the v a rietie s of

self- k now le d ge
a nna lisa c o liv a
Palgrave Innovations in Philosophy

Series Editors
Vincent Hendricks
Philosophy/MEF
University of Copenhagen
Copenhagen, Denmark

Duncan Pritchard
Department of Philosophy
University of Edinburgh
Edinburgh, UK
Palgrave Innovations in Philosophy is a new series of monographs. Each
book in the series will constitute the ‘new wave’ of philosophy, both in
terms of its topic and the research profile of the author. The books will be
concerned with exciting new research topics of particular contemporary
interest, and will include topics at the intersection of Philosophy and
other research areas. They will be written by up-and-coming young phi-
losophers who have already established a strong research profile and who
are clearly going to be leading researchers of the future. Each monograph
in this series will provide an overview of the research area in question
while at the same time significantly advancing the debate on this topic
and giving the reader a sense of where this debate might be heading next.
The books in the series would be of interest to researchers and advanced
students within philosophy and its neighboring scientific environments.

More information about this series at

http://www.springer.com/series/14689
 Annalisa Coliva

The Varieties of
Self-Knowledge
Annalisa Coliva
Department of Philosophy
Irvine, California, USA

Palgrave Innovations in Philosophy

ISBN 978-1-137-32612-6 ISBN 978-1-137-32613-3 (eBook)
DOI 10.1057/978-1-137-32613-3

Library of Congress Control Number: 2016936088

© The Editor(s) (if applicable) and The Author(s) 2016

The author(s) has/have asserted their right(s) to be identified as the author(s) of this work in accordance
with the Copyright, Designs and Patents Act 1988.
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of
illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and trans-
mission or information storage and retrieval, electronic adaptation, computer software, or by similar or
dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, express or implied, with respect to the material contained herein or for any
errors or omissions that may have been made.

Cover illustration: © Vincenzo Dragani / Alamy Stock Photo

Printed on acid-free paper

This Palgrave Macmillan imprint is published by Springer Nature

The registered company is Macmillan Publishers Ltd. London
 Series Editors’ Preface

Palgrave Innovations in Philosophy is a series of short monographs. Each

book will constitute the ‘new wave’ of pure or applied philosophy, in terms
of both topic and research angle, and will be concerned with ‘hot’ new
research areas in philosophy and neighbouring intellectual disciplines.
These monographs will provide an overview of an emerging area while
significantly advancing the debate on this topic and giving the reader a
sense of where this debate might be heading next. While the series will
devote attention to core topics of philosophy, it will also feature books
with an interdisciplinary outlook, as we believe that many of the most
exciting developments in our discipline involve a fusion of philosophy
with other subjects.

Vincent Hendricks, Copenhagen

Duncan Pritchard, Edinburgh

v
 Credits

Chapter 2 §§1-2 and Chapter 6, §4 draw on “Critical notice of Tyler

Burge’s Origins of Objectivity”, Disputatio 4/33, 2012, pp. 515–530 and
on “One variety of self-knowledge: constitutivism as constructivism”, in
A. Coliva (ed.) The Self and Self-Knowledge, Oxford, Oxford University
Press, 2012, pp. 212–242.
Chapter 5 §§1, 3.1 contain parts of “Peacocke’s self-knowledge”, Ratio
21, 2008, pp. 13–27 and of “Review of Jordi Fernández’ Transparent
Minds”, Theoria 81, 2014, pp. 442–445.
The Appendix contains “How to commit Moore’s paradox”, The
Journal of Philosophy CXII/4, pp. 169-192, 2016.

vii
 Acknowledgments

This book has developed over several years, since I first became interested
in the topic of self-knowledge during my stay at Columbia University
thanks to a Fulbright Research Fellowship and a Fellowship of the Italian
Academy at Columbia University in 2002–2004. I would like to thank
my host there, Akeel Bilgrami, for his encouragement as well as for regu-
lar conversations, at his place, on this and related topics. Thanks are due
also to Andreas Kemmerling, my host in Heidelberg, where I was visit-
ing in 2006–2007 under the auspices of an Alexander von Humboldt
Fellowship and where I further developed my ideas on the topic of self-
knowledge. I would also like to thank the participants at two workshops
on “The Self and Self-Knowledge” that I organised in Bigorio and at the
Institute of Philosophy in London, in 2004 and 2008, respectively, which
proved invaluable in shaping my ideas on this topic: Dorit Bar-On, Akeel
Bilgrami, John Campbell, Jane Heal, Conor McHugh, Lucy O’Brien,
Ted Parent, Christopher Peacocke, Eva Picardi, Jim Pryor, Ursula Renz,
Carol Rovane, Martine Nida-Rümelin, Barry Smith, Gianfranco Soldati,
Paul Snowdon and Crispin Wright. The workshops were made pos-
sible by the generous contributions of the Fonds National Suisse, the
Institute of Philosophy and the British Academy, which I hereby thank
for their support. Thanks are also due especially to Akeel Bilgrami, Alan
Millar and Margaret Gilbert for their comments on my work on Moore’s

ix
x Acknowledgments

paradox, and to Hanoch Ben-Yami for detailed comments on the section

on Wittgenstein.
An entire first draft of the book was presented at the COGITO Research
Center in the spring and fall of 2015. I therefore take this opportunity to
thank, in particular, Sebastiano Moruzzi, Giorgio Volpe, Francesco Spada
and Luca Zanetti for their insightful comments. Thanks are also due to
Carla Bagnoli, Maria Teresa Bradascio, Filippo Ferrari, Alessia Pasquali
and Patrizia Pedrini for occasional attendance at those seminars and their
comments on those occasions. I would also like to thank Paolo Leonardi
and Eva Picardi for providing such a congenial and stimulating research
environment at COGITO.
In the spring 2016 I presented a whole draft of the book in my graduate
seminar at UCI. I would like to thank students in attendance, particulary
Kyle Banick, John Fensel and Adam Sanders, for interesting discussions
and insightful comments, which helped me a lot to finalize the manu-
script. Last but not least, I would like to thank Duncan Pritchard for
his support and willingness to include this monograph in the Palgrave
Innovations in Philosophy series.
 Abbreviations

BlB Wittgenstein L. (1958) The Blue Book (Oxford: Blackwell).

PI Wittgenstein L. (1953) Philosophical Investigations (Oxford:
Blackwell).

xi
 Contents

1 Introduction 1
Bibliography 18

2 Varieties of Mental States 19

1 Perceptions and Sensations 20
1.1 The Objectivity of Perceptual Representation 22
1.2 Perceptual Contents 23
1.3 Sensory States and Sensations 25
2 Two Kinds of Propositional Attitudes: Dispositions and
Commitments 26
2.1 Propositional Attitudes as Dispositions 27
2.2 Propositional Attitudes as Commitments 31
3 Emotions 38
3.1 Emotions as Sensations 39
3.2 Emotions as Value Judgements 40
3.3 Emotions as Felt Bodily Attitudes 40
3.4 Emotions as Perceptions of Evaluative
Properties 42
3.5 The Borderline View of Emotions 46
4 Summary 47
Bibliography 48

xiii
xiv Contents

3 Varieties of Self-knowledge 51
1 First-personal Self-knowledge 52
1.1 Groundlessness 52
1.2 Transparency 58
1.3 Authority 62
2 Counterexamples from Content Externalism and
Cognitive Sciences? 67
3 Third-personal Self-knowledge 69
4 Summary 74
Bibliography 75

4 Epistemically Robust Accounts 77

1 Inner Sense Theories: Armstrong and Lycan 78
2 Inferential Theories: Gopnik and Cassam 84
3 Simulation Theories: Goldman and Gordon 88
4 Summary 95
Bibliography 96

5 Epistemically Weak Accounts 99

1 Peacocke’s Rational Internalism 100
2 Burge’s Rational Externalism 111
3 Evans’s Transparency Method 119
3.1 Fernández’s Epistemic Account 119
3.2 Moran’s Deliberative Account 122
4 Summary 128
Bibliography 130

6 Expressivism About Self-knowledge 133

1 At the Origins of Expressivism: Wittgenstein 134
2 Bar-On’s Neo-expressivism 151
3 Summary 159
Bibliography 160
 Contents xv

7 Constitutive Theories 163

1 The Left-to-Right Side of the Constitutive Thesis:
Shoemaker 167
2 The Right-to-Left Side of the Constitutive Thesis:
Wright 174
3 The Two Sides of the Constitutive Thesis: Bilgrami 183
4 A Metaphysically Robust Kind of Constitutivism:
Coliva 188
4.1 The First Half of the Constitutive Thesis:
Transparency 188
4.2 Objections from Empirical Psychology 194
4.3 The Second Half of the Constitutive Thesis:
Authority 197
5 Summary 212
Bibliography 215

8 Pluralism About Self-knowledge 217

1 Propositional Attitudes as Commitments:
The Limits of the Constitutive Account 219
2 Sensations, Basic Emotions, Perceptions and
Experiences: Constitutivism Meets Expressivism 222
2.1 Sensations 222
2.2 Basic Emotions 229
2.3 Perceptions and Perceptual Experiences 231
3 Propositional Attitudes as Dispositions and Complex
Emotions: Third-personal Self-knowledge 232
4 Summary 240

Appendix: Moore’s Paradox 243

Moorean and Wittgensteinian Analyses 245
The Constraints on Any Feasible Account of Moore’s
Paradox 252
What Moore’s Paradox Isn’t About: Jane’s Odd Case 253
What Moore’s Paradox Is About—First Pass 254
xvi Contents

What Moore’s Paradox Is About—Second Pass 260

An Objection 265
Bibliography 269

Index 279
 1
Introduction

The main and novel idea presented in this book is that self-knowledge—
that is, our knowledge of our own mental states—comes in many ways.
We have first-personal knowledge of our own mental states when, for
instance, we are immediately aware of our occurrent sensations. By con-
trast, we have third-personal knowledge when, for example, we realise
that we enjoy a given mental state by reflecting on our behaviour and by
inferring to its likely cause. Even when distinctively first-personal knowl-
edge is at stake, it must be kept in mind that we have a variety of mental
states. For instance, we enjoy sensations, such as pains and tickles, which
have a characteristic phenomenology, but also perceptions that have both
a phenomenal and a representational content; we have propositional atti-
tudes, such as beliefs, desires and intentions, and these come in various
fashions—that is, as dispositions and as commitments—hence, as the
result of one’s own deliberations based on considering evidence for or
against a given proposition or course of action. Finally, we enjoy emo-
tions, whose nature still escapes philosophical consensus. Such a variety
of mental states invites caution in propounding single, all-encompassing
accounts of how we may know each of these types of mental state. In par-
ticular, although it is clear that sensations and at least some emotions have

© The Editor(s) (if applicable) and The Author(s) 2016 1

A. Coliva, The Varieties of Self-Knowledge,
DOI 10.1057/978-1-137-32613-3_1
2 A. Coliva

a distinctive phenomenology and can also be had by creatures who can-

not self-ascribe them, it is more difficult to maintain that propositional
attitudes have an intrinsic phenomenology which can distinguish wishes
from hopes, say, or beliefs from acceptances, and so on. Perceptions too
have their typical phenomenology, but they also provide a representa-
tion of the environment around the perceiver, or of her body, which is
independent of the exercise of concepts, at least when “basic” perceptions
are at stake. Hence, they can be enjoyed by creatures who are incapable
of self-ascribing them. By contrast, for propositional attitudes as com-
mitments, it makes sense to hold that they can be constituted at least in
part by their very self-ascription, like when one deliberates by judging “I
intend to do such and so” and there does not seem to be any room for the
suggestion that one would thereby be tracking a pre-existing intention.
Even if one were convinced that first-personal self-knowledge is ulti-
mately an epistemic relation between a subject and a proposition or a
state of affairs, one should be open to the possibility that the methods
whereby one gets to know in a first-personal way one’s own mind can
vary depending on the kind of mental state at issue. Hence, one should
be open to pluralism about the methods whereby we gain self-knowledge,
which go from being inferential, or even based on simulative routines,
when third-personal self-knowledge is at stake, to possibly a variety of
means when first-personal self-knowledge is at stake, depending on the
kind of mental state one would know in such a way.
However, as will become apparent, in the case of so-called first-personal
self-knowledge, there is no real epistemic relation between a subject and a
proposition or state of affairs. So, to talk of “knowledge” in this connection
is somewhat a misnomer, brought about by habit—in particular, by the
philosophical habit of using that term and of mistaking certain conceptual
truths, which entrain, in appropriate conditions, the indubitability of one’s
own psychological self-ascriptions, as in fact due to a peculiar epistemic
relation holding between a subject and her own mental states.
Thus, the kind of pluralism propounded in this book is both a plural-
ism of methods when genuine knowledge of one’s mind is at stake—
that is, when third-personal self-knowledge is at stake—and of “states”,
as we may put it, for lack of a better term, when we are dealing with
first-personal self-knowledge.  Thus, in that connection, the term “self-
knowledge” is in fact used as a shorthand for a set of conceptual truths
 1 Introduction 3

which can be variously redeemed. Hence, in some cases, through a vari-

ety of methods, a state of knowledge of one’s own mind is obtained and
expressed through the relevant psychological self-ascriptions. In some
other cases, the relevant self-ascriptions (which in some cases may even
be superficially identical to the ones which express third-personal self-
knowledge)  express a different and non-epistemic kind of state, which
varies from merely showing or exhibiting one’s mind to bringing about
the relevant first-order mental states.
It is fair to say, however, that although by now a lot of philosophers work-
ing on self-knowledge—particularly on first-personal self-knowledge—are
aware of the limitations in scope of their preferred accounts1 and there-
fore are at least implicitly committed to pluralism about self-knowledge
(particularly of methods but perhaps, in some cases, of both methods
and states), they have been reluctant to embrace it explicitly.2 For some
reason, which seems mostly to reveal a monistic prejudice, they seem to
think that if their preferred theory has only limited application, it is not
interesting (or not interesting enough). Subject to a craving for gener-
ality, which, as said, is likely due to a deep-seated monistic prejudice,
they often attempt to extend their preferred theory of self-knowledge to
mental states that are, after all, resilient to the treatment, thus ending up
weakening their own accounts.3 Sometimes, in this vein, they realise that
the attempt to generalise their preferred accounts stumbles, in particular,
against the asymmetry between first- and third-personal self-knowledge;
as a consequence, they are led to denying it or to making it a difference
in degree rather than in kind.4 Or else, they tend to consider of limited
philosophical interest and significance the kinds of self-knowledge they
knowledgeably do not account for. Thus, you find theorists mostly inter-

1
A lot of them start out by presenting themselves as offering accounts of our knowledge of just this
or that kind of mental state.
2
A notable exception is Boyle 2009, 2011a. His kind of pluralism, however, is more limited than
the one defended in this book, for he mainly stresses the difference between first-personal self-
knowledge of propositional attitudes as commitments and of one’s passive mental states, such as
sensations and perceptions. Furthermore, he thinks that knowledge of our own beliefs is more
fundamental than any other kind of first-personal self-knowledge.  No such priority claim is
defended in this book.
3
Several authors whose views will be considered in the following chapters are subject to this criti-
cism, as we shall see.
4
Gopnik 1983 and Cassam 2014 are a case in point.
4 A. Coliva

ested in first-personal self-knowledge who downplay the importance of an

inquiry into third-personal self-knowledge, typically on the grounds that
it would not be especially interesting from an epistemological point of
view.5 Conversely, those who offer an account that works mostly for third-
personal self-knowledge, and who realise that they cannot fully account
for first-personal self-knowledge, insist on the irrelevance of the latter par-
ticularly to personal development vis-à-vis the importance of the former.6
Hence, the implicit bias towards monism can have various effects, going
from leading one to the pursuit of generality at the expense of credibility,
to the denial of structural differences between first- and third-personal
self-knowledge, or, finally, to being chauvinist with respect to those forms
of self-knowledge one admittedly cannot account for.
Therefore, the present book unashamedly buys into pluralism about
self-knowledge. It does so by first presenting in some detail the plu-
rality of mental states we enjoy and their intrinsic differences. It then
defends the existence of a deep asymmetry—that is, an asymmetry in
kind and not merely in degree—between first- and third-personal forms
of self-knowledge. It then reviews several theories of first-personal self-
knowledge, discussing their various pitfalls but also accepting those ker-
nels of truth they have, when they have them. In the last chapter, they are
put at the service of a pluralistic account of self-knowledge, both of meth-
ods, in particular when third-personal self-knowledge is at stake, and of
states, since, as anticipated, in many cases of so-called first-personal self-
knowledge, the relevant psychological self-ascriptions do not depend on,
and do not express, the obtaining of a genuine epistemic relationship
between a subject and her own first-order mental states, as it were.
In the second chapter, titled “Varieties of Mental States”, we introduce
the variety of mental states we enjoy. We explore and propose a system-
atisation of the complex geography of the mental. We first distinguish
between sensations and perceptions, by reference to the fact that only
the latter have correctness conditions, while allowing that their contents,
at least in the case of “basic” perceptions, may be entertained also by
creatures who do not possess the concepts necessary to their canonical

