itis no wonder the search was in vain. [Perry 1977:489]

If Perry is right to interpret him in this way, Frege is guilty of a rather crudeattempt at dodging the problem idexicals pose for his theory: if the descriptionassociated with ‘I’ is incommunicable, Frege can hardly be expected to produceit. However, as Perry remarks, a private description provides no real advanceover public ones with respect to solving the problem of indexicals. Nor canwhatever threadbare plausibility the view may draw from various other philo-sophers’ doctrines concerning the privacy of ‘the subject of thought’ withstandits extension to other indexicals. Similar considerations apply in the cases of‘now’ and ‘here’, but who would want to claim that these words have special,primitive, incommunicable senses?

III

The obvious response for a Fregean faced with the objections raised by Perry isto deny that the Fregean sense of an expression must be given by, mean thesame as, or be expressible in terms of a definite description. Others may have


subscribed to such a view,10 but nothing in what Frege says requires that sensemust be understood in this way.11 Indeed, the same passage that Perry takes tobe a tacit admission of failure to find a thought-completing description in thecase of ‘I’ is sometimes cited as evidence in support of an understanding ofFrege according to which he denied that the sense of a singular term can alwaysbe expressed with a definite description.12 If Frege is not a descriptivist, a mainpremise of Perry’s argument is false.

This is not yet to show that Perry’s conclusion is false. Unlike propernames, an indexical has associated with it a simple descriptive rule, which, in agiven context of utterance, determines what an utterance of that indexical refersto. Moreover, when we have stated the rule we have stated the conventionalsignificance— themeaning—of that indexical. But such ameaning rule is nota Fregean sense.

The defender of Frege must give up, at least in the case of indexicals, theidentification of sense and conventional significance. Now, there is no questionthat, for example, ‘I’ has a significance that does not vary from one speaker tothe next, that is known to all competent speakers, and that governs linguisticcommunication. It is incumbent upon the Fregean to explain why Perry iswrong to suppose that, at least in the case of indexicals, the notion of sense issuperfluous.

Such an explanation must begin with a distinction Frege himself did notemphasize, namely, the distinction between a linguistic expression regarded asa type, and a particular token of that expression. Suppose you return from a tripto find a message on your answering machine, which consists entirely of anunrecognizably garbled voice uttering the sentence ‘I am having a great time,call me when you get back.’ If you are a competent speaker of English, you willunderstand the sentence recorded on the machine; if asked, you could, forexample, paraphrase it, construct a scenario in which it might be uttered, etc. Tounderstand the sentence in this way— as a type— and to grasp the contribution

10. Carnap, for example, held that the contribution a singular termmakes to the proposition containingit is a descriptive individual concept, construed as an amalgamation of properties (Carnap 1947: §§ 4–9).

11. The chief text, for those who would attribute such a view to Frege, draws on a footnote to ‘On SenseandReference’ (Frege 1892:27n). The passagedoes show that Frege sometimes cited definite descriptionsas embodying the senses of proper names. But there is no passage in Frege’s writings which advances thethesis that the sense of a proper name can always be so expressed.

12. Frege’s foremost interpreter, Michael Dummett, has long campaigned against the view that Fregeheld a description theory of names. See, for example, Frege: Philosophy of Language (1973:97–98,110–111). The objections which Dummett there raises against the descriptivist reading of Frege areindependent of the issue of indexicality.

42 Richard DeVidi

which its constituent, indexical singular terms contribute to its meaning, it issufficient to exercise knowledge of the conventional significance of theindexicals. Under these circumstances, it would make no sense to ask whospoke the sentence, where ‘here’ is, or even who is being addressed. However,when you receive the message, what you are really interested in is how you areto understand this particular utterance of the sentence — the token sentence.But to understand the particular utterance, while it is perhaps necessary, it ishardly sufficient to understand the conventional significance of the singularterms involved. To understand the utterance, more than a grasp of the conven-tional significance of the indexicals is required, for these do not by themselvesprovide enough for one who grasps them to determine who left the message orwhere ‘here’ is.

Indeed, anyone who allows that an indexical sentence can be used toexpress something which is true or false — whether it is called a statement, aproposition, or a thought — must allow that the particular utterance has asignificance, a kind of meaning, distinct from its significance as a type expres-sion. And the contribution the indexical makes to the expression of the thought(statement, or proposition) must itself be a kind of meaning. It is this point thatcommits Perry to holding that the reference of an indexical— the contributionit makes to a proposition it is used to express— is a kind of meaning. The samepoint also establishes that there is, contra Perry, theoretical space for Fregeansense in an account of indexicals.

The meaning of an indexical expression considered as a type expression,what Perry calls its role, is not sufficient to account for the meaning of particu-lar, token utterances of the expression. What is missing, in the case of theanswering machine message, is the means to identify the one who uttered it. ‘I’cannot have failed to refer to the person (whoever it was) who uttered thesentence. However, it is clearly possible for the person who heard the messageto fail to understand just what was said.

IV

At the beginning of the previous section I mentioned that Frege’s remarks on ‘I’have been cited in support of a non-descriptivist interpretation of his theory ofsense. One such interpretation has been advanced by Gareth Evans. Frege, inthe passage at issue, explains the use of ‘I’ in terms of ‘a special and primitiveway’ in which ‘everyone is presented to himself ’ and in which he is given to no


one else (1918/19: 66). Evans takes this to underwrite a reading of Fregeaccording to which the metaphorical language of ‘modes of presentation’ is tobe replaced with more literal talk of ‘ways of thinking’, that to grasp the sense ofan expression is to think of its reference in a certain way (Evans 1982:16).13

What Frege says about incommunicable thoughts does little to discount it.Indeed, it seems quite natural to say that the way I think of myself is differentfrom the way anyone else thinks of me, and that if it is this way of thinking thatis a constituent of thoughts I have about myself, it is no wonder that no one elsecan think them.

That this conception of thoughts is uppermost in Frege’s mind when hemakes his remarks about ‘I’ is borne out by the two paragraphs which precedeit. In these paragraphs, Frege gives an example intended to show that differentthoughts may be obtained from the same sentence, even a sentence which isapparently unambiguous and which contains no indexical words. Frege gives usthe case of Dr. Gustav Lauben, whom Leo Peter identifies by his presentoccupation and residence, but whomHerbert Garner identifies by his date andplace of birth. According to Frege, this entails that each of them expresses adifferent thought when he says ‘Dr. Gustav Lauben has been wounded’. Fregegoes so far as to say that in this case, ‘Herbert Garner and Leo Peter do notspeak the same language’ (Frege 1918/19: 65). The picture of sense which Fregegives in this case is one in which the senses of proper names vary, strictlyspeaking, from speaker to speaker, and where there is no community-widesense but only a community-wide reference, and in general, names are peculiarto idiolects.

Here, Frege views the sense that a speaker attaches to a singular term asconstituting the manner in which its reference is given to him, that is, his meansof identifying it or picking it out. In accordance with this, Evans wishes toinclude in an account of the sense of the first person pronoun a description ofhow a subject thinks of himself. Perry also acknowledges the need for ‘ways ofthinking’. For example, when Bob says ‘I am cold’, on Perry’s account Bobhimself is a constituent of the proposition expressed. Some way of distinguish-ing Bob’s epistemic state from Marie’s when she says, of Bob, ‘He is cold’ isrequired, for each expresses the same proposition. Perry holds that Bob’s way of

13. Evans appears to give little weight to Frege’s oft-expressed antipsychologism concerning sense. InEvans’ view, the sole purpose of introducing the notion of sense was to account for certain notions weemploy in ‘ordinary propositional attitude psychology,’ including cognitive value, knowledge, thought,understanding, and so forth (Evans 1982:13, 18, 19, 24).

44 Richard DeVidi

thinking the proposition is distinguished from Marie’s in that he thinks it byentertaining the meaning in the language, the conventional significance, of ‘I’.Evans rejects this proposal, on the grounds that it makes no acknowledgmentof the need to explain what it is to think of a person as oneself. He dismisses as‘neither necessary nor sufficient’ the suggestion that ‘self-conscious thoughtdepends upon the inner exploitation of certain public linguistic devices.’

In The Varieties of Reference, Evans provides a framework for reference inwhich reference to self must be incorporated. Evans considers Frege to havedeveloped the first systematic theory of language, and presents his own accountof reference as a development, with criticisms and emendations, of a Fregeanmodel. Evans’ framework is informed by his adherence to two fundamentalprinciples. The first of these he calls Russell’s Principle: that in order to refer toan object one must know which object it is, and to know this is to be able todistinguish that object from all others. The second principle is the GeneralityConstraint, according to which an understanding of a thought ‘a is F�’ presup-poses an understanding of what it is for other objects to be F, and for a toinstantiate other properties.14

Evidently, it is Russell’s Principle that is at the basis of Evans’ criticism ofPerry. Evans equates the identifying knowledge spoken of in the principle withFregean sense. For Evans, there are no utterances of singular terms that do notcarry a Fregean sense, understood as the way in which the object is presented bythat use of the term, and it is an illusion to believe, as Perry does, that there areuses of a term that serve only to introduce the bare referent of the term. Someterms — definite descriptions and names introduced by a stipulation — havedescriptive senses. But all other terms, including most names and all indexicals,are said to be Russellian singular terms, in that they refer without the mediationof a descriptive sense. Evans says that ‘in order to understand a referringexpression used in this way, the hearer must link up the utterance with someinformation in his possession,’ (Evans 1982:305).

In the case of the first person pronoun, Evans holds that my capacity tothink of myself as ‘I’ is based upon my capacity to identify myself as an objectof a certain kind. Evans is not a Cartesian: he does not base first person refer-ence on a special awareness of some internal self, directly apprehended. Rather,

14. The Generality Constraint accords well with Frege’s Context Principle and his views on thecompositionality of thoughts. Russell’s Principle, in particular its attribution to Frege, has been thesource of some controversy. Evans takes it to imply the existence of the object referred to; but Fregeclearly allows, in ‘On Sense and Reference’ and inmany other places, that there are terms that have sensebut lack a reference.


he treats it as securing reference to an external object, a person, on the basis ofinformation about that object. What underwrites first person reference is thevarious ways I have of gaining knowledge of myself as a spatiotemporal objectthat is an element of the objective order, on the basis of our outer perception,and also, Evans says, through a

general capacity to perceive our own bodies, although this can be broken downinto several distinguishable capacities: our proprioceptive sense, our sense ofbalance, of heat and cold, and of pressure (1982:220).

Together these different ways I have of knowing myself enable me to identifymyself and distinguish myself from all other objects.

V

There is something wrong with Frege’s account of the first person that Evansdoes not acknowledge. Frege maintains that, when, for example, Dr. GustavLauben thinks to himself ‘I have been wounded’, his thought is one which noone but he can grasp. If, however, he says out loud to someone else, say LeoPeter, ‘I have been wounded’, themode of presentation in which he is presentedto himself will be one which is available to other people to grasp. FollowingDummett (1981:123) let’s call these two uses of ‘I’ the use of ‘I’ in soliloquy andthe use of ‘I’ in communication. As Dummett points out, it is quite implausibleto suppose that the word ‘I’ expresses a different sense when used to say tooneself, ‘I have been wounded’, and when the same words are directed atsomeone else. Suppose Lauben tells Peter of the events in which he was in-volved, and in the course of the telling he says, ‘At that moment I said tomyself,“I have been wounded”.’ The reported utterance must have employed the ‘I’ ofsoliloquy; but those who hear the story will have no difficulty knowing whatthought Lauben had.

How can they know this? Evidently, the hearer’s way of thinking of Laubencannot be just the same as the way in which he thinks of himself. But if Laubenexpresses a mode of presentation that is distinct from themode of presentationthat is part of the thought Peter understands, the rule that the mode of presen-tation is a constituent of the thought that is expressed and grasped would seemto be called into question. Neither Frege nor Evans shows any willingness togive up this principle. Instead, they grasp the nettle and hold that Peter doesnot, strictly speaking, understand what Lauben said, for only Lauben can grasp

46 Richard DeVidi

the private and incommunicable sense of his own utterance of ‘I’.Perry is quite right to point out that the notion of an incommunicable sense

is out of character with the rest of Frege’s views on language. Throughout hiswritings, from theGrundlagen (1884) up to and including ‘The Thought’, Fregeinveighed against ‘psychologism’. Fregemaintained an extremely sharp distinc-tion between the subjective and the objective, and harshly criticized views thatidentified the content of an assertion with anything subjective, that is, with thecontents of themind. His chief objection to such views is that such contents areincommunicable. Even in ‘The Thought’ Frege emphasizes that it will not do tofirst explain what it is for a single individual to attach a particular sense to anexpression— for example, the Pythagorean theorem (cf. Frege 1918/19: 68)—and then state what it is for her to be correct in doing so, as far as the languagespoken by the community is concerned. Tomaintain that thoughts are commu-nicable, what is required is an explanation of what it is for a word to have acertain sense in the common language, and an explanation of what it is for anindividual speaker to understand it rightly or wrongly. This view is in sometension with the claim that ‘I’ thoughts are incommunicable.

Whether the doctrine of incommunicable senses is a lapse or a settled viewthat requires emendation, Frege himself provides the resources for an alterna-tive account of the matter. When Frege requires a non-metaphorical — andnon-psychologistic— account of sense, he appeals to the notion of a condition,something like an ideal procedure for determining its reference. AlthoughGrundgesetze, vol. I §32 (Frege 1893), is the only passage in which Frege uses theword ‘condition’, he expounds the very same conception in numerous places.

On this conception, the aim of a theory of sense will be to specify theknowledge a speaker or hearer must have, in virtue of her command of thelanguage, if communication is to be possible. There is no reason to think thatthe thought processes of the individual must play a primary role in such anaccount, rather than, say, the practices of the community. In order to knowwhat thought Lauben expresses when he says ‘I have been wounded’, whetherto himself or others, no more is required (apart from knowledge of the predi-cate) than knowledge of the conventional significance of ‘I’ plus knowledge ofthe relevant features of the context, for example, who uttered it. Knowledge ofthe conventional significance of ‘I’ does not, by itself, constitute the grasping ofa Fregean sense. At most, it can yield an understanding of the sentence in whichit occurs as a type, which is not the same as understanding the sentence as atoken utterance . To grasp the thought expressed by a sentence in which the firstperson pronoun occurs, one must also recognize the particular person who


utters it — whether oneself or someone else — where ‘recognize’ connotes nomore than a bare recognition that that person made the utterance. That is allthat our knowledge of the public language demands.

VI

Had Frege recognized the distinction between speaker’s meaning and linguisticmeaning, perhaps he would not have had recourse to incommunicable senses.This distinction is most obviously required in order to explain individualspeakers’ imperfect knowledge of their language. It is also required to accountfor the connections which a speaker makes between a name and its bearerwhich he does not suppose to underlie its use in the language and which are notof a kind to do so. If we are concerned with the precise content of a belief thatsomeone expresses by means of a sentence, it is the individual’s understandingof the words that determines this — not the meaning of the sentence in thecommon language.

Once the distinction between speaker’s meaning and linguistic meaning ismade, it becomes possible to recognize two components in the explanation ofthe understanding, by a speaker or hearer, of a thought expression involving thefirst person pronoun. A thought, considered as the content of a belief, cannotbe identified with the sense of a sentence of the common language, but ratheris to be identified with the subject’s personal understanding of it. Evans’ appealto self awareness is not necessary to account for the former; but it may, and iscertainly intended to play a role in the explanation of the latter.

To conclude, I will say a few words about why, even considered as part of atheory of beliefs, Evans’ account of first person reference is suspect. Evansrejected Perry’s suggestion that the conventional significance of ‘I’ plays anyrole in self conscious thought as neither necessary nor sufficient. He emphasizesinstead one’s capacity to exploit the various ways one has of gaining informa-tion about oneself, both through outer perception and through our propriocep-tive sense, sense of balance, and so forth, in order to refer to oneself. Evans isnot unaware that such an account, which models first person reference onperceptual demonstrative reference, is vulnerable to certain standard objections(Evans 1982:215–216; this volume, 105). ‘I’ has certain peculiar features: one’suse of ‘I’ is immune to reference failure and to failure through misidentificationof the referent. Moreover, competent use of ‘I’ can survive both the total loss ofbeliefs about oneself, for example, one’s nature, history, and spatiotemporal

48 Richard DeVidi

location, and even the acquisition of massively false beliefs about oneself. Itseems that self-conscious thought cannot depend upon information supplied byperception of one’s environment and body, for it can go on in the absence ofsuch things.

In response to this objection, Evans maintains that it is necessary that thesubject at least ‘be disposed to have such thinking controlled by informationwhichmay become available to him in each of the relevant ways’ (1982:215–216;this volume, 105). Any attempt at showing that such an epistemological under-pinning is not essential to first person thought is therefore subject to theresponse that it has confused the possibility of a non-activated disposition withno disposition at all. While the appeal to dispositions provides Evans with a tidyresponse to a serious objection, it is less than evident that giving such disposi-tions the role Evans demands of them can be motivated independently of theneed to defend the general view. In particular, it is not clear how the fact thatone is disposed to have his thinking controlled by information which maybecome available could explain how it is that one succeeds in identifyinghimself when no such information is available. The fact that I am disposed tohave my thinking controlled by information received from an oncomingbaseball does little to help me identify it when my eyes are closed.

Let’s suppose, contra Evans, that the indexical rule governing ‘I’ does playa role in self conscious thought. The indexical rule by itself seems sufficient toexplain why competent use of ‘I’ is guaranteed against failure throughmisiden-tification or reference failure, and why it can survive both the loss of beliefsabout oneself or the acquisition of massively false beliefs about oneself. Theindexical rule can, without appeal to non-activated dispositions, account for thefirst person pronoun’s propensity to stick to the subject. To say this is not toclaim to have obviated the need for an account of the various avenues throughwhich we attain information about ourselves — a complete account of selfconsciousness would require it. But if our concern is first person reference, asexpressed publicly or as it occurs in thought, the indexical rule seems to leavethese ways in which one attains information about oneself on a par with theways in which one recognizes any other speaker.

Section II

Classical Contributions

Chapter 4

‘He’

A study in the logic of self-consciousness*

Hector-Neri CastañedaUniversity of Indiana

Introduction

The word ‘he’ is a device for talking about persons, beings who enjoy, evenindulge in, self-awareness. The word ‘he’ and the phrase ‘he himself ’ aresometimes used to refer to the entity known, thought of, by the person whoknows, thinks of, himself. We say, e.g., “He believes (knows, says, argues,claims) that he (himself) is healthy (rich, tall, heavy, Napoleon, a victim).” Thisuse of ‘he’ (to be called the S-use of ‘he’) as a pointer to the object of someone’sself-knowledge, self-belief, self-conjecture, is the main topic of this study.1

* This essay was concluded while the author was doing research supported by the National ScienceFoundation Grant No. G.S. 828. [Editors’ note: With some misgivings, we have strictly retained theoriginal wording, punctuation and use of quotation marks except when it is clearly in error.]

1. The main topic of this essay is almost brand new. My first glimpse of its complexity occurred whenI was trying to formulate what I called “Meaning postulates of ‘pain’” in ‘The Private LanguageArguments’ in C. Rollins (ed.)Knowledge and Experience (University of Pittsburgh Press, 1963). But I amdeeply indebted to Jaakko Hintikka for having fully awakenedme to this topic. For my critical review ofhisKnowledge and Belief (Cornell University Press, 1962) for The Journal of Symbolic Logic, Vol.29, No.3(September 1964), I had to study his formal treatment of statements expressed in sentences of the form“The man who is in fact a knows that he is a.” Hintikka points out (p.159) that “Ka(a=a)” is not asymbolization of that sentence form. He proposes to symbolize it as “($x)(Ka(x=a))” (p.159). Thissymbolization is satisfactory inasmuch as Hintikka has, very ingeniously, taken the free but bindableindividual variables occurring in contexts of the form ‘Ka(… )’ to range over objects or persons knownto a. Thus, ‘($x)(Ka(x=a))’ can be read as ‘There is a person known to a such that a knows that such aperson is a.’ There is, surely, a sense of ‘knowing a person’ in which for a to know the person who is infact is a is to know himself.

Yet Hintikka’s very reading makes his calculus inadequate to handle several important statements.It cannot handle, e.g., contingent statements expressed in sentences of the form ‘There is an object suchthat a does not know that it exists.’ Here the individual variable ‘it’ occurs free in the clause ‘such that adoes not know that it exists.’ Hence, by Hintikka’s reading, that ‘it’ must refer to an object known to a,such statements would be self-contradictory. For another difficulty let ‘a�’ stand for ‘The Editor of Soul.’

52 Hector-Neri Castañeda

My purpose here is to provide an exhaustive discussion of, and a rigoroustreatment of the logic of, the S-used third-person pronoun. My major conten-tions are: (a) that the S-uses of ‘he’ are quite different from the other uses of thethird-person pronoun; (b) that the S-uses of ‘he’ constitute the employment ofa unique logical category, which is not analyzable in terms of any other type ofreferringmechanism (i.e., the other uses of ‘he’, other personal pronouns, propernames, demonstratives, and definite descriptions); (c) that in each sentencecontaining tokens of the S-used ‘he’ there is at least one such token which is notanalyzable, but there may be other tokens which are analyzable in terms of anunanalyzable tokenof the S-used ‘he’; (d) that the complex logic of the S-used ‘he’is governed by the principles (H*), (H*1)–(H*3), whose formulation is a highpoint of the paper; (e) that the first-person pronoun is also an unanalyzablecategory, even though some tokens of ‘I’ can be analyzed in terms of someunanalyzable token of ‘I’, in accordance with principle (P); (f) that the widelyaccepted rule of detachment ‘From “x knows that p” one is allowed to infer thatp�’ breaks down when the statement that p is expressible in a sentence contain-ing an S-used ‘he’; (g) that a valid substitute for the above rule is (K*).

The results of this investigation have important consequences for thephilosophy of mind, which will be discussed in a separate paper. In yet anotherpaper I will defend theses parallel to (a)–(g) for the case of expressions like‘then’ and ‘there’ when used in oratio obliqua, for instance, in “At place P andtime t x believed that it was then raining there”.

Suppose now that Smith has never seen his image or pictures in photographs, mirrors, ponds, etc.Suppose that at time t Smith does not know that he has been appointed the Editor of Soul and that at the comes to know that the man whose photograph lies on a certain table is the new Editor of Soul,without Smith realizing that he himself is theman in the photograph. In this situation, “There is a personsuch that the Editor of Soul knows that that person is the Editor of Soul,without the Editor knowing thathe himself is that person” is true. This statement cannot, however, be symbolized in Hintikka’s calculus.The obvious candidate ‘($x)(Ka(x=a)&~Ka(x=a))’ is strictly a formal contradiction. What is needed issomething like ‘($xa)(Ka(x=a)&~Ka(x=himself)),’ where the expression ‘himself ’ has ‘a�’ as its logicalantecedent. This is precisely the initial insight into the peculiar syncategorematic character of thepronoun ‘he*’ that led to the claims and principles put forward in Sections 4 and 5 below.

Hintikka’s calculus (together with his ingenious reading of free variables) is, furthermore,inadequate to handle the complexities of the pronoun ‘he*’. See footnote 15 below.

Some time after I had finished this paper I came across P. T. Geach’s “On Belief About Oneself,”Analysis, vol. 18 (1957), pp. 23–4. Here Geach formulates three important things. First, “a=b and bbelieves that b is Φ” entails “a believes that b is Φ”. Second, “a=b and b believes that b is Φ” does notentail “a believes that a isΦ”. Third, “a=b and b believes that he himself isΦ” does entail “a believes thathe himself isΦ”. Thus, Geach should be credited (as far as I know) with having posed the problem of thelogic of the pronoun ‘he*’ for the first time.

My greatest debt is to P. T. Geach who saved me from several errors and suggested objections andcorrections.

A study in the logic of self-consciousness 53

1. The ‘he’ of self-consciousness

Let us, first, demarcate our field of enquiry. We want to study the uses of ‘he(him, his)’ and ‘she (her, hers)’ in which the pronoun refers to the object of aperson’s knowledge, beliefs, thoughts, assertions, about himself. We shall, forsimplicity, speak of the ‘he’ of self-consciousness, or of the S-use of ‘he’, or ofthe pronoun ‘he*’. We shall put an asterisk after a form of the third-personpronoun to indicate that it exemplifies an S-use. The ‘he*’ pronoun appears, forinstance, in sentences like ‘Arthur believes that he* is happy’ and ‘Mary claimsthat she* knows that Paul loves her*’. For simplicity we shall concentrate ourdiscussion mainly on statements about cognitive attitudes or acts, but ourinvestigation also applies to linguistic acts which attribute a self-reference tosomeone.We shall speak of cognitive verbs to refer to verbs expressing cognitiveacts or attitudes or dispositions (‘think,’ ‘believe,’ ‘know,’ ‘suppose,’ ‘infer’) aswell as to linguistic acts of the assertive or quasi-assertive kind (‘claim,’ ‘hold,’‘state,’ ‘say,’ ‘deny,’ ‘argue,’ etc.)

It is only a linguistic freak that ‘he’ in the sense of ‘he*’ looks exactly like thethird-person pronoun ‘he’, which occurs, for instance, in ‘Arthur came, but heknew nobody he saw; he left early.’ This can be seen simply by glancing at theother uses of the third-person pronoun. But before taking this glance we shallintroduce other simplifying conventions.

1. Single quotes will be used to form names of sentences or expressions.

2. Indentation of, as well as double quotes around, a sentence will be em-ployed to form a name of some statement that has, or could have, beenmade bymeans of a normal utterance (in normal contexts and with its ordinary mean-ing) of a token of the sentence indented, or quoted. We shall assume that thename so produced names the same statement throughout the present investiga-tion, but we shall assume nothing about the method for picking out thestatement in question.

3. Numerals prefixed to indented sentences will sometimes refer to theindented sentence and sometimes to the statement formulated with thatsentence; the context will make clear which one is meant.

4. We shall sometimes speak of ‘he’ (or ‘I’) as short for ‘a pronominal expres-sion used to formulate third-person (or first-person) reference.’

5. We shall use ‘tokenn’ and ‘token in the narrow sense�’ to refer to each of thetokens of each form of a pronoun or verb; e.g., the tokensn of ‘I’ are not, but


contrast with, the tokens, of ‘me’; similarly, a tokenn of ‘runs’ is not, butcontrasts with, a tokenn of ‘run’ or ‘ran’.

6. We shall call all tokensn of a pronoun or verb, regardless of inflection,tokensw, or tokens in the wider sense of the pronoun in question. Thus, eachtokenn of ‘I’, each tokenn of ‘me’, each tokenn of ‘my’, each tokenn of ‘mine’,and each tokenn of ‘myself ’ are all tokensw of the pronoun ‘I (me, my, mine,myself)’ or simply, ‘I’.

It seems that, aside from its S-uses, the pronoun ‘he (him, his, she, her, hers)’is normally used in a way that falls in one or other of the following categories:

A.�The third-person pronoun is sometimes a proxy for the ostensive demonstra-tive description ‘that (this) man’ or ‘that (this) woman’. This is probably the casewhen someone says, e.g., “Look, he is dragging her”. The pronoun is sometimesa proxy for a quasi-ostensive demonstrative description. This is the case, forinstance, when a person sees a picture (photograph, bust, replica) of anotherperson and asserts, e.g., “She is beautiful!” Here the uttered tokenn of ‘she’simply stands for the description ‘the woman whose picture (photograph, bust,replica) this is’.

B.�‘He’ is perhaps used sometimes as an (ostensive) demonstrative pronoun, i.e.,as a substitute for no definite description, but merely to point to a personsingled out from the remaining objects present in one’s current experience. Inthis use the third-person pronoun is merely a colorful proxy for ‘this (that)’.

C.�‘He’ is used sometimes as a mere part of a universal quantifier, as, e.g., whenone says “He who marries young…” The pronoun ‘he’ is also used as a merevariable bound by a universal, or existential, quantifier to which it refers back.Examples of these uses are, respectively, found in normal utterances of sentenc-es of the form ‘Anyone who marries young is such that he…’ and ‘Somebodycame when I was out and he returned my book’.

D.�The third-person pronoun is frequently employed as a relative pronoun, i.e.,as a proxy for a name or description which precedes or follows it. An exampleis furnished by the assertion “If Arthur comes late, he (i.e., Arthur) will call.”Clearly, this statement is the same as the statement that one would make, in thesame circumstances, by asserting “If Arthur comes late, Arthur will call” or byasserting “If he (Arthur) comes late, Arthur will call.”

E.�‘He’ is, perhaps occasionally, a proxy for the definite ostensive description‘that (this) body’. This would be the case if the sentence ‘He weighs 185


pounds’, for instance, were used to formulate the statement “That (this) body[pointing to the body of some person] weighs 185 pounds.”

F.�‘He’ is often employed as a place-holder for some unspecified descriptionwhich refers to a previously mentioned object. This is typical of constructionsin oratio obliqua, i.e., clauses subordinated to cognitive verbs in the sensecharacterized above. Examples are “Paul said (believes, knows) of (someonewho in point of fact is2) Mary that she is happy.” These statements must bedistinguished from the statements “Paul said (believes, knows) that Mary ishappy.” The former are (nearly) the same as the statements: “There is a proper-tyΦ such that Mary is the only person who isΦ and Paul said (believes, knows)that the only person who isΦ is happy.” The most characteristic pattern of the(F)-‘he’ is ‘… believes (knows, thinks, asserts, holds) of y that… he (she)…’, asjust illustrated. But Helen Cartwright3 has furnished an example which devi-ates from this pattern, viz., “Paul saw Mary and believes that she is happy.”

G.�The pronoun ‘he’ is often used to indicate what Russell called the largerscope of a description. For instance, “If the author of Principia Mathematicaremembers it, he will write to you about it” is analyzable (as Russell thought) as,or merely presupposes (as Strawson wants it), “There is just one author ofPrincipia Mathematica such that: if he remembers it, he will write you about it”(where the new occurrences of ‘he’ are variables of quantification). On the otherhand, “If the author of Principia Mathematica remembers it, the author ofPrincipia Mathematica will write you about it” seems to be analyzable as, or topresuppose, “If there is just one author of Principia Mathematica and heremembers it, then there is just one author of Principia Mathematica and he willwrite you about it” (where again the new occurrences of ‘he’ are variables ofquantification).

Here we are not claiming that these six uses of ‘he’ are really all distinct andnon-overlapping.We are simply not interested in examining them for their own

2. This parenthetical clause was inserted on the suggestion of Professor Wilfrid Sellars. I hope that itmakes it clear that I do not claim that the sentence ‘Paul believes of Mary that she is happy’ is never usedin ordinary speech to make the same statement that is more properly made with the sentence ‘Paulbelieves that Mary is happy’. I claim only that the former sometimes is used, or can at any rate be usedto make a statement such that (1) it does not imply that the proposition Paul takes to be true is “Maryis happy”, and (2) it implies that Paul takes as true some proposition of the form “Z is happy”, where ‘Z’stands for some way he uses to refer to a certain person, who happens to be Mary.

3. At the October 13th, 1964 Wayne State University Colloquium, at which the earlier version of thispaper was discussed.


sakes.We want to show only that they are quite different from the S-uses of ‘he’.Suppose that a man called Privatus informs his friend Gaskon that

(1) The Editor of Soul knows that he* is a millionaire.

It is immediately clear that Privatus’ use of ‘he himself ’ or of ‘he*’ is not aquantifier or a variable referring back to a quantifier. Hence, Privatus’ use of‘he*’ is not an instance of use (C). But as P.T. Geach pointed out to me, ‘he*’can also be an instance of use (C). It plays this role in “Someone thinks that he*is a genius.”

It is extremely doubtful that the Editor of Soul or Privatus or Gaskon thinkof a mere body as a millionaire. But even if they all did, we may suppose that inthis case they all think of persons. Thus, Privatus’ use of ‘he*’ is not a proxy for‘this (that) body’.

We can, thus, disregard uses (C) and (E) entirely.The tokenn of ‘he*’ in (1) is not a proxy for ‘the Editor of Soul�’. If it were

statement (1) would be the same statement as:

(2) The Editor of Soul knows that the Editor of Soul is a millionaire.

But (2) is not the same statement as (1). For (1) does not entail (2). TheEditor of Soul may know that he himself is a millionaire while failing to knowthat he himself is the Editor of Soul, because, say, he believes that the Editor ofSoul is poverty-stricken Richard Penniless. Indeed, (2) also fails to entail (1). Tosee this suppose that on January 15, 1965, the man just appointed to theEditorship of Soul does not yet know of his appointment, and that he has reada probated will by which an eccentric businessman bequeathed several millionsto the man who happens to be the Editor of Soul on that day. Thus, Privatus’use of ‘he himself ’ or ‘he*’ just cannot be a proxy for ‘the Editor of Soul�’.

Clearly, the same considerations apply to any name or description thatreplaces ‘the Editor of Soul�’ as well as to any tokenw of ‘he*’. Hence we concludethat a use of a tokenw of ‘he*’ is not an instance of use (D).

We can discuss uses (A) and (B) together. If ‘he*’ were used demonstrative-ly meaning ‘this (that) man’ or simply ‘this (that)’, the demonstrative referencewould have to be made by the speaker. But Privatus can truly and correctly sayto Gaskon “The Editor of Soul knows (believes, thinks) that he* is a million-aire”, even when the Editor of Soul is wholly outside Privatus’ and Gaskon’sexperience, e.g., when the Editor is locked up alone inside a spaceship travelingnear the end of the Milky Way. Hence, the pronoun ‘he*’ is not simply thestrictly third-person pronoun ‘he’ used demonstratively.


Use (F) of ‘he’ does resemble the use of ‘he*’ in important respects. Forsimplicity we shall speak of the (F)-‘he’ to refer to the tokens of ‘he’ which areinstances of use (F). To start with, the (F)-‘he’ and ‘he*’ can appear only in aconstruction having a clause in oratio obliqua. In the second place, each tokenof ‘he*’ or of the (F)-‘he’ must have an antecedent description or name orpronoun, which determines to whom the token refers. Thirdly, in neither caseis the token in question a proxy for its antecedent. Clearly,

(3) Paul believes of (someone who is in point of fact) Mary that she is happy

is not the same statement as

(4) Paul believes that Mary is happy.

Likewise, the statement:

(5) Mary believes that she* is happy

is not the same as

(6) Mary believes that Mary is happy.

Yet there are crucial differences between ‘he*’ and the (F)-‘he’. In the firstplace, (4) and “Mary exists” together entail (3); but neither (5) and “Maryexists” together entail (6), nor (6) and “Mary exists” together entail (5). In thesecond place, the (F)-‘he’ can always be analyzed away, while ‘he*’ can never beanalyzed away, as we shall establish in Sections 2 and 4 below.

Consider (3) again. Clearly (3) entails that what Paul believes is expressiblein a sentence of the form “Z is happy”, where ‘Z’ stands for some unspecifiedway of referring to Mary that Paul employs. (Obviously, the way in questionmay very well be the name ‘Mary’ itself.) If Paul were to express his belief, hewould employ some name or demonstrative or description referring to Mary.But (3) does not decide this point. Statement (3) is merely the statement thatfor some person Paul refers to as Z, Paul believes that Z is happy, while we referto that person as Mary.

This matter of referring to someone as Z, or as Mary, or as the man nextdoor, is a murky business which we shall not clarify in this paper. Here we shalltake the concept way of univocally referring to X as Z as primitive. We do notcommit ourselves here to any of the following natural ways of analyzing thisconcept. Way 1: We abandon the idea of a universe of discourse constituted byparticulars, and make our variables of quantification range over guises (orguised particulars, i.e., particulars qua satisfying some condition). Then we


analyze, e.g., (3) as “There is a guise, call it x, such that x= (the guise) Mary, and(the guise) Paul believes that x is happy.” Way 2: We keep our individualvariables ranging over particulars qua particulars, but analyze ‘ways of referring’in terms of Russellian descriptions. Then, (3), for instance, becomes analyzedas “There is a property Φ such that just Mary is Φ and Paul believes that theonly thing which is Φ is happy.” Way 3: With Frege we distinguish sense anddenotation for names, demonstratives, etc., and introduce variables rangingover individual senses; we form a term denoting a particular by underscoring anexpression denoting an individual sense. Then (3) can be analyzed as “There isan individual sense Z such that Z=Mary and Paul believes that Z is happy.”4

Likewise, we shall not analyze here the perplexing phrases ‘believes (knows,thinks, says, etc.) that Z is…’ where ‘Z’ is a variable. It makes no difference atthis juncture whether ‘Z’ is a variable ranging over particulars, guises, orindividual senses. The natural temptation is to analyze such phrases in terms ofquantification over propositions or statements. For instance, “Paul believes thatZ is happy” would be treated as an abbreviation of “There is a propositionwhich is the object of Paul’s belief and which is constituted by the entity Z(whatever it may be, individual sense, guise, or whatnot) as subject and theproperty happiness as predicate.” Other philosophers and logicians wouldprefer an analysis in terms of sentences. But in this essay we do not have todecide this issue.

The problems we are dealing with in this essay are essentially independentof one’s ultimate views on reference to particulars and on the role of variablesinside cognitive contexts. We want here to formulate an analysis of the (F)-‘he’as well as partial analyzes of some special uses of ‘I’ and of ‘he*’ in terms of ourprimitive ‘ways of univocally referring to X as Z’. We grant, of course, that ouranalyses are rightly regarded as incomplete inasmuch as our present primitiveneeds to be analyzed.

Let ‘Φ(heY)’ be a sentence containing one of or more tokensw of the (F)-‘he’, which refer to a person Y; let ‘Φ(Z)’ be the result of replacing each of thesetokensw by tokensw of ‘Z’; let ‘E’ stand for any of the cognitive verbs (in thesense explained above). Then:

4. I am indebted to Edmund Gettier andWilfrid Sellars for pointing out the need for saying somethingabout my primitive way of referring so that it would not be wholly obscure. I hope that with thispromissory note I may be allowed to continue using an apparent quantification over ways of referring.Sellars was very helpful in emphasizing the need for mentioning way 3.


(P) A statement of the form “X E’s of Y thatΦ (heY)” is the same statement asthe corresponding one (i.e., the one that has the same variables standingfor the same terms where these terms have the same references and mean-ings and logical positions) (i) “There is a way of referring univocally to Yas Z and X E’s that Z isΦ(Z)”, or alternatively, the statement correspond-ing in the same way, (ii) “There is a way of referring univocally to Y as Z,which is different from each of the ways of referring to Y as W1, W2,…, asWn, and X E’s that Φ(Z)”, where for each ‘Wi’ the clause represented by‘Φ(heY)’ contains some sub-clause of the form ‘(…=Wi)’.

Alternative (ii) is included in principle (P) in order to meet a counter exampleproposed by Nuel Belnap and developed by Charles Chihara against alternatives(i).5 The counter-example is this: consider the case of Jaakko Hintikka6 andNuel Belnap7 who believe that every statement of the form “a=a” is (analytical-ly) true. Suppose that, say, Hintikka believes that the morning star = themorning star and that he has no other belief about the morning star. Since themorning star is in fact identical with the evening star, there is a way of referringunivocally to the evening star as Z andHintikka believes that Z is identical withthe morning star. Yet it would be false in ordinary language that Hintikkabelieves of (some object which in point of fact is) the evening star that it isidentical with the morning star. It seems to me that this case conforms toalternative (ii). On the other hand, there are people8 who would say in the caseof Hintikka as just described that Hintikka does believe of something, which isin point of fact identical with the evening star, that it is identical with themorning star. For these persons alternative (i) suffices.

Now, my central point here is that the (F)-‘he’ is fully analyzable along thelines of principle (P). Whether alternative (i) suffices or whether other alterna-tives besides (i) and (ii) are required is immaterial here. Alternative (ii) merelyintroduces conditions on the way of referring to an entity that is attributed toa person. And I can easily conceive of ordinary statements expressed throughsentences with tokensw of the (F)-‘he’ in which other restrictions on a way of

5. During the discussion of the earlier version of the paper at the University of Pittsburgh PhilosophyColloquium, on November 20, 1964.

6. See J. Hintikka, “Toward a Theory of Definite Descriptions,” Analysis (Oxford), Vol. 19 (1958), pp.78–95.

7. See Belnap’s review ofHintikka’s articlementioned in footnote 6 inThe Journal of Symbolic Logic, vol.25 (1960), pp. 88–9.

8. For example, Wilfrid Sellars and Bruce Aune during the discussion mentioned in footnote 5.


referring to an entity are implicit. Yet my primary contention stands unaffected:The (F)-‘he’ can be analyzed away.

To sum up, the S-uses of ‘he’ are logically different from the uses (A)–(F)of the third-person pronoun.

2. ‘He*’, descriptions, names, and demonstrative reference

We have seen that when Privatus asserts “The Editor of Soul believes that he* isa millionaire”, Privatus’ tokenw of ‘he*’ is not a proxy for the description ‘TheEditor of Soul�’. More generally, Privatus’ tokenw of ‘he*’ is not replaceable byany other description or name of the Editor of Soul (or of any other person orthings), which does not include another tokenw of ‘he*’.9 Suppose that ‘theperson Φ’ is a definite description, with no tokensw of ‘he*’, of the Editor ofSoul. Clearly the statement “The Editor of Soul believes that the person Φ is amillionaire” does not entail “The Editor of Soul believes that he* is a million-aire.” Since there are no tokensw of ‘he*’ in ‘The Editor of Soul believes that theperson Φ is a millionaire’, this sentence does not make an assertion of self-belief. The same holds for names. But more importantly, when Privatus asserts“The Editor of Soul believes that he* is a millionaire”, Privatus does notattribute to the Editor the possession of any way of referring to himself asidefrom his ability to use the pronoun ‘I’ or his ability to be conscious of himself.The latter ability is the only way of referring to himself that Privatus mustattribute to the Editor for his statement to be true. Hence, the statement “TheEditor of Soul believes that he* is a millionaire” does not entail any statement ofthe form “The Editor of Soul believes thatΦ is a millionaire”, where ‘Φ’ standsfor a name or description not containing tokensw, of ‘he*’.

It is also apparent that Privatus’ token of ‘he*’ is not a proxy for a tokenw of‘I’ or ‘You’. On the one hand, Privatus’ statement “The Editor of Soul believesthat I am a millionaire” is obviously quite different from his statement “TheEditor of Soul believes that he* is a millionaire.” On the other hand, Privatus’statement “The Editor of Soul believes that you are amillionaire” does not entail,and a fortiori is different from, his statement “The Editor of Soulbelieves that he*

9. I am indebted to Richard Cartwright for the following example of a description that can replace atokenw of ‘he*’: ‘the person (or entity) identical with himself*’. Clearly, there is an infinite chain ofdescriptions generated from this: ‘the entity identical with the entity identical with… the entity identicalwith himself*’.


is a millionaire,” even if Privatus uses the word ‘you’ to talk and refer to theEditor of Soul. The latter may simply be ignorant that he* is the Editor of Soul.

Now, all the above considerations apply to any tokenw of ‘he’ regardless ofwhat name or description happens to be its logical and grammatical antecedent.Thus, we conclude that the pronoun ‘he*’ is never replaceable by a name or adescription not containing tokensw of ‘he*’. This suggests that ‘he*’ is a purelyreferential word. But demonstrative pronouns seem to be purely referential:they, it is often said, seize the objects they refer to directly, without attributingto these objects any feature or characteristic relation. They have denotation, butnot sense, it is said; they, indeed, seem like the closest approximation to thelogician’s ideal of logical names. One is tempted to think that, although ‘he*’ isnot always a demonstrative pronoun, nevertheless it must be analyzed, orunderstood, in terms of the demonstrative uses, (A) and (B), of ‘he’. In particu-lar it might be thought that the analysis of ‘he*’ could be worked out of fourrelated prongs:

i. ‘he*’ in sentences of the form ‘X E’s (e.g., believes, knows, thinks) that he*isΦ’, used assertively, corresponds to some use of the demonstrative ‘he’ bythe person X;

ii. if one asserts “X believes (knows, thinks) that he is Φ” [or “X believes(knows, thinks) of him that he is Φ”] and uses ‘he’ [or ‘him’] demonstra-tively to refer to the person X, then one has asserted “X believes (knows,thinks) that he* is Φ”;

iii. if one asserts truly “X believes (knows, thinks) that he is Φ” using ‘he’demonstratively to refer to the person X and the person X is in a position totake one’s use of ‘he’ as a demonstrative use, then X believes (knows,thinks) that he* is Φ;

iv. if a person X asserts “he isΦ”, using ‘he’ purely demonstratively to refer toX, then X believes that he* isΦ.

Claim (i) is false. To begin with, it is not uncommon that a man, sayPrivatus, asserts truly something like “The Editor of Soul believes (claims,knows) that he* is a millionaire”, in spite of the fact that the Editor of Soul mayhave never expressed his belief through any sentence so that he has never usedthe word ‘he’ demonstratively in a sentence that would have to correspond toPrivatus’ use of ‘he*’. Nevertheless, suppose that the Editor of Soul has justasserted what, according to Privatus, he believes (claims, knows). Most likely hesaid “I am a millionaire.” Thus, Privatus’ use of ‘he*’ may correspond to theEditor of Soul’s use of ‘I’, and need not correspond to this Editor’s use of the


demonstrative ‘he’. That it can never correspond to a demonstrative use of ‘he’will be established with the refutation of (iv).

Claim (ii) is also false. Suppose that Privatus says to Gaskon:

(7) The Editor of Soul believes of him [using this word purely demonstra-tively, with a pointing to the Editor of Soul, as a proxy for ‘this’] that heis a millionaire.

[Here the occurrence of ‘he’ is an instance of use (F).] By asserting (7) Privatusdoes not tell Gaskon what exactly it is that the Editor of Soul believes to be thecase, i.e., the proposition or statement that the Editor takes to be true. Privatusjust informs Gaskon that under some way or other of referring to the man he ispointing to, the Editor of Soul believes that man to be a millionaire. Hence, thetruth of (7) does not require that the Editor of Soul thinks of the man Privatusis pointing to as the same as he* (the Editor himself). Thus, (7) may be truebecause the Editor believes, e.g., that a man with a unique scar on his back is amillionaire and this man happens to be the one Privatus is pointing to. But theEditor of Soul need not realize that he* is the man with the unique scar, or thathe* is a millionaire. He may not even know of Privatus’ statement.

Suppose that Privatus says to Gaskon:

(8) The Editor of Soul believes that he (using this word purely demonstra-tively, with a pointing to the Editor of Soul) is a millionaire.

Suppose now: (a) that Privatus has never before seen the Editor of Soul and thathe cannot recognize him, (b) that Privatus knows of the Editor’s belief becausehe has just talked with him on the telephone, (c) that the Editor informedPrivatus that a man with a unique scar on his forehead is a millionaire, (d) thatsuch a man is the one Privatus is pointing to when he asserts (8), and (e) thatPrivatus does not know that the Editor is the man with that unique scar.Clearly, from (a)–(e) it follows that Privatus’ statement (8) is not the same ashis statement “The Editor of Soul believes that he*is a millionaire.”

Claim (iii) is equally false, in spite of its greater plausibility. Consider onceagain Privatus’ statement (8) to Gaskon. As above, suppose (a)–(e). However,suppose now: (f) that the Editor of Soul is near Privatus and overhears hisstatement (8), and (g) that the Editor is in a position to identify the manPrivatus is pointing to. In short, the Editor is in a position to take Privatus’demonstrative use of ‘he’ as a demonstrative use of ‘he’. Can the Editor fail torealize that he* is the one being pointed to by Privatus? It seems to me that hecan. This is very unlikely, I grant, but not impossible. For suppose (h) that


Privatus knows that the man he talked to on the telephone is the new Editor ofSoul, while this man does not know of his* own appointment yet, and (i) thatthere is a set of mirrors arranged with the right kind of angulation, so that theEditor of Soul sees, in a mirror, Privatus pointing to the man with the uniquescar. Suppose further, as we may, (j) that the Editor does not realize that he* isseeing himself* in the mirror, so that when he himself points to the man withthe unique scar (by pointing to his own mirror image), he may say “He is amillionaire”, without realizing that he is pointing to himself* (or his* ownmirror image). Thus, Privatus’ assertion of (8) may be both true and under-stood by the Editor of Soul without the Editor of Soul believing that he* is amillionaire. Hence, (a)–(j) yield the falsity of claim (iii).

It might be replied that in the preceding fantasy Privatus uses ‘he’ demon-stratively to refer to the Editor of Soul, while this Editor uses ‘he’ demonstra-tively to refer to his mirror image, so that the Editor does not really takePrivatus’ use of ‘he’ in exactly the same way, i.e., to refer exactly to the sameentity. This reply is unsound. No doubt, the Editor of Soul points to his mirrorimage, but only because in this case he is primarily pointing to the Editor ofSoul. Surely he is not referring to a mirror image, but to the man he sees, evenif he only sees the man indirectly via the mirror image. (Ordinary usage isagainst the reply. We do say things like: “The sheriff saw the outlaw in themirror and shot first.”)

At any rate, it is only an empirical, though perhaps physiologically neces-sary, fact that one sees the physical world from the top of one’s nose as the focusof the perspective one finds in one’s visual perceptions. We can easily imaginea universe in which one’s focus of perspective is located several feet away fromone’s nose. We can also imagine this focus changing from time to time,according, perhaps, to certain happenings in one’s brain. In such a universeone’s focus of perspective might be on the left of one’s body, say at one mo-ment, and later it might be in front with one’s own body among the objects onesees. Of course, one would know that a certain body is one*’s body in the usualway, namely, by feeling kinesthetic sensations, pains, itches, etc., in that body.But at moments in which all his bodily sensations were non-existing, or dulland he were not attending to them, one identical twin, for instance, wouldsometimes bemomentarily in doubt as to which of two similar bodies was his*.He could get out of his doubt, for instance, by walking or trying to grabsomething: the body in which he felt the sensations of effort, pressure, etc.would be his body. Now, in a world of this sort nothing need impede Privatus,Gaskon, and the other people talking with them from being very similar to the


Editor of Soul in both bodily and vestiary appearance. Let us call all this (k).Here again we may suppose (a)–(h) as the background of facts behind

Privatus’ statement (8). Suppose (l) that while asleep, the Editor has beencarried to a chair near Privatus, so that his body is surrounded by very similarbodies. Suppose (m) that the Editor and Privatus both know that the uniquemark on the forehead of the new millionaire is to be produced suddenly as aremote effect of some drug, or food, or drink. Thus, we may imagine (n) thatthe Editor wakes up at the very moment the mark appears, (o) that he andPrivatus both see the mark at once and (p) that Privatus points to the body withthe mark, in order to point to the person whose body it is and says to Gaskon“The Editor of Soul believes that he is a millionaire.” Now if (q) the Editor ofSoul dies of a heart attack before he has any kinesthetic sensation through whichhe can identify the body pointed to by Privatus as his*, he will never come tobelieve that he* is a millionaire. Thus, (a)–(h), and (k)–(q) establish the falsityof claim (iii).

The two situations that yield the falsity of claim (iii) also yield the falsity ofclaim (iv). In the first situation, the Editor of Soul may see his body (via hismirror image) and say “He is a millionaire” without realizing at the time that heis pointing to himself through pointing to his body (via his mirror image). Inthe second situation, the Editor may say “He is a millionaire” on seeing themark on his forehead, and if he does not point and is not aware of any bodilysensation at that very moment, he may die, by (q), without realizing that he wastalking of himself.

The crucial weapon in the refutation of claim (iii) is the fact, whichphilosophers (especially Hume and Kant) have known all along, that there is noobject of experience that one could perceive as the self that is doing the perceiv-ing. However it is that one identifies an object of experience as oneself, whenev-er one does, one identifies an object in experience with a thing which is not partof the experience, and this thing is the one to which the person in question willrefer by ‘I’ (or its translation in other languages), and another person will referto by ‘he*’, or ‘he himself ’ in the special S-use.

Carl Ginet has proposed10 an ingenious analysis of ‘he*’ in terms of thepronoun ‘I’, which seems to preserve the directly referring role of ‘he*’ bydumping it, so to speak, on the demonstrative reference of the first-personpronoun. Ginet writes:

10. In his comments on this paper when it was presented at the meeting of the Michigan Academy inMarch 1965.


for any sentence of the form “X believes that he* is H” [there is] a correspond-ing sentence that contains no form of “he*” but that would in most circum-stances make the same statement. The corresponding sentence that will do thejob, I suggest, is the one of the form “X believes (to be true) the propositionthat X would express if X were to say, ‘I am H’” or, perhaps more clearly, “If Xwere to say ‘I am H’ he would express what he (X) believes.”

This suggestion looks very plausible at first sight. But it faces some seriousdifficulties.

Merely saying ‘I am H’ does not guarantee that any proposition is ex-pressed. We must, then, construe Ginet’s “saying ‘I am H’” as something like“assertively uttering ‘I am H’ and nothing else.” But this does not suffice. Forone can express many different propositions by assertively uttering ‘I am H’.What Ginet’s formula lacks is a precise specification of the proposition that Xbelieves when he believes that he* is H. That is to say, Ginet’s ‘saying “I am H”’must be understood as “assertively uttering ‘I amH’ and nothing else, when thesentence ‘I amH’means what it normally means.” But the analysis requires thatwe unpack the clause ‘what it normally means’. Evidently, when this is done weare going to come out with something like “assertively uttering ‘I am H’ andnothing else, when the word ‘I’ is used by X to refer to himself.” Thus, thisanalysis of ‘h*’ in terms of ‘I’ is at bottom circular.

On the other hand, suppose that Ginet’s formula is sufficient to specifypropositions without circularity. Then there is the crucial trouble, pointed outto me by Robert S. Sleigh, that the subjunctive proposition “If X were to say ‘IamH’ he would express a true proposition” is true if X is a truth-telling person,even if in fact X does not believe that he is H. The analysis is, then, too broad.

There is also the fact that “X believes that he* is H” does not entail thatthere are any sentences or that ‘I am H’ is a sentence in some language, or that‘I’ is a word. But Ginet’s analysans does require that ‘I am H’ be a sentence and‘I’ be a word in some language.

In sum, the S-uses of ‘he’ or ‘he himself ’, that is, the use of the pronoun ‘he*’cannot be analyzed in terms of the demonstrative references of the strictly third-person pronoun ‘he’. The only demonstrative reference of ‘he*’ is bound up withthat pertaining to the first-person pronoun ‘I’.

3. ‘I’

In order to analyze in detail the connection between a use of ‘he*’ and animplicit use of ‘I’, we need some grasp of the logic of ‘I’. For our purpose it


suffices here to discuss just a few of the most general features of the logic of ‘I’.We consider first the use of ‘I’ in oratio recta, i.e., in clauses which are neithersubordinated to cognitive verbs (in the sense we gave this term in section I) norenclosed within quotation marks.11 This use of ‘I’ is characterized by thefollowing properties.

I. Like all other demonstratively used pronouns, be they called demonstrat-ives, be they called personal pronouns, ‘I’ has a referential priority over all namesand descriptions of objects. A name, or a description, correctly usedmay fail torefer to the object to which it purports to refer because there may be not one,but many objects which have the name in question, or the properties men-tioned in description. A demonstrative, however, cannot fail to refer because ofa multiplicity of candidates for reference. Once the demonstrative is correctlytendered, there is for its user at most one candidate reference by it.

II. The pronoun ‘I’ has an ontological priority over all names and descriptions.A correct use of ‘I’ cannot fail to refer to the object it purports to refer. Somephilosophers would argue that every demonstratively used pronoun has thispriority over all names and descriptions. They could claim that if a token of ademonstrative is correctly used in oratio recta, of necessity, the token inquestion would succeed in referring to something in the speaker’s currentexperience. They would argue that even in the case of a delusion, if, e.g., a manhonestly asserts “That is a dog”, his token of ‘that’ successfully picks out acomplex of sense-data, or a region of physical (or psychological) space, or a setof features, or what not. Here, however, we do not have to go into this matter.Whether ‘I’ alone or all demonstratively used pronouns have this ontologicalpriority is of no consequence for the present investigation. What matters to ushere is that ‘I’ does have this priority over names and descriptions.

III. The pronoun ‘I’ and all descriptions and, to a certain extent, names have anepistemological priority over all the other demonstratively used pronouns. Inorder to keep knowledge or belief, or in order merely to rethink, of the objectsoriginally apprehended bymeans of demonstratives onemust reformulate one’sknowledge or belief, or thought, of those objects. Onemust replace each purelydemonstrative reference by a reference in terms of descriptions or names, or in

11. Here I am not regarding modal contexts as oratio obliqua. For our discussion of ‘I’ and ‘he*’ thereis no difference between, e.g., “X loves me” or “It is possible that X loves me.” In neither case is there anappearance of attributing to X a first-personal way of referring to another person, as there is, e.g., in“Peter believes that I love him.”


terms of the demonstrative ‘I’. If Privatus asserts “This is blue”, perhaps with apointing, he seems both to single out an object in his experience and to attributeto it nothing but blueness. Let this be as it may. The crucial thing, however, isthat later on, when the object is no longer in his presence, the pronoun ‘this’ hasto yield to a name or description of the object Privatus called ‘this’. Demon-stratives are necessarily eliminable for their users. The only exception is thedemonstrative ‘I’. Nobody can at all keep knowledge or belief of whateverinformation about himself he receives, unless hemanages to replace every singlereference to himself in terms of descriptions or names, or in terms of otherdemonstratives (like ‘you’, ‘he’, ‘this’), by a reference in terms of ‘I (me, my,mine, myself)’. This does not mean, of course, that whenever, e.g., Privatushears “Privatus is Φ,” he is to perform a physically, or psychologically, distin-guishable act of translation: “That is, I am Φ.” The point is a logical one.Privatus cannot remember, or merely consider later on, that he* isΦ, unless heremembers, or considers, what he would formulate by saying “I am Φ” or“Privatus is Φ and I am Privatus.” At least the statements of identity “I amPrivatus” or “I am the one who…” must include an ineliminable use of ‘I’ forPrivatus. If he only entertains or thinks the statements, without actually makingany assertion, we shall speak of his making an implicit use of ‘I’.

IV. But the epistemological priority of the demonstrative ‘I’ is only partial.Everybody else must replace a person’s references to himself in terms of ‘I (me,my, mine, myself)’ by references in terms of some description or name of theperson in question.

These are all trivial features of ‘I’, names, descriptions and the other demonstra-tive pronouns. But they have some important consequences. For instance, eventhough a demonstrative seems merely to denote an object, it has a sense for itsuser, which consists in its being a place-holder for some description or other, towhich, by (III), it must yield.

Some philosophers may be tempted to claim that every demonstrative hasas its sense a set of descriptions, namely, the descriptions to which it can, ormust, yield in future references to the same entity the demonstrative denotes.They may be tempted to this claim by the natural principle that every statementcan be repeated, i.e., reasserted or at least re-thought. Thus, if Privatus says“This is Φ” he can be supposed to be capable of making this very same state-ment any time he wishes to. But since when the objects of his experience havechanged, he can no longer call ‘this’ the same object he called ‘this’ before, hemust employ some description, and this description, one is tempted to think, is


part of the sense of his original tokenw of ‘this’. This seems an extreme viewwhich we need neither attack nor defend here. Indeed, we need not even takeissue here with the weaker view that the sense of a tokenw of a demonstrative isgiven by some descriptions pinpointing the object as one experienced at thetime or place it was referred to demonstratively. Descriptions of this type are,e.g., “The cow I saw yesterday” and “The man I touched just a little while ago.”The view in question need not object to the sense of a tokenw of a demonstra-tive being given in terms of other tokenw of the same or other demonstratives.This view does not have to include the thesis that demonstratives can beanalyzed away in terms of descriptions.

Let us move on to oratio obliqua. In the case of tokensw of the pronoun ‘I’occurring in clauses subordinated to cognitive verbs (in the sense of ‘cognitiveverb’ explained in section I), we must distinguish two types.

I-i.�There are (F)-uses of ‘I’. These are found in statements of the form “Xbelieves (thinks, knows, asserts, argues, etc.) of me that I (me, my, mine,myself)…”, here the second tokenw of ‘I’ is characterized both by having a logicaland grammatical antecedent, which is a tokenn of ‘me’, and by being a place-holder for some unspecified description or name of the person to whom ‘me’refers.

I-ii.�There are directly self-referring uses of ‘I’. These are found in statementsof the form “X believes (thinks, knows, argues, etc.) that … I (me, my, mine,myself)…”, where (at least some) tokensw of ‘I’ have no grammatical or logicalantecedent, and, thus, refer directly to the speaker, instead of referring to theperson referred to by a first-person antecedent.

Let us consider type (I-i). By analogy with the (F)-uses of ‘he’ we mayexpect the (F)-uses of ‘I’ to be eliminable for the speaker in essentially the sameway. This is in fact what happens. But in the case of ‘I’ there is a new reason tosuspect that an (F)-tokenw must be eliminable. By (IV) above, nobody can referto another person by means of ‘I’. Thus, when someone, call him Y, asserts (orthinks), e.g., “X believes of me that I am Φ”, the tokenw of ‘I’ here cannotrepresent in and by itself the sort or reference to Y that the person X canmake.12 It must, perforce, stand for some way of referring to Y that is available

12. I am indebted to Norman Kretzmann for the following teasing use of ‘I’ which looks like acounterexample to this claim. Suppose that there is a play about Privatus and that Privatus is in theaudience. Suppose further that the actor representing Privatus is losing his moustache and that Privatusreferring to the actor says “I am losing my moustache.”


to X. In general, let ‘E’ stand for a cognitive verb; let ‘Φ(I)’ stand for a sentencecontaining some (F)-tokensw T of ‘I’ whose antecedent is a tokenn of ‘me’ notoccurring in ‘Φ(I)’, and let ‘Φ(Z)’ be obtained by replacing in ‘Φ(I)’ everytokenw T by a tokenw of ‘Z’. Then:

(P¢) A statement of the form “X E’s of me that Φ(I)” is the same as the state-ment of the corresponding form “There is a way of referring to a certainperson as Z, I am that person and X E’s that Φ(Z).”

Suppose, for example, that Privatus asserts:

(3) Paul believes of Mary that she is happy.

Here Privatus’ tokenn of ‘she’ is an instance of the (F)-use of ‘he’. Supposefurther that Mary wants to say about herself* what Privatus has said about her.She will say:

(9) Paul believes of me that I am happy.

Here Mary’s second tokenw of ‘I’ corresponds to Privatus’ tokenn of ‘she’. It isan instance of the (F)-use of ‘I’. It is at bottom not an authentic first-person useof ‘I’. Its first-person role is wholly derivative: it consists merely in having as itslogical and grammatical antecedent Mary’s tokenn of the authentically first-person ‘me’. ‘I’ in ‘that I am happy’ does not refer by itself to the speaker. Justas (3) does not ascribe to Paul the belief that Mary is happy, likewise (9) doesnot ascribe to Paul the belief that “I (Mary) am happy” is true.

When Mary utters (9), her tokenn of ‘I’ functions simply as a place-holderfor some unspecified name or description which refers to the person referred toby Mary’s tokenn of ‘me’. This is the complete analysis of this (F)-use of ‘I’.Hence Mary’s (F)-used tokenn of ‘I’ is eliminable for Mary by principle (P¢).Mary’s statement (9) is the statement she would have made had she said:

(9a) There is a way of referring univocally to a certain person as Z, I am thatperson, and Paul believes that Z is happy.

For contrast, note that Mary’s token of ‘me’ in (9) is not in oratio obliqua.Marycannot eliminate this tokenw of ‘I’ from (9) : it reappears in (9a) in the clause ‘Iam that person’.

Let us proceed to type (I-ii). The main features of a use of ‘I’ of type (I-ii)are determined by (III) and (IV). In the first place, since by (III) a person’s usesof ‘I’ in oratio recta are for him ineliminable, these uses are also ineliminablewhen they are preceded by a cognitive prefix of the form ‘I (believe, know,


think, claim, argue, asserted, etc.)’. The person who is said to do the believing(knowing, thinking, arguing, asserting, etc.)mustmake genuine self-referencesby means of ‘I’. In the second place, by (IV) a person must replace somebodyelse’s, say Y’s, uses of ‘I’ by descriptions. Hence, every tokenw of ‘I’ in a clausesubordinated to a prefix of the form ‘X believes (knows, thinks, says, etc.)’,where ‘X’ is not the pronoun ‘I’, can only be a place-holder for a way ofreferring to Y which is available to X.

To simplify let us define a cognitive prefix as a clause containing a cognitiveverb and ending in ‘that’.13 For instance, ‘While staring at Mary, in the kitch-en, John told Irene that he would love her forever’ has as its cognitive prefix‘while staring at Mary, in the kitchen, John told Irene that’. Thus, we have that:

V. If a tokenw of ‘I’ of type (I-ii) occurs oratio obliqua immediately subordinat-ed to a cognitive prefix containing a tokenw of ‘I’, then the former is eliminablefor the person it refers to if the latter is also eliminable for the same person.

VI. A tokenw of ‘I’ of type (I-ii) is eliminable for the person it refers to, provid-ed that the oratio obliqua in which it occurs is subordinated to a cognitive prefixcontaining no tokensw of ‘I’.

That (VI) cannot be strengthened to a statement about a sufficient andnecessary condition was shown to me by Geach. In “I believe that I am amillionaire, and Gaskon believes that he* is a millionaire” the second tokenw of‘I’ is eliminable, even though subordinated to an ineliminable tokenw of ‘I’,inasmuch as the statement is, as Geach pointed out, equivalent to “Each of twopersons, Gaskon and me, believes that he* is a millionaire.” Since in this case atokenw of ‘I’ is eliminable for a tokenw of ‘he*’ we shall distinguish it from stricteliminability which consists in being eliminable, salva propositione, by some-thing other than tokens of the S-pronoun ‘he’, whatever these may be. Thus,strict eliminability implies eliminability simpliciter. Similarly, for ‘he*’ stricteliminability is eliminability in terms other than tokensw of ‘I’.

Concerning the elimination procedure we must note that there are twocases of tokensw of ‘I’ eliminable for their users: (a) tokensw which are notsubordinated to cognitive prefixes with eliminable tokensw of ‘I’, and (b)tokensw which are subordinated to cognitive prefixes with eliminable tokenswof ‘I’. The procedure of elimination for case (b) involves the procedure for case

13. This shift from cognitive verb to cognitive prefixes was required to meet an objection of Geach’s toprinciple (P≤) below.


(a). Suppose, for instance, that Privatus asserts “Jones believes that I know thatI am happy.” The first token of ‘I’ is a place-holder for some description orname of Privatus known to Jones; let it be ‘Z’. But the second token of ‘I’ is nota place-holder for ‘Z’. For Privatus has not asserted something like “There is away of referring to me as Z, and Jones believes that Z knows that Z is happy.”Surely, the man known to Jones as Z may fail to know himself* as Z, and this issomething which Jones may be expected to know, and Privatus may also knowthat Jones knows it. But, more importantly, Privatus’ statement makes the claimthat Jones ascribes self-knowledge to Privatus, i.e., a knowledge in terms of thepurely referential first-person way. Privatus’ assertion is, at bottom, “There is away of univocally referring to a person as Z, I am Z, Jones can identify Z in therelevant respect, and Jones believes that Z knows that he* is happy.” The mainpoint is that the elimination of tokens of ‘I’ in oratio obliqua, is in case (b), to bemade in terms of fresh tokens of ‘he*’.

Let us formulate the general principle of elimination for cases (a) and (b) atonce. Let ‘E’ stand for a cognitive verb; let ‘Φ(I)’ stand for a sentence or clausecontaining tokensw of ‘I’ which are not subordinated to any cognitive verb in‘Φ(I)’; let ‘Φ(Z)’ stand for the result obtained from ‘Φ(I)’ by: (1) replacing alltokensw of ‘I’ not subordinated to any cognitive verb in ‘Φ(I)’ by tokens of Z,(2) replacing all tokensw of ‘I’ subordinated in ‘Φ(I)’ only to cognitive prefixescontaining tokensw of ‘I’ by tokensw of ‘he*’ whose antecedent is Z, and (3) byreplacing nothing else. Then:

(P≤) A statement of the form of “X E’s that Φ(I)” is the same as the corre-sponding statement of the form “There is a way of referring to a certainperson as Z, X can identify Z (in the relevant respect, or knows who Zis), I am Z, and X E’s that Φ(Z).”

There is a serious obscurity about (P≤), namely, the obscurity surroundingthe notion of identification. What counts as identifying a person is somethingthat cries out for analysis. However, this analysis lies too far afield from us here.All we can say at this juncture is that there seem to be different criteria foridentifying a person, depending on the circumstances of the persons involved,in particular the circumstances linking the identifier and the identified person.Thus, suppose that a certain crime has been committed. In this case in order toknow who the criminal is, a detective must be able to know at least how to bringthe criminal to jail or to court. On the other hand, for a high-school student toknow who the President of the United States is he needs only to be able toanswer certain questions that his teacher may put to him.


Consider Mary’s statement:

(10) Paul believes that I am happy.

Clearly, the reference to Mary by means of ‘I’ is one that Paul just cannot make(assuming ordinary meanings, of course). By (III) and (IV), whatever Paulbelieves of Mary, he has to believe it by thinking that a certain person Z ishappy. The claim that Z is (the same as) Mary is one that Paul need not make,nor is it one that Mary’s statement (10) attributes to him. Indeed, the claim thatMary is Z is one that Mary herself need, or does, not make when she asserts(10). Thus, we have to analyze (10) very much in the same way that we analyzestatements made by means of sentences containing tokens of the (F)-‘he’. Butnot quite. Mary’s statement (10) is not identical with her statement (9), “Paulbelieves of me that I am happy”; hence, we are not to analyze (10) as (9a) above.

The difference between (9) and (10) is very intriguing. But we do not haveto dwell upon it here. We need note only that it lies in the claim, made by (10)but not by (9), that Paul’s way of referring toMary as Z enables him to pick out,or identify, Mary as the person whom he believes to be happy. This is preciselywhat principle (P≤) asserts.

Now, by (V)–(VI) a tokenw T of ‘I’ subordinated to a cognitive verb whosesubject is not a token of ‘I’ is eliminable for the person referred to by the tokenwT, even if T is subordinated to a cognitive verb whose subject is another tokenwT¢ of ‘I’. For example, suppose that Privatus asserts:

(11) I believe that the Editor of Soul knows that Mary believes that I am amillionaire.

Here the first tokenn of ‘I’ is in oratio recta, and by (III) it is ineliminable forPrivatus. The second tokenn of ‘I’ is, by (VI), eliminable even for Privatus. Byapplying (P≤) to (11) Privatus could eliminate the latter tokenn of ‘I’ in favourof a new tokenw of ‘I’ that would be subordinated to the prefix ‘I believe that’.Hence, this new tokenn would also be ineliminable for Privatus, by (V). But allthese tokensw of ‘I’ would be eliminable for Privatus if he were to subordinate(11) or its equivalent, by (P≤), to a prefix of the form ‘X E’s that’. Yet, again, theeliminable tokensw of ‘I’ would give rise to ineliminable tokensw of ‘I’ by virtueof (P≤). The whole thing is simply that the only ineliminable tokens of ‘I’ for theuser of ‘I’ are (1) those occurring in oratio recta and (2) those in oratio obliquasubordinated only to prefixes of the form ‘I E that’.


4. ‘He*’ and ‘I’

We must now discuss the eliminability of tokensw of ‘I’ for a person who,instead of making them, hears them. By (IV) above, the tokensw of ‘I’ in oratiorecta are necessarily eliminable for anyone who does not make them. But thetokensw of ‘I’ in oratio obliqua that are ineliminable for the person whom thetokensw refer to are not, for another person, eliminable in terms of descriptionsor names of the former persons. For instance, suppose that Privatus asserts “Ibelieve that I am a millionaire.” For everybody else Privatus’ first tokenn of ‘I’must yield to some description or name of Privatus, e.g., ‘Privatus’. But hissecond tokenn of ‘I’ must be replaced by a tokenw ‘he*’. His statement wouldyield for us the statement “Privatus believes that he* is a millionaire.” Ingeneral, the precise correspondence between ‘he*’ and ‘I’ is simply this:

(H*) A use of ‘I’ in oratio obliqua that is ineliminable for the person it refers to,who uses it, corresponds to, in the sense that it must yield to, anotherperson’s ineliminable use of ‘he*’ in oratio obliqua.

At this juncture the question arises: Can the pronoun ‘he*’ appear in oratiorecta? It might seem obvious that the answer to this question is in the affirma-tive. There are in fact three general considerations which might at first sightseem to support an affirmative answer. First, when a person formulates what hebelieves (knows, thinks, etc.) he says ‘I…,’ and other persons say ‘He…’; second,in a sentence of the form ‘X believes (knows, thinks, says, etc.) that he* …’ theperson X is precisely the one ‘he*’ refers to; third, what one knows is true, thusif X knows that he* is Φ, it seems, simpliciter, that he* is Φ.

To be sure, when a person X hears Y say “I amΦ”, X will understand whathe heard as “He is Φ”. But from this it does not follow this token of ‘he’ is atoken of ‘he*’. Indeed, there is a good reason for suspecting that here we do nothave a token of ‘he*’. X’s token of ‘he’ must be a demonstrative if it is to refer toY univocally and without describing him. But we saw in Section 2 that thepronoun ‘he*’ is not demonstrative. Thus, the first reason for the claim that‘he*’ appears in oratio recta provides, rather, evidence against this claim.

Doubtless, in the sentence ‘X believes that he* isΦ’ the token of ‘he*’ doesrefer to the person X. This suggests both that we could have statements of theform “he* is X” and that whoever believed or asserted “X believes (knows,thinks, says, etc.) that he* is Φ” should certainly know the truth of the corre-sponding statement “he* is X”. It would seem, then, that tokens of ‘he*’ canappear in oratio recta. However, if there were complete statements of the form


“he* is X”, these statements could be known to be true by the person X himself.But suppose that X, who does not know that he* is X, does assert (or just thinks,for that matter) “X believes (knows, thinks, says, etc.) that he* is X”. That is, Xcould come to know that he* is X simply by thinking that the man X has someproperty or other. For instance, the heaviest man of Europe could come toknow that he* weighs more than anybody else without resorting at all to scalesand comparison of weights! This absurdity arises simply from allowing thetokensw of ‘he*’ to function as independent symbols, i.e., as referring devices intheir 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 X”. That is, a sentence containing a tokenw of ‘he*’ can, given ordinarymeanings, formulate a statement only if the tokenw in question has an anteced-ent in the same sentence.

Let us turn now to the third reason for supposing that ‘he*’ can appear inoratio recta. Doubtlessly, the following principle is true:

(K) A statement of the form “X knows that …” entails that the statementdenoted by the sentence filling the blank ‘…’ is true.

Thus, the statement “The Editor of Soul knows that he* is a millionaire” entailsthat the statement (or proposition) denoted by the clause ‘he* is a millionaire’is true. But (K) must be distinguished from two rules of inference, which seemto follow immediately from (K). They are:

(K.1) From a statement of the form “X knows that p” you may infer the corre-sponding statement that p.

(K.2) If a sentence of the form ‘X knows that p�’ formulates a statement youaccept, then you may detach the sentence (or clause) S represented by ‘p�’and use S by itself to make the statement which S formulates as part of thelarger sentence ‘X knows that p�’ provided that S contains no tokensw ofeither first or second-person pronouns.

Clearly neither principle (K) nor rule (K.1) says anything about the kind ofsentence through which the proposition p is to be formulated. Only (K.2), ofthe three, establishes a condition about sentences that could support the claimthat tokensw of ‘he*’ can appear, correctly, in oratio recta. But I want to arguethat rule (K.2) is invalid: it leads to contradictions. Moreover, I want to arguethat rule (K.1) is also in need of revision.

Let us consider (K.2) first. It is a fact that in ordinary life one often draws,or can draw anyway, valid inferences like:


(11) The Editor of Soul knows that he himself [i.e., he*] is a millionaire.

hence,

(12) He is a millionaire.

This inference seems validated by (K.2). One feels tempted to say that the tokenof ‘he’ in (12) is simply an instance of the pronoun ‘he*’ which appears in (11).Indeed, since the above inference is valid, the token of ‘he’ in any token ofsentence (12) is a token of ‘he*’ if and only if the above inference is validated by(K.2). If the inference is validated by (K.2), then (12) contains an independentuse of ‘he*’ in oratio recta.

The validity of the inference “(11), hence (12)” does not, of course, dependat all on who draws it. Thus, the role of (K.2) and the character of the token of‘he’ in (12) remain unaltered if Privatus draws the inference, even though

(13) It is not the case that Privatus believes that he* is a millionaire.

Suppose further that Privatus believes that (11) and draws out loud, or inwriting, the inference from (11) to (12). Clearly, while drawing the inferencePrivatus believes both the premise and the conclusion to be true. Make now theassumption that the occurrence of ‘he’ in (12) is an instance of ‘he*’. Thenwhile drawing the inference from (11) and (12), Privatus believes that both (11)the Editor of Soul knows that he himself is a millionaire and (12) he* is amillionaire. But then the latter ‘he*’ refers back to ‘Privatus’, so that we caninfer that Privatus believes that he* is a millionaire. Since this result contradicts(13), the assumption is false. The use of ‘he’ in (12), as part of the inferencefrom (11), is the same regardless of whether or not the Editor of Soul believes(or knows) that he* is the Editor of Soul. Hence, we may conclude that in nocase is the token of ‘he’ in (12) a token of ‘he*’. This being the case, the infer-ence “(11), hence (12)” is not validated by rule (K.2).

Evidently, the same arguments apply to any statement of the form “X E’sthat Φ(he*).” Thus, we may regard as established that:

(H*1) The pronoun ‘he*’ is strictly a subordinate pronoun: it is by itself anincomplete, or syncategorematic, symbol, and every sentence or clausecontaining a tokenw of ‘he*’ which is not in oratio obliqua, is also anincomplete or syncategorematic sentence or clause.

Let us return to the inference “(11), hence (12)”. The inference is palpablyvalid. But what kind of ‘he’ is, then, the one appearing in (12)? It seems to methat (12) has a token of ‘he’ which is an instance of use (D), i.e., a relative


pronoun which is here a proxy for the definite description ‘The Editor of Soul�’.In other words, the conclusion of the inference “(11), hence (12)” is thestatement “(He=) The Editor of Soul is a millionaire.” But then, the inference“(11), hence (12)” is not validated by rule (K.1)! By the arguments of Section 2above, the tokenw of ‘he*’ in (11) does not have a sense that can be given by anyother use of the third-person pronoun, or any name, or any definite descrip-tion. Thus, the question arises: Is there a way at all of formulating the proposi-tion which, by (11), the Editor of Soul knows to be true, so that we can applyrule (K.1) to (11)?

The proposition that the Editor of Soul knows to be true, if (11) is true, isone which the Editor can formulate by saying: “I am a millionaire.” But eventhough the Editor of Soul would not be inferring a false conclusion from a truepremise, the inference “The Editor of Soul knows that he* is a millionaire,hence I am a millionaire” is even for him not validated by an application of rule(K.1). On the other hand, the inference “I am the Editor of Soul, the Editor ofSoul knows that he* is a millionaire, hence I am a millionaire” is valid, but it is,obviously, not validated by rule (K.1).14

There is just one other referring expression that has to be considered: thepronoun ‘you’. Let us assume that any of us can formulate the statement that,according to (11), the Editor of Soul knows to be true by uttering the sentence“You are a millionaire.” But the inference “The Editor of Soul knows that he*is a millionaire, hence you are a millionaire” is not valid, and, a fortiori, not anapplication of rule (K.1). On the other hand, “The Editor of Soul knows thathe* is a millionaire, and you are the Editor of Soul, hence, you are a millionaire”is a valid reference, but it is not validated by (K.1). [Ed. note: The original has‘reference’ but the context seems to demand ‘inference’.] Actually, I doubt verymuch that it is always possible to formulate the very same proposition denotedby a clause of the form ‘he*’ isΦ in oratio obliqua by means of the correspond-ing oratio recta sentence of the form ‘You are Φ’. Consider, for instance, thestatement “Just before his death Caesar thought that he* was to be crownedking.” It seems to me that the statement “Caesar, you were to be crowned king”is not precisely the one Caesar thought to be true. For one thing, the latter

14. I am very grateful to both Richard Cartwright and Robert Sleigh, Jr. for having impressed upon meduring thediscussionmentioned above in footnote 3 the need for distinguishing betweenprinciples (K.l)and (K.2). They also pointed out some errors in my confused discussion of (K.l) and (K.2), e.g., that Ihad not discussed the case of ‘you’, as I do below in this version. However, neither Cartwright nor Sleighare responsible for the present formulation of (K.1) or (K.2).


statement entails or implies or presupposes that somebody is talking to Caesar,while the statement (or proposition) that Caesar thought to be true does notcarry that implication or presupposition or entailment.

At any rate, even if the proposition that, by (11), the Editor of Soul knowsto be true can be formulated by sentences in oratio recta containing the pro-noun ‘you’, we have seen that rule (K.1) cannot apply to (11). Obviously,similar considerations hold for any other third-person statement ascribing self-consciousness to someone. Hence

(H*2) The generally accepted rule (K.1) is invalid: it fails for statements ex-pressible in sentences containing tokensw of ‘he*’.

(H*2) raises a serious task, namely, that of formulating a set of true principleswhich are to replace the widely accepted (K.1).

By (H*) and by (V) and (VI) of Section 3, we should expect the tokensw of‘he*’ that are ineliminable in a sentence S to become eliminable when S isconcatenated to a prefix of the form ‘X thinks (believes, knows, says, etc.)’. Aswe shall see in the sequel, this is precisely the case even when the occurrence of‘X’ in this prefix-form stands for another token of the expression which is theantecedent of a tokensw of ‘he*’ in S. To say it at once, the principle governingthe ineliminability of ‘he*’ is:

(H*3) A tokenw T of ‘he*’ is strictly ineliminable for its user in two types ofcases, and only in these two types of cases: (1) T occurs in an oratio obli-qua subordinated to just one cognitive prefix containing the antecedentof T; (2) T occurs in an oratio obliqua subordinated to n + 1 cognitiveprefixes such that the very first one, from the left, has the antecedent A ofT, and the other n verbs have tokensw of ‘he*’ whose antecedent is also A.

By (H*3) the following tokensw of ‘he*’ are ineliminable for their users:“Privatus believes that he* is happy”, and “Alexander believed that he* knewthat he* once thought that he* was a god.” (H*3) separates the tokensw of ‘he*’that are eliminable for their users from those which are not; but it does notfurnish the elimination procedures. By (H*)–(H*3), these procedures aresimilar to those employed for the elimination of ‘I’, but there is the importantdifference already noted in Section 3 that many a tokenw of ‘I’ has to be elimi-nated in favor of a tokenw of ‘he*’. We shall discuss the general principles forthe elimination of ‘he*’ by considering some simple examples with ‘he*’. As inthe case of ‘I’, there are two cases: (a) some tokenw of ‘he*’ are eliminable froma subordinate clause at the cost of introducing another tokenw of ‘he*’ in the


main clause, and (b) some tokenw of ‘he*’ are eliminable in the sense that theyare replaced by tokenw of ‘he*’ in the very same subordinate clause, but with adifferent antecedent.

A simple example of case (a) is this. Suppose that Privatus asserts:

(14) The Editor of Soul believes that the Editor of Soul believes that he* is amillionaire.

It is not evident what exactly it is that Privatus has asserted. His utterance has aninteresting ambiguity. Imagine the Editor of Soul himself making the same, orvery similar, statement. He has a choice between

(15) I believe that the Editor of Soul believes that I am a millionaire,

and

(16) I believe that the Editor of Soul believes that he* is millionaire.

Of course, he can also say:

(17) I believe that I believe that I am a millionaire.

What interests us now is the contrast between (15) and (16). The occurrence of‘he*’ in (16) stands for an occurrence of ‘I’, as it were, two steps removed: itstands for a tokenw of ‘I’ that the Editor of Soulmentioned by the Editor of Soulcould produce to say “I am millionaire.” Thus, in the case of (14), we mustdistinguish two cases. On the one hand, the tokenw of ‘he*’ occurring in (14)may have as its antecedent the second occurrence, from it to the left, of thephrase ‘the Editor of Soul�’. This is the syntax of the sentence that Privatusmeant to utter, if he meant tomake the statement, (18), below analogous to theEditor of Soul’s statement (15). On the other hand, the occurrence of ‘he*’ in(14) may have as its antecedent the first occurrence, from it to the left, of ‘theEditor of Soul�’. This is the syntax of the sentence Privatus meant to utter, if hemeant to make the statement, (19) below, analogous to (16). Let us affixnumerals to a tokenw of ‘he*’, in such a way that we can determine whichexpression is the antecedent of the tokenw in question by counting, to the left ofit, the cognitive verbs to which it is subordinated. Thus, our two interpretationsof (14) can be written as follows:

(18) The Editor of Soul believes that the Editor of Soul believes that he*2 is amillionaire;


(19) The Editor of Soul believes that the Editor of Soul believes that he*1 is amillionaire.15

By principle (P≤) above, the second occurrence of ‘I’ in (15) is eliminable.Hence, by (H*) the tokenw of ‘he*’ in (18) should also be eliminable in a similarway. Thus, (18) is the same as the statement:

(18a) The Editor of Soul believes that there is a way of univocally referring to acertain person as Z, that he* is Z, and that the Editor of Soul both canidentify Z in the relevant respect and believes that Z is a millionaire.

On the other hand, by (H*3), neither (19) nor (18a) has occurrences of ‘he*’that are eliminable for their users. Note, incidentally, that a first cousin of (18),‘The Editor of Soul believes that the Editor of Soul believes of him*1 that he is amillionaire’ also has an ineliminable occurrence of ‘he*’: the antecedent of‘him*1’ is the very first occurrence of ‘the Editor of Soul�’ (the last tokenn of ‘he’is an instance of the (F)-use).

An example of case (b) appears in

(20) The Editor of Soul believes that Privatus believes that he*2 knows thathe*3 is the Editor of Soul.

Given our convention on subscripts, the two tokensw of ‘he*’ in (20) have ‘theEditor of Soul�’ as their antecedent. Statement (20) is the same as the statement

(20a) The Editor of Soul believes: that there is a way of referring to a certainperson as Z, that he*1 is Z, and that Privatus can both identify Z in therelevant respect and believes that Z knows that he*1 is the Editor of Soul.

By our convention, in (20a) the first occurrence of ‘he*1’ has ‘the Editor of Soul�’as antecedent, while the second occurrence has ‘Z’ as antecedent. By (H*3)neither occurrence is eliminable for the speaker.

15. The distinction between (18) and (19) cannot be formulated in Hintikka’s calculus mentioned infootnote 1. There is nothing in this calculus that corresponds to the criss-crossing references mirroredby our subscript notation. Yet the possibility of criss-crossing is central to the pronoun ‘he*’.

Chapter 5

Self-reference and self-awareness*

Sydney S. ShoemakerCornell University

If we consider the logical powers of first-person statements and the role playedby the first-person pronoun in communication, nothing seems clearer than thatin all first-person statements, including “avowals,” the word ‘I’ functions as asingular term or singular referring expression. Statements expressed by thesentence “I feel pain” have it in common with those expressed by sentences like“He feels pain” and “Jones feels pain” that they contradict the proposition“Nobody feels pain” and entail the proposition “Someone feels pain.” In theseand other ways “I feel pain” behaves logically as a value of the propositionalfunction “X feels pain.” Moreover, in all first-person statements, including“psychological” or “experience” statements, the word ‘I’ serves the function ofidentifying for the audience the subject to which the predicate of the statementmust apply if the statement is to be true (what it indicates, of course, is that thesubject is the speaker, the maker of the statement). And this is precisely thefunction of a referring expression.

Yet philosophers have often found the referring role of ‘I’ perplexing, andsome have been led to hold that in at least some of its uses it is not a referringexpression at all. Thus Wittgenstein reportedly held at one time that “I havetoothache” and “He has toothache” are not values of a common propositionalfunction, that in “I have toothache” the word ‘I’ does not “denote a possessor,”and that “Just as no (physical) eye is involved in seeing, so no Ego is involved inthinking or in having toothache.” He is also reported to have viewed withapproval Lichtenberg’s saying that instead of “I think” we ought to say “It

* Presented in an APA symposium on Self and Reference, December 28, 1968; see Michael Woods,‘Reference and Self-identification,’ The Journal of Philosophy, LXV, 19 (Oct. 3, 1968): 568–578.

In writing this paper I had the advantage of having read, or heard, unpublished papers on its topicby Norman Malcolm and Keith Gunderson. I am also grateful to Hector-Neri Castañeda, HarryFrankfurt, and Margaret Wilson for helpful comments and criticisms.

82 Sydney S. Shoemaker

thinks” (with ‘it’ used as it is in “It is snowing”).1 At apparently the oppositeextreme from this, yet stemming from much the same sources, is the view that‘I’ refers to a “transcendental ego,” an entity that is in principle inaccessible tosense experience. In this paper I shall try to diagnose the source of, and todispel, some of the mysteriousness which surrounds the use of the word ‘I’ andwhich underlies the perennial attractiveness of such unacceptable views aboutthe self and self-reference.

I

In the Blue Book Wittgenstein distinguished “two different uses of the word ‘I’(or ‘my’),” which he calls “the use as object” and “the use as subject.” Asexamples of the first of these he gives such sentences as “My arm is broken” and“I have grown six inches.” As examples of the second he gives “I see so and so,”“I try to lift my arm,” “I think it will rain,” and “I have toothache.” He goes onto say: “One can point to the differences between these two categories by saying:The cases of the first category involved the recognition of a particular person,and there is in these cases the possibility of an error, or as I should rather put it:the possibility of an error has been provided for… On the other hand, there isno question of recognizing a person when I say I have tooth-ache. To ask ‘areyou sure it is you who have pains?’ would be nonsensical.”2

It is important to see that the distinctionWittgenstein is drawing here is notthe controversial distinction between ‘corrigible’ and ‘incorrigible’ first-personstatements. It is easy to overlook this, forWittgenstein’s examples of “the use assubject” are mostly statements that many philosophers have held to be incorri-gible. But Wittgenstein’s point is not that these statements are totally immuneto error, though he may have believed this to be true of some of them, but israther that they are immune to a certain sort of error: they are immune to errordue to a misrecognition of a person, or, as I shall put it, they are immune toerror through misidentification relative to the first-person pronouns. It is theuse of ‘I’ in such statements, i.e., its use “as subject,” that philosophers havefound puzzling.

1. SeeG. E.Moore, ‘Wittgenstein’s Lectures in 1930–33,’ Philosophical Papers (London: Allen&Unwin,1959): 306–310.

2. The Blue and Brown Books (New York: Oxford, 1958): 66–67.

Self-reference and self-awareness 83

If I say “I am bleeding,” it can happen that what I say is false even though Iam giving expression to the knowledge that a certain person is bleeding; it maybe that I do see a bleeding arm or leg, but that because my body is tangled upwith that of someone (e.g., we are wrestling) or because I am seeing myidentical twin or double in a mirror, I ammistaken in thinking the person whois bleeding to be myself. Such statements are subject to error throughmisidenti-fication relative to the first-person pronouns, where to say that a statement ‘a isΦ’ is subject to error through misidentification relative to the term ‘a�’ meansthat the following is possible: the speaker knows some particular thing to beΦ,but makes the mistake of asserting ‘a is Φ’ because, and only because, hemistakenly thinks that the thing he knows to be Φ is what ‘a�’ refers to. Thestatement “I feel pain” is not subject to error throughmisidentification relativeto ‘I’: it cannot happen that I am mistaken in saying “I feel pain” because,although I do know of someone that feels pain, I am mistaken in thinking thatperson to be myself. But this is also true of first-person statements that areclearly not incorrigible; I can be mistaken in saying “I see a canary,” since I canbe mistaken in thinking that what I see is a canary or (in the case of hallucina-tion) that there is anything at all that I see, but it cannot happen I that I ammistaken in saying this because I have misidentified as myself the person I knowto see a canary. And whereas the statement “My arm is moving” is subject toerror through misidentification, the statement “I am waving my arm” is not.

First-person statements that are immune to error throughmisidentificationin the sense just defined, those in which ‘I’ is used “as subject,” could be said tohave “absolute immunity” to error through misidentification. A statement like“I am facing a table” does not have this sort of immunity, for we can imaginecircumstances in which someone might make this statement on the basis ofhaving misidentified someone else (e.g., the person he sees in a mirror) ashimself. But there will be no possibility of such amisidentification if onemakesthis statement on the basis of seeing a table in front of one in the ordinary way(without aid of mirrors, etc.); let us say that when made in this way the state-ment has “circumstantial immunity” to error throughmisidentification relativeto ‘I’. It would appear that, when a self-ascription is circumstantially immuneto error through misidentification, this is always because the speaker knows orbelieves it to be true as a consequence of some other self-ascription, which thespeaker knows or is entitled to believe, that is absolutely immune to errorthrough misidentification; e.g., in the circumstances just imagined the proposi-tion “I am facing a table” would be known or believed as a consequence of the


proposition “I see a table in the center of my field of vision.”3

If I say “I feel pain” or “I see a canary,” I may be identifying for someoneelse the person of whom I am saying that he feels pain or sees a canary. Butthere is also a sense in which my reference does not involve an identification.My use of the word ‘I’ as the subject of my statement is not due to my havingidentified as myself something of which I know, or believe, or wish to say, thatthe predicate of my statement applies to it.

But to say that self-reference in these cases does not involve identificationdoes not adequately capture what is peculiar to it, for these first-person state-ments are not alone in involving reference without identification. Consider twocases in which I might say “This is red.” Suppose that I am selling neckties, thata customer wants a red necktie, and that I believe I have put a particular red silknecktie on a shelf of the showcase that is visible to the customer but not to me.Putting my hand on a necktie on that shelf, and feeling it to be silk, I might say“This one is red.” Here it could be said that I have identified, correctly orincorrectly, the object I refer to in saying ‘this’ as the object I ‘have in mind,’i.e., as the object of which I wish to say that it is red. We can contrast this witha case in which I simply point to a necktie that I see and say “This is red.” In thelatter case there is, in the present sense, no identification and hence no possibil-ity of misidentification. In the first case I intend to refer to a certain red necktieI believe to be on the shelf, but the there is also a sense in which I intend torefer, and do refer, to the necktie actually on the shelf, and there is a possibilityof a disparity betweenmy intended reference andmy actual reference. But therecan be no such disparity in the second case. In this case my intention is simplyto refer to one (a specific one) of the objects I see, and in such a case thespeaker’s intention determines what the reference of his demonstrative pro-noun is and that reference cannot be other than what he intends it to be.

But now let us compare this sort of reference without identification withthat which occurs in first-person statements. The rules governing the use of ademonstrative like ‘this’ do not by themselves determine what its reference is onany given occasion of its use; this is determined, as we have noted, by the

3. A qualification is needed here. Someone who lacks the concept of seeing and the concept of a field ofvision and who, therefore, is in one sense incapable of believing that he sees a table in the center of hisfield of vision, might nevertheless make (and be entitled to make) the statement “I am facing a table” inthe circumstances imagined, i.e., when it is in fact true of him that he sees a table in the center of his fieldof vision. For our present purposes we can perhaps stretch the notion of being entitled to believe that pto cover the case of someone who lacks the concepts needed to express ‘p�’ but who would be entitled tobelieve that p if only he had these concepts.


speaker’s intentions. When a man says “This is red,” there are generally anynumber of things to which he could be referring without misusing the word‘this’, and there is of course no requirement that different tokens of ‘this’ in aman’s discourse should all refer to the same thing. This permits the reference of‘this’ on a particular occasion to be fixed by the speaker’s intention to say of aparticular thing that it is red, i.e., fixed in such a way that it can refer to nothingother than that thing and, consequently, in such a way that his statement “Thisis red” does involve an identification. But we cannot explain in this way thereference without identification that occurs in first-person statements. One canchoose whether or not to use the word ‘I’, but the rules governing the use of thisword determine once and for all what its reference is to be on any givenoccasion its use, namely, that its reference is to the speaker, and leaves nolatitude to the speaker’s intentions in the determination of its reference.

There are other important differences between ‘I’ and demonstratives like‘this’. Although there are cases in which the reference of a demonstrative cannotbe other than what the speaker intends it to be, there is in even these cases thelogical possibility of failure of reference; it may happen, e.g., in cases of halluci-nation, that there simply is no object to which a speaker can truly be said to bereferring in saying “This is red.” But there is, as Descartes’ ‘cogito argument’brings out, no such possibility of failure of reference in the use of the word ‘I’.Again, if I retain in my memory an item of knowledge which at the time of itsacquisition I could have expressed by saying “This is red” and if I wish at a latertime to express that knowledge, it will not do for me simply to utter the past-tense version of the sentence that originally expressed it. In the expression ofmymemory knowledge the word ‘this’ will typically give way to a description ofsome kind, e.g., ‘the thing that was in front of me’, or ‘the thing I was lookingat’. I may of course say at a later time “This was red,” pointing to somethingthen in front of me, but this statement will involve an identification and will besubject to error throughmisidentification. But the appropriate way of expressingthe retained (memory) knowledge that at the timeof its acquisitionwas expressedby the sentence “I see a canary” is to utter the past-tense version of that sen-tence, namely, “I saw a canary.” This, if said on the basis of memory, does notinvolve an identification and is not subject to error through misidentification.4

4. See Hector-Neri Castañeda’s paper ‘“He”: A Study in the Logic of Self-consciousness,’ Ratio, VIII, 2(December 1966): 145, [this volume, xxx] for a closely related point. I have discussed the immunity toerror through misidentification of first-person memory statements in my forthcoming paper ‘Personsand Their Pasts.’ See also ch. IV of my book Self-Knowledge and Self-ldentity (Ithaca, N.Y.: Cornell,1963).


I think that many philosophers have assumed that where self-reference doesnot involve identification it must involve the sort of demonstrative referencethat occurs when one says “This is red” of something one sees. If one makes thisassumption, but then notices that ‘I’ is no more a demonstrative pronoun thanit is a name or ‘disguised description,’ one will quite naturally conclude that inits use “as subject” the word ‘I’ is not a referring expression at all, or asWittgen-stein put it, that it does not “denote a possessor.”

II

Philosophers who have reflected on the “use as subject” of the first-personpronouns have often been inclined to say such things as that one cannot be anobject to oneself, that one’s self is not one of the things one can find or encoun-ter in the world. Themost commonly drawn conclusion, of course, is that one’sself, what one “calls ‘I’,” cannot be any of the physical or material things onefinds in the world. But as is well known to readers of Hume and Kant, amongothers, it is also widely denied that any immaterial object of experience could bethe subject of thought and experience. These views lead naturally to theconclusion that ‘I’ does not refer, that there is no self, or that the self is some-how not “in the world.”

In the Blue Book (op. cit.: 74)Wittgenstein observed that in “I feel pains” we“can’t substitute for ‘I’ a description of a body.” And Thomas Nagel has recentlypointed out, in effect, that there is no description at all which is free of token-reflexive expressions and which can be substituted for ‘I’; no matter howdetailed a token-reflexive-free description of a person is, and whether or not itis couched in physicalistic terms, it cannot possibly entail that I am thatperson.5 Inspired by these considerations, someone might reason as follows:“Nothing that I find in the world can be myself (or Self), for there is nothingthat I could observe or establish concerning any object I find in the world fromwhich I could conclude that it is myself.” This would clearly be a very badargument, for even if its premise were true it would not establish that I cannotfind what is in fact myself in the world; it would only establish that if I found

5. See Nagel’s ‘Physicalism,’ Philosophical Review LXXIV, 3 (July 1965): 353–355.Since the abovewaswritten it has come tomy attention that Castañeda has argued, very persuasive-

ly, that there is no description, not even one containing indexicals (other, presumably, than the first-person pronouns themselves), that can be substituted for ‘I’. See his ‘“He,”’ [1966; this volume] and his‘Indicators and Quasi-Indicators,’ American Philosophical Quarterly, IV, 2 (April 1967): 87 and 95.


what is myself in the world I could not know that it was myself that I had found.But it does not even establish that, for its premise is false. It is true that there isno token-reflexive-free description of any person from which it would followthat that person is myself, but there is no reason why, in establishing whethersomeone is myself, I should be limited to facts about him that can be describedwithout the use of token-reflexive expressions.

In our world there seldom occurs anything that it would be natural todescribe as “finding oneself in the world” or “being an object to oneself.” It isrelatively seldom that we observe ourselves in the ways in which we observeothers. But we can easily imagine a world full of reflecting surfaces, in whichmost seeing involves the intervention of one or moremirrors between what seesand what is seen. And one can perhaps imagine a world in which light raysfollow curved paths, or a non-Euclidean world in which light rays following‘straight’ paths sometimes return to their point of origin. In such worlds onecould be, visually at least, an object to oneself in the way in which others areobjects to one. It is clear that there would be no guarantee, in such a world, that,when observing oneself, one would know that it was oneself one was observing.But it is also clear that there would be no reason in principle why one shouldnot find this out. Presumably one could find it out in much the way in which,in our world, one finds that it is oneself one is seeing in a mirror.

But while there can occur something that is describable as “finding oneselfin the world” or “being an object to oneself,” it is not and could not be on thebasis of this that one makes the first-person statements in which ‘I’ is used “assubject.” It is clear, to begin with, that not every self-ascription could begrounded on an identification of a presented object as oneself. Identifyingsomething as oneself would have to involve either (a) finding something to betrue of it that one independently knows to be true of oneself, i.e., somethingthat identifies it as oneself, or (b) finding that it stands to oneself in somerelationship (e.g., being in the same place as) in which only oneself could standto one. In either case it would involve possessing self-knowledge— the knowl-edge that one has a certain identifying feature, or the knowledge that one standsin a certain relationship to the presented object — which could not itself begrounded on the identification in question. This self-knowledgemight in somecases be grounded on some other identification, but the supposition that everyitem of self-knowledge rests on an identification leads to a vicious infiniteregress. But in any case, and this is perhaps the most important point, theidentification of a presented object as oneself would have to go together withthe possibility of misidentification, and it is precisely the absence of this


possibility that characterizes the use of ‘I’ that concerns us. I think that this isone of the main sources of the mistaken opinion that one cannot be an objectto oneself, which in turn is a source of the view that ‘I’ does not refer. One feelsthat if one ever encounters the referent of ‘I’ in experience, this ought to occuron those occasions on which one’s right to say ‘I’ is most secure, and thatnothing that is not an object to one on those occasions can be its referent at all.

I have just said that identification necessarily goes together with thepossibility of misidentification. It is clear enough why this is so in the case of theidentification as oneself of a flesh-and-blood person who is observed by meansof ordinary sense perception, but it may be questioned whether an identifica-tion of a self as oneself would be subject to error if selves were conceived asintrospectible immaterial substances. And even if my ‘self ’ is a flesh-and-bloodperson, why shouldn’t it be accessible to me (itself) in a way in which it is notaccessible to others, so that in knowing that what is presented to me is present-ed in this special way— from the inside, as it were — I would know that it canbe nothing other than myself?

Now there is a perfectly good sense in which my self is accessible to me ina way in which it is not to others. There are predicates which I apply to others,and which others apply to me, on the basis of observations of behavior, butwhich I do not ascribe to myself on this basis, and these predicates are preciselythose the self-ascription of which is immune to error throughmisidentification.I see nothing wrong with describing the self-ascription of such predicates asmanifestations of self-knowledge or self-awareness. But it is plainly not theoccurrence of self-awareness in this sense that has been denied by those philoso-phers who have denied that one is an object to oneself; e.g., it is not what Humedenied when he said: “I can never catch myself at any time without a percep-tion, and never can observe anything but the perception.”6 What those philos-ophers have wanted to deny, and rightly so, is that this self-awareness is to beexplained in a certain way. They have wanted to deny that there is an experienc-ing or perceiving of one’s self that explains one’s awareness that one is, forexample, in pain in a way analogous to that in which one’s sense perception ofJohn explains one’s knowledge that John has a beard. An essential part of theexplanation of my perceptual awareness that John has a beard is the fact that theobserved properties of the man I perceive, together with other things I know,are sufficient to identify him for me as John. If the awareness that I am in pain

6. A Treatise of Human Nature, ed. by L. A. Selby-Bigge (Oxford, 1888): 252.


had an explanation analogous to this, it would have to be that I “perceive,” by“inner sense,” something whose “observed properties” identify it to me asmyself. And if the supposition that the perception is by “inner sense” is sup-posed to preclude the possibility of misidentification, presumably this must bebecause it guarantees that the perceived self would have a property, namely, theproperty of being an object of my inner sense, which no self other than myselfcould (logically) have and by which I could infallibly identify it as myself. Butof course, in order to identify a self as myself by its possession of this property,I would have to know that I observe it by inner sense, and this self-knowledge,being the ground of my identification of the self as myself, could not itself begrounded on that identification. Yet if it were possible in this one case for myself-knowledge not to be grounded on an identification of a self as myself, thereseems to be no reason at all why this should not be possible in other cases, e.g.,in the case of my knowledge that I feel pain or my knowledge that I see a canary.Thus the supposition that there is observation by inner sense of oneself —where this is something that is supposed to explain, and therefore cannot besimply equated with, the ability to self-ascribe those predicates whose self-ascription is immune to error throughmisidentification— is at best a superflu-ous hypothesis: it explains nothing that cannot be just as easily, and moreeconomically, explained without it.

Yet despite these considerations, it can seem puzzling that self-awareness,of the sort we are concerned with, does not involve being presented to oneselfas an object. I cannot see the redness of a thing without seeing the thing that isred, and it would seem that it should be equally impossible to be aware of a stateof oneself without being aware of that which has that state, i.e., oneself. It mayseem to follow from the view I have been advancing that one is aware of thepredicates of self-ascriptions, or aware of the instantiation of these predicates,without being aware of their subject, i.e., that in which they are instantiated.Thus we have Hume’s view, that one observes “perceptions” but not anythingthat has them, and the view that one sometimes finds in discussions of Descar-tes’ Cogito, that one seems, mysteriously, to be aware of thoughts, or of think-ing, but not of that which thinks. If this strikes one as impossible, and if one isnevertheless persuaded that what is called “self-awareness” does not involvebeing aware of oneself as an object, it may seem that the only possible conclu-sion is that, when used in first-person statements, the expressions ‘feel pain’,‘am angry’, ‘see a tree’, etc., are pseudo-predicates, like ‘is raining’, and that, inits use as subject, the word ‘I’ is a pseudo-subject, like the ‘it’ in “It is raining.”

I think that the main source of trouble here is a tendency to think of


awareness as a kind of perception, i.e., to think of it on the model of sense-perception. I have been denying that self-awareness involves any sort ofperception of oneself, but this should not be taken to mean that in making ajudgment like “I feel pain” one is aware of anything less than the fact that onedoes, oneself, feel pain; in being aware that one feels pain one is, tautologically,aware, not simply that the attribute feel(s) pain is instantiated, but that it isinstantiated in oneself. What makes the matter seem puzzling, I think, is thatone starts off by trying to construe self-awareness on the model of the observa-tional knowledge that a perceived thing has a certain sensory property, and failsto abandon this model, or to abandon it completely and consistently, when onebecomes persuaded that self-awareness does not involve any sort of perceptionof one’s self, i.e., does not involve what I have called “being presented to oneselfas an object.” One tries to construe one’s knowledge of the instantiation of theattribute ascribed in a self-ascription on the model of a case in which one seesor otherwise observes the instantiation of a sensory attribute, like redness, whileat the same time denying that one perceives that in which the attribute isinstantiated. And this, of course, leads to incoherence. The way out of thisincoherence is to abandon completely, not just in part, the perceptual model ofself-knowledge.

What perhaps makes it difficult to abandon the perceptual model is the factthat it can seem to be implied by the very vocabulary we use to express certainpsychological predicates. Thus, for example, we speak of a person as feeling apain in his back or an itch on his nose, and there is an almost irresistibletemptation to construe the cases thus described as cases of someone perceivinga particular of a certain sort — a private, mental, object. But even if we do soconstrue them, this does not really support the perceptual model. The attributeself-ascribed in the statement “I feel a pain” is that of feeling or having a pain,not that of being a pain or of being painful. And whether or not we construe‘feel’ as a perceptual verb and allow that I can be said to feel something to be apain, I can hardly be described as feeling anything to have the attribute of feelinga pain or as feeling this attribute to be instantiated. Our language may suggestthat pains are perceived, but it does not suggest—and it seems tome clearly notto be true— that one perceives the feeling or the ‘having’ of one’s pains.

III

If one finds it puzzling that there can be the sort of self-reference that occursin the use “as subject” of the first-person pronouns, i.e., that there can be self-


ascriptions that are absolutely immune to error throughmisidentification andare not based on self-observation, one should reflect on the fact that if thiswere not possible there would be much else, and much that we take forgranted, that would also not be possible. The question of how it is possible thatthere should be such self-reference is equivalent to the question of how it ispossible that there should be predicates, or attributes, the self-ascription ofwhich is immune to error through misidentification. And this question seemsto me to be at the root of the larger question of how it is possible that thereshould be psychological attributes, or, what is almost though not quite thesame thing, the question of how it is possible that there should be what Straw-son has called “P-predicates.” It has often been held to be one of the definingfeatures of the realm of the mental, or the psychological, that each personknows of his own mental or psychological states in a way in which no otherperson could know of them.We can put what is true in this by saying that thereis an important and central class of psychological predicates, let us call them“P*-predicates,” each of which can be known to be instantiated in such a waythat knowing it to be instantiated in that way is equivalent to knowing it to beinstantiated in oneself.7 There are psychological predicates that are not P*-predicates — e.g., “is highly intelligent.” But I think that those which are notP*-predicates are classified as psychological predicates only because they arerelated in certain ways to those which are; e.g., they are predicable only of thingsof which some P*-predicates are also predicable, and many of them ascribedispositions that manifest themselves in the having of P*-predicates.8 If this isright, the question of how it is possible that there should be psychologicalpredicates turns essentially on the question of how it is possible that thereshould be P*-predicates, and this is the same as the question of how it is

7. A more explicit formulation is this: Φ is a P*-predicate if and only if there is a way w of knowing Φto be instantiated such that, necessarily, S knowsΦ to be instantiated in wayw if and only if S knows thathe himself is Φ. It is a consequence of this that, although self-ascriptions of P*-predicates need not beincorrigible and although it is not necessarily the case that if a P*-predicate applies to a person thatperson knows that it applies to him, it is necessarily the case that if a person knows that a P*-predicateapplies to him he knows that it applies to him in the “special way” appropriate to that predicate (whichdoes not preclude that he should also be in a position to know that it applies to him in other ways, i.e.,ways in which others might know that it applies to him). Thus if one construes ‘feeling pain’ as the“special way” in which a person knows that he is in pain and holds that it is possible for a person to bein pain and know that he is in pain (e.g., on the basis of his behavior) without feeling pain, one shouldhold that the predicate of ‘is (am) in pain’ is not a P*-predicate and that its self-ascription is not immuneto error through misidentification. But I can see no reason for holding this view.

8. Strawson makes closely related points about what he refers to as “some important classes ofP-predicates”; see his Individuals (New York: Doubleday, 1959):104–107.


possible that there should be predicates the self-ascription of which is absolutelyimmune to error throughmisidentification.

There is another question that seems to me to turn essentially on thisquestion, namely, the question of how it is possible that there should be a first-person pronoun at all; i.e., how it is possible that people should be able toemploy a referring expression whose meaning is given by the rule that it refersto the person who uses it. There, I think, is an important sense in which the“use as subject” of the first-person pronouns is more fundamental than their“use as object.”

It is possible to imagine a people who speak a primitive language containinga first-person pronoun but no P*-predicates — let us suppose that the onlypredicates in this language, and thus the only predicates self-ascribed by itsspeakers, are what Strawson calls “M-predicates,” i.e., predicates that do not“imply the possession of consciousness on the part of that to which they areascribed” (105). As I have already noted, it is possible for there to be self-ascriptions involving self-identification only if there are some self-ascriptionsthat do not involve self-identification. Now there are M-predicates, e.g., “isfacing a table,” which can in some circumstances be self-ascribed withoutidentification. But in order to describe the circumstances in which such self-ascriptions could occur and in order to formulate the grounds of such self-ascriptions, it would be necessary to employ predicates, P*-predicates, thatcould not be expressed in our imaginary language. A speaker of this languagewould have to learn to self-ascribe suchM-predicates as ‘is facing a table’ underjust those circumstances in which he would be entitled to self-ascribe certainP*-predicates, e.g., ‘sees a table in the center of one’s field of vision’, if only hehad these P*-predicates in his vocabulary. And if he can be taught to self-ascribean M-predicate in this way, thus showing that he can discriminate betweencases in which a certain P*-predicate applies to him and cases in which it doesnot, there would seem to be no reason in principle why he could not be taughtto self-ascribe the P*-predicate itself. I think we can say that anyone who canself-ascribe any predicate whatever thereby shows that he is potentially capableof self-ascribing some P*-predicates, and that if he is presently incapable ofdoing so this is due simply to a correctable lack in his vocabulary or his stock ofconcepts. Something similar can be said of other sorts of reference. It is acondition of someone’s being able to make a demonstrative reference, of thesort that does not involve identification, that he should in some way perceivethe object referred to. Anyone who can correctly employ referring expressionsof the form ‘this so and so’ thereby shows that he is potentially capable of self-


ascribing P*-predicates of the form ‘perceives a so and so’.There is another way of indicating the priority I am claiming for the use “as

subject” of ‘I’ over its use “as object.” The clearest cases of the use “as object”are those in which the predicate self-ascribed is anM-predicate. Nowwhere ‘Φ’is an M-predicate, to say that I am Φ is to say that my body is Φ. And if askedwhat it means to call a body “my body” I could say something like this: “Mybody is the body from whose eyes I see, the body whose mouth emits soundswhen I speak, the body whose arm goes up when I raise my arm, the body thathas something pressing against it when I feel pressure, so on.” All the uses of ‘I’that occur in this explanation of the meaning of the phrase ‘my body’, which inturn can be used to explicate the use “as object” of the first-person pronouns inthe self-ascription ofM-predicates, are themselves uses “as subject”. To put thisin another way, M-predicates are mine in virtue of being connected in a certainway with P*-predicates that are mine.9

There is, I think, a tendency to find the use “as subject” of ‘I’ mysteriousand to think that it is perhaps not reference at all, because it cannot be assimi-lated to other sorts of reference, e.g., to the use “as object” of ‘I’ or to demon-strative reference, the latter being taken as paradigms of unproblematicreference. This tendency ought not to survive the realization that these othersorts of reference are possible only because this sort of self-reference, thatinvolving the use “as subject” of ‘I’, is possible. There is, I think, an importantsense in which each person’s system of reference has that person himself as itsanchoring point, and it is important for an understanding of the notion ofreference, and also for an understanding of the notion of the mental, that weunderstand why and how this is so.

9. In ‘Physicalism’ (op. cit.) Thomas Nagel mentions this view, that “My physical states are onlyderivatively mine, since they are states of a body which is mine in virtue of being related in theappropriate way to my psychological states,” as a source of the view that the subject of psychologicalstates cannot be the body. The reasoning is that since the psychological states “are mine in an original,and not merely derivative, sense,” their subject “cannot be the body which is derivatively mine.” I thinkthat the answer to this is that it is only under a certain description, namely, qua subject of certainM-predicates, that my body is “derivatively mine,” and that this is compatible with it, that same thing,being “nonderivatively mine” under some other description, e.g., “subject of my thoughts andexperiences”. I should mention that Nagel does not endorse the reasoning that I am here rejecting.

Chapter 6

Self-identification*

Gareth EvansUniversity of Oxford

1. Introductory

I approach the subject-matter of the present chapter with some trepidation.‘I’-thoughts give rise to the most challenging philosophical questions, whichhave exercised the most considerable philosophers, including Descartes, Kant,and Wittgenstein, and I have no illusion that I am able to answer these ques-tions. (For one thing, there can be no complete understanding of self-identifica-tion without an understanding of the self-ascription of mental predicates; andno adequate understanding of the self-ascription of mental predicates withoutan account of the significance of those predicates — in short, without anaccount of the mind.)

However, there are reasons why a work on the general theory of referencecannot simply ignore the problems of self-identification, even if it does notpurport to give a definitive answer to them. It cannot be assumed that thismode of identification, which is anyway known to give rise to difficult ques-tions, can be fitted into whatever general framework has been constructedfor understanding thoughts about particular objects. And while it would bepresumptuous to look for much in the way of dividends in the understandingof the general philosophical problems of the self, it is not presumptuous toexamine self- identification in the hope that light will be cast upon the modesof identification which we have been considering, and that some of the ideasthrown up in the course of the previous chapter may have application. For,despite considerable differences, ‘I’-thoughts are thoughts of the same generalcharacter as ‘here’ thoughts and ‘this’-thoughts.

* This chapter reproduces Chapter Seven of Evans 1982 in its entirety, except for four short appendicesat the end of the original.

96 Gareth Evans

Let me begin by explaining very generally what I take this similarity toconsist in.

When we are interested in ‘I’-thoughts, we are interested in thoughts whichmight typically be expressed with the use of the first-person pronoun.1 We arenot interested in all thoughts which a subject may have ‘about himself ’, forpresumably a person may think about someone who is in fact himself withoutrealizing that he is doing so. Oedipus was thinking about Oedipus, that is to say,himself, when he thought that the slayer of Laius should be killed; but Oedipuswas not thinking about himself ‘self-consciously’ (this is just a label for the kindof thinking which interests us), because he did not realize that he was the slayerof Laius.2

What is it for Oedipus to realize that he is the slayer of Laius? One thingseems clear: it is not to realize that the Φ is the slayer of Laius, for any descrip-tive conceptΦ. It is not to realize that the son of Jocasta is the slayer of Laius, orthat the man who answered the riddle of the Sphinx is the slayer of Laius,because Oedipusmight realize these things without realizing that he is the slayerof Laius (not knowing that, or having forgotten that, he is the son of Jocasta orthe man who answered the Sphinx’s riddle); and he might realize that he is theslayer of Laius without realizing these things, for the same reason. (This is tooshort an argument on a difficult point, but it has been considerably filled out inthe literature, and will be substantiated in the course of this chapter.)3

There seem to be at least two indispensable consequences which we shouldexpect from such a realization. In the first place, Oedipus must appreciate therelevance, to propositions of the form ‘The slayer of Laius is F�’, of the variousspecial ways he has (as every person does) of gaining knowledge about himself.Secondly, Oedipus must realize how to act upon propositions of the form ‘The

1. Though it seems to me to be completely inessential that there should exist such a device in thesubject’s language.

2. However, in order to avoid unnecessary prolixity, I shall use phrases like ‘think of oneself ’ and ‘Ideaof oneself ’ in the ‘intensional’ rather than the ‘extensional’ sense, so that they are equivalent to ‘think ofoneself self-consciously’ and ‘self-conscious Idea of oneself ’.

3. See, e.g., Hector-Neri Castañeda, ‘“He”: a Study in the Logic of Self-Consciousness’, Ratio 8 (1966),130–157 [this volume, xxx]; and ‘Indicators and Quasi-Indicators’, American Philosophical Quarterly 4(1967), 85–100. The point needs making with some care. Descriptions like ‘the person in front of this�’,or ‘the person here now,’, might seem to yield counterexamples. But it would be pointless to object onthis score. Counting these as counterexamples would require a priority of ‘this’ and ‘here’ over ‘I’, whichis indefensible, as will become obvious. (In any case the definition of ‘I’ in terms of ‘here’ would behopeless: one can think of onself while hurtling through space, and consequently unable to secure a gripon anything for one’s ‘here’-thoughts: see Evans 1982, §6.3.)

Self-identification 97

slayer of Laius is F�’. For example, if Oedipus believes that they are looking forthe slayer of Laius, and if he does not wish to be apprehended, then he shouldmake himself scarce.4 That is to say, the Idea which one has of oneself involvesthe same kinds of elements as we discerned in the case of, say, ‘here’ (Evans1982, §6.3): an element involving sensitivity of thoughts to certain information,and an element involving the way in which thoughts are manifested in action.Such an analysis would certainly explain the widely recognized irreducibility ofself-conscious thoughts to thoughts involving definite descriptions: for nodescriptive thoughts could guarantee the existence of these special dispositions.

Of course, it must be recognized immediately that there are crucial differ-ences. We clearly do have ways of gaining knowledge of ourselves, and‘I’-thoughts are thoughts which are controlled, or are disposed to be controlled,by information gained in these ways. But the ways in which we are sensitive tothe states of ourselves are both more varied and more complex than thesensitivity to places that underlies our ‘here’-Ideas. One quite unprecedentedfeature is the way in which ‘I’-thoughts depend upon the knowledge we have inmemory of our past states (see §5). And an evenmore important difference liesin the fact that the essence of ‘I’ is self-reference. This means that ‘I’-thoughtsare thoughts in which a subject of thought and action is thinking about himself— i.e. about a subject of thought and action. It is true that I manifest self-conscious thought, like ‘here’-thought, in action; but I manifest it, not inknowing which object to act upon, but in acting. (I do not move myself; Imyself move.) Equally, I do not merely have knowledge of myself, as I mighthave knowledge of a place: I have knowledge of myself as someone who hasknowledge and who makes judgements, including those judgements I makeabout myself.

Nevertheless, despite these important differences, the ingredients aresufficiently familiar for us to believe that we might get to know our way about.

It is worth clearing away, at this early stage, a curious idea, expressed indifferent ways by Geach and by Strawson, to the effect that the interest of ‘I’ is

4. This element in an account of ‘I’ is stressed in much recent work: e.g. John Perry, ‘Frege onDemonstratives’,The Philosophical Review 86:474–97, and ‘The Problem of the Essential Indexical’,Nous13 (1979): 3–21 [this volume]; David Lewis, ‘Attitudes De Dicto and De Se�’, Philosophical Review 88(1979): 513–543. Neglect, in this work, of the other element produces a strangely one-sided effect —‘strangely’, because the other element is just as striking, and clearly parallel, and also because thedominant conception of the identification of empirical content concentrates exclusively on the input orevidential side of things. This chapter will partly redress the balance by rather neglecting the actioncomponent.

98 Gareth Evans

exclusively the interest of a communicative device— that is, in effect, that thereare no Ideas corresponding to the pronoun and available to be exercised inthinking. According to Geach,5 Descartes in his solitary meditations had no needof ‘I’ in such judgements as ‘I’m getting into an awful muddle’; he could havejudged ‘This is really a dreadful muddle!’ Similarly, Strawson6 suggests that it isright to speak of self-ascription, for instance of pain, only because one tells othersthat one is in pain; otherwise one’s judgement can simply be ‘There is a pain’. Boththese philosophers are preoccupied with the fact that there is no need for me totell myself who it is who is getting into a muddle, or who is in pain.

But there is a mistake here. Reference, as a communicative phenomenon,involves getting an audience to think of the right object (the intended object).Obviously, thinking of an object does not consist in getting oneself to think ofthe right object (the intended object). But surely this cannot show that there isno such thing as thinking of an object, in a certain way, outside of communica-tive contexts.