5
Most theorists whose views will be discussed in the following chapters do that.
6
Cassam 2014, who defends inferentialism, is a case in point. For a critical assessment, see
Coliva 2015a.
 1 Introduction 5

specification. We then move on to propositional attitudes and distinguish

between beliefs, desires and intentions as dispositions and as commit-
ments. Whereas the former may be independent of judgement and may
well be unconscious, the latter depend on judging either that P is the case
or that P would be good to have or do (in light of one’s further goals).
For such a reason, these mental states may also be called “judgement-
dependent” propositional attitudes. Moreover, they constitutively involve
the ability to accept criticism or of being self-critical if one does not live
up to them. Afterwards, we consider the complex case of emotions. As
is well known, there are, nowadays, a number of different and compet-
ing accounts, which range from identifying emotions with sensations to
equating them with evaluative judgements or with perceptions of values
or finally with felt bodily attitudes. It will be argued that none of these
theories seems entirely satisfactory, although a detailed treatment of each
falls beyond the scope of this book. It will be claimed that if this is the
case, it is really tempting to consider emotions as sui generis mental states,
sharing some features of other mental states while not reducing to any of
them. On this view, common sense would, after all, be right in consider-
ing emotions to be different from all other mental states and in grouping
them under one special category.
In chapter 3, titled “Varieties of Self-Knowledge”, we turn to the
characteristic traits of first-personal self-knowledge—namely, so-called
“transparency”, “authority” and “groundlessness”. At first approximation,
transparency amounts to the idea that subjects who possess the relevant
concepts, as well as being rational and possessed of normal intelligence,
are such that when they enjoy a given mental state they are immediately
in a position to self-ascribe it. Authority, in contrast, has it that subjects’
psychological self-ascriptions are correct, at least in the normal run of
cases. Finally, according to “groundlessness”, subjects’ psychological self-
ascriptions are not based on the observation of their own mental states
or on inference to the best explanation starting from their own observed
behaviour and possibly further aspects of their own psychology. In fact,
each element in this triad admits of different readings and in the chapter
we go to some length in discussing them. Furthermore, their domain of
application has to be properly limited and we spend time showing how that
should be done. The key idea defended in this chapter is that transparency,
6 A. Coliva

authority and groundlessness are not contingent but necessary and a priori
aspects of what goes by the name of first-personal self-knowledge. For mas-
sive failures at this kind of self-knowledge would display either the lack of
the relevant psychological concepts or failures at rationality. Rationality, in
this connection, has to be understood in a “thick”, rather than in a “thin”,
sense. The latter amounts to the idea that we are critical reasoners insofar
as we revise our propositional attitudes and goals on the basis of counter-
vailing reasons. However, I agree with several philosophers (Christopher
Peacocke, Dorit Bar-On and Quassim Cassam, just to mention a few)
who, contra Sydney Shoemaker and Tyler Burge, do not think that self-
knowledge is necessary for being critical reasoners. If that is the notion of
rationality one has in mind, then lack of self-knowledge will not make one
necessarily irrational. Yet we also have a thick notion of rationality, accord-
ing to which making certain psychological self-ascriptions and behaving in
ways which run systematically against them would impugn the idea that
we are confronted with a normal subject, up to the point of rendering her
pronouncements onto herself irrelevant, a mere flatus vocis devoid of any
significance, if not of meaning altogether. These characteristic traits of first-
personal self-knowledge will also be defended against possible objections
stemming from recent findings in cognitive sciences and from scepticism
about knowledge of the content of our own propositional attitudes deriv-
ing from the endorsement of semantic externalism. For instance, several
studies in cognitive science tend to show that we do not have knowledge of
our own character traits, that we are bad at affective forecasting—that is, at
figuring out how we would actually feel if some relevant change happened
to our lives—and, finally, that we are really poor at identifying the causes
of our decisions and further behaviour. None of this, however, shows that
we never have essentially first-personal self-knowledge. Rather, it shows
that its scope is limited and does not extend to our deep-seated and future
dispositions, or to the causal relations among our various mental states,
which are known, if and when they are, in a third-personal way. Yet all this
is compatible with the fact that we have essentially first-personal knowl-
edge of a wide range of mental states, such as our sensations, perceptions,
basic emotions and propositional attitudes as commitments.
Furthermore, some theorists take the rise of content externalism to be
incompatible with at least authoritative self-knowledge regarding one’s
 1 Introduction 7

current propositional attitudes, for, if externalism is correct, a subject

may think of being thinking a water thought, say, when she is in fact
entertaining a thought about twater, due to her actual causal connections
with an environment in which lakes, rivers and seas are in fact filled in
with XYZ, rather than H2O. Let us grant, for the sake of argument, that
externalism is correct. Let us further suppose that our subject is actually
thinking a twater thought, unbeknownst to her. Still, she would seem to
have essentially first-personal access to it, even if she may be wrong about
its actual content. Hence, she would still have transparent access to the
fact that she is entertaining a thought, rather than a hope or a wish; and
her access would still be groundless—that is, it would be based neither on
observation nor on inference. Finally, she would still be authoritative with
respect to its seeming or apparent content. (Some theorists would call it
“narrow content” and would happily acknowledge its existence alongside
with “wide content”, but we need not take a position about it here.)
In keeping with the characteristic aspects of the Palgrave Innovations in
Philosophy series, the volume then presents and critically discusses vari-
ous accounts of such privileged self-knowledge that have been proposed,
with special emphasis on contemporary versions of each of these theories.
Hence, in the fourth chapter, titled “Epistemically Robust Accounts”,
we start by considering the inner-sense account of self-knowledge. This
model tends to equate self-knowledge with forms of knowledge based
on outer observation, though granting a subject’s privileged access to her
own mental states. In particular, its contemporary versions, due mostly
to David Armstrong and William Lycan, claim that we have a reliable
inner mechanism that “scans” our first-order mental states and produces
the corresponding second-order ones. The chief objection will be that the
model presupposes a crude form of reliabilism that severs the constitutive
connection between self-knowledge, rationality and concepts’ possession.
We then turn to inferentialist accounts of self-knowledge. The infer-
entialist model tends to assimilate self-knowledge to knowledge of other
people’s mental states. Recently, it has been taken up and partially re-
fashioned by Alison Gopnik, who has developed a “theory-theory”
account. Within the first 4 years of life, children acquire and develop a
little theory of the mind, which they apply to both themselves and oth-
ers, in order to (self-)ascribe mental states starting from the observation
8 A. Coliva

of overt behaviour (or other “inner promptings”). Her views have given
rise to a heated debate, at the interface of philosophy of mind, psychol-
ogy and neuroscience, between supporters of the theory-theory approach
and partisans of so-called “simulation” theories, such as Alvin Goldman
and Robert Gordon. According to simulation theorists, who are other-
wise divided on many issues, knowledge of other people’s mental states
is not based on the application of a theory but on the simulation of the
other person’s point of view, which gives rise to a psychological ascription
based on what one oneself would feel and think if one were in the other
person’s shoes. These views are exposed and critically examined. The main
objection against the inferentialist account is that it implausibly assimi-
lates first-personal self-knowledge to knowledge of other people’s mental
states. Furthermore, it runs the risk of providing a circular account of self-
knowledge and it succumbs as soon as one tries, like in Quassim Cassam’s
recent version of it, to make it transcend its proper domain of applica-
tion. The main criticism against simulation theories, in contrast, is that
they are in fact unclear about how we would get knowledge of our own
minds, on the basis of which we should then gain knowledge of other
people’s mental states, and risk falling back onto other, problematical
models of self-knowledge (such as the inner-sense model). Simulation
theorists, in particular Gordon, also have interesting but underdevel-
oped views about the nature and acquisition of psychological concepts,
such as the concept of belief and of other propositional attitudes. Still,
both inferentialism and simulative accounts have important things to say
about some instances of third-personal self-knowledge, such as knowl-
edge of our deep-seated dispositions and the kind of self-knowledge we
can gain through affective forecasting.
In the fifth chapter, called “Epistemically Weak Accounts”, we intro-
duce and assess various models, which are united in claiming that
self-knowledge is indeed a kind of modest, yet genuinely cognitive
achievement, while trying to avoid construing self-knowledge either as
due to the operations of an inner scanning mechanism or as a form of
inferential knowledge. For this reason, they are called “epistemic” mod-
els. They come in various fashions, however. Some of them can be traced
back to some remarks by Gareth Evans in The Varieties of Reference.
According to Evans, in order to know our own beliefs, we need only
 1 Introduction 9

to look outward, see whether we can answer “yes” to the question as

to whether P is the case, and then preface P with “I believe”. Recently,
Evans’s insights have been developed especially by Richard Moran but
also by André Gallois, Jordi Fernández and Alex Byrne. We pay special
attention to Fernández’s and Moran’s more thorough accounts. Both are
found wanting, even though for different reasons. The former is criticised
for implausibly claiming that the evidence which justifies one’s belief in P
is also the one that justifies one’s self-ascription of that belief. The latter,
in contrast, is criticised for not offering any suitable explanation of why
self-knowledge of our propositional attitudes should, after all, count as
an epistemic achievement and for tending to equate first-personal self-
knowledge with making up one’s mind. Intuitively, however, we also have
first-personal self-knowledge of several mental states which are not the
result of any deliberation on our part, such as sensations, perceptions and
(at least basic) emotions.7
Significantly different, yet still epistemic, accounts have been proposed
by Peacocke and Burge. Peacocke, in particular, places crucial emphasis
on the fact that first-order propositional attitudes have a characteristic
phenomenology. Accordingly, there is something that it is like to judge
that P, for instance. We are therefore aware of our judgement that P, qua
such a judgement and, by tacitly applying the rule that if one judges that P,
one believes it, we correctly self-ascribe the corresponding belief. Burge’s
account, finally, takes self-knowledge to be a requirement of rationality
(in a “thin” sense): in order to be rational thinkers, we must be prepared
to revise our beliefs on the basis of countervailing evidence. Hence, we
are entitled—that is, non-discursively justified—to self-ascribe them.
Such a second-order belief, in turn, amounts to knowledge since it is true
and justified (albeit non-discursively).
The main objection against those epistemic accounts that devote special
attention to inner phenomenology is that such a distinctive phenomenol-
ogy does not really differentiate between various kinds of propositional
attitudes. For instance, it is difficult to say what distinguishes hopes
from wishes at the phenomenological level. This will have a direct bear-
ing on Peacocke’s position. For, if the phenomenology is not sufficiently

7
For a criticism of Byrne’s development of Evans’s ideas, see Boyle 2011a.
10 A. Coliva

fine-grained to license a specific psychological attribution, it cannot be

appealed to in order to explain self-knowledge along the lines proposed
by Peacocke. Furthermore, it is claimed, against Peacocke’s account, that
it runs the risk of providing a circular account of our knowledge of our
propositional attitudes. For, if, in order to avoid the previous problem,
it posits a subject’s antecedent knowledge of her own beliefs (or of other
related propositional attitudes such as judgements vis-à-vis beliefs), it
would actually presuppose self-knowledge rather than explain it.
In addition, Burge’s account is criticised mainly either for implausibly
claiming that “thin” rationality requires knowledge of the kind of attitude
one is enjoying or for resting on an ad hoc notion of rationality which
compromises the interest of his theory. Moreover, claiming that self-
knowledge is constitutive of being a reasoner does not provide an epis-
temic account of it. It merely points out an a priori connection. Indeed,
if Burge were to supplement his account by saying that one gets to know
one’s attitudes through the operation of some reliable cognitive mecha-
nism, the epistemic aspects of his account of self-knowledge would be
dangerously close to crude reliabilist theories of self-knowledge, already
presented and criticised in the previous chapter.
In Chap. 6—“Expressivism About Self-Knowledge”—we move on to
expressivist accounts of our knowledge of our own mental states. The
basic, underlying idea is that self-ascriptions of mental properties are ways
of expressing our own minds other than in natural and instinctive ways,
such as by means of cries and laughter or other behavioral manifestations.
After presenting and critically examining Wittgenstein’s approach, which
is at the origins of expressivist positions, as well as of some aspects of
constitutive ones (that are reviewed in Chap. 7), we dwell on Dorit Bar-
On’s recent and powerful defence of that model. Though generally sym-
pathetic to that approach, we highlight the fact that, after all, it seems
much better suited to account for our knowledge of sensations rather than
of propositional attitudes, and certainly it cannot be generalised across
the board to provide an all-encompassing account of our knowledge of
our minds. In particular, it does not explain those cases in which our
first-order mental states originate in our self-ascriptions, like when we
deliberate “I intend to φ” or judge “I judge/opine/wish … that P” and
there does not seem to be room for the idea that we would thereby be
 1 Introduction 11

expressing a pre-existing mental state. Nor does it explain how we can

actually have knowledge, obtained through a cognitive achievement, of
a lot of dispositional mental states we enjoy. Furthermore, difficulties
emerge as soon as one tries to combine expressivism with the view that
first-personal self-knowledge is, after all, the result of some sort of cogni-
tive achievement, like in Bar-On’s account. For if the model presupposes
the existence of an inner scanning mechanism, it falls prey to the objec-
tions raised against inner-sense theories. If, in contrast, it presupposes
some other kind of epistemic access to one’s own first-order mental states,
it succumbs to the difficulties presented against Burge’s idea that we are
entitled to our psychological self-ascriptions. Bar-On’s new “expressive
entitlements”, moreover, are reviewed and found wanting. Hence, the
supposed advantage of expressivism over its rivals, which should allegedly
consist in avoiding observationalism, inferentialism and other unpalatable
accounts of the epistemology of first-personal self-knowledge, is spoiled.
Still, expressivism has something important to say about our “knowledge”
of our own sensations and basic emotions; moreover, it can be extended to
our “knowledge” of our own perceptions and can offer interesting insights
about the nature and the acquisition of several psychological concepts.
These insights will be built upon in the final chapter of the volume.
In the seventh chapter, so-called “constitutive” accounts of self-
knowledge are dealt with. At the heart of this kind of approach lie two
main ideas. First, that first-personal self-knowledge is not the result of
any cognitive achievement but rather consists in some conceptual truths,
corresponding to transparency, authority and groundlessness (see Chap.
3), which can be variously redeemed. Hence, properly speaking, self-
knowledge is not really a form of knowledge. This result is indirectly sup-
ported by the failure of the various attempts to account for first-personal
self-knowledge as a real cognitive accomplishment examined in previous
chapters. Second, proper constitutive positions are characterised by two
metaphysical claims. The first one is that, under specifiable conditions,
first- and second-order mental states do not have separate existence. The
second is that, at least in part and under specifiable conditions, our first-
order mental states are constituted by their very self-ascription.
The model has been defended in various ways starting with  Sydney
Shoemaker’s pioneering work, through Crispin Wright’s and Jane Heal’s
12 A. Coliva

linguistic version of constitutivism, up to Akeel Bilgrami’s agential ver-

sion of constitutivism. A profitable way of presenting their debate is to
see them as according different priorities to either side of the following
biconditional, known as the constitutive thesis, and as providing differ-
ent characterisations of its C-conditions: Given C, one believes/desires/
intends that P/to φ iff one believes (or judges) that one believes/desires/
intends that P/to φ.
According to Shoemaker, priority must be given to its left-to-right side
and the C-conditions must be characterised by reference to subjects who
possess normal intelligence and rationality and are endowed with the rele-
vant psychological concepts. According to Wright,8 in contrast, the right-
to-left side is the fundamental one and the C-conditions must refer to
the communal linguistic practice of making psychological avowals, which
are usually taken as authoritative. Finally, according to Bilgrami, the two
sides of the biconditional are on a par and the C-conditions must make
reference to the fact that the mental states at issue are such that it makes
sense to regard the subject as responsible for them—that is, to be either
blame- or praise-worthy for them. Each of these positions is presented and
found wanting either for resting on dubious a priori claims regarding, for
instance, the necessity of self-knowledge for being a reasoner or for failing
to vindicate the central metaphysical contentions of constitutivism.
We then introduce a metaphysically robust brand of constitutivism,
which is claimed to hold for only a very limited class of mental states—
namely, for those propositional attitudes as commitments we undertake by
deliberating what to believe, desire, intend to do, and so on, on the basis of
evaluating (or at least of being able to evaluate) evidence in favour of P/φ-
ing or of its desirability or advisability. When these propositional attitudes
are at stake and the subject is endowed with the relevant psychological
concepts, which are acquired “blindly”,9 both sides of the biconditional
hold as a matter of conceptual necessity, and, in particular, the right-to-left
side actually makes good the second metaphysical commitment charac-
teristic of constitutive accounts. Thus, adult human beings actually have