Indispensable though those familiar ingredients (an information compo-nent and an action component) are in any account of the Ideas we have ofourselves, our previous reflections (Chapter 6) have made it sufficiently clearthat they cannot constitute an exhaustive account of our ‘I’-Ideas. So long as wefocus upon judgements which a person might make about himself upon thebasis of the relevant ways of gaining knowledge, the inadequacy may not strikeus. A subject’s knowledge of what it is for the thought ‘I am in pain’ to be truemay appear to be exhausted by his capacity to decide, simply upon the basis ofhow he feels, whether or not it is true— and similarly in the case of all the otherways of gaining knowledge about ourselves. However, our view of ourselves isnot Idealistic: we are perfectly capable of grasping propositions about ourselveswhich we are quite incapable of deciding, or even offering grounds for. I cangrasp the thought that I was breast-fed, for example, or that I was unhappy onmy first birthday, or that I tossed and turned in my sleep last night, or that Ishall be dragged unconscious through the streets of Chicago, or that I shall die.In other words, our thinking about ourselves conforms to the GeneralityConstraint.7 And this means that one’s Idea of oneself must also comprise, over

5. Mental Acts, Ch 26.

6. Individuals, pp. 99–100.

7. This thought is diametrically opposed to a line of thought of Wittgenstein’s, in which he encouragedus to look at first-person psychological statements in a way that brought out their similarity to groans ofpain — i.e. precisely to think of them as unstructured responses to situations. (He was well aware that


and above the information-link and the action link, a knowledge of what itwould be for an identity of the form I = δt to be true, where δt is a fundamen-tal identification of a person: an identification of a person which—unlike one’s‘I’ identification— is of a kind which could be available to someone else. Onlyif this is so can one’s general understanding of what it is for a person to satisfythe predicates ‘ξ is dead’, ‘ξ is breast-fed’, ‘ξ is unhappy’, etc. be coupled withone’s Idea of oneself to yield an understanding of what it would be for oneselfto satisfy these predicates. (The tenses in the various examples will be taken careof by our knowledge of what it is for δt = δ¢t¢ to be true, when t π t¢�; i.e. by ourgrasp of the identity-conditions of persons over time: cf. Evans 1982, §4.3.)

My insistence that ‘I’-Ideas be recognized to conform to the GeneralityConstraint is correlative with Strawson’s parallel insistence that the Ideas (orconcepts) of properties of consciousness obey a similar constraint: namely (toput it in my terms) that anyone who has a grasp of the concept of being F mustbe able to understand what it is for an arbitrary proposition of the form a is Fto be true (where a is an Idea which he possesses of an object).8 The GeneralityConstraint requires us to see the thought that a is F as lying at the intersectionof two series of thoughts: the thoughts that a is F, that a is G, that a is H,…, onthe one hand, and the thoughts that a is F, that b is F, that c is F,…, on theother. Strawson has explored the consequences of one kind of generality, and Iam exploring the consequences of the other.

It is vital to remember this feature of our thought about ourselves.‘I’-thoughts are not, as is sometimes suggested, restricted to thoughts aboutstates of affairs ‘from the point of view of the subject’. Nor can the thoughts Ihave been discussing be hived off from genuine self-conscious thought, forexample by suggesting that by ‘I will die’, I mean that Gareth Evans will die. Notat all; there is just as much of a gap between the knowledge that Gareth Evanswill die and the self-conscious realization that I will die as there is between anythought to the effect that theΦ is F and the self-conscious thought that I am F.

It is not wholly inaccurate to say that I grasp such an eventuality by think-ing of myself in the way that I think of others; this is just another way of sayingthat the fundamental level of thought about persons is involved. But it is ofcourse essential that I am aware that the person of whom I am so thinking ismyself�; certainly I must have in mind what it is for δ is dead to be true, for

this would enable him not to think about certain issues.) See Philosophical Investigations (Blackwell,Oxford, 1953): §§404–406.

8. See Individuals, p. 99.

100 Gareth Evans

arbitrary δ, but I must also have in mind what it is for δ = I to be true. Mythought aboutmyself does satisfy the Generality Constraint; and this is becauseI can make sense of identifying a person, conceived from the standpoint of anobjective view of the world, as myself.9

It has been suggested that we do not in fact understand what it is for suchan identity to be true. Thomas Nagel has written:

I can conceive impersonally my house burning down, and the individual T. N.standing before it, feeling hot andmiserable, and looking hot andmiserable tobystanders … If I add to all this the premiss that I am T.N., I will imaginefeeling hot andmiserable,.seeing the sympathetic bystanders, etc.; but this is notto imagine anything happening differently.10

Earlier he wrote:

The addition of this premiss makes a great difference in how [the] world isconceived, but no difference in what is conceived to be the case.11

It is upon this basis that Nagel suggests the existence of an unbridgeable gulfbetween subjective and objective — between, as I should put it, propositionsabout persons and objects formulated at the fundamental level of thought, andpropositions formulated with such Ideas as ‘I’, ‘here’, and ‘this’. Nagel envisagesthe universe considered sub specie aeternitas, and then wonders how to incorpo-rate into that model the fact that such-and-such a person is me. Since thisidentification does not seem to make any difference to the model — nothing isdifferently conceived — Nagel suggests that we cannot really understand whatit is for such an identity-proposition to be true.

But in fact I have already implicitly explained what is involved in graspingsuch an identity-proposition — in knowing what it is for such an identity-proposition to be true. It seems to me clear that as we conceive of persons, theyare distinguished from one another by fundamental grounds of difference of thesame kind as those which distinguish other physical things, and that a funda-mental identification of a person involves a consideration of him as the person

9. Even if we consider cases in which the subject is normally in a position to that he satisfies a predicate,we do not take an Idealistic view of such judgements. There remains a gap— the ever-present possibilityof error — between evidence and conclusion. The idea that I can identify myself with a personobjectively conceived is often mis-expressed, e.g., in terms of the idea that I realize that I am an objectto others (also an object of outer sense, as Kant says: Critique of Pure Reason, B415). This misleadinglyimports an ideal verificationist construal of the point. (See Evans 1982, §§4.2, 6.3.)

10. The Possibility of Altruism (Clarendon Press, Oxford, 1970),: 103.

11. Ibid.


occupying such-and-such a spatio-temporal location.12 Consequently, toknow what it is for δt = I to be true, for arbitrary δt,, is to know what isinvolved in locating oneself in a spatio-temporal map of the world.13 (SeeEvans 1982, §6.3.) Such an identity-proposition need not make any differenceto how the spatio-temporal map of the world is conceived, but it will make agreat difference to how the subject’s immediate environment is conceived.(Nagel was looking for the impact in the wrong place.) It is true that we cannotstate in non-indexical terms what it is for the identity-proposition to be true;but why should we suppose that everything that is true can be represented inthat way? Nagel may conclude that propositions like I am δt are not objective-ly true — true from the standpoint of eternity. I should not feel obliged toquarrel with this, since it is indeed true that such a proposition is capable ofbeing grasped only by the person who can formulate it; so if ‘objective’ means‘graspable by anyone’, such identifications are indeed not objectively true.

Nagel’s suggestion — that we do not really understand what it is for us tobe identical with objects conceived to be parts of the objective spatio-temporalframework — surely must be wrong. Were it correct, our thinking aboutourselves could not conform to the Generality Constraint. We would then haveto suppose that we have an Idealist conception of the self.14 Conversely, just asour thoughts about ourselves require the intelligibility of this link with theworld thought of ‘objectively’, so our ‘objective’ thought about the world alsorequires the intelligibility of this link. For no one can be credited with an‘objective’ model of the world if he does not grasp that he is modeling the worldhe is in— that he has a location somewhere in the model, as do the things thathe can see. Nothing can be a cognitive map unless it can he used as a map —unless the world as perceived, and the world as mapped, can be identified. Forthis reason, I think that the gulf between the ‘subjective’ and ‘objective’ modesof thought which Nagel tries to set up is spurious. Each is indispensably boundup with the other.

Despite the important differences, then, between ‘I’ and ‘this’ and ‘here’, thegeneral structure of our account of these Ideas is the same. In particular, a

12. I stress that I am speaking of our ordinary scheme of thought. There are other conceptions — e.g.,conceptions of control systems or information stores — which might serve some of our purposes, andwhich would involve different fundamental grounds of difference.

13. (See §4.)

14. This would be the same as saying that ‘I’ does not refer to anything. (This would be a reason for theextraordinary conclusion of G.E.M.Anscombe, ‘The First Person’, in Samuel Guttenplan, ed., Mind andLanguage, pp. 45–65.

102 Gareth Evans

subject’s self-conscious thought about himself must be informed (or must atleast he liable to be informed) by information which the subject may gain ofhimself in each of a range of ways of gaining knowledge of himself; and at thesame time the subject must know which object it is of which he thus has, or iscapable of having, knowledge.

Now even if it is commonly agreed that there is a connection betweenthinking of oneself self-consciously and thinking in ways that are liable to beinformed by certain kinds of evidence or information about oneself which oneis capable of acquiring,15 there is widespread misunderstanding about whatkinds of evidence or information they are. For it is widely believed that self-conscious thought is exclusively thought informed by the knowledge that onemay have about one’s own mental life. I shall show that this is quite incorrect.

It is true, as I said earlier, that the essence of self-consciousness is self-reference, that is to say, thinking, by a subject of judgements, about himself, andhence, necessarily, about a subject of judgements. (This means that we shall nothave an adequate model of self-consciousness until our model provides for thethought, by a subject, of his own judgements. Without that, no matter howmuch it mimicked our own use of ‘I’, the model would always be open to asceptical challenge: how can it be guaranteed that the subject is referring tohimself?) It follows that in a self-conscious thought, the subject must think ofan object in a way that permits it to be characterized as the subject of that verythought. But it certainly does not follow that he must think of himself as theauthor of that very thought — if, indeed, such a thing is intelligible; nor, moregenerally, that he must think of himself exclusively as an author of judgements,or even as a possessor of a mental life. On the contrary, we shall see that ourself-conscious thoughts about ourselves also rest upon various ways we have ofgaining knowledge of ourselves as physical things. If there is to be a divisionbetween the mental and the physical, it is a division which is spanned by theIdeas we have of ourselves.

Before moving on to an elaboration of the ways of gaining knowledge ofourselves on which our ‘I’-thoughts depend, I want to give a warning about adanger inherent in all reflections about self-consciousness.

Up to this point, we have been able to take the subject of thought, and his

15. “Just as I cannot know what form the evidence of a fire will take unless I know whether the fire ispast, present, or future, so I cannot know what to expect in the way of evidence that one of the personsin a group has been poisoned unless I know whether it is I or someone else.” Nagel, The Possibility ofAltruism, p. 103.


identity, for granted. We have been able to say such things as ‘The subject isdisposed to treat this or that state of information as germane to such-and-sucha thought’. For instance, we might have said that the subject’s knowledge ofwhat it is for such-and-such a future-tense proposition to be true depends onhis ability to tell, later, whether or not it is true. Now there is no harm incontinuing this way of proceeding when we come to consider self-identifica-tion: indeed it is unavoidable. But we must realize what we are doing. We arebuilding the subject’s identity over time into the description of his situation.This maymake it appear that he has an infallible knowledge of what is involvedin this identity; but the appearance is nothing but an artefact of our way ofdescribing the situation.16

Let me give two connected examples of the kind of mistake I wish to warnagainst. G.E.M. Anscombe, in her fascinating paper on ‘I’-thoughts,17 observesthat it is not possible for the subject to identify different things by the various‘I’-identifications he makes over time. It is not possible for there to be an‘unnoticed substitution’, so that he thinks that he is identifying the same thing,when in fact he is not doing so. This corresponds to nothing in the case ofrepeated identifications of objects other than the subject himself, and such alogical guarantee of correctness makes Miss Anscombe suspicious; it is one ofthe reasons she gives for her extraordinary conclusion that self-consciousthought is not thought about an object at all — that the self is not an object. But,of course, the ‘logical guarantee’ is simply produced by Miss Anscombe’s wayof describing the situation, in terms of one and the same subject havingthoughts at various times. It is a simple tautology that, if it is correct to describethe situation thus, the self-identifications are all identifications of the same self,and hence it cannot be a reason for anything.18

In the second example, the mistake seems to give a subject an infallibleknowledge of what it is for a state of affairs to concern his own future. Pursuingthe style of description which served us adequately in the discussion of thesubject’s grasp of future tense propositions about other objects, we might saysomething like this: the subject’s knowledge of what is involved in a future state

16. I think Kant may have had this phenomenon inmind, as much as anything which depends speciallyupon memory, when he spoke, in the Third Paralogism, of the “logical identity of the ‘I’” (Critique ofPure Reason, A363).

17. ‘The First Person’, op. cit.

18. There are phenomena (having to do with memory) which are superficially similar to the spuriousone that Miss Anscombe describes: see §5.

104 Gareth Evans

of affairs concerning himself can depend upon his ability, when the time comes,to decide whether or not the state of affairs obtains. We might say: certainly,when time t comes, he will know whether or not the hypothesis that he ex-pressed earlier by ‘I’ll be in pain at t�’ was or was not correct, just by whether ornot he is in pain at t�; so all he has to envisage, when he envisages the future stateof affairs of his being in pain, is a future pain. What more could possibly heinvolved? For this certainly seems to be a foolproof method for verifying theprediction. It is not possible for the subject to have got hold of the wrongperson at time t.

Now, I do not mean to deny that there is something correct about this, asa description of the subject’s envisagings about his own future.19 But what issuspicious is the complete adequacy we have built into his conception by ourway of describing it. Of course it is not possible for the subject to have got holdof the wrong person — as the case is described, there is a logical guarantee ofadequacy. But this is, again, an artefact of our way of describing the situation:it certainly does not show that, just by envisaging future situations, a subject hasa complete and clear conception of what it is for a future state of affairs toinvolve himself. The ‘method of verification’ has a presupposition. Of course wemust not say (using the ordinary vocabulary): it presupposes that the subjectremains the same over time. But it presupposes that the subject who exists at tand ‘remembers’ the hypothesis expressed earlier is the person who made thehypothesis, and hence is the person whom it concerns. And this is something ofwhich he can have no genuine logical guarantee.

Forewarned against these errors, in what follows I shall continue to use theordinary vocabulary.

2. Immunity to error through misidentification

If an analogy is to be sought between self-identification and one of the modesof identification we considered in the last chapter, it is ‘here’ rather than ‘this’which provides the closer parallel. Just as it is not necessary, if a subject is to bethinking about a place as ‘here’, that he actually have any information deriving

19. My point is not to deny that there is such a thing as criterionless self-ascription of anticipatedproperties. (I think this is simply the other side of the same coin as criterionless self-ascription ofremembered properties, for which see §5.) But I do want to deny that this is a matter of a logicalguarantee of an identity assumption.


from it, so it seems not to be necessary, if a subject is to think about himselfself-consciously, that he actually have any information about himself. A subjectmay be amnesiac and anaesthetized, and his senses may be prevented fromfunctioning; yet he may still be able to think about himself, wondering, forexample, why he is not receiving information in the usual ways.20 But it wouldbe as wrong to conclude from this that self-consciousness can be explainedwithout reference to the various ways that subjects have of gaining knowledgeabout themselves, as we decided it would be to draw the parallel conclusionabout ‘here’ (see Evans 1982, §6.3). It is essential, if a subject is to be thinkingabout himself self-consciously, that he be disposed to have such thinkingcontrolled by information which may become available to him in each of therelevant ways. Or, at least, so I shall argue.21

So to argue is to claim that each of these ways of gaining knowledge ofourselves gives rise to judgements which exhibit the phenomenon I called in thelast chapter ‘immunity to error through misidentification’ (see Evans 1982,§6.6). And certainly there seem to be indications that this phenomenon doesarise. To take the stock kind of example, when the first component is expressiveof knowledge which the subject has about his own states, available to him in thenormal way, and not taken by him to be knowledge which he has gained, ormay have gained, in any other way,22 the utterance ‘Someone seems to seesomething red, but is it I who seem to see something red?’ does not appear tomake sense. But the phenomenon appears to be more widespread than the stockexamples. For example, it seems equally not to make sense for a subject to utter‘Someone’s legs are crossed, but is it I whose legs are crossed?’, when the firstcomponent is expressive of knowledge which the subject has gained about theposition of his limbs, available to him in the normal way.

Unfortunately, many philosophers give the quite mistaken impression thatit is only our knowledge of our satisfaction of mental properties which gives rise

20. See Anscombe,’The First Person’: 57–58. This is another basis on which she attempts to found herextraordinary conclusion. See n. 21 below.

21. It is not surprising that ‘I’ follows themodel of ‘here’ rather than ‘this’. The explanation is somewhatsimilar to the explanation for ‘here’ (see Evans 1982, §6.3). A subject does not need to have informationactually available to him in any of the relevant ways in order to know that there is just one object towhich he is thus dispositionally related. (This undermines Miss Anscombe’s argument about theanaesthetizedman, that since no object presented tohim, there is no room for him to use a demonstrativeexpression referring to himself.)

22. See Evans 1982, §6.6 for an explanation of the need for this qualification (which I shall drop infuture).

106 Gareth Evans

to judgements exhibiting immunity to error through misidentification.23 Thisis tantamount to the claim that self-conscious thought rests only upon theknowledge we have of ourselves as mental or spiritual beings. And this in turngenerates the unfortunate, and quite inaccurate, impression that in thinking ofoneself self-consciously, one is paradigmatically thinking about oneself as thebearer of mental properties, or as a mind — so that our ‘I’-thoughts leave itopen, as a possibility, that we are perhaps nothing but a mind. In order toeradicate this impression, we must go back to its source: a remarkable passageof Wittgenstein’s where the phenomenon of immunity to error throughmisidentification is noticed for the first time.

Now the idea that the real I lives in my body is connected with the peculiargrammar of the word ‘I’, and the misunderstandings this grammar is liable togive rise to. There are two different cases in the use of the word ‘I’ (or ‘my’)which I might call ‘the use as object’ and ‘the use as subject’. Examples of thefirst kind of use are these: ‘My arm is broken’,’I have grown six inches’, ‘I havea bump on my forehead’, ‘The wind blows my hair about’. Examples of thesecond kind are: ‘I see so and so’, ‘I hear so and so’, ‘I try to lift my arm’, ‘Ithink it will rain’, ‘I have a toothache’. One can point to the difference betweenthese two categories by saying: The cases of the first category involve therecognition of a particular person, and there is in these cases the possibility ofan error, or as I should rather put it: The possibility of an error has beenprovided for… It is possible that, say in an accident, I should feel a pain in myarm, see a broken arm at my side, and think it is mine, when really it is myneighbour’s. And I could, looking into a mirror, mistake a bump on hisforehead for one onmine. On the other hand there is no question of recogniz-ing a person when I say I have toothache. To ask ‘are you sure that it’s you whohave pains?’ would be nonsensical … And now this way of stating our ideasuggests itself that it is impossible that in making the statement ‘I have atoothache’ I should have mistaken another person for myself as it is to moanwith pain by mistake, having mistaken someone else for me. To say, ‘I havepain’ is no more a statement about a particular person than moaning is. ‘Butsurely the word “I” in the mouth of a man refers to the man who says it; itpoints to himself …’. But it was quite superfluous to point to himself.24

23. See Strawson, The Bounds of Sense�: 164–165: both of his examples relate to mental self-ascription.See also Shoemaker, ‘Self-Reference and Self-Awareness’ [this volume]. Shoemaker does note that thereare other kinds of statement that exhibit this immunity, e.g., ‘I am facing a table’ (cf. §3 below); but heargues that their possession of the property is only derivative, in virtue of the involvement of thepsychological judgement ‘There is a table in my field of vision’: see p. 557.

24. The Blue and Brown Books [1933–4]: 66–67.


It is worth briefly observing, first, that where there is immunity to error throughmisidentification, Wittgenstein draws the conclusion that the word ‘I’ is notbeing used to refer to (talk about) a particular object (a person). This seems tobe just a mistake. For we have seen (Evans 1982, §6.6) that immunity to errorthrough misidentification is a straightforward consequence of demonstrativeidentification; it will exist whenever a subject’s Idea of an object depends uponhis ways of gaining knowledge about it. And demonstrative identification is,precisely, a way in which a thought can concern (be about) an object.25

The word ‘identify’ can do us a disservice here. In one sense, anyone whothinks about an object identifies that object (in thought): this is the senseinvolved in the use I have just made of the phrase ‘demonstrative identifica-tion’. It is quite another matter, as we saw, in effect, in Evans 1982, §6.6, for thethought to involve an identification component — for the thought to beidentification-dependent. There is a danger of moving from the fact that thereis no identification in the latter sense (that no criteria of recognition arebrought to bear, and so forth) to the conclusion that there is no identification inthe former sense. I am not sure thatWittgenstein altogether avoids this danger.

But this conclusion is not our present concern, but rather, Wittgenstein’streatment of examples like ‘The wind is blowing my hair’. For it was thistreatment which gave rise to the widespread belief that the phenomenon ofimmunity to error through misidentification, which is so central to the notionof self-consciousness, does not extend to self-ascriptions of physical properties.But of course it does. There is a way of knowing that the property of ξ�’s hairbeing blown by the wind is currently instantiated, such that when the firstcomponent expresses knowledge gained in this way, the utterance ‘The wind isblowing someone’s hair, but is it my hair that the wind is blowing?’ will notmake sense.Wittgenstein’s discussion does not take sufficient account of the factthat the property of being immune to error throughmisidentification is not one

25. One can perhaps imagine why Wittgenstein drew the conclusion. The information one seems tobe presented with is simply that such-and-such a property is instantiated; there does not appear to beanything in the information to tell one which object instantiates the property (as there might be if therecognizable appearance of an object was also presented). So if one was talking about a particularobject (person) one would be going beyond the information given. But we saw that in the case ofdemonstrative identification this conclusion will not follow if the subject can be credited with aknowledge of which object the information concerns; which he would nave, in that case, by virtue ofbeing able to locate it. Nowwe have not yet enquired into the question in what out knowledge of whichperson we are consists. But it seems reasonable to suppose that we have such knowledge; and in thecontext of that supposition, the point about the content of the information by itself goes only a veryshort distance. (See, further, §4 below).

108 Gareth Evans

which applies to propositions simpliciter, but one which applies only to judge-mentsmade upon this or that basis. Once we appreciate this relativity to a basis,which arguablymust be taken into account in the case of mental self-ascriptionas well, the fact that there are cases involving the self-ascription of physicalpredicates in which ‘the possibility of error has been provided for’ will be seennot to impugn the fact that there are cases in which it just as clearly has not.

It may be suggested that Wittgenstein is concerned with a different notion:that his question is not whether there is a way of knowing that one instantiatessome property which generates immunity to error through misidentificationin my sense, but whether there is a way of knowing it which does not. (Only ifthis last is not so will we have immunity to error through misidentification inthe strong sense with which, according to this suggestion,Wittgenstein is con-cerned.) After all, it may be said, it is the possibility of discovering that what isin fact one’s own arm is bent, without knowing that it is one’s own, to whichhe appeals.

But, first, the evidence that this is Wittgenstein’s concern is uncertain. Therelevant direct statement is

The cases of the first category involve the recognition of a particular person,and there is in these cases the possibility of an error, or as I should rather putit: ‘The possibility of an error has been provided for.’

And this statement simply cannot be correctly used to mark off a category ofpropositions identified solely in terms of the predicate involved, independentlyof the question how one comes to know that the predicate is instantiated.

Secondly, one cannotmake this sort of absolute claim of immunity to errorthrough misidentification for mental self-ascriptions either — at least not self-ascriptions of the kind Wittgenstein chooses, which includes ‘I see so-and-so’and ‘I hear so-and-so’. Consider a case in which I have reason to believe that myfactual information may be misleading; it feels as if I am touching a piece ofcloth, and my relevant visual information is restricted to seeing, in a mirror, alarge number of hands reaching out and touching nothing, and one handtouching a piece of cloth. Here it makes sense for me to say ‘Someone is feelinga piece of cloth, but is it I?’ (One cannot produce this kind of case for mentalpredicates whose self-ascription is absolutely incorrigible. But what is theinterest of this fact?)

Thirdly, the stronger notion is in any case much less interesting. It is highlyimportant that our ‘I’-Ideas are such that judgements controlled by certain waysof gaining knowledge of ourselves as physical and spatial things are immune to


error throughmisidentification: that the hearing of the relevant information on‘I’-thoughts rests upon no argument, or identification, but is simply constitu-tive of our having an ‘I’-Idea. (The fact that these ways of gaining knowledge ofourselves must enter into the informational component of a functional charac-terization of our ‘I’-Ideas— of what it is to think of oneself self-consciously—is the most powerful antidote to a Cartesian conception of the self.)

Our task now is to investigate more precisely the way we have of gainingknowledge of ourselves upon which our ‘I’-thoughts depend. In the next twosections, I shall consider the bases of physical and mental self-knowledge;although, as we shall see, they can be separated only artificially.

3. Bodily self-ascription

I shall discuss two ways we have of gaining knowledge of our physical states andproperties, both of which give rise to the phenomenon of immunity to errorthrough misidentification.

In the first place, we have what might be described as a general capacity toperceive our own bodies, although this can be broken down into severaldistinguishable capacities: our proprioceptive sense, our sense of balance, ofheat and cold, and of pressure.26 Each of these modes of perception appearsto give rise to judgements which are immune to error through misidentifica-tion. None of the following utterances appears to make sense when the firstcomponent expresses knowledge gained in the appropriate way: ‘Someone’slegs are crossed, but is it my legs that are crossed?’; ‘Someone is hot and sticky,but is it I who am hot and sticky?’; ‘Someone is being pushed, but is it I whoam being pushed?’ There just does not appear to be a gap between the subject’shaving information (or appearing to have information), in the appropriateway, that the property of being F is instantiated, and his having information(or appearing to have information) that he is F�; for him to have, or to appearto have, the information that the property is instantiated just is for it to appearto him that he is F.

26. I do not include in this list perceptual knowledge of the physical self based upon executing certaincharacteristic movements, e.g., looking down, or feeling one’s body with the sort of motions one uses inwashing oneself. But actually I am not at all persuaded that judgements so based ought not to appear onthe list, because I am not at all persuaded that they depend upon an identification component.

110 Gareth Evans

If we attempt to impose some articulation in these cases, seeing I am F asbased on b is F and I am b , then we run into considerable difficulties (of akind with which we are now familiar: see Evans 1982, §6.6). The articulationmight be recommended on the basis of the possibility of a deviant causal chain,linking the subject’s brain appropriately with someone else’s body, in such away that he is in fact registering information from that other body. As before(Evans 1982, §6.6), I claim that this possibility merely shows the possibility ofan error; it does not show that ordinary judgements of the kind in question areidentification-dependent.

In the first place, we cannot think of the kinaesthetic and proprioceptivesystem as gaining knowledge of truths about the condition of a body whichleaves the question of the identity of the body open. If the subject does notknow that he has his legs bent (say) on this basis (because he is in the situationdescribed), then he does not know anything on this basis. (To judge thatsomeone has his legs bent would be a wild shot in the dark.)27

In the second place, there are problems about the Ideas that would beinvolved in the supposed identification component. The supposed Idea b couldbe adequate only if it involved identification by description, on the lines of ‘thebody from which I hereby have information’. Such an Idea would certainly beinvolved in one’s thinking if one knew one was in the abnormal situationdescribed; but it is surely too sophisticated to be discerned as an element in thenormal case of judgements of the kind we are considering. And, to turn to theother side of the supposed identification component: if our Ideas of ourselveswere such as to leave room for such an identification component that is, if theydid not have the legitimacy of this kind of physical self-ascription, withoutneed for argument or identification, built in at the foundation — then it isquite unclear how they could ever allow for the identification of the self as aphysical thing at all.28

The second way of gaining knowledge of our physical properties has animportance in our thought about ourselves which it is difficult to exaggerate.29

27. It would not be a wild shot in the dark if the subject had been told that he was linked up appropriate-ly with someone else’s body. But then he would be in the position of knowing that the information wasnot being received in the normal way. Cf. Evans 1982, §6.6.

28. [See § 4]

29. It is a symptomof the importance of thismode of self-knowledge to our conception of ourselves thatit, or some shadow of it, is preserved in even themost metaphysical accounts of the self, in which the selfis regarded as the origin of the perceptual field, or as a point of view on the world. See Wittgenstein,Tractatus Logico-Philosophicus [1921]:5.6–5.641; and ‘Wittgenstein’s Notes for Lectures on “Private


(We have already touched on it at several points: see Evans 1982, §§6.3, 6.6.) Ihave in mind the way in which we are able to know our position, orientation,and relation to other objects in the world upon the basis of our perceptions ofthe world. Included here are such things as: knowing that one is in one’s ownbedroom by perceiving and recognizing the room and its contents; knowingthat one is moving in a train by seeing the world slide by; knowing that there isa tree in front of one, or to the right or left, by seeing it; and so on. Once again,none of the following utterances appears to make sense when the first compo-nent expresses knowledge gained in this way: ‘Someone is in my bedroom, butis it I?’; ‘Someone is moving, but is it I?’; ‘Someone is standing in front of a tree,but is it I?’

The explanation of the importance of this way of gaining knowledge is nothard to find. Any thinker who has an idea of an objective spatial world — anidea of a world of objects and phenomena which can be perceived but which arenot dependent on being perceived for their existence — must be able to thinkof his perception of the world as being simultaneously due to his position in theworld, and to the condition of the world at that position.30 The very idea of aperceivable, objective, spatial world brings with it the idea of the subject asbeing in the world, with the course of his perceptions due to his changingposition in the world and to the more or less stable way the world is. The ideathat there is an objective world and the idea that the subject is somewherecannot be separated, and where he is is given by what he can perceive. For thepurposes of an almost entirely arbitrary, though traditional, distinction betweenthe mental and the physical, we are leaving such self-ascriptions as ‘I am seeinga tree’ until the next section; but in fact they constitute an indispensable part ofthe little theory which is required for the idea of an objective spatial world. ‘Iperceive such-and-such, such-and-such holds at p�; so (probably) I am at p�’; ‘Iperceive such-and-such, I am at p, so such-and-such holds at p�’; ‘I am at p,such-and-such does not hold at p, so I can’t really be perceiving such-and-such,even though it appears that I am’; ‘I was at p a moment ago, so I can only havegot as far as p�’, so I should expect to perceive such-and-such’. These argumentsexploit principles connecting the subject’s position, the course of his percep-tions, and the speed and continuity of his movement through space; and thechild must learn to trip round and round those principles, so that he comes to

Experience” and “Sense Data”’ [1935?]: 271–320, at p. 299: “But I am in a favoured position. I am thecentre of the world.”

30. See ‘Things Without the Mind’ for more on this.

112 Gareth Evans

think effortlessly in these ways.31

In §1 I suggested that our knowledge of what it is for I am δt to be true,where δt is a fundamental identification of a person (conceived of therefore, asan element of the objective spatial order), consists in our knowledge of what itis for us to be located at a position in space. In Evans 1982, §6.3 I argued thatthis in turn can be regarded as consisting in a practical capacity to locateourselves in space by means of exactly the kinds of patterns of reasoning that Ihave just described. It is this capacity which enables us to make sense of the ideathat we ourselves are elements in the objective order; and this is what is re-quired for our thoughts about ourselves to conform to the Generality Con-straint (Evans 1982, §4.3, §1).

Now if this is right, we can see that the perception-based judgements aboutour position and our relations to other things which we are discussing must beidentification-free.32 If we try to regard them as identification-dependent, weshall run into the same difficulties as before. First, there is no knowing about theposition, orientation, etc., of some physical object in the ways in question, insuch a way that it is left an open question which object it is. And second, thereis the familiar problem about the supposed identification component. Theobject in question could be identified only by description, as the object aboutwhose position, orientation, etc., information is being obtained in the ways inquestion; such an Idea would indeed be adequate, and might be used incircumstances which were abnormal or taken to be abnormal, but hardly figuresin the normal case. And if we try to give an account of the Idea the subject hasof himself, leaving it an open question whether the object whose position andrelations can be known about in this way is the subject, it becomes quiteproblematic how the subject could ever make sense of the thought that he islocated somewhere.33

The considerations of this section tell against the common idea that ourconception of ourselves ‘from the first-person perspective’ is a conception of a

31. Do we really have to go any further than this in order to answer Strawson’s questions (Individuals,p. 93): “(1) Why are one’s states of consciousness ascribed to anything at all? And (2) Why are theyascribed to the very same thing as certain corporeal characteristics, a certain physical situation, etc.?”

32. It is difficult enough to make sense of the Cartesian position, in which this is not regarded as a wayof gaining knowledge about oneself at all (only about somethingwhich one ‘has’). But I confess to findingmyself utterly defeated by the suggestion that while this is a way of gaining knowledge of oneself, it is onewhich exploits an identification.

33. It was precisely for this reason that I argue (Evans 1982, §6.3) that ‘here’ should not be regarded asdefined in terms of ‘I’ (as if ‘I’ had a priority). In fact ‘I’ and ‘here’ are exactly correlative: the samecapacity underlies understanding of both, namely knowledge of what it is for I am at p to be true.


thinking, feeling, and perceiving thing, and not necessarily of a physical thinglocated in space. The theoretical significance of immunity to error throughmisidentification is that it shows that evidence of certain kinds bears onthoughts involving ‘I’-Ideas directly and immediately. It is a fact about ‘I’-Ideasof objects that evidence of these kinds pertains to thoughts about those objects.(The immediate bearing of such evidence would have to be part of a functionalcharacterization of what it is to have an ‘I’-Idea.) Thus the cases of immunity toerror through misidentification that we have considered in this section revealthat our conception of ourselves is firmly anti-Cartesian: our ‘I’-Ideas are Ideasof bearers of physical no less than mental properties.

4. Mental self-ascription

My discussion of the ways in which we have knowledge of our own mentalstates will be extremely incomplete. My purpose is simply to bring to mindsome of the main features of this kind of self-knowledge, so that we can have atleast a rough idea of how it can be incorporated into our Idea of ourselves. Infact, I shall concentrate upon the ways we have of knowing what we believe andwhat we experience, for I believe that if we get these right, we shall have a goodmodel of self-knowledge (or introspection) to follow in other cases. In particu-lar, I shall quite avoid the idea of this kind of self-knowledge as a form ofperception —mysterious in being incapable of delivering inaccurate results.

Wittgenstein is reported to have said in an Oxford discussion:

If a man says to me, looking at the sky, ‘I think it is going to rain, therefore Iexist’, I do not understand him.34

The contribution is certainly gnomic; but I think Wittgenstein was trying toundermine the temptation to adopt a Cartesian position, by forcing us to lookmore closely at the nature of our knowledge of our ownmental properties, and,in particular, by forcing us to abandon the idea that it always involves an inwardglance at the states and doings of something to which only the person himselfhas access. The crucial point is the one I have italicized: in making a self-ascription of belief, one’s eyes are, so to speak, or occasionally literally, directedoutward—upon the world. If someone asksme ‘Do you think there is going to

34. ChristopherCoope, PeterGeach, TimothyPorts, andRogerWhite, eds.,A Wittgensteinian Workbook(Blackwell, Oxford, 1971): 21. (My emphasis.)

114 Gareth Evans

be a third world war?’, I must attend, in answering him, to precisely the sameoutward phenomena as I would attend to if I were answering the question ‘Willthere be a third world war?’ I get myself in a position to answer the questionwhether I believe that p by putting into operation whatever procedure I have foranswering the question whether p. (There is no question of my applying aprocedure for determining beliefs to something, and hence no question of mypossibly applying the procedure to the wrong thing.) If a judging subject appliesthis procedure, then necessarily he will gain knowledge of one of his ownmental states: even themost determined sceptic cannot find here a gap in whichto insert his knife.

We can encapsulate this procedure for answering questions about what onebelieves in the following simple rule: whenever you are in a position to assertthat p, you are ipso facto in a position to assert ‘I believe that p�’. But it seemspretty clear that mastery of this procedure cannot constitute a full understand-ing of the content of the judgement ‘I believe that p�’.35 Understanding of thecontent of the judgement must involve possession of the psychological conceptexpressed by ‘ξ believes that p�’, which the subject must conceive as capable ofbeing instantiated otherwise than by himself.36 Involvement of this concept inthe judgement would bemanifested by an appreciation of the fact that the kindsof evidence which he is prepared to recognize as relevant to the ascription of thepredicate to others bear also upon the truth of his claim, and a willingness torecognize, as relevant to the ascription of the predicate to others, evidence oftheir having executed the same procedure—making a judgement as to whetherp — which underlies his own self-ascription. Without this background, wemight say, we secure no genuine ‘I think’ (‘think that p�’) to accompany histhought (‘p�’): the ‘I think’ which accompanies all his thoughts is purely formal.But adding the backgroundmakes no difference to themethod of self-ascription:in particular, we continue to have no need for the idea of the inward glance.

35. ‘I believe that p�’ admits of a distinction between internal and external negation. ‘It is not the casethat I believe that p�’ can be the of an openmind. This enables us to express one side of the central notionof objectivity — the idea that truth transcends my knowledge or belief: it may be that p, it may be thatnot-p�; I do not know, and have to be withdrawn in the circumstance that it is not the case that p. Sucha state of affairs — expressible in the past tense by ‘I believed that p, but it was not the case that p�’ —cannot be believed to obtain currently, but its possibility makes sense. This is the other side of the ideaof objectivity: although I believe that p, it may be the case that not-p. In short, learning the differencebetween ‘I believe that p�’ and ‘p�’ involves learning the different ways in which the two sentences embedunder various operators: crucially negation, modality, and the past tense.

36. And by himself at other times.


The self-ascription of perceptual experiences follows a different model, onewe can understand only if we begin by considering the ordinary situation inwhich a subject is perceiving the world and making judgements about it. Ingeneral, we may regard a perceptual experience as an informational state of thesubject: it has a certain content —the world is represented a certain way— andhence it permits of a non-derivative classification as true or false. For an internalstate to be so regarded, it must have appropriate connections with behaviour—it must have a certain motive force upon the actions or the subject. This motiveforce can be countermanded, in the case of more sophisticated organisms(concept-exercising and reasoning organisms), by judgements based upon otherconsiderations. In the case of such organisms, the internal states which have acontent by virtue of their phylogenetically more ancient connections with themotor system also serve as input to the concept-exercising and reasoningsystem. Judgements are then based upon (reliably caused by) these internalstates; when this is the case we can speak of the information being ‘accessible’ tothe subject, and, indeed, of the existence of conscious experience.

The informational states which a subject acquires through perception arenon-conceptual, or non-conceptualized. Judgements based upon such statesnecessarily involve conceptualization: in moving from a perceptual experienceto a judgement about the world (usually expressible in some verbal form), onewill be exercising basic conceptual skills.37 But this formulation (in terms ofmoving from an experience to a judgement) must not be allowed to obscure thegeneral picture. Although the subject’s judgements are based upon his experi-ence (i.e. upon the unconceptualized information available to him), hisjudgements are not about the informational state. The process of conceptualiza-tion or judgement takes the subject from his being in one kind of informationalstate (with a content of a certain kind, namely, non-conceptual content) to hisbeing in another kind of cognitive state (with a content of a different kind,namely, conceptual content). So when the subject wishes to make absolutelysure that his judgement is correct, he gazes again at the world (thereby produc-ing, or reproducing, an informational state in himself), he does not in any sensegaze at, or concentrate upon, his internal state. His internal state cannot in anysense become an object to him. (He is in it.)

However, a subject can gain knowledge of his internal informational statesin a very simple way: by re-using precisely those skills of conceptualization that

37. For the difference between non-conceptual and conceptual states, cf. Evans 1982, §§5.2, 6.3.

116 Gareth Evans

he uses to make judgements about the world. Here is how he can do it. He goesthrough exactly the same procedure as he would go through if he were trying tomake a judgement about how it is at this place now, but excluding any knowl-edge he has of an extraneous kind. (That is, he seeks to determine what hewould judge if he did not have such extraneous information.)38 The result willnecessarily be closely correlated with the content of the informational statewhich he is in at that time. Now he may prefix this result with the operator ‘Itseems to me as though …’. This is a way of producing in himself, and givingexpression to, a cognitive state whose content is systematically dependent uponthe content of the informational state, and the systematic dependence is a basisfor him to claim knowledge of the informational state. But in no sense has thatstate become an object to him: there is nothing that constitutes ‘perceiving thatstate’. What this means is that there is no informational state which stands to theinternal state as that internal state stands to the state of the world.

Once again, describing this procedure cannot constitute a complete accountof what it is to have this capacity for self-knowledge. The subject who genuinelyhas this capacity for self-knowledge must understand the content of hisjudgement ‘It seems to me as though p�’, and his understanding of it mustdetermine it to have a content different from that of the judgement ‘Possibly p�’,or ‘Going by appearances, p�’. This requires a background of a sort analogous tothat mentioned above in connection with self-ascription of belief. As myallusion in that context will have suggested, I believe we may have here aninterpretation ofKant’s remark about the transcendental ‘I think’which accompa-nies all our perceptions.39 Without the background, we have at most a formal ‘Ithink’; it yields nothing until embedded within a satisfactory theory.40

The procedure I have described does not produce infallible knowledge ofthe informational state, for mistakes of the kind that occur when the subjectmakes judgements about the world can also produce inaccuracies when thesame procedure is reused for this different purpose. For example, consider acase in which a subject sees ten points of light arranged in a circle, but reportsthat there are eleven points of light arranged in a circle, because he has made a

38. For ‘extraneous’, see Dummett, ‘What is a Theory of Meaning?’ (II): 95. Obviously the subject canengage in this procedure when he believes his perceptual information is illusory; the rule then obligeshim to pretend or suppose that it is not, and ask himself what he would say or judge in that case.

39. Critique of Pure Reason, B 131–2.

40. The point is that ‘I think’ (or ‘it seems to me’) acquires structure (‘ξ thinks’ or ‘it seems to ξ�’, with‘I’ in the argument-place) only when it is related to (at least possible) other exemplifications of the samepredicate.


mistake in counting, forgetting where he began. Such a mistake can clearlyoccur again when the subject re-uses the procedure in order to gain knowledgeof his internal state: his report ‘I seem to see eleven points of light arranged ina circle’ is just wrong. However, when the subject conceptualizes his experiencein terms of some very elementary concept, such as a simple colour concept like‘red’, it is not easy to make sense of his making a mistake. Concentrating uponthis kind of case, and feeling extremely suspicious of the idea of a judgementwhich is about something distinct from itself, yet which cannot be wrong, somephilosophers have adopted the contention that the existence of an internal infor-mational state is constituted by the subject’s disposition to make certain judge-ments. But this is both extremely implausible in itself and quite unnecessary.

The proposal is implausible, because it is not the case that we simply findourselves with a yen to apply some concept — a conviction that it has applica-tion in the immediate vicinity. Nothing could more falsify the facts of thesituation. Further, no account of what it is to be in a non-conceptual informa-tional state can be given in terms of dispositions to exercise concepts unlessthose concepts are assumed to be endlessly fine-grained; and does this makesense? Do we really understand the proposal that we have as many colourconcepts as there are shades of colour that we can sensibly discriminate?

The proposal is unnecessary for the following reason. Logically infallibleknowledge does indicate that the state judged about and the judgement are not(as Hume would have said) distinct existences. But there are two ways in whichthis can be acknowledged. Either the state can be regarded as constituted bydispositions to make certain judgements — and this, as we have seen, is notvery plausible; or, alternatively, the judgement’s being a judgement with acertain content can be regarded as constituted by its being a response to thatstate. And on reflection, the latter is far the more plausible option: suchinfallibility as there is arises because we regard it as a necessary condition for thesubject to possess these simple observational concepts that he be disposed toapply them when he has certain experiences. This sort of infallibility is ratherlimited and uninteresting. And it is of a quite different kind from that whicharises in the case of the self-ascription of belief.

However, in an important respect these two ways of gaining self-knowledgeare similar: namely that neither conforms to the description ‘looking within’. Inthe case we have just been describing, the subject’s concentration, as with self-ascription of belief, is on the outside world: how does he, or would he, judge itto be? The cases are different in that in this case there is something (namely aninternal, informational state of the subject), distinct from his judgement, to

118 Gareth Evans

which his judgement aims to be faithful. But it is something necessarily ap-proached in the roundabout way I have described.

We have taken account, then, of the following two facts. First, in a state ofinformation on the basis of which a subject may ascribe to himself an experi-ence as of seeing, say, a tree, what he observes (if anything) is only the tree, nothis own informational state. (But let me remind you that the procedure I havedescribed, of re-using the conceptual skills which one uses in order to makejudgements about the world, is not by itself enough for the capacity to ascribeexperiences to oneself.) Second, any informational state in which the subjecthas information about the world is ipso facto a state in which he has informa-tion about himself; of the kind we are discussing, available to him. It is of theutmost importance to appreciate that in order to understand the self-ascrip-tion of experience we need to postulate no special faculty of inner sense orinternal self-scanning.41

Not all our reports of experience have the character we have been consider-ing, because not all our characterizations of internal informational statesdescribe them in terms of their content. Although ‘It feels to me as thoughthousands of little pins are lightly touching my skin’ and ‘It feels to me asthough my legs are crossed’ are reports of the same kind as those we have beenconsidering, ‘I feel a pain in my foot’ and ‘I feel an itch in my foot’ are not. Butonce we have the general framework of bodily perception and bodily sensations(understood as informational states, which may be illusory, whose contentconcerns the body and its states and positions), then it is perhaps not toodifficult to fit these in. There is no reason why an informational state shouldhave only informational properties. Thus we can say that when a subject feels anitch, he perceives (or appears to perceive) a part of his body in a way whichmakes him very much want to scratch, and when he is in pain, he perceives (orappears to perceive) a part of his body in a way which is awful. (One thing thatcan be said for this approach is that it does explain why pains and itches arenecessarily felt in a particular part of the body.)

The features of these modes of self-knowledge have given rise to certainillusions about the self. Hume said:

For my part, when I enter most intimately into what I call myself, I alwaysstumble on some particular perception or other, of heat or cold, light or shade,

41. There is a faculty of internal self-scanning: namely bodily perception (see §3).


love or hatred, pain or pleasure. I can never catch myself at any time without aperception, and never can observe anything but the perception.42

Now in fact we have totally rejected the background of perceptual metaphor inwhich Hume casts his point, and in particular we have invoked nothing thatcould be construed as stumbling on perceptions; those inner states of thesubject that we spoke of cannot intelligibly be regarded as objects of his internalgaze. However, there is something in Hume’s point; indeed, in a manner ofspeaking, it becomes even stronger when the metaphor is dispensed with. Forwhat we are aware of, when we know that we see a tree, is nothing but a tree. Infact, we only have to be aware of some state of the world in order to be in aposition to make an assertion about ourselves.

Now this might raise the following perplexity. How can it be that we canhave knowledge of a state of affairs which involves a substantial and persistingself, simply by being aware of (still worse, by merely appearing to be aware of)a state of the world?43

We must agree that we cannot get something for nothing. So the anxietywill be lessened only by showing how the accounting is done—where the ideaof the persisting empirical self comes from. Nothing more than the originalstate of awareness — awareness, simply, of a tree — is called for on the side ofawareness, for a subject to gain knowledge of himself thereby. But certainlysomethingmore than the sheer awareness is called for: the perceptual state mustoccur in the context of certain kinds of knowledge and understanding on thepart of the subject. (Otherwise, we might say as before, the ‘I think’ whichaccompanies the subject’s perceptions is purely formal, or empty.) No judge-ment will have the content of a psychological self-ascription, unless the judgercan he regarded as ascribing to himself a property which he can conceive asbeing satisfied by a being not necessarily himself — a state of affairs which hewill have to conceive as involving a persisting subject of experience. He canknow that a state of affairs of the relevant type obtains simply by being aware ofa tree, but he must conceive the state of affairs that he then knows to obtain asa state of affairs of precisely that type. And this means that he must conceive of

42. A Treatise of Human Nature, Book I, Part IV, chapter VI: 239 in the Everyman edition (Dent,London. 1911).

43. See G. E.Moore. ‘Wittgenstein’s Lectures in 1930–33’ (III), Mind lxiv (1955): 1–27, at p. 13: “… andhe said of what he called ‘visual sensations’ generally, and in particular of what he called ‘the visual field,that “the idea of a person doesn’t enter into the description of it, just as a [physical] eye doesn’t enterinto the description of what is seen.” (See n. 25 above.)

120 Gareth Evans

himself, the subject to whom the property is ascribed, as a being of the kindwhich he envisages when he simply envisages someone seeing a tree— that is tosay, a persisting subject of experience, located in space and time.44

We see, then, that the two applications of the Generality Constraint— oneimposing the requirement of generality upon the concept — ξ sees a tree, andthe other imposing a generality upon the Idea ‘I’ — work together, forcing thesubject to think of himself as an element of the objective order.

There is a parallel for this situation. For we want to allow, equally, that asubject can know that he is in front of a house simply by perceiving a house.Certainly what he perceives comprises no element corresponding to ‘I’ in thejudgement ‘I am in front of a house’: he is simply aware of a house. But if we areto interpret a judgement made upon this basis as having the content ‘I am infront of a house’, we must have reason to suppose that the subject regardshimself as recognizing the existence of a state of affairs of precisely the samekind as obtains when, for instance, a car is in front of a house. So what heenvisages, or judges, certainly comprises two elements spatially related, al-though what he sees does not. (This only goes to show that it is not a good idea,in attempting to determine the content of a person’s judgement, to examinenothing but the content of the perceptions which can legitimately give rise to it.)

Presumably it goes without saying that both the ways of gaining knowledgeof ourselves that I have discussed in this section give rise to judgements whichare immune to error through misidentification. When the first componentexpresses knowledge gained in one of these ways, it does not make sense for thesubject to utter ‘Someone believes that p, but is it I who believe that p?’, or ‘Itseems to someone that there’s something red in front of him, but does it seemto me that there is something red in front of me?’

We have reached, I hope, a point upon which everyone can agree: someonewho understands a term as referring to himself must be disposed to regard, asrelevant to the truth or falsity of certain utterances involving that term, theoccurrence of certain experiences which he is in a position immediately torecognize. It would appear that nothing is easier than to test for the existenceof this disposition: we can stimulate the subject in various ways, and see howhis evaluation of the relevant sentences is affected. I have tried to explain whythe sheer existence of this disposition does not by itself guarantee that the

44. For the fact that certain judgements about the self can be identification-free, combined with theclaim that links with empirical criteria of personal identity are nevertheless not severed in the case ofsuch judgements, see Strawson, The Bounds of Sense, p. 165.


subject has an adequate Idea of himself, but it is clearly an indispensableelement in any such Idea.

However, it needs to be treated with the greatest care, formis-statements andmisunderstandings of this element provide a strong pressure towards solipsism.

A subject’s knowledge of what is in question when, for example, he is inpain, or when he sees a tree, can seem to be very similar to his knowledge ofwhat is in question when an observable state of the world obtains. When aneclipse of the sun occurs, he might say, such-and-such experiences are to beexpected; and he knows which they are — he can manifest this knowledge byhis performance in suitable tests. Equally, he might say, when he is in pain,such-and-such experiences are to be expected; and he knows which ones theyare. There is certainly a difference between his being in pain and anyone else’sbeing in pain; and this difference is one which he can detect, just as he candetect the difference between an eclipse of the sun and an eclipse of the moon.

Now we might think of the solipsist as someone who attempts to use thisasymmetry in order to state, at least to his own satisfaction, what he means by‘I’. Rather as one might differentiate the sun from the moon by saying that thesun is that object an eclipse of which makes it reasonable to expect such-and-such experiences, whereas the moon is that object an eclipse of which makes itreasonable to expect such-and-such different experiences, the solipsist thinkshe can say: I am that object such that when it is in pain something frightful isto be expected.

Consider a pair of different observable possibilities: say a certain armchairbeing red and it being green. A suitably equipped subject can tell the differencebetween these states of affairs, because of the difference in the way they appear.And this is how the solipsist construes the asymmetry from which he starts. Hesupposes that he can adopt the same impersonal style of description in charac-terizing the asymmetry he is concerned with — speaking of what kind ofexperiences are to be expected. Now we can say that a green armchair and a redarmchair can be distinguished by a difference in the way they appear (by adifference in the experiences which are to be expected), because we are speakingof what may be expected by any observer (or at least any normal observer). Butthat is exactly not the kind of difference with which the solipsist is concerned.What he can legitimately say is only this: ‘I can tell the difference between astate of affairs involving myself and a state of affairs involving someone else bythe difference in the ways in which they appear to me’�; thus he can say ‘WhenI am in pain, something frightful happens to me�’. Of course, no one in his rightmind would want to say that, because it is tautological. And the solipsist’s error

122 Gareth Evans

lies precisely in his desire to extract something informative (non-tautological)about his use of ‘I’ from the asymmetry.

He is of course right that there is something substantial which must havebeen learned by someone who uses the first-person pronoun properly, and thatit is manifested in his responding differently to the kind of differences fromwhich the solipsist begins. But this knowledge ispractical�: it cannot be transformedinto a substantial account of what the subject means by ‘I’. It is the solipsist’sattempt to transform it into something substantial that yields his desire to say“by ‘I’ I mean the person such that, when he is in pain, something frightful is tobe expected”.45 And the project of wringing something substantial out of thedifference is obviously ruined when the impersonal mode of description isreplaced by the appropriate relativized description, yielding a tautology.

5. Memory

Memory is not a way of gaining knowledge but a faculty of retaining knowledge;so this next section is not parallel to the last two. As I remarked earlier (§1 ), therelation of memory to ‘I’-thinking introduces quite novel elements, over andabove the sort of feature involved in ‘here’-Ideas and ‘this’-Ideas.

Let me make a preliminary identification of the point I have in mind, tobegin fixing ideas, by reference to language. (This formulation will not in factstandmuch scrutiny.)Wemight say: if a subject remembers, at time t¢, being ina position at time I to assert ‘I am F�’, then he is in a position, without furtherinformation, to assert ‘I was F (at t�)’. (Hence if he is in a position at t¢ to assert‘I am now G�’, he is in a position to assert ‘Something was F (at t�) and is nowG�’.) There is no such simple rule relating memory to ‘here’, ‘this’, or ‘now’.

It is better to put the point in terms of how a person’s belief system isorganized to take account of the passage of time: a subject which David Kaplanhas called ‘cognitive dynamics’.46 (This allows us to eliminate any appear-ance that the subject arrives at the past-tense judgement by some inference,the premiss of which is that he was in a position to make such-and-such a

45. Incidentally, although it is sentience that the solipsist denies to others, it is not at all essential that hisdefinition of ‘I’ be restricted to the attempt to extract something substantial from our capacities formental self-ascription. Hemight just as well have said: I am the person such that, when his legs are bent,this is felt: or: I am the person such that, when he moves, changes of such-and-such a kind are to beobserved. The solipsist need not be a dualist.

46. In ‘Demonstratives’, Kaplan 1989b.


judgement; which would require us to raise the question of the structure andnature of the premiss.)

It is a precondition of rationality that information acquired at one timeshould be available to the subject later: hence, given that in a rational creatureinformation (and misinformation) generates beliefs, that beliefs should persist.(And the capacity to retain information is memory.) If we take a belief state, asI think we should, to be a disposition to have certain thoughts or to makecertain judgements, then we can say that any rational being must have acognitive system which brings it about that the dispositions to make judge-ments he has at one time should be systematically dependent on the disposi-tions to make judgements he had at earlier times. (There being, presumably, asingle persisting structural feature of the nervous system underlying both setsof dispositions.)

Persistence of belief does not always, or indeed even usually, involve persis-tence of a disposition to make the same judgements (if judgements are individ-uated simply in terms of the forms of words which would express them). Forinstance, the persistence of a belief that I would have manifested at some timeby the judgement ‘John is now angry’ involves the disposition to judge later notthat John is now angry but that John was then angry.47

There is indeed no general guarantee that a belief will persist. Suppose Iperceive an object and judge ‘This is F�’, and then lose track of it. There is noguarantee that there will be available to me a past-tense demonstrative Idea (seeEvans 1982, §5.5), enabling me to judge ‘That was F�’. (If such an Idea isavailable to me, it will need to draw on conceptual material not present in theoriginal demonstrative identification.)

When a subject keeps track of a place as he moves (or does not move), orkeeps track of an object as he or it moves (or not), I think we should regard theslightly varying forms of the judgements he is disposed to make as manifesta-tions of a single persisting belief (a continuing acceptance of the same thought).Success here certainly depends upon a skill (the ability to keep track). Butthinking a thought inevitably takes time, and this kind of skill must be seen asgenerally underlying demonstrative judgements. I cannot see the later membersof a series of judgements ‘It’s Φ here’, ‘It’s Φ there’, …, made while one movesabout, keeping track of the place at which one has ascertained that it is Φ, asbased upon an identification. Similarly with times, and series of judgements like

47. Why call it the persistence of the same belief? Because no new evidence is needed to sustain it.

124 Gareth Evans

‘It is Φ now’, ‘It was Φ a moment ago’, ‘It was Φ a while back’,…48

Now, if we consider ‘I’ in this context, we see that the cognitive dynamics of‘I’-Ideas are peculiarly simple. We can isolate the cognitive dynamics of ‘I’-Ideasin particular by considering tenseless predicates: if a subject has, at time t, abelief which he might manifest in judging ‘I am F�’ (where ‘ξ is F�’ is tenseless),then the tendency of belief to persist means that there is a non-negligibleprobability (depending on the gap between t and t¢�) of his still having, at a latertime t¢, the disposition to judge ‘I am F�’. If we now introduce the considerationof tenses (etc.), we introduce other aspects of cognitive dynamics: combinedwith the aspect we have hitherto isolated, this yields (for instance): if a subjecthas at t a belief which he might then manifest in judging ‘I am now F�’, thenthere is a non-negligible probability of his having, at a later time t¢, a dispositionto judge ‘I was previously F�’. The later manifestation of the belief still employs‘I’ (contrast the need to shift from ‘here’ to ‘there’ as one moves but keeps trackof a place, or from ‘now’ to ‘then’ as one keeps track of a time receding into thepast); and, so far as the ‘I’-Idea is concerned, the later dispositions to judge flowout of the earlier dispositions to judge, without the need for any skill or care(not to lose track of something) on the part of the subject.

We can put this point by using, once again, the terminology of immunity toerror through misidentification: a past-tense judgement, ‘I was F�’, is not basedupon a pair of propositions, That person was F and I am that person (where‘that person’ captures an identification of an object with respect to a pasttime).49

It would be a mistake to think that this holds only when the disposition tomake the past-tense judgement ‘I was F�’ manifests the persistence of a belieforiginally acquired in one of the special ways of gaining knowledge of ourselvesdiscussed in §3 and §4. For even when there is an articulation underlying theformation of the original belief, it remains the case that the unthinking opera-tion of cognitive dynamics will yield a subsequent disposition tomake a suitablepast-tense judgement, without there being any need for an identificationcomponent in the process whereby the dispositions to make judgements flowinto one another.We need, therefore, a more sophisticated notion of immunity

48. [See Evans 1982, §1 of the Appendix to Chapter 6.]

49. An identification of an object (an Idea, a) is an identification with respect to the present if it followsfrom the truth of a is F that it is now the case that ($x�)(x is F�), no matter what is substituted for ‘F’. Anidentification is an identification with respect to the past if there is no such general implication. (Ofcourse ‘That man is now F�’ has the implication even if ‘That man’ is a past-tense demonstrative; but thesame identification can figure in judgements that do not have the implication, e.g. ‘That man was F’.)


to error through misidentification than before: one which can highlight theabsence of an identification of a certain kind, or at a certain point in the processthat issues in a judgement, rather than the absence of any identification at all.

This feature of ‘I’-Ideas means that they span past and present in a novelway.50 (A ‘this’-Idea will typically span a period of time: Evans 1982, §§6.1,6.4. But its doing so rests upon the exercise of a skill on the subject’s part.Moreover, the time span of an ‘I’-Idea may be quite extensive.) An ‘I’-Idea givesrise to thoughts dependent upon information received over a period of time.(This opens up a new possibility of ill-groundedness: see Evans 1982, §5.4.Whether or not a subject essaying the employment in thought of such an Ideais in fact thinking of anything depends upon whether there is just one thingwhich is both what the relevant previous beliefs, now retained, were about andwhat his current self-oriented ways of gaining knowledge concern. (See §7).51

It is possible to regard this feature of ‘I’-Ideas as part of the informationalcomponent of a functional characterization. A possessor of an ‘I’-Idea has acapacity to ascribe past-tense properties to himself on a special basis: namely thememory of the basis appropriate for an earlier present-tense judgement. Thisbrings out a similarity with the informational components we have alreadyencountered. But we must be careful not to lose sight of the fact that memoryis not a way of gaining knowledge.

So far in this section we have been exclusively concerned with the retentionof belief. Thus we might consider a subject who forms the belief that a tree isburning, on the basis of perceptual information he is receiving. We have seenhow, by perceiving the world, the subject is ipso facto put in a position to gainvarious pieces of knowledge about himself: being in a position to assert ‘A treeis now burning’, he would also be in a position to assert ‘I am facing a tree thatis burning’, ‘I see a tree that is burning’, more weakly ‘I seem to see a tree thatis burning’, etc.52 Suppose he forms self-conscious beliefs of this latter kind;then, just as the belief initially expressible by ‘A tree is now burning’ may persist

50. And expectation will bring in the future. (See §1.)

51. If there is really no identification involved in the past-tense judgements we are considering, then insuch a judgement, ‘I was F�’, we cannot think of the ‘I’-Idea as effecting a purely present-tense identifica-tion of its object (so that in the event of a serious mismatch of the subject’s apparent memories with hisown past, we could say that he has, perfectly determinately, a particular object in view, but has a whollyinaccurate conception of its past). The identification is not exclusively present-tense: while it is true that‘I was F�’ entails ‘There is now something which was F�’, it is no less true that ‘I am F�’ entails ‘There wassomething which is now F’.

52. See §3, §4.

126 Gareth Evans

to a later time, at which it needs to be expressed by, say, ‘A tree was burning lastnight’, so the self-conscious beliefs formed on the basis of the same originalinformational state may persist, needing to be expressed as ‘I was facing a treethat was burning’, etc.

We should take note, however, of the possibility of cases in which theoperation of memory takes place purely at the level of the informationalsystem. In this sort of case, what memory ensures is the subject’s possession ofa non-conceptual informational state, whose content corresponds in a certainrespect with that of some earlier informational state of the subject (a perceptu-al state); although its content differs from that of the antecedent perceptualstate in that, if the subject is in the memory state, it seems to him that such-and-such was the case. (That is, memory states, even of this kind, are not free-floating images whose reference to the past is read into them by reasoning onthe part of the subject).53

Now just as the non-conceptual informational states involved in perceptionput a subject in a position to acquire present-tense self-knowledge by theexercise of his conceptual capacities, so these non-conceptual informationalstates put a subject in a position to acquire past-tense self-knowledge by theexercise of his conceptual capacities. A subject can form beliefs which he wouldexpress by ‘I was facing a tree that was burning last night’, etc., in this way, onthe basis of a non-conceptual memory state, without needing to have had thedisposition to make the corresponding present-tense judgements (withouthaving had the beliefs) at the time of the original perception.

This sort of memory is extremely important. It is frequently said thatmemory provides us, in the first instance, with information about our pastexperiences; but this is certainly quite wrong about the kind of operation ofmemory that I have just described: we no more have, in memory, informationwhich is primarily about our past experiences than we have, in perception,information which is primarily about our present experiences. Just as percep-tionmust be regarded as a capacity for gaining information about the world, somemory must be regarded as a capacity for retaining information about theworld. The truism about perception is not upset by the fact that there areoccasions on which we are in a perceptual state without gaining information

53. How there can be such informational states, with a content which concerns the past and frequentlymore or less specific past times, and how the behaviour of the subject must be dependent upon suchstates for such a content to be ascribed to them these are different questions, which it is fortunately notnecessary, for my current purposes, to answer.


about the world — for example, when we hallucinate. In the same way, theparallel truism about memory should not be upset by the fact that there areoccasions on which we use our memories without being in a position to makeany knowledgeable past-tense judgements about the world — as when weremember a hallucination. In the first case we gain misinformation about theworld; in the second case we retain misinformation about the world. (Ofcourse, when the information we retain was originally derived from our ownbodies, then the memory state can be said to be primarily about ourselves: forexample, I may remember in this way that I was pushed and pulled last night.But obviously this is a special case.)54

This kind of operation of memory — in which a self-conscious past-tensejudgement is based upon a non-conceptual memory state involving informa-tion about some past state of the world, in a way analogous to that in which aself-conscious present-tense judgement might have been based on the non-conceptual perceptual state whose informational content has been retained bythe memory (even though no self-conscious present-tense judgement was sobased) — seems to exemplify the phenomenon of immunity to error throughmisidentification. When the first component expresses knowledge which thesubject has gained (and does not suppose he has not gained, or may not havegained) in this way, it does not appear to make sense for him to say ‘Someonesaw a tree burning last night, but was it I?’, or ‘Someonewas in front of a tree lastnight, but was it I?’, or ‘Someone was pushed and pulled last night, but was it I?’

Shoemaker, however, has recently argued that this is not so: the appearancethat judgements about oneself based on memory are identification-free isclaimed to arise from the trivial linguistic fact that we would not describe aperson whose information about the past was not originally acquired by himselfas remembering.55

But it is not true that the apparent identification-freedom in question is amere appearance, wholly due to this linguistic phenomenon. It is true thatStrawson rather lays himself open to the accusation of trading on the linguisticphenomenon, when he tries to illustrate the identification-freedom in questionby considering the deviance of the utterance ‘I distinctly remember that inner

54. Remembering an episode in this way is frequently described as remembering it ‘from the inside’.One remembers an episode involving oneself ‘from the inside’ if one retains information of a charactersuch that if one were to possess its ‘present tense’ counterpart now, one would thereby be enabled tomake the various first-person judgements one’s capacity to make which at the time was constitutive ofone’s involvement in the episode.

55. See ‘Persons and Their Pasts’.

128 Gareth Evans

experience occurring, but did it occur to me?’56 But the claim can be made, asI made it above, without explicitly using the notion of remembering in the firstcomponent (‘Someone was F’, not ‘I remember that someone was F’), andhence without any necessary reliance on linguistic restrictions imposed by theconcept of remembering.

Shoemaker’s argument that memory judgements about oneself are identifi-cation-dependent, and hence not immune to error through misidentification,is this. We can imagine a case in which a subject’s apparent memories arecausally derived, not from past informational states of his own, but fromsomeone else’s past informational states. In such a situation the subject wouldhave information about past states of the world, and, carried upon the back ofthat information, information about past states of some observer of the world;but he would not have information about past states of his own, but ratherinformation about the past states of the person from whom his apparentmemories causally derive.57 Such a situation would arise if a perfect duplicateof a person (including his brain) were made: in this case the new person wouldappear to remember all the events which the original could remember — andappear, moreover, to remember them ‘from the inside’.

We could not legitimately speak of the subject’s remembering a treeburning, or of his remembering seeing a tree burning, or his remembering beingpummeled; this is because of the restriction built into the term ‘remember’, onwhichwehave just been commenting. So Shoemaker introduces the term ‘quasi-remembering’, explained in such a way that we can say that the subject quasi-remembers the tree burning and quasi-remembers seeing the tree burn.

Given that quasi-remembering (q-remembering) is a possible situation, itwould appear that a subject might grasp that it is a possible situation, and evenbelieve, perhaps for good reason, that he is in it. Such a subject would seem tobe able to utter, perfectly significantly, ‘Someone stood in front of a burningtree, but it was not I’ — even when the first component expresses informationwhich he has in his memory.58 And a second subject might genuinely havereason to doubt whether this was his situation, so that he might say ‘Someonewatched a tree burning, but was it I?’ being unsure, for example, how long agothe event occurred, and so whether it occurred after the time at which he started

56. The Bounds of Sense, p. 165.

57. Similarly for the ‘retention’ of beliefs (the operation of memory considered at the beginning of thissection).

58. Here ‘memory’ includes q-remembering.


his independent existence.This argument is of a kind with which we are already familiar, and we have

seen that it fails to establish its conclusion. (See Evans 1982, §6.6 on places, and§3 on physical self-ascription.) The argument presents us with a case in whichinformation apparently possessed about an object in a certain way — thenormal way — can underlie a judgement about an object which is based uponan identification. But it certainly does not follow from this that judgementsabout an object, based upon this way of possessing information about it, mustbe based upon an identification. And it seems to me that we cannot regard theordinary memory judgements which we make about ourselves as articulableinto the two components, ‘That man was in front of a burning tree’ and ‘I amthat man’. The argument against the present suggestion parallels the argumentI offered against a similar suggestion in the case of ‘here’ (Evans 1982, §6.6).

In that case, I argued that it is not intelligible to challenge the directrelevance to ‘here’-judgements of perceptual information acquired in thenormal way — on pain of not having an adequate ‘here’-Idea of a place, or atleast not one which is recognizably like the one we have, to sustain the chal-lenge. For I argued that what enables us to credit a thinker with an adequateknowledge of which place is in question in his ‘here’-judgements is his capacityto identify his whereabouts in the spatial framework. This depends upon hiscapacity to use his perceptions, and the changing course of his perceptions, todetermine his position; and this in turn depends upon a willingness to allowcurrent (normal) perceptions to bear upon the question how things stand here(where he is).

There is a similar difficulty in the present case about the nature of the‘I’-Idea that would figure in the supposed identification component. I havesuggested (see §3) that in general a subject’s possession of an adequate Idea ofhimself depends upon the same capacity that underlies his ability to use‘here’-Ideas— the capacity to determine his position in the objective order. Butthis capacity depends no more upon current perception than upon a subject’swillingness to use his memory, in order to bring information about the courseof his past (though usually recent) perceptions to bear upon the question of hispast position, and thereby upon the question of his current position. Self-location cannot in general be a momentary thing. For one thing, places are notalways immediately recognizable. For another, self-location crucially dependsupon the axiom that the subject moves continuously through space; and thataxiom can be brought to bear upon particular questions of location only if thesubject has the capacity to retain information about his previous perceptions,

130 Gareth Evans

and to use that information in making judgements about his past, and therebyhis present, position. If we imagine a subject who cannot retain information formore than a few seconds, or, equally, a subject who refuses to argue from suchretained information as he possesses to propositions about his past position,then we have imagined a subject who just does not have the practical capacityto locate himself in space.

Furthermore, there is an equal difficulty with the other term of the identity.We are supposed to be able to regard the normal situation in which a subjectjudges that he was F as being the result of two judgements, one of them beingthat he is the same person as …— but as who? We must see the judgement ‘Iwas seeing a tree burn’ as resting on the judgements ‘That man was seeing a treeburn’ and ‘I am that man’— but now what Idea of a man is represented by ‘thatman’?We know that the sheer existence of an information-link does not sufficefor an adequate Idea. As before, we might consider an Idea that is adequate byvirtue of involving a description: that man is identified as the man whose pastinformational states are causally responsible for these apparent memories. Nowit is perfectly all right to ascribe such an adequate Idea to a person who has beenapprised of his situation (or who has anyway been brought, perhaps falsely, tothink of his situation) as being that of a duplicate. But it is surely far-fetched inthe extreme to suppose that such an Idea is generally involved in our past-tenseself-ascriptions.

There is another reason for being dissatisfied with the picture whichemerges when we think of judgements about one’s past, based upon memory,as identification-dependent. The picture requires us to be able to think ofmemory as a way of having knowledge of an object which leaves its identity (inparticular, its identity with the subject) an open question. It will be agreed thatthe normal subject will have knowledge of his own past, but this will be seen as theresult of two pieces of knowledge, expressible as ‘Thatmanwas F�’ and ‘I am thatman’; and it must be the case that the subject could have knowledge of the truthof the first component, independently of the truth of the second component. Butit does not appear to me to be possible to think of memory as an ‘identity-neutral’ way of having knowledge of the past states of a person. (And this is notthe result of some trivial linguistic truth about what we would call ‘memory’.)

Suppose we surgically ‘transfer the memories’ from the brain of subject S¢to the brain of a subject S, and suppose S does not know that this has happened.S will, of course, make judgements about his past in the normal way. Butsuppose that he discovers that he was not F, and he was not G, … — that ingeneral his memory cannot be relied upon as an accurate record of his past.


Suppose that, fantastically, he then retreats to making general past-tensejudgements: ‘Someone was F, and was G …’. These judgements could not possiblyconstitute knowledge. Even to be intelligible in putting them forward, S wouldhave to offer what had actually happened, or something very like it, as ahypothesis; but he could not possibly be said to know that it was true. It wouldbe a sheer guess. Consequently he could not be said to know anything based onit. (We must remember that it is not sufficient for knowledge that a true beliefbe causally dependent on the facts which render it true.) And if S cannot knowthe truth of these general judgements, he certainly cannot be said to know thetruth of the supposed judgements ‘That man was F�’, etc., which are strongerthan they are.59

So the theory does not fit the facts. Memory is not a way of possessingknowledge about an object of a kind which leaves open the question of theidentity of that object. If a subject has, in virtue of the operations of his memo-ry, knowledge of the past states of a subject, then that subject is himself.

Of course, it is possible for memory to serve as the basis for a different wayof having knowledge about the past; if a subject knows that his apparentmemories are systematically correlated with the past states of some other object(via being systematically correlated with its memories), then he can infer fromthe existence of a present apparent memory to the past state of that object. Butit surely cannot be suggested that this is the normal operation ofmemory— thatwe infer, via a general belief about the correlation between our present apparentmemories and our past states, from the present apparentmemories to those paststates. There is here a fundamental asymmetry between two ways of gainingknowledge of the past: the one depends upon a general belief, but the other,being underwritten by evolution, does not. Indeed, I should be prepared to arguethat if one attempted to arrive at past-tense judgements only in the inferentialway (in the absence, that is, of a capacity for direct memory), one would not beequipped to understand the conclusions of the supposed inferences.

We have found, therefore, no compelling reason for giving up the view thatour Ideas of ourselves do not permit a gap to open up between knowing, invirtue of the operation of memory, that someone saw a tree burning, andknowing that it was oneself who saw a tree burning. For a subject to haveinformation (or misinformation) in this way, to the effect that someone saw atree, just is for the subject to have information (or misinformation) to the effect

59. This argument should be capable of being recast in terms of the subject’s having knowledge of statesof the world, e.g., that a tree was burning.

132 Gareth Evans

that he saw a tree. For a subject to have an apparent memory of a tree burningis for it to seem to him that a tree burned, and by the same token, for it to seemto him that he saw a tree burn. As Shoemaker said in an earlier work:

I do not express my memory of eating eggs for breakfast by saying ‘I had eggsfor breakfast’ because I think that it is convenient, or advisable, that suchmemories be expressed in this way; I do so because my memory is, precisely, amemory that I had eggs for breakfast.60

Shoemaker’s later work has made it seem to some philosophers that this mustbe wrong. Derek Parfit, for instance, has written:

When I seem to remember an experience, I do indeed seem to rememberhaving it. But it cannot be a part of what I seem to remember about thisexperience that I, the person who now seems to remember it, am the personwho had this experience. That I am is something I automatically assume. (Myapparent memories sometimes come to me simply as the belief that I had acertain experience.) But it is something that I am justified in assuming onlybecause I do not in fact have q-memories of other people’s experiences.61

But our earlier reflections enable us to see what is wrong with this passage:namely that it assumes that the identification which ‘I’ effects (for me now) isan exclusively present-tense identification. (Notice the gloss on ‘I’: ‘the personwho now seems to remember [the experience]’.) Whereas we saw at thebeginning of this section that it is of the essence of an ‘I’-Idea that it effects anidentification which spans past and present.62

Since I think many philosophers would dispute the conclusion I havereached, it may be worth saying a word about why they might do so.

Some of them have been influenced by an observation by BernardWilliamsabout the imagination.63 To imagine being in the West Indies (as opposed toimagining someone being in the West Indies) presumably involves producingin oneself informational states of roughly the kind which might underlie thefirst-person judgement ‘I am in the West Indies’: visual impressions of palm

60. Self-Knowledge and Self-Identity (Cornell University Press, Ithaca, 1963): 33–34.

61. ‘Personal Identity’, Philosophical Review lxxx (1971): 3–27, at p. 15.

62. This has a connection with the idea that memory suffices for personal identity. My judgement (orseeming judgement) ‘I was F�’, made on the basis of apparent memory, can go wrong only in certainways. Either there is no past state of affairs of which the apparent memory constitutes information (noone was — relevantly — F�); or the object involved in the sate of affairs was indeed myself; or, finally, Ido not exist. (This last possibility will seem paradoxical; but see §6.)

63. ‘Imagination and the Self ’, in Problems of the Self (CUP, Cambridge, 1973): 26–45.


trees and steel bands; auditory impressions of the band’s music interminglingwith the breaking of the waves; perhaps suitable kinaesthetic impressions oflying on a beach with the hot sand under one’s body. NowWilliams argues thatimagining being in the West Indies, in this way, is not necessarily to imaginethat one is in the West Indies, or to imagine oneself being in the West Indies. Ithink this is correct: it does not follow, from the fact that a certain piece ofimagining involves certain states of information which would, if they actuallyoccurred, warrant the judgement that p, that the piece of imagining can bedescribed as imagining that p. (I think this can be seen from the fact that it maybe just as wrong to describe the subject as imagining that the place he occupies— here — is in the West Indies, even though the same imagined states ofinformation would, if actual, warrant the judgement that here is in the WestIndies.) So there is a difference between what might be described as imaginingbeing in the West Indies (from the inside), and imagining one’s being in theWest Indies; and I think this has encouraged some philosophers to think thereis a similar gap between remembering, or seeming to remember, being in theWest Indies (from the inside) and remembering, or seeming to remember, one’sbeing in the West Indies.

Williams’s observation does preclude anyone from mounting a purelygrammatical argument for the identity of these informational states; but itcannot show their distinctness. After all, imagining a tree is not necessarilyimagining a tree here, even though seeming to see a tree is, necessarily, seemingto see a tree here. Williams’s point tells us something important about theimagination, but I do not see that it shows us anything about memory.

An equivocation on the term ‘quasi-memory’ (‘q-memory’) providesanother reason why philosophers have not accepted the equation I havesuggested. The notion of q-memory was originally introduced in this way(making it perfectly intelligible): a subject q-remembers an event e if and onlyif (i) he has an apparent memory of such an event, and (ii) that apparentmemory in fact embodies information deriving from the perception of thatevent by a person who is not necessarily himself. Given the notion thus intro-duced, we are able to say that a subject q-remembers an event that he did notwitness, and in consequence that he q-remembers witnessing an event, being infront of a tree, etc., when he did not witness the event, was not in front of a tree,etc. (Of course introducing such a definition leaves the question of the contentof memory states quite untouched; it can still be right to say, as I have, that anapparent memory ofΦ-ing is necessarily an apparentmemory of oneselfΦ-ing.)But now it is somehow supposed that the intelligibility of the notion of q-memory,

134 Gareth Evans

thus introduced, demonstrates the possibility of a faculty which is both like ourmemory in giving subjects knowledge of the past, and unlike it in that the contentof the memory states in no way encroaches upon the question of whose past isconcerned. The informational states of a q-memory faculty announce them-selves, so to speak, as merely q-memories, so that it seems to the subject thatsomeone or other Φ-ed without its in any way seeming to him that he Φ-ed.

Obviously this is a fallacy. I can introduce the term ‘q-perceive’, in such away that a subject can be said to q-perceive a tree provided that he seems to seea tree as a causal result of a process which takes that tree as input, whether ornot that tree is where the subject is disposed to locate it either in space or intime. But, by this purely linguistic manoeuvre, I have not shown the intelligibil-ity of a faculty of q-perception: one which involves informational states whosecontent is simply of the existence, somewhere in space and time, of such-and-such a kind of thing. The manoeuvre does not show the possibility of itsperceptually seeming to the subject that there is a tree without its seeming tohim that there is a tree where he is.

6. The possibility of reference failure

It seems to me to be a corollary of the reflections in this chapter that ourordinary thoughts about ourselves are liable to many different kinds of failings,and that the Cartesian assumption that such thoughts are always guaranteed tohave an object cannot be sustained.

I have emphasized that a subject’s Idea of himself does not require him tohave a current conception of himself; what is required, in the exceptionalcircumstances in which the various avenues of self-knowledge are blocked, isthat the subject be disposed to accept any information accessible in those waysas germane to the thoughts we regard as manifesting self-consciousness. But inthe normal situation, of course, these dispositions are exercised, and he has anevolving conception of himself, embodying information derived in the variousways, and partly retained in memory, which informs his thoughts abouthimself. As with other thoughts which are information-based, there is apresupposition that there is just one thing from which the various elements ofthe conception derive.

Now it appears relatively easy to elaborate examples in which this presup-position is not true. The ‘memory-transfer’ case we discussed in §5 provides asimple example of this kind: provided that the subject was ignorant of the


situation, he would bring both present-tense physical and psychologicalinformation and past-tense physical and psychological information to bearupon his self-conscious reflections, and there would be no one thing fromwhich both kinds of information derived. The subject would be in a muddle,rather as can happen when a subject uses a demonstrative in connection withtwo information-links (via different sense-modalities).64 Of course the subjectcould be apprised of the situation and draw in his horns, inhibiting the tempta-tion to make past-tense self-ascriptions upon the basis of his apparent memo-ries. But the fact that a more guarded mode of thought can have an object inthis situation certainly does not establish that his original thought did.65

A similar kind of situation would arise if the subject received kinaestheticinformation from a body other than that at the origin of his egocentric space; orif the subject’s actions were in fact manifested in a body distinct from the bodywhich he perceives and from which he perceives.

Amore alarming kind of situation would arise if there were in fact no bodywhich could be regarded as the subject’s body. Consider, for example, theperennial nightmare: the idea that a human brain might exist, from birth, in avat, subjected by clever scientists to a complex series of hallucinations (includingkinaesthetic hallucinations), of a kind which would enable the brain to developnormal cognitive faculties. (I shall pretend that I am convinced that this specula-tion is fully intelligible.) Here we have a case where a considerable element of thesubject’s conception of himself, both present and past, derives from nothing. Inall his physical self-ascriptions, there is simply nothing fromwhich his informa-tion derives. When he thinks he is moving, or that his legs are bent, there isnothing of whose physical condition he is, even inaccurately, informed.

In this case, unlike the others, there does not appear to me to be anythingto which the subject can intelligibly retreat. For if the subject were apprised ofthe facts, he would have to abandon the Idea of himself as the occupant of aposition in space, determinable upon the basis of the course of his perceptions.He would have to attempt to think of himself as nowhere. And this means thathe has lost the essential basis of his knowing which element in the objectiveorder he is. He would not, therefore, have an adequate Idea of himself.

64. Suppose one uses the expression ‘this cup’ when one is seeing a cup and feeling a cup (in fact, thoughone does not know it, two cups).

65. In considering this case, we must guard against being tugged in the direction of a purely linguisticor communicative interpretation of the subject’s ‘I’, correlative with someone else’s ‘you’ addressed tohim. See below.

136 Gareth Evans

It may be conceded that, if a subject is to have an adequate Idea of himself,he must be able to make sense of his identity with an element of the objectiveorder; but objected that in this case there is a perfectly good such element,namely a brain, with which the subject, in this nightmarish situation, canidentify himself. But canwemake sense of a self-conscious conception of himselfon the part of the subject, whichmakes such an identification possible—whichpermits him to understand, and possibly accept, ‘I am a brain’? The subject is tothink ‘Somewhere in the world there is a small parcel of grey matter, wrinkled,moist, and soft, about three inches high, and that is me.’ I think that we can feela resistance to this identification, of the same kind which Nagel, and others,wrongly supposed to arise upon the contemplation of our identity with aparticular living animal in theworld (see §1). For this is unlike the ordinary case.There is nothing in this subject’s self-conception which speaks for this identity.In the ordinary case, a subject will be able to make such judgements as ‘I amsitting on a bench in a park facing a round pond, withmy legs slightly bent’; andit is these elements of one’s self-conception— as intimate as any others— thatallow the identification of oneself with a human being, located at such-and-sucha position in space and time, to get a grip. But obviously the physical side of theconception of himself which our unfortunate subject possesses does not encour-age any identification of himself with a brain; and anyway, we are now consider-ing a case in which such elements are extruded from the subject’s conception ofhimself, being, as he now realizes, without any foundation in fact. So theidentification is, so to speak, wholly theoretical, and it remains quite obscurewhat mode of thinking about himself renders it even thinkable.

Many people will regard the remarks I have made about this case as quiteunintelligible. For I have spoken of the subject thinking … and wondering …,while at the same time I have denied that his ‘I’-thoughts-whether revised orunrevised-have an object. Surely (people will object) this is unintelligible, for ifthere is a subject, thinking ‘I’-thoughts, then his ‘I’-thoughts will concernhimself. But why is it thought that self-identification, or thoughts about oneself,are as simple as this — so that, whenever there is a subject, then there is at leastone thing he can unproblematically think about, namely, himself?

This idea comes, I think, from a false analogy with the functioning of ‘I’ asa communicative device.66 For it seems perfectly true that provided there is a

66. Encouraged, perhaps, by the following train of thought: having allowed ‘I’-thoughts to theanaesthetized amnesiac, we suppose that all self-conscious thoughts involve no more than his‘I’-thoughts do— that we may exclude from the basis of our own self-conscious thoughts all the actual


speaker, he can refer to himself, using the device of self-reference ‘I’. (If such adevice does not exist in the language, it can be introduced unproblematically.)But what this means is that it is always possible for an audience to think of thereferent of a token of such a device as the person who uttered that token. Thatis, the audience identifies the referent, in the first instance, by description, as theutterer of certain sounds (or the producer of a certain inscription)— even if thenotion of ‘uttering’ is sufficiently sophisticated to allow the utterer to beregarded as possibly distinct from the person fromwhose lips the words issue.67

To attempt to apply the analogy, we should have to suppose that self-reference in thought is achieved because we think of ourselves (at any time) asthe thinker of this thought, or possibly, as the subject of these experiences. In fact,we already have a considerable amount of evidence that we do not actuallythink of ourselves in this way: the suggestion neglects the way in which ourIdeas of ourselves rest upon various avenues of self-knowledge whose bearingupon ‘I’-thoughts cannot be regarded as underwritten by any identification. Butnor is it clear that the suggestion describes a possible way in which anyone couldthink of himself. We have proceeded too far for us to allow such demonstrativesas are here invoked — this thought, these experiences — to pass withoutscrutiny. For this is the point at which the analogy with the interpretation of ‘I’in speech must break down. It is perfectly intelligible for someone to identifycertain words demonstratively — to have a perfectly clear understanding ofwhich words are in question—without having any idea of their author, or evenof whether any such person, or just one such person, exists. Hence he can knowwhich person is in question in his interpretation of ‘I’, for, knowing whichsounds are in question, he knows what it is to identify a person as their author.But the analogy does not work, precisely because it is not possible to have anadequate Idea of certain pains or thoughts independently of an adequateconception of the person whose thoughts or pains they are. Mental events aredistinguished from one another, and from all other things, by reference to thedistinctness of the person to, or in, whom they occur. So the sheer demonstra-tive element — what we might attempt to regard as an information-link —cannot by itself constitute an adequate Idea of such things. A subject can gaze

information that we take ourselves to have of ourselves, thereby excluding the multiplicity of informa-tion-links that gives rise to the possibility of ill-groundedness. But it is really not obvious that thesupposition is correct. Is it supposed that the Idea I associate with the name ‘Gareth Evans’ is one whichwould genuinely permit it to be a discovery that I am Gareth Evans?

67. See Anscombe, ‘The First Person’: 60.

138 Gareth Evans

inwardly with all the intensity he can muster, and repeat to himself ‘this pain’,‘this pain’, as he concentrates upon his pain, but he will not thereby be able toknow which pain is in question unless this provides him with a basis foridentifying the pain with a pain conceived as an element in the objective order— which means a pain conceived as the pain of this or that person in theobjective order. Consequently, he cannot have an adequate Idea of these mentalevents unless he knows to which person they are happening. So he cannotidentify himself by reference to them.68

I do not see, then, that it is absurd to suppose that there might be a subjectof thought who is not in a position to identify himself, and whose attempts atself-identification fail to net any object at all.

I have used an extreme case to make this point, rather than the morefamiliar kind of fantasy, in which a brain which was once the controlling organof a human being is extracted from the body and kept alive in vitro, because Iam less clear about whether the subject in the more familiar case should be saidto have no adequate Idea of himself. In order to see why this is open to ques-tion, we must reflect a moment about the situation of a man who is paralysed,and who has lost the use of his senses. Now, terrible though this situation mustbe, we need not automatically think that the subject has lost the capacity tomake sense of an identification of himself with an element in the objectiveorder; we might want to say that he retains knowledge of how to make such anidentification, even though practically it is now beyond him. We might bewilling tomake such a statement because we are willing to accept the condition-al judgement that, if the normal use of his body and of his senses were restoredto him, he could locate himself. (Similarly, even though his knowledge of theright-hand side rests upon dispositions which he cannot nowmanifest, we mightbe prepared to allow that he has a perfectly determinate side of himself inmind.) It is certainly less clear, but it is at least arguable, that the same kind ofthing should be said of a disembodied brain. Being one of us, the victim wouldof course think of himself as somewhere�; he would think ‘I could find out whereI am, if only this damned darkness would lift’. And perhaps the dispositionsremaining to the subject— dispositions which could be exercised only if he werere-embodied —might be thought to allow us to say that he had a quite definite

68. His situation is rather like that of someone who can perceive certain things but has no idea of wherethey are. But notice that the let-out available for that subject is not available for this one: he cannot thinkof the thoughts or pains as ‘the pains of which I currently have information’ or ‘the pains which arecausally responsible for these informational states’ (which informational states?).


place inmind as the one he occupied: a place which would in fact be the position(or close to the position) of his brain; and it might be thought that this providesa basis, if only a slender basis, for an identification of himself with his brain.

If this is so, it rests upon a contingent fact about our construction, namelythat the control centre of the human animal is located within it. It seemspossible to envisage organisms whose control centre is outside the body, andconnected to it by communication links capable of spanning a considerabledistance.69 An organism of this kind could have an Idea of itself like our own,but if it did, it would be unable to cope with the situation that would arise whenthe control centre survived the destruction of the body it controlled. Thinkinglike us, the subject could of course have to regard itself as somewhere, but inthis case it would not make any sense to identify a particular place in the worldas the place it thought of as here. The place occupied by the control centre iscertainly not the subject’s here�; and even if we counterfactually suppose thecontrol centre re-equipped with a body, there is no particular place where thatbody would have to be. Because its ‘here’ picks out no place, there is no bit ofmatter, no persisting thing, which the subject’s Idea of itself permits us toregard as what it identifies as itself. Here, then, we have a very clear situation inwhich a subject of thought could not think of itself as ‘I’; its ‘I’ — its habitualmode of thought about itself is simply inadequate for this situation. (It forcesthe subject to think in ways which are no longer appropriate.)70 This casehelps us to see that the reason we do not find the ‘disembodied brain in a vat’case very disturbing, conceptually, is that the brain is also the last remainingpart of the subject’s body. (The case is often presented as a limiting case ofamputation.) A tiny foothold is thus provided for the idea that the subject iswhere the brain is, and hence for the idea that the brain is what the subject is.

7. Conclusions

The preceding reflections constitute only the most preliminary approach to thisdifficult subject, but I shall try to draw the threads together by stating sometentative conclusions.

69. I have to confess that it is difficult to understand how such beings could have evolved naturally. Butperhaps we can think entirely of artificially produced specimens.

70. If we can think of this subject, it is with the aid of newer and more sophisticated machinery; not bythinking ourselves into his situation.

140 Gareth Evans

1. Our self-conscious thoughts about ourselves are irreducible to any othermode of thought; in particular, they cannot be regarded as involving theidentification of an object by any description. Self-conscious thought aboutoneself is thought informed, or at least liable to be informed, by informationwhich the subject has (or can normally acquire) about himself in a variety ofdifferent ways. There is always a gap between grasping that the Φ is F, andgrasping that I am F — there is always room for the realization that I am the Φbecause there is no description Φ such that grasping that the Φ is F guaranteesthe subject’s realization of the bearing of the information which he has, or is ina position to acquire, in these various ways, upon the proposition that the Φ isF. (There is just as much of a gap between the subject’s grasp of ‘TheΦ is F�’ andof ‘I am F�’ when he is amnesiac, anaesthetized, and without the use of his sensesas there is when he currently has a conception of himself acquired in thesevarious ways.) Furthermore, the way in which the subject knows which objectis in question — his grasp of what it is for an identity-proposition of the formI am δt to be true — cannot be reduced to knowledge of what it is for δt =the Φ to be true. The only plausible candidate for an instance of ‘the Φ�’ thatwould falsify these claims is ‘the person here’. But this produces the quitemisleading impression that there is some priority of ‘here’ over ‘I’, whereasone’s ‘I’-Idea and one’s ‘here’-Ideas are really two sides of a single capacity, eachwholly dependent upon the other. Both ‘I’-thoughts and ‘here’-thoughts areways in which the subject’s capacity to locate himself in the objective spatialorder is exploited. (See Evans 1982, §6.3.)

2. Despite recent philosophical claims to the contrary, our thoughts aboutourselves are about objects — elements of reality. We are, and can make senseof ourselves as, elements of the objective order of things. Our thinking aboutourselves conforms to the Generality Constraint — we are able to conceive ofendless states of affairs involving ourselves, and what we conceive is notnecessarily what it is like for us, or what it will be like for us, to be aware of, orbe in a position to know the existence of, such a state of affairs. Therefore weare not Idealists about ourselves, and this means that we can and must think ofourselves as elements of the objective order. All the peculiarities we havenoticed about ‘I’-thoughts are consistent with, and, indeed, at points encourage,the idea that there is a living human being which those thoughts concern.


3. Our thoughts about ourselves are in no way hospitable to Cartesianism.Our customary use of ‘I’ simply spans the gap between the mental and thephysical, and is no more intimately connected with one aspect of our self-conception than the other.

4. The Ideas we have of ourselves, like almost all Ideas we have, rest uponcertain empirical presuppositions, and are simply inappropriate to certaindescribable situations in which these presuppositions are false.

Chapter 7

The problem of the essential indexical

John PerryStanford University

I once followed a trail of sugar on a supermarket floor, pushing my cart downthe aisle on one side of a tall counter and back up the aisle on the other, seekingthe shopper with the torn sack to tell him he was making amess.With each triparound the counter, the trail became thicker. But I seemed unable to catch up.Finally it dawned on me. I was the shopper I was trying to catch.

I believed at the outset that the shopper with a torn sack was making a mess.And I was right. But I did not believe that I was making a mess. That seems tobe something I came to believe. And when I came to believe that, I stoppedfollowing the trail around the counter and rearranged the torn sack in my cart.My change in beliefs seems to explain my change in behavior. My aim in thispaper is to make a key point about the characterization of this change, and ofbeliefs in general.

At first, characterizing the change seems easy. My beliefs changed, didn’tthey, in that I came to have a new one, namely, that I am making a mess. Butthings are not so simple.

The reason they are not is the importance of the word ‘I’ in my expressionof what I came to believe.When we replace it with other designations of me, weno longer have an explanation of my behavior and so, it seems, no longer anattribution of the same belief. It seems to be an essential indexical. But withoutsuch a replacement, all we have to identify the belief is the sentence ‘I ammaking a mess.’ But that sentence by itself does not seem to identify the crucialbelief, for if someone else had said it, they would have expressed a differentbelief, a false one.

I argue that the essential indexical poses a problem for various otherwiseplausible accounts of belief. I first argue that it is a problem for the view thatbelief is a relation between subjects and propositions conceived as bearers oftruth and falsity. The problem is not solved merely by replacing or supple-menting this with a notion of de re belief. Nor is it solved by moving to a

144 John Perry

notion of a proposition that, rather than true or false absolutely is only true orfalse at an index or in a context (at a time, for a speaker, say). Its solutionrequires us to make a sharp distinction between objects of belief and beliefstates, and to realize that the connection between them is not so intimate asmight have been supposed.

Locating Beliefs

I want to introduce two more examples. In the first, a professor, who desires toattend the department meeting on time and believes correctly that it begins atnoon, sits motionless in his office at that time. Suddenly, he begins to move.What explains his action? A change in belief. He believed all along that thedepartment meeting starts at noon; he came to believe, as he would have put it,that it starts now.

The author of the book Hiker’s Guide to the Desolation Wilderness stands inthe wilderness beside Gilmore Lake, looking at the Mt. Tallac trail as it leavesthe lake and climbs the mountain. He desires to leave the wilderness. Hebelieves that the best way out fromGilmore Lake is to follow theMt. Tallac trailup the mountain to Cathedral Peaks trail, on to the Floating Island trail,emerging at Spring Creek Tract Road. But he does not move. He is lost. He isnot sure whether he is standing beside Gilmore Lake, looking at Mt. Tallac, orbeside Clyde Lake looking at Jack’s Peak, or beside Eagle Lake looking at one oftheMaggie peaks. Then he begins to move along the Mt. Tallac trail. If asked, hewould have explained the crucial change in his beliefs this way: ‘I came tobelieve that this is the Mt. Tallac trail and that is Gilmore Lake.’

In these three cases, the subjects in explaining their actions would useindexicals to characterize certain beliefs they came to have. These indexicals areessential, in that replacement of them by other terms destroys the force of theexplanation, or at least requires certain assumptions to be made to preserve it.

Suppose I had said, in the manner of de Gaulle, ‘I came to believe that JohnPerry is making a mess.’ I would no longer have explained why I stopped andlooked in my own cart. To explain that, I would have to add, ‘and I believe thatI am John Perry,’ bringing in the indexical again. After all, suppose I had reallygiven my explanation in the manner of de Gaulle, and said ‘I came to believethat de Gaulle is making a mess.’ That would not have explained my stoppingat all. But it really would have explained it every bit as much as ‘I came tobelieve John Perry is making a mess.’ For if I added ‘and I believe that I am de

The problem of the essential indexical 145

Gaulle,’ the explanations would be on par. The only reason ‘I came to believeJohn Perry is making a mess’ seems to explain my action is our natural assump-tion that I did believe I was John Perry and did not believe I was de Gaulle. Soreplacing the indexical ‘I’ with another term designating the same person reallydoes, as claimed, destroy the explanation.

Similarly, our professor, as he sets off down the hall, might say ‘I believe themeeting starts at noon.’ In accepting the former as an explanation, we would beassuming he believes it is now noon. If he believed it was now 5 P.M., he wouldnot have explained his departure by citing his belief that the meeting starts atnoon, unless he was a member of a department with very long meetings. Afterall, he believed that the meeting started at noon all along, so that belief canhardly explain a change in his behavior. Basically similar remarks apply to thelost author.

I shall use the term ‘locating beliefs’ to refer to one’s beliefs about whereone is, when it is, and who one is. Such beliefs seem essentially indexical.Imagine two lost campers who trust the same guidebook but disagree aboutwhere they are. If we were to try to characterize the beliefs of these camperswithout the use of indexicals, it would seem impossible to bring out thisdisagreement. If, for example, we characterized their beliefs by the set of ‘eternalsentences’ drawn from the guidebook they would mark ‘true,’ there is no reasonto suppose that the sets would differ. They couldmark all of the same sentences‘true,’ and still disagree in their locating beliefs. It seems that there has to besome indexical element in the characterization of their beliefs to bring out thisdisagreement. But as we shall see, there is no room for this indexical element inthe traditional way of looking at belief, and even when its necessity is recog-nized, it is not easy to see how to fit it in.

The Doctrine of Propositions

I shall first consider how the problem appears to a traditional way of thinkingof belief. The doctrines I describe were held by Frege, but I shall put them in away that does not incorporate his terminology or the details of his view. Thistraditional way, which I call the ‘doctrine of propositions,’ has three maintenets. The first is that belief is a relation between a subject and an object, thelatter being denoted, in a canonical belief report, by a that-clause. So ‘Carterbelieves that Atlanta is the capital of Georgia’ reports that a certain relation,believing, obtains between Carter and a certain object — at least in a suitably

146 John Perry

wide sense of the object — that Atlanta is the capital of Georgia. These objectsare called propositions.

The second and the third tenets concern such objects. The second is thatthey have a truth-value in an absolute sense, as opposed tomerely being true fora person or at a time. The third has to do with how we individuate them. It isnecessary, for that S and that S´ to be the same, that they have the same truth-value. But it is not sufficient, for that the sea is salty and that milk is white arenot the same proposition. It is necessary that they have the same truth condi-tion, in the sense that they attribute to the same objects the same relation. Butthis also is not sufficient, for that Atlanta is the capital of Georgia and thatAtlanta is the capital of the largest state east of the Mississippi are not the sameproposition. Carter, it seems, might believe the first but not the second.Propositions must not only have the same truth-value and concern the sameobjects and relations, but also involve the same concepts. For Frege, this meantthat if that S = that S’, S and S’ must have the same sense. Others might eschewsenses in favor of properties and relations, yet others take concepts to be justwords, so that sameness of propositions is just sameness of sentences. Whatthese approaches have in common is the insistence that propositions must beindividuated in amore ‘fine-grained’ way than is provided by truth-value or thenotion of truth conditions employed above.

The Problem

It is clear that the essential indexical is a problem for the doctrine of proposi-tions. What answer can it give to the question, ‘What did I come to believewhen I straightened up the sugar?’ The sentence ‘I ammaking a mess’ does notidentify a proposition. For this sentence is not true or false absolutely, but onlyas said by one person or another; had another shopper said it when I did, hewould have been wrong. So the sentence by which I identify what I came tobelieve does not identify, by itself, a proposition. There is a missing conceptualingredient: a sense for which I am the reference, or a complex of properties Ialone have, or a singular term that refers to no one but me. To identify theproposition I came to believe, the advocate of the doctrine of propositions mustidentify this missing conceptual ingredient.

An advocate of the doctrine of propositions, his attention drawn to index-icals, might take this attitude towards them: they are communicative shortcuts.Just before I straightened up the sack I must have come to believe some


propositions with the structure α is making a mess,where α is some concept thatI alone ‘fit’ (to pick a phrase neutral among the different notions of a concept).When I say ‘I believe I ammaking a mess,’ my hearers know that I believe somesuch proposition of this form; which one in particular is not important for thepurposes at hand.

If this is correct, we should be able to identify the proposition I came tobelieve, even if doing so is not necessary for ordinary communicative purposes.But then the doctrine of propositions is in trouble, for any candidate will fallprey to the problemsmentioned above. If that α is making a mess is what I cameto believe, then ‘I came to believe that A is making a mess,’ where A expressedα, should be an even better explanation than the original, where I used ‘I’ as acommunicative shortcut. But, as we saw, any such explanation will be defective,working only on the assumption that I believed that I was α.

To this it might be replied that though there may be no replacement for ‘I’that generally preserves explanatory force, all that needs to be claimed is thatthere is such a replacement on each occasion. The picture is this. On eachoccasion that I use ‘I,’ there is some concept I have in mind that fits me unique-ly, and which is the missing conceptual ingredient in the proposition thatremains incompletely identified when I characterize my beliefs. The concept Iuse to think of myself is not necessarily the same each time I do so, and ofcourse I must use a different one than others do, since it must fit me and notthem. Because there is no general way of replacing the ‘I’ with a term that getsat the missing ingredient, the challenge to do so in response to a particularexample is temporarily embarrassing. But the doctrine of propositions does notrequire a general answer.

This strategy does not work for two reasons. First, even if I was thinking ofmyself as, say, the only bearded philosopher in a Safeway store west of theMississippi, the fact that I came to believe that the only such philosopher wasmaking a mess explains my action only on the assumption that I believed thatI was the only such philosopher, which brings in the indexical again. Second, inorder to provide me with an appropriate proposition as the object of belief, themissing conceptual ingredient will have to fit me. Suppose I was thinking ofmyself in the way described, but that I was not bearded and was not in aSafeway store— I had forgotten that I had shaved and gone to the A&P instead.Then the proposition supplied by this strategy would be false, while what I cameto believe, that I was making a mess, was true.

This strategy assumes that whenever I have a belief I would characterize byusing a sentence with an indexical d,

148 John Perry

I believe that … d …

that there is some conceptual ingredient c, such that it is also true that,

I believe that d is c

and that, on this second point, I am right. But there is no reason to believe thiswould always be so. Each time I say ‘I believe it is now time to rake the leaves,’I need not have some concept that uniquely fits the time at which I speak.

From the point of view of the doctrine of propositions, belief reports suchas ‘I believe that I am making a mess’ are deficient, for there is a missingconceptual ingredient. From the point of view of locating beliefs, there issomething lacking in the propositions offered by the doctrine, a missingindexical ingredient.

The problem of the essential indexical reveals that something is badlywrong with the traditional doctrine of propositions. But the traditional doctrinehas its competitors anyway, in response to philosophical pressures from otherdirections. Perhaps attention to these alternative or supplementary models ofbelief will provide a solution to our problem.

De Re Belief

One development in the philosophy of belief seems quite promising in thisrespect. It involves qualifying the third tenet of the doctrine of propositions, toallow a sort of proposition individuated by an object or sequence of objects, anda part of a proposition of the earlier sort. The motivation for this qualificationor supplementation comes from a type of belief report, which gives rise to thesame problem, that of the missing conceptual ingredient, as does the problemof the essential indexical.

The third tenet of the doctrine of propositions is motivated by the failure ofsubstitutivity of coreferential terms within the that-clause following ‘believes.’But there seems to be a sort of belief report, or a way of understanding somebelief reports, that allows such substitution, and such successful substitutionbecomes a problem for a theory designed to explain its failure. For supposePatrick believes that, as he would put it, the dean is wise. Patrick does not knowFrank, much less know that he lives next to the dean, and yet I might in certaincircumstances say ‘Patrick believes Frank’s neighbor is wise.’ Or I might say‘There is someone whom Patrick believes to be wise,’ and later on identify that


someone as ‘Frank’s neighbor.’ The legitimacy of this cannot be understood onthe unqualified doctrine of propositions; I seem to have gone from one proposi-tion, that the dean of the school is wise, to another, that Frank’s neighbor is wise;but the fact that Patrick believes the first seems to be no reason he shouldbelieve the second. And the quantification into the belief report seems to makeno sense at all on the doctrine of propositions, for the report does not relatePatrick to an individual known variously as ‘the dean’ and ‘Frank’s neighbor,’but only with a concept expressed by the first of these terms.

The problem here is just that of a missing conceptual ingredient. It lookedin the original report as if Patrick was being said to stand in the relation of abelief to a certain proposition, a part of which was a conceptual ingredientexpressed by the words ‘the dean.’ But if I am permitted to exchange thosewords for others, ‘Frank’s neighbor,’ which are not conceptually equivalent,then apparently the initial part of the proposition he was credited with belief inwas not the conceptual ingredient identified by ‘the dean’ after all. So whatproposition was it Patrick was originally credited with belief in? And ‘There issomeone such that Patrick believes that he is wise’ seems to credit Patrick withbelief in a proposition, without telling us which one. For after the ‘believes’ wehave only ‘he is wise,’ where the ‘he’ does not give us an appropriate conceptualingredient, but functions as a variable ranging over individuals.

We do seem in some circumstances to allow such substitutivity, and makeready sense of quantification into belief reports. So the doctrine of propositionsmust be qualified.We can look upon this sort of belief as involving a relation toa new sort of proposition, consisting of an object or sequence of objects and aconceptual ingredient, a part of a proposition of the original kind, or what wemight call an ‘open proposition.’ This sort of belief and this kind of propositionwe call ‘de re,’ the sort of belief and the sort of proposition that fits the originaldoctrine, ‘de dicto.’ Taken this way, we analyze ‘Patrick believes that the dean ofthe school is wise,’ as reporting a relation between Patrick and a propositionconsisting of a certain person variously describable as ‘the dean’ and ‘Frank’sneighbor’ and something, that x is wise, which would yield a proposition withthe addition of an appropriate conceptual ingredient. Since the dean himself,and not just a concept expressed by the words ‘the dean’ is involved, substitu-tion holds and quantification makes sense.

Here, as in the case of the essential indexical, we were faced with a missingconceptual ingredient. Perhaps, then, this modification of the third tenet willsolve the earlier problem as well. But it will not. Even if we suppose— as I thinkwe should— thatwhen I said ‘I believe that I ammaking amess’ I was reporting a

150 John Perry

de re belief, our problem will remain.One problem emerges when we look at accounts that have been offered of

the conditions under which a person has a de re belief. The most influentialtreatments of de re belief have tried to explain it in terms of de dicto belief orsomething like it. Some terminological regimentation is helpful here. Let uscouch reports of de re belief in terms ‘X believes of a that he is so and so,’reserving the simpler ‘X believes that a is so and so’ for de dicto belief. Thesimplest account of de re belief in terms of de dicto belief is this:

X believes of y that he is so and so

just in case

there is a concept α such that α fits y and X believes that α is so and so.

Now it is clear that if this is our analysis of de re belief, the problem of theessential indexical is still with us. For we are faced with the same problem wehad before. I can believe that I am making a mess, even if there is no concept αsuch that I alone fit α and I believe that α is making a mess. Since I do not haveany de dicto belief of the sort, on this account I do not have a de re belief of theright sort either. So, even allowing de re belief, we still do not have an accountof the belief I acquired.

Now this simple account of de re belief has not won many adherents,because it is commonly held that de re belief is a more interesting notion thanit allows. This proposal trivializes it. Suppose Nixon is the next President. SinceI believe that the next President will be the next President, I would on thisproposal believe of Nixon that he is the next President, even though I amthoroughly convinced that Nixon will not be the next President.1

To get a more interesting or useful notion of de re belief, philosophers havesuggested that there are limitations on the conceptual ingredient involved in thede dicto belief that yields the de re belief. Kaplan, for example, requires not onlythat there be some α such that I believe that αwill be the next President and thatα denotes Nixon, for me to believe of Nixon that he will be the next President,but also that α be a vivid name of Nixon for me (1969, 225ff). Hintikka requiresthat α denote the same individual in every possible world compatible with whatI believe (1967, 40ff). Each of these philosophers explains these notions in sucha way that in the circumstances imagined, I would not believe of Nixon that heis the next President.

1. For the classic discussion of these problems, see Quine (1966).


However well these proposals deal with other phenomena connected withde re belief, they cannot help with the problem of the essential indexical. Theytighten the requirements laid down by the original proposal, but those wereapparently already too restrictive. If in order to believe that I ammaking amessI need not have any conceptual ingredient α that fits me, a fortiori I am notrequired to have one that is a vivid name of myself for me, or one that picks outthe same individual in every possible world compatible with what I believe.

Perhaps this simply shows that the approach of explaining de re belief interms of de dicto belief is incorrect. I think it does show that. But even so, theproblem remains. Suppose we do not insist on an account of de re belief interms of de dicto belief, but merely suppose that whenever we ascribe a belief,and cannot find a suitable complete proposition to serve as the object becauseof a missing conceptual ingredient, we are dealing with de re belief. Then wewill ascribe a de re belief to me in the supermarket, I believed of John Perrythat he was making a mess. But it will not be my having such a de re belief thatexplains my action.

Suppose there were mirrors at either end of the counter so that as I pushedmy cart down the aisle in pursuit I sawmyself in the mirror. I take what I see tobe the reflection of the messy shopper going up the aisle on the other side, notrealizing that what I am really seeing is a reflection of a reflection of myself. Ipoint and say, truly, ‘I believe that he is making a mess.’ In trying to find asuitable proposition forme to believe, we would be faced with the same sorts ofproblems we had with my earlier report, in which I used ‘I’ instead of ‘he.’ Wewould not be able to eliminate an indexical element in the term referring to me.So here we have de re belief; I believe of John Perry that he is making a mess.But then that I believe of John Perry that he is making a mess does not explainmy stopping; in the imagined circumstances I would accelerate, as would theshopper I was trying to catch. But then, even granting that when I say ‘I believethat I ammaking a mess’ I attribute to myself a certain de re belief, the belief ofJohn Perry that he is making a mess, our problem remains.

If we look at it with the notion of a locating belief in mind, the failure of theintroduction of de re belief to solve our problems is not surprising. De repropositions remain nonindexical. Propositions individuated in part by objectsremain as insensitive to what is essential in locating beliefs as those individuatedwholly by concepts. Saying that I believed of John Perry that he was making amess leaves out the crucial change, that I came to think of the messy shopper notmerely as the shopper with the torn sack, or theman in themirror, but as me.

152 John Perry

Relativized Propositions

It seems that to deal with essential indexicality we must somehow incorporatethe indexical element into what is believed, the object of belief. If we do so, wecome up against the second tenet of the doctrine of propositions, that suchobjects are true or false absolutely. But the tools for abandoning this tenet havebeen provided in recent treatments of the semantics of modality, tense, andindexicality. So this seems a promising direction.

In possible-worlds semantics for necessity and possibility we have thenotion of truth at a world. In a way this does not involve a new notion of aproposition and in a way it does. When Frege insisted that his ‘thoughts’ weretrue or false absolutely, he did not mean that they had the same truth-value inall possible worlds. Had he used a possible worlds framework, he would havehad their truth-values vary from world to world, and simply insisted on adeterminate truth-value in each world and in particular in the actual world. Ina way, then, taking propositions to be functions from possible worlds to truth-values is just a way of looking at the old notion of a proposition.

Still, this way of looking at it invites generalization that takes us away fromthe old notion. From a technical point of view, the essential idea is that aproposition is, or is represented by, a function from an index to a truth-value;when we get away from modality, this same technical idea may be useful,though something other than possible worlds are taken as indices. To deal withtemporal operators, we can use the notion of truth at a time. Here the indiceswill be times, and our propositions will be functions from times to truth-values.For example, that Elizabeth is Queen of England is a proposition true in 1960 butnot in 1940. Hence ‘At some time or other Elizabeth is Queen of England’ istrue, simpliciter.2

Now consider “I am making a mess.’ Rather than thinking of this aspartially identifying an absolutely true proposition, with the ‘I’ showing theplace of the missing conceptual ingredient, why not think of it as completelyidentifying a newfangled proposition, that is true or false only at a person? Moreprecisely, it is one that is true or false at a time and a person, since though truewhen I said it, it has since occasionally been false.

If we ignore possibility and necessity, it seems that regarding propositionsas functions to truth-values from indices that are pairs of persons and times will

2. See Montague (1974), especially ‘Pragmatics’, and Scott (1970).


do the trick, and that so doing will allow us to exploit relations between elementswithin the indices to formulate rules that bring out differences between indexicals.‘I am tired now’ is true at the pair consisting of the person a and the time t ifand only if a is tired at t, while ‘You will be tired’ is true at the same index if andonly if the addressee of a at t is tired at some time later than t.

Does this way of looking at the matter solve the problem of the essentialindexical? I say ‘I believe that I am making a mess.’ On our amended doctrineof propositions, this ascribes a relation between me and that I am making amess, which is a function from indices to truth- values. The belief report seemsto completely specify the relativized proposition involved; there is no missingconceptual ingredient. So the problem must be solved.

But it is not. I believed that a certain proposition, that I am making a messwas true — true for me. So belief that this proposition was true for me thendoes not differentiate me from some other shopper, who believes that I ammaking a mess was true for John Perry. So this belief cannot be what explainsmy stopping and searching my cart for the torn sack. Once we have adoptedthese newfangled propositions, which are only true at times for persons, wehave to admit also that we believe them as true for persons at times, and notabsolutely. And then our problem returns.

Clearly an important distinction must be made. All believing is done bypersons at times, or so we may suppose. But the time of belief and the persondoing the believing cannot be generally identified with the person and timerelative to which the propositions believed is held true. You now believe thatthat I am making a mess was true for me, then, but you certainly do not believeit is true for you now, unless you are reading this in a supermarket. Let us callyou and now the context of belief, andme and then the context of evaluation. Thecontext of belief may be the same as the context of evaluation, but need not be.

Now the mere fact that I believed that proposition that I am making a messto be true for someone at some time did not explain my stopping the cart. Youbelieve so now, and doubtless have no more desire to mess up supermarketsthan I did. But you are not bending over to straighten up a sack of sugar.

The fact that I believed this proposition true for Perry at the time he was inthe supermarket does not explain my behavior either. For so did the othershopper. And you also now believe this proposition was true for Perry at thetime he was in the supermarket.

The important difference seems to be that for me the context of belief wasjust the context of evaluation, but for the other shopper it was not and for youit is not. But this does not do the trick either.

154 John Perry

Consider our tardy professor. He is doing research on indexicals, and haswritten on the board ‘My meeting starts now.’ He believes that the propositionexpressed by this sentence is true at noon for him. He has believed so for hours,and at noon the context of belief comes to be the context of evaluation. Thesefacts give us no reason to expect him to move.

Or suppose I think tomyself that the personmaking themess should say so.Turning my attention to the proposition, I certainly believe that I am making amess is true for the person who ought to be saying it (or the person in themirror, or the person at the end of the trail of sugar) at that time. The contextof evaluation is just the context of belief. But there is no reason to suppose Iwould stop my cart.

One supposes that in these cases the problem is that the context of belief isnot believed to be the context of evaluation. But formulating the required beliefwill simply bring up the problem of the essential indexical again. Clearly andcorrectly we want the tardy professor, when he finally sees he must be off to themeeting, to be ready to say ‘I believe that the time at which it is true that themeeting starts now is now.’ On the present proposal, we analyze the belief hethereby ascribes to himself as belief in the proposition that the time at which itis true that the meeting starts now is now. But he certainly can believe at noonthat this whole proposition is true at noon, without being ready to say ‘It isstarting now’ and leave. We do not yet have a solution to the problem of theessential indexical.

Limited Accessibility

One may take all that has been said so far as an argument for the existence ofa special class of propositions, propositions of limited accessibility. For whathave we really shown? All attempts to find a formula of the form ‘A is makinga mess,’ with which any of us at any time could express what I believed, havefailed. But one might argue that we can hardly suppose that there was notanything that I believed; surely I believed just that proposition which I ex-pressed, on that occasion, with the words ‘I am making a mess.’ That wecannot find a sentence that always expresses this proposition when said byanyone does not show that it does not exist. Rather it should lead us to theconclusion that there is a class of propositions that can only be expressed inspecial circumstances. In particular, only I could express the proposition Iexpressed when I said ‘I am making a mess.’ Others can see, perhaps by


analogy with their own case, that there is a proposition that I express, but it isin a sense inaccessible to them.

Similarly, at noon on the day of the meeting, we could all express theproposition the tardy professor expressed with the words ‘The meeting startsnow.’ But once that time has passed, the proposition becomes inaccessible. Wecan still identify it as the proposition that was expressed by those words at thattime. But we cannot express it with those words any longer, for with eachpassingmoment they express a different proposition. And we can find no otherwords to express it.

The advocate of such a stock of propositions of limited accessibility may notneed to bring in special propositions accessible only at certain places. For it isplausible to suppose that other indexicals can be eliminated in favor of ‘I’ and‘now.’ Perhaps ‘That is Gilmore Lake’ just comes to ‘What I see now in front ofme is Gilmore Lake.’ But elimination of either ‘I’ or ‘now’ in favor of the otherseems impossible.

Such a theory of propositions of limited accessibility seems acceptable, evenattractive, to some philosophers.3 Its acceptability or attractiveness will dependon other parts of one’s metaphysics; if one finds plausible reasons elsewhere forbelieving in a universe that has, in addition to our common world, myriads ofprivate perspectives, the idea of propositions of limited accessibility will fit rightin.4 I have no knock-down argument against such propositions, or the meta-physical schemes that find room for them. But I believe only in a commonactual world. And I do not think the phenomenon of essential indexicalityforces me to abandon this view.

The Obvious Solution?

Let us return to the device of the true/false exam. Suppose the lost author hadbeen given such an exam before and after he figured out where he was. Wouldwe expect any differences in his answers? Not so long as the statements con-tained no indexicals. ‘Mt. Tallac is higher than either of the Maggie Peaks’would have beenmarked the same way before and after, the same way he wouldhave marked it at home in Berkeley. His mark on that sentence would tell us

3. Frege seems to accept something like it, as necessary for dealing with ‘I’ (1918).

4. See Castañeda (1977), especially section II.

156 John Perry

nothing about where he thought he was. But if the exam were to contain suchsentences as ‘That is Gilmore Lake in front of me,’ we would expect a dramaticchange, from ‘False’ or ‘Unsure’ to ‘True.’

Imagine such an exam given to various lost campers in different parts of theWilderness. We could classify the campers by their answers, and such a classifi-cation would be valuable for prediction and explanation. Of all the camperswhomarked ‘This is Gilmore Lake’ with ‘True,’ we would say they believed thatthey were at Gilmore Lake. And we should expect them to act accordingly; ifthey possessed the standard guidebook and wished to leave the Wilderness, wemight expect what is, given one way of looking at it, the same behavior: takingthe path up the mountain above the shallow end of the lake before them.

Now consider all the good-hearted people who have ever been in a super-market, noticed sugar on the floor, and been ready to say ‘I ammaking a mess.’They all have something important in common, something that leads us toexpect their next action to be that of looking into their grocery carts in searchof the torn sack. Or consider all the responsible professors who have everuttered ‘The department meeting is starting now.’ They too have somethingimportant in common; they are in a state that will lead those just down the hallto go to the meeting, those across campus to curse and feel guilty, those onleave to smile.

What the members within these various groups have in common is notwhat they believe. There is no de dicto proposition that all the campers orshoppers or professors believe. And there is no person whom all the shoppersbelieve to bemaking amess, no lake all the campers believe to be Gilmore Lake,and no time at which all the professors believe their meetings to be starting.

We are clearly classifying the shoppers, campers, and professors into groupscorresponding to what we have been calling ‘relativized propositions’ —abstract objects corresponding to sentences containing indexicals. But whatmembers of each group have in common, which makes the groups significant,is not belief that a certain relativized proposition is true. Such belief, as we saw,is belief that such a proposition is true at some context of evaluation. Now all ofthe shoppers believe that that I am making a mess is true at some context ofevaluation or other, but so does everyone else who has ever given it a moment’sthought. And similar remarks apply to the campers and the professors.

If believing the same relativized proposition is not what the members ofeach of the groups have in common with one another, why is it being used as aprinciple of classification? I propose we look at things in this way. The shoppers,for example, are all in a certain belief state, a state that, given normal desires and


other belief states they can be expected to be in, will lead each of them toexamine his cart. But although they are all in the same belief state (not the sametotal belief state, of course), they do not all have the same belief (believe thesame thing, have the relation of belief to the same object).

We use sentences with indexicals or relativized propositions to individuatebelief states, for the purposes of classifying believers in ways useful for explana-tion and prediction. That is, belief states individuated in this way enter into ourcommonsense theory about human behavior and more sophisticated theoriesemerging from it. We expect all good-hearted people in the state that leadsthem to say ‘I ammaking amess’ to examine their grocery carts, nomatter whatbelief they have in virtue of being in that state. That we individuate belief statesin this way doubtless has something to do with the fact that one criterion forbeing in the states we postulate — at least for articulate, sincere adults — isbeing disposed to utter the indexical sentence in question. A good philosophyof mind should explain this in detail; my aim is merely to get clear about whatit is that needs explaining.

The proposal, then, is that there is not an identity, or even an isomorphiccorrespondence, but only a systematic relationship between the belief states oneis in and what one thereby believes. The opposite assumption, that belief statesshould be classified by propositions believed, seems to be built right intotraditional philosophies of belief. Given this assumption, whenever we havebelievers in the same belief state, we must expect to find a proposition they allbelieve, and differences in belief state lead us to expect a difference in proposi-tion believed. The bulk of this paper consisted in following such leads tonowhere (or to propositions of limited accessibility).