8
Heal 2002 too defends this position, but our exposition will focus on Wright’s more thorough
account.
9
That is to say, by being drilled to substitute their immediate avowal, “P”, “P would be good to
have”, “I will φ”, with the corresponding psychological one—that is, “I believe that P”, “I want/
desire that P”, “I intend to φ”.
 1 Introduction 13

two ways of forming commitments, either by judging their contents or by

directly self-ascribing them. In the latter case, then, authority is secured in
a much stronger way, since the psychological self-ascription is actually self-
verifying. Furthermore, the account is supplemented by an explanation
of how we acquire and canonically deploy the relevant psychological con-
cepts, which does away with the idea that psychological concepts are either
tags for mental states one should already have in view or a priori rules one
should self-consciously apply, often by having in view either other mental
states or even the very mental states one would thereby categorise. This
account, in turn, helps to make good the first metaphysical claim at the
heart of constitutive positions—namely, that when subjects are rational,
intelligent and conceptually endowed, first-order mental states and their
self-ascriptions do not have separate existence. For the latter are seen as
replacements of instinctive and direct forms of expression of one’s ongoing
first-order mental states, which are integral to those very first-order mental
states, rather than as different mental states entered by making judgements
about already-singled-out first-order mental states.
Such a position is then defended against the objection that we may be
self-deceived and thus ascribe to ourselves a mental state—particularly a
propositional attitude—we in fact lack. The key move consists in deny-
ing—following Bilgrami’s lead—that self-deception is a case in which
one goes wrong about one’s first-order mental states. Rather, it consists
in having two mutually inconsistent propositional attitudes—one as a
commitment and one as a disposition—which give rise to a subject’s
somewhat irrational behaviour. Yet one’s self-ascription of the commit-
ment is actually correct, even if one happens to behave in ways which run
contrary to it because of one’s counter dispositions.
In the eighth and final chapter, called “Pluralism About Self-
Knowledge”, a pluralist account of self-knowledge is put forward. As
the discussion in Chap. 7 makes apparent, constitutive accounts can
hold in their full-blooded version for only our (so-called) knowledge of
our propositional attitudes as commitments. By contrast, it is argued
that knowledge of one’s own propositional attitudes as dispositions is
achieved through a variety of methods. Hence, we sometimes know them
through inference to the best explanation—in the same way in which we
can know of other people’s mental states by inferring to them from their
owners’ overt behaviour and by exploiting some general theory of the
14 A. Coliva

mind. However, only in one’s own case can the inference be based on rel-
evant inner promptings, such as sensations, emotions and further mental
states. In some other cases, instead, it can depend on deploying simulative
methods, like when we engage in affective forecasting. Sometimes, we gain
knowledge of our minds by relying on other people's judgements about
us. Our self-knowledge is therefore achieved though testimony. Finally,
knowledge of our dispositional mental states can be obtained by means of
the self-conscious deployment of highly dispositional psychological con-
cepts. In this case, there is inferential reasoning going on, but it is not a
kind of inference to the best explanation. Rather, it consists in subsuming
some aspects of one’s overall behaviour and mental states under a concept
by self-consciously exploiting its characteristic notes.
Strong constitutive accounts have limited purchase also because, con-
trary to what some of their supporters hold, they do not extend to past self-
ascriptions of propositional attitudes as commitments, which are known,
when they are, on the basis of mnestic evidence. Still, it is true that being
able to remember one’s past mental actions, or indeed other mental states,
as well as one’s own past actions, is constitutive of being a cognitively well-
functioning human being. Yet that does not mean that we can account for
our knowledge of these past mental states along constitutivist lines.
Moreover, strong constitutive accounts are not apt to explain self-
knowledge of our sensations and of other mental states that have a dis-
tinctive phenomenology and that are clearly independent of our ability to
self-ascribe them,10 such as bodily sensations, basic emotions, perceptions
and perceptual experiences. Here, the most promising account will have
to forsake the second metaphysical claim at the heart of strong constitu-
tive explanations, according to which psychological self-ascriptions can
at least partially constitute the first-order mental states they ascribe to a
subject. What remains are simply the other characteristic claims of con-
stitutive positions, according to which conceptually competent creatures
are authoritative, at least in the normal run of cases, with respect to their
own mental states and are immediately in a position to self-ascribe them
without observing either their own mental states or their overt behav-
iour. These first-order mental states, however, can exist independently

10
Pace McDowell 1994.
 1 Introduction 15

of their self-ascription. Hence, the allegedly epistemic problem of self-

knowledge becomes the problem of explaining how the relevant concepts
are acquired and canonically applied without falling back into observa-
tional or inferential models. Expressivism is once again crucial in this
connection because it allows one to avoid these pitfalls. In particular, the
idea is put forward that when we deal with self-ascriptions of sensations
and occurrent basic emotions, which have a distinctive (often bodily)
phenomenology, possessing the relevant concepts is the result of having
been drilled to substitute their more immediate expressions with verbal
behaviour. This conceptual drilling is what gives rise to their character-
istic first-personal “knowledge”. Yet the latter is crucially not the result
of any, however modest, cognitive achievement. Hence, the use of the
term “knowledge” in this connection is more the—“grammatical”, as
Wittgenstein would have it—signal of the absence of room for sensible
doubt and ignorance (at least in the normal run of cases) rather than
the mark of a genuinely epistemic relationship between a subject and
her own sensations and basic emotions. Furthermore, seeing the avowal
as a replacement of more instinctive forms of behaviour helps vindicate
the claim that the first-order mental state and its self-ascription are not
separate existences.
Similarly, we propose an expressivist account of our “knowledge” of
our own perceptions which is held to originate in blind drilling. The
idea, once more, is that we first learn to voice their contents and, on that
basis, we are drilled to express ourselves by prefacing such contents with
“I see that” or “I hear that”, and so on. Therefore, our knowledge of our
perceptions does not usually require us to attend to our experiences and
to identify them as seeings (or hearings, etc.) either directly or through
the application of a little psychological theory.
The case of non-basic emotions is different. While basic emotions like
fear can be conceptualized and expressed just like expressivism recom-
mends, with more complex emotions, such as jealousy or envy, we usually
know them by attending to a complexity of factors, such as their charac-
teristic phenomenological aspects (if and when they have them) as well as
our own behaviour in contextually salient occasions. Moreover, we usu-
ally infer from these data to their likely causes, such as the love for a given
person or the envy for her success, and so on. Indeed, our application of
16 A. Coliva

this little theory may often take place in rapid and almost unnoticeable
ways but only because we are already proficient in applying it. Indeed,
genealogically or in new, unexpected cases, it will require time and effort
and possibly help from a third party. For we may well be at a loss about
how to interpret the pool of data about ourselves we may have collected.
That is to say, we may need the intervention of another person to be in a
position to infer that our characteristic feelings and behaviour are signs of
love or envy. Moreover, a lot of our third-personal self-knowledge, such
as affective forecasting or knowledge of our deep-seated dispositions, will
depend on simulating relevant aspects of a given situation to see how we
would react to it, thereby acquiring some insight into our own nature
and character. Reading novels and watching movies can achieve similar
results insofar as we may identify with the protagonists or be prompted to
simulate salient aspects of the plot to see how we would react if we found
ourselves in those situations.
Finally, it should be stressed that, contrary to the kind of self-verifying
self-ascriptions that have commitments as contents, in all cases in which
psychological self-ascriptions substitute more instinctive forms of behav-
iour, there is, however, limited room for error. Owing to slips of the
tongue or to somewhat impaired cognitive conditions, a subject could
actually voice sensations, basic emotions or perceptions she is not actually
enjoying. Yet constitutivism can take care of these possibilities by appro-
priately specifying the relevant C-conditions. By contrast, when the self-
ascription of dispositions or of non-basic emotions is at stake, there is no
default presumption that a subject should be authoritative with respect
to them. For she will be as exposed to error as she would be if she were
applying her psychological theory in order to get knowledge of another
person’s mental states.
At least since Shoemaker’s work, an account of self-knowledge has been
taken to have a bearing on the perplexing yet fundamental phenomenon
of Moore’s paradox—the paradox, that is, consisting in judging “P, but
I do not believe it” or “I believe that P, but it is not the case that P”.
Accordingly, in the Appendix, the proposed account of commitments and
their distinctively first-personal self-knowledge is brought to bear on it.
In particular, it is claimed that only by countenancing propositional atti-
tudes as commitments can Moore’s paradox so much as exist. By contrast,
 1 Introduction 17

if one took its doxastic conjuncts to express (the lack of ) beliefs as dispo-
sitions, the paradox would, surprisingly, disappear. Indeed, the case of a
self-deceived subject who discovers her self-deception can perfectly well
illustrate the point. For one may find oneself in a position in which one
would coherently assert “I believe that my husband is unfaithful to me,
but he is not”; where the first conjunct expresses a disposition one has
found out by observing one’s own behaviour and by inferring to its likely
cause, and the second conjunct expresses one’s belief as a commitment,
given one’s knowledge of one’s spouse’s loyal behaviour. By contrast, it
would seem that if, by uttering (or judging) that very sentence, one were
trying, through its first conjunct, to express a commitment, its second
conjunct would actually undo it. This, in fact, would generate a Moorean
paradox. The interesting and novel result is that the existence of Moore’s
paradox can be secured only by countenancing essentially normative
mental states such as commitments.
Hence, to conclude: what goes by the name of “self-knowledge” is a
blend of disparate factors. Sometimes psychological self-ascriptions actu-
ally constitute the corresponding first-order mental states and although
one cannot fail to “know” them, it is not because one entertains a par-
ticular epistemic relation to one’s first-order mental states. Rather, it is
because the self-ascription brings them about and therefore is necessar-
ily authoritative. Some other time, our psychological self-ascriptions are
alternative ways of giving expression to mental states, which can exist
independently of them, resulting from being drilled to substitute their
immediate expression with the relevant linguistic behaviour. Still, under
appropriately specified C-conditions, being in a position immediately to
self-ascribe them and being correct in one’s self-ascription are guaran-
teed to hold a priori and as a matter of conceptual necessity. Finally, in
many cases, self-knowledge is actually the result of the application to
one’s own case of a little psychological theory or of simulative strategies
or, indeed, of an inferential deployment of highly dispositional psycho-
logical concepts, or may depend on testimonial evidence. Only in these
latter cases would self-knowledge be the result of some kind of cognitive
achievement and the term “knowledge” would, accordingly, express an
epistemic relation between a subject and her own mental states. In all
other cases, by contrast, the term “knowledge” would signal rather the
18 A. Coliva

fact that there is no room for error, when self-verifying self-ascriptions are
at stake, or at least not in the normal run of cases, when we are dealing
with self-ascriptions of sensations, basic emotions, perceptions and per-
ceptual experiences. Either way, self-knowledge is valuable either because
of its constitutive links with (“thick”) rationality, concepts’ possession,
and, at least in some cases, responsible agency, or because it can help us
have a better, more integrated and unitary life. Small surprise, then, that
Western philosophy since its inception appropriated the dictum of the
oracle of Delphi, “Know thyself ”.11

Bibliography
Boyle, M. (2009). Two kinds of self-knowledge. Philosophy and Phenomenological
Research, 77(1), 133–164.
Boyle, M. (2011a). Transparent self-knowledge. Aristotelian Society Supplementary
Volume, 85(1), 223–241.
Cassam, Q. (2014). Self-knowledge for humans. Oxford: Oxford University Press.
Coliva, A. (2015a). Review of Quassim Cassam Self-knowledge for humans,
Analysis, doi: 10.1093/analys/anv078.
Gopnik, A. (1983). How we know our minds: The illusion of first-person
knowledge of intentionality. Behavioral and Brain Sciences, 16, 1–15.
Reprinted in Goldman, A. (ed) (1993). Readings in philosophy and cognitive
science. Cambridge MA: MIT Press.
Heal, J. (2002). First person authority, Proceedings of the Aristotelian Society,
102(1), 1–19. Reprinted in (2003). Mind, Reason and Imagination (pp. 273–
288). Cambridge: Cambridge University Press.Please provide page range for
Heal (2002). 1–19
McDowell, J. (1994). Mind and world. Cambridge MA: Harvard University
Press.

11
According to the style of Palgrave Innovations in Philosophy series, while retaining the ambition of
presenting a novel account, the volume also contains a detailed discussion of some prominent
contemporary theories of self-knowledge. However, it does not address the issue of the first person,
even though having a concept of oneself is necessary in order to make psychological self-ascriptions.
The topic is extremely complex and will deserve a volume in its own right. I hope to be able to write
it before too long.
 2
Varieties of Mental States

In this chapter, we explore and propose a systematisation of the com-

plex geography of the mental. We first distinguish between sensations
and perceptions (§1). We then move on to propositional attitudes and
distinguish between beliefs, desires and intentions as “dispositions” and
as “commitments” (§2). Finally (§3), we consider the complex case of
emotions, whose nature still escapes philosophical consensus. After pre-
senting and criticising several contemporary accounts, we put forward a
borderline view of emotions.
This overview shows that we enjoy a variety of mental states, whose
intrinsic features are extremely different. This paves the way to the claim,
at the heart of this book, that single, all-encompassing accounts of how
we can know our minds are unlikely to be successful, for what that knowl-
edge is about is a heterogeneous mix. Just as not many will expect a uni-
form account of how we can gain knowledge of such diverse objects as
truths about physical objects, moral truths and mathematical ones, say,
so it is unrealistic to think that mental states, whose intrinsic features are
very dissimilar, should be known in the same way, simply because they are
all mental states. It would be like thinking that a uniform account is owed
of how we know that there is a hand where we seem to see it or that there

© The Editor(s) (if applicable) and The Author(s) 2016 19

A. Coliva, The Varieties of Self-Knowledge,
DOI 10.1057/978-1-137-32613-3_2
20 A. Coliva

is a planet orbiting around the Sun, in between the Earth and Jupiter, or
that there is a black hole in a given portion of the universe. Even if the
objects of such knowledge are truths about physical entities, the ways
in which we do know them are clearly different—ranging from simply
observational to instrumental and robustly inferential. Equally, realising
the variety of mental states we enjoy will familiarise ourselves with the
idea that the ways in which we know them may be substantially different.

1 Perceptions and Sensations
Throughout their lives, human beings usually enjoy sensations such
as pains and tickles, on the one hand, and perceptions like seeing that
there is a book on the desk or hearing that one’s child is crying, on the
other. Is there a substantial difference between sensations and percep-
tions? According to Tyler Burge’s account of perception (2010), which
is the best developed one to date, the crucial difference between them is
that only perceptions are objective representations of the physical world
around a perceiver. That is to say, only perceptions are representations
of external objects, shapes, colours, and so on and as of objects, shapes
and colours out there—that is to say, as not mere variations in proximal
stimuli. It is therefore useful to present his views in some detail.1
According to Burge, in order to answer constitutive questions about
when objectivity starts, two main methodological assumptions should be
made. The first one is a firmly anti-individualist conception of the mind.
The second one consists in deploying an empirically informed methodol-
ogy. Accordingly, Burge rejects reductionism of the mental to the physical
and looks at the best and most mature science concerning, particularly,
perception to see, firstly, what conceptual categories it makes use of and,
secondly, what kind of data it provides us with regarding the issue of when,
at the ontogenetic and phylogenetic levels, objective representation starts.

1
In what follows, I will not dwell on those aspects of Burge’s overall proposal I am not entirely
happy with, such as the notion of a priori knowledge he seems to subscribe to, with the result that
we would know a priori certain natural norms regarding perception, or his account of smell and
taste according to which these senses do not afford perceptions but merely sensations. For a presen-
tation and a criticism of these aspects of his account, see Coliva 2012a.
 2 Varieties of Mental States 21

The upshot is that, according to Burge, representation is an irreducible

and non-eliminable mental kind, characterised by the possession of verid-
icality conditions. That is to say, representations, by their very nature,
have conditions for being correct or incorrect. Therefore, they are neither
reducible to other mental states which lack veridicality conditions, such
as sensory states, nor eliminable in favour of non-psychological states,
such as merely causally reliable or information theoretic processing, or
even teleological co-variation. Hence, representation is a psychologically
robust kind, which, moreover, is routinely used in scientific explanations.
Such a psychological kind is instantiated in perception, thought and lan-
guage. However, according to Burge, it is in perception that it makes its
first appearance, as a close examination of perceptual psychology reveals.
Furthermore, perceptual psychology makes clear that perception is com-
mon to human adults, human infants and a large number of creatures in
the animal kingdom. The reason why perception is so widespread is that
it does not require a representation by the individual of the conditions
which make it possible for it to have perceptual representations. The lat-
ter constitutively involve patterns of causal relations with the environ-
ment around a perceiver, which help determine the specific natures of
these perceptual states. Such causal relations, however, are not themselves
representational.
An important aspect of Burge’s anti-individualism consists in hold-
ing that any representational state requires bearing “certain ‘associational’
relations to some veridical representational states” (2010, p.  68). Yet,
according to Burge, representations are not object-dependent or, more
generally, referent-dependent. Hence, contrary to what disjunctivists
about perception hold, there can be genuinely perceptual states that are
not veridical. Still, they are representations all the same and, moreover, are
constitutively dependent on the fact that subjects are (or have been, even
philogenetically) in causal connection with some aspects of the physi-
cal environment around them. Furthermore, anti-individualism helps
explain why representational mental states have veridicality conditions
that are non-accidentally fulfilled by elements in the environment and,
connectedly, why they can sometimes be erroneous. What is more, Burge
claims that “the science of perceptual psychology presupposes anti-
individualism about perception” and hence it “makes anti-individualism
22 A. Coliva

about perception empirically specific” (2010, p. 87). For perceptual psy-

chology contributes law-like generalisations that explain “the processes by
which perceptual states with specific veridicality conditions are formed
from specific types of proximal stimulations” (2010, p.  88) as well as
cases of perceptual illusion. The difference between veridical and illusory
perception generally depends on differences in the actual, occurrent dis-
tal antecedents of a given type of proximal stimulation. Hence, causal
relations between perceptual states and their representata are presupposed
by scientific explanation. Moreover, the central problem of perceptual
psychology—the so-called “underdetermination problem”, according to
which the same proximal stimulations are compatible with several differ-
ent physical causes—is solved when the principles that govern the forma-
tion by perceptual systems of veridical perceptual states are discovered.
Veridical perceptual states, in turn, are individuated by their relations
to environmental entities. Hence, the solution of the central problem
of perceptual psychology presupposes anti-individualism. Furthermore,
according to Burge, the laws that govern perceptual systems are never
attributable as acts to the perceiver, not even implicitly. They are com-
putational formation principles, “inaccessible to consciousness and not
under the perceiver’s control” (2010, p. 94). They operate at the subper-
sonal level, although their results are constitutively attributable to the
whole perceiver, despite not being necessarily conscious. Hence, once
more, perceptual psychology is anti-individualist, insofar as it does not
require a subject to be able to represent the conditions which make per-
ception possible. Finally, perceptual systems are domain-specific, (par-
tially) encapsulated from other cognitive systems, although they can
interact with other systems, and are shared across a wide number of spe-
cies. All these aspects further support the view that perceptual psychology
is deeply committed to anti-individualism.