Consider a believer whose belief states are characterized by a structure ofsentences with indexicals or relativized propositions (those marked ‘true’ in avery comprehensive exam, if we are dealing with an articulate, sincere adult).This structure, together with the context of belief— the time and identity of thespeaker — will yield a structure of de re propositions. The sequence of objectswill consist of the values that the indexicals take in the context. The openpropositions will be those yielded by the relativized proposition when shorn ofits indexical elements. These are what the person believes, in virtue of being inthe states he is in, when and where he is in them.5

5. This two-tiered structure of belief states and propositions was suggested by David Kaplan’s system ofcharacters and contents (1979).While Kaplan’smotivations for the distinctionwere basically semantical,it seems tome that the present considerations also supply an epistemological motivation for it. (See also

158 John Perry

This latter structure is important, and classifications of believers by whatthey believe are appropriate for many purposes. For example, usually, when abeliever moves from context to context, his belief states adjust to preservebeliefs held. As time passes, I go from the state corresponding to ‘The meetingwill begin’ to the one corresponding to ‘The meeting is beginning’ and finallyto ‘The meeting has begun.’ All along I believe of noon that it is when themeeting begins. But I believe it in different ways. And to these different ways ofbelieving the same thing, different actions are appropriate: preparation,movement, apology. Of course, if the change of context is not noted, theadjustment of belief states will not occur, and a wholesale change from believingtruly to believing falsely may occur. This is what happened to Rip van Winkle.He awakes in the same belief states he fell asleep in twenty years earlier, unad-justed to the dramatic change in context, and so with a whole new set of beliefs,such as that he is a young man, mostly false.

We have here a metaphysically benign form of limited accessibility. Anyoneat any time can have access to any proposition. But not in any way. Anyone canbelieve of John Perry that he is making a mess. And anyone can be in the beliefstate classified by the sentence ‘I am making a mess.’ But only I can have thatbelief by being in that state.

There is room in this scheme for de dicto propositions, for the characteriza-tion of one’s belief states may include sentences without any indexical element.If there are any, they could appear on the exam. For this part of the structure,the hypothesis of perfect correspondence would be correct.

A more radical proposal would do away with objects of belief entirely. Wewould think of belief as a system of relations of various degrees between personsand other objects. Rather than saying I believed in the de re propositionconsisting of me and the open proposition, x is making a mess, we would saythat I stand in the relation, believing to be making a mess, to myself. There aremany ways to stand in this relation to myself, that is, a variety of belief states Imight be in. And these would be classified by sentences with indexicals. On thisview, de dicto belief, already demoted from its central place in the philosophy ofbelief, might be seen asmerely an illusion, engendered by the implicit nature ofmuch indexicality.

To say that belief states must be distinguished from objects of belief, cannotbe individuated in terms of them, and are what is crucial for the explanation of

Kaplan 1989b.)


action, is not to give a full-fledged account of belief, or even a sketchy one.Similarly, to say that we must distinguish the object seen from the state of theseeing subject, and that the latter is crucial for the explanation of action guidedby vision, is not to offer a full-fledged account of vision. But just as the argu-ments from illusion and perceptual relativity teach us that no philosophy ofperception can be plausible that is not cognizant of this last distinction, theproblem of the essential indexical should teach us that no philosophy of beliefcan be plausible that does not take account of the first.6

6. Versions of this paper were read at philosophy department colloquia at UCLA, Claremont GraduateSchool, and Stanford University, to the Washington State University at Bellingham PhilosophyConference, and to the Meeting of Alberta Philosophy Departments. I am indebted to philosophersparticipating in these colloquia for many helpful criticisms and comments. I owe a special debt toMichael Bratman and Dagfinn Føllesdal for detailed comments on the penultimate version. Most of theideas in this paper were developed while I held a fellowship from the Guggenheim Foundation and wason sabbatical leave from Stanford University, and I thank both for their support.

Section III

Recent Work

Chapter 8

The myth of mental indexicals*

Ruth Garrett MillikanUniversity of Connecticut

Thesis: So-called ‘essential indexicals’ in thought are indeed essential, but theyare not indexical. It is not their semantics that distinguishes them but theirfunction, their psychological role.

A strong contemporary current runs to the effect that the ability of an agent toproject knowledge of the world into relevant action in the world depends on theability to think indexical thoughts. For example, if I wish to get to Boston, itmay be helpful to know that the 8:25 train goes there. But I cannot put thisknowledge to use unless I also come to know, at some point, that there [a placeindexed via perception] is the 8:25 train. Similarly, should my life be endan-gered by an approaching bear, it might help me to know it. But it will not beenough for me to know of this danger to me under some impersonal descrip-tion of me, such as ‘the person sitting in Bruno’s favorite berry patch’ or evenunder the name ‘Ruth Millikan,’ unless I further know that I am the person inBruno’s favorite berry patch or that I am Ruth Millikan (I might not know, forexample, should I be amnesiac). But this kind of thought — there is the 8:25; Iam Ruth Millikan — is, it is supposed, indexical. Thus Dennett, summing upthe literature, remarks, “lndexicality of sentences appears to be the linguisticcounterpart of that relativity to a subjective point of view that is a hallmark ofmental states” (Dennett 1987:132).1 He clarifies, using a (ubiquitous) quota-

* With the kind permission of the editors of Nous, this is a revised and expanded version of 1990: ‘TheMyth of the Essential Indexical’ Nous 24.5: 723–734. The substantive additions are mainly toward theend of the essay.

1. Dennett cites Castañeda 1966, this volume; 1967; 1968; Perry 1977; 1979, this volume; Kaplan 1989b;and Lewis 1979. Another clear example is McGinn 1983. There are also clear gestures toward such athesis, alas, in my 1984.

164 Ruth Garrett Millikan

tion in which Perry remarks, “When you and I have beliefs under the commoncharacter of ‘A bear is about to attack me’ we behave similarly, … [whereas]when you and I both apprehend that I am about to be attacked by a bear webehave differently” (Perry 1977:494). That is, our behaviors hinge not somuchon the objects of our thoughts, on their propositional contents, as on whatKaplan calls the ‘character,’ in this case the indexical type, of our thoughts.Kaplan too identifies “the context sensitivity of character” with what he calls“the context sensitivity of mental states” and remarks, “Dare I call it egoorientation?” (Kaplan 1989b:531).

No, he should not dare. For it is not indexical thoughts that serve to orientan agent in his world. A picture that holds us captive portrays the index as apointing finger, showing the direction of its referent from here, so that we mayact from here regarding it. Internalized, the pointing finger is a pointingthought, guiding action towards its object. But, I will argue, first, it is not truefor the general case that the relation an indexical or the interpreter of anindexical bears to the indexical’s referent is a relation that needs to be takenaccount of during action. Second, conversely, it is not true for the general casethat those relations of self to world that one must take into account in order toact in the world are relations of the sort that an indexical or the interpreter of anindexical bears to the indexical’s referent. Third, it is no part of the job of anindexical token to signify the relation either of itself to its referent or of itsinterpreter to its referent. Fourth, conversely, inner signs that do signifyrelations between agent and world as needed for action are not as such index-ical. Finally, if an agent employs a mental term to represent herself, this inprinciple cannot be a mental indexical: there can be no thought that has the(Kaplan-style) character of ‘I.’ Nor are there thoughts with the character ofeither ‘here’ or ‘now.’ It is not just that for indexicals there is no simple Fregeancorrespondence between possible sentence meanings and possible thoughttypes. The whole genre of indexicals is simply missing from thought.

An indexical sign has no constant referent, no referent qua sign type.Tokens of an indexical type have referents when they are situated in appropriatecontexts. An appropriate context contains something bearing a designatedrelation to the indexical token, which something is thereby that token’s referent(e.g., a person, an object) or is thereby that variant in world affairs that thetoken indexes (e.g., a time, a place, a property). This designated relation for a

The myth of mental indexicals 165

given indexical type I call its indexical adapting relation.2 The indexical adapt-ing relation for ‘I,’ for example, is being the producer of the token; for ‘you’ itis being the addressee of the sentence containing the token; for ‘here,’ being aposition near the origination point of the token; for demonstratives, beingsuitably related in any of various conventional ways to, say, a gesture accompa-nying the token, to other words bearing certain relations to the token; and soforth.3 Thus themeaning of an indexical type can be thought of as expressed bya function from token context to token referent. Kaplan (1989b) calls thisfunction (or close enough) the ‘character’ of the indexical type. It is a thought’s‘character’ in this sense — I shall say ‘karacter’ — that is taken by Kaplan andmany others to connect directly with action, with behavior. “We use themanner of presentation, the character, to individuate psychological states, inexplaining and predicting action” (Kaplan 1989b:532).

First let us examine the relation of the referent of an indexical token to thetoken’s interpreter: is this relation relevant to action? If there were such thingsas mental indexicals, the mental indexical token would be inside the interpreter.Let us begin instead with the easier case of public-language indexicals, wherethe full structure of the relation of interpreter to indexed referent is out in theopen. For public-language indexicals, it is evident that there are actually tworelations to be considered. First, there is the relation the indexical token bearsto its referent as dictated by the karacter of the sign: this is the ‘indexicaladapting relation’ for the sign. Second, there is the relation the interpreter bearsto the indexical token. Different interpreters may, of course, bear quite differentrelations to the same indexical token, hence to its referent. Should a publicindexical serve to alert or accommodate its interpreter to the relation of itsreferent to the interpreter, it is clear that the interpreter would have to sum twoprior relations to find this relation: the interpreter’s relation to the token plusthe token’s relation to the referent. The same structure is still there, though lessevidently, when the relation of interpreter to sign remains constant, the signremaining inside the interpreter.

The first thing to notice is that this pair of relations does not as such ornecessarily yield a sum relevant to action. I’ll give two examples of failure tosum in a relevant way. These should be enough to make the general point. The

2. The reason for this terminology is explained inmy 1984: chs. 2 and 10. Notice also that I amnot using‘context’ quite in Kaplan’s recommended way.

3. For a discussion of the various kinds of indexicals and their adapting relations, see my 1984: ch. 10.


first occurs when the indexical adapting relation is being a (certain sort of) causeof the indexical token. The second occurs when the referent or variant indexedby the indexical token is a type or kind rather than a particular.

Suppose that you receive an undated postcard from Barcelona signed byAlvin that says, “I am leaving for a few days in Rome.” You know what thereferent’s, Alvin’s, relation to the indexical token ‘I’ is: the referent wrote it, hewas its cause; that’s what makes him the referent. And you know what yourrelation to that token of ‘I’ is: you have it in your hand. But this yields no clueconcerning your relation to the referent, to Alvin. At least, it yields no clueconcerning any salient relation, any relation you are likely to need to takeaccount of in order to act regarding Alvin. The given relation of you to thetoken plus the given relation of the token to its referent has, as it were, novector sum.

Nor does it help to move the sign that indexes its cause to the inside of theinterpreter. Suppose that it were true that your thought-tokens ‘Iris Murdoch’were indexical tokens, referring (as do tokens of public language ‘I’) to theirsalient cause, which was in this case the cause, Murdoch herself, of the firstancestor token of ‘Iris Murdoch’ produced at Murdoch’s baptismal ceremony.(I am not recommending this theory of thoughts of Murdoch.) Thinking this(supposedly) indexical thought, even if it involved understanding the nature ofthis indexical adapting relation between thought and referent exactly, wouldnot reveal to you any salient relation you presently bore to Murdoch. It wouldnot help you to take action towards Murdoch.

Taking what is perhaps a more plausible case, consider the popular theorythat percepts are mental indexicals referring to their salient causes. The fact thatthe perceiver contains the percept plus the fact that the percept was caused bythe perceived, by the referent, does not sum to a determinate usable relationbetween perceiver and perceived. Think of seeing an object through a set oftrick mirrors. You perceive the object alright, but you perceive it as in adifferent spatial relation to you than it in fact bears. The bare fact that the objectperceived equals what causes your percept does notmean you can locate it, thatyou grasp its relation to you as needed for action. It is not true in general thenthat the indexical adapting relation for an indexical sign is one that it helps totake into account when engaging in action toward the referents of its tokens.Indexicality as such seems to have nothing to do with orientation for action.

On the other hand, a veridical percept generally does show some relation orrelations between the perceiver and the perceived that it might be necessary oruseful for the perceiver to take into account during action toward the perceived.


Specifically, there are usually (mathematical-style) transformations of thepercept that would correspond systematically to transformations of the spatialrelation of perceiver to perceived. For example, one sees how far and in whatdirection one is from an object. But these relations are not determined byadding the fact that the percept is inside the perceiver to the fact that theperceived is a cause of the percept; it is not a resultant of those facts. The spatialrelation is independently shown in the percept. Because percepts often do showcertain relations to the self of the objects perceived and because you often doneed to take those relations into account in order to act, perception is oftenessential for action. But this has nothing to do with indexicality in perception.Possibly the percept that shows relations is also indexical (I think not — seenote 8 below, and also my 1997) but showing relations would not make itindexical nor would being indexical make it represent any relations. (Relationsshown in my percepts are, of course, relations to me. Soon I will raise thequestion whether, in order to show a relation to me, the percept must index meor my place, but that is a separate question.)

The irrelevance of indexical adapting relations to action shows up especiallyclearly when what is indexed is a type or kind: that color, that word type, thatspecies, that metal, and so forth. Similarly, if quotation marks are indexicals(Davidson 1979, My 1984: ch. 13) or if intentional contexts (‘believes that….’‘wishes to…,’ etc.) are indexicals (Davidson 1968, My 1984: ch. 13, Boër andLycan 1986), these must index types rather than tokens. Conceivably, in thesecases the indexical token brings the interpreter into some sort of non-vector-sum relation to the type that is indexed. But how would a grasp of this relationhelp the interpreter to act in relation to the indexed type?

There does not seem to be anything about indexical adapting relations perse that makes them especially relevant to action. Conversely, the sorts ofrelations between self and world that an agent must take into account are notrelations that adapt any indexicals. To act, I must, of course, take account of thenature and disposition of things in my world relative to my powers of action.The example we all think of first is taking account of my spatial relations to thethings I would act on, for I must act on them frommy place in relation to them.(Notice that this is contingent: with the power of telekinesis, I might not needto take account of these relations.) An important way of knowing the places ofthings to be acted on relative to me is through perception. But if the perceptualrepresentations of these spatial relations are easy to confuse with indexicals,most action-relevant relations that one must take account of certainly are not.Consider, for example, my grasp of the size and weight of things I would act on


relative to my size and strength, of various distances relative to my reachingpowers, climbing powers, leaping, throwing, walking, running, and shoutingpowers, and so forth. Nor is it true that all action is based on perception. I neednot perceive my arm in order intentionally to raise it, or perceive my eyelid inorder intentionally to blink it, and if I should come to know that the triggerreleasing the catch to the door of my jail cell is directly under my left indexfinger (say, the kindly guard tells me), I don’t need to perceive anything at all inorder to act so as to free myself. I need only know how to depress my left indexfinger. Indexicality certainly is not ‘the linguistic counterpart of that relativityto a subjective point of view that is a hallmark of mental states,’ then— at leastnot in so far as the subjective point of view is the point of view needed for action.

It is not indexicals that orient me in my world for action. This is apparentalso from the fact that the indexical adapting relation for an indexical is not arelation that is expressed or shown by the indexical. Suppose that the relationthat an indexical bears to its referent should turn out to be relevant to action.Still, it is not the job of an indexical to tell of or to display this relation betweenitself and its referent.

The karacter of the indexical does not correspond to any part of its proposi-tional content. Instead, to interpret an indexical, one must have prior knowl-edge of which item it is that bears the relevant adapting relation to the indexicaltoken. Onemust know this independently and prior to successfully interpretingthe indexical. One must already know both that this referent exists and that itis appropriately related to the indexical token. One does not find this out byinterpreting the indexical; one needs already to know it in order to interpret theindexical. For example, a token of ‘I’ does not tell me who the originator of thattoken is, that it is, say, Alvin, nor even that it has an originator and that this isrelevant. Rather, if I am to understand a token of ‘I,’ I must already know whothe speaker is and that knowing this is relevant. That is why Alvin had to signhis postcard, and why I had to learn English to understand what it said. Similar-ly, a token of ‘here’ does not tell me where it is. To understand ‘here,’ I mustindependently know what place the token is in, or was in when originated. A‘here’ shouted in the dark is of no use to a person with one deaf ear who cannotlocalize sounds.

Similarly, turning an example of Perry’s on its side, suppose that a postcardarrives with illegible postmark, return address, and signature saying, “I amhaving a good time now.” Perry says that the “truth conditions” of this inscrip-tion are, merely, that “the person who wrote the postcard was having a goodtime at the time he or she wrote it” (1988:9). But the ‘truth conditions,’


understood this way, express only the reader’s knowledge of the karacter, ratherthan expressing any of the propositional content, of the inscription. By analogyPerry should say that the sentence ‘John has brown hair’ has as ‘truth con-ditions’ that whatever person named ‘John’ it is being used to denote has brownhair. But this is to confuse knowing how to find out what it means withknowing how to find out whether it is true. It is to confuse pre-semantics withsemantics. It is not part of the job of a sentence to tell you what does or wouldmake it meaningful. The content of ‘I am having a good time’ concerns only itsactual writer and time and place of writing. Clearly, in order to get to thatcontent it would be necessary to know who wrote it, when, and where. If theinterpreter lacks this knowledge, none of the intended message gets through.Nor, of course, does the message contain that information.

Exactly the same principle applies to the most paradigmatic of indexicals:‘that’ accompanied by a pointing finger. The pointing finger is understood onlyif what it points at is visible or otherwise independently identifiable. Or supposeit is the job of ‘that’ to point out a direction, ‘that way.’ The interpreter musthave a clear view of the surroundings so as to see in what direction the fingerpoints. The interpreter must be able independently to identify that direction,not necessarily with a name (‘east, west’) but, say, via an ability to track it, toknow what it would be to continue following that selfsame direction, asopposed to turning away from it. To know what an indexical points to, toidentify the indexed, requires that one have a prior route to thinking of thatobject, a route other than via the indexical token, and that one grasp this priorroute as arriving at the same object as bears the (priorly known) indexicaladapting relation to the indexical token. Indexicals do not tell what they pointat. It is their interpreters that do the telling. Indexicals do not tell what is intheir contexts. The context of an indexical is what determines its content.

Nor should we be confused by the fact that it is often possible to use a signto obtain information that it is not the function of the sign to convey. Forexample, you can use any public language sign as evidence that there existed aperson, who spoke a certain language, at its point of origin— like footprints inthe sand. Similarly, you could use Perry’s partly illegible postcard as evidencethat there existed a person who wrote the postcard and who was having a goodtime at that time. You could reach this conclusion, as Perry has suggested, bymaking the assumption that the sentence on the postcard is true. If the postcardhad said, ‘I will meet … in Rome,’ the blank filled in with an illegible name, onthe same assumption you could infer that someone had at some time planned tomeet someone else in Rome, and so forth. But it was not the purpose of the


postcard to convey this general proposition. Its purpose was to tell about, say,Alvin. Similarly, from a pointing finger accompanied by the sentence ‘This is acarpenter ant’, you may gather that close to the end of the finger is a carpenterant, even though you cannot see it from there. But that is not what the speakeraims to impart. The speaker intends you to see what is a carpenter ant.

Nor should we be confused by the fact that sentences whose public mean-ings are indexical can also intentionally be turned to nonindexical purposes byindividual users. Consider an anonymous threat over the telephone, ‘I’ll see thatyou die,’ or the childwho says, ‘This is what you are getting for Christmas,’ whilecoyly holding it behind her back. These are indexical sentence types, but they arenot serving indexical purposes. They are not functioning in a normal way.4

All of these uses of language are possible. But what defines the indexical useof a sign is that its context is used by the interpreter to determine the content, todetermine the referent, not talked about in the content. A representation thattold of its own relation to something else would not be indexical but self-referential, and it would be its content, not its karacter, that told of the rela-tion.5 Similarly, an inner representation that told or showed the relation ofitself to the world would not on that account be indexical.

Still, wouldn’t a representation telling of something’s relation to me thatwas crucial for action, say the spatial relation of something to me, have torepresent me, and wouldn’t any representation of me used in this way have tobe indexical? Colin McGinn says, “All the [essential] indexicals are linked withI, and the I mode of presentation is subjective in character because it comprisesthe special perspective a person has on himself. Very roughly, we can say that tothink of something indexically is to think of it in relation to me, as I ampresented to myself in self-consciousness” (1983:17). It will not be enough, asubstantial literature agrees, that an agent entertain representations the contentof which concerns the relation of herself to the world. That might be done bythe use of relation terms along with an innerMillian name that the agent has forherself, or alongwith any description that happens to catch her uniquely.6Whatis required is that the agent recognize any such nameor description as a name for

4. That is, they are not serving their stabilizing functions. See my 1984: chs. 3 and 4.

5. I accept Kaplan’s remarks on Reichenbach’s confusion of indexicality with self-reference as definitive,a confusion embodied in Reichenbach’s term ‘token reflexive.’ See Kaplan 1989b:519–520.

6. By a ‘Millian name’ I mean one about the semantics of which nothing can be said beyond that it is aname with such and such a referent. The semantics of mental names of this sort, their psychologicalpossibility, and how they get their referents are discussed in my 1984; 1993, ch. 4; 1994 and especially1998a, 1998b.


herself, that she identify its contentwith that of her inner term ‘I.’ Onlywhen shegrasps that the person so positioned in the world is ‘I’ can she act from a knowl-edge of that position. And this grasp requires thinking an indexical thought.

Now it is trivial that if I am to react in a special and different way to theknowledge that I, RM, am positioned so in the world, a way quite unlike how Iwould react knowing anyone else was positioned so in the world, then my innerway of representing RM must bear a very special and unique relation to mydispositions to act. But what does that have to do with indexicality? My inner wayof representing RM is obviously not just an ordinary name in my mentalvocabulary. It hooks up with my know-hows, with my abilities and dispositionsto act, in a rather special way. Conceivably, I might also have ordinary mentalnames for RM, or mental descriptions, that didn’t hook up with these know-hows, because I didn’t recognize them as having the same content as this specialRM representation, just as I might think ‘Cicero’ and then ‘Tully’ withoutknowing these were thoughts of the same person. My way of representing RMis indeed special. Let us call it ‘RM’s active self-thought’ or ‘@RM,’ for itrepresents a person whom I know, as thought of that way, how to manipulatedirectly; I know how to effect her behavior. But in order to know how to managethis person, why would I need to think indexical thoughts?What has know-howto do with indexicality?7

An indexical term is one whose referent varies with context, being identi-fied, for each token, by the fact that it is what bears a certain relation, theindexical’s adapting relation, to the token. Applying this principle to indexicalsin thought, a thought would be indexical if its context determined its referent,and if there were normal procedures for identifying this referent, that is, fordetermining with what prior or independent thought tokens it coincided incontent, procedures depending on the fact that the referent bore the indexicaladapting relation to the token. That is, these procedures would work onlybecause the referent bore the adapting relation to the token. Determining for athought token with what other thought tokens it coincides in content isdetermining, paradigmatically, which other term tokens it can be paired with toserve jointly as a middle term during inference. (Determining this correctly isbest thought of as an ability or know-how rather than as knowledge that.) Toillustrate, if there were such things as Millian names in a language of thought,

7. I give accounts of abilities or ‘know-hows,’ calling them ‘competencies,’ in my 1993: ch. 11; 1994.Abilities express biological purposes and, as such, are very different from mere causal dispositions.


an obvious procedure for identifying the referent of a mental Millian nametoken would be to pair it with other tokens of the same mental type. But theprocedure for pairing a mental indexical token with other tokens having thesame referent would have to be routed via the context of the token— as I knowwho is tired when Alvin says ‘I am tired’ through an independent identificationof the person who has produced this token of ‘I.’ Otherwise, the mental termwould not be functioning indexically.8

Is that the sort of way that my mental term ‘@RM,’ the term that bears thatquite unique and special relation to my dispositions to act, hooks up with itsreferent, hooks up with me? Do I succeed in identifying the content of varioustokens of my mental ‘@RM,’ that is, do I succeed in reidentifying myself, onlybecause I grasp for each token of ‘@RM’ independently that it bears a certain

8. In the earlier version of this essay (‘TheMyth of the Essential Indexical’), I offered the following as anexample of indexicality in perception, but I now think that I was mistaken:

It is plausible that in perception the percept is about, refers to, its cause, that is, to the causeof the percept token. What the (veridical) percept token shows is certain properties of thiscause plus, often, the spatial relation of that cause to the perceiving subject. But the perceptis not about the generality that there exists a something of a certain character so related to theperceiver; it does not, as it were, translate with an existential quantifier. Rather, it is about, itis a percept of, its particular cause. This particular aboutness is expressed through the abilitythat the normal perceiver has to track the particular referent with eyes, head, and, if necessary,feet in order to accumulate more information about it. This process involves identifying or,what is the same, reidentifying the tracked object, for it involves using a series of percepts ofit, of the same thing, conjointly (compare the function of a middle term) so as to extractinformation presumed to be about just one thing, about one particular individual. And themethod of determining that these various percepts belong together as percepts of the same—the method of tracking — is routed through the fact that the perceived was the cause of thepercept. It was the cause of the percept in accordance with a certain way of causing normal forthat kind of perception, and it will accordingly cause later percepts in a traceable pattern,other percepts with the same referent.

I no longer think that the cause of a percept functions qua cause to determine any part of its intentionalcontent. The confusion arises because the only verbs of perception that we have are ‘success’ or‘achievement’ verbs, such as ‘see,’ ‘hear’ and ‘perceive,’ parallel to ‘remember,’ ‘know’ and ‘realize’ in therealm of conception. Just as you do not ‘remember’ it if it did not happen, you do not ‘see’ or ‘perceive’it if it did not cause your perception. In the realm of conception, however, we have non-success verbslike ‘believe’ and ‘think,’ making us easily aware that the intentional object of conception is not alwaysits actual source. For example, the object of beliefmay not bewhat is actually being remembered. Imightbelieve that Aunt Nellie once took me to the movies yet actually be remembering Aunt Alice’s doing so.Suppose we invent a verb ‘to visage’ for the perceptual realm to parallel the conceptual verb ‘to believe.’Then we have a way of saying that I may sometimes visage things that are not the actual causes of myperceptions. I might visage Aunt Nellie in the distance when it is actually Aunt Alice I am seeing.Compatibly, just as we can say ‘I remembered it wrongly,’ we can say ‘I saw it wrongly.’ The intentionalobject of a perception is what is visaged, and this is not always what actually causes the perception. Thepercept is not an indexical representation of its cause.


adapting relation to me? Isn’t it more reasonable that my mental ‘@RM’ issimply a mental proper name? I take different tokens of ‘@RM’ to refer to thesame not because of their individual contexts, not as a result of some relationeach of these tokens independently bears to me, but simply because they aretokens of the same type.

If the thought ‘@RM’ were indexical in my system ofmental representation,then its referent would have to be identified via its context. Correlatively, itsreferent would have to shift in accordance with context. And what sort ofcontext would that be?

Perhaps we are supposing the relevant context to be the mind ‘@RM’appears in. (Devitt: ‘The reference of “I” is determined by the head it is in’[1984:400].) Are we supposing, then, that in my language of thought, in myinner system of representation, tokens of ‘@RM’ might appear in your head sothat I must check whose head ‘@RM’ appears in before identifying its content?Or are we supposing, perhaps, that my mental language is some sort of univer-sal language, one selfsame language that all people speak in their heads, so thatrather than ‘@RM,’ I must think ‘I,’ the self-name in universal Mentalese? Buteven if this were the case (maybe Jerry Fodor thinks that it is), in what sensewould the self name be indexical? Certainly there would be no interpreter forwhom it would be indexical. Or is the claim that it would be indexical for God,or for an intrusive mind or brain reader?

But the language of thought, if there is such, is not God’s language, norbrain-reader language, but the thinker’s language. Godmight read tokens of theuniversal self name, tokens of mental ‘I,’ indexically, determining the referenceof each by first noting whose head it was in. Similarly, I might ‘read a chame-leon’s back’ descriptively, as a natural sign telling what color the chameleon hasbeen sitting on, although the chameleon’s color has no descriptive meaning forthe chameleon. The universal self name would not be an indexical for the selveswho named themselves with it, and when read by someone else, it couldfunction only as a natural sign, not as a sign in the language of thought.

So my mental ‘I,’ my ‘@RM,’ is not an indexical. More reasonable (though,I will soon argue, probably still incorrect) would be to take it as a (Millian)name for me; your ‘I,’ which may well have quite a different mental shape, as a(Millian) name for you. But supposing this to be so, there is still the questionwhy using the public indexical ‘I’ seems to express this mental name in a waythat using one’s name or a description of oneself does not. If I say to you ‘I wasborn in Philadelphia’, I express to you that the very agent whose presence youare now in was born in Philadelphia, whereas if I say ‘Ruth Millikan was born


in Philadelphia’, you will not understand that it concerns the person presentunless you happen to know my name. And it also seems true that I might say‘RuthMillikan was born in Philadelphia’ without knowing I was RuthMillikan,whereas saying ‘I was born in Philadelphia’ shows that I believe of the verybearer of my active self name,‘@RM,’ that she was born in Philadelphia. And yetthis is not quite accurate. What is true is that for me to say anything at all is forme literally to put words in the mouth of this body, so that if we grant that Iunderstand English and also understand which body it is I control, we must alsogrant that I realize that any ‘I’ that I intentionally produce will refer to thecontroller of this body. Is it possible to be deluded about which body I controlso that I might say ‘I was born in Philadelphia’ expecting the words to emergeout of someone else’s mouth? If not, that would be an empirical fact about theimpossibility of certain kinds of neurological damage or disturbance. Thepsychological literature shows that a great manymental disturbances that seeminconceivable in fact are occasionally realized.

A more interesting question is how to understand that a name for myselfmight be ‘active.’ A way to begin, I believe, is with the notion of a ‘pushmi-pullyu representation’ or ‘PPR’ (Millikan 1996). A PPR is a representation thatis fact stating and directive at the same time, or better, is undifferentiated asbetween these two modes. The simplest examples are found in nature. Thebeaver’s tail splash tells that there is danger and also tells other beavers to diveunder. The dance of the honey bee tells where there is nectar and also tells otherbees where to go. I have argued that human intentions are inner PPRs, repre-senting future facts about oneself that may need to be considered in furtherplanning and at the same time guiding action towards realization of those veryfacts (1996). One’s active self name, then, is the PPR name for oneself thatoccurs in the inner representations that are one’s intentions.

But this is still not quite right, as I think. I suspect that the self is notroutinely represented at all either in one’s expectation about one’s future or inone’s intentions to act. Notice that in soliloquy there is no explicit reference tothe self: ‘To be, or not to be;’ ‘And now to bed.’ Also, when A hands to B a formthat is to direct B’s intentions, no explicit reference is made to B’s self: ‘Closethe door, please;’ ‘Be quick now.’ Similarly, when I see or otherwise perceive thespatial relation of an object to myself, often I need not perceive any portion ofmyself in order to act with regard to it. For example, to walk towards the churchat the end of the square, I need not perceive my legs or any other part of myself.And yet, I have said, my spatial relations to other things are an important partof what is represented in perception. How can this be?


Once again, bee dances (Of course! What else?) are the key. What a beedance shows is the direction, relative to hive and sun, in which there is nectar.But there are no variables in the bee dance that show the hive, or the sun, orthat it is nectar that is being represented. For example, there is no aspect of thebee dance that, if varied or replaced, would show the relation of nectar to hiveand moon, or the relation of danger to hive and sun. Similarly, the visualpercept shows the spatial disposition of other objects relative to where I am, butthere is no variable in it that might be replaced to show the spatial dispositionsof these objects relative to any other spatial point of view. I can imagine and Ican conceive from spatial points of view I do not currently occupy, but I cannotperceive from other points of view. That is of the nature of perceptual represen-tation, designed, in the first instance, to guide action.

Similarly, my intentions are not designed to guide anyone’s actions but myown. Hence they have no need explicitly to represent me. I do not have to takeinto account variations in whose head a token of ‘@RM’ appears, nor variationsin whose action it is supposed to guide. But, once again, this inarticulateness inhow the self is represented has nothing to do with indexicality.

But we have not disposed yet of quite all of the myths. What about mentalindexicals corresponding to ‘here’ and ‘now’? Corresponding to ‘here’ arethoughts of things understood as being close to me. Thus ‘here’ does notexpress an indexical thought but merely a thought of an impure relationalproperty. Thoughts of impure relations, whether the relata are explicitlyrepresented or not, are not, as such, indexical. Being more careful, however,what we express with ‘here’ are thoughts of things understood to be close tomenow. Perry tells us, “it is plausible to suppose that other indexicals can beeliminated in favor of ‘I’ and ‘now’….. But elimination of either ‘I’ or ‘now’ infavor of the other seems impossible.” I have tried to show how to eliminate themental indexical ‘I’. Can we also eliminate ‘now’?

Using Perry’s example, if the absent minded professor intends to go to thedepartment meeting and knows that is starts at noon, even though it is in factnoon, the professor may not move towards the meeting for he may not realizethat it is noon now. Whatever is a mental correlate of ‘now,’ like the mentalcorrelate of ‘I,’ appears also to be, as such, an element in PPRs. From ‘I amgoing to the meeting at noon’ coupled with ‘It is now noon’ is derived thethought ‘I am now going to the meeting.’ And like a bee dance, the thought ‘Iam now going to the meeting’ serves both to represent what is happening andto cause it to happen. That accounts for the action-producing characteristic ofthe thought. Considering this thought in so far as it represents to its thinker


both when and who is going, we have also eliminated the indexicality of thepart showing who. We did this by pointing out that the context of whateverpart shows the who is constant for a given thinker. The same thinker does notthink in this way of a variety of who’s. But in whatever way the thinker thinksof when he is going, surely the same thinker does think of many differentwhens. Moreover, which when a thinker is thinking of in this way surelychanges with the time of the thought. Doesn’t it follow that the thought mustbe indexical?

Surprisingly, no. Consider a scale drawing of a building where onecentimeter represents one meter. Compare a scale drawing of a small insectwhere one centimeter represents one millimeter. Now consider a scale drawingof a cross-stitch pattern where one centimeter represents one centimeter.Surely none of these representations contain indexicals. On the last drawing,one centimeter does indeed represents one centimeter, but it does not indexitself, nor it is ‘token reflexive.’ Similarly, children’s marking pens are coloredon the outside to indicate the color on the inside, red on the outside standingfor red on the inside, blue standing for blue, and so forth. Or perhaps thecolors on the inside are a shade lighter than the ones on the outside. Anotherexample is the relative places of dots on a map which show the relative placesof cities on the earth’s surface, geometrical relations on the map indicating thesame geometrical relations on the earth. This kind of representational systembears a strong resemblance to systems exhibiting compositionality, possibletransformations (mathematical sense) of the representation correspondingsystematically to possible transformations of the represented. One centimeterlonger corresponds to one meter longer, or to one millimeter longer, or to onecentimeter longer; one shade redder corresponds to one shade redder, twice asfar from represents twice as far from, three dots in an isosceles trianglerepresents three cities in an isosceles triangle, and so forth. Surely there is nohint of indexicality here.

Now consider the beaver’s danger signal. It is not indexical, but the placeof the splash represents the place of the danger and the time of the splashrepresents the time of the danger. Similarly, the absent minded professor’sthought that now is the time of the meeting represents the time of the meetingwith the time of the thought. It is not indexical. The professor need not haveany independent hold on the relation of the thought to the time of the meeting— as one would need an independent hold on who it was that wrote ‘I amhaving a good time’— in order fully to appreciate his thought that the meetingis now. If he has mistaken the time and, having discovered this, sometime later


again thinks that the meeting is now, he does not think the same indexicalthought type in another context. He thinks a different thought altogether.Similarly, if the cartographer corrects himself by moving the dot for Chicagoto a different location, he does not use the same indexical representation typein a different context. He uses a different representation altogether though, ofcourse, one from the same representational system.

Chapter 9

Thinking about myself

Maite EzcurdiaUniversidad Nacional Autónoma de México

Frege held that upon having a self-thought — a thought or belief expressed inpublic language by a subject with ‘I’ — a subject is presented to herself in aparticular and primitive way in which she is presented to no other.1 Taking‘mode of presentation’ loosely, we may include under Frege’s suggestion otherviews. All of these accept, quite generally and roughly, that when one has a self-thought2 one thinks of oneself as one’s self, that is, as a subject which is one’sown, which — so to speak — one knows how to move.3 Disagreements ariseon the issue of what this mode of presentation amounts to. Some hold that it isa particular sense (Frege, 1918) or concept given by a certain mode of identifi-cation (Evans, 1982, this volume; Peacocke, 1983), whereas others think it is away of believing or role (Perry, 1979; this volume) or a non-communicablemode of presentation (Récanati, 1993). I shall group all these accounts togetherunder the label the mode-of-presentation view.

Recently two further proposals about what is involved in having an‘I’-thought have emerged: they deny an ‘I’-thought involves anything like amode of presentation of thinking of oneself as one’s self. The first proposal is

1. See Frege, 1918.

2. I am taking ‘thoughts’ here in a non-Fregean way, hence not to mean Fregean propositions. I take‘thoughts’ to refer to psychological states, like believing, doubting, etc.

3. On the Fregean approach one might take this as a primitive mode of presentation, but one may alsowant to reduce this mode of thinking to other modes of thinking like thinking of oneself as this person.(For arguments against some candidates that reduce it see Burge, 1998.) Although I think that the self-mode-of-presentation is non-reducible, I shall not argue for this here.

The phrase ‘the subject one knows how to move’ (which I use throughout) is meant to include theknowledge that a subject has of herself as a subject not only of movement or action, but also of thoughtand perception. Such knowledge may be characterised as a know-how and as a knowing what it is like.I use the phrase for want of an adequate short expression that involves action, thought and perception.

180 Maite Ezcurdia

the no-mental-representation view and is held primarily by Mellor (1988).4 Theview is that upon having a self-thought a subject does not have mode ofpresentation ormental representation of herself; all she has is an object, namely,herself. The second proposal is the Millian view and has been put forth byMillikan.5 The view is that although on having a self-thought one does have amental representation of oneself, there is no mode of presentation mediatingbetween that representation and what it represents. The view is essentially thatthere are different mental counterparts for ‘I’, each behaving as a Millian namefor the subject in whose head each occurs.6

I think these more recent views are mistaken, and that the mode-of-presentation view is essentially right. Self-thoughts do require a subject to thinkof herself under a particular mode of presentation, viz., as her own self. Inpaving the way for an argument in favour of this, I shall argue against the otherviews. I begin by saying what a theory of self-thoughts must account for inorder to show how it is that both the no-representation and the Millian viewsfail to do this. My conclusion will be threefold: that the mode-of-presentationline of thought is essentially correct, that the mental representation which is acounterpart for ‘I’ is very likely indexical like the public language ‘I’, and thatthe mode of presentation of the mental counterpart for ‘I’ need not be the sameas the one involved in the semantic rule or character of the public language ‘I’.

I

There are two features of self-thoughts that distinguish them from what I shallcall ‘name-thoughts’, viz., thoughts reported in public language with the use ofnames for the subject.

4. Perry (1986) holds this view for what he calls “basic beliefs”. His view is that in the case of basic beliefscontext suffices for determining or fixing the object of belief, so that no mental representation of thatobject is necessary. The argument I shall be presenting against Mellor will also count against Perry’saccount of basic beliefs, though I shall not spell this out here.

5. See Millikan, 1990, revised version this volume. The view discussed here is from the 1990 version.Given Fodor’s recent general account of modes of presentation as mental representations —Mentaleseexpressions — whose semantics is given solely in terms of reference, he might be interpreted as anadvocate of this view. (For some hints towards this see Fodor, 1994, and for the full view see Fodor, 1998.But contrast it with Fodor, 1987.)

6. There is a sense in which both these views agree with Frege’s claim that upon having thoughts aboutmyself I am thinking of myself in a primitive way in which no one else can think of me. On the no-representationview, theonlypersonwhodoesnotneed tohave a representation inorder to thinkaboutmeis me, whereas others do need representations of me. On theMillian view, the only subject who can thinkaboutme through aMillianmental representation, through an active self-name (asMillikan calls it), isme.

Thinking about myself 181

a.�The self-knowledge feature�: self-thoughts may suffice for knowledge aboutoneself whereas name-thoughts do not.7

b.�The self-locating feature�: self-thoughts, in particular, self-beliefs, are neces-sary for a subject to be motivated into action because they locate the subject forthe subject whereas name-thoughts do not.

a.�In attributing to myself the property of thinking, I may think of (or refer to)myself in two sorts of ways either by thinking (1) or by thinking (2).

(1) Maite is thinking.8

(2) I am thinking.

Both are attributions to myself of a certain property, but only my believing (2)suffices for knowledge. Only my same act of thinking (2) suffices as justificationor warrant for my knowing it to be true. This is not so with my thinking (1).For I may have forgotten that I am Maite. Thus, my thinking (1) does notsuffice as justification for my knowing that it is true — though it does sufficefor making it true. This is because whereas in my having a name-thought I canbelieve the object of thought is someone different from me, my having a self-thought ensures that I recognize or realize which is the object of the thought,that I recognize or realize that that thought is about me.9

7. Self-knowledge as discussed by Burge (1988 and 1996) includes attributions of thoughts to ourselveslike knowing that I am thinking that water is liquid. Here I present only attributions of properties (likethat of thinking), and not of full thoughts, which a subject makes to herself. However, the feature of self-knowledge should be extendable to the cases considered by Burge.

8. (1) to (6) can be taken to be either utterances by me or sentences relative to contexts of use in whichI am the producer and which report the contents of my thoughts at the time and place of having them.I shall sometimes speak as if utterances, tokens or occurrences of public language indexical-types arewhat do the referring and as if tokens or occurrences of Mentalese indexical-types do the referring(where occurrences are just indexical-types relative to contexts of use). In the case of public languageindexicals, Kaplan (1977:522–3; 1989b:584) argued that, for the sake of a formal semantic account, weneed occurrences or syntactical types relative to contexts of use to be the bearers of reference. Recently,García-Carpintero (1998) has challenged this view in favour of the view that in the case of publiclanguage indexical utterances or events (and hence tokens) are the bearers of reference. I shall not takea stand here on what the bearers of reference actually are, nor shall I address to what extent Kaplan’s orGarcia-Carpintero’s considerations apply in the case ofMentalese indexicals (if there are any). These aretopics for another paper.

9. Notice that the point here is not that self-thoughts are guaranteed a res whereas name-thoughts arenot. For the distinction I am drawing is not between self-thoughts and any thought whose content isadequately reported in public language with the use of a name, but rather between self-thoughts andmyown name-thoughts. In this case both are guaranteed an object of thought. (See Rovane, 1987, andAnscombe, 1975, for a discussion of the general distinction between thoughts whose content is reportedwith the use of a name and self-thoughts, or more precisely, between utterances containing a name and

182 Maite Ezcurdia

b.�Suppose that I believe the content expressed by (3), a sentence producedtoday at 6:30pm.

(3) Maite must start reading her paper here and now.

Suppose further that I begin reading my paper. Does my believing (3) give mea reason to do so? Since I may not know that I am Maite, (3) alone could notgive me a reason to act. I must have another belief. I must believe in additionthat I amMaite. Thus, a self-thought is needed for my believing (3) to give mea reason to act. Contrast this with my believing the content expressed by (4), asentence uttered by me today at 6:30pm.

(4) I must start reading my paper here and now.

My believing (4), that I must start reading my paper here and now, does notrequire that I know that Maite and I are the same person. My believing thecontent expressed by (4) suffices for my having a reason to act here andnow.10 The reason is that my self-belief locates me in a way in which myname-belief does not locate me, though the contents of both beliefs have thesame truth-conditions.

Any account of self-thoughts must rescue these two features. The challengefor the no-representation and the Millian views is whether they can do so.

It is tempting to think that the locating feature of self-thoughts suffices forshowing that the mode-of-presentation view is essentially right, and that theno-representation and Millian views are wrong. However, this temptationmust be resisted.

Perry’s (1979; this volume) argument for the indispensability of indexicalbeliefs for actions might be mistakenly taken to show this. Perry argues thatbeliefs reported with certain indexicals are essential for intentional actions,because they report beliefs which locate certain objects or variants for thesubject. The beliefs in question locate the subject, the present time and thepresent place for the subject, and are reported in public language with the use

‘I’-utterances.)

10. Strictly speaking my self-belief (4) does not suffice for giving me a reason to act since other mentalstates like desires are necessary. The difference between my believing (3) and (4) can be brought outmore precisely in the following way. There is some set S of mental states (desires, beliefs, etc.) which Ihave such that (i) it alone does not suffice for givingme a reason to begin readingmy paper, (ii) togetherwithmy self-belief (4) it does suffice for giving me a reason to begin readingmy paper, and (iii) togetherwith my belief (3) it does not suffice for giving me a reason to start reading my paper. For my believing(3) and the set S to jointly give me a reason to act, I need the belief that I am Maite.


of ‘I’, ‘now’ and ‘here’.11 According to Perry, the fact that such beliefs arenecessary for having a reason to act depends not on what they are beliefs about,but rather on the ways in which the objects of beliefs (the subject, time andplace) are believed or presented. For example, in the case of ‘I’-thoughts Perrybelieves that for me to locate myself, I need to locate a subject as me, as thesubject whom I know how tomove. So according to Perry, forme to locateme intheway essential for having a reason to act, it is necessary that I be presentedwithmyself asmyself, as a self which ismy own or as the subject I know how tomove.

Although Perry is right in thinking that ‘I’-beliefs locate me in a way that isnecessary for my having a reason to act and are reported in public languagewith the use of ‘I’, no argument has yet been given for thinking that for thosebeliefs to serve that task theymust involve amode of presentation of the subjectas herself, or as the subject of thought, action and perception whom she herselfknows how to move. Hence, Perry’s argument as it stands cannot be taken torule out the Millian and no-representation views.

II

Advocates of the no-representation view take it that there need not be a mentalrepresentation for ‘I’ because context guarantees the object the thought is about.Context is believed to fix or determine what the thought is about. It is thiscontextual guarantee for the case of self-thoughts which I shall be questioninghere, in particular the contextual guarantee suggested by Mellor.

Mellor believes that when thinking about an object or variant one onlyneeds a mental representation of it as a causal substitute or surrogate for it.When I believe that my neighbour’s car is red, I must have a mental representa-tion of that car for my belief to have the causal powers it has over my othermental states and actions concerning that car. My neighbour’s car is at somespatio-temporal distance from me so it is precluded from causing me to act orform other mental states (beliefs, desires, etc) that concern it. This is why, onMellor’s view, I require a mental representation of that car when I think aboutit. Nevertheless, Mellor believes this does not happen with self-beliefs: I do notneed a mental representation of myself for my self-beliefs to have the causal

11. Not all ‘here’-thoughts are thus indispensable. Utterances in which ‘here’ occurs demonstratively—for example, when pointing to a map in uttering ‘We are here’ — do not express contents of thoughtswhich are thus essential for action.

184 Maite Ezcurdia

powers they have in making me act or have other mental states about me. Thereis, according to Mellor, sufficient spatio-temporal contiguity between a self-belief and its res to guarantee that self-beliefs are able to causally interact withrespect to its res, viz., the subject who has that belief.12 This contiguity he callscausal contiguity.13

Mellor intends causal contiguity to do two things. To fix or determine theres of self-beliefs, and to ensure that such beliefs have the causal powers theyhave insofar as the only person they can cause to act with respect to herself isthe subject for she is both the subject who has that belief and the res of thebelief. If causal contiguity were to do this then we would have an account ofself-thoughts that would vindicate their self-knowledge and self-locatingfeatures. My believing (2) would just entail that my belief is about me and notabout any other subject for I would be the only object my belief would becausally contiguous with. In having no representation for anyone else but justhaving me, I would be compelled to recognize that it was me that the belief wasabout. Hence, it appears that no mental representation is required for mybelieving (2) to constitute self-knowledge. Furthermore, if I directly figure ina self-belief, I could not fail to locate myself upon having a self-belief. Thus,the self-locating feature would also appear to be rescued. However, these arejust appearances.

There are basically two problems with Mellor’s no-representation view. (A)It is incomplete insofar as it falls short of accounting for the difference betweenself-beliefs and name-beliefs in a way that shows that whilst self-beliefs suffice

12. There is another reason why Mellor might want to accept the no-representation view: if there arebeliefs which involve a self mode of presentation then there are subjective beliefs; and given that beliefscause other mental states and/or actions and that for Mellor facts are the causal relata, then there aresubjective facts. But Mellor denies that there are such facts on pain of contradiction. (See Mellor 1981,Chapter 6, and 1988.) Thus he must deny that there are any subjective beliefs, hence any beliefs whichinvolve a self mode of presentation.

13. Putting the matter a bit crudely, Mellor suggests that when I have an ‘I’-belief, I do not need arepresentation of myself. I need only myself: I represent me.

Subjective beliefs […] need no causal surrogates, no internal representations of the agents[…] they refer to. And that’s why they pose no problems of reference: the relation of referencein subjective beliefs is simply that of identity. [Mellor, 1988:90]

This opens up the logical possibility of two views: the no-representation view and another which holdsthat I do have a representation of myself but that representation is just me, the object it represents. Isthere a substantial difference between these? It seems not. The objections that follow will apply to eitherof these, though I shall frame them against the no-representation view only. Just as context will notguarantee that I am the res of my self-beliefs, it will not guarantee that I am the representation involvedin self-beliefs.


for self-knowledge and self-location, name-beliefs don’t. But even when the no-representation theorist could adequately complete his account to bring out thisdifference, he would still be faced with a second insuperable problem. (B)Causal contiguity alone cannot fix or determine that my self-thoughts are aboutme, for there are too many entities (objects or variants — times, places, etc)with which I am causally contiguous.

A. If causal contiguity is what secures the object of a self-belief, then why is itnecessary to have a mental representation when I have a name-belief? Why is itnecessary that I have a mental representation that corresponds to the publiclanguage ‘Maite’ given that there is causal contiguity betweenmy belief and theobject of my belief (i.e., enough spatio-temporal contiguity for me to act uponthe object of my belief)? If (as some assume) the only semantic role of a nameis to refer to the object which it names such that its only contribution at thelevel of thought is its referent, then given that there is spatio-temporal contigu-ity between a name-belief and the object named it should follow that there is noneed for a mental representation that corresponds to the name. Because I canfail to realize that a name-thought is about me, Mellor must assume that I doneed a mental representation when I have a name-thought in contrast with myself-thoughts. But if this is so, thenMellor’s claim that causal contiguity sufficesfor supposing that there is no need for a mental representation of the object ofthought must be wrong. There must be some other reason for supposing thatself-thoughts do not require a representation whilst name-thoughts of oneselfdo. There must be something my self-thoughts do that my name-thoughts donot do or vice versa.

There are two lines that one could adopt here. One could reject the viewthat a name’s sole contribution at the level of thought is its referent or onecould show that the causal contiguity of name-beliefs with their res is differentfrom the one involved in ‘I’-beliefs. One way to distinguish the causal contigu-ity of self-beliefs from that of name-beliefs would be to say that in having‘I’-beliefs I am causally contiguous with myself as myself whereas I am not thuscausally contiguous with myself upon having ‘Maite’-beliefs. But this alreadyappeals to a mode of presentation of myself, hence this response is not availableto Mellor.14 The line he may adopt is the first one: to give an account of the

14. There is a reading of Mellor under which for him what distinguishes self-thoughts from name-thoughts is their functional/causal roles, the way in which they affect behaviour. This difference infunctional roles is not taken by Mellor to be a difference in content: “beliefs are distinguished from oneanother not only by their contents but by how they affect behaviour” (Mellor, 1988:85). Nonetheless,

186 Maite Ezcurdia

role of names such that they require a mental representation or a mode ofpresentation in the subject whichmakes it possible for her not to know that sheis the object of her name-beliefs, so that they fail to locate her for herself. In thiscase Mellor must identify the other semantic or psychological contribution ofnames to thoughts which requires that there be mental representations ascounterparts for them. But until Mellor can distinguish appropriately self-thoughts from name-thoughts, his account remains incomplete.

What emerges from this discussion is that causal contiguity in not enoughto allow us to do without a mental representation. More is needed. Thisbecomesmore evident when we try to distinguish ‘I’-beliefs from ‘now’-beliefs.

B. Given that there is causal contiguity between the present time and a‘now’-belief — that is, between the time at which a subject has a belief and thatbelief which is reported in public language at the time of having it with ‘now’—Mellor takes it that when an individual has a ‘now’ - thought there is no needfor her to have amental representation of the present time. In an analogous wayin which I just figure in my self-thoughts, the time at which I have a‘now’-thought just figures in that thought. But if this is so, how are we todistinguish ‘I’-beliefs from ‘now’-beliefs? How am I to distinguish my beliefabout the present time to the effect that it is C from one about myself to theeffect that I am C? One way would be via the concept C itself. If C is a conceptthat is not satisfiable by times but only by subjects or is satisfiable only by timesbut not by subjects, then C would dictate or fix the intentional object of thebelief. Suppose that I believe that a is in pain, where ‘a�’ is just a place-holder foreither myself or the present time. The concept of being in pain would determinethat my belief is about me and not about the present time, for only subjects canbe in pain (times can’t) and I am the only subject with which I am causallycontiguous. However, this would already entail that I recognize myself in acertain way, that I think of myself, if not as a self, at least as a sentient being.Thus, the no-representation theorist cannot accept this answer. He cannotallow the predicated concept that figures in a belief to be what determines theintentional object of the belief. He needs a different account.

this might still amount to a difference in modes of presentation. The suggestion is that for Mellor thereis a difference between the way I am causally contiguous with myself upon having a self-thought fromthe way I am causally contiguous with myself upon having name-thoughts. But, as said before, thedifference in these two ways of being causally contiguous with myself can be read as there being twodifferent modes of presenting or representing myself, and hence as entailing either a mode of presenta-tion or a Millian account.


Suppose that there is a concept F which could be satisfied by both a subjectand a time, and that a subject S attributes F to herself. Were S not to have amode of presentation or a representation of herself or the present time, thenthere would be no way of telling to what she was attributing F, to the presenttime or to herself. But ex hypothesi there was no indeterminacy in S�’s attribu-tion: S was attributing F to herself. Thus, S must have a way of distinguishingbetween the present time and herself, and that must be either through amentalrepresentation or a mode of presentation or both.

Take the concept being hot. I can believe that I am hot and I can also believethat it is hot now. These beliefs are different. I may well believe that I am hotbecause I have too many clothes on or because I have a temperature withoutthereby believing that it is hot now. I may also believe that it is hot now uponlooking at the weather report without believing that I am hot since I’ve justcome out of the pool. Suppose then that I believe (5).

(5) It is hot now.

Howwould my believing (5) differ frommy believing (6) on the no-representa-tion view?

(6) I am hot.15

The no-representation theorist has no way of explaining whymy belief is abouta time and not about myself. For him, the belief expressed with (5) is hopelesslyindeterminate. The indeterminacy lies between (5) expressing my belief that itis hot now and it expressing my belief that I am hot. But ex hypothesi, there is nosuch indeterminacy. My belief is about a time and not about myself. Insofar asboth I and the present time are causally contiguous with my beliefs, the no-representation theorist is committed to attributing an indeterminacy in bothmy ‘I’ andmy ‘now’-beliefs which is not there. Causal contiguity cannot then bewhat secures, fixes or determines the intentional object of a self-belief, and

15. I intend the predicates in (5) and (6) to express the same concept of being hot and not for one toexpress the concept of being hot whilst the other the concept of feeling hot. My believing (6) may welldiffer from my believing that I feel hot when I have a high temperature and, nonetheless, feel cold. Theno-representation theoristmightwant to argue that the predicates in (5) and (6) cannot express the sameconcept because the concept of being hot, as opposed to that of feeling hot, is inappropriately applied tome for it can only appropriately apply to my body. However, as suggested above, the no-representationtheorist cannot argue from the inappropriateness or appropriateness of a concept for this alreadysuggests that I have a conception of myself, a mode of presentation of myself, as the appropriate orinappropriate object to which the concept may be applied. In this case the mode of presentation inquestion is (an undesirable) one of presenting me as different from my body .

188 Maite Ezcurdia

hence an account of self-thoughts that relies solely on it cannot vindicate theirself-knowledge or self-locating features.

Still the no-representation theorist could rejoin thus: “But you havesupposed that being hot is a monadic concept or property, that is, a concept orproperty that applies to a single entity, when it is not. Rather it is dyadicproperty/relation or concept, in particular it is a concept that applies to certainentities relative to a time. (In public language ‘being hot’ is a two-place predi-cate.) Thus, the belief which you reported with (6) is a belief that you are hotrelative to a time, namely, the time at which you have the belief. Hence, a moreappropriate report of the content of your belief at the time you had it wouldhave been your uttering (6¢).

(6¢) I am hot now.

And the belief which you reported with (5) is really a belief that a place is hotrelative to the time of the belief. The pronoun in (5) should be taken as a nounphrase referring to a place, and not as a pleonastic element.16 Hence, there isno indeterminacy concerning the object of your belief when you believe (5) or(6). In the latter case, your belief is about yourself and the present time, whereasin the former case it is about a place and the present time.”

There are two things to say about this rejoinder. The first is that therejoinder is allowed to say that in order to have a complete thought-content, theconcept being hot requires that there be two things which are related. But whatit is not allowed to do is to restrict which sort of thing is the one that it isrelating, because that would already require the subject who has the belief torecognize that the relata are the right sort of thing, and hence to recognize themas something. It would require me to recognize myself as the right sort of thing,as the sort of thing which can be hot. And this, as we saw before, is somethingthat the no-representation theorist cannot accept.

The second is that even if we grant that being hot is a dyadic concept theindeterminacy in the intentional objects of self-beliefs returns. If there is anyplace that I am thinking about when believing (5) it is the one referred to by‘here’ when produced by me upon having that belief.17 I could then report mybelief (5) thus:

16. That is, not as a noun phrase which makes no semantic contribution to the uttered sentence andmerely appears because English requires each of its sentences to have a formal subject.

17. ‘Here’ is being used in (5) as a pure indexical, i.e., in a non-demonstrative way. See footnote 11above.


(5¢) *Here is hot now.

(Or more grammatically — though notice that the ‘it’ remains:

(5≤) It is hot now here.)

Thus, my belief would be about the place that I am at when I believe (5) or (5¢).But that place, just as the time of my belief, is also causally contiguous with mybelief. At the time of believing (5¢) I am spatio-temporally contiguous with theplace I occupy, with what ‘here’ in (5¢) refers to. And given that such a belief canonly cause me to act from that place, then— followingMellor’s line of thought— there need not be a mental representation for that place. The present place,the current time and the subject of a belief are all causally contiguous with thatbelief. The time of the belief is the time at which the subject has that belief, andthe place of the belief is just the place where the subject is located when she hasthat belief. Thus, there should be no need for the subject to have a mentalrepresentation in order to think of herself, of the time at which she has thethought or of the place she is at when having that thought.

Although Mellor does not explicitly acknowledge that the place which thesubject occupies at the time of belief is also causally contiguous with the belief,this is certainly so. But this opens a further source of indeterminacy about theres of a belief. Granting that being hot is a dyadic concept relating a time andsomething else, how is the no-representation theorist to distinguish mybelieving (5¢) from my believing (6¢)? The no-representation view would stillhave to acknowledge an indeterminacy in my believing (5¢), in whether it isabout me or the place I am at. And notice that in being unable to do away withthe indeterminacy, it is unable to account for the causal powers of the beliefs inquestion, for why they are ones that concern me and not the place I am at orvice versa. Thus, in being unable to avoid an indeterminacy in the res of self-beliefs, the no-representation theorist is unable to rescue the self-locating andself-knowledge features of self-thoughts. Consequently, it is unable to give atrue account of what is involved in self-thoughts.

The source of the problem lies in the assumption that in having a ‘now’, ‘I’or ‘here’-thought, context — whether it be restricted by causal contiguity orsomething else— fixes or determines the res (object or variant) of my thought.Contexts (even those that select only causally contiguous features) involvemany entities, so the subject must have a way of discerning one amongst themfor her to have a thought about one of those elements. Context alone cannotensure that my ‘I’-thoughts are about myself, and not about the present time or

190 Maite Ezcurdia

the place I am at when having such thoughts. I need something more: a mentalrepresentation or a mode of presentation can do the job. A mental representa-tion may be such that it serves only to represent a certain feature of a context,and a mode of presentation may just present a particular feature of a givencontext. If the mode of presentation in having ‘I’-thoughts is one that presentsa subject as me, as myself, then there is no possibility that it present the presenttime or the place I occupy.

III

Given that the Millian view accepts that there are some mental representationsfor subjects, others for present times and others for current places, it does notface the indeterminacy problem which the no-representation view faces. Theview is that upon having a self-thought a subject thinks of herself with amentalrepresentation which is a Millian name for that subject, that is, with nomode ofpresentation mediating between the name and the subject. This claim makessense only if we assume that mental representations behave in a way similar topublic language expressions. It is under the assumption of a language ofthought that Millikan questions whether the mental counterparts for ‘I’ couldbe indexical and argues in favour of their Millian status. Roughly, Mentalese isa rule-governed system of language-like representations with physical shape-like properties as well as syntactic and semantic properties. It is under theassumption that Mentalese is sufficiently like public languages that it makessense to speak of indexical andMillianmental representations, of mental termswhose reference may vary relative to a context and of mental terms which aremere tags for their objects of reference. Let us then assume a language ofthought of this sort even if only for the sake of argument.