1.1 The Objectivity of Perceptual Representation

The crucial issue addressed by Burge is what it means to say that per-
ception affords an objective representation. “Objective” as used here
connotes being a product of objectification, which “is formation of a
state with a representational content that is as of a subject matter beyond
 2 Varieties of Mental States 23

idiosyncratic, proximal, or subjective features of the individual” (2010,

p. 397), comprising entities in one’s physical environment and also one’s
own body. According to Burge, in order to perform objectification, the
system must discriminate one shape from the other, but also shapes from
other relevant elements, which are environmentally salient and could
have an impact on the needs and activities of the perceiver—similarly,
for the perception of bodies, which must be discriminated from events,
properties, and so on. Perception is still objective even if the perceptual
system is incapable of discriminating these elements from illusions, prox-
imal stimulations, abstract kinds, undetached entity parts, and so on. For
the latter do not figure as relevant alternatives in a causal account of the
formation of the perceptual states or figure in natural biological explana-
tions of functional individual needs and activities. Hence, “the perceiver’s
objectifying discriminatory abilities determine the nature and content of
his perceptual abilities only within this larger environmental and etho-
logical framework” (2010, pp. 407, 466).
Another fundamental facet of objectification, according to Burge, con-
sists in the exercise of perceptual constancies, which allow us, for instance,
to perceive a colour as the same even if it is presented in different ways,
like when a white wall is perceived as having the same colour although it
is unevenly illuminated. Again, perceptual constancies are at work when
we perceive a given object as the same while we move further away (or
nearer) to it, thus undergoing different proximal stimuli. According to
Burge, perceptual constancies are necessary and sufficient for the system’s
being a perceptual system (2010, p. 413).

1.2 Perceptual Contents

Perceptions have representational contents, according to Burge. The latter

are abstract kinds that fix conditions under which a psychological state is
veridical. All perceptual representational contents are structured—that is,
they have singular and general elements. The latter perceptually indicate
certain types or attributes—roundness, to the right of, and so on—and
attribute them to particulars. Burge calls them “perceptual attributives”
(2010, p. 380). Perception, however, singles out also particulars: not only
24 A. Coliva

bodies or events but also specific, contextually determined instances of

properties and relations. These singular elements are labelled “singular
perceptual applications”. Both perceptual attributives and singular per-
ceptual applications are semantically relevant: the former can rightly or
wrongly indicate types or attributes or rightly or wrongly attribute them
to particulars; the latter, in contrast, could fail to refer.
A close examination of perceptual psychology supports the view that
the elements of perceptual contents are not objects and properties but
perceptual modes of presentation of them. Specific objects and proper-
ties are relevant only in order to determine whether a given perceptual
representational state is veridical. Moreover, according to Burge, there is
a structural difference between perceptual and propositional content. The
former necessarily involves singular, context-determined elements, which
are categorised or grouped from a contextually bound perspective. What
is not yet present, however, is the separation of attributions from singular
reference, to arrive at propositional predication. “A capacity for such a
separation is a central aspect of achieving the specific context indepen-
dence and generality that are embodied in pure attribution, propositional
thought and rational inference” (2010, p. 541). Moreover, the content of
perception is similar to a map or a sketch from an egocentric perspective.2
This is not the form of a proposition. In addition, while the transforma-
tions of perceptual states do not depend on the individual, the transfor-
mations among propositions—for instance, in inference—are normally
acts by the individual. Furthermore, there is no perception of logical con-
stants, while real propositional contents involve logical operations.
Finally, according to Burge, perceptual attributives are limited; they
concern shape, spatial relations, colour, motion, texture, possibly danger,
food, conspecifics, and so on. Burge calls these attributives “perceptu-
ally basic” (546). Perceptual beliefs containing only conceptualisations of
perceptually basic attributives are called “basic perceptual beliefs” (ibid.).
Many of our perceptual beliefs employ concepts, which go beyond the
range of perceptually basic ones, like the concepts of baseball bat, CD
player, and so on. However, according to Burge, “in any particular appli-

2
In this respect, Burge’s account of the content of perception is similar to Christopher Peacocke’s
scenario content. Cf. Peacocke 1992, Chap. 3.
 2 Varieties of Mental States 25

cation (…) the broader type of perceptual beliefs ultimately relies on con-
ceptualisations of basic perceptual attributives” (ibid.). In propositional
thought about perceived particulars, the singular elements are inherited
from perception and embedded in an inferential structure, which may
involve also quantificational elements, although it need not do so.3

1.3 Sensory States and Sensations

Sensory states are different from perceptual ones in that they are not
objective. That is to say, they do not represent external elements of real-
ity as such. Take smell, for instance. In many cases, it seems to afford a
manifold of stimuli without attributing them to some relevantly stable
distal cause. For such a reason, sensory states do not have veridicality
conditions. They are not correct or incorrect, true or false representations
of something out there. They are merely subjective variations in proximal
stimuli.
Sensory states, however, are not necessarily sensations. That is to say,
there need not be a subjectively conscious phenomenal aspect to them.
A creature may be hard-wired such that her sensory systems may register
variations in temperature or pressure on her skin, so as to give rise to cer-
tain bodily movements, say, without her being conscious or experiencing
a sensation of cold or heat or of increased or decreased bodily pressure.
Similarly, perceptions, according to Burge, need not involve a conscious,
phenomenal element, even if they are objective representations with
veridicality conditions. Therefore, blind sight, for instance, would be a
case of perception. Of course, this is not to say that this conscious, phe-
nomenal aspect is not present in many, or even most, human perceptions.

3
Burge compellingly criticises Elisabeth Spelke’s claim that bodies are not represented as such in
perception. Moreover, he convincingly argues that cohesion, solidity, boundedness and spatio-
temporal continuity are properties, which can be represented as such in perception. According to
Burge, the ability to discriminate three-dimensional figures from a background and to represent
them as cohesive and bounded, together with the ability to track objects perceptually over time
(although not necessarily in motion or behind occlusions), is “constitutively necessary to visually
representing bodies as such” (2010, pp.  456, 458–459). By contrast, he thinks that perceptual
attributions of solidity are not necessary to that end, even if they are sufficient for it. Notice, more-
over, that, according to Burge, the ability to perceive bodies as such is not necessary for objective
perceptual representation, although it is central to the development of our conceptual system.
26 A. Coliva

It means merely that it need not be present for a creature to be able to

perceive aspects of her environment.
Furthermore, it should be noticed that even if bodily sensations are—
by definition—felt in one’s body, they are not representations of it and
its physical properties. I can feel pain in my knee, but this does not pro-
vide me with a representation of the knee—that is, of its location, exten-
sion and shape. Nor do they have veridicality conditions. Hence, I may
be mistaken about where my pain is located and even feel it in a limb
I no longer have. The sensation, however, remains, notwithstanding its
erroneous localisation. Of course, there may be (quite unusual) cases in
which it seems to one as if one is in pain when one is not or in which one
takes a tickle for a pain. But that does not hinder the fact that if there is a
given sensation, then one would be enjoying it, even if one were mistaken
regarding its localisation or its conceptualisation.
One aspect Burge does not touch upon, but which is worth mention-
ing in this connection, is the presence of qualia—that is to say, the spe-
cific phenomenal aspect of a given sensation. It seems quite intuitive that
a painful sensation feels different from a sensation of cold or heat or that
the smell of coffee feels different from the one of chocolate or vanilla. It
seems compatible with everything we have been saying so far that, for
creatures who can enjoy sensations, they may also feel different to them,
without requiring any conceptualisation on their part, while their iden-
tification as sensations of pain (as opposed to sensations of cold or heat)
or as smells of coffee (as opposed to smells of chocolate or vanilla) would
depend on the exercise of concepts.

2 Two Kinds of Propositional Attitudes:

Dispositions and Commitments
Human beings not only enjoy sensations and perceptions but are capable
of thoughts. They believe that the sun will rise tomorrow, they desire to
live a pleasant life, they hope their children will flourish, and so on. They
are therefore capable of having various attitudes with respect to different
propositional contents they can entertain, thanks to the possession of the
relevant concepts. That is to say, one cannot believe, desire or fear that
 2 Varieties of Mental States 27

tomorrow it will rain, unless one has the concepts necessary to entertain
that proposition.4
Subjects also orchestrate thoughts in inferences. Hence, if one believes
that all humans are mortal and that Socrates is human, one will believe
that he is mortal. If one believes that tomorrow it will rain and one
wishes to avoid getting wet, (ceteris paribus) one will go out only with an
umbrella. Practical inferences are characteristically sensitive not only to
the contents of one’s thoughts but also to the attitude one has towards
them. For if, instead of wishing not to get wet, one desired it, one’s actions
ought rationally to be different.
Recourse to subjects’ attitudes is routinely made in order to explain
and predict their behaviour. For instance, I can explain why Anna voted
for a given party by reference to her convictions and desires for herself
and the society she lives in. Alternatively, knowing her convictions and
desires, I can predict how she will vote at the next general elections.

2.1 Propositional Attitudes as Dispositions

However, on reflection and contrary to what mainstream philosophy of

mind seems to hold, our notion of an intentional mental state is not
univocal. On the one hand, there are intentional mental states that
we might call “mental states as dispositions” or “non-judgement-sensitive
mental states”.5 On the other, there are what may be called “intentional
mental states as commitments” or “judgement-sensitive mental states”
(we will dwell on them in the next section). Admittedly, the class of dis-

4
It is a further issue, which will not concern us here, whether those concepts are the constituents of
the proposition itself, or whether the latter consists merely of objects, properties and relations and
whether concepts are needed merely to entertain it in thought, or finally whether it is a set of pos-
sible worlds—that is, all and only those worlds in which it will rain tomorrow in a given place—or
are indeed unstructured abstract entities, which can be entertained just in case one has the relevant
concepts.
Similarly, it is a further issue, which we will not address in this volume, which conditions must be
satisfied in order for a subject to possess the relevant concepts. For an overview and critical assess-
ment of contemporary theories of concepts, see Coliva 2006.
5
Bilgrami (2006, Chap. 5) distinguishes between mental states as “dispositions” and as “commit-
ments”. Scanlon 1998 and Moran 2001 between “brute” or “non-judgement sensitive” and “judge-
ment-sensitive” mental states.
28 A. Coliva

positional propositional attitudes is very heterogeneous, and its width

may be hard to determine exactly. We will presently offer a characteri-
sation of it. Yet it is important to keep in mind that, in this volume,
the expression “dispositional propositional attitudes” is not contrasted,
as usual, with the expression “occurrent propositional attitudes” (unless
otherwise indicated). That is, the contrast we would like to capture is not
between the ongoing belief that my mother is 76 years old and the very
same content believed, dispositionally, even when I am asleep or I am
engaging in activities, which involve thinking about altogether different
contents. For reasons we will see in the next section, the occurrent belief
that my mother is 76 years old would be a commitment and it would
normally give rise to appropriate first-order dispositions, such that, if,
while engaged in a different activity, I were asked how old my mother is,
I would answer “76”. Nor is it necessarily a contrast between unconscious
and conscious propositional attitudes. Some of our unconscious disposi-
tional propositional attitudes, as we shall see, can become conscious, yet
they would still differ from commitments. So, here is a series of features
that characterise dispositional propositional attitudes:

(a) these mental states are not the result of a conscious deliberation, like
a judgement, on a subject’s part, based on considering and, in par-
ticular, on assessing (or even being able to assess) evidence in favour of
P (or of P is worth pursuing, it would be good if P happened, etc.);
(b) these mental states are not within one’s direct control, being rather
something one finds oneself saddled with;
(c) hence, these mental states are not something one will be held ratio-
nally responsible for.

Some examples of mental states that will satisfy these conditions are
(i) mental states that are not formed by having assessed evidence in favour
of P (in the case of beliefs) or of P would be good to have (in the case of
desires, intentions and hopes). Brute urges and needs—insofar as some
of them may (debatably) be considered propositional attitudes—would
fall into this category, but also those propositional attitudes that might
be ascribed to a-conceptual creatures to make sense of their intelligent
behaviour and that, while responsive to some evidence, are not dependent
 2 Varieties of Mental States 29

on its appraisal. For, ex hypothesi, these creatures do not have the concepts
necessary to grasp those contents, let alone the ones needed to assess the
evidence in their favour.6
(ii) Mental states that are attributed to subjects to make sense of their
behaviour, of which they themselves may be entirely ignorant. The lat-
ter class of mental states may comprise, but is not exhausted by, uncon-
scious mental states of a Freudian kind, which, however, can be operative
in shaping a subject’s behaviour.7 The idea here is that a subject may
form these beliefs and desires as dispositions, as reactions to experiences
undergone in early infancy and be entirely oblivious to them, while they
do shape much of her overt behaviour. This is not to say that she may
not acquire knowledge of them through therapy, say. Clearly, however,
that would be a case of third-personal self-knowledge. Deep-seated biases
would be another case in point, like gender preferences in offering certain
kinds of job preferably to male (or female) candidates. Again, one may
act on the basis of such a bias and could eventually recognise it, but if
one did recognise it, the first-order mental state, if still in place, would
remain dispositional, at least in the normal run of cases. Indeed, one may
continue to act on its basis while sincerely judging the opposite.8
(iii) Also several propositional attitudes that are neither biases nor
Freudian mental states but are self-attributed on the basis of an act of self-
interpretation, by finding them out through the observation of one’s own
behaviour and other immediately self-known mental states, will fall into
this category. For self-interpretation, when successful, makes one aware
of a mental state that is already there yet is not “one’s making” but rather
something one finds oneself saddled with. A nice example, though not
a case of propositional attitude, is provided by Jane Austen in her novel
Emma, when the protagonist finds out about her love for Mr. Knightley,

6
This makes it disputable that they could have the relevant beliefs as well, if those depended on
having the concepts necessary to grasp the propositions which constitute their contents. If one were
in the grip of such preoccupations, then a-conceptual creatures could at least be granted with
proto-beliefs, desires and intentions. See Dummett (1996, Chap. 12).
7
There may also be mental states which are attributed from a third party to make sense of a subject’s
behaviour, which are unconscious yet are not of a Freudian nature. The example discussed in (iii)
would be a case in point if, instead of being self-ascribed, the mental state were ascribed by another
person.
8
We will discuss this possibility in the context of our treatment of self-deception.
30 A. Coliva

her long-lasting friend, by reflection and inference on her own imme-

diately available feelings of jealousy at the prospect that Mr. Knightley
could return another woman’s feelings.9 Since this example will be dis-
cussed again in the following, it is worth quoting it in full:

Emma’s eyes were instantly withdrawn; and she sat silently meditating in a
fixed attitude, for a few minutes. A few minutes were sufficient for making
her acquainted with her own heart. A mind like hers, once opening to
suspicion, made rapid progress. She touched—she admitted—she acknowl-
edged the whole truth. Why was it so much the worse that Harriet should
be in love with Mr. Knightley than with Mr. Churchill? Why was the evil
so dreadfully increased by Harriet’s having some hope of return? It darted
through her, with the speed of an arrow, that Mr. Knightley must marry
no-one but herself.

As the converse of self-interpretation, (iv) there may be mental states,

which one can predict—either through inference or simulation—will
assail one, in given circumstances, which, however, will not be within
one’s direct control. Perhaps, owing to one’s long-lasting self-observations
or through an act of simulation, one will know that if one were to work
in an unsupportive environment for a while, one would start losing one’s
self-confidence and believing that one’s work is meaningless or of poor
quality. The characteristic feature of these mental states—in this case, the
belief that unless one’s work gains some kind of external recognition it is
not worthy—is that one would seem to find oneself saddled with them,
even if one were rationally able to find reasons that should make one
think differently. This does not contrast with (c), that is the fact that one
will not be held rationally responsible for these mental states. For indeed
one would not be held rationally responsible for having them but only for
not trying to get rid of them, when rationally unmotivated, once becom-
ing aware of them through self-interpretation.

9
The example is presented and discussed in Wright 1998, pp. 15–16, borrowed from Tanney 1996.
Analogous examples could easily be construed for the case of propositional attitudes. Giorgio Volpe
has kindly pointed out to me that also Schopenhauer, in On Freedom of the Human Will, holds the
view that a person’s character traits are known to her through reflection and inference on her past
behaviour.
 2 Varieties of Mental States 31

Another example could be the one of (v) propositional attitudes

formed on the basis of habit. My actions show that I confidently believe
that there is a floor behind the door of my bedroom. Or one’s surprised
reaction at the sight of a black sheep in a field, after seeing only white
sheep throughout one’s life up to that point, can show that one has so far
believed as a disposition, and as a result of habit, that all sheep are white.
Indeed, one might even know otherwise, but one’s surprise would betray
one’s dispositional belief.
There may be more instances of propositional attitudes as dispositions,
but the important point is that these would all be propositional attitudes
manifested in the relevant first-order dispositions to act, while meeting con-
ditions (a–c). Let us now turn to a different kind of propositional attitudes.