Let us call the mental counterpart(s) for ‘I’ self-representation(s). On theMillian view, each subject has her own self-representation, her own counterpartfor ‘I’, which differs from other subjects’ self-representations insofar as each isa Millian name. Each self-representation is meant to ‘hook up’ with thecorresponding subject’s dispositions to act, think and perceive, in a way thatvindicates the self-locating feature. Whereas ‘I’ is an indexical term which doesinvolve a mode of presentation, its Mentalese counterparts, according to theMillian, are not indexical and do not involve any mode of presentation.Millikan’s own strategy in favour of the Millian view is to argue that self-representations are not indexical so that they could not have the sort of mode


of presentation which indexicals have, and then to propose that self-representa-tions are Millian names.18

A term is indexical if and only if it varies in reference with a relevantvariation in the context in which it is used and it does so in virtue of its seman-tic rule or character. What determines the relevant variation in contexts is thesemantic rule for the indexical. The rule for ‘I’ determines that relative to aparticular context of use it refer to the producer or utterer, so two contexts willbe relevantly different if and only if the producers or utterers in them aredifferent. The semantic rules of indexicals involve modes of presentation oftheir referents which figure in or just constitute such rules. The rule for ‘I’determines that it refer in a given context of use to the producer under a certainmode of presentation. That mode of presentation may just be one whichpresents its referent as the speaker or utterer (Barwise and Perry, 1981; Récanati1993), as herself (Rovane, 1987) or simply as the producer (see V below).19

18. It is difficult to see what Millikan’s reasons are, but I think what follows is a fair reconstruction ofthem. The presentation in the main text assumes as little as possible of Millikan’s views of language andthought as biological categories. For those interested in her theoretical framework, themain text shouldbe read paying close attention to footnotes 19, 21 and 22.

19. Although Millikan does mention an indexical’s character, she frames her discussion explicitly interms of an indexical sign’s adapting relation. The latter is a relation which has as its domain persons,objects, times, places or properties, and indexical signs as its range. It maps entities like persons, objects,times, places or properties onto indexical tokens, and it achieves this in virtue of the semantic rule ofindexicals. It is in virtue of the indexical adapting relation for a sign that times, places, objects, peopleand properties get to be the referents of token indexicals.

Tokens of an indexical type have referents when they are situated in appropriate contexts. Anappropriate context contains something bearing a designated relation to the indexical token,which something is, thereby, that token’s referent (e.g., a person, an object), or is, thereby,that variant in world affairs that the token indexes (e.g., a time, a place, a property). Thisdesignated relation for a given indexical type I call its indexical adapting relation. The indexicaladapting relation for ‘I’, for example, is being the producer of the token … . [Millikan1990:725; this volume, 168–9]

An indexical adapting relation can be seen, roughly, as that relation which an object or variant bears toan indexical token when such a sign functions according to what it was ‘designed’ to do or, moreprecisely, according to its relational proper function. The object or variant in such cases becomes theindexical token’s adaptor, and the indexical is thus said to be for the moment adapted and to have anadapted proper function or what Millikan called “an adapted sense” (Millikan, 1984). (For a detaileddiscussion of this see Millikan, 1984, Chapters 2 and 10.)

Because it is through the semantic rules or character of indexicals that indexical adapting relationsachieve the relevant mapping of referents onto indexical tokens, because I believe that arguments infavour of the Millian position should stand or fall independently of Millikan’s controversial project ofexplaining language and thought in terms of biological categories, and for the sake of simplicity, I shallspeak only of the semantic rules of indexicals. In section IV we shall see more clearly to what extent avariant of the Millian position is plausible only under an account of thought in essentially biologicalterms.

192 Maite Ezcurdia

According to Millikan, self-representations are not indexical because (i)they cannot vary in reference in the way that indexicals do, and because (ii) theusers of those self-representations do not identify their referents indexically. Toshow (i), Millikan implicitly assumes that indexicals must satisfy what I call theindexical ability requirement, whilst to show (ii), she explicitly assumes what Icall the epistemological requirement.

i. How could a self-representation occur in relevantly different contexts suchthat it vary in reference? It is certainly possible that the same physical sign-shapebe instantiated in both your and in my head, and that it refer to me when it isproduced in my head and to you when produced in your head. But this doesnot suffice to show that the tokens produced in you and those in me are tokensof the same indexical sign.20 One reason for saying this might be that it wouldnot have been shown yet that our different syntactical tokens had obtained theirreferents in the same sort of way, that they had obtained their different referentsvia the same semantic rule. But this is not the reason given by Millikan.

Millikan thinks that a term is indexical only if it satisfies the indexicalability requirement�: if T is an indexical term then a user of the language shouldbe able to use T indexically, that is, she must be able to produce and/or interpretT as having different referents in relevantly different contexts. So even if thesyntactical form of all occurrences of self-representations were identical, forMillikan this would not count as evidence that my and your self-representa-tions were occurrences of an indexical term, not — at least — until oneshowed that a thinker could produce and/or interpret one of those self-representations to refer to something other than herself. But according to her,no one could produce a token of her own self-representation to refer toanything other than herself, nor could a thinker interpret self-representationsas varying in reference, not — at least — within the language of that self-representation, within the language of thought.

According to Millikan, the only way I could envisage myself producing aself-representation to refer to you would be by producing it in your head, butI certainly can’t do that. I can only produce tokens of Mentalese in my head.And the same goes for everyone else. Thus, on her view, no thinker couldproduce a self-representation to refer to someone else. But even if a subjectcould not produce a self-representation indexically, could she interpret it

20. As noted in footnote 8 above, I shall sometimes speak both as if tokens and as ifMentalese indexical-types relative to contexts of use are what do the referring. I do not take either side of the debate.


indexically?Given that Millikan does not think that being of the same syntactical type

is evidence that a term is indexical, there is no harm in supposing that thesyntactical form ‘ME’ is the form of all self-representations. Let us indicate thisby subscripting it with ‘U’ thus ‘MEU’, keeping ‘ME’ for token self-representa-tions (nomatter whose self-representations they are). According toMillikan, nothinker could interpret ‘MEU’ indexically for no thinker has immediate accessto Mentalese expressions produced by others. God or a brain-reader might beable to interpret instances of ‘MEU’ as referring to different subjects when theyoccurred in different heads, identifying different subjects as the referents of‘MEU’ in relevantly different contexts. But the fact that a term can be usedindexically by someone whose language is not the original language of that termis no indication that the term in the original language is indexical. Uponinterpreting different occurrences of self-representations as having differentreferents, God and the brain-reader are not doing so from within the languageof thought, but rather outside it. According to Millikan, they take the differentoccurrences of ‘MEU’ as evidence that they are signs for the subjects in whosehead they occur in the same way in which one would take footprints to beevidence of someone’s having been there.21 Self-representations are terms inthinkers’ language of thought, so if no thinker can either interpret or producethat term indexically, i.e., to refer to different selves, then on Millikan’s view‘MEU’ cannot be indexical. Even if evolution had ensured that every thinkerused tokens of the syntactic type ‘MEU’ as their self-representations, this wouldstill not make ‘MEU’ an indexical term.

ii. Millikan thinks that if a term T is indexical then the subject who uses Tmeaningfully must first identify or know the context C in which it is used,hence identify or know the different elements of C, so that she can then

21. The syntactical shape interpreted by God and/or the brain reader would be functioning more as anatural sign which they interpret as signs for the subjects in whose head they appear. To this effectMillikan writes:

[…] the language of thought, if there is such, is not God’s language nor the brain-reader’slanguage but the thinker’s language. Godmight read tokens of the universal self-[representa-tion], tokens of the mental ‘I’, indexically, determining the reference of each by first noticingwhose head it was in. Similarly, I might ‘read a chameleon’s back’ descriptively, as a naturalsign telling what colour the chameleon has been sitting on, although the chameleon’s colourhas no descriptive meaning for the chameleon. The universal self-[representation] [i.e.,‘MEU’] would not be an indexical for the selves who named themselves with it, andwhen readby someone else it would function only as a natural sign, not as a sign in the language ofthought. [Millikan, 1990:732; this volume, 177]

194 Maite Ezcurdia

identify one of those elements as the referent of T in C via T�’s semantic rule.This is the epistemological requirement.22 If ‘MEU’ were indexical, then, in thelight of the epistemological requirement, I would have to know the context,and in particular the head in which a token of ‘MEU’ occurs, before I couldidentify its referent in that context. Prior to my being able to think of myselfwith my self-representation I would have to have some other way of identify-ing myself, I would need to have some other identificatory knowledge ofmyself. But I do not usually identify the context in which my self-representa-tion occurs. From this Millikan concludes that subjects do not satisfy theepistemological requirement with respect to self-representations, and hencethat such representations are not indexical.

There are three objections against Millikan’s view. Two (A and B) aredirected against the arguments objecting to the indexicality of self-representa-tions and so against the motivation for the Millian view; whilst the third ( viz.,C) is directed against the view itself. (A) The epistemological ability require-ment on which (ii) is based is not a requirement for a term to be indexical. (B)Even if the indexical ability requirement turns out to be a requirement for aterm to be indexical, it is unclear that users of self-representations do not satisfyit. (C) Finally, in order to satisfy the self-locating and self-knowledge features ofself-thoughts, the Millian supposes that self-representations have a particularkind of psychological role. However, in order to fulfil that psychological role,self-representations must present the subject in a particular way, under aparticular mode of presentation.

A. Although self-representations fail to satisfy the epistemological requirementthatMillikan assumes for indexicals, this does not establish that self-representa-tions are not indexical. For it is not true that the public language ‘I’, which is anindexical term, satisfies it, and so that it is indeed a requirement for a term to beindexical. In order to meaningfully utter ‘I’ or to interpret a meaningfulutterance of it, it must be the case — according to Millikan — that I know the

22. Millikan says:

To interpret an indexical, one must have prior knowledge of, one must already knowindependently and ahead of time, what item bears the indexical’s adapting relation to theindexical token. Onemust already know that it exists and how it is related to the token, henceto the interpreter. One does not find this out by interpreting the indexical; one needs alreadyto know it in order to interpret the indexical. [Millikan, 1990:727–8; this volume, 172]

The way she motivates this is via an appeal to the way in which we interpret others use of indexicals. Butshe intends this to apply to producers too: they must have a way of identifying the context before theycan meaningfully produce an indexical.


context in which it occurs, and that via this knowledge and the semantic rule for‘I’ I identify which of the elements in the context is the referent. But it is nottrue that when I use ‘I’ to refer to myself I must first identify the context inwhich that ‘I’ occurs. There are cases where I may have no other way of identify-ing the context in which ‘I’ occurs, for example when I am amnesiac andparalysed, and yet I still meaningfully ask:What happened tome?,Where am I?or Who am I?23 In this case I have no way of identifying the contexts of myutterances prior to my uttering them. Thus, in order to meaningfully utter ‘I’ Ineed not identify the context first in order to know to whom it is referring. Thisdoes not entail, however, that the public language ‘I’ is nonindexical since thesemantic rule of ‘I’ still allows it to change reference with a relevant differencein context. Hence, the fact that one need not go via the context first in order toidentify the referent of a self-representation does not show that it is notindexical.24 Thus, the epistemological requirement, if a requirement at all, isnot one for a term to be indexical.

B. In contrast with the epistemological requirement, the indexical ‘I’ doesappear to satisfy the indexical ability requirement. It is true that in the same wayin which I cannot produce a self-representation to refer to anyone or anything

23. Granted that if I can utter this then I am not completely paralysed. It is enough if I am paralysed tothe extent of only being able to utter those words. Someone in an Evans-like spirit might want to denythat in these cases I do refer to me with ‘I’ by arguing that I may fail to have discriminating knowledgeof myself, that I may fail to locate myself. But notice that this would entail that I fail to have a thoughtabout myself (in Evans’s terminology: an Idea of myself). So even if I failed to refer to myself with ‘I’ inthis situation, an analogy between my self-representation and my use of ‘I’ would still hold. Accordingto this, I can use ‘I’ appropriately only if I am able to have discriminating knowledge of myself. The factthat I fail to have such knowledge of myself not only entails that I fail to have a self-thought but also thatI fail to refer to me with ‘I’. (Although Evans himself considers cases in which there is failure of reference(Evans, 1982:249–255; this volume, 134–139), he would not intend the cases outlined above to be oneswhere there is failure of reference or of thought (Evans, 1982:215–220; this volume, 104–9). RecentlyO’Brien (1995a) has argued that these cases cannot be accommodated within Evans’s general theory ofreference as ones where the subject is successful in having a self-thought.)

24. Other counterexamples to the epistemological requirement are ourmental representations of ‘now’and ‘here’. I may have a ‘now’-thought or a ‘here’-thought about a certain time and a certain placewithout being able to identify or think of (and hence know) the relevant time or place in any other way(perhaps because I have just been woken up and I am paralysed and amnesiac). Thus, I may have suchthoughts without being able to identify in advance the context in which they occur. In these cases, theonly way I can identify a certain place or time is by having those ‘here’- and ‘now’-thoughts, thus Icannot identify those places and times as anything else. But in spite of this, my ‘now’- and my‘here’-thoughts, my ‘now’ and ‘here’ mental representations, are indexical for they do differ in referencewith a relevant difference in context. (Millikan, surprisingly, claims that ‘here’-thoughts are not indexicalon the basis that they fail to satisfy the epistemological requirement (1990:731), but such thoughts areones that do clearly vary in reference!)

196 Maite Ezcurdia

else but myself, I cannot produce ‘I’ to refer to anyone but myself. However, invirtue of being a user of English, I can interpret ‘I’ as referring to someone elsewhen others produce it. Thus, on Millikan’s view in satisfying the indexicalability requirement ‘I’ is indexical, but ‘MEU’ is not. As with the epistemologicalrequirement, it is unclear that the indexical ability requirement is a requirementfor a term to be indexical. We said that a term is indexical if and only if it variesin reference with a relevant variation in context according to the semantic ruleof the term. From this nothing follows concerning our practical abilities forproducing or interpreting it indexically. More argument is needed, and certain-ly Millikan (1984) has some to offer. We shall consider them briefly in IV. Buteven if she were right about the indexical ability requirement, she is not right inthinking that it could never be satisfied by a thinker. We can envisage a situa-tion in which I could interpret someone else’s self-representation as referring toherself, and hence a situation in which the proposed disanalogy between self-representations and indexicals like ‘I’ fails.

Suppose that Mary’s brain and my brain were wired up in a way such thatI could receive Mary’s ownMentalese sentences. In order to avoid confusing theMentalese sentences coming from Mary from those coming from me, theywould be displayed in a particular area of my mind/brain allocated to Mary’sthoughts under my Mentalese phrase ‘Mary’s thoughts’ as heading. So when Ireceived Mary’s Mentalese sentence ‘ME is tired’, it would come under the areaof Mary’s thoughts and I would not confuse it with mine. Given where thatMentalese sentence occurred, I would interpret that token of ‘MEU’ as referringto Mary, i.e., to someone else and not to me. Because it is possible that Iinterpret ‘MEU’ indexically, sometimes referring tome and sometimes toMary,within the same language of thought, the indexical ability requirement issatisfied contrary to what Millikan supposes.

C. Arguments (A) and (B) have left open the possibility that ‘MEU’ is index-ical. In order to show that it is one would need to show that all its instances areof the same semantic type, in particular, that they pick out their referents in thesame way, by applying the same semantic rule. However, in order to show thattheMillian position is wrong and themode-of-presentation view correct, we donot need to go as far as showing that self-representations are indexical. All we needto show is that they involve a mode of presentation of the subjects which theyrepresent. This we can do by looking at self-representations’ psychological roles.

If my self-representation is to enable my self-thoughts to have the self-locating feature, then it must have the psychological role of hooking up with the


dispositions to act, think and perceive, of the person whom (so to speak) I knowhow to move. But notice that every self-representation must have the same kindof psychological (or causal-functional) role relative to the subject in whose headeach occurs, viz., the psychological role of hooking up with the dispositions toact, think and perceive of the subject whom she knows how to move. There isthen a kind of psychological role which all self-representations share. Given thissameness in psychological role, we can say that when a token of ‘MEU’ isproduced in me it locates me in the same sort of way as other tokens of it whenproduced in others locate them.

There are essentially two ways in which we could specify the psychologicalrole which all self-representations share: via (X) or via (ME), where ‘herself ’ isneutral over gender and the index ‘x�’ indicates an anaphoric relation with theprevious ‘x�’.

(X) For all x, each token of ‘MEU’ in x presents x.(ME) For all x, each token of ‘MEU’ in x presents herselfx.

Whereas (ME) specifies a psychological role that presents each subject asherself, that is, under a mode of presentation, (X) does not. Would (X) sufficefor specifying the psychological role of ‘MEU’ in a way that brings out thedifference between self-beliefs and name-beliefs? I think not. Take my‘Maite’-thoughts. These would have the psychological role of presenting me,and this role would be an instance of (X), but they are not thoughts which hookup with my dispositions to act such that they locate me in a way that suffices togive me a reason to act, or with my dispositions to think (to judge or believe)such that it suffices for self-knowledge.25 The reason is mainly that (X) leavesout an important fact about self-thoughts of which the self-knowledge and self-locating features are manifestations, namely, that such thoughts involve a self-conscious element. (X) specifies a psychological role in which a representationpresents a subject, but not one in which the subject realizes that the object beingrepresented is herself. It is this self-conscious element what makes my merebelieving that I am thinking constitute knowledge. It is the self-realization thatthe person who should start reading her paper here and now is myself thatmakes me act accordingly here and now. Thus, if our account of self-represen-tations is to capture the self-locating and self-knowledge features, then the

25. If (X) were to accurately specify the psychological role of self-representations, then there would beno way of distinguishing name-thoughts of mine like (1) from self-thoughts like (2), and hence the self-knowledge feature of self-thoughts would not be vindicated.

198 Maite Ezcurdia

accurate specification of their psychological role will have to show that theyhook up with the right dispositions, namely, with the dispositions to act, thinkand perceive of the subject whom she herself knows how to move.26 (ME) doesthis by claiming that a self-representation presents a subject as herself. But itdoes so at a cost: at the cost of presenting the subject under a certain mode ofpresentation. There may well be other candidates whichmay serve to specify thepsychological role of self-representations. But whatever those candidates maybe, they must show the self-conscious aspect involved in self-representations.And to show the requisite awareness which a subject has of herself, they willrequire to show that the subject is presented to herself in a way in which — asFrege put it — no other is presented to her. This is just what the mode ofpresentation view suggests and is contrary towhat theMillian account supposes.

IV

What we have said so far suffices for showing that both the Millian and the no-representation views are mistaken, and that the mode-of-presentation view is ingeneral correct. But we have not yet said enough to show that self-representa-tions are indexical. We could have a version of the Millian view re-emerging:the Millian* version. This version would not oppose the mode of presentationview insofar as it would accept that there is a mode of presentation associatedwith each self-representation (something our originalMillian account denied),but it would say that despite a self-representation having a mode of presenta-tion, all it does semantically is to refer to or represent a self. TheMillian* accountwould say that themodes of presentation accompanying self-representations areonly psychologically relevant, not semantically relevant. Semantically, self-representations are Millian names.27 28 Put another way, the Millian* viewwould hold that although there is a conception (mode of presentation) associat-ed with the self-concept (self-representation), it is not semantically constitutiveof it.

26. See footnote 3.

27. So far I have said nothing against Millikan’s general point that what it is about ‘now’, ‘here’ and ‘I’-thoughts that is necessary tomotivate us into action is their psychological roles andnot somuchwhetherthey are or not indexical. I agree with her on this point. What I challenge is the idea that such thoughtsare not also indexical, and that in the case of self-representations there is no corresponding mode ofpresentation.

28. Both the Millian and the Millian* accounts would agree on this last point.


We said above that in order to show that self-representations are notMillian names, but rather that they are different occurrences of the sameindexical, we needed to show that all self-representations were of the samesemantic type, that they all shared the same semantic rule, and thereby referredin the same way. From the discussion of self-representations’ psychologicalroles, we know that they all involve the same kind of mode of presentation, thatthey all present different selves in the same manner. We now need to show thatthis kind of mode of presentation is involved in the semantic rule of MEU, inparticular, that the way of presenting different selves constitutes or is part of theway in which eachME refers to each self. What we need to show then is that thesemantic rule for ‘MEU’ is (ME¢) and not (X¢), that is, that the mode of presen-tation involved in the psychological role is semantically relevant for determin-ing the different referents of ‘MEU’.

(ME¢) For all x, each token of ‘MEU’ in x refers to herselfx.(X¢) For all x, each token of ‘MEU’ in x refers to x.

The issue lies in the way in which self-representations get their reference. It isnatural to say that self-representations get their reference in virtue of thepsychological role they play in the subject’s psyche, i.e., in virtue of theirpresenting something as one’s self. This entails that the mode of presentationinvolved in their psychological role already figures in the semantic rule of self-representations. Hence, (ME¢) and not (X¢) must give us the correct specifica-tion of the semantic rule of self-representations or the self-representation-type.And notice that such a rule determines that the self-representation-type isindexical, that is, that if it is tokened in relevantly different contexts, then itsreferents are different. The relevant difference in context is just the differenceof self in which ‘MEU’ is tokened.

To suppose that the psychological role of self-representations is semantical-ly relevant does not entail a confusion of presemantic conditions with semanticconditions. A particular act of baptism, for example, may have been a pre-semantic condition for a public language name to refer but it is not a semanticcondition for it to refer. Or a description can serve to fix the referent of such aname, yet it is not part of the semantic condition or semantic rule of that name.But the reason why this is so is twofold: (a�) there are numerous acts of baptismand reference-fixing descriptions which could have served to fix the referent ofthe name whilst the referent (and in general the semantics) of the nameremained the same, and (b�) once the act of baptism has taken place or thedescription fixed a reference they are (so to speak) left in the past. After fixing

200 Maite Ezcurdia

the reference via a description or baptism, the description and the baptism ceaseto play a role in determining the referent of the name. Nevertheless, neither (a�)nor (b�) are true of the psychological role of a self-representation. There are notnumerous psychological roles whichmy self-representation could have on painof ceasing to be a representation of me that rescued the self-locating and self-knowledge features. Thus (a�) is not satisfied. Furthermore, my self-representa-tion continues to play a role in hooking up with me in a certain way, and it isthis way which determines that it hooks up with me and not with some otherentity (not with time, for example). Hence, the psychological role of my self-representation plays a constant role in determining the referent of my self-representation. If my self-representation failed to hook upwithmy dispositions,with the subject I know how tomove, then not only would it lose its psycholog-ical role, it would lose its referent too. For how could it refer to me?Why shouldit? Given the initial act of baptism, if the name ‘Maite’ is a Millian name, itcontinues to be my name even if it fails to play the psychological role in mewhich it plays right now. But in virtue of what would my self-representation‘ME’ (if it survives) continue to be a representation of me if it were to lose itspsychological role? (ME¢) must then be the semantic rule for all self-representa-tions, for the self-representation type. And given (ME¢), it would seem that theself-representation type ‘MEU’ just is an indexical term, a term whose tokensdiffer in reference with a difference in context.

Before considering an objection to my account of the semantics of ‘MEU’,let me summarize what I think happens with self-representations. The rolewhich my self-representation has in my psyche is to hook up with the disposi-tions to act, think and perceive, of the person I know how to move; and it is invirtue of this role that my self-representation can refer only to me. It refers tome because of and in the way in which I am presented by its psychological role,namely, as the person I know how to move. A self-representation of any otherperson has a similar role in that person’s psyche for that person. In her, her self-representation refers to her and to no one else. It refers to her as that person’sself. Thus any other person’s self-representation has the same kind of psycho-logical role as my own self-representation. The psychological roles of differentself-representations of different persons all have the role of hooking up with thedispositions to act, think and perceive, of each of those persons in a certain way— as the self or person each knows how to move. Thus, self-representationspresent different selves in the same way, they present each self as her own self toeach subject. And this mode of presentation constitutes the way in which thereference of each self-representation is determined, it constitutes the semantic


rule for the self-representation type. If so, the representation is indexical.29

Someone might object to my account by claiming that I am committingmyself to a particular view of howmental representations get to bemeaningful,viz., to the view that the semantics of mental representations are determined bythose representations’ psychological roles (or causal-functional role). However,I am not committing myself to that much. All I am committing myself to is tothe view that the semantics of certain kinds of mental representations are, atleast partly, determined by their psychological role. I do not take a stand eitheron whether the reference of a Mentalese term is in general determined by whattypically causes such tokens or by its causal-functional (psychological) role. Thereference of any Mentalese term may be obtained in terms of what typicallycauses it, but this is not to say that this exhausts the semantics of that term. Insome cases, like the Mentalese terms for ‘cat’, ‘dog’, ‘water’, it may be so. Butnot in the case of the Mentalese counterparts for ‘now’, ‘here’, other indexicalMentalese terms, and self-representations. In the case of themental representa-tions for ‘now’ and ‘here’, though their reference may vary with a variation incontext, we could say that what typically causes them are certain times andcertain places, and so that they get their reference (partly) in virtue of whattypically causes them. However, this does not exhaust their semantics, for theirsemantics is not exhausted by what they refer to. They refer to times and placesin a special way, viz., as the present time and as the current place. And whatdetermines this way of referring to them is just those terms’ psychological (orcausal-functional) role. Thus, even if we thought that in general Mentaleseterms get their semantics through what typically cause them, we would have toaccept that the semantics of certain Mentalese terms is not exhausted by whattypically causes them. What I have said above can be seen as an argument forthe claim that the self-representation-type should be taken as a case wheresemantics is not exhausted by what typically causes it, but is also given by theway in which what typically causes it is presented.

29. The way in which the semantic rule of the self-representation is interpreted will depend on thegeneral view we may hold about thought. We may, for example, think along neo-Fregean lines that thetoken mode of presentation of myself that determines its referent is a sense and so is constitutive of thecontent of my self-thoughts. We may deny this and adopt instead a two-factor theory of content, wherethe semantic rule is what semantically determines the content of the thought but it is not part of thatcontent, where the content of the thought is just its truth-conditions. In this case the semantic rulewould amount to what Fodor (1987) calls “narrow content”, to a function from contexts to broadcontent.

202 Maite Ezcurdia

What is true, nonetheless, is that in claiming that certainmental representa-tions get their semantics at least partly in virtue of their psychological role I amcommittingmyself to the claim that the semantics of such representations is notdetermined by their biological functions, as Millikan supposes. Were they to bethus determined, then the Millian* view (though not the Millian view) wouldbe right. Very roughly, according to Millikan, a term’s semantics is given by itsproper function, by the function that it was designed to carry out. On her view,a proper function F of a device D with a trait C is that which C has in the pastcaused and its having caused F is what explains why D-type devices haveproliferated. If biological functions are what determine the semantics of mentalrepresentations, then prima facie a self-representation’s proper function is notthat it be indexical. The mechanisms that have aided in its proliferation are notones where it varies in reference with a variation in context because they are notones where the users of self-representations have produced or interpreted themas varying in reference (the indexical ability requirement).30 What has aidedin the proliferation of self-representations is their serving the psychological rolespecified by (ME), i.e., the psychological role responsible for securing the self-locating and self-knowledge features for the selves of which they are representa-tions. Thus, under this account of the semantics of thought, self-representationsare not indexical, and the Millian* view is right.

To argue against a biological account of the semantics of mental representa-tions is no easy task. I shall not attempt it here. All I shall say is that the Millian*view is right only if a particular and controversial account of how thought-representations get their semantics is right. In contrast, the view I am espousingdoes not depend on such an account nor on any particular view of howMentalese terms in general get their semantics.

V

I have argued that both the no-representation and the Millian views fail asaccounts of self-thoughts. I have also argued that for self-thoughts to vindicateboth the self-knowledge and self-locating features, they must present the subject

30. The need for the indexical ability requirement as a requirement for indexical terms on Millikan’sview would lie then in finding out what historically its proper function has been, and hence in what hasaided in the proliferation of indexical signs. An indexical sign has proliferated just because it has beenused (produced or interpreted) indexically, as referring to different objects or variants; therefore, its(relational) proper function is to vary in reference with a variation in context.


as herself, as that subject which she knows how to move. Thus, the mode ofpresentation approach is essentially right.

Furthermore, I have argued that it is plausible to assume that the Mentaleseterm for oneself is indexical, insofar as it is plausible to assume that it has asemantic rule that allows it to vary in reference with a relevant variation incontext. It is thus tempting to think that the semantic rule for that term, theway it refers to subjects, is the same as the way in which the public language ‘I’refers to subjects. It is tempting to think that (ME¢) not only gives us thesemantic rule for ‘MEU’ but also for ‘I’. But this need not be so. Consider this.Is it necessary for anyone to interpret a token of ‘I’ as referring to somethingthat is a self? I am not sure that our intuitions are really clear on this point.Some might think that we are committed to thinking that any specification ofthe semantic rule for ‘I’ presupposes that what it refers to is a subject.31 Foreven if we take the rule in question to be one which states that a token of ‘I’refer to the speaker or the utterer of it, we are already committed to the viewthat the referents are subjects, are beings who can engage in intentional actionslike speaking or uttering. But I think there is a neutral specification of thesemantic rule of ‘I’, and that until our intuitions are clear on whether all tokensof ‘I’ must refer to selves, to persons, we will do best to adopt it.

If we say that the semantic rule for ‘I’ is that a token of it refer to theproducer of the token, we are not thereby committed to subjects or selves beingthe producers of the tokens. We would take on no commitments as to thenature of the producer of the tokens. This rule also seems to fare well wheninterpreting people’s tokens of ‘I’. As O’Brien (1995b) has suggested, ininterpreting a subject’s token of ‘I’, I bring to bear additional information, viz.,that the producer in this case is a subject. In general, we interpret tokens of ‘I’as being produced by and hence referring to people, to selves, because most ofthe occurrences of ‘I’ we are exposed to are produced by subjects. Thus,although the self-representation-typemay be indexical like the public language‘I’, its semantic rule is different from the semantic rule for ‘I’.32

31. For example, O’Brien, 1995b, and Rovane, 1987.

32. For a suggestion of an account of the relation between self-thoughts and uses of ‘I’ which aregoverned by the rule being-the-speaker-of, see O’Brien, 1995b. I think that if such an account does endupworking for when the semantic rule is being-the-speaker-of, it will alsowork for being-the-producer-of.

Thanks to audiences at Birmingham, Granada, King’s College London, Mexico and ST Andrewswho listened and commented on previous versions´of the present paper. I am grateful in particular toTyler Burge, Fraser MacBride, Lucy O’Brien and Gianfranco Soldati for comments and discussion.

Chapter 10

Introspective misidentification

An I for an I

Melinda Hogan and Raymond MartinUniversity of British Columbia / University of Maryland

Distinguish four discoveries one might make by introspecting: first, that amental state has certain features, say, that an emotion is pleasurable. Second,that a mental state with certain features is tokened at the present moment, say,that a pleasurable emotion is now occurring. Third, that a mental state withcertain features is tokened at the present moment in a self — that a pleasurableemotion that is occurring now is someone’s. Fourth, that a mental state withcertain features is tokened at the present moment in one’s own self — that apleasurable emotion that is occurring now and that is someone’s, is one’s own.

Is it possible that by introspecting a person should make any of the firstthree of these discoveries without also making the fourth? Some philosophershold that it is not possible. They hold that any introspectively-based judgementof the form ‘I am E’ is immune to error through misidentification of thepossessor of the property E. We call this the no-misidentification view. Some whohold this view maintain that this kind of immunity to error shows that intro-spection is not a form of perception.1

We shall present several counter-examples to the no-misidentification view.The examples are of situations in which one could have introspective access toa property’s occurring, think that this property is a property of oneself, andeither be wrong about this or else fail to have conclusive reasons for it. In thefirst three examples, the protagonist is wrong in thinking that the property is aproperty of himself because the property in question is a property of someone

1. Wittgenstein is a case in point. He holds that this kind of immunity to error shows that introspectionis not a form of perception in which one is aware of oneself as the object of introspective awareness, andthat judgments based on introspection are not about a particular person (Wittgenstein 1933–4: 66–67).See also Gareth Evans’ (1982, this volume) discussion of the issue.

206 Melinda Hogan and Raymond Martin

else. In the last example, the protagonist fails to have conclusive reasons forthinking that the property is a property of himself either because the propertycould be a property of someone else or because it could be a property of no one,that is, of no person or self.

The no-misidentification view leaves it open that introspectively-basedjudgments can bemistaken in ways other than that the possessor of the proper-ty introspected has been misidentified. There are at least two sorts of situationsin which mistaken judgments could arise. First, one could confuse the mentalproperty that is discovered in introspection with another property. For instance,one could falsely judge, ‘I want to be fair’, where one’s mistake was to confusea vengeful desire with a desire to do what is fair. On the no-misidentificationview one could not falsely judge, ‘I want to be fair’, where one’s mistakeconsisted in taking the possessor of the desire to be oneself. So, on the no-misidentification view, one canmake a mistake about what the mental propertyis but not about who or what has the property.

Second, one could falsely characterize the possessor of a mental property,and thus, in a different sense of ‘who?’, get it wrong about who has the property.For instance, one could think to oneself, ‘I, theMayor of Toronto, am presentlyhaving a titillating emotion ‘ and be mistaken in thinking that the person whois titillated is Mayor of Toronto. To concede that in this way one could falselycharacterize the possessor of the mental property, however, is different fromconceding that in forming a first-person judgment based on introspection onecould fail to identify the possessor of the property.

The Mayor-thought (among other things) supplies an answer to thequestion, ‘who? ‘ in a sense of ‘who? ‘ that asks for a further characterization ofthe object in question once it has been identified (picked out). (Who is thatmasked man?) The type of judgment that is thought to be immune to errorthrough misidentification supplies an answer to the question ‘who?’ in a senseof ‘who?’ that asks for an object that will fit a given characterization to beidentified (picked out) in the first place. (Whose turn is it to do the dishes?)

We assume that the possibility of mistakes in identifying the object per-ceived is a constant feature of all perception. So if introspection is a kind ofperception in which the object perceived is the possessor of the property ofwhich one is introspectively aware, then when one makes an introspectively-based judgment ordinarily it would be possible to misidentify the possessor ofthat property. For instance, when, based on visual perception, one judges thatone’s arm is sunburned, one can get it right about the sunburn but wrong aboutwhose arm it is. An analogous mistake in a judgment based on introspection

Introspective misidentification 207

would be for one to judge that it is oneself who is hungry or nervous whenreally it is someone else.2

Even if introspection were a kind of perception, introspectively basedjudgments still might be immune to error through misidentification. Forinstance, for some physiolopgical reason, introspectionmight provide informa-tion only about the introspector, that is, the one who forms the introspectively-based judgment, ‘I am E’. Similarly, in the case of visual perception, a personmay be physically restrained so that he can see only himself. However, thosewho hold the no-misidentification view for the reasons we are here consideringwill think that this sort of guarantee against error is the wrong sort of guarantee.In introspectively-based judgments, they will say, the reason one does not makea mistake about who or what has the property of which one is introspectivelyaware is not that God puts blinders on introspectors so as to ensure that one’sintrospective gaze is necessarily fixed only upon oneself; this cannot be thereason since the self is not the object of introspective awareness.

So if there were a guarantee against misidentification of the sort that thosewho for such reasons hold the no-misidentification view say that there is, whatwould it consist in? It is hard to say. To our knowledge, no proponent of theno-misidentification view has ever explained what it would consist in, unless itis just this: that if a property is discovered by introspection, thenmerely becauseit is discovered by introspection (and not, say, because introspection yieldsinformation only about oneself or for any other reason), one is correct injudging that the property belongs to oneself. Sometimes it is suggested, vaguely,that this is ensured by ‘a rule of language’. The question, then, is whether thereis any such guarantee. We do not think that there is.

In support of our view, we shall now present four counterexamples to theno-misidentification view. They illustrate different ways in which informationabout mental states might be structured. To save space, we shall present theseexamples briskly and without the usual stage-setting. There will be ways offurther specifying the examples so that they do not support our view. Our claimis that there are also straightforward ways of specifying them so that they do.

2. Surprisingly, Michael Ayers seems to deny the possibility of both of these kinds of mistake:“It could be argued that, if my body were so presented, then I could be aware of the disposition of a bodyand yet bemistaken as to which body was the object of my awareness, which is absurd. In fact, of course,the impossibility of misidentifying which body I am peculiarly aware of gives no grounds for denyingthat I have a particular sensory awareness of my own body. We do have bodily sensations such as (mostspectacularly) pain, and we cannot intelligibly wonder whether the body so presented is our own orsomeone else’s.” [Ayers 1991:288]


Example 1. John is telepathic. If through introspection he is aware of theoccurrence of mental properties, then in the absence of contextual clues, he isin a position to infer only that someone has the properties, not that they are hisown. John is also a hardened egoist. He has never had a sympathetic feeling foranother person. In fact, because of how he is constituted psychologically he isincapable of having a sympathetic feeling. But he does not know that. On theoccasion in question John discovers through introspection that someone ishaving a sympathetic feeling. The ordinarily reliable contextual clues arepresent but this time they lead him astray. He judges that it is himself who ishaving the sympathetic feeling — actually it is someone else.

Consider an analog: Machines M and N are at a coarse-grained level ofanalysis functionally isomorphic to creatures with minds. M is hooked up tomachine N in such a way that M can detect changes of state in N. M can alsodetect changes in itself. Certain changes of state occurring in N are what it is forN to have the ‘mental’ property E. But it is only in the context of N’s internalorganization, not M’s, that these changes of state would be occurrences of E. Inthe past N but not M has had E. As it happens, M is incapable of having E. Thisinformation is not available to M. On the occasion in question M detects N’shaving E but does not register the information that it is N that has E. InsteadMprints out, ‘For the first time, I am having E�!’ In so doing, M misidentifies thepossessor of E.

Objection�: If M is not merely serving as N’s mouthpiece, then either M isnot detecting N’s ‘mental’ properties in such a way that the detection amountsto introspection or else N’s ‘mental’ properties must also be M’s, and so M iscorrect in judging ‘I am E�’. Hence, Mmakes no mistake (unless in judging ‘I amE’, M implicitly ‘assumes’ that M is the only one who is E, or who is experienc-ing that particular token of E�). There is an analogous problem in the case of thetelepathy example.

Reply�: First, why assume that if a mental property is detected in such a waythat the detection amounts to introspection, then the introspector must beamong those who possess it? So far as we are aware, there is no evidence abouthow people process information that supports the view that anything like thisassumption is required to explain either how introspection works or how itdiffers from other forms of information-gathering.

Second, whatever distinction onemight plausibly draw between introspec-tion and other sorts of information-gathering, it still has to be shown that one’smerely introspecting a mental property is sufficient for one’s possessing it. It isnot. The introspector’s possession of the property must be compatible with his


possession of whatever other properties he possesses. Failure to satisfy thiscondition could keepM from possessing E regardless of how he came upon theinformation that E was occurring.

Example 2. First, the proto-example. Paul is super-macho. He has neveracknowledged even to himself that he has ever felt fear. Eventually, throughpsychotherapy, he discovers that when he gets angry in a certain way thenalmost certainly he is afraid. As a consequence of this discovery he acquires theability to discover through introspection that he is afraid by discoveringthrough introspection that he is angry.

Now the example itself. Change the proto-example so that what Paul learnsin psychotherapy is that when he gets angry in the relevant way then almostcertainly either he is afraid or his daughter is afraid. Subsequently, when Paullearns through introspection that he is angry he has hunches about whether itis himself or his daughter who is afraid. These hunches, while generally reliable,are, occasionally mistaken. Paul is unaware that his hunches are ever mistaken.On one occasion he learns introspectively that either he or his daughter is afraidand, under the influence of one of his hunches, he then takes himself to be theone who is afraid. He gets it wrong. It was his daughter.

Objection�: Paul is not really introspecting since his access to the informationthat fear is occurring is indirect.

Reply: The distinction between direct and indirect needs to be explained. Itmay come to this: that Paul knows that fear is occurring by knowing that angeris occurring, whereas he does not know that anger is occurring by knowing thatfear is occurring; rather, he just knows it. But the problem with this account ofthe direct/indirect distinction is that one knows that anger is occurring byknowing lots of other things, for instance, by knowing what anger is. So evenPaul’s introspective awareness of anger is indirect.

Even if the distinction between direct and indirect were tenable, for theobjection to work some reason would have to be given why direct access isrequired for introspection.

Example 3. Sam’s brain is removed from his head and divided into func-tionally equivalent halves, each capable of sustaining his full psychology. Eachof these halves is then transplanted into its own brainless body, which except forbeing brainless is qualitatively identical to Sam’s body just prior to the surgeryand, hence, also qualitatively identical to the other brainless body. Once Sam’sbrain is removed (and just prior to its being transplanted), his body is de-stroyed. One of the survivors of the operation thinks that he is the only survi-vor. So, he thinks that he is Sam. He quasi-remembers having become aware


just before the operation, and through introspection, that he felt nervous. Hebelieves that it was himself who felt nervous. He is wrong. It was Sam.3

Objection 1�: An assumption of the example is that the person who quasi-remembers is not Sam. Some philosophers would disagree.4

Reply�: Granted. So this example is for those who think that the person whoquasi-remembers is not Sam.5

Objection 2�: The quasi-rememberer is not attempting to base his identifica-tion of the person who was nervous on observed similarities between himselfand Sam or on anything like a perceptual tracking of a self over time. Hence, hisidentification is not based on the sort of evidence on which ordinary perceptualidentifications are based. The only thing that would count against the no-misidentification view would be attempted and failed identifications that arebased on that sort of evidence. Such misidentifications would be ones thatmischaracterize the possessor. So, even though the quasi-rememberer refers tohimself whenhe judges that he felt nervous, and even though hewas not the onewho felt nervous, this is not a genuine case of introspective misidentification.

Generally, in the case of the identifications we make in connection with ourown memories, the fact that “our… knowledge of the world supports a pre-sumption that ‘fission’ and the like do not in fact occur” (Shoemaker 1986:131)is what entitles us to identify with the person whose experiences we remember(and so take ourselves to be that person). We do not use our memories tocompare observed similarities between selves at different times or to perceptual-ly track selves. We simply have the experience of remembering (or quasi-remembering) the introspecting of a mental property and thereby identify with,and identify, the possessor of the introspected mental property.

Reply. First, for the point at issue, it is not enough that our knowledge of theworld supports a presumption that fission and the like do not occur (and thusgives us grounds to identify with the person whose experiences we remember).Unless our knowledge also supports the presumption that fission is physically(and not just technologically) impossible, it leaves it an open question whetherfissionmight someday occur, hence, also open whether in a fission situation, in

3. This sort of example and the replies to it which we distinguish will be familiar to those who are awareof the post-1960’s debate over personal identity and what matters in survival. See, for example, Parfit1971, Perry 1972, Lewis 1976, Parfit 1976, Lewis 1983, Parfit 1984, Shoemaker 1984, Sosa 1990, Rovane1990, Unger 1991.

4. Unless he has changed his mind, David Lewis would disagree (Lewis 1976).

5. Unless they have changed their minds, Parfit, Shoemaker, Unger, and Sosa would think that theperson who quasi-remembers is not Sam. See note 3.


forming a judgment based on introspection, misidentification of the possessorof the mental property is possible. The reply to the counterexample in effectchanges the no-misidentification view to the claim that provided fission casesdo not occur, judgments based on introspection are immune to error throughmisidentification.

Second, the objector to the example changes the no-misidentification viewin another way as well. The objector changes it to the view that misidentifica-tion that comes about in certain ways is impossible, whereas the original (andmore interesting) view is that misidentification per se is impossible.

The objector claims that in introspection one employs no observed similari-ties to identify the possessor of a given mental property. But even if the objectorwere right about this, that would not show that this example is not a counter-example. The no-misidentification view is not the view that in introspectionone employs no observed similarities to identify the possessor of the intro-spected property, but rather the view that one cannot in judgments based onintrospectionmake amistake about the possessor of the property introspected.On the assumption that the fission-descendant in the example is not Sam butnevertheless learns through introspection that Samwas nervous just before theoperation, the example is designed to show, and does show, that in introspec-tion one can make a mistake about the possessor of the property introspected.It is because the example shows this that it is a counterexample to the no-misidentification view. To be a counterexample, the example does not also haveto show that in introspection one employs observed similarities to make ajudgment about the possessor of a mental property.

In fact, the objection involves a confusion between two senses of ‘who?’.The objector’s suggestion is, first, that it is a background condition for anintrospectively based judgment’s being reliable that fission is not occurring, andsecond that since what would otherwise be a misidentification (after all, thefission-descendant does get it wrong) is not due to the introspector’s mis-characterizing the possessor of the mental property (e.g. as one who did notundergo fission), the example is not really one of misidentification after all:since the introspector does notmischaracterize, he does notmisidentify. But therelevant sense of ‘who?’ in the question ‘Who was nervous?’ is the one thatmerely asks for the possessor of the property of having been nervous to bepicked out whether or not this is accomplished by having an accurate character-ization. And in the counter-example the wrong person is picked out as the onewho was nervous.

Third, there is nothing that prevents a quasi-rememberer or for that matter


even an ordinary rememberer from basing his identification on observedsimilarities between selves (or person-stages) observed at different times.Imagine, for instance, that, contrary to the original example, the fission-descendant knows or believes that he may not be the only survivor. But he alsoquasi-remembers mental states, other than that of being nervous, that Sam hadjust before the operation. And he takes these to be just like mental states that,based on current introspection, he thinks are typical of him. So, he infers(recklessly) on the basis of observed similarities between selves observed atdifferent times that he must be Sam. He gets it wrong.

Example 4. George went to the hospital for an operation during which hewas mistakenly given a curare-like drug for an anaesthetic. As a consequence,throughout the operation, although he experienced intense pain, he wasparalyzed physically and unable to protest. In addition, his organism wasparalyzed cognitively in ways that affected its ability to process sensory informa-tion in making decisions. On learning after his recovery that the curare-likedrug was mistakenly used, George decided to sue the hospital.

At the trial, George foolishly admitted that although he can now rememberthe experiencing of the pain, so far as he knows, once the operation wascompleted he suffered no physical or psychological scars from that experienceof pain’s having occurred; for instance his memory of the pain is not now and,so far as he or anyone knows, never was distressful to him. In exploiting thisadmission, a lawyer representing the hospital claimed that although a mistakewas made as a consequence of which George’s organism experienced pain, sofar as anyone knows George was unharmed. The reason for this, the lawyerclaimed, is that while the pain may or may not have been experienced by a self,the experience of the pain is (and always was) so unintegrated with the rest ofGeorge physically and psychologically that George did not experience it. HenceGeorge deserves no compensation.

Objection�: Someone felt the pain and since George remembers having feltit, George is the only candidate. So, George felt the pain.

Reply�: The dispute is over the question whether painful states of theorganism that endured the pain were George’s psychological states. One way inwhich they could fail to have been George’s is that the organism at the time itendured the pain was not a person (or self) at all. Another way is that theorganismwas a person but was so thinly connected to George’s normal self thatit was not George. A third way is that the organism was George but its experi-ences of pain— particularly its experiences of the pain’s hurtfulness—were sounintegrated with the rest of George’s psychology that they could count as states


of the organism but not as states of George. We do not claim that it is clear thatthe pain was not George’s. Our point is that in the light of these alternativepossibilities it is not clear that it was.

In asserting that George remembers having felt the pain, the objector begsthe question. What George remembers now is the occurrence of pain, or theexperiencing of the pain. But from his remembering the occurrence of pain, orthe experiencing of pain, it does not follow that he remembers that he felt pain.

For there to be adequate grounds for the no-misidentification view, itcannot be perfectly sensible for some fully informed and fully rational person todoubt whether the pain was George’s. But imagine that one of the jurors isphilosophically sophisticated, as sensible and as rational as can be, and under-stands perfectly whatever is known about the physical and psychological detailsof the case. It is easy to imagine also that she does not think that the evidenceshows conclusively that the pain was George’s. We do not say that this judg-ment of hers is the only sensible response to this case. But surely it is a perfectlysensible response nonetheless. If the no-misidentification view were true, itshould not be.

Chapter 11

First-person reference, representationalindependence, and self-knowledge*

Christopher PeacockeUniversity of Oxford

My aim here is to identify a certain distinctive feature of some first-personthoughts, and to discuss its explanation and its significance. The challenge ofproviding a good account of the nature of first-person thought has drawn forthsome of the most justly celebrated and dazzling contributions of philosophy. Ithink that some of these contributions are dazzling not only in the intensity ofthe light they cast, but also in their havingmade it hard to see the true nature ofthis feature. The feature has often been mischaracterized in one way or another,and theories have been built on these mischaracterizations. Sometimes anappreciation that an earlier characterization is defective has been replaced by anew mischaracterization. Before I go any further, I had better say what I thinkthe feature is.

1. Representational independence

Consider first the everyday case in which a person forms a belief with thecontent ‘I am in front of a door’, and does so for the reason that he sees a doorahead of him. His visual experience represents him as bearing a certain spatialrelation to the door. This is so even if he cannot see or otherwise experience his

* I have been helped by discussions at the meeting on Direct Reference and Indexicals in March 1994at the Zentrum für interdisziplinaire Forschung of the University of Bielefield; at the Karlovy-Vary 1994conference on reference; and at my seminar in New York University in 1996. Comments from JohnCampbell, Paul Horwich, Thomas Nagel, Derek Parfit, John Perry, Stephen Schiffer, Paul Snowdon,TimothyWilliamson, CrispinWright, and Eddy Zemach have been particularly helpful. The later stagesof the preparation of this material for publication were carried out while I held a Personal ResearchProfessorship funded by the Leverhulme Trust, whose support I very gratefully acknowledge.

216 Christopher Peacocke

own body on this particular occasion— taking his experience at face value, hewould judge that he is in front of a door. Whether or not he perceives his ownbody on this particular occasion, the way the visual experience represents theworld as being is one which justifies acceptance of the first-person contentendorsed in his belief ‘I am in front of a door’. (We assume he possesses theconcepts in the content of this belief.) More generally, when a person forms aperceptual belief ‘I am F’, he does so because his experience itself has thecontent ‘I am F’, or has some content which justifies the content ‘I am F’. Thispoint would be agreed by theorists who otherwise disagree about what philo-sophical theory is to be given of an experience’s having a first-person content.I will return to that last point.

We can more generally consider examples in which the thinker’s reason formaking his judgment ‘I am F’ is his being in some state which represents acertain content C as correct. In the class of examples I want to consider, thecontent C may, but need not, be the same as the content ‘I am F’. Sometheorists believe that perceptual states have nonconceptual contents which aredistinct in kind from the contents which feature in beliefs (Evans 1982, thisvolume; Peacocke 1992). What I have to say is orthogonal to that issue. In theclass of examples on which I want to focus, it will be the case that the content C,even if it is distinct from the content ‘I am F’, is still one from which it isrational to move to the content ‘I am F’.

In our opening example, it is a perceptual experience which is the content-possessing state which gives a reason for forming the first-person belief. But thegeneral property I specified can be present in cases in which the reason-givingstate is not an experience (and still not a belief either). If you are sitting at yourdesk in the night, and a power cut turns out all lighting, you may still in thedark be able to keep track of your movements relative to the room you are in.You may walk across the room, and then form the belief ‘I am in front of adoor’. You have a faculty which allows you to keep track of your location, andthe exercise of the faculty on this occasion results in its seeming to you that youare in front of the door. This seeming is not a perceptual experience. Nor is it tobe identified with a belief. You may know that your tracking faculty mis-functions in a specific way. In those circumstances, your judgments mayoverrule the deliverances of your inaccurate faculty, and you may form thebelief that the door is probably two feet to the left of you. When, however, youare in the dark and you are taking the spatial faculty’s deliverances at their facevalue, your reason for forming your belief ‘I am in front of the door’ is a belief-independent state, and one which represents it as being the case that you are in

First-person reference, representational independence, and self-knowledge 217

front of the door.We can say that a use of the first person, in a particular belief ‘I am F’, is

representationally dependent if

i. there is a content C which is the content of one of the thinker’s currentmental states, a state which represents C as correct;

ii. the content C either is, or justifies, the content ‘I am F’; andiii. the thinker forms the belief ‘I am F’ by taking the mental state mentioned

in (i) at face value, in respect of its content C.

The uses of ‘I’ in the two beliefs ‘I am in front of the door’ in each of our twoopening examples, based on visual perception in the one case and the deliv-erances of the spatial faculty on the other, are both representationally depen-dent uses. In the usual case of knowledge of one’s bodily and spatial propertiesresulting from perception, clause (iii) will be fulfilled because in such cases thesubject is taking the content of the representational state at face value.

The characterization (i)–(iii) says what it is for a use of the first person inthought to be representationally dependent. We could similarly give a charac-terization of what it is for an utterance, on a particular occasion, of the first-person pronoun to be representationally dependent.

The second clause, that it is the content C which makes reasonable thejudgment ‘I am F’, is crucial to the notion of representational dependence.What is required is that the content C is one which (if indeed it is not the verycontent ‘I am F’ itself) can be certified as standing in some justificationalrelation to the first-person content ‘I am F’, without any need to mention thekind of mental state of which it is the content. We can assess the transition fromthe content C to ‘I am F’ in abstraction frommental states of which they may bethe contents. That one truth-evaluable content can stand in justificationalrelations to another is something which is a presupposition of such disciplinesas logic, probability theory and confirmation theory. For the notions of non-conceptual representational content that have been introduced in the morerecent literature, similar theories of justificational relations, both amongstthemselves, and in relation to other contents, could also be developed. Arepresentationally dependent use of the first person is so-called in part becausethe thinker is in a state which represents the content C as correct, and thecontent represented as correct is one which stands in such a justificationalrelation to the content of the belief ‘I am F’.

From the standpoint of someone engaged in giving an account of masteryand understanding, representationally dependent uses form a theoretical natural


kind. For these uses, there is the possibility of giving a partial account ofmasteryof these uses by describing the rational sensitivity of certain beliefs involvingthem to the content of the representational states with the justifying content.

Representational dependence is not a notion which is definitionallyrestricted to the first person. Quite generally for any mode of presentation m,we can say that a use of m, in a particular belief ‘m is F�’, is representationallydependent if there is a content C having the properties which result fromclauses (i)–(iii) above by substituting ‘m�’ for ‘I’. That is, the content C must bethe content of one of the thinker’s current representational mental states; itmust in itself make reasonable judgment of the content ‘m is F�’; and the thinkermust form the belief ‘m is F�’ because he is accepting the content C of thatrepresentational mental state. A use of a perceptual-demonstrative such as ‘thatcar’, in an ordinary perceptual-demonstrative belief like ‘That car is travellingfast’, is representationally dependent.

A representationally independent use of the first person, in a belief ‘I am F’,is one meeting this condition: the person who has the belief is in a state whichis his reason for forming the belief, but the conditions (i)–(iii) are not met; noris there a set of contents and set of mental states with those contents whichcollectively meet the conditions (i)–(iii). Representational independence is thecrucial notion which I want to put to work in this paper.

The uses of the first person in beliefs with the following contents can, inappropriate circumstances, be representationally independent:

I see the phone is on the table.I remember attending the birthday party.I remember that Russell was born in 1872.I am beginning to dream.I fear that the motion will not be carried.

To say that such uses are representationally independent is not to say that thebeliefs in question are not held for reasons. On the contrary, the thinker’sreason for believing ‘I see the phone is on the table’ is his visual experience as ofthe phone’s being there. It is arguable that there are also reasons in the othercases on this list. What matters for a representationally independent use of thefirst person is, rather, the nature of the reason, and its relation to the content ofthe belief in question. When a person predicates the complex concept

sees the phone to be on the table


of himself, it is not the content of the reason-giving experience, viz., that thephone is on the table, that makes rational a predication of seeing of himself. Itdoes of course contribute to the rationality of attributing a particular content tothe seeing. But the concept of seeing does not feature in the content of theexperience which represents the phone as being on the table (nor does anynonconceptual analogue of the notion of seeing feature in it). If we considerjust the relation between the representational contents of the experience and thejudgment themselves, and how their correctness requires the things andproperties to which they refer to be, then we should say this: the content

The phone is on the table.

does not in itself support the content

I see the phone is on the table.

It is true that a thinker might reflect on the conditions for his own use of aperceptual-demonstrative such as ‘that phone’ to be a correct use. He mightreflect on these conditions, and infer from them, together with the correctnessof his use of it on this particular occasion, that he must be seeing the phone. Butclearly an ordinary, nonphilosophical person’s knowledge that he sees thephone to be on the table is not reached via any such philosophical route. Theordinary person’s self-ascription of the experience rests on the occurrence of theexperience itself, rather than any philosophical reflection about conditions ofuse of demonstratives.

The normal case of psychological self-ascription contrasts, then, with therepresentationally dependent cases with which we started. When the content ‘Iam in front of a door’ is believed on the basis of perceptual experience, thecontents of the experience and of the belief based upon it are not merelyconfirming, they are identical, if we allow perceptions to have a first-personconceptual component. Even if we do not, whatever notion of nonconceptualrepresentational content is endorsed, it will not have much plausibility unlessthe content attributed to an experience when one is in front of the door isregarded at least as justifying the content ‘I am in front of a door’.

Still, it may be objected that the distinction between representationallydependent and independent cases falls apart when we recognize a level ofnonconceptual representational content of experience, or perhaps a moreprimitive kind of conceptual content, instances of which can justify ‘I am infront of a door’. A more primitive content C0 may for instance be one whichdoes not involve the possessor of a state with content C0 thinking of himself as


a persisting material object. Such a primitive content C0 might be captured veryroughly by the content of an utterance ‘There’s a door in front of this point’,where the demonstrative ‘this point’ refers to the point from which the subjectseems to be perceiving. When such contents are recognized, an objector maysay, then in the case which I have classified as representationally dependent,both of the following points hold. First, the contents C0 and ‘I am in front of adoor’ are distinct, as the content of the visual experience of seeing (which doesnot itself involve seeing) is distinct from the content of the judgment ‘I see thephone is on the table’. Second, the objector may say, in both the representa-tionally independent and the representationally dependent cases, there is a basisfor the resulting judgment. In both kinds of case, we have causal and rationaldependence of the final belief upon the occurrence of the visual experience, andin both cases, the causation involves reasons. In both kinds of case, the subject’sreason for his final belief is the occurrence of the justifying experience. So what,the objector asks, is the distinction?

I reply that even when contents such as C0 are acknowledged, there remainsa distinction between the cases. In the representationally dependent cases, wecan split the explanation of the subject’s final acceptance of ‘I am in front of adoor’ into two components. There is, first, his endorsement of the moreprimitive content C0 of his experience or other representational state. Second,there is the transition from C0 to the content ‘I am in front of a door’. Thislatter transition is ratifiable as legitimate independently of any endorsement ofthe initial content. (Similarly, in the different case of inference, we can ratify atransition from premises to conclusion as valid regardless of whether thethinker accepts the premises or not). The thinker’s willingness to make thislegitimate transition is an essential part of the explanation of his final belief. Itis crucial to the distinction I am advocating that the transition is one from acontent, something evaluable as correct or incorrect. The transition from theobjector’s content C0 to ‘I am in front of a door’ is also of this kind. By contrast,in the representationally independent case, there are not two correspondingcomponents in the explanation of the final acceptance of ‘I see the phone is onthe table’. There is in the normal case of psychological self-ascription notransition from one content to a second content, ‘I see the phone is on thetable’. Nor indeed in the general case need there be endorsement of the contentof the representational reason-giving state.

The justificational relations which hold between contents may depend upona subject’s situation and history. There may not be a necessary relation betweenthe correctness of ‘There is a door in front of this point’ and ‘I— a continuant,


material person — am in front of this door’. But for a person many aspects ofwhose thought depend upon his embodiment, and who has reasonably usedsuch forms of thought in the past, there will be an entitlement to make thattransition between contents. The fact that justification or entitlement candepend upon contingent facts about the thinker does not destroy the distinctionbetween representationally dependent and independent uses of the first person.

Four salient points about representationally independent uses of the firstperson emerge even from the introductory materials and distinctions we havebefore us so far.

i. Representational independence of a use of the first person is to be distin-guished from immunity to error through misidentification, the phenomenondefined in the discussions of first-person thought in Wittgenstein (1933–4),Shoemaker (1968, this volume), and Evans (1982, this volume). To reactivateour memory traces about this distinction, I quote Shoemaker’s characterization:

to say that a statement ‘a is Φ” is subject to error through misidentificationrelative to the term ‘a�’ means that the following is possible: the speaker knowssome particular thing to be Φ, but makes the mistake of asserting ‘a is Φ�’because, and only because, he mistakenly thinks that the thing he knows to beΦ is what ‘a’ refers to [Shoemaker 1968:557; this volume, 83].

Actually, this formulation would include the quite irrelevant case in which thethinker makes an error because he fails to understand the term ‘a�’ correctly. Solet us transform this definition to one concerning the level of thought, byreplacing ‘asserting’ and ‘statement’ by ‘believing’ and ‘belief ’. The kind ofmistake in question would then be that of believing that the thing he knows tobe Φ is identical with a. The variables are now used for modes of presentation,or thought-constituents.

As the connoisseurs of this literature will remember, Shoemaker also drawstwo further distinctions within beliefs which are immune to error throughmisidentification. One is the distinction between the circumstantially immune,and the absolutely immune. A judgment is circumstantially immune to errorthrough misidentification relative to the first person if it is immune to suchmisidentification errors when made on a certain basis (Shoemaker 1968:557,this volume, 83). When made on some other basis, it may be subject to errorsof identification. The other distinction is that between de facto immunity toerror through misidentification and stronger forms of immunity. A judgment‘I once played in a concert’ made on the basis of a memory, from the inside, asof playing an instrument in a concert, has only de facto immunity. In a world in


which people have apparent, and appropriately causally related, memories ofothers’ experiences, there would be no such immunity (Shoemaker 1970).