2.2 Propositional Attitudes as Commitments

Manifestly, adult human beings also have propositional attitudes that

depend on a judgement based on the assessment of the evidence at sub-
jects’ disposal and that, for this reason, are within their control and for
which they are held rationally responsible. Call them “intentional mental
states as commitments” or “judgement-sensitive mental states”.10 Although
the word “commitment” may have become common currency in philo-
sophical literature nowadays,11 there is still little agreement among its
users about its meaning. For our purposes here, what is essential to com-
mitments and makes them, in effect, very close to “judgement-sensitive”
beliefs, desires, intentions, wishes, hopes and so on is the following:

(a’) that they are the result of an action—the mental action of judging
that P is the case (or worth pursuing/having)—on the subject’s part,

10
Cf. Bilgrami 2006, p. 213; Scanlon 1998, Chap. 1; Moran 2001, p. 116.
11
Bilgrami makes extensive use of the term; Robert Brandom too, although he is more interested
in stressing the social dimension of commitments than the former (or indeed myself ). Furthermore,
it is not my contention, somehow built in to the very notion of a commitment, that one should
have knowledge of all the logical consequences of one’s own beliefs and further propositional atti-
tudes. As Bilgrami points out (2006, pp. 371–372, fn. 7, but see also pp. 376–377, fn. 20), the
origin of the use of this term to refer to intentional states (or at least to a class of them) goes back
to Isaac Levi.
32 A. Coliva

on the basis of considering and hence of assessing evidence for P (is

worth pursuing/having)12;
(b’) that these mental states are (at least13) normatively constrained—that
is, they must respond to the principles governing theoretical and
practical reasoning;
(c’) and, in particular, they are so constrained (also) from the subject’s own
point of view;
(d’) that they are mental states for which the subject is held rationally
responsible.14

Hence, propositional attitudes as commitments depend on a subject’s

deliberation with respect to P, in the case of belief, of “P ought to be pur-
sued” and of “It would be good (for me) if P were the case” in the case of
desires and intentions, hopes and wishes, and so on, based on consider-
ing and evaluating evidence for P or for “P ought to be pursued”, and
so on. Judging that P is the case is constitutive of believing that P as a
commitment. One cannot have or initiate the latter without the former.

12
This is the main difference between the present account of commitments and Bilgrami’s. For, in
his view, commitments are not dependent on a subject’s judgement.
13
One may even hold that they are intrinsically normative and not merely—as it were, externally—
constrained by normative principles. This is indeed the view that I favour and that will be put to
use in the diagnosis of Moore’s paradox (see Appendix). There is no need to take a stance on it at
this stage, though, for the less committal view would still do, in order to mark out propositional
attitudes as commitments from propositional attitudes as dispositions.
14
This is the constraint Bilgrami identifies as essential to commitments, from which, on his view,
(b’) and (c’) follow. However, he gives a moral or evaluative twist that it is best to resist. For, in his
view, would one be held not only rationally responsible for one’s commitments but also accountable
at large. For instance, one might be reproached or resented for having certain commitments (Cf.
Bilgrami 2006, p. 226). However, specified in the way Bilgrami characterises it, (d’) is not sufficient
to mark out the contrast between commitments and dispositions, because one can criticise or be
criticised, and accept to be criticised, also (for) one’s own dispositions, such as the disposition to
smoke, or, to take a more loaded example, for wanting to get rid of other male opponents as a result
of an unresolved Oedipus complex. But, surely, neither mental state is the result of a subject’s
action, for which one can be held rationally responsible, although one may be considered “badly”—
in Bilgrami’s extended sense of the term—for having it. It is then not by chance that, as a matter of
fact, Bilgrami ends up endorsing the view that “we do have transparent self-knowledge of mental
dispositions” (Bilgrami 2006, p. 287). I find this conclusion unpalatable, for, surely, when we do
get knowledge of our unconscious mental states we obtain it through a process of self-interpreta-
tion or of analysis (that may or may not be guided by a therapist) relevantly similar to the ways in
which we may come to attribute mental states to others. So, it seems to me that whatever knowl-
edge we may eventually gain of our unconscious mental states, it is not “transparent” and is actually
grounded in observation and inference.
 2 Varieties of Mental States 33

Yet beliefs as commitments involve also dispositional elements, which are

not involved in the mere act of judging that P is the case. For example,
one ought to be disposed to use P as a premise in practical and theoretical
reasoning and accept criticism or be self-critical if it were shown that P is
not supported by sufficient evidence and, similarly, in the other cases of
propositional attitudes as commitments and their relations to judgement.
Moreover, the evaluation of evidence may not always be explicit. For
we sometimes form beliefs (and other mental states) as commitments
immediately on the basis of this or that available evidence. However,
one ought to be disposed to produce such evidence, were one requested
to, and ought to withdraw from one’s belief as a commitment (or other
propositional attitudes of that sort), were it shown that the original evi-
dence is either bad or insufficient evidence in its favour. Furthermore, to
say that propositional attitudes as commitments are the result of a sub-
ject’s deliberation based on considering evidence should not involve us in
any form of doxastic voluntarism, as we will soon see. Moreover, it should
readily be acknowledged that there will be forms of local holism between
mental states as commitments. Indeed, viewing also desires, intentions
and other propositional attitudes, beside beliefs, as commitments, tightly
connects them with believing that their contents are worth pursuing or
would be good for one if actualised.
In addition, these mental states are (at least) normatively constrained
in the sense that they are subject to norms governing practical and theo-
retical reasoning. In the case of belief, as already remarked upon in pass-
ing, should countervailing evidence come in, a subject ought to withdraw
from holding P, “P ought to be pursued” and “It would be good (for me)
if P were the case” and so on. Thus, one ought to withdraw from one’s
belief that P is the case or from one’s desire/intention/hope/wish that P
should obtain. In the practical case, it is clear that practical syllogisms
are sensitive not only to the propositional contents entertained but also
to the way in which these contents are entertained. Hence, for instance,
if a subject believes that it is raining, desires not to get wet and happens
to carry an umbrella, she ought, ceteris paribus, to open it, whereas she
ought not to have that behaviour did she merely wish it was raining.
Finally, since such “oughts” would have to be appreciated by the subject
herself, were she not to withdraw from her beliefs and further propositional
34 A. Coliva

attitudes as commitments in light of counter-evidence, not only would she

incur rational criticism but also she should accept to incur it, since she would
not have—the phrase comes in handy—“lived up to her commitments”.
Now, some clarifications are in order. First, to say that beliefs (as well
as other propositional attitudes) as commitments are the result of a delib-
eration and fall within one’s responsibility does not “involve us in any
sort of voluntarism about [their] formation […], any more than we need
to see ordinary argument with others as aiming at getting one’s interlocu-
tor somehow to adopt a new belief by sheer act of arbitrary will”.15 That
is to say, forming certain mental states by considering relevant pieces of
evidence is a rational action yet not an act of arbitrary will. Indeed, being
aware of evidence in favour of P, and being unaware of countervailing
evidence, ought to give rise to one’s judgement that P is the case and thus
to the corresponding belief (similar considerations would hold for other
propositional attitudes and the kind of evidence with respect to which
they are rational responses). Hence, it would be a sign of irrationality not
to form that belief, given sufficient (known) evidence for P.
I said that it would be a sign of irrationality not to form certain mental
states as commitments, given certain pieces of evidence, and this might
invite the objection that, at least in the case of intentions, one could fail
to form them, upon having suitable evidence in their favour, because of
mere weakness of will. My view is that weakness of will could only prevent
one from acting on the basis of a given intention but would not impinge
on its formation. One might then suggest that despite having evidence
in favour of “It would be good (for me) if P were the case”, a (lucid and
attentive) subject could fail to form the corresponding intention because
of other considerations. Hence, rationality does not require forming the
relevant intentions upon having at one’s disposal certain pieces of evi-
dence. That is all right, but it merely shows that the subject did not have
sufficient reasons to form a given intention. Hence, it is obvious, and for
this reason this qualification will be omitted in the following, that when
I claim that there are intentions as commitments that are based on evi-
dence, I will in fact be talking of those intentions we form on the basis
of sufficient reasons for them. Again, this shows that, inevitably, there will

15
Moran 2001, p. 120. Cf. Moran 2003 and Shoemaker 2003, Postscript.
 2 Varieties of Mental States 35

be forms of local holism in the formation of one’s mental states, which is

all to be expected. So, if one did not form such a belief, one should incur
and ought to accept to incur criticism. Conversely, forming the belief
that P when no evidence in favour of P is at one’s disposal, let alone con-
trary, undefeated evidence is available to one, would not be the rational
thing for one to do. If one nonetheless did it, then one should accept and
ought to agree to accept criticism.
Secondly, it is important to emphasise that a subject may not always
arrive at her beliefs and further propositional attitudes because of con-
scious consideration of the evidence for P (or else, for P ought to be pur-
sued, or for “it would be good (for me) if P were the case”). All the
present account of commitments requires is that she must be able to offer
her evidence in support of her beliefs and judgement-sensitive desires and
other propositional attitudes if asked to give her grounds for them—that
is, the reasons why she holds them, not the reasons why she believes she
does.16 This is part of what distinguishes commitments from dispositions,
which, though perhaps based on evidence, are not based on the assess-
ment—not even the potential assessment—of such evidence.
Moreover, there may be cases in which one has forgotten the evidence
on which one first based one’s judgement that P. Again, nothing in the
present proposal entails that a subject should remember these details.
Rather, she still counts as having a belief as a commitment that P, say, if
her belief in P originated in an act of judgement that P, based on what-
ever evidence was available to her at that time and the belief has been
sustained throughout that time.
Furthermore, the kind of evidence at a subject’s disposal could also
be testimonial in nature. Thus, for instance, I believe as a commitment
that Napoleon was defeated at Waterloo clearly not because I witnessed
the event nor because I am a knowledgeable historian who is aware of
primary historical sources and has checked them thoroughly. Rather, I
have a belief as a commitment to that effect because I formed it on the
basis of testimony and by trusting the teachers and textbooks that passed
that information on to me. In principle, I could cite that testimonial
evidence, if asked to provide my reasons for my belief, and I should be

16
Cf. Moran 2001, p. 116.
36 A. Coliva

prepared to withdraw from it if it were shown to me that the books and

teachers from whom I got it were systematically unreliable.
To forestall possible misunderstandings, it is important to stress that to
say that these propositional attitudes as commitments are rationally held
simply means to say that they are held for reasons; it does not entail that
they are held for good reasons. Indeed, one’s reasons may be bad or even
absurd ones. So, for instance, I can wrongly think that it is raining given
my present visual representation of a portion of sky outside my window,
when it is not. I can believe as a commitment that the solution to a given
arithmetical problems is “10” when in fact it is not, owing to my poor,
too fast and casual reasoning. In all these cases, I have reasons for my
beliefs, I can produce them if requested, and I can accept criticism up
to the point of realising that they are indeed bad reasons and that I had
better withdraw my beliefs. One can even be afflicted by wishful thinking
and claim the possession of deep and mysterious reasons for one’s belief
or even profess one’s faith and, to the extent that that involves beliefs as a
commitments, do so precisely because what is believed is indeed absurd.17
Again, this is not to say that deep and mysterious reasons or absurd ones
are good reasons; still, these are the reasons some people would have and
would have considered in order to form a judgement and thus a belief as a
commitment in P. As long as they are prepared to offer them, if requested,
and as long as they realise that their reasons may be challenged, even if
they think their reasons may not be defeated, they would certainly have
reasons which are good enough to have beliefs as commitments, even if
they are not—arguably—objectively good reasons.
Thirdly, it may sound surprising that desires could also be seen as
brought about by rational deliberation. However, it is important to keep
in mind that here we are not concerned with what we might call “brute”
desires such as lust or hunger but only with rationally held ones. Yet,
since these “oughts” are what distinguish beliefs and desires as commit-
ments from mere drives and brute dispositions, if I believe/desire that P
as a commitment, then I ought to do so on the basis of evidence and ought
to withhold from it in case contrary evidence came up. Obviously, no

17
“Credo quia absurdum” is usually attributed to Tertullian, even if he did not write it, and is often
associated with fideistic positions.
 2 Varieties of Mental States 37

such “oughts” hold for drives, like the urge to smoke a cigarette after din-
ner, which will persist no matter what amount of counter-evidence will
be considered against the advisability of such a practice, or for brute dis-
positions, such as the disposition to form a certain thought upon hearing
a given word, tune, and so on. Hence, for instance, one may rationally
desire to provide one’s child with the best possible education. If one does
so, it will be for reasons, which, as such, may be further assessed. Were
it to turn out—quite implausibly—that countervailing considerations
should outweigh that desire, one should withhold from it, if rational.
Finally, it has to be registered that the distinction between proposi-
tional attitudes as commitments and as dispositions may be misleading,
insofar as it might suggest the idea that commitments exclude having
behavioural dispositions (at large). However, in the present understand-
ing of them, propositional attitudes as commitments will, in fact, be
accompanied by the relevant set of first-order dispositions, at least in
the normal run of cases. Hence, to exemplify with the case of beliefs,
the term “belief ” does not have “divided reference”, as Moran alleges
against Georges Rey, like the term “jade”, say.18 That is to say, it does not
pick out two different kinds of mental states, namely dispositions, on the
one hand, and commitments, on the other. Rather, there is considerable
overlap between beliefs as dispositions and as commitments to justify the
idea that they are more like two species of the same genus and therefore it
is no mere accident that the term “belief ” is, in fact, applied to both. For
instance, if I believe that P as a commitment, then I ought to be disposed
(ceteris paribus) to use P as a premise in a piece of practical or theoretical
reasoning, whether that reasoning is conscious or not. Hence, I should
have at least some of the same behavioural dispositions I would have if
I merely believed as a disposition that P. Similarly, in the case of a desire
as a commitment that P, I ought to be disposed (ceteris paribus) to seek
means to make P happen, just as I would do with a desire as a disposition.
Yet the aetiology, the modes of knowledge and some of the possible con-
sequences of having commitments and dispositions would be different.
However, only in the case of beliefs and other propositional attitudes
as commitments, if I did not actualise such first-order dispositions, I

18
Moran 2001, p. 87.
38 A. Coliva

then should be prepared to accept criticism for having failed to live up

to my commitments, or be self-critical myself. Hence, it is not enough
merely to declare “P˝, say, to be credited with the belief as a commitment
that P. It is also necessary and sufficient, in order to count as having the
relevant commitment, to display the second-order disposition to accept
criticism for not having actualised the appropriate first-order ones. This
aspect may escape notice because we tend to conflate two senses of “com-
mitment”—one social (or third-personal), as it were, and one personal
(or first-personal). For instance, if I say “I will buy you an ice cream” to
my son, I will be taken by him (and possibly other people in attendance)
to be binding myself to that course of action (ceteris paribus).19 Other
people therefore can ascribe to me the commitment to buying him an ice
cream. However, I may have lied to him, in which case I would have no
personal commitment to doing what I said I would.20 What we have been
concerned with here, however, are only personal commitments, which
require, for their existence, the dispositions we have been reviewing and
are not incurred simply by voicing them. Hence, only propositional atti-
tudes as commitments can be had without behaving as required by them,
as long as one were prepared to be self-critical or to accept criticism for
not doing it.

3 Emotions
Besides having sensations such as heat and cold, perceptions like see-
ing a table in front of one, and propositional attitudes as dispositions
and commitments, human beings are capable of experiencing emotions,
such as fear of a barking dog in front of one, or love for one’s beloved,
or hatred towards a harming enemy, and so on. Although common sense
groups them separately from other mental states and has a specific name
for them, philosophical analyses have usually tried to reduce emotions to

19
This distinction will become crucial in our analysis of Moore’s paradox (see Appendix).
20
I may be reproached for having lied to him, of course, but this would be a moral judgement,
which does not change the fact that I did not have any personal commitment in the first place.
 2 Varieties of Mental States 39

other kinds of mental states, like sensations, judgements, perceptions. We

will briefly review these attempts and show them wanting. The underly-
ing idea, therefore, is that perhaps the time has come to owe common
sense its due and start living with the idea that emotions are not reducible
to other kinds of mental states.

3.1 Emotions as Sensations

David Hume (1739–1740) and William James (1890) are usually consid-
ered to have held the view that emotions are sensations. Hence, the fear
of the barking dog in front of me would be a particular sensation. The
love for my beloved, a characteristic feeling and the hatred for a danger-
ous enemy another kind of feeling.
Various objections have been raised against such an account—for
instance, that it is not always the case that an emotion is accompanied
by a sensation. If love is an emotion, then my love for my children is not
a sensation. Rather, it is a complex mental state, which involves particu-
lar attitudes of care and attention. Similarly, if depression is an emotion,
then it is unclear what kind of sensation it involves. Rather, it seems to
display itself in characteristic attitudes towards people and events or even
life itself.
Moreover, jealousy and envy, for instance, may be akin inasmuch
as one’s feelings are involved, but they clearly differ insofar as the for-
mer concerns the attentions that one’s beloved reciprocates or displays
towards another person, whereas the latter concerns whatever another
person, not necessarily a beloved one, possesses and one would like to
have for oneself.
Sensations, moreover, are not representational, for they lack veridical-
ity conditions; nor are they objective, in the sense of being of objects and
properties represented as such, as we saw in §1.1. Even basic emotions,
such as fear, however, seem to have some form of representational con-
tent. For instance, my fear of the barking dog in front of me seems to
present that dog as dangerous to me and to have, on top of such a repre-
sentational content, a particular phenomenal aspect to it.
40 A. Coliva

3.2 Emotions as Value Judgements

Stoics are said to have held a completely different account of the emo-
tions, one recently upheld and developed by Martha Nussbaum (2001).
Accordingly, emotions are taken to be value judgements. Hence, for
instance, my fear of the barking dog in front of me would, in fact,
amount to the judgement, however implicit that might be, that the dog
is dangerous to me. My love for my beloved would be identical to the
judgement that he is important to me. Finally, my hatred for a dangerous
enemy would consist in the judgement that he is mischievous. Feelings,
when they are present, are added on top of the judgement but are not
constitutive of the emotion.
Surely, the proposal does well in accounting for the representational
elements of emotions. For they are taken to be judgements, hence propo-
sitional mental states, with genuine truth conditions. Yet it strikes as too
intellectualistic. After all, in order to make judgements, one needs con-
cepts, at the very least, if not also the ability to perform inferences and to
gather and assess evidence for or against their contents. Arguably, how-
ever, basic emotions are present in infants and animals. Yet it is dubious
that they would possess the relevant concepts, let alone make characteris-
tic inferences and weigh reasons for or against their contents.
Moreover, it is often the case that experiencing a certain emotion is
the basis for making a certain judgement. For instance, it is the fear I am
feeling of the barking dog in front of me which makes me judge (rightly
or wrongly) that it is dangerous. Furthermore, despite my fear, I can issue
a contrary judgement, if, for instance, I have independent and stronger
reasons to think that, after all, it is not dangerous. Yet, although normally
a better judgement overrides a worse one, I can still fear the dog, even if
I judge that, after all, it is not dangerous.

3.3 Emotions as Felt Bodily Attitudes

Perhaps the most recent account of emotions is the one due to Julien
Deonna and Fabrice Teroni (2012), who develop suggestions that can be
found in Edouard Claparède (1928). Accordingly, my fear of the barking
 2 Varieties of Mental States 41

dog in front of me would be a specific attitude, felt in my body, provoked

by the sight of certain physical features of the dog, such as its big teeth
and aggressive attitude. My love for my beloved would be a characteristic
attitude felt in my body provoked by the sight of his physical appearance
or of his kindness, whereas my hatred for a dangerous enemy would be a
different attitude felt in my body caused by the perception of its aggres-
sive or deceitful behaviour towards me.
This account is less cognitively demanding than the one previously
reviewed and assigns a constitutive role to feelings, that do not figure
merely as possible accompaniments of the judgement or, in this case,
of the attitude. Yet, if the attitude is a propositional one, it will require
concepts to be entertained. If it is merely a bodily reaction, in contrast,
it would have to be clearly specified for each kind of emotion and would
have to be always present. Still, neither the former nor the latter condi-
tion seems to be satisfied in all cases of emotions. As we saw, envy and
jealousy may be felt in one’s body in similar ways, despite being different
emotions. Moreover, there are emotions that do not seem to give rise to
distinctive bodily attitudes. Depression may be a case in point, but even
the love for one’s children.
Moreover, one can see the big teeth of the dog and notice its aggres-
sive attitude and yet feel no fear towards it. It seems that one will assume
certain bodily attitudes only if the dog is somehow experienced as dan-
gerous. This, however, poses the problem of how we can perceive the
dangerousness of that animal. After all, as we shall see in the following,
we do not seem to be endowed with a faculty dedicated to the perception
of evaluative properties.
Furthermore, the attitudes we have towards certain people, creatures
or events are often motivated by the kind of emotions they provoke in
us. For instance, my defensive bodily attitude towards the barking dog
in front of me seems to be caused by the fear it provokes in me. The
bodily attitude seems to be brought about by the emotion and not iden-
tical to it. In fact, having been told that one should not display one’s
fear in front of a dog, I may fake a completely self-confident bodily atti-
tude. Nonetheless, I could be scared of the dog. Hence, there seem to be
good reasons to reject the idea that emotions are identical to felt bodily
attitudes.
42 A. Coliva

3.4 Emotions as Perceptions of Evaluative

Properties

The most popular account of emotions at present is perhaps the one,

propounded by theorists such as Ronald De Sousa, Christine Tappolet,
Christopher Peacocke and Robert Audi, according to which emotions are
perceptions of evaluative properties. Hence, my fear of the barking dog in
front of me would be the perception of its dangerousness, my love for my
beloved the perception of his lovability, and my hatred towards a danger-
ous enemy the perception of his mischievousness.
The theory does well in avoiding the troubles that affect the judgement
theory of emotions. For concepts are not necessary in order to have at
least basic perceptions, as we saw in §1.1. Similarly, one could fear the
barking dog in front of one without having the concept of danger, or love
someone without having the concept of being lovable (no matter how
one unpacks it), or hate someone without having the concept of mischie-
vousness (no matter how one specifies it).
Furthermore, the perceptual account explains the possibility for emo-
tions to justify the corresponding beliefs, just as perceptions can justify
the corresponding judgements. For instance, perceiving the dog as dan-
gerous justifies one’s judgement “That dog is dangerous”, just as perceiv-
ing a table in front of one justifies one’s judgement “Here is a table”.
Several objections have been raised against the perceptual account,
however. For instance, it has been pointed out that we do not have a
dedicated faculty for the perception of evaluative properties. If, then, we
should take the theory to be merely suggesting an analogy between per-
ceptions and emotions, it would be too vague and underspecified to be
illuminating.
Some supporters of the perceptual account of the emotions, such as
Audi, have claimed that it should be equated to seeing-as rather than to
merely seeing. Yet seeing-as, in its turn, is quite a poorly understood phe-
nomenon. Moreover, some cases of seeing-as clearly involve the posses-
sion of concepts. For instance, in the course of a geometrical proof, I can
see a segment now as the side of a triangle, now as a line that may inter-
sect two parallels (Coliva 2012b). None of this would be possible without
having the relevant concepts. Hence, to be told that emotions are cases in
 2 Varieties of Mental States 43

which we see something as dangerous, lovable, or wicked does not carry

much traction. Furthermore, if it turned out to depend on the possession
of the relevant concepts, it would spoil the main advantage of the present
proposal over the judgement account of the emotions.
Other objections often raised against the perceptual account are that
emotions can take place even when the object is not present, but that
perceptions cannot, and can lack an object whereas perceptions need one.
Usually, perceptual theorists appeal to imagination to overcome the for-
mer objection and hold that one would perceive everything as dangerous,
say, if one felt generically scared.
Michael Brady (2013) has recently objected that, contrary to per-
ceptions, emotions are not reasons for the corresponding evaluative
judgements. Rather, by occupying and consuming our attention, they
prompt us to look for reasons for our evaluative judgements.
In response, I think perceptual theorists should notice that the objec-
tion rests on a dubious conception of perceptual justification, accord-
ing to which perceptual experiences as such justify the corresponding
beliefs. A perceptual experience just by itself, however, makes equally
more probable “Here is a table”, say, and “I am a BIV presently hal-
lucinating a table”. The reason why it seems otherwise is that perceptual
experiences amount to justifications only thanks to certain background
assumptions, which rest most of the times implicit, such as “There is an
external world”, “My sense organs usually work fine”, “I am not a BIV
or otherwise massively deceived by an evil demon”, and so on. Perceptual
experiences are therefore necessary but not sufficient for having percep-
tual justifications.21
Given such an account of perceptual justification, the parallel between
perceptions and emotions could be restored. For what happens, when
assailed by an emotion, such as a sudden fear provoked by experiencing
a creeping sound while half-asleep at night, is that subjects would check
whether the background assumptions on which emotional justifications
rest, such as being cognitively lucid and alert, and the failure of alternative
21
I have defended this view at length in Coliva 2015. The rejection of the idea that perceptual
judgements are immediately justified by one’s perceptual experiences can be found in several other
theorists, who, however, provide a different positive account of the structure of perceptual justifica-
tion. Among them, see Wright 1985, 2004, and Davies 1998, 2004.
44 A. Coliva

explanations of the sound just heard, are satisfied. Hence, the emotion
is there, yet knowledge of the complexity of the conditions on which
it would accurately track a real danger prompts the subject to consider
whether they are satisfied. Only when one has reasons to think they are
does the emotion give a subject a reason for the corresponding evaluative
judgement. This, however, is structurally identical to what happens in the
case of perception. In perception, however, the background conditions
that allow us to form the relevant justifications, upon having certain per-
ceptual experiences, are less complex and more easily in place. That is why
there seems to be a difference between the two cases. To see how this need
not be so, however, consider a subject who knowingly lives in fake-barn
country. In that case, she could not take her perceptual experiences at face
value to form the corresponding empirical beliefs. In fake-barn country,
moreover, acting just on the basis of one’s visual experience would be
unjustified and therefore irrational.
A further objection raised against perceptual accounts is that they can-
not make sense of so-called “why questions” with respect to one’s emotions
(Brady 2013; Deonna and Teroni 2012). For instance, we may well ask
“Why do you fear that dog?”, thereby looking for an explanation of the
reasons of one’s fear, when, for instance, it is clear that the dog in question
is totally harmless. Yet, the objection goes, we cannot raise a similar kind
of question in the case of a perception. “Why do you see a dog?” could be
taken only as an inquiry into the physical causes of that perception.
Now, supporters of the objection tend to equate the reasons one would
cite in response to the why question with the correctness conditions of
the emotion. Thus, for instance, given my fear of the dog, the question
“Why do you fear that dog?” could be answered by saying “Because it is
dangerous”. Its being dangerous would be, at once, the reason of my fear
and what would make that emotion correct and therefore rational, on
this view. Alternatively, if the dog were not dangerous, that would ipso
facto prove my emotion not only incorrect but also irrational. According
to the supporters of the objection, perceptions are neither rational nor
irrational, but merely correct or incorrect. Hence, they conclude, emo-
tions cannot be perceptions, because the former can, as such, be rational
or irrational whereas the latter are neither justified nor unjustified or,
more generally, assessable along rational dimensions.
 2 Varieties of Mental States 45

Notice that if the objection moved from a sound premise, namely that
emotions are rational (or irrational), that would mean that they would be
assessed as rational (or irrational) as such, not just via (or because of) the
judgements based on them. If that were the case, emotions would be very
different from sensations and perceptions, which are not assessable as ratio-
nal or irrational, and would be much more similar to judgements, beliefs
and other propositional attitudes, which are assessable along that dimension.
Yet, in response to the objection, supporters of the perceptual model
would be well advised to notice that if reasons are identical to correctness
conditions, then perceptions too would, in some sense, be justified, when
correct, or unjustified, when incorrect, and therefore would be rational
or irrational. Of course, that strikes us as odd, but the onus is on the
opponents’ shoulders. For they would have to show that, while having
correctness conditions, those conditions cannot be reasons of one’s per-
ceptions, while they can be reasons for one’s emotions. This seems to be
a difficult task. Moreover, it seems misguided and brought about by too
hasty an equation between correctness conditions and reasons or justi-
fications we had better avoid. Think of beliefs, which paradigmatically
have truth conditions. Yet they can be justified, on most accounts of
justification, even when occasionally false, and, conversely, be true while
unjustified. Yet, if one does not equate correctness conditions and reasons
or justifications, the “why questions” objection evaporates. For both per-
ceptions and emotions would have correctness conditions and yet the fact
that one may assess the rationality of only the latter would not prove that
emotions are, as such, rational or irrational.
The “why questions” objection is therefore misguided, but it stresses
something important, namely that we do assess emotions along ratio-
nal dimensions. The interesting question—independently of whether the
issue can be used to attack perceptual theories of the emotions—is how
to explain such an attitude. The first option would be to say that we are
speaking loosely. What we should say, rather, is that a given emotion is
either correct or incorrect vis-à-vis the event we are witnessing or imag-
ining. It is correct when it tracks the corresponding evaluative property,
incorrect otherwise. In this case, the analogy with perceptions would be
re-established and this would show, once more, that the “why questions”
objection brings no grist to the anti-perceptual theorists’ mill.
46 A. Coliva

Alternatively, one may suggest that emotions are assessed along a ratio-
nal dimension not as such, but derivatively. That is to say, they are con-
sidered rational when they are in line with our best judgement, irrational
otherwise. For instance, fearing a small, harmless dog is irrational because
it contrasts with our best judgement, whereas fearing a big, aggressive-
looking dog is rational because it is in keeping with our best judgement.
On such a view, emotions would differ from perceptions, for we do not
say that perceptions that are/are not in keeping with our best judgement
are rational/irrational, but, rather, veridical or delusionary. Hence, quite
ironically, it is only by renouncing the idea that emotions are, as such,
rational/irrational that one can appeal to their rational assessment to
draw a contrast between emotions and perceptions. Personally, I favour
this latter interpretation. Nevertheless, it would merely add to the other
ones reviewed in this section, which decisively speak against the percep-
tual account of the emotions.

3.5 The Borderline View of Emotions

The models reviewed so far share a common trait. They all try to reduce
emotions to something else—that is, either to sensations or judgements
or to bodily attitudes or perceptions. Perhaps it is by no chance that they
all fail. The idea I wish to put forward is that given the extant state of the
art, it is tempting to think that emotions should not be reduced to any
other allegedly better-understood mental state. Perhaps we had better start
taking the measure of the fact that our conceptual scheme is indicative of
something right, when it groups various mental states under the category of
emotions, rather than do without it, at least at a reflexive, theoretical level.
In this connection, it is useful to exploit the analogy with border-
line mental illness. Borderline patients are not those who are borderline
between mental illness and sanity. Rather, they manifest symptoms, in
various combinations and proportions, which belong to the characteristic
traits of neurosis, on the one hand, and psychosis on the other, without
having only those that belong to one or the other of these two catego-
ries. Borderline patients, therefore, are mentally disturbed, but they are
neither neurotic nor psychotic. They belong to a third, autonomous cat-
egory while sharing symptoms belonging to the other two.
 2 Varieties of Mental States 47

The idea is that emotions should be understood along similar lines—

that is to say, as different mental states with respect to sensations, judge-
ments, bodily attitudes and perceptions. Yet they exhibit some traits in
common with each of these other mental states. For instance, they often
involve characteristic feelings and sensations; like judgements, they can be
assessed—derivatively, as we saw—along a rational dimension; in some
cases, they do give rise to distinctive bodily attitudes; and, in many cases,
they, like perceptions, have intentional objects and representational con-
tent and seem to be independent of the possession of the relevant concepts.
Interestingly, on the borderline view of the emotions, different emo-
tions may share different features with other kinds of mental states. As we
saw at length before, certain emotions do not seem to have a distinctive
phenomenology or, indeed, a phenomenological dimension to them at
all. In other cases, in contrast, the kinds of sensations experienced are very
specific. Some emotions can be had by unsophisticated subjects, who do
not have a developed conceptual repertoire; some others require quite a
sophisticated one (think, for instance, of the intense emotion, character-
ised by a mixture of pleasure and admiration, produced by appreciating
the power of a subtle philosophical distinction, which immediately sheds
new light on a given domain).
If the borderline view is on the right track, then emotions cannot be
reduced to other kinds of mental states, although it may be important
to study them one by one to clarify their multifarious traits and points
of similarity and divergence with respect to other kinds of mental states.
This way, the various “family resemblances” they share will be brought
to light and it will no longer seem so surprising that mental states as
disparate as love, fear, hatred, jealousy, envy or depression are all grouped
together as emotions. Just as it is not surprising that chess, tennis, playing
with dolls or Russian roulette are all considered games, even if they share
no specific set of necessary and sufficient conditions.

4 Summary
In this chapter, we have seen the variety of mental states we can enjoy.
They range from non-representational ones such as sensations, to repre-
sentational yet non-conceptual ones like perceptions, to representational
48 A. Coliva

and conceptual ones like propositional attitudes, which come as dispo-

sitions, that need not be conscious and that do not depend on a sub-
ject’s deliberation vis-à-vis the available evidence, and as commitments,
which depend on such a deliberation and are normatively constrained.
We have then seen how emotions belong to yet a further category that
shares similarities in various directions with the other kinds of mental
states reviewed. Such a variety of objects of knowledge suggests that also
the way in which each of them can be known may be different. This paves
the way to the main thesis of this book, to which we now turn.

Bibliography
Bilgrami, A. (2006). Self-knowledge and resentment. Cambridge, MA: Harvard
University Press.
Brady, M. (2013). Emotional insight. Oxford: Oxford University Press.
Burge, T. (2010). Origins of objectivity. Oxford: Oxford University Press.
Claparède, E. (1928). Emotions and feelings. In M. L. Reynert (Ed.), Feelings
and emotions: The Wittenberg symposium (pp. 124–139). Worchester: Clark
University Press.
Coliva, A. (2006). I Concetti. Roma: Carocci.
Coliva, A. (2012a). Critical notice of Tyler Burge origins of objectivity.
Disputatio, 4(33), 515–530.
Coliva, A. (2012b). Human diagrammatic reasoning and seeing-as. Synthese,
186(1), 121–148.
Coliva, A. (2015). Extended rationality. A hinge epistemology. London: Palgrave
Macmillan.
Davies, M. (1998). Externalism, architecturalism and epistemic warrant, in
Wright C., Smith B. and Macdonald C. (eds), 321–361.
Davies, M. (2004). Epistemic entitlement, warrant transmission and easy
knowledge. The Aristotelian Society Supplementary Volume, 78, 213–245.
Deonna, J., & Teroni, F. (2012). The emotions. A philosophical introduction.
London/New York: Routledge.
Dummett, M. (1996). Origins of analytic philosophy. Cambridge MA: Harvard
University Press.
Moran, R. (2001). Authority and estrangement. Princeton: Princeton University
Press.
 2 Varieties of Mental States 49

Moran, R. (2003). Responses to O’Brien and Shoemaker. European Journal of

Philosophy, 11(3), 402–419.
Nussbaum, M. (2001). Upheavals of thought. The intelligence of emotions.
Cambridge: Cambridge University Press.
Peacocke, C. (1992). A study of concepts. Cambridge MA: MIT Press.
Scanlon, T. (1998). What we owe to each other. Cambridge MA: Harvard
University Press.
Shoemaker, S. (2003). Moran on self-knowledge. European Journal of Philosophy,
11(3), 391–400.
Tanney, J. (1996). A constructivist picture of self-knowledge. Philosophy, 71,
405–422.
Wright, C. (1985). Facts and certainty. Proceedings of the British Academy, 71,
429–472.
Wright, C. (1998). Self-knowledge. The Wittgensteinian legacy, in Wright C.,
Smith B. and Macdonald C. (eds), 15–45.Wright C., Smith B. and
Macdonald C. (1998) (eds)  Knowing our own minds, Oxford, Clarendon
Press.
Wright C., Smith B. and Macdonald C. (1998) (eds) Knowing our own minds,
Oxford, Clarendon Press.
 3
Varieties of Self-knowledge

In this chapter, we first present the characteristic features of what counts

as truly first-personal self-knowledge—namely, groundlessness, transpar-
ency and authority (§1). The key idea defended in the following is that
they are not contingent but necessary and a priori aspects of what goes
by the name of “(first-personal) self-knowledge”. For massive failures at
self-knowledge would display either the lack of the relevant psychological
concepts or failures at rationality, understood in a “thick” sense, which
is accordingly specified. These characteristic traits of first-personal self-
knowledge are then defended against possible objections stemming from
scepticism about knowledge of the content of our own propositional
attitudes deriving from the endorsement of semantic externalism and
from recent findings in cognitive sciences (§2). It is argued that none of
this shows that we never have essentially first-personal self-knowledge.
Rather, it shows that its scope is limited and does not extend to our deep-
seated and future dispositions, to the dispositional elements of our feel-
ings and emotions, and to the causal relations among our various mental
states, which are known, if and when they are, in a third-personal way.
The various modes of third-personal self-knowledge are then presented
and discussed (§3). Yet all this is compatible with the fact that we have

© The Editor(s) (if applicable) and The Author(s) 2016 51

A. Coliva, The Varieties of Self-Knowledge,
DOI 10.1057/978-1-137-32613-3_3
52 A. Coliva

essentially first-personal knowledge of a wide range of mental states and

at least of their “narrow” content, if and when they have it, such as our
ongoing sensations, perceptions, basic emotions and propositional atti-
tudes as commitments.