Shoemaker himself, and later Evans in The Varieties of Reference, made aconvincing case that the belief ‘I am in front of a door’, when judged on thebasis of an experience of being in front of a door, is immune to error throughmisidentification. This is an example of Shoemaker’s ‘circumstantial’ and defacto immunity: Belief in such a content is immune when made on the basis ofperceptual experience, in the circumstances of the actual world. Yet this was ourparadigm, initial example of a representationally dependent use in thought ofthe first person. So such immunity to error throughmisidentification of a first-person belief is not sufficient for it to be representationally independent.

A converse proposition is, though, very plausible, and supported by theaccount I will offer below. This is the proposition that all representationallyindependent uses of the first person in which the first-person belief is knowl-edge are also uses which are immune to error through misidentification.Certainly, a representationally independent use cannot be based on an identity-elimination inference from the contents of two other mental states. Such aninferential transition would require that the contents of the reason-giving statesgive reasons for his first-person belief, and that is contrary to the definition ofrepresentational independence.

When we have rather more in front of us, I will return to address thequestion of whether absolute, ‘logical’ immunity to error through misidentifica-tion is sufficient for representational independence. I believe it is not sufficient.

ii. Is infallibility at least necessary for representational independence? Our listof contents which, believed in suitable circumstances, involve representationallyindependent uses of the first person included contents of the form ‘I rememberthat p�’ and ‘I see that p�’. A thinker is not infallible about these contents.

It may be objected that such contents should not be on the list anyway. Itmay be said that in judging these fallible contents, the thinker takes for grantedor presupposes the first-person content that he is perceiving properly (in therelevant modality), or the first-person content that his memory is functioningproperly. I agree that these are taken for granted; but that does not imply thatthe examples should be excluded from the list of representationally independentuses. If the use of the first person in ‘I see that p�’ is representationally depen-dent, by what content C is it justified? Perhaps the proposal is that it is justifiedjointly by ‘I have an experience as of its being the case that p�’ and ‘I am perceiv-ing properly’. Here the former content, ‘I have an experience…’ would have to


be the content of a belief. This raises the following question.Can it ever be that a seeing justifies the one who sees in self-ascribing a

seeing? Or is it always the case that what is justified is, in the first instance, onlya self-ascription of an experience, from which one must infer, with the addi-tional information that one is perceiving properly, that one is also genuinelyseeing? It looks as if the latter will have to be the position of the person whoinsists that knowledgeable representationally independent uses of the firstperson must be infallible. But the position is highly problematic. In particular,it is difficult to square with a plausible view of ordinary perceptual knowledge.There would be widespread agreement that ordinary perceptual knowledge,while justified by perceptual experience, does not have to result from inferencefrom the thinker’s current experiences. A seeing can, in suitable circumstances,justify the thinker’s belief ‘That tree has leaves’. Why can it not, in the samecircumstances, equally justify the thinker’s belief ‘I see that tree has leaves’? Itseems that any doubt that it can would equally spill over into a doubt as towhether noninferential perceptual knowledge is possible either. It would clearlytake us far afield to pursue this here. We can, though, note the prima facieplausibility of this conditional: If noninferential perceptual knowledge of theworld is possible, then so also must be knowledgeable, representationallyindependent uses of ‘I see that p�’. If this is right, then infallibility is not neces-sary for representational independence.

iii. To say that a use of the first person is representationally independent is notin any way to imply that it does not refer. The characterization of a use asrepresentationally independent has to do with the absence of a certain kind ofreason for forming the belief in which the first-person way of thinking isemployed. The conclusion that representationally independent uses do not referwould follow if it were a necessary condition for a use of the first person to referthat the thinker’s reasons for accepting it be of the kind characterized in thedefinition of representational dependence. But I cannot see any reason forbelieving that, andmany against. One reason against is that forms of reasoninginvolving the first person, and whose validity requires that the first personrefers, seem valid regardless of whether the use of the first person in a premiseis representationally dependent or not. Nor in my judgment do we have anysatisfactory model for the semantic evaluation of first-person thoughts whichdoes not employ the notion of reference. Representationally independent usesof the first person, like all other uses, conform to the principle that they refer tothe thinker of the thought in which they occur. I will return later in this paperto nonreferential treatments of the first person.


iv. When a mental state provides a reason for a belief involving a representa-tionally independent use of the first person, the content of that state cannotfully explain what makes the first-person belief rational (when it is so). Thismuch is a consequence of the very definition of a representationally indepen-dent use. It follows that to find an account of mastery and understanding ofthese uses we must look elsewhere.

The phenomenon that I want to highlight is not just that there are truebeliefs which contain representationally independent uses of the first person.That contents of the sort listed in the examples above of representationallyindependent uses of the first person are on occasion true is not obviouslysomething which should by itself prompt a search for a philosophical theory.The mere truth of these contents does not depend upon the uses of the firstperson being representationally independent. They will have the truth valuesthey do regardless of how they come to be believed. The phenomenon I want tohighlight is rather the fact that beliefs containing representationally indepen-dent uses of the first person can, in suitable circumstances, be knowledge. Whatexplains this fact?

2. Delta theories

I now turn to consider a class of theories which aim to explain how beliefsinvolving representationally independent uses of the first person can amount toknowledge. Theories in this class differ from one another in the way in whichthey elaborate important aspects of the explanation, but the broad structure ofthe answer they give is common to all of them. According to these approaches,when a person self-ascribes an experience (say), it is the occurrence of theexperience itself which is part of the person’s reason for making a judgmentwhose content contains the concept of experience. That it is an experience hejudges himself to be having is rationally explained in part by the fact that he ishaving an experience. The case contrasts with a representationally dependentexample, in which the content judged does not go beyond what is justified bythe content of the experience.

Under this treatment of the representationally independent case, thethinker’s reason for making his judgment that he has an experience of a certainkind is not some thought to the effect that the experience occurs, the experienceas thought of under a certain Fregean Sinn. Any such thought is part of theexplanandum, not the explanans. The explanans is just the occurrence of the


experience itself to its subject. No thought or representation of himself as thesubject of the experience enters his reasons for his judgment. I think this is whatis right in Shoemaker’s remark that “When one is introspectively aware of one’sthoughts, feelings, beliefs and desires, one is not presented to oneself as a flesh-and-blood person, and one does not seem to be presented to oneself as anobject at all” (Shoemaker 1984:102). There is no reason to object to theformulation that the thinker judges the content because he has the experience,for this explanans does not involve any thought or conception of himself ashaving the experience in the statement of his reason for making the judgment.It is worth noting that in that minimal sense, we can equally truly say of anobservational judgment about the external world that the thinker makes itbecause he has an experience of a certain sort. But when a thinker makes anobservational judgment, his reasons for making it do not need to include somethought to the effect that he has a certain type of experience.

Now the occurrence of an experience, or of any other particular consciousstate, can give an immediate reason for forming a belief (or doing anything else)only to the subject who has the experience or conscious state. Suppose a personhas a visual experience as of the phone being on the table, and that the occur-rence of this experience is his reason for forming the belief ‘I have an experienceas of the phone being on the table’. Then the owner of the experience will be thesame as the subject referred to in the first-person component of the belief. Thesituation can be diagrammed as shown in Figure 1.

Because of the shape of this diagram, I call accounts of the general kind I

coconsciousness relation

judgment‘I ψ that p’

conscious mental state or eventof type ψ and content p

the subject

relation ofreference

relation ofownership by

Figure 1.

have been outlining delta accounts. Delta accounts may differ over how it is that


a conscious state can provide a reason for forming a belief, notably overconscious beliefs and other propositional attitudes, which have been found farmore problematic than the self-ascription of experiences. I will return to thesematters; our present concern is with the explanation offered by delta accountsof the representationally independent phenomena in question.

Why is a first-person belief formed in the way described in a delta accountalso a true belief? The explanationmust involve the following elements: the factthat when a belief is formed in this way, the owner of the conscious state will bethe reference of the first-person component of the self-ascription; and someexplanation of why the person is right about the kind of mental state it is whichhe has. The explanation of the correctness of the belief formed does not need toinvolve the correctness of the first-person content of some other representa-tional state. It may do so if the attitude in question is factive, as with self-ascriptions of genuine perception, knowledge, or memory (and in those casesit must involve more, too). But it is only because these attitudes are (at least)factive that these elements must be involved in the explanation of correctness ofthose particular kinds of self-ascription. They will not, for instance, be presentin accounts of the correctness of self-ascriptions of belief or experience.

This contrasts sharply with the explanation to be given of why a representa-tionally dependent first-person belief based on perception is true. It would beimpossible to explain why judging ‘I’m in front of a door’ on the basis of aperception with that first-person content (or whatever justifies it) yields a truebelief, in the ordinary circumstances in which it does, without alluding to thecorrectness of the first-person representational content of the perceptualexperience (or to the correctness of that content’s nonconceptual ground). Acorresponding point holds for first-person beliefs formed as a result of use ofthe faculty we mentioned for keeping track of location in the dark, and for anyother method of reaching representationally dependent first-person beliefs.

We need, though, an explanation not just of truth but of the status of theresultant beliefs as knowledge, when they are formed in the way described by adelta account. Delta accounts have the resources to meet this demand, and todo so in a way which goes beyond mere reliability requirements, to meet thedemand for justification and rationality. Consider a thinker who makes thetransition from a conscious state or event of type ψ with content p to the belief‘I ψ that p,’ where his reason for forming the belief is the occurrence of this veryconscious state or event. It is an a priori truth that any use of the first-personconcept in thought refers to the thinker of the thought. So we can conclude ona priori grounds that when the first-person belief about his mental state is


formed for these reasons, it will be true. That is, from facts about the nature ofthe first-person concept, we can infer that transitions of this sort, when madefor the specified reasons, will lead to true beliefs. This is somewhat analogous tothe way in which we can infer from the nature of the logical concepts involved,and the way their semantic values are fixed, that certain logical transitions arevalid. In both the delta case and the logical case, what underwrites the rationali-ty of the transitions, and what would be mentioned in a full account of whythey are justified, goes well beyond what is involved in mere de facto reliability.

This is part of the way in which a delta-theorist can discharge the obligationto show that the relevant self-ascriptive beliefs are knowledge, but it is notmeant to imply that we do not equally have justification and knowledge in therepresentationally dependent cases too. We do. The point is rather that anaccount of how the status of knowledge is achieved in the representationallydependent cases will be rather different. It must, in the representationallydependent cases, involve an account of how the thinker is justified in taking thecontent of the initial state at face value. It will also not involve the same kind ofpivotal and special role of the a priori principle that any use of ‘I’ in thoughtrefers to its thinker, in the way that there is such a special role in the treatmentjust outlined of justification for the representationally independent cases.

Any delta account must be filled out with a detailed theory in the philoso-phy of mind and epistemology of how conscious states justify self-ascriptions.It would take us too far from representational independence to pursue thisthoroughly here. But I do want to indicate a few aspects of the location of sucha position in the space of possible positions, in a way intended to indicate thatthe idea is not a total nonstarter.

Delta accounts are not required to assimilate conscious occurrent proposi-tional attitudes to the different category of sensations and perceptual represen-tational states. Conscious occurrent attitudes form a category of consciousstates in their own right. A delta account is committed only to the modest viewthat a conscious occurrent propositional attitude is something which can givea thinker a reason for forming another attitude. To deny this would be to countordinary conscious inference as impossible.

In conscious inference, a thinker’s acceptance of certain premises is part ofthe cause and reason for his acceptance of a certain conclusion. Rational thoughtinvolves one’s new judgments being sensitive to which contents one alreadyaccepts. In a self-ascription of an attitude involving a representationally inde-pendent use of the first person, we have sensitivity not just to the content of theconscious attitudes one already possesses, but towhich attitudes they are. Indeed


conscious practical reasoning already involves some primitive sensitivity to thedistinction between one’s own beliefs, one’s desires, and one’s intentions. Theseshould be relatively uncontroversial truths, and acknowledging them cannotinvolve assimilating conscious attitudes to the sensational or to the perceptual.

According to the delta accounts, a thinker’s reason for judging (for in-stance) ‘I remember attending the party’ can be his prior conscious apparentpersonal memory of attending the party. The transition this involves from anevent with a certain content that p to one with a corresponding content ‘Iremember doing so-and-so’ should not be construed as an inference from thememory to the self-ascription, any more than the transition from a perceptualexperience to a belief should be construed as inferential. Rather, there is arational sensitivity to the distinction between those of one’s states which arememories, and those which are not, a sensitivity which is already employed inordinary first level conscious thought and practical inference. This preexistentsensitivity is exploited by someone who has the concept of memory, and self-ascribes personal memories in the manner given in a delta account. Similarremarks apply to other conscious attitudes.

In short, the required theory in the philosophy of mind will treat theknowledge expressed in representationally independent psychological self-ascriptions neither on a perceptual model, nor an inferential model, but as athird kind of case with its own distinctive characteristics. Further discussion ofthese issues can be found in Peacocke (1997b).

Two disclaimers should be made on behalf of delta accounts. There isindeed a sense in which a delta account explains how a representationallyindependent ‘I think’ is capable of accompanying any of a thinker’s consciousstates. But a delta account cannot amount to an elucidation of the relations ofownership or of co-consciousness. On the contrary, delta accounts simplypresuppose those relations. The most that can be said is that if a delta accountis correct, a good explication of these relations must leave room for it.

The other disclaimer is that the delta account offers nopositive support for ‘noownership’ theories of mental states. A misconstrual of the phenomenon ofrepresentationally independent usesmight encourage no-ownership theorists, butit would take a misconstrual to do it. The phenomena as identified so far are atthe level of sense, not reference. A delta account explains the status as knowl-edge of representationally independent beliefs while agreeing that consciousstates are owned. So if anything, delta accounts are rather in the first instanceammunition for the opponent of the no-ownership theorist. These highlydistinctive phenomena can be explained without resort to a no-ownership account.


The no-ownership theorist may in fact enter these discussions as anobjector to what has so far been said. If the no-ownership theorist is right inthinking that there is some important respect in which experiences and otherconscious events are not owned, then clearly there is a need to say much moreabout the transition from a conscious event itself to a thought in which somemental property is self-ascribed. If the self-ascription is taken straightforwardly,as involving reference to a subject, the no-ownership theorist may worry thatthe transition is unsound. He may say that delta accounts give a good explana-tion of why some beliefs containing representationally independent uses of thefirst-person amount to knowledge, if it is taken for granted that conscious stateshave to have subjects. If there are subjects of conscious states, the thinker of theself-ascription must, for a priori reasons, be the same as the subject of themental event providing his reason for the self-ascription. But this justificationof an identity presupposes the existence of such a subject, and this is preciselywhat the no-ownership theorist questions. There is certainly an issue to beaddressed here, and I will return to it in Section 5. First we need to understandrepresentational independence and its consequences better.

3. Representational independence outside the first person?

It is not obvious that the abstract structure exhibited by a delta account isrestricted to judgments involving the first person. We can hope to learn moreabout the phenomenon of representational independence by consideringwhether the phenomenon extends beyond the first person. The natural place tolook is among the other indexical modes of presentation.

The abstract structure of a delta account involves three relations and threeterms. One term is an event of a certain kind, or a state of affairs, and thesecond is a judgment made rational by the first term. For each of these twoterms, there are two different relations, corresponding to the lower sides of theinverted delta, in which they stand to the same thing, corresponding to thethird, lowest node of the inverted delta. There is also a third relation betweenthe first two terms. This triangular structure seems to be instantiated if we take,for instance, a particular event — say, a particular political demonstration —as the first term; a perceptual demonstrative judgment ‘That demonstration ishappening now’ as the second term; and the time t of the event as the thirdterm. The corresponding relations are given in Figure 2 below.We could set upa similar example for a spatial judgment ‘That demonstration is happening


here’, if the spatial demonstrative is sufficiently broadly understood. But I will

the time t

cotemporalityrelation

has as its temporalreference

occurs at

the judgment‘that demonstration is

happening now’the event

(the political demonstration)

Figure 2.

concentrate on the temporal example, since similar points will apply to thespatial case.

The temporal diagram instantiates the same structure as that for the first-person case, but does the situation it depicts really result in a belief which is notmerely knowledge, but also involves a representationally independent use of‘now’? A problem surfaces at the top left-hand vertex of the temporal diagram.It is not just the event of the demonstration itself, it is rather the thinker’sperception of the event, which is his reason for making the present-tensejudgment. But when we bring in the perception of the event, we must remem-ber that the perceptual experience itself has a present-tense content. It repre-sents to the perceiver the event as occurring then — at the time of the experi-ence. It is because this is so that if we introduce a large time lag in the produc-tion of someone’s experiences, he is subject to a perceptual illusion. When thereis a large time lag, it looks as if things are occurring now to the subject of theexperience, when in fact they are not. This present-tense component of percep-tual experience undermines any claim of the use of the present-tense constitu-ent in the judgment ‘That demonstration is occurring now’ to be representa-tionally independent. In making this judgment, the thinker is simply endorsingpart of the present-tense content of his perceptual experience. So this use of‘now’ is after all representationally dependent.

We can envisage a subject for whom not only are there huge time delays inthe operation of the perceptual mechanism, but who also knows this to be hissituation. When this subject has an experience of representational kind T, hecan still use the temporal demonstratives [thatT time] of the sort I discussed in


Sense and Content (Peacocke 1983). This demonstrative refers to the time ofoccurrence of the events perceived, regardless of whether this time is identicalwith the time of occurrence of the perceptual experience. But I do not see howthese circumstances will yield us a representationally independent use of atemporal demonstrative which also results in knowledge. There are two kindsof case to consider. First, suppose a thinker who uses this special temporaldemonstrative happens to believe that, on this particular occasion, [thatT time]is in fact now. If he forms the belief ‘That event is happening now’ in thesecomplex circumstances, it will rest on the identity belief ‘[thatT time] = now’.The resulting belief will not be representationally independent, for it will rest inpart on a rational inference, one of whose premises contains the present-tensemode of presentation. In the second case, we can consider the same person whoknows of the real possibility of massive delay, and suppose he judges thecontent ‘That demonstration is happening at [thatT time]’. Is this use of [thatTtime] representationally independent? No, for in these circumstances, theperceptual experience does still represent the demonstration as occurring at[thatT time]. The position is structurally just the same as it is for the present-tense component ‘now’ in the normal situation.

An important difference between these temporal cases, then, and the first-person representationally independent uses is that when an experience, or someother conscious state, gives a thinker a reason for forming a belief, we do not—and for many independent reasons must not— regard the thinker as presentedwith that state or event only through perception. The relation between athinker’s judgments about some of his mental states and the states themselvesdiffers in this respect from the relation between a thinker’s judgments aboutexternal states of affairs and those states of affairs themselves.

This does not mean that representationally independent temporal index-icals cannot be involved in knowledge. On the contrary, when for example anexperience is knowledgeably self-ascribed, and the use of the first person isrepresentationally independent, the resulting judgment has a present-tensecontent: ‘I have such-and-such kind of experience now’. This use of ‘now’ isrepresentationally independent. But there do not seem to be autonomousexamples wholly underived from the mental cases. Certainly, the existence ofstructural delta-like diagrams like that above for the temporal case is not byitself sufficient to show that there are knowledgeable representationally inde-pendent uses outside the mental realm.

It begins to look as if two generalizations about representational indepen-dence are true. First, if a property is knowledgeably self-ascribed with a repre-


sentationally independent use of the first person, the property is either apsychological property or is inferred from such a self-ascription of a psychologi-cal property. Descartes’s use of ‘I’ in ‘I exist’, as the conclusion of the cogitoargument, is a knowledgeable self-ascription of a nonpsychological property,but it manifestly meets the condition of having been inferred from such a self-ascription of a psychological property, viz., that of thinking (in Descartes’srather general sense). Psychological self-ascriptions seem to have an explan-atorily prior role within the class of knowledgeable, representationally indepen-dent self-ascriptions. If this conjecture is true, it would be a vindication of thecommon intuition that there is a close connection between some mentalphenomena and a certain distinctive use of the first person.

The second general conjecture, given the earlier reflections about represen-tationally independent uses of other concepts, is that any knowledgeable,representationally independent use of a concept will be in a content which iseither a psychological self-ascription or is soundly inferred from one.

I have concentrated on knowledgeable, representationally independent usesof the first person in beliefs of a particular kind, viz., those which result fromself-ascription of a psychological state or event which is rationally based uponan occurrence of a conscious state or event of the very same kind as is self-ascribed. This is mainly because I think that such examples have been the mostinfluential historically. I do not, however, want to leave the impression that allknowledgeable, representationally independent uses are of this kind. Evans, forinstance, mentions an ascent-procedure for the self-ascription of belief: theprocedure of judging ‘I believe that p�’ if, putting into operation whateverprocedure he has for judging whether p, the subject comes to the conclusionthat p (1982:225; this volume, 114). Self-ascriptions of belief reached by thisprocedure will be knowledgeable, the use of the first person in them will berepresentationally independent; and it can be true that the self-ascription ismade for reasons (though of course these will be reasons for thinking the worldis a certain way). But it is also true that in these cases we can give the analogueof a delta-account. We do not, though, need to appeal here to the relation ofcoconsciousness which labels the horizontal line in our diagram. The verynature of the ascent procedure ensures that the referent of ‘I’ in beliefs self-ascribed bymeans of the procedure is the same as the person who has the beliefself-ascribed.


4. An illusion and its source

The writings of Kant, Schopenhauer, and the early Wittgenstein contain theclaim that there is a ‘transcendental’ or ‘metaphysical’ subject of experience andthought. This transcendental subject is said to be distinct from the ordinaryperson who has the experiences and thoughts. I want to argue that the notionof a transcendental subject of experience results from an unjustified andincorrect projection from the realm of sense to the realm of reference. There isa genuinely distinctive representationally independent use of the first-personconcept. It does not follow that such representationally independent uses of ‘I’refer to a distinctive kind of thing, something not referred to in represen-tationally dependent uses of the first person.

The transcendental subject in Kant’s thought, to remind ourselves, is notwhat is denied in Kant’s critique of ‘rational psychology’. Rational psychologywas said by Kant to have used unsound arguments for such properties as thesimplicity and immortality of the soul, but the critique given of rationalpsychology does not involve denying the existence of the transcendental subjectof experience. On the contrary, Kant’s critique is written from the standpointof one who endorses the existence of the transcendental subject but disputesthat the rational psychologist’s methods are legitimate means of coming toknow anything about it. It is also quite clear from the text elsewhere that Kantendorses the existence of the transcendental subject: ‘I have knowledge thatwhat is substantial in me is the transcendental subject’ (Kant 1781/7, B426–7;see also A355, footnote A478–9/B506–7, and A492/BS20). In any case, even if itsexistence were not explicitly endorsed, transcendental idealism could hardly beformulated without commitment to its existence.

Now suppose someone thinks that representationally independent, knowl-edgeable uses of the first person do refer but holds also that they refer tosomething which can be known about only in knowledge involving representa-tionally independent uses of the first person. What might one expect such aperson to hold about the entity to which he takes such uses of ‘I’ to refer?

I suggest that he would hold just about what Kant in fact held of thetranscendental subject. Kant wrote of the transcendental subject that

It is known only through the thoughts which are its predicates, and of it apartfrom them, we cannot have any concept whatsoever, but can only revolve in aperpetual circle, since any judgment upon it has already made use of its


representation. And the reason why this inconvenience is inseparably boundup with it, is that consciousness in itself is not a representation distinguishinga particular object, but a form of representation in general [A346 = B404].

This is a brilliant but difficult passage. I think that we can understand it betterusing the distinctions and theses I have proposed. I suggested that all knowl-edgeable self-ascriptions which involve a representationally independent use ofthe first person rest ultimately on psychological self-ascriptions containingrepresentationally independent uses of the first person. So if the idea of atranscendental subject of experience has its origins in the phenomenon ofrepresentational independence, it is understandable that one who believes insuch a subject should say that it ‘is known only through the thoughts which areits predicates’.

It is true that in this passage Kant mentions only thoughts, and not theother conscious states which can also be ascribed in representationally indepen-dent uses. But in other passages, ‘thoughts, feeling, desire, or resolution’ are allincluded as what Kant would call ‘predicates of inner sense, representations andthought’ (A358, A359). This suggests that Kant would have been quite happy,if pressed, to replace ‘known only through its thoughts’ in the displayed passageby ‘known only through its thoughts and other conscious states’.

The account of representationally independent use I have been giving alsoallows us to see a ground in that phenomenon for the claim Kant makes in thesecond part of the displayed passage. This is the claim that the reason why thetranscendental subject can be known only through its thoughts is that con-sciousness in itself is not a representation distinguishing a particular object, buta form of representation in general. Precisely what distinguishes a representa-tionally independent use of ‘I’ when a thinker knowledgeably thinks ‘I am F’ isthat the thinker’s reason for self-applying the predicate F is not that one of hisconscious states has the content ‘I am F’ — his reason is not given by a ‘repre-sentation distinguishing a particular object’ as F. It is rather the occurrence ofa certain kind of conscious state itself which is his reason for making thejudgment. It is the kind of state he is in, ‘a form of representation in general’which gives his reason for the particular psychological predication which isbeing made in thought.

This general diagnosis of the inclination to postulate a transcendentalsubject of experience also bears upon Kant’s view that the categories areinapplicable to this postulated subject. Kant wrote:


The subject of the categories cannot by thinking the categories acquire aconcept of itself as an object of the categories. For in order to think them, itspure selfconsciousness, which is what was to be explained, must itself bepresupposed [B422].

What is true is that in judging knowledgeably ‘I am thinking about such-and-such object as having so-and-so properties (involving the categories)’, or even‘I am thinking about the categories’, the use of ‘I’ is representationally indepen-dent. In such knowledge, the subject is not thought of in a way which makes itmanifest that it is subject to the categories. Equally, it is also true that anyknowledge of a content ‘m is F�’, where F involves one of the categories, andwhere this knowledge is based on the perception that m is F, will be representa-tionally dependent in respect of the use of the singular concept m.

It obviously does not follow that what a representationally independent useof ‘I’ refers to is something which is not subject to the categories. The slide fromtruths about the level of uses of the first-person concept, to very questionableclaims about the level of reference, is particularly striking in such phrases as this:

the transcendental subject of all inner appearances, which is not itself appear-ance and consequently not given as object, and in which none of the categories… meet with the conditions required for their application [footnote A478–9= B506–7].

From the fact that in certain self-ascriptions the reference of ‘I’ is not then givenas an object, it does not follow that it is not subject to the categories.

On the present diagnosis, the critique of Kant’s notion of the transcendentalsubject would at certain points be close to Kant’s own highly effective critiqueof rational psychology. Kant wrote that ‘the simplicity of the representation ofa subject is not eo ipso knowledge of the simplicity of the subject’ (A355).Equally, to employ in knowledge a representation of a subject which does notrepresent it as located in the spatiotemporal world is not eo ipso to haveknowledge that the subject is not located in the spatiotemporal world.

In effect, by appealing to a transcendental subject, Kant is saying that anexotic metaphysics can help to explain a genuine epistemological phenomenon:the existence of knowledgeable, representationally independent uses of the firstperson. It was a deep insight that these uses exist. It was also an insight that,here as elsewhere, the epistemology and the metaphysics of the first personmust be integrated. Nonetheless, Kant’s proposed solutionmakes things worse.Presumably we must attribute his departure from the standards he rightlyapplied against the rational psychologist in part to his prior acceptance of


transcendental idealism.The views of the early Wittgenstein on what he called ‘the metaphysical

subject’ (Wittgenstein 1921:5.641) were clearly somewhat different from thoseof Kant, and it is certainly an interesting question what exactly they were. Butbecause the texts are so brief, this would have to involve exegetical speculationand hypothesis that would take us too far from our main present concerns. Iturn instead to address this question: What is the most fundamental explana-tion of the dual temptations either to suppose that in some of its uses, ‘I’ doesnot refer at all, or that it refers to something extraordinary, something otherthan a person?

In The Blue Book, Wittgenstein’s later self famously explained the tempta-tion as resulting from amisconstrual of the genuine phenomenon of immunityto error through misidentification of certain first-person judgments (Wittgen-stein 1933–4: 67). This explanation was illuminatingly elaborated by Shoemak-er; and further investigated by Evans. Immunity to error through misidentifica-tion is a necessary part of the explanation; but I want to argue that it is not thefull explanation, nor is it fundamental.

Suppose a thinker makes a perceptual-demonstrative judgment ‘that[perceptually given] F is G’, and in doing so applies an observational concept,on the basis of his current experience, to an object presented in the sameexperience. When such a judgment comes to be made in that way, it is immuneto error through misidentification relative to the demonstrative component‘that F’. The judgment does not rest on any identity belief ‘that F is identicalwith such-and-such’, and so there is no room for error resulting from the falsityof such a belief. Such immunity from error through misidentification does notin these cases involve any temptation to regard the object demonstrativelythought about as ‘not in the spatiotemporal world’ — on the contrary. So thesource of any temptation in the first-person case cannot be immunity to errorthrough misidentification alone. Something else must be involved.

Is the explanation then the combination of immunity to error throughmisidentification with the infallibility of the judgment made, given the reasonsfor which it is made? We already know this combination cannot be a necessarycondition for the illusion to arise, if it be granted that self-attributions ofmemory and knowledge give rise to it, for those attributions are not infallible.The combination of immunity to error throughmisidentification and infallibil-ity is also not sufficient for this kind of illusion either. Consider the judgment‘This pain is acute’, where the thinker’s reason for making the judgment is theacuteness of the pain he experiences. This is both immune to error through


misidentification relative to the demonstrative component ‘this pain’, and it isinfallible if anything is. But there is not the same strength, or kind, of temptationthat there has been in the first-person case to suppose that ‘this pain’ does notrefer, or refers to something extraordinary. The same example shows that addinga third requirement to the explanation, to the effect that in the circumstances theterm in question is guaranteed to have reference, is also not yet enough.

One feature differentiating those self-ascriptions which have led to prob-lematic claims about the first person from these most recent cases is this. In therecent cases, a certain experience, either sensation or perception, is required forthe demonstrative component ‘that F’ — the perceptual demonstrative — or‘this pain’ to become available to the thinker. But in the former cases, there isno experience, neither perception nor sensation, whose presence is requiredbefore the first-person becomes available for such uses in thought. When aperson makes a judgment of the form ‘I am in mental state M’, and does so forthe reason that he is in the conscious mental state M, it is not that mental stateM which makes available the first-person component of his thought. It couldnot be, since the very same first-person component is employed in the differentcircumstances when he knowledgeably judges ‘I am not in mental state M’. Thischaracteristic differentiates the first person from both perceptual demon-stratives and demonstrative thought about one’s own sensations. We can labelthis characteristic by saying that ‘I’, unlike these other demonstratives, isexperientially independent.

Uses of ‘I’ in certain knowledgeable present-tense psychological self-ascriptions have then at least these three characteristics in the circumstances inwhich they are made:

i. they have guaranteed reference;ii. they are immune to misidentification errors; andiii. they are experientially independent.

Even with this threefold combination, however, it is hard to believe we havereached the source of the illusion that in some of its uses, ‘I’ either does notrefer, or refers to something extraordinary. Properties (i) and (iii) on this list,i.e. guaranteed reference and experiential independence, are present in any useof the first person in thought, in whatever circumstances and with whateverpredicate the first person is combined. As we noted, the second property on thislist is present, at least in the form of what Shoemaker calls ‘circumstantial’immunity, for such perceptually based spatial thoughts as ‘I am in front of adoor’, and these thoughts, par excellence, locate one in the objective order. It is


not as if, when we confine our attention to the circumstances in which suchspatial thoughts are immune to error through misidentification, we are moretempted by the illusion—we are not. In any case, it is not particularly transpar-ent why the combination of properties (i) to (iii) should lead to this particularsyndrome of illusions about the first person.

I suggest that a better explanation of the illusion is one which makesessential reference to the fact that uses of the first person which produce theillusion are representationally independent. There is no pull towards theillusions in uses of the first person in judgments which are representationallydependent, and whose justifying mental states have a representational contentwhich clearly represents the subject as an item in the objective spatio-temporalorder. As we just noted, there is no such pull even if the resulting judgment iscircumstantially immune to error through misidentification. In knowledgewhich involves a representationally independent use of the first person, on theother hand, the subject’s reason for making the judgment is not one whichinvolves the exercise of the conception of himself as located in the objectiveorder. He must of course be an element in the objective order, for many anincoherence would result from the supposition that he is not. It may also bearguable that he must have some conception of himself as located in theobjective order. Nonetheless, no such conception is actually exercised by thethinker when he makes a knowledgeable, representationally independentjudgment about himself, and a delta account makes clear how this is possible.

5. Is self-knowledge subjectless?

I turn now to the task postponed at the end of Section 2. This is the task ofconstructing an account which simultaneouslymeets two desiderata. It address-es the objection we imagined coming from the theorist who holds that con-scious mental events are fundamentally subjectless. It also gives an evaluationof the transition from the occurrence of a conscious event or state to thecorresponding self-ascription.

The account must have at least three strands. It will have to have a meta-physical strand if it is to address the claims of the no-subject theorist. It willhave to have an epistemological strand, in its assessment of whether thetransitions from psychological states to their self-ascription are justifiedtransitions. Finally the account will have to draw on part of a theory of contentfor thought and language, for a theory of content is what is needed to explain


any links there may be between the metaphysics and the issue of justification.We can distinguish at least three different kinds of account of the transition

from a conscious state to its self-ascription.

1. According to accounts of the first kind, the transition from a psychologicalstate to its self-ascription is justified, but at some fundamental level, there is nogenuine reference made with ‘I’ in thought or language in the self-ascription. Inthis latter respect, positions of the first kind are in agreement with the viewexpressed in Wittgenstein’s Philosophical Remarks, when he wrote

One of the most misleading representational techniques in our language is theuse of the word ‘I’, particularly when it is used in representing immediateexperience, as in ‘I can see a red patch’…. It would be instructive to replace thisway of speaking by another in which immediate experience would be repre-sented without using the personal pronoun; for then we’d be able to see thatthe previous representation wasn’t essential to the facts….We could adopt thefollowing way of representing matters: if I, L.W., have toothache, then that isexpressed by means of the proposition ‘There is toothache’…. It’s evident thatthis way of speaking is equivalent to ours when it comes to questions ofintelligibility and freedom from ambiguity [Wittgenstein 1975: §§57–8].

Of course, the later Wittgenstein would almost certainly not be happy with aconception of conscious events under which they can be reasons for a thinker’sdoing something, including forming a belief. Positions of this first general kindare adopting only the quoted idea of the first person as nonreferential, andneed not be accepting either the middle or the later Wittgenstein’s views onexperience.

2. Positions of the second kind hold that the function of the first person inthought and language is indeed to make reference. For just this reason, thesecond position continues, the transition from the occurrence of a consciousevent to a corresponding self-ascription is not justified. This is precisely thestance of Lichtenberg’s criticism of Descartes’s cogito, that Descartes wasentitled only to the premise ‘There is thinking going on’.

3. While accounts of kind (1) aim to validate the relevant transitions to a self-ascription by reducing the content of the self-ascription, accounts of kind (2)do not reduce it. Accounts of both of these first two kinds are implicitly takingit that one cannot simultaneously have both the correctness of the transitionand take it that the function of ‘I’ is to refer. The third species of position rejectsthis implicit assumption which is common to the first two positions.


Positions of the third kind agree with the second position that the function ofthe first person is to refer, but they agree with the first position that the transi-tion is justified. Accounts which are of this third kind subdivide into thosewhich we can classify as concessive to the no-subject theorist and those whichare nonconcessive. The concessive accounts agree that conscious mental eventsare fundamentally subjectless but aim to construct a referent for the first personfrommaterials available at that allegedly subjectless level, and in such a way thatthe transition is justified. The nonconcessive accounts deny the existence of anysubjectless level, and aim to say why the transition is justified without presup-posing the existence of any such subjectless level. I will be advocating a non-concessive form of this third kind of position.

The very idea of a conscious state or event involves the existence of a subjectof the state or event. A conscious state, as Nagel (1979) said, is one such thatthere is something it is like to be in that state; and this must mean something itis like for the subject. Reference to the subject of conscious states is essential inelucidating what it is for them to be conscious. The subject here cannot bedismissed as a mere intentional object. It is not just as if there is something it islike for the subject — there really is something it is like for the subject, whenenjoying a conscious state. It is not at all easy to see how to give an account ofwhat it is for a state to be conscious in a wholly impersonal fashion.

BernardWilliams has emphasized that from the holding of the two ‘imper-sonal’ conditions that there is a thinking that p and there is a thinking that q, itdoes not follow that there is a thinking that p and q (Williams 1978:95–101).For someone who accepts the link just stated between consciousness and theexistence of subject, this gap is just what one would expect. For there to be aconscious thinking that p is for there to be a conscious state or event such thatthere’s something it’s like for its subject; and for there to be a consciousthinking that q is for there to be a conscious state or event such that there’ssomething it’s like for its subject. It does not follow that these are the samesubject, and so does not follow that there is a conscious thinking that p and q.

If the consciousness of a state has to be elucidated in terms which requirethe existence of a subject of that state, certain kinds of reductionism aboutsubjects will be undercut. For certain reductionists about the subject of experi-ence and thought, conscious events and states are supposed to be the verybuilding blocks in terms of which the existence of subjects is analysed. Ifconsciousness itself has to be elucidated in terms which already require theexistence of a subject, the materials drawn on by the reduction would bepresupposing what the reduction purports to elucidate.


In response to this objection, the reductionist may pursue certain analogies.An academic society, or a club, may, and perhaps must, be involved in events ofsuch a kind that only a society or club could be involved in them. This wouldhold for the admission of new members, changing its rules, and any other caseof taking action as a society or club. This evidently does not rule out thepossibility of an ontological reductionist’s attitude to these societies and clubsas sets of persons related to one another in complex ways. Yet structurally, thissituation is very different from the challenge which the connection betweenconsciousness and subjects poses for certain kinds of reductionist. The personsin terms of whom the reductionist wants to explain the existence of societiesand clubs are not themselves plausibly analysed in terms of their relations tosocieties and clubs.

These reflectionsmay have a very Cartesian feel about them. Indeed, if theyare correct, they support Descartes on two points. First, Descartes did, just fromhis enjoyment of conscious states, have a right to the premise of the cogito.Second, the transition he made from the premise to the conclusion ‘I exist’ wasa sound transition. But the reflections so far obviously do not license all ofDescartes’s conclusions about the nature of the referent of the first person. Thenature of the subject whose existence is entailed by the occurrence of consciousstates is rather something for further investigation.

Some conclusions can be drawn, though, just from the austere descriptionof the subject as the enjoyer of conscious states.

a.�If the subject is capable of remembering its own earlier states, it must besomething capable of persisting through time. This follows just from the factthat correctness of a memory requires that the subject of the memory state —that is, the one who is remembering — be identical with the one who in thememory is represented, from the first-person viewpoint, as having been thus-and-so. Since the subject who has current experiences is identical with thesubject who remembers, it follows that the subject who experiences must becapable of persisting through time. This subject must have the requiredcomplex capacities for memory storage, retrieval, and temporal representation.

b.�Some conscious states can give a subject reasons for making spatial judg-ments about itself, such as ‘I am underneath this tree’. In suitable circumstanc-es, the judgments so made amount to knowledge. When they are knowledge, itfollows that the subject of experience must have a spatial location in the world.When the spatial judgment is knowledge, the perceptual experience whichmakesit rational does notmerely support a weaker conclusion of the form ‘Something


bearing such-and-such relation tome is underneath this tree’. It supports a first-person content which locates the subject itself in the spatial world.

c.�It is the same subject who both experiences and acts (indeed some forms ofthinking are actions). Since the subject of conscious states is also identical withthe subject who acts, the conscious subject who is also an agent must have thecomplex of causal powers required for attempts and for actions based onreasons.

d.�There will be constitutive links between certain types of contents featuringin the mental states a subject may enjoy, and the subject’s environmentallyindividuated relations and capacities. This is the general thrust of externalistdevelopments in the philosophy of mind and language.

In the theses (a)–(d) we have the starting point for an investigation whichproceeds from relatively Cartesian data about the properties of the subject ofexperience, and by way of various constitutive principles about those properties,draws substantial conclusions about the nature of the subject of experience.There is a great deal more to be said about all of (a)–(d), in particular about theidea of a person as a persisting object in the world. Further pursuit of theinvestigation may also illuminate the issue of why there is an internal connec-tion between consciousness and the existence of subjects. At this late stage ofthe paper, however, instead of pursuing that project, I confine myself to thebearing of these reflections upon the transition from a conscious state to its self-ascription, and upon the kinds of position (1)–(3) distinguished above.

These reflections seem to me to support the nonconcessive form of thethird of the three kinds of position distinguished, that is, the position accordingto which the transition in question is justified, and there is genuine referencewith the use of the first person in these self-ascriptions. The transition isjustified in part because if these reflections are correct, conscious states alreadyinvolve the existence of a subject of the conscious states. The nonconcessiveform of the position is supported because there is no subjectless level ofconscious phenomena from which a constructed reference for the first personmight be built up.

Lastly, that it is the subject of conscious states to which the first-personrefers, when used in the self-ascriptions, follows from two other truths wenoted. One of the truths was the a priori principle of the theory of reference andcontent, and it states that any use of the first person in thought refers to thethinker of the thought. The other truth is that the subject of conscious statesand experiences which give immediate, non-inferential reasons for self-ascrip-


tions is the thinker. From this totality of principles it follows that any use of thefirst person in a self-ascription of a conscious state rationally made because sucha conscious state occurs must be a use which refers to the subject of the con-scious state. This is a fuller account of why the transition is justified.

Wemight conceive of an objector who complains that though the consider-ations to date show there must be a subject of conscious states, the first personin thought and language does not refer to it. Given the reasoning we haveemployed to reach our conclusion, this objector will have to deny either that thefirst person in thought refers to the thinker of the thought, or that the subjectof conscious states is also the thinker. Neither of these options is attractive.

The same considerations tell equally against positions of the first kind, thepositions which agree that the transition is justified but say that there is nogenuine reference made with ‘I’. The arguments have already been given, but itis worth reflecting on how they get a grip against the positive proposal that islikely to accompany positions of the first kind. Defenders of the first kind ofposition will be tempted by apparently subjectless formulations, such as theprinciple that a particular thought, on a particular occasion, with the content ‘Iam in pain’ is true just in case there is a pain coconscious with that thought.Since coconsciousness presupposes consciousness, and that according to thepresent reflections presupposes a subject, the objection would be that this isindeed only apparently a subjectless formulation. What is required for such aparticular thought ‘I am in pain’ to be true involves the existence of a subject ofthe pain. These points are all on the side of the truth condition of ‘I am in pain’.They enter again when we consider the justification of the transition to athought ‘I am in pain’, for it must involve the occurrence of the pain, and this,according to these reflections, must involve a subject, the same subject as thethinker of the thought.

These points are entirely consistent with the recognition of some genuinelyfeature-placing thoughts or sentences. It is indeed true that one can no moremake sense of the possibility of rain without a location than one can of painwithout a subject. The truth of this observation should not lead one to construethe ‘it’ in ‘It is raining’ as making reference to a place. The reference to a placeis rather provided for by words and phrases following ‘It’s raining’, like ‘here’,‘in Oxford’. It is this position which also accepts quantifiers, as in ‘It’s rainingeverywhere in Europe’. The metaphysical truth that there is no rain without itslocation is entirely consistent with treating these phrases, rather than thegrammatical subject, as the referential devices. In psychological self-attribu-tions, by contrast, it is the grammatical subject which occupies the slot that can


equally be filled with other singular terms and with quantifiers.1

The second kind of position on the transition was the Lichtenbergianstance that ‘I’-thoughts are true only if ‘I’ refers, and the transition is unjusti-fied. It is equally undermined by the considerations I have offered. From thestandpoint I have been developing in this paper, it is natural to suspect that theLichtenbergian position results from an illegitimate slide. There are twodifferent relations in which a thinker may stand to a given mental state whichprovides the reason for a particular thinker’s psychological self-ascription. Forthe thinker who is in the conscious state, that state can give a reason forjudgments and other actions without his having to identify the owner of thestate in advance at all (not even in a first-personal way), and without hishaving to think about the relations in which he stands to it. All the same, thestate is his, it can provide reasons for him only because it is so, and it willcommonly be thought of as possessed by a particular thinker when otherpeople think about the state. It may also be thought about in indirect wayswhich differ even more from the way in which it is given to its owner. Thenonsequitur is to move from these very different agreed relations in which itsowner and other thinkers stand to a given conscious state, to the conclusionthat there are two different states of affairs here, ‘one more substantial than theother,’ in Williams’s phrase (1978:95) — one of them involving a thinker, andthe other not. An appreciation of the way in which a subject’s conscious statesdo not themselves involve his thought of himself as having them should not bepumped up into the claim that a conscious state could occur without irreduc-ibly having an owner. The delta accounts of Section 2 above themselves showhow the transition from simply being in a conscious state to the judgment thatone is in it is warranted, and they do so without commitment to any level ofunowned thoughts and mental states.

The positions (1)–(3) are all at least in part about the level of reference.However not every writer who has insisted upon the ‘subjectless’ character ofcertain psychological self-ascriptions has beenmaking a claim about the level ofreference. Some have been making claims primarily about the level of sense.The claim has been that in the first-person self-ascription, no concept of oneselfis or need be deployed, or perhaps no concept of a certain kind, or again noconcept that involves thinking of oneself as one object among potentially manyothers. Perry (1990:1993) develops one such position. Such theorists need not

1. For further pertinent grammatical considerations toward the same conclusions as those reached inthis paper, see Katz (1987).


have any quarrel with the theses of this paper. It is open to at least some of themto accept a delta account. Their claims about concepts should not, though, beseen as simply orthogonal to the direction of this paper. A psychological self-ascription made for the reason that the thinker is in a certain conscious staterequires the thinker to have the concept of the psychological state in question.It is a substantive and important philosophical question, and one of pertinencealso to psychology, what kind of thing a person must think of himself as beingif he is to be able to ascribe psychological states to himself. It is not at all clearthat a positive account of mastery of psychological notions can leave any roomfor ‘no-concept’ treatments of the first-person element in psychological self-ascriptions. The issue should be on our agenda.

Chapter 12

The constructed and the secret self

William SeagerScarborough College, University of Toronto

In a justly famous article, Bernard Williams (1970) opposes two decisivelydifferent descriptions of what seem to be identical situations. The first descrip-tion is that of the science fictional case of ‘swapping selves’: we are to imaginea machine that can somehow extract information about the physical bases ofmemories, character traits, intellectual and creative abilities, etc. from a humanbrain and then ‘transfer’ these to another brain (presumably by reconfiguringit to encode the information retrieved from the first brain). Thus the machinecould also exchange all mental characteristics between two brains. To make thethought experiment vivid, Williams imagines that a pair of persons, A and B,must choose — and they must choose selfishly and before the exchange —between either receiving a million dollars (Williams’s example is actually$100,000, but we must allow for inflation) or being severely tortured, wherethese choices come into effect after the operation. But A and B can only choosewho receives these outcomes by pointing to the present body of either A or B.Thus if A thinks that he will be transferred to B’s present body, he will do wellto choose that his own present body be the one that will be tortured and that theperson who will inhabit B’s present body shall be the one to receive the milliondollars. Of course, B’s situation is the same as A’s, but reversed.

The other description is less hopeful. Here, A is told that he shall betortured tomorrow, but only after a series of preliminary indignities areinflicted upon him. He shall have his memories erased, his character, abilities,etc. effaced, followed by the implantation in his brain of a set of false memories,a false character, etc. Then he shall be mercilessly tortured.

The point is that, leaving aside tendentious talk of ‘transferring the self ’, orquestion begging accounts of who shall be tortured, both descriptions can beseen to describe the same set of events. On the other hand, it seems unquestion-able that after the operation there is a fact of the matter about who is who.Either A is still in the A-body, or he is in the B-body (or perhaps, if unlikely, he

248 William Seager

is nowhere, having been destroyed by the operation). Reflection on this thoughtexperiment provides strong grounds for the former disjunct if we antecedentlysubscribe to some form of materialism by which one’s memories, character,abilities, personality traits, etc. all depend upon the brain.1 After all, we allowthat memories can fade, be manipulated or altered and that one’s character canbe radically changed, and we know that such alterations can occur as the resultof physical changes in the brain.2 It seems a good principle that when a rangeof some object’s properties depends upon an underlying substrate then theidentity of the object goes with the identity of the substrate rather than with theproperties themselves. Such a principle is not universally valid; there arecertainly entities whose identities consist in their ‘structure’ no matter whatsubstrate supports them. But we don’t think that personal identity is like that.The Toronto Blue Jays could be dis-incorporated, utterly eliminated from themajor leagues and then be reconstituted as the same team, even though withdifferent players, different coaches and playing in a different stadium (thoughI guess it has to be in Toronto). That cannot happen to me, though I can beresurrected, if the very same relevant substrate is brought back into being.

If we take this view of personal identity and reject dualism then we shouldidentify persons with their bodies (or, better, their brains). And yet, of course, thescience fiction cases retain a certain attractiveness. Is this attraction merely anillusion, encouraged by the long years of extreme ignorance about the physicalsubstrate of the self and a variety of influential religious doctrines that exploited it?I think it is a kind of illusion, but one that it is productive to examine. The dualfacts that we allow for survival across severe alteration of memory and personal-ity, and yet are also willing to entertain the possibility of identity-transfersuggests that we have an idea of a self somehow intermediate between the selfcomposed of memory and character and the self constituted by the substrate.Such an intermediate self could be shuttled from substrate to substrate, thus

1. The situation is more complex for a dualist. It is possible, I guess, that a non-material self could jumpfrom one body to another (as many have believed in one way or another), and so it is possible that ourimagined brain operation could facilitate such jumps (though why it should have such a remarkable,extra-physical effect is another question). But notice that we are arguing over the nature of the substratehere, so the principle to be enunciated below remains valid for dualist or materialist.

2. I mention two famous and extreme cases. In 1848, Phineas Gage had a three foot bar of iron passthrough his brain as the result of an accidental explosion. Gage survived but his character was severelyaltered, completely destroying his old life (see Damasio 1994 for this as well as many other examples).A terrible case of induced amnesia resulted from an operation on ‘HM’ in 1953. The operation, designedto relieve epilepsy (which to some extent it did), involved removal of much of HM’s hippocampus andobliterated HM’s ability to lay down any new memories (see Schacter 1996).

The constructed and the secret self 249

allowing for transference of self, but it could also remain the same despitechanges in the more superficial personal qualities of memory and character.

I think this idea is less confused than it appears. In fact, a version of it is anatural outgrowth of a representational theory of the mind, which almostdemands that we allow for different ‘sorts’ of selves.

The function of the mind is to help us survive and flourish in the world andit does this in large measure with the help of representations of that world.When we think about what to do, we plot the outcomes of possible actions,whose perceived relevance is gauged relative to our present beliefs about thestate of the world, and whose envisaged outcomes depend upon these beliefs aswell as beliefs about how the world will change because of these actions. Thevery possibility of such an activity depends upon our being able to formrepresentations of the way the world is, and the way the world will be after ouractions, as well as our being able to appreciate the differences between theserepresentations. It is difficult not to embrace a representational model of themind which also allows for representations at much less sophisticated cognitivelevels than conscious planning. And it is tempting to assert that what markedthe emergence of mind and consciousness was exactly the ability to representthe world coupled to the ‘ability’ to behave in accordance with the representa-tion rather than in more direct response to the world itself.

We tend to think that animals have minds because they do things that lendthemselves to a representationalist explanation (or, as philosophers say,explanations based upon the ascription of intentional mental states). Why is thedog barking up that tree? Because it thinks the squirrel is still there (though wehave seen the squirrel depart). The dog’s representation of the world hasdiverged from reality, but its behaviour is governed by how it represents theworld, not the world.

Similarly, machines are more ‘intelligent’ the more they act according torepresentations of the world, the less they simply respond (even in complexways) to the world. AI researchers have explicitly copied our intuitive ideasabout the representational mind in their model of action. Since this model isvery simple, it is a worthy metaphor for mind and the problems it highlightsturn out to be very deep. Let’s outline the model, for every feature of it is ripewith philosophical fruit.

The model is based on the notion of ‘generate and search’: generate a spaceof possibilities and search for the ones that are useful. The space of possibilitiesare the ‘worlds’ that the system can reach via its actions; usefulness is measuredrelative to the system’s goals. Here is a picture:

250 William Seager

The Worlds at t1

Worlds at t2

The World Now

The PossibleActions

G

Figure 1.

This is the ‘game tree’ of life. Relative to the current representation of the world,certain actions are possible now, which generate representations of possibleworlds whose nature depends upon the actions and the presumed current stateof the world. The system can get to the worlds, or at least thinks it can get to theworlds, that it can generate by its actions. The action selected is the one thatleads to a goal-world (marked with the ‘G’ in the diagram). You will recognisethat this is exactly how you play tic-tac-toe.

The diagram tends to imply that the deliberation process is computationallyintensive, but the model is more accommodating than appearances suggest. Isit likely that a centre-fielder computes the trajectory of a fly ball from datagathered as the ball is hit, produces a game tree of movements and selects inadvance a path through the tree that ends with a world in which the ball is in hisglove? Hardly. It has been noticed that when the ball is coming down towardsthe fielder, he will be where the ball ends up if he moves so that the ball alwaysappears to be coming straight toward him (neither drifting up or down, right orleft in the visual field). Performing this feat does not seem to require anycomputation of Newtonian trajectories. But the process of keeping the ballcentred, as it were, in the visual field can be modelled by our picture of deliber-ation. We need only suppose there is a representation of whether the ball isdrifting, and the direction of the drift. The relevant actions are those that reducethe drift, and we have a short loop connecting goal worlds to the representationof the real world.

Nonetheless, this model of thought and action raises very difficult ques-tions. That it has revealed how serious these problems are and also introducednew problems are virtues of the model (see Dennett 1984).

For example, any finite system has a necessarily limited representationalcapacity, which means that its representations are incomplete. Even Borges’(1954) famous conceit of the map drawn at 1:1 scale must suffer from severe


incompleteness. Incompleteness entails selection, immediately raising a host ofsignificant problems about the ways that such selection can be made. In theabstract, these problems are easy to solve: represent those aspects of the worldthat are relevant to the job the representation is supposed to do. In practice, ithas turned out to be almost impossible to find a general resolution of theseissues. It may be that the constellation of problems of relevance, of which thenotorious Frame Problem is one prominent example, are insoluble from withinthe paradigm of computational cognitive science and artificial intelligence (foran argument to this effect see Horgan and Tienson 1996).

I have no solution to offer but I would like to point out that the ‘paradigm-shift’ towards connectionism which is often urged in the face of the problemsof relevance does not by itself undermine the representational theory of mind.It is true that some champions of connectionism, usually from the group thatfavours the so-called dynamical systems approach to cognition, have an anti-representationalist agenda (for an introduction to the dynamical systems viewof cognition, see van Gelder 1995). However, it seems clear that the problemsof relevance are problems of representational ‘updating’ and need not beconstrued as problems with the representational approach itself, nor even withthe general ‘game tree’ model of cognition. It is only if we regard classicalcomputational mechanisms as the onlymethod by which the internal represen-tation can be updated that the problems of relevance arise as a problem for therepresentational approach. But there is no need to make this assumption. Ingeneral, a dynamical systems approach models cognition as a set of interactingvariables, perhaps thought of as exerting ‘forces’ on one another that cantypically be described in a set of differential equations. The representation canbe regarded as the ‘position’ of the system in an abstract space and trajectoriesthrough this space can be seen as representational updating. It is an excitingpossibility — though no more than a possibility — that cognition (or at leastcertain elements of cognition) can be modelled in this way and that the ‘cogni-tive forces’ at work updating the system’s representation could solve theproblems of relevance, complexity and combinatorial explosion that beset thecomputational approach. Such a view of cognition would remain within thecamp of the representational theory of mind.3

Furthermore, it appears evident that the mind does employ representations

3. For an example of the dynamical systems approach to cognition that illustrates how it can beintegrated with a representational theory of mind, see the work on ‘decision field theory’ by Townsendand Busemeyer (1995).

252 William Seager

in cognition and that, short of endorsing a radical eliminativism, any natural-ized understanding of the intentional mental states will both use and accountfor representational states. When you plan your day, for example, you activelyconstruct and compare a variety of representations of the world as it may be atvarious times in the future, which depend upon how the world is represented tobe now as well on how you think the world will change depending upon yourprojected actions. These representations can be false, but they cannot seriouslybe thought to be non-existent. Much of your consciousness consists in theapprehension of the significance (both in the representational and ‘affective’sense) of these representations, along with, though much more rarely, anapprehension of the representations themselves. There must be as well a neuralimplementation of these processes of apprehension, and it is very hard to seehow these implementations could fail to have representational functions(though how they get these functions is another and much vexed question). Butanything with a representational function is a representation.

Be that as it may, it is the nature of the self from an assumed perspective ofa representational theory of mind that I want to focus on here. In particular, Iwant to explore the way that ‘dual’ conceptions of the self naturally emergefrom a representational approach to the mind. For reasons that will appear, Ilabel the two conceptions of the self the ‘secret self ’ and the ‘constructed self ’.Let’s consider the former first.

If a system is going to successfully manoeuver through the world by use ofthe game tree model adumbrated above, it will require a representation of itself.Of course, this is not to say that all systems that deploy representations in themodulation of their behaviour will perforce contain a self-representation. Allsorts of simple control devices — from thermostats to houseflies — regulatetheir behaviour on the basis of more or less sophisticated representations of theworld with no need for a self-representation. But more complex systems thatgenerate representations of the future and various ‘paths’ (which are sequencesof representations) to those represented futures and whichmust compare theircurrent representation of the world with these paths need to know aboutthemselves: where they are, and will be, in the world, what properties they arecurrently possessed of (and how these will change via time and action), whatobjects they can interact with throughout these paths towards the future and soon. Self-knowledge is self-representation. Just as in general the possibility ofillusion is like a sign: ‘representations at work’ so too self-illusion is a sign ofself-representation and it is all too evident that cognizers can suffer from a


variety of illusions about themselves.4

There is, however, a ‘core’ self-representation which is very special. It isneeded to account for certain peculiarities of what is called indexical knowledgeas well as the relation between such knowledge and non-indexical knowledge,action and perception. The notion of indexical knowledge arises from theappreciation that even exhaustive knowledge of the ‘objective facts’ will notnecessarily lead to any knowledge about oneself as such. This can be bestexplicated by a rather fanciful example (which I borrow from John Perry 1979,this volume). Imagine that you are in the ultimate library; every fact about theworld can be found within its books. Despite this, if you lack indexical knowl-edge, the library will not reveal to you such mundane facts as who you are,where you are andwhat time it is. Suppose your name is X. You read in the bookbefore you that X is in the ultimate library. But you, lacking indexical knowledge,don’t know that you are X. Another book says that X is reading in the ultimatelibrary at 2:30 pm on December 23rd 1998, but since you don’t know that youare X, you can’t decode this to discover what time it is right now. Arguably, unlessyou already possess some indexical knowledge, you cannot deduce any suchknowledge even from the complete data base of the ultimate library.

Perhaps this seems an implausibly strong claim. After all, as you browse inthe ultimate library you might come across a book which reads ‘X is readingbook Y and X is the only person reading book Y and there is only one copy ofY in the ultimate library (and, by the way, there is only one ultimate library andthat is the only place where Y can be found)’. You flip to the cover and noticethat you are holding book Y, and hence you discover that you are X.5 However,in order to make this inference you do need one antecedent piece of indexicalknowledge: I, myself, am holding book Y. Of course, it is not hard, normally, toget this sort of indexical knowledge. It is an interesting question, though, justhow such indexical knowledge is generated.

I propose that the answer is that there exists a special self-representation (ora self-representational sub-system within the overall cognitive economy) whose

4. It is possible that some illusions about the self are actually beneficial. Generally speaking, people thinkthat they are rather nicer, smarter, more attractive and leading more worthwhile lives than they reallyare, but this illusion might improve their lives. In this case, clear headedness might only increase thesuicide rate.

5. Note that you will never find a sentence saying ‘X is reading this book’ for that is a piece of indexicalknowledge. Maps that can’t move (such as the ones in shopping malls) exploit the fact that theirimmobility renders one item of indexical knowledge unusually secure: you are here (with an ‘x’ to markthe spot).

254 William Seager

function is to transform incoming information into indexical knowledge. Thisrepresentation is not distinguished by its carrying any information about theself. If this self-representation was individuated by the way it represented theself, it would not be the locus of indexical knowledge, for the way it representsthe self could then be expressed in non-indexical terms. No list of facts aboutmyself, such as ‘Seager lives in Canada’, ‘Seager has three children’ etc. couldreveal to me that I am Seager. The self-representation I am positing is primitiveand information free. Its distinctive role is to embody indexical knowledge, notby explicitly encoding the information but rather by the way it integratesperception and action with the information already within the system.

Suppose that someone shouts “Seager’s pants are on fire”. I hear thisinteresting piece of news and my internal representation is appropriatelyupdated, perhaps by the addition of the following item: there is an x, x’s nameis ‘Seager’ and x’s pants are on fire.6 But I won’t be motivated to do anythingabout this unless this information is somehow connected to me, to my con-cerns. This is the job of the special representation of the self. If, for example, Iknow that I am Seager, then getting the information that Seager’s pants are onfire will lead to the conclusion that my pants are on fire. And if I know that mypants are on fire, you can be sure I’ll do something about it. (Motivation canarise in a variety of ways but it always depends on the linkage to the self thatindexical knowledge provides; in the example just given, if I don’t think I amSeager but do think that I am near Seager, I will still be motivated to act — orshould be motivated at least — but now my action will be to try to help Seagerin some way.)

Logically, we can think of indexical knowledge as working in a perfectlystandard way. The posited self-representation can be thought of as a kind ofname which functions in inference like any other term, so the inference from‘Seager’s pants are on fire’ to ‘my pants are on fire’ proceeds by substituting*self* for ‘Seager’ as licensed by the indexical knowledge that I am Seager (as in

6. I am not assuming that the internal representation is literally composed of sentences in mockpredicate logic. I have no idea how our representational machinery is structured; probably it exploits ahuge number of distinctmodes of representation. I use the sentence format for ease of presentation only.Within this perspective it is natural to regard the self-representation as a kind of name, but one that hasa unique function as described above. There are advantages to looking at it this way, notably ease ofcomprehension and explication. But we are not forced to adopt a ‘language of thought’ conception ofcognition to endorse the idea of a special sort of self-representation. The posited self-representation isdistinctive in the way it links together knowledge, perception and action, and this function could becarried out by a wide variety of representational systems. And it seems that in any complex cognitivesystem such a function must be carried out.


‘Seager’s pants are on fire’, ‘*self* = Seager’, therefore, ‘*self*’s pants are onfire’). It is possible to develop a formal logic in which such self-representationsfunction appropriately as a model of indexical knowledge (see Seager 1990).

What we should focus on here however is the special features of the self-representation which suits it to the job of encoding indexical knowledge. Thesefeatures are the ‘direct’ link between the self-representation and both percep-tion and action. If I represent myself as in danger I will act; if I represent Seagerbeing in danger then whether I act or not depends upon whether I know that Iam Seager. The self-representation is, so to speak, the ground-floor representa-tion which links me to the world. But I need have no access to this representa-tion and in fact I have no direct access to it. I had to posit its existence.

There is something of a paradox here. Insofar as I can consciously representmy self-representation to myself it becomes a bearer of information aboutmyself which may or may not be linked to perception and action. It becomes aname for myself like any other which must itself by linked to the self-represen-tation before I would be motivated to do anything with respect to its consciousrepresentation. This paradox arises because there is nothing to the self-repre-sentation except its links to action, perception and other knowledge. It can bethought about only by creating a representation of it, which is not identical to it.

Nonetheless, I think the possession of this kind of self-representation isnecessary for us to have any sort of fundamental sense of self, for it is whatunifies our awareness of the world.7 Our consciousness of the world is unifiedin the sense that, for everything that we are aware of, we are aware of thosethings relative to ourselves, as things that we, ourselves, perceive in one way oranother and that we can act ‘towards’.8 It is rather crude to say that so far as Iam aware the world is exhaustively catalogued as the things that matter more orless to me, but it is not altogether inaccurate. I think this is Wittgenstein’s pointwhen he says: “what brings the self into philosophy is the fact that ‘the world ismy world’” (1921:5.641). There also seem to be connections between this ideaand Kant’s transcendental unity of apperception. If we regard Kant’s I think(which can accompany all other representations) as a gesture towards what Ihave been calling indexical knowledge, the similarity is striking (see Brook 1994,

7. Thus I think that simpler creatures that interact with the world without having a self-representation— the sort of creature considered briefly above— cannot develop a sense of self. Of course, muchmorethan possession of the kind of primitive self-representation posited here is needed for a sense of self.What more is required will be discussed below.

8. Here perception must be taken to include remembering and imagining, and action must include‘thinking about’.

256 William Seager

especially ch. 4, for more detailed links between Kant’s philosophy and this wayof looking at self-representation). A representation is mine if it is possible for itto be linked to the self-representation, and anything which I am aware of eithervia perception or actionmust meet this condition or be capable of meeting thiscondition. Consider once more the example of the burning pants. What I can‘notice’ about the situation is whatever about it that can forge links to my self-representation. Although I am in fact Seager, I can’t be aware that my pants areon fire just by knowing that Seager’s pants are on fire unless I know that I amSeager. It is always through the link to the self-representation that I becomeaware of features of the world around me.

The peculiar ‘position’ of the self-representation is expressed in anotherremark of Wittgenstein’s: “the subject does not belong to the world: rather, it isa limit of the world” (1921:5.633). Thinking of Wittgenstein’s ‘subject’ as ourposited self-representation, it is a limit in the sense that nothing unrelated to itcan ever enter the world as perceived or conceived. In another sense, it might bethought of as the centre of my world, utterly invisible because everything is seenfrom its vantage point.9

This special self-representation is what I called above the secret self, for it isinvisible to the subject and we know it only through postulation. Yet of coursewe know much about ourselves. But this is a different self, which I call theconstructed self.

There are two fundamental assessments of the world that the secret-selfmust be responsible for: what is happening, and how does it matter to me. Thelatter is far more important than the former, for the only things that I reallyneed to know about are the things that might matter to me. It is tempting tospeculate that the secret-self trades in speed, so is satisfied with rough assess-ments of truth and quick judgements of value. Given this speculation it isfurther tempting to locate the secret-self in the ‘lower’ brain, in the so-calledlimbic system, perhaps, for one more definite possibility, the perceptual andmotor pathways that run through the amygdala, for which there is abundantevidence of rough, quick and decisive ‘assessment’ of truth and value (seeLedoux 1996, ch. 6, for some fascinating data on fear conditioning and thedistinction between the operation of the ‘thalamo-amygdala’ pathways and the‘cortico-amygdala’ pathways). One last speculation would then stress the

9. A better metaphor which rather nicely combines both the idea of centre and limit might be anadaptation of the oldmystical view ofGod. The subject is a sphere whose centre is everywhere andwhosecircumference is nowhere.


significance of the relation between the ‘old-low’ brain and the ‘new-high’brain, perhaps going so far as to locate the systems responsible for the construct-ed self (see below) within the latter even as we locate the secret-self systemswithin the former.

It can also be pointed out that by their nature there could in fact be manysecret selves within any subject. These would be distinct centres of awareness,possibly quite disjoint from one another (though nothing prevents the con-structed selves — to be discussed below— which will be associated with thesesecret selves from knowing about the other constructed selves). These secretselves would be distinguished primarily by what they ‘regard’ as true andvaluable (see below for the central significance of these notions for both kindsof selves), but they might also differ in the sorts or amount of cognitive resourc-es available to them (e.g. memories, skills, etc.).

Let us ask how we know about ourselves, our mental selves that is (not ourbodies), our thoughts and feelings, hopes and dreams as well as the memoriesthat anchor us to our lives. It is correct but facile to reply that we know our-selves by introspection. We then have to wonder how introspection works. Theanswer to this explains why we should call the self we know by introspection theconstructed self and reveals the relation between the two selves.

What is in my opinion the best theory of introspection has recently beendeveloped by Fred Dretske (1995, ch. 2). His theory is incomplete however, andI want both to review it and extend it to amore general theory of introspection.

Dretske’s idea is that introspection is a form of what he calls ‘displacedperception’ which is simply learning about one thing by perceiving somethingelse. An example he uses is learning that the postman has arrived by perceptionof the dog’s barking. To get such knowledge one must hear the dog and onemust also know what the dog’s barking signifies. Introspective knowledge of ourown perceptual states similarly requires that we perceive but also that we knowthat perceiving is a mental act. All knowledge — at least all declarative knowl-edge such as introspection delivers — is conceptual and so requires an appro-priate field of concepts for its formulation. In the case of introspective knowl-edge, what concepts would these be? Since introspective knowledge is knowl-edge of the mind, they must be mentalistic concepts, concepts of mental states.So introspective knowledge requires the field of concepts that taken togetherform our notion of the mind. I don’t think it does any harm to use the familiarlabel of folk psychology for this body of concepts along with their associatedgrounds for application. I know that I am perceiving red, when I am perceivingred, because I can apply the concept of perceiving red to my perception of red.

258 William Seager

I don’t need to perceive my perceiving (as certain internal scanner theories ofintrospection have maintained, see for example Armstrong 1968) to make thisapplication any more than I need to perceive my perceiving of a barking dog toapply the concept of ‘barking dog’ to that object.

Of course, I do need to be perceiving red to make the introspective applica-tion of the concept ‘perceiving red.’ In fact, I have to be consciously perceiving,for if I was not conscious of the colour I would have no ground for asserting myintrospective knowledge claim (nothing, as it were, to apply my concept to for,as Kant famously said, concepts without intuitions are empty). Withoutconsciousness there is no evidence on which to ground the introspectiveknowledge claim. We can, I suppose, still imagine bizarre science fiction caseswhere I come to know that I am, somehow, unconsciously perceiving red, butthis knowledge would not be introspective knowledge just because there is noconscious mental state to provide the grounds for any introspective knowledge.The point can be made in a partial definition of introspection as self-knowledgeof a mental state on the basis of one’s state of consciousness engaging one’smentalistic conceptual machinery.

Although my conscious states provide what can be called ‘evidence’ for myjudgements of introspection, it would be misleading to say that I infer frommystate of consciousness to an introspective judgement about that state of con-sciousness. For this would imply that I already know what my state of con-sciousness is, which would be to say that I have already introspected. If I inferfrom anything here, it is from the way the world is presented to me (somethingI know without introspection). It requires additional conceptual equipment togo from the presentation to the knowledge that the world is being presented tome. I need the concept of ‘conscious presentation’, which is not needed just forthe world to be presented to me. So, when Dretske talks of ‘displaced percep-tion’ as a model for introspection we should not think of the displacement asinvolving a move from an awareness of a mental state to a secondary awarenessof that state (or yet another mental state), but rather as a move from anawareness of a non-mental state (or object, scene, bodily condition, etc.) to theawareness of the mental state of being aware of that non-mental thing.10

10. I suppose we could allow that there are some mental states that can be perceived in some wayindependent of introspection. Imyself can see little need for this, but if one wanted to assert that feelings(of pain, fear, anger, etc.) were perceptions of mental entities themodel could accommodate that desire.These perceptions would not be introspections for they would not be awarenesses of the mental as such(obviously one can feel angry without having any introspective knowledge that one is angry, or is feelingangry). I think it is preferable to regard sensations as a kind of perception of the bodywhich we are aware


If we think in terms of inference, we require at least two beliefs: the input toand the output of the inference. If the input of the inference was something likethe belief ‘I am aware of a tiger in front of me’ we would have implicitlyappealed to introspective knowledge for we are claiming that I already knowabout my awareness of the tiger. The account offered would thus be circular.We should instead insist that the input belief is ‘a tiger is in front of me’ and theoutput belief is ‘I am aware of a tiger before me’. Think of children. At an earlyage they can form the belief that a tiger is in front of them; it takes moreconceptual sophistication for them to know that they are aware (or are visuallyaware) of a tiger in front of them. Such increased sophistication is very impor-tant for it allows children to entertain the possibility that they might be havingnothing more than the mere visual experience as of a tiger in front of them andthat others might have a divergent experience of the world. The ability tocomprehend the epistemic distance between the world and the experience of theworld is not some kind of benighted proto-Cartesianism; it is a vital steptowards self-consciousness and an awareness of one’s own identity.

The key to understanding this position on introspection is always to bear inmind that when we perceive we do not perceive a perceptual state but rather weperceive what the perceptual state represents. Seeing a tiger involves a represen-tation of a tiger but it does not involve seeing that representation. Thought is thesame; when we think, we are aware — in the first instance at least — of thethought’s content, not of the thought itself. To adapt a remark of Dretske’s(1995:100–101), mental representations are the things we are conscious with,not the things we are conscious of. Although it is venerable, the idea that we arereally aware of our mental states instead of being aware of what they representis as confused as the idea that we can only talk about words because we have touse words whenever we talk. ‘Talking about X’ involves the use of words but itdoes not require that we talk about those words in order to talk about X. Just so,seeing a tiger demands the use of representations (of tigers) but it does notrequire that we see those representations. The fact that perception can beillusory or hallucinatory is of no more significance than the fact that we canutter falsehoods. Obviously, there is no reason at all to think that the sentence

of as we are aware of the rest of the world (though via distinctive sensory channels). Thus, for example,the sensations of anger involve an awareness of events in the body, as well as strong awareness of valueand dis-value (a vitally important component of consciousness which will be discussed below). Anawareness of anger as such requires the level of conceptual sophistication which permits introspection.Evenmany adults are sometimes in states where they have the feelings of anger but don’t know that theyare angry (and similarly for other emotions).

260 William Seager

‘tigers live on the moon’ is really about its own words just because it is false.The theory is not, then, that one infers from a knowledge of one’s state of

consciousness to introspective conclusions about that state, though Dretske’sexamples sometimes unfortunately tend to suggest such an inferential mod-el.11 The ‘evidence’ needed for introspective knowledge of our own perceptualstates is simply our own conscious perception of theworld, not a consciousnessof that consciousness. One can, however, suffer perceptual delusions, illusionsand hallucinations. Perceptual consciousness remains throughout all of these,and may be indistinguishable from veridical perception, and so introspectiveknowledge of our own perceptual states also remains possible, though suchknowledge will inherit the illusory aspect. Thus my introspective claim that Iam perceiving a horse can be in error no less thanmy perceptual claim that thereis a horse in front of me, but I can weaken my introspective claim (e.g., inphilosopher’s jargon, to something like ‘I am in a horse-perceiving-like percep-tual state’) just as I can weaken my perceptual claim (e.g. to ‘there seems to bea horse in front of me’). But my introspective judgement arises in both sorts ofcase not via conscious inference but rather through the application of theappropriatementalistic concepts (and, of course, the application of a concept isnot itself a conscious actwhich itself requires some kind of reflective deliberation— such a view leads to an obvious vicious regress).12 Once I’ve got the conceptof ‘perceiving a horse’ I can apply it to the state of perceiving a horse — that iswhat ‘having a concept’ is —and the ‘input’ to this application is the consciousperception of the horse (not a perception of the perception of the horse).

On this model, the evident fact that animals and children (at least veryyoung children) are incapable of introspection is neatly explained by their lackof the proper field of mentalistic concepts, the lack of a ‘theory’ of the mind.They consciously perceive the world, but don’t know that that is what they aredoing and it is this lack of conceptual machinery that precludes introspection.Their inability to introspect does not stem from a lack of some special internalscanner within their brains (as Armstrong and perhaps Churchland wouldappear to be forced to aver, see Armstrong 1968, Churchland 1985), nor is it, as

11. This causes Dretske some difficulty when he tries to explain how introspective knowledge is more‘direct’ or ‘secure’ than other sorts of knowledge (see Dretske 1995:60�ff.). I will discuss this a little morebelow.

12. It remains entirely possible that the sub-personal story of concept application will be one ofinference-like cognitive processes that marshal evidence and ‘test’ hypotheses, but the subject iscompletely unaware of these putative operations. The conceptually informed world is ‘delivered’ to uswithout any conscious effort on our part.


Ryle would have it (see 1949, ch. 6), that they cannot perceive their ownbehaviour nor ‘hear’ what they say to themselves (children talk to themselveslong before they can introspect). To the extent that the conceptual system thatmakes up our theory of the mind is a relatively late acquisition, to that extentintrospective knowledge will itself be a late acquisition. There is evidence that,in all its fullness, the theory of mind is acquired pretty late, typically after agefour (see Perner 1991, Gopnik 1993). It is also perhaps worth mentioning thatthe lack of introspective abilities will not preclude children from makingperceptual judgements, for these depend upon a field of concepts which applyto the world, not to the mind itself — and these concepts come first and early.

Note also that if the ‘theory of mind’ is not only a late developmental eventbut a late cultural acquisition then we can infer that early hominids wereincapable of introspection, and hence were entirely un-self-conscious. Theinfamous speculations of Julian Jaynes (1976) can be fitted into this model ofintrospection in interesting ways. To see Jaynes as claiming that it is self-consciousness rather than consciousness itself that is a recent, indeed very recent,development, and even a recent invention, is to make his views rather moreplausible than they otherwise appear. It is fascinating to speculate that peoplebegan, for example, talking to themselves before they knew what they weredoing, so they had no way of understanding what was happening to them(Jaynes argues that such people would have assigned such voices to thirdparties; they were the ‘voices of the gods’).13

Dretske’s account, as he presents it, is restricted to introspective knowledgeof perceptual states but this is only a small province within the realm of self-knowledge. What I want to do now is extend the account to cover otherphenomenal states as well as intentional mental states. The case of otherphenomenal states requires only a trivial extension. Introspective knowledge of,for example, our own pains requires what we might call first-order conscious-ness of the pain (which exists in animals no less than ourselves),14 plus the

13. An anecdote about my own daughter perhaps bears this out. When she was three she oftencomplained about not being able to sleep. When we asked her about this she gave the strange explana-tion that her ‘ears were talking to her’. I believe that she was talking to herself but didn’t as yet have thenotion of such an activity and so could not understand what was going on, and was rather naturallyupset by it.

14. It is important to bear in mind that feeling a pain is not a matter of introspection. Feeling a pain isto be thought of as analogous to seeing a horse (in fact, I believe that feeling a pain is a kind ofperception, a perception of the body by a highly specialized sensory system). The phrase ‘feel a pain’suggests an awareness of a mental state, but there is no awareness of a pain as a mental state as such�; weare just aware of the terrible feeling.

262 William Seager

knowledge that this sort of thing is a pain, or the explicitly second-orderknowledge I am in a state that hurts or something along these lines— the exactextent of knowledge of the mind required to underwrite introspective knowl-edge is of course somewhat vague. Since the phenomenal states provide, bydefinition, a range of characteristic conscious experience, the displaced con-sciousness model can straightforwardly apply to them. It is the extension tointentional mental states which is more problematic. It will be achieved by anextension of the realm of the perceptible. What I mean can perhaps best beillustrated by an example.

Suppose I write down a list of simple sums, like, 2+7=9, 12+3=15, 8+5=14,etc. I could ask someone to put a check-mark beside the ones that were true andthis would be a trivial task so long as my subject knew a little about simplearithmetic. It would not be a task demanding introspection. But I could insteadask my subject to check off the sums which he believed to be true. This would beno less trivial so long as my subject understood just a little bit about what beliefswere (as well as simple arithmetic). What the subject must minimally under-stand is an elementary principle of folk psychology, which I can write in adistinctly non-elementary form as: the object of belief qua belief is the true. Thesecond of my tasks involves introspection, albeit at a rather primitive level. Butseeing that 2+7=9 is correct, or is true, is not in itself an act of introspection anymore than is seeing that a zebra is striped. To see what is true, we need toinvestigate the world, not ourselves.15 This investigation can occur in theimagination, or via memory, so there is an appropriate internal source for theevidence needed to provide the full range of introspective knowledge of ourown belief states. When we ‘look’ inside ourselves we don’t see beliefs lined upalong our mental hallways, but we can discover truths there and if we dodiscover a truth we have at exactly the same time trapped a belief. Most thingsare less certain than elementary sums and so we may wonder about our ownbeliefs insofar as we wonder what is true, and thus wemay be unsure about ourown beliefs.

The other basic category of intentional states is desire. The general sort ofevidence needed for introspective knowledge here is value. The picture neededfor the account of introspection I am urging requires that when we look aroundthe world, we not only see objects with their various perceptible properties, butwe also perceive a field of values. Is it true that the objects around us come

15. Except, pedantry insists, when we ourselves are the object of the search for knowledge; this will notalways be the search for introspective knowledge.


graded in their value to us as they enter consciousness? I think even a littlereflection upon experience shows that it is, although this is a multifarious valuewhich is constantly changing in response to all sorts of changes within our-selves. We can use a variant of the arithmetic example to show this. Suppose wereplace our list of simple sums with a tray of various small items: some nails, olddry leaves and some sweets and we now ask a hungry subject to pick out thegood ones, or the ones that are good to eat. Any subject, over the age of roughly1.5, could accomplish this. Some (slightly) more sophisticated subjects could beasked to pick out the ones they desire to eat. The first task does not require anyintrospection, or self-knowledge to be successfully carried out. The second one,as I mean it to be interpreted, is a task involving introspection. (It does requiresome interpretation since we would normally use the phrase ‘pick out the onesyou want’ to specify the first task rather than the second— a significant fact thatactually supports the view of introspection I am putting forth.)16 One canconsult one’s own desires about a field of objects before one selects, but this issimply to gauge the objects’ values from the point of view of ascribing desires tooneself, just as taking up a point of view in which one talks of one’s own beliefsis to gauge the truthfulness of a variety of propositions (or whatever the abstractobject of belief is taken to be).17

There are, of course, a myriad of intentional states beyond belief and desire.But it may be that they can mostly be defined in terms of belief and desire, orare just various forms of belief and desire. For example, wishing for p is, moreor less, to desire p and to believe that p is unlikely to be (or come) true (for anice attempt to define a large range of emotions basically in terms of belief anddesire, see Descartes, 1649). For the really complex interweaving of high levelintentional states, the theory I am offering admittedly begins to give out butthere a self-interpretation theory like Ryle’s seems more reasonable. When

16. This slight difficulty of interpretation also occurs in perceptual contexts. We often ask for informa-tion about the world in explicitly mentalistic terms. For example, if we want to know if a zebra is in theroom we can ask: do you see a zebra. This is not meant to be a request for an introspective search of ourinformant’s perceptual states, but simply a way of asking if a zebra is visible (it is, of course, by way ofsuchmodes of speech that the theory ofmindwhich grounds introspection is passed on to our children).

17. Moore’s ‘paradox’, that it is somehow senseless for me to say ‘p is true but I don’t believe it,’ eventhough a lot of other people say this of me all the time, is grist for the mill of this account of introspec-tion. On other views of introspection, such as the internal scanner view, Moore’s paradox is somewhattroubling for it would seem that I ought to be able to scanmy belief states independently of assessing thetruth of any proposition and so I could, one might think, quite easily discover that I don’t believesomething which I can see, so to speak, to be true. The impossibility of this must be given a rather ad hocexplanation in an inner scanner theory.

264 William Seager

Mary is trying to decide if she really does love John, she must engage in morethan the mere assessment of truth and value. But these two fundamentalassessments remain at the core of her self-knowledge. If, for example, sheimagines life with John, or goes over the way he acts in a variety of situations,she is assessing truth and value within the imaginary or remembered scenes.

I suggest that there are three classes of ‘elementary acts of mind’ or con-sciousness which are required to underwrite this view of introspection. Since wecan be in error about the ‘external significance’ of all three, I will describe themin terms of seemings. They are: phenomenal seemings (which encompasses bothour conscious perceptual states and all our ‘feelings’, including pains and otherbodily sensations), truth seemings and value seemings. Each one provides aroute to introspective knowledge about our ownmental states, not via any sortof direct or privileged access but simply by way of the application ofmentalisticconcepts (drawn from our ‘theory’ of mental states) to these seeming. Knowingabout the mind, I can that I am in a perceptual state of seeing red when I lookat, say, the Canadian flag; knowing about the mind, I can know that I am inpain when I feel the twinge; knowing about the mind, I can know that I believethat 2+2=4 when I understand the truth of this sum; knowing about the mind,I can know that I desire a chocolate when I sense the goodness (relative to thepurpose of eating) of the candy before me. To the extent that the three elemen-tary acts are indeed constituents of consciousness, we have a reasonable groundfor the extension of Dretske’s view to the whole of our mental lives. And it doesseem to me evident that we are or can be conscious of perceptual properties,truth and value.

Dretske (1995:60�ff.) worries that this view of introspection makes intro-spective knowledge a species of inferential knowledge, the indirectness of whichwould to many be objectionably implausible. Dretske seems to believe that thefact that non-veridical perceptions can ground introspective knowledge no lessthan veridical ones ‘neutralizes the objection’ (p.60). He goes on to say that: “ifthis is inferential knowledge, it is a strange case of inference: the premises donot have to be true to establish the conclusion” (p.61). This is a strange way toput the point. Surely the ‘premises’ here are the seemings I have noted aboveand, of course, it is true that I seem to see red even when my perception is non-veridical. We are not really more directly aware of our own mental states thanwe are aware of the world around us, but within the realm of introspectiveknowledge we have usually already taken back the epistemic commitment to theveridicality of the perceptual state; at least we are not interested in its veridic-ality but rather in the perceptual state itself.


The appearance of a direct introspective awareness of our ownmental statesis to be primarily explained by the fact that the ‘inference’ from how the worldis presented to us to claims about our own mental states is not (or not usually)dependent upon a conscious deliberation but is rather simply the sub-personalapplication of the mentalistic concepts to their appropriate objects. It is likevision itself — when I see a computer keyboard in front of me I do not ‘infer’ tothe keyboard from an ‘appearance as of a keyboard’ plus assumptions about thereality of the external world and all the causal relations that link me to it. Theconcept just springs into my mind and I see the keyboard as a keyboard.Similarly, when I feel a pain I don’t have to think about whether I am experienc-ing a pain; it just springs into my mind that I am.

There are also three features of introspection and consciousness thatconspire to enhance the sense of directness in introspection. The first is that wecan confuse consciousness with introspection. There is no introspectioninvolved in being conscious of the elementary seemings that make up the ‘field’of the mind but, since we, as sophisticated beings in possession of the conceptof mind, are aware that we are aware, it can seem that we are directly appre-hending our own minds. One can take up a detached view of one’s experienceand view it as experience — everything then becomes introspectible since oneis regarding everything one is aware of as a manifestation of mind. Then theintrospective move fromworld to mind ‘disappears’ in rather the same way thata constant background noise can disappear from consciousness, simply becauseit is so ubiquitous. The second feature that makes introspection seem so directis that we tend to mix together our mental states with the state of the world inour speech; we use the phrase ‘I believe …’ to report the truth of something(and vice versa); we use the phrase ‘I see …’ to report that something is beforeus, and similarly we conflate desire with goodness. We come to see the worldthrough our theory of the mind and we are encouraged in this tendency (seenote 16 above).18 But the third and most important feature is just that thetheory of the mind we use to make introspective judgments is second nature to

18. A word of warning. Philosophers have often confused consciousness with introspection. Forexample, Locke defines consciousness as “the perception of what passes in aman’s ownmind” (1690, bk.2, ch. 1: 115) and more recently Armstrong echoes this with “consciousness … is simply awareness ofour own state of mind” (1968:94). Introspection is entirely distinct from consciousness however.Introspectiondependsuponconsciousness but consciousness doesnot dependupon introspectionor thepossession of the ability to introspect. I appealed to consciousness in the so-called elementary acts ofmind which feed evidence to our introspective abilities, which are grounded in our knowledge of atheory of mind. But obviously I did not explain consciousness itself. And there is no prospect ofexplaining consciousness by explaining introspection or self-knowledge.

266 William Seager

us. We are thoroughly trained in its application from a very early age. Wecompletely and naturally absorb the elementary principles that we believe whatis true, want what is good and are seeing what is visible to us.

Into the gap between the thing and the application of a concept to the thingerror can creep. A rustling movement in the bush can prompt the firm convic-tion that a tiger is approaching; here the gap is wide and error is frequent. Howcan we measure the gap between experiencing a pain and noting that it is a painthat I feel? In one sense (much remarked upon recently) the gap can be as largeas we like, for we can imagine that the subject does not properly ‘grasp’ theconcept of pain. This should be uncontroversial. In another sense however thegap can shrink to almost nothing, for introspective knowledge can avail itself ofthe conceptual resource of demonstrative reference. If I feel a strange sensation,I can always introspect it as a ‘something which is happening (to me)’. Toentertain the suggestion that I don’t understand the concepts involved in suchaminimal introspective view is to come very close to undermining thought itself.

Our important self-knowledge is introspective knowledge. It is knowledgeabout our own beliefs and desires, our pain and joys and most important of allour own memories. To see these as our own, or part of our selves, we mustunderstand what a mind is that it can contain such wonders, and we must learnto recognise them so as to reflect upon them as defining parts of ourselves. Ibelieve that long before we could introspect we felt pain and joy, we could thinkand plan (though introspection lets us feel more keenly, and sometimes to thinkand plan more clearly). At some point in the distant past, the concepts whichmake up the overarching concept of mind must have been invented, as allconcepts must be invented.19 Probably it was not a sudden invention but aslow accretion of ideas (there is after all some evidence that certain non-humananimals, higher primates in particular, have some kind of idea of self and somesense of their fellows’mental states (see Humphrey 1984, Cherney and Seyfarth1990; for a skeptical account, see Heyes 1994). Surely, there was no inventor offolk psychology but there might have been (see Sellars’s [1956] myth of Jones).The concept of mind was constructed and we learned to apply its many sub-concepts, just as we constructed and learned to apply so many other concepts.The interlocking notions of belief, desire, and feeling are the most precious

19. It is perhaps conceivable that some concepts are innate, and have been generated by evolution. Butit is not very plausible; concepts have social and normative dimensions that seem foreign to purelyevolutionary processes. Categories can be evolutionarily implanted, and can do some of the jobs thatconcepts do, but nothing as rich as our theory of mind could be reduced to categorization.


legacy we have from our remote ancestors, as crucial today as, perhaps, 50,000years ago (or whenever they were invented — possibly much more recently).

The self that we know through introspection is a construct. This does notfollow from the fact that the concept of mind is a construct. Every concept is aconstruct, and a social construct if you like, but that does not make the thingsto which the concepts apply constructs (still less social constructs). The conceptof a quasar is of recent construction but quasars are not.When we introspect wefind mental states, and these are not constructs. But the idea of self has strangepowers, for we, being first conscious and second self-conscious beings, canconstruct ourselves in its image. Quasars cannot do this; they are stuck withbeingwhat they are.We, in complete distinction fromperhaps everything else inthe world, canmake ourselves into something for which we had only the image.

What is important to the self? Memories are perhaps the most criticalelement of selfhood. We treasure our memories, we build systems of externalaids to their retention, and we integrate our memories into our own sense ofourselves (we are taught to do so). We also transform ourmemories in more orless subtle ways. One way we transform them — by selection, distortion, de-and over-emphasis— functions to re-form our own lives (or our images of ourlives) into coherent narrative structures (see Dennett 1992, Bruner 1991). IanHacking has emphasised that conscious beings have the peculiar power of beingable to make themselves into exemplars of some theoretical picture of con-sciousness or personhood (see Hacking 1995). It is in this strong sense thatone’s self is a construct, constructed both by oneself continually through life(though usually pretty well set fairly early on) and by the surrounding people(no doubt most notably one’s parents) as they exist within a culture that has itsown stake in what constitutes a proper self and its own power to bring suchselves into being.

But always lurking under the constructed self is the unconstructed secretself, continually monitoring every event for personal significance, ceaselesslylinking incoming information to oneself. The secret self is nothing in itself andis vastly influenced by the constructed self (since the construction will influenceone’s own assessment of value and significance). Such influence is far fromcomplete control, however. The assessments of the secret self have someindependence, which appears at many levels of cognition. At a fairly low level,in illusions such as the Müller-Lyer, our senses tell us that one line is longerthan the other, and though we know that it is not so we still see one line aslonger. At much higher levels, we can tell ourselves that we believe a proposi-tion but know that we do not; we can insist to ourselves that something is

268 William Seager

valueless while knowing that we still desire it. Such conflicts reveal a connectionto the secret self, for the values of objects are gauged in relation to myself andmy own concerns; they are thus linked to the basic self-representation whichunderlies the consciousness of value. More abstract assessments of value involvemuchmore sophisticated and self-conscious appreciation of the structure of theworld and my place in it, and this is the home of the constructed self.

I think that we have some sense of ourselves as a ‘centre’ of perception,truth and value which retains some independence from the self we know byintrospection. This centre remains through changes in memory and personality,for without the secret self we are utterly disengaged from the world and couldneither act in the world nor even be aware of it. The independence of the secretself gives some currency to the idea of a self separate from the constructed selfwhich could, conceivably, be transferred from body to body without thetransfer of the memories, capacities and personality traits characteristic of theconstructed self. Such an idea I think helps explain the intuitive appeal of thescience fiction cases that began this essay. The idea is however spurious, for thesecret self has no identitywhich could persist apart from the bodily mechanismswhich underlie the links between perception and action. There is no way asecret self could be transferred from one body to another; everyone’s secret selfis equivalently empty of any traits that could mark it as belonging to one personrather than another. On the other hand, the constructed self could be trans-ferred, but only in the sense of duplicating its structures in another.

So in the end we should agree with Williams’s final assessment that theidentity of persons is the identity of their bodies, while appreciating the groundfor the intuition that there is somehow something — a kind of self — under-neath the constructed self. Paradoxically, the secret self is at once the source ofeverything that is mine in experience, but bears absolutely no relation toanything that makes me distinctively me. For what makes me me is the con-structed self.

Bibliography

Anscombe, G.E.M. 1975. The first person. InMind and Language, S. Guttenplan, ed. Oxford:Clarendon Press: 45–65.

Armstrong, D. A Materialist Theory of the Mind. London: Routledge, Kegan Paul.Ayers, M. 1991. Locke, Vol.2. London: Routledge, Kegan Paul.Barwise, J. and J. Perry. 1981. Situations and Attitudes. Journal of Philosophy 78: 669–691.Bennett, J. 1974. Kant’s Dialectic. Cambridge: Cambridge University Press.Boër, S.E. and W.G. Lycan.1986. Knowing Who. Cambridge MA: MIT Press/A Bradford

Book.Borges, J.1954/1972. On Exactitude in Science. In A Universal History of Infamy, N. Gio-

vanni, trans. New York: Dutton.Brook, A. 1975. Imagination, possibility and personal identity. American Philosophical

Quarterly 12: 185–98.———. 1994. Kant and the Mind. New York: Cambridge University Press.Bruner, J. 1991. The Narrative Construction of Reality. Critical Inquiry 1: 1–21.Burge, T. 1988. Individualism and Self-Knowledge. Journal of Philosophy 85: 649–663.———. 1996. Our Entitlement to Self-Knowledge. Supplementary Proceedings of the

Aristotelian Society 96: 91–116.———. 1998. Reason and the First Person. In C. Wright, B.C. Smith, and C. Macdonald,

eds. Knowing Our Own Minds. Oxford: Oxford University Press.Cassam, Q., ed. 1994. Self-Knowledge. New York: Oxford University Press.Castañeda, H.-N. 1964. Review of Hintikka’s Knowledge and Belief in Journal of Symbolic

Logic 29.———. 1966, this volume. ‘He’: A study in the logic of self-consciousness. Ratio 8:130–57.———. 1967. Indicators and quasi-indicators. American Philosophical Quarterly 4:85–100.———. 1968. On the logic of attributions of self-knowledge to others. Journal of Philosophy

65:439–56.———. 1963. The Private Language Arguments. In C. Rollins, ed. Knowledge and Experience.

Pittsburgh: University of Pittsburgh Press.———. 1977. Perception, belief, and the structure of physical objects and consciousness.

Synthese 35: 285–351.Cheney, D.L. and Seyfarth, R.M. 1990. How Monkeys See the World: Inside the Mind of

Another Species. Chicago: University of Chicago Press.Chisholm, R. 1981. The First Person. Brighton: The Harvester Press.Churchland, P.M. 1985. Reduction, Qualia, and the Direct Introspection of Brain States.

Journal of Philosophy 82: 8–28.Coope, C., Geach, P., Ports, T., and White, R. eds. 1971. A Wittgensteinian Workbook.

Oxford: Blackwell.

270 Bibliography

Damasio, A. 1994. Descartes’ Error: Emotion, Reason, and the Human Brain. New York: G.P.Putnam.

Davidson, D. 1968. On saying that. Synthese 19:130–146.———. 1979. Quotation. Theory and Decision 11:27–40.Dennett, D.C. 1978. Where am I? In Brainstorms. Hassocks: Harvester Press: 310–323.———. 1984. Cognitive Wheels: The Frame Problem of AI. In C. Hookway, ed., Minds,

Machines and Evolution. Cambridge: Cambridge University Press: 129–151.———. 1987. Beyond Belief. In The Intentional Stance. Cambridge MA: MIT Press/A

Bradford Book: 117–202.———. 1992. The Self as Narrative Center of Gravity. In F. Kessel, P. Cole and D. Johnson,

eds., Self and Consciousness: Multiple Perspectives. Hillsdale, NJ: Erlbaum.Descartes, R. 1649. Passions of the Soul, E. Haldane and G. Ross, trans. and eds. The Philo-

sophical Works of Descartes. Cambridge: Cambridge University Press, 1982.Devitt, M. 1984. Thoughts and their ascription. In Causation and Causal Theories, Midwest

Studies in Philosophy IX, P.A. French, T.E. Uehling andH.K.Wettstein, eds.,Minneapo-lis, MN: University of Minnesota Press: 383–420.

Donnellan, K. 1966. Reference and definite descriptions. The Philosophical Review 75:281–304.

———. 1974. Speaking of nothing. The Philosophical Review 83:3–31.Dretske, F. 1995. Naturalizing the Mind. Cambridge, MA: MIT Press/A Bradford Book.Dummett, M. 1973. Frege: Philosophy of Language. London: Duckworth.———. 1976.What is a Theory of Meaning? (II), in Gareth Evans and JohnMcDowell, eds.,

Truth and Meaning. Oxford: Clarendon Press: 67–137.———. 1978. Truth and Other Enigmas. Cambridge, MA: Harvard University Press.———. 1981. The Interpretation of Frege’s Philosophy. London: Duckworth.Evans, G. 1981. Understanding Demonstratives. In Meaning and Understanding, H. Paret

and J. Bouvresse, eds. New York: De Gruyter, 280–303.———. 1982. The Varieties of Reference. J. McDowell, ed. Oxford: Clarendon Press (Chapter

VII; this volume).Ezcurdia, M. this volume. Thinking about myself.Fodor, J. 1987. Psychosemantics. Cambridge, MA: MIT Press/A Bradford Book.———. 1994. The Elm and the Expert. Cambridge, MA: MIT Press/A Bradford Book.———. 1998. Concepts: Where Cognitive Science Went Wrong. Cambridge, MA: MIT Press/A

Bradford Book.Frege, G. 1892. Über Sinn und Bedeutung. Zeitschrift für Philosophie und philosophische kritik

100:25–50. H. Kaal, trans. In McGuinness, 1984:157–77.———. 1918/19. Der Gedanke: eine logische Untersuchung. Beiträge zur Philosophie der

deutschen Idealismus I:58–77. H. Kaal, trans. In McGuinness, 1984:351–372.García-Carpintero, M. 1998. Indexicals as Token-Reflexives. Mind, 107: 529–563.Geach, P.T. 1957. On Belief About Oneself. Analysis 18: 23–24.———. 1957. Mental Acts. London: Routledge, Kegan Paul.Gopnik, A. 1993. How We Know Our Minds: The Illusion of First-Person Knowledge of

Intentionality. Behavioral and Brain Science 16: 1–14. Reprinted in Alvin Goldman, ed.1993. Readings in Philosophy and Cognitive Science. Cambridge, MA: MIT Press/ABradford Book: 315–46.

Bibliography 271

Guttenplan, S., ed. 1975. Mind and Language. Oxford: Clarendon Press.Hacking, Ian. 1995. Rewriting the Soul. Princeton: Princeton University Press.Hermes, H., F. Kambartel, and F. Kaulbach, eds., P. Long and R. White, trans. 1979. Gottlob

Frege: Posthumous Writings. Chicago: The University of Chicago Press.Heyes, C.M. 1994. Reflections on Self-recognition in Primates. Animal Behaviour 47: 909–19.Hintikka, J. 1958. Toward a Theory of Definite Descriptions. Analysis 19: 78–95.———. 1962. Knowledge and Belief. Cornell: Cornell University Press.———. 1967. Individuals, possible worlds, and epistemic logic. Nous 1: 33–62.Horgan, T. and J. Tienson. 1996. Connectionism and the Philosophy of Psychology. Cambridge,

MA: MIT Press/A Bradford Book.Hume, D. 1911. A Treatise of Human Nature. London: Dent Publishing Co.Humphrey, N. 1984. Consciousness Regained: Chapters in the Development of Mind. Oxford:

Oxford University Press.Jaynes, J. 1976. The Origin of Consciousness in the Breakdown of the Bicameral Mind. Boston:

Houghton Mifflin.Kant, I. 1798.Anthropology from a Pragmatic Point of View (Ak VII). MaryMcGregor, trans.,

The Hague: Martinus Nijhoff, 1974 [‘Ak’ stands for Kant’s Gesammelte Schriften, ed.Koniglichen Preussischen Akademie der Wissenschaften, 29 vols. Berlin: Walter deGruyter et al., 1902–, commonly known as the Akademie edition].

———. 1781/87. Critique of Pure Reason. Norman Kemp Smith, trans. in Immanuel Kant’sCritique of Pure Reason. London: Macmillan Co. Ltd., 1963. Also trans. Werner S.Pluhar. Indianapolis, IN: Hackett Publishing Co., 1996, and Paul Guyer and AllanWoods, Cambridge: Cambridge University Press, 1998.

———. 1783. Prolegomena to Any Future Metaphysics. (Ak IV) P. Carus, trans. Indianapolis,IN: Library of the Liberal Arts, 1950.

Kaplan, D. 1969. Quantifying in, in Donald Davidson and Jaako Hintikka, eds., Words andObjections. Dordrecht: Reidel: 206–242.

———. 1989a. Afterthoughts. In Themes from Kaplan, J. Almog, J. Perry, and H. Wettstein,eds. New York: Oxford University Press, 1989:565–614.

———. 1989b. Demonstratives. In Themes from Kaplan, J. Almog, J. Perry, and H. Wett-stein, eds. New York: Oxford University Press, 1989:481–563.

Katz, J. 1987. Descartes’s Cogito. Pacific Philosophical Quarterly 68: 175–96.Lewis, D. 1976. Survival and Identity. In A.O. Rorty, ed. The Identities of Persons. Berkeley:

University of California.———.1979. Attitudes de dicto and de se. The Philosophical Review 88:513–43.———.1983. Postscripts to ‘Survival and Identity’. In Philosophical Papers Vol. 1. New York:

Oxford.Locke, J. 1690. An Essay Concerning Human Understanding. London: Basset. (Page references

in the text are to the Nidditch edition, Oxford: Oxford University Press.)McGinn, C. 1983. The Subjective View. Oxford: Clarendon Press.McGuinness, B., ed., H. Kaal, trans. 1980. Gottlob Frege: Philosophical and Mathematical

Correspondence. Oxford: Basil Blackwell.McGuinness, B., ed., 1984. Gottlob Frege: Collected papers. Oxford: Basil Blackwell.Mellor, H. 1988. I and Now. Proceedings of the Aristotelian Society 89: 79–94.———. 1981. Real Time. Cambridge: Cambridge University Press.

272 Bibliography

Millikan, R.G. 1984. Language, Thought and Other Biological Categories. Cambridge MA:MIT Press/A Bradford Book.

———. 1990. The Myth of the Essential Indexical. Noûs 24: 723–734.———. 1993. White Queen Psychology and Other Essays for Alice. Cambridge MA: MIT

Press/A Bradford Book.———. 1994. On Unclear and Indistinct Ideas. In James Tomberlin, ed., Philosophical

Perspectives, vol. VIII. Atascadero CA: Ridgeview Publishing: 75–100.———. 1996. Pushmi-pullyu representations. In James Tomberlin, ed., Philosophical

Perspectives, vol. IX. Atascadero CA: Ridgeview Publishing: 185–200. Reprinted in L.May and M. Friedman, eds.1996. Mind and Morals. Cambridge MA: MIT Press/ABradford Book:145–161.

———. 1997. Images of identity: in search of modes of presentation. Mind 106: 499–519.———. 1998a. A common structure for concepts of individuals, stuffs, and basic kinds:

more mama, more milk andmore mouse. Behavioral and Brain Sciences 22.1, Feb 1998.Reprinted in E. Margolis and S. Laurence eds. 1998. Concepts: Core Readings. CambridgeMA: MIT Press/A Bradford Book.

———. 1998b.With enemies like this I don’t need friends: author’s response. Behavioral andBrain Sciences 22.1, Feb 1998.

Montagne, R. 1974. Formal Philosophy. New Haven: Yale University Press.Moore, G.E. 1955. Wittgenstein’s Lectures in 1930–33 (III). Mind 64: 1–27.Nagel, T. 1965. Physicalism. Philosophical Review 74: 339–56.———. 1970. The Possibility of Altruism. Oxford: Clarendon Press.———.1979a. Brain Bisection and the Unity of Consciousness. In Mortal Questions

Cambridge: Cambridge University Press: 147–164.———. 1979b. What Is It Like to Be a Bat? In Mortal Questions. Cambridge: Cambridge

University Press: 165–180.O’Brien, L.F. 1995a. Evans on Self-Identification. Noûs 29: 232–247.———. 1995b. The Problem of Self-Identification. Proceedings of the Aristotelian Society 95:

235–251.Parfit, D. 1971. Personal Identity. Philosophical Review 80: 3–27.———. 1976. Lewis, Perry and What Matters. In A.O. Rorty, ed. The Identities of Persons

Berkeley: University of California Press.———. 1984. Reasons and Persons. Oxford: Clarendon Press.Peacocke, C. 1983. Sense and Content: Experience, Thought and Their Relations. Oxford:

Oxford University Press.———. 1992.A Study of Concepts.Cambridge, MA:MIT Press/A Bradford Book.———. 1997a, this volume. First-person Reference, Representational Independence, and

Self-knowledge. Also in Wolfgang Kunne, Albert Newen, and Martin Anduschus, eds.Direct Reference, Indexicality, and Propositional Attitudes. Stanford, CA: CSLI Publica-tions.

———. 1997b. Conscious Attitudes, Attention and Self-Knowledge. In C. MacDonald, B.Smith, and C.Wright, eds. Knowing Our Own Minds: Essays on Self-Knowledge. Oxford:Oxford University Press.

Perner, J. 1991. Understanding the Representational Mind. Cambridge, MA: MIT Press/ABradford Book.

Bibliography 273

Perry, J. 1972. Can the Self Divide? The Journal of Philosophy 69: 463–88.———. 1977. Frege on demonstratives. Philosophical Review 86:474–97. Reprinted in J.

Perry. 1993. The Problem of the Essential Indexical. New York: Oxford University Press:3–26.

———. 1979, this volume. The problem of the essential indexical. Noûs 13:3–21. Reprintedin J. Perry. 1993. The Problem of the Essential Indexical. New York: Oxford UniversityPress: 33–53.

———. 1986. Thoughts without Representation. Supplementary Proceedings of the Aristote-lian Society 60: 137–52. Reprinted in J. Perry. 1993. The Problem of the EssentialIndexical. New York: Oxford University Press: 205–219.

———. 1988. Cognitive significance and new theories of reference. Noûs 22:1–18. Reprintedin J. Perry. 1993. The Problem of the Essential Indexical. New York: Oxford UniversityPress: 227–245.

———. 1990. Individuals in informational and intentional contexts. In E. Villaneuva, ed.Information, Semantics, and Epistemology. Cambridge: Basil Blackwell: 172–89. Reprint-ed in J. Perry. 1993. The Problem of the Essential Indexical. New York: Oxford UniversityPress: 279–300.

———. 1990. Self-Notions. Logos 11: 17–31.Pippin, R. 1987. Kant on the Spontaneity of Mind. Canadian Journal of Philosophy 17:

449–76.Powell, C.T. 1990. Kant’s Theory of Self-Consciousness. Oxford University Press.Quine, W. V. O. 1966. Quantifiers and propositional attitudes. In Ways of Paradox. New

York: Random House: 183–194.Quinton, A.M. 1962. Spaces and Times. Philosophy 37: 130–147.Récanati, F. 1993. Direct Reference. Oxford: Blackwells.Rosenthal, D. 1991. The Nature of Mind. Oxford: Oxford University Press.Rovane, C. 1987. The Epistemology of First Person Reference. Journal of Philosophy 84:

147–167.———. 1990. Branching Self-Consciousness. Philosophical Review 99: 355–95.Ryle, G. 1949. The Concept of Mind. London: Hutchinson & Co.Schacter, D. 1996. Searching for Memory: The Brain, the Mind and the Past. New York: Basic

Books.Scott, D. 1970. Advice onmodal logic. In Karel Lambert, ed., Philosophical Problems in Logic.

Dordrecht: Reidel: 143–173.Seager, W. 1990. The Logic of Lost Lingens. Journal of Philosophical Logic 19: 407–428.Sellars, W. 1956. Empiricism and the Philosophy of Mind. In H. Feigl and M. Scriven, eds.

Minnesota Studies in the Philosophy of Science I. Minneapolis, MN: University ofMinnesota Press. Reprinted in Sellars, W. 1963. Science, Perception and Reality. London:Routledge, Kegan Paul.

———. 1970–1. “ … this I or he or it (the thing) which thinks … .” Proceedings of theAmerican Philosophical Association 44: 5–31.

Shoemaker, S. 1963. Self-Knowledge and Self-Identity. Ithaca, NY: Cornell University Press.———. 1968, this volume. Self-reference and self-awareness. Journal of Philosophy 65:

555–67.———. 1970. Persons and their Pasts. American Philosophical Quarterly 7: 269–285.

274 Bibliography

———. 1984. Personal Identity: AMaterialist’s Account. In S. Shoemaker and R. Swinburne,eds. Personal Identity. Oxford: Oxford University Press.

———. 1986. Introspection and the Self. Midwest Studies in Philosophy X: 101–120.Reprinted in Cassam 1994; page-references are to Cassam.

Sosa, E. 1990. Surviving Matters. Noûs 24: 305–30.Strawson, P.F. 1959. Individuals. London: Methuen.———.1966. The Bounds of Sense. London: MethuenTownsend, J. and J. Busemeyer. 1995. Dynamic Representation of Decision-Making. In R.

Port and T. van Gelder, eds. Mind as Motion. Cambridge, MA: MIT Press/A BradfordBook.

Unger, P. 1991. Identity, Consciousness, and Value. New York: Oxford University Press.Van Gelder, T. 1995. What Might Cognition Be, If Not Computation? Journal of Philosophy

91: 345–381.Williams, B. 1970. The Self and the Future. In Problems of the Self. Cambridge: Cambridge

University Press: 46–63.———. 1973. Imagination and the Self. In Problems of the Self. Cambridge: Cambridge

University Press: 26–45.———. 1978. Descartes: The Project of Pure Enquiry. Harmondsworth, Middlesex: Penguin.Wittgenstein, L.1921. Tractatus Logico-Philosophicus. D. Pears and B. McGuinness, trans.

London: Routledge, Kegan Paul, 1961.———. 1935–6. Wittgenstein’s Notes for Lectures on ‘Private Experience’ and ‘Sense Data’.

Philosophical Review 77, 1968:271–320.———. 1933–4. The Blue and Brown Books, 2nd ed. Oxford: Blackwell, 1969.———. 1953. Philosophical Investigations. G.E.M. Anscombe, trans. Oxford: Blackwell.———. 1975. Philosophical Remarks. R. Rhees, ed., R. Hargreaves and R. White, trans.

Oxford: Blackwell.

Index of names

AAnscombe, E., 28 n.15, 32–33, 101 n.14,

103, 105 n.20, 137 n.67, 181 n. 9Armstrong, D., 258, 260, 265 n.18Aune, B., 59 n.8Ayers, M., 207 n.2

BBarwise, J., 191Belnap, N., 59Bennett, J., 25Boër, S., 167Borges, J., 250Bratman, M., 159 n.6Brook, A., 2, 32, 255–256Bruner, J., 267Burge, T., 179 n.3, 181 n.7, 203 n.32Busemeyer, J., 251 n.3

CCampbell, J., 215n.Carnap, R., 41 n.10Cartwright, H., 55Cartwright, R., 60 n.9, 76 n.14Castañeda, H-N., 1, 2, 3, 6, 9, 10 n.2,

15–16, 27, 31–32, 34–35, 38 n. 7,81n, 85 n.4, 86 n.5, 96 n.3, 155 n.4,163 n.1

Cheney, D.L., 266Chihara, C., 59Churchland, P., 260

DDamasio, A., 248 n.2Davidson, D., 167Dennett, D., 163, 250, 267

Descartes, R., 85, 89, 95, 98, 232, 239, 241,263

DeVidi, R., 2–3, 23 n.11Devitt, M., 173Dretske, F., 16, 19–20, 257–261, 264Dummett, M., 41 n.12, 45, 116 n.38

EEvans, G., 1, 3, 4, 9, 28 n.15, 29 n.16, 30,

35–36, 39 n.8, 42–45, 47–48, 179,195 n.23, 205 n.1, 216, 221–222, 236

Ezcurdia, M., 4, 5, 11, 29

FFodor, J., 173, 180 n.5, 201 n.29Frllesdal, D., 159 n.6Frankfurt, H., 81n.Frege, G., 2–3, 15, 31–48, 58, 145, 152,

155 n.3, 179, 180 n.6, 198

GGage, P., 248 n.2Garcia-Carpintero, M. 181 n.8Geach, P.T., 52 n.1, 56, 70, 97–98Gettier, E., 58 n.4Ginet, C., 64–65Gopnik, A., 261Gunderson, K., 81n.Guyer, P., 9 n.1

HHacking, I., 267Heyes, C.M., 266Hintikka, J., 51 n.1, 59, 79 n.15, 150Hogan, M., 5Horgan, T., 251

276 Index of names

Horwich, P., 215n.Hume, D., 38, 64, 86, 88, 89, 117–119Humphrey, N., 266Husserl, E., 15

JJaynes, J., 261

KKant, I., 2, 6, 9–30, 64, 86, 95, 100 n.9, 103

n.16, 116, 233–236, 255–256, 258Kaplan, D., 122, 150, 157 n.5, 163 n.1,

164–165, 170 n.5, 181 n.8Katz, J., 244 n.1Kemp Smith, N., 9 n.1, 17 n.9Kretzmann, N., 68 n.12

LLeibniz, G.W., 13 n.5Lewis, D., 97 n.4, 163 n.1, 210 n.3Lichtenberg, G.C., 239Locke, J., 265 n.18Lycan, W., 167

MMacBride, F., 203 n.32Malcolm, N., 81n.Martin, R., 5McGinn, C., 163 n.1, 170Mellor, D., 5, 180, 183–186, 189Millikan, R.G., 4–5, 180, 190–196, 198

n.27, 202Montague, R., 152 n.2Moore, G.E., 82 n.1, 119 n.43, 263 n.17

NNagel, T., 10 n. 2, 86 n.5, 93 n.9, 100–101,

102 n.15, 136, 215n., 240

OO’Brien, L.F., 195 n.23, 203

PParfit, D., 29, 132, 210 n.3, 215n.

Peacocke, C., 2, 3, 4, 5–6, 11, 12 n.4, 16,19, 179

Perry, J., 1, 4, 8, 10 n.2, 27, 35–42, 43, 44,46, 48, 97 n.4, 163 n.1, 164, 168–169,175, 179, 180 n.4, 182–183, 191, 210n.3, 215n., 244, 253

Perner, J., 261Pippin, R., 21Powell, C.T., 21, 28 n.15

QQuine, W.V.O., 31, 150 n.1

RRécanati, F., 179, 191Reichenbach, H., 170 n.5Rosenthal, D., 16, 19Rovane, C., 181 n.9, 191, 203 n.31Russell, B., 31, 40, 55Ryle, G., 261, 263

SSchachter, D., 248 n.2Schiffer, S., 215n.Schopenhauer, A., 6, 233Scott, D., 152 n.2Seager, W., 6Sellars, W., 21, 55 n.2, 58 n.4, 59 n.8, 266Seyfarth, R.M., 266Shoemaker, S., 1, 2, 3–4, 6, 9, 10–11,

14–16, 16 n.7, 21, 25, 27–28, 31–32,35, 128, 132, 210, 221–222, 225,236–237

Sleigh, R.S., 65, 76n.14Snowdon, P., 215n.Soldati, G., 203 n.32Sosa, E., 210 n.3Strawson, P.F., 23, 55, 91–92, 97–98, 99,

106 n.23, 112 n.31, 120 n.44,127–128

TTienson, J., 251Townsend, J., 251 n.3

Index of names 277

UUnger, P., 210 n.3

VVan Gelder, T., 251

WWilliams, B., 6, 132–133, 240, 244, 247,

268Williamson, T., 215n.Wilson, M., 81n.

Wittgenstein, L., 2, 3, 6, 15, 16 n.7, 21, 22,25, 32, 81–82, 86, 95, 98 n.7,106–108, 110 n.29, 113, 205 n.1, 221,233, 236, 239, 255–256

Wood, A., 9 n.1Woods, M., 81n.Wright, C., 215n.

ZZemach, E., 215n.

In the series ADVANCES IN CONSCIOUSNESS RESEARCH (AiCR) the following titleshave been published thus far or are scheduled for publication:

1. GLOBUS, Gordon G.: The Postmodern Brain. 1995.2. ELLIS, Ralph D.: Questioning Consciousness. The interplay of imagery, cognition, and

emotion in the human brain. 1995.3. JIBU, Mari and Kunio YASUE: Quantum Brain Dynamics and Consciousness. An intro-

duction. 1995.4. HARDCASTLE, Valerie Gray: Locating Consciousness. 1995.5. STUBENBERG, Leopold: Consciousness and Qualia. 1998.6. GENNARO, Rocco J.: Consciousness and Self-Consciousness. A defense of the higher-order

thought theory of consciousness. 1996.7. MAC CORMAC, Earl and Maxim I. STAMENOV (eds): Fractals of Brain, Fractals of

Mind. In search of a symmetry bond. 1996.8. GROSSENBACHER, Peter G. (ed.): Finding Consciousness in the Brain. A neurocognitive

approach. 2001.9. Ó NUALLÁIN, Seán, Paul MC KEVITT and Eoghan MAC AOGÁIN (eds): Two Sciences

of Mind. Readings in cognitive science and consciousness. 1997.10. NEWTON, Natika: Foundations of Understanding. 1996.11. PYLKKÖ, Pauli: The Aconceptual Mind. Heideggerian themes in holistic naturalism. 1998.12. STAMENOV, Maxim I. (ed.): Language Structure, Discourse and the Access to Conscious-

ness. 1997.13. VELMANS, Max (ed.): Investigating Phenomenal Consciousness. Methodologies and Maps.

2000.14. SHEETS-JOHNSTONE, Maxine: The Primacy of Movement. 1999.15. CHALLIS, Bradford H. and Boris M. VELICHKOVSKY (eds.): Stratification in Cogni-

tion and Consciousness. 1999.16. ELLIS, Ralph D. and Natika NEWTON (eds.): The Caldron of Consciousness. Motivation,

affect and self-organization – An anthology. 2000.17. HUTTO, Daniel D.: The Presence of Mind. 1999.18. PALMER, Gary B. and Debra J. OCCHI (eds.): Languages of Sentiment. Cultural con-

structions of emotional substrates. 1999.19. DAUTENHAHN, Kerstin (ed.): Human Cognition and Social Agent Technology. 2000.20. KUNZENDORF, Robert G. and Benjamin WALLACE (eds.): Individual Differences in

Conscious Experience. 2000.21. HUTTO, Daniel D.: Beyond Physicalism. 2000.22. ROSSETTI, Yves and Antti REVONSUO (eds.): Beyond Dissociation. Interaction be-

tween dissociated implicit and explicit processing. 2000.23. ZAHAVI, Dan (ed.): Exploring the Self. Philosophical and psychopathological perspectives

on self-experience. 2000.24. ROVEE-COLLIER, Carolyn, Harlene HAYNE and Michael COLOMBO: The Develop-

ment of Implicit and Explicit Memory. 2000.25. BACHMANN, Talis: Microgenetic Approach to the Conscious Mind. 2000.26. Ó NUALLÁIN, Seán (ed.): Spatial Cognition. Selected papers from Mind III, Annual

Conference of the Cognitive Science Society of Ireland, 1998. 2000.27. McMILLAN, John and Grant R. GILLETT: Consciousness and Intentionality. 2001.

28. ZACHAR, Peter: Psychological Concepts and Biological Psychiatry. A philosophical analy-sis. 2000.

29. VAN LOOCKE, Philip (ed.): The Physical Nature of Consciousness. 2001.30. BROOK, Andrew and Richard C. DeVIDI (eds.): Self-reference and Self-awareness. 2001.31. RAKOVER, Sam S. and Baruch CAHLON: Face Recognition. Cognitive and computa-

tional processes. 2001.32. VITIELLO, Giuseppe: My Double Unveiled. The dissipative quantum model of the brain.

2001.33. YASUE, Kunio, Mari JIBU and Tarcisio DELLA SENTA (eds.): No Matter, Never Mind.

Proceedings of Toward a Science of Consciousness: Fundamental Approaches, Tokyo, 1999.2001.

34. FETZER, James H.(ed.): Consciousness Evolving. n.y.p.35. Mc KEVITT, Paul, Seán Ó NUALLÁIN and Conn Mulvihill (eds.): Language, Vision,

and Music. Selected papers from the 8th International Workshop on the Cognitive Science ofNatural Language Processing, Galway, 1999. n.y.p.

36. PERRY, Elaine, Heather ASHTON and Allan YOUNG (eds.): Neurochemistry of Con-sciousness. Neurotransmitters in mind. 2001.

37. PYLKKÄNEN, Paavo and Tere VADÉN (eds.): Dimensions of Conscious Experience.2001.

38. SALZARULO, Piero and Gianluca FICCA (eds.): Awakening and Sleep-Wake CycleAcross Development. n.y.p.

39. BARTSCH, Renate: Consciousness Emerging. The dynamics of perception, imagination,action, memory, thought, and language. n.y.p.

40. MANDLER, George: Consciousness Recovered. Psychological functions and origins ofconscious thought. n.y.p.

41. ALBERTAZZI, Liliana (ed.): Unfolding Perceptual Continua. n.y.p.

Date post:	21-May-2017
Category:	Documents
Upload:	justin-lee
View:	219 times
Download:	0 times

[Andrew Brook, Richard C. Devidi] Self-Reference and Self Awarness

Documents