1 First-personal Self-knowledge
Knowledge is a factive propositional attitude: if S knows that P, it is the
case that P. As is well known, there is no consensus regarding its analysis or
even if such analysis is possible in the first place.1 For our purposes, how-
ever, it will do to take it to consist in having a true and suitably justified
belief. Knowledge therefore requires the possession of the concepts neces-
sary to entertain the proposition that P, which is the content of one’s belief.
In general, empirical knowledge is either observational or inferential or
a mixture of the two.2 Suppose I want to know what colour the curtains
of my lounge are. I look and see they are pale yellow. Suppose I want to
know whether Socrates is mortal. I reason like this: Socrates is a man, all
men are mortal, therefore Socrates is mortal. Finally, suppose I want to
know where the north is. I look at the sky, see the sun right in front of
me, check my watch and find out it is noon and therefore infer that the
north is in the opposite direction.

1.1 Groundlessness

Self-knowledge, which consists in the ability correctly to self-ascribe

mental states, like “I am in pain”, “I see a canary”, “I believe it will rain
tomorrow”, “I am scared of that dog”, seems neither inferential nor obser-
vational. For, of course, we sometimes infer our own mental states from
other ones. Recall, for instance, the passage, mentioned in Chap. 2 (§2),
1
Nozick 1981, Pritchard 2005, Sosa 2007 and Williamson 2000 are among the most prominent
figures who maintain either a given analysis of knowledge or the idea that knowledge cannot be
analysed.
2
Self-knowledge is knowledge of our own mental states, which, though not (necessarily) type-
identical with physical states, are contingent states of affairs. Hence, knowledge of them cannot be
a priori.
 3 Varieties of Self-knowledge 53

in Jane Austen’s Emma in which the protagonist realises her love for Mr.
Knightley. However, for the inference to get started at all, Emma must
already have knowledge of those mental states of hers that figure as con-
tents of the premises of her reasoning, such as, say, “I feel terrible about
the fact that Mr. Knightley could return Harriet’s feelings for him”. On
pain of an infinite regress, there must be knowledge of our own mental
states that is not inferential. Somewhere down the line, we must be able
to know our own mental states directly.3
Moreover, quite intuitively, self-knowledge cannot be a matter of
observation either: firstly, mental states are just not the kind of thing one
could observe4. Secondly, the Cartesian picture of an inner eye, which is
supposed to observe mental states that are luminously presented in the
mental arena, is more a recipe for trouble than a viable explanation of the
sense in which self-knowledge could be observational. Very briefly: it is a
recipe for trouble because it would lead to solipsism—the idea that each
of us is caught up in her own world insulated from anyone else, being
unable to know whether others have mental states and are, therefore, full-
fledged persons5. After all—the train of thought would go—how would I
know that other people have mental states at all if those mental states are
intrinsically private to them and foreclosed to me? It may really become
a recipe for catastrophe when taken a step further and taken to involve
conceiving of our psychological language as private. Since Wittgenstein,
however, private language has become synonymous with incoherence.
The idea is this: any language is a rule-governed practice and it is essential
to it that there be a distinction between correct and incorrect applications
of the terms. Still, if the meanings of our psychological vocabulary (or
indeed the contents of our psychological concepts)  are constituted by
mental states that are private to each individual, then whatever seems to
one to be the right application of the psychological term “S” is right6. This

3
Cf. Wright 1998, p. 16.
4
This point has been convincingly argued for by Ryle 1949and Shoemaker1996b
5
This point has been vigorously argued for by Ryle 1949.
6
Wittgenstein’s remarks in the Philosophical Investigations about the very possibility of a private
language are profitably read as responses to Russell’s observations on introspection in Russell (1912,
Ch. 5) and Russell (1921). Russell’s views have somehow been resurected by Gertler (2011a) and
Chalmers (2004). Assuming, for the sake of argument, that no criterion of correctness were necessary
54 A. Coliva

just means that the distinction between being right/seeming right, which
is much the same as the distinction between correct/incorrect applica-
tions of a word (or a concept), has vanished. Yet, as we have just seen,
that distinction is essential to there being a (psychological) language (and
concepts) at all. Thus, Cartesianism about self-knowledge may take us to
a conception of our psychological language, which turns it into a non-
language. However, we do have a psychological language—after all, we
tell each other about our own thoughts and feelings, sensations and emo-
tions all the time and we seem to understand each other well. Therefore,
it seems safe to conclude that Cartesianism will not do. If Cartesianism
has to go, so has the idea that self-knowledge could be observational.
Hence, the first two constraints on any feasible account of self-
knowledge are:

1. avoidance of the observational model and

2. avoidance of the inferential one.

Collectively, (1) and (2) amount to the idea that self-knowledge is

groundless. A word of caution is in order. Groundlessness consists in the
idea that self-knowledge is not the result of any substantial cognitive
achievement, such as observing or inferring from a symptom to its likely
cause. It does not necessarily involve the idea that one’s psychological self-
ascriptions are not based on anything, such as the very experience one is
undergoing when, for instance, one is in pain and avows it. Hence, on such
a minimal characterisation of groundlessness, such a feature of first-per-
sonal self-knowledge does not rule out by definition the possibility of weak
epistemic accounts of self-knowledge, which we will review in Chap. 6.

for concepts’ possession (as well as for linguistic meaning), it would seem possible to form very short-
lived, demonstrative concepts of the form that S, which would exploit the presence of the very
mental state one is having. Some theorists would object to the very idea of such short-lived concepts,
requiring the possibility of redeployment in order to grant the existence of a genuine concept in the
first place. If so, Wittgenstein’s critique would apply.
If one did not require multiple applications for the existence of the relevant demonstrative con-
cepts, one would avoid Wittgenstein’s objection or at least would considerably deflate it. But then
one would have to live with the unpalatable consequence of having only short-lived, in fact singular
psychological concepts. That would make it difficult to understand how we could ever make identity
judgements about different token mental states, or comparative judgements about them over time.
 3 Varieties of Self-knowledge 55

However, it has to be registered that many philosophers think that the

ongoing experience, which would allegedly ground one’s psychological self-
ascription, would not play the role of genuine evidence, or of a reason, for
one’s self-ascription. Accordingly, self-ascribing pain, while undergoing a
painful sensation, would simply amount to recognising what is pain as pain
or even to giving immediate expression to one’s ongoing pain through con-
ceptual or linguistic material or both. It would not amount to having an
independent mental state in view which could, as such, corroborate the truth
of one’s self-ascription. For having that mental state “in view”, in such a way
as to ground one’s self-ascription, would, presumably, be identical to, or at
least extremely close to, already recognising it as the mental state it is. But
recognising something for pain is just—or is very close to—judging “I am in
pain”. The justification of the relevant self-ascription that should be under-
written by the mental state would thus be either straightforwardly circular or
based on a mental state too tightly connected to the self-ascription one would
like to justify thereby. Moreover, various theorists are wary of the idea that
mental states have a distinctive phenomenology, which would enable one
to distinguish each of them from other ones, just on the basis of having the
relevant mental state. These theorists would therefore subscribe to a stronger
notion of groundlessness, according to which mental self-ascriptions, char-
acteristic of first-personal self-knowledge, not only fail to be the product
of observation and inference but are not based—in any epistemologically
relevant way—on one’s previous awareness of ongoing mental states.

Groundlessnessweak: first-personal self-knowledge is neither observational

nor inferential.
Groundlessnessstrong: first-personal self-knowledge is neither observa-
tional nor inferential, nor is it epistemologically based on one’s previ-
ous awareness of one’s ongoing mental states.

In what follows, we will see how these different intuitions can be devel-
oped into different accounts of self-knowledge. At this stage, it is enough
to register them and to point out that, in a minimal sense, groundlessness
amounts to the idea that first-personal self-knowledge is not the result of
any substantial cognitive achievement such as inner perception or infer-
ence to the best explanation.
56 A. Coliva

Is groundlessness, in the weak sense just characterised, a contingent

feature of our first-personal self-knowledge or else a necessary and a priori
aspect of it? Suppose that a subject, endowed with the relevant concepts
and quite proficient in applying them to herself and others, did so, in her
own case, by always observing her behaviour and inferring to its likely
cause. She would not know that she is in pain because she feels it, but
because she would see herself scream and moan, after getting injured, and
would infer that she must be in pain. She would not know that she intends
to buy some groceries because she is aware of that intention, but because
she would realise that she has gone to the shops and done her shopping.
Such a subject would strike us as someone incapable of having genuine
first-personal self-knowledge and in some sense as alienated from herself.
The word “alienation” may sound grandiose in this connection, but, mini-
mally, it would mean that the subject would bear to herself—that is, to her
mental states—the kind of relation she bears to those of others. However,
what makes a mental state of ours, ours, is, besides its origin in our brain
activity, the fact that it is known to us in a way which differs from how
we know of other people’s mental states, and in which no one else can
know of it, so as to figure among the subject’s motivations for her further
actions. The failure of groundlessness would therefore result in a lack of
“ownership”, or “mine-ness”, as we might put it, over the mental states
one would be inferring to have. Or, again, in the case of sensations and
other mental states with a distinctive phenomenology, not knowing them
directly, but by having to infer to them, starting from the observation of
one’s overt behaviour, by application of a little theory, therefore hypoth-
esising having them, would cast doubt on the fact that one is capable of
enjoying sensations and other phenomenologically salient mental states at
all. Moreover, when intentions (and other propositional attitudes relevant
to action) are at stake, a subject who would know of them only inferen-
tially, from reflection on her behaviour and by inference to its best expla-
nation, besides being “alienated” from her actions, would actually strike
us as not fully responsible for them. For she would know of her motives
only afterwards, after having witnessed herself act in certain ways. Yet, at
the time when the action was accomplished, she would not know either
what action she was in fact doing, or why. She would not know her action,
because, prior to having interpreted it, it would just be a movement or a
piece of overt behaviour to her. Moreover, failing to have immediate access
 3 Varieties of Self-knowledge 57

to her intentions, she would not know why, for instance, she happens to
be pushing a cart around in a supermarket—as she witnesses herself doing.
She would not act but be acted, and at any instance of the question “Why
did you do/are you doing that?”, she would be entitled to respond “I do
not know. Let me find out” and she would then be entitled to preface
whatever answer she might wish to give, based on inference to the best
explanation, with “Probably I did/am doing that because …” or “It is my
hypothesis that I did/am doing this because …”. Now, these responses,
though understandable from the point of view of a subject like the one
just described, would actually strike us as a symptom of mental illness if
they were given by an actual subject. They would strike us as reports on
another subject, who is the subject of one’s actions, which would happen
to live in one’s body.7 On a “thick” notion of being a rational agent, who
can be considered rationally responsible for (at least the vast majority of )
her actions, knowing of them and their motivations only through infer-
ence to the best explanation and in the way of a hypothesis would actu-
ally represent a major departure from rationality.8 Hence, groundlessness
seems to be an element of genuine first-personal self-knowledge, which in
turn is a constitutive element of being a responsible agent.
Now, to claim that groundlessnessweak is constitutive of first-personal
self-knowledge and to redeem that claim by means of a priori consider-
ations having to do, as we have seen, with the notion of sensation (and
possibly other phenomenologically salient mental states), and with that
of being rational agents who can be held responsible for their actions,
does not mean to prejudge the issue of how a satisfactory account of first-
personal self-knowledge can go about trying to meet this desideratum. In
particular, it is still possible to propose a fully epistemic account of first-
personal self-knowledge that aims to respect this requirement. It is only
by the lights of non-epistemic theories of first-personal self-knowledge
that groundlessness, just like the other features of first-personal self-
knowledge we will examine in the following, cannot be accounted for in
epistemic terms. By the lights of those theories, groundlessness has to be

7
Similar considerations are advanced in Shoemaker 1996b, as we shall see in Chap. 7 (§1).
8
As stressed in Chap. 2, in connection with propositional attitudes as commitments, to be rational
agents does not mean acting necessarily for good reasons. It means, however, to have knowledge of
the actions one is performing and of their motivations, at least for the most part, so as to be in a
position to be held responsible for them.
58 A. Coliva

considered an a priori feature of what goes by the name of first-personal

self-knowledge, it is not underwritten by any epistemic achievement on a
subject’s part, and can be redeemed only on the basis of a priori reflection
on a set of mutually interdependent concepts.

1.2 Transparency

Consider a sensation of cold or the deliberation to cook risotto for dinner.

If someone asked you “Are you feeling cold?” or “What do you intend to do
for dinner?”, you would be in a position to answer their question immedi-
ately. Our mental states seem to be directly, or transparently, known to us.
More precisely, their occurrence is usually of a piece with one’s awareness
of them. According to some theorists (Shoemaker 1996b), they are “self-
intimating” (the label goes back to Ryle 1949). If one has them, one would
be ipso facto aware of them. Similarly, the occurrent painful sensation or
fear one might have is of a piece with one’s awareness of them. Again, one’s
seeing a PC in front of one is of a piece with one’s awareness of seeing it.
Surely, however, the occurrence of states of affairs out there—even the most
banal ones—is not of a piece with one’s awareness of them. The trivial fact
that there is some pigeon flapping its wings right now outside the window,
for instance, is not of a piece with my awareness of it. I will have to look
out of the window and recognise the bird as a pigeon and so on. Nor is the
occurrence of other people’s mental states of a piece with one’s awareness of
them. Hence, for example, no matter how good one might be at figuring
out what is crossing someone else’s mind right now, that someone else is
feeling bored, is being perplexed or is annoyed by what they are reading is
not something one can be immediately aware of. As said, it is something
one will have to figure out by taking into account facial expressions, sighs,
bodily movements; connecting them with general knowledge of what those
reactions are an expression of; and finally inferring that they are bored,
perplexed or annoyed. Transparency, then, is one of the features that set
first-personal self-knowledge apart from all other kinds of knowledge.
Notice that transparency as presented so far is not equivalent to the
idea, first put forward by Gareth Evans (1982), that in order to know
whether one believes that P, one’s eyes should be directed not inward but
outward—that is to say, towards P, to see whether it is the case and, if
 3 Varieties of Self-knowledge 59

it is, to preface it with “I believe that”. This recipe for accounting for
our knowledge of propositional attitudes is often called the “transparency
method” because it bypasses one’s inner world, as it were, to direct one’s
mind and attention to P itself and to the reasons for or against P that
should lead us either to believe it or not. We will address this model of self-
knowledge in Chap. 5. The reason why we put it aside until then is that,
as remarked, Evans’s method is supposed to give us an account of how we
can have knowledge of our minds and, in particular, of beliefs and possibly
other propositional attitudes. Therefore, it cannot be taken as one of the
neutral conditions that all accounts of self-knowledge should satisfy.
Some theorists are unconvinced that transparency is a characteristic
feature of our own mental states (see Snowdon 2012 for a criticism of
transparency with respect to sensations, emotions and passing thoughts).
They point out, for instance, that higher-order mammals and infants do
not seem to be capable of self-ascribing sensations and basic emotions,
or even perceptions, although they can enjoy them. To such a worry, one
might respond by noticing that we have been talking of transparency
mostly in terms of awareness and it seems safe to hold that, at least in
the case of sensations, occurrent emotions with a characteristic ongoing
phenomenology, and conscious perceptions, their occurrence seems to
coincide with one’s awareness of them, even if one does not possess the
concepts necessary to self-ascribe them. More precisely:

Transparencyweak: if one has a given mental state M, one is aware of it.

That is to say, the mental state M is phenomenologically salient to the
subject.

Even so, it is clear that transparency does not hold unconditionally,

but only for those mental states which have a distinctive phenomenol-
ogy—leaving aside for the moment the width of this class—and are not
purely dispositional. Furthermore, subjects undergoing these mental
states will have to be cognitively lucid, attentive and alert.
However, one might object to this characterisation by pointing out that
it captures self-awareness rather than self-knowledge. For the latter, con-
trary to the former, consists in the ability correctly to make the relevant
psychological self-ascriptions. Furthermore, one might want to capture
60 A. Coliva

the idea that also propositional attitudes which may not have a distinctive
phenomenology can be transparently self-known. As we saw at the very
beginning, having formed the intention to cook risotto for dinner, one
may be immediately in a position to self-ascribe that thought, even if,
arguably, there is no distinctive phenomenology to forming such inten-
tions. In such a scenario, the likely move would be to conditionalise trans-
parency to the obtaining of the relevant (C-)conditions. Namely, subjects
should be conceptually endowed, besides being cognitively lucid, attentive
and alert. By means of such a conditionalisation, then, usual counterex-
amples to transparency, often presented in the literature on self-knowl-
edge, would easily be dispensed with. The case of animals and infants
would be countered by the requirement that subjects should possess the
relevant concepts. If, in addition, a subject is under the effect of drugs,
has her attention occupied by a pressing task, or is distracted or distressed
in various ways, she may be having sensations, emotions and perceptions
and yet not be aware of them. Given the conditionalisation just proposed,
these cases would not represent counterexamples to transparency.
A separate remark is apposite in relation to the Freudian idea that
we may have many unconscious propositional attitudes, such as beliefs,
desires and intentions, of which, by definition, we are totally unware.
Obviously, there is no denying that this might be the case. However,
Freudian mental states are mental dispositions (see Chap. 2, §2). Hence,
one may add to the C-conditions that when propositional attitudes are
at stake, they should be beliefs, desires and intentions as commitments,
not as dispositions. We will come back to this issue in due course (see
Chap. 7, §§3–4). For now, it suffices to have shown how transparency
can be understood and limited, so has to hold for at least a good range
of mental states.

Transparencystrong: Given C-conditions (including concepts’ possession,

cognitive well-functioning, alertness and attentiveness and to the
exclusion of unconscious and purely dispositional mental states), if
one has a given mental state M, one will be in a position to judge or
believe (or both) that one has it.

Now we should consider whether transparency is a contingent feature of

first-personal self-knowledge or, rather, a necessary and a priori aspect of it.
 3 Varieties of Self-knowledge 61

Let us start with sensations, perceptions and basic emotions. As we saw in

Chap. 2, one may distinguish between sensory states, with no representa-
tional content and therefore without correctness conditions, which simply
register proximal stimuli, and sensations properly so regarded. Only the
latter are necessarily conscious—that is, such that having them coincides
with being aware of them. That is to say, transparency is, on this reading,
a necessary and a priori aspect of what goes by the name of “sensation”.
Similar considerations may perhaps be put forward in the case of at least
those emotions that have a distinctive phenomenology to them. However,
the claim that transparency holds a priori, at least for some kinds of mental
state, would then seem to depend on a terminological choice. Furthermore,
one may want to insist that, for instance, the soldier who has been march-
ing for days does feel pain, even if he is not aware of it because his fatigue
and stress are so conspicuous that he is oblivious to them and just keeps
walking. It is at this stage that the C-conditions we have introduced in the
characterisation of transparencystrong are helpful to delimit those cases in
which it makes sense to think that our knowledge of our ongoing sensa-
tions meets the transparency requirement. If a subject were always unable,
while feeling pain and being attentive and alert, to judge that she is, this
would cast doubt on the fact that she does possess the relevant concepts;
alternatively, if we had reasons to think that she does have those concepts,
her inability to self-ascribe an ongoing pain would cast doubt on her cogni-
tive well-functioning to the point of impairing the idea that we are dealing
with a rational subject. Furthermore, she would not be in a position to
take responsibility for her actions. She would, for instance, try to cure her
wound, but if asked why, she would not know, even after she screamed and
moaned and while having, ex hypothesi, the relevant psychological concepts.
The case of perceptions is clearly different, for, alongside with conscious
perceptions, whose occurrence would be of a piece with one’s awareness of
them, we have acknowledged (see Chap. 2, §1) the possibility of uncon-
scious perceptions. Blind-sight would be a case in point. Hence, there
would be perceptions whose occurrence would not be of a piece with one’s
awareness of them. To such a challenge, we can respond by noticing that
clearly unconscious perceptions fall out of first-personal self-knowledge.
For we would know of them by observation and inference, by having wit-
nessed ourselves act in ways which can be made sense of only by presup-
posing that we did have those perceptions. By contrast, if we are dealing
62 A. Coliva

with conscious perceptions, then their occurrence would be of a piece

with one’s awareness of them. Furthermore, if endowed with the relevant
concepts while being cognitively alert and attentive, we would be in a
position immediately to judge (or believe) that we are having them.
It is worth noting that, given the strong link between perception and
action, the actions we would perform on the basis of unconscious per-
ceptions would fall out of responsible agency. The subject affected by
blind-sight does not know either that she is catching a ball as she does
that, or why. Hence, she cannot be held responsible for it. Transparency is
therefore a necessary and a priori element only of conscious perceptions,
which, in turn, are constitutive of responsible outer actions.
Similarly, a subject who, having formed an intention or other proposi-
tional attitudes as commitments, were systematically unable to self-ascribe
them, either would be lacking the relevant concepts or, if there were rea-
sons to think she possessed them, would seem to be somehow mentally
deranged. For she would systematically respond “I do not know” to the
question “Do you believe/desire that P/intend to φ?” after in fact having
asserted (or judged) “P”, “P is worth doing” and “I will φ” while having
the relevant psychological concepts.
Once more, claiming that transparency is a constitutive feature of
first-personal self-knowledge and that it can be redeemed on the basis of
a priori considerations concerning a range of interconnected concepts,
such as the concept of responsible agency, rationality and first-personal
self-knowledge, does not preclude the possibility of giving an epistemic
account of it. It is only by the lights of non-epistemic accounts of first-
personal self-knowledge that this requirement is not underwritten by any
kind of epistemic achievement on a subject’s part and is, in fact, just a
conceptual truth concerning a set of interlocking concepts.

1.3 Authority

Authority, in contrast, amounts to this. If you are sincere and compe-

tent with respect to the concepts you use to express your mental states,
nobody can—rationally—cast any doubt on your avowals. If you answer
my question “What are you thinking?” by saying “I think that summers
 3 Varieties of Self-knowledge 63

in Italy are really too hot” and you are sincere and know how to use ‘I’,
‘think’, ‘summers’, ‘Italy’, …, then nobody could challenge you by say-
ing “Are you sure that this is what you are thinking?”. Of course, one
can challenge the subject’s grounds for believing that summers in Italy
are too hot, by pointing out that it is August 15, you are in Italy and the
temperature is only 15 °C. Yet one cannot challenge the subject’s own
believing that she believes it. By challenging the grounds in favour of
that content’s belief, one can lead a subject to revise her belief, but this
possibility depends on the (implicit) acknowledgement that the subject
does believe—erroneously, from your point of view—that summers in
Italy are too hot and that she knows that much about herself. Similarly,
if I sincerely avow a painful sensation, it would be pointless for someone
to challenge me by saying “Are you sure that is what you are feeling?”. To
the restatement of my avowal, I could only add that I am being sincere
and that I can speak English. Therefore, subjects are authoritative with
respect to their own mental states: if they say (or judge) that they have a
certain mental state M, then they have it.
Surely, however, if someone asked you “What is the weather like?” or
“How is your mother?”, from your sincere and conceptually competent
answers “It is raining” or “She is very sad since she retired recently and
she feels useless”, it wouldn’t have followed at all that it would be inap-
propriate for someone to challenge you by saying, for instance, “Are you
sure? You haven’t taken a look out of the window and one of those relax-
ation CDs that make all the noises of nature is on”. Or: “Are you sure? I
saw your mum last night at a party, dancing to ‘Staying Alive’ with your
dad, and she did not look sad at all”. Therefore, authority is another
feature that sets self-knowledge apart from all other kinds of knowledge.
Again, some theorists are wary of authority (see again Snowdon 2012).
With respect to sensations, for instance, they point out that one can
envisage a situation where a subject is asked to locate her sensation and is
uncertain about that. This, however, would merely show that we are not
authoritative with respect to the bodily location of our ongoing sensa-
tions. Yet it would not show that we are not authoritative with respect to
the fact that we are undergoing them.
Another case often discussed in the literature is the one of a subject
who, by dreading the pain a dentist will produce by messing about her
64 A. Coliva

mouth, claims to be feeling pain already when the dentist has not even
touched her. This case can be taken care of by specifying the C-conditions
under which authority is supposed to hold. In fact, the subject should
be cognitively lucid. Arguably, however, intense fear can impair one’s
cognitive functions. Once again, the possibility, on specific occasions, of
mistaken self-ascriptions of sensations should not be taken as a counter-
example to authority, but only as imposing the need of properly charac-
terising its extent.
Another prima facie difficult case to handle is the one of confused or
vague sensations, such as itches that border pain. A subject could then
be unclear whether she is undergoing one or the other kind of sensation
and make avowals she would then like to retract. This, however, would
not be a counterexample to authority, because we are, in fact, dealing
with a confused or vague sensation, which escapes definite categorisation.
Hence, we vacillate between two concepts we do possess, when both of
them seem to apply to it to some extent. In this sense, our self-ascriptions
may be imprecise and retractable. Yet this does not show that, at least
when sensations are not of a confused nature, we would not be authori-
tative with respect to them. More generally, the ubiquitous presence of
vagueness in our concepts should not be taken—by itself—as a serious
challenge to the possibility of first-personal self-knowledge any more
than the ubiquitous presence of vagueness in the domain of colour is
normally taken as a challenge to the possibility of correct colour percep-
tions and judgements.
Another possibility is to be wrong about the content of one’s percep-
tual appearances. For instance, one can say that one’s after image contains
five red dots, when in fact it contains only four. However, authority can
be maintained even in this case because determining the number of dots
involves counting and one may go astray in doing it. Still, one would be
authoritative with respect to the fact of having an after image.
Interestingly, Eric Schwitzgebel (2008) has recently argued that we are
not authoritative with respect to our inner feelings and emotions. The
idea is that we may be bad at judging that the feelings we are currently
experiencing are feelings of anger, say, when in fact this would be clear to
a third party. It should be admitted that sometimes we do not realise what
the feelings we may experience are symptoms of. This, however, could be
 3 Varieties of Self-knowledge 65

due to poor performance in connecting one’s ongoing feelings to the dis-

positional elements of the relevant emotion. Hence, it would be a case of
mistaken self-interpretation. That is to say, it would be a case of lacking
or of going astray in one’s third-personal knowledge of one’s own mental
states, rather than a case of failure at first-personal knowledge of them.
Finally, moving on to propositional attitudes, it may happen that one
says that P, or that P would be good to have, thus manifesting the belief
or the desire that P, and yet behaves in ways that run contrary to one’s
professed beliefs and desires. The phenomenon is known as self-deception
and is taken to be the most powerful counterexample to authority. There
are three possible answers on behalf of the supporters of authority. First,
one may complicate the C-conditions to exclude cases of self-deception
from the range of propositional attitudes one is authoritative about. This
move, however, seems quite ad hoc; hence, it would be preferable to avoid
it. Second, one may propose an alternative account of self-deception,
which is compatible with the retention of authority over one’s propo-
sitional attitudes (we will explore this possibility in Chap. 7, §§3–4).
Finally, one may hold that authority can admit of exceptions, yet argue
that it is present in a significant amount of self-ascriptions about one’s
own intentional mental states.
Hence, we face the possibility of two different readings of author-
ity, which differ in strength. A minimal one has it that we are mostly—
though not always—authoritative with respect to a specified class of
mental states we can enjoy and, in particular, that self-deception may
lead us to wrong psychological self-ascriptions. A stronger reading, in
contrast, has it that, as a matter of fact, there are no relevant exceptions to
authority, once we have appropriately specified the class of mental states
for which it holds and have independently accounted for self-deception
in ways which explain away the impression that it should pose a challenge
to authority. Accordingly, we would have the following two possible read-
ings of authority:

Authorityweak: Given C-conditions (including concepts’ possession, cog-

nitive well-functioning, alertness and attentiveness), if one judges to
have a mental state M (save for dispositional ones or for the disposi-
tional elements of some mental states), one will usually have it.
66 A. Coliva

Authoritystrong: Given C-conditions (including concepts’ possession, cog-

nitive well-functioning, alertness and attentiveness), if one judges to
have a mental state M (save for dispositional ones or for the disposi-
tional elements of some mental states), one will always have it.

Be that as it may, authority too would seem to be a necessary and a

priori feature of first-personal self-knowledge. For, if one were systemati-
cally proven wrong in one’s psychological self-ascriptions, doubt would
be cast upon one’s possession of the relevant concepts. Alternatively, if
one wished to maintain that a constantly mistaken subject (with respect
to her own mental states) could still be said to have the relevant psy-
chological concepts, perhaps because she retains the ability to apply
them to other people’s mental states, her persistent self-deception would
impair her rationality. For she would avow certain mental states while
she would systematically behave in ways which run contrary to them.
Hence, we could no longer make sense of her linguistic and non-linguis-
tic behaviour.
Once again, to claim that authority is constitutive of first-personal
self-knowledge and that it is a priori connected with other notions,
such as the one of being a rational agent who can be held responsible
for her actions, does not preclude the possibility of accounting for it
in epistemic terms. It is only in the perspective of non-epistemic theo-
rists that those attempts fail and that authority is not underwritten by
any epistemic achievement on a subject’s part but is purely a fallout of
the necessary and a priori links holding between some interconnected
concepts.
Groundlessness, transparency and authority, therefore, are constitu-
tive features of first-personal self-knowledge and are constitutively tied
to what it means to possess psychological concepts as well as to what it
means for someone to be a rational subject, who can be held responsible
for her own actions.9 Any satisfactory account of first-personal self-
knowledge will have to meet these desiderata and respect the idea that,
for the appropriate kinds of mental state in suitable conditions, if they are
not met, that will impair a subject’s rationality.

9
Whether, in turn, they are also constitutive elements of what it means to be a self or a subject at
all, insofar as they are constitutive of being critical reasoners and subjects of moral norms as Burge
2011 maintains, is a further issue, which would need a separate treatment.
 3 Varieties of Self-knowledge 67

2 Counterexamples from Content
Externalism and Cognitive Sciences?
Some theorists (Boghossian 1989 and McKinsey 1991) take the rise of con-
tent externalism to be incompatible with at least authoritative self-knowledge
regarding one’s current propositional attitudes. For, if externalism is correct,
a subject may think of being thinking a water thought, say, when she is in
fact entertaining a thought about twater because of her actual causal connec-
tions with an environment in which lakes, rivers and seas are in fact filled in
with XYZ rather than H2O. Hence, it would seem to her to be thinking a
water thought, while she would not. Therefore, she would lack knowledge
of her current thought. For, arguably, knowledge of one’s thought would
entail knowledge of its nature—its being a belief rather than a hope or a
desire—and of its content—namely, its being about water and not twater.
Let us grant, for the sake of argument, that content externalism is cor-
rect. Let us further suppose that, unbeknownst to her, our subject is actu-
ally thinking a twater thought. Still, she would seem to have essentially
first-personal access to it even if she may be wrong about its actual con-
tent. Hence, she would still have transparent access to the fact that she is
entertaining a thought rather than a hope or a wish, and her access would
still be groundless. That is, it would be based neither on observation nor
on inference. Finally, she would still be authoritative with respect to its
seeming or apparent content. Some theorists would call it “narrow con-
tent” and would happily acknowledge its existence alongside with “wide
content” (Putnam 1975). Hence, although strong authority may be
impaired by content externalism, a weaker kind of authority, at least with
respect to the kind of mental states one is enjoying and the narrow con-
tent of one’s propositional attitudes, would be preserved. It is that kind of
authority that, being fundamental to psychological explanation in a way
in which, arguably, authority over the wide content of one’s propositional
attitudes is not, needs to be explained by any account of first-personal
self-knowledge and that we will try to account for in what follows.10
Let us now move on to the challenges to first-personal self-knowledge
raised in recent years by studies in cognitive psychology. They tend to chal-
lenge the idea that we are authoritative with respect to our own mental

10
For a review of classical positions on this topic, see Parent 2013.
68 A. Coliva

states. In what follows, I will draw extensively on Brie Gertler’s (2011a)

excellent discussion of the topic.11 First, there are studies that impugn the
idea that we have privileged access to our character traits, such as jealousy
or loyalty. People tend to deny being jealous and to affirm to be loyal, even
if they actually behave in ways which run contrary to their psychological
self-ascriptions. Second, it has been shown that we are bad at “affective
forecasting”.12 That is to say, we predict that a life-changing event, such
as winning the lottery, will permanently affect our psychological attitude
towards life. Studies have shown, however, that after a short period in
which one’s moods are elated, people revert to their “happiness baseline”.
In both cases, we may have privileged access to our resolutions or inten-
tions and expectations, such as the intention to be loyal, or the expectation
that winning the lottery will make us permanently happy, and yet we are
not authoritative with respect to whether we will actually behave in the
way we intend or expect we will behave. Gertler rightly notices that both
character traits and affective forecasting involve access to our own psycho-
logical dispositions. Dispositions involve counterfactual conditions and it
may well be the case that we are bad at determining whether we will satisfy
them or not. This clearly limits the scope of first-personal knowledge of
our own mental states but does not rule it out.
Other studies show that we are bad not only at determining our psy-
chological dispositions but also at indicating the causes or motivations of
our actions.13 In fact, they often remain unconscious, by operating at the
subpersonal level. Hence, we do have first-personal knowledge of certain
occurrent thoughts, such as one’s desire to have ice cream, which we take
to be the cause of our action of going towards the fridge to have some,
but in fact that thought may not be the real cause of our action. Again,
Gertler rightly notices that, since Hume, causes have been shown not
to be directly observable and rather to be the result of theorising about
established correlations between different types of events. Therefore, it
should not be surprising that we lack authoritative access to the causes
of our actions. Still, we are under the illusion of having a distinctively
first-personal access to them because we do have such privileged access to

11
See Gertler 2011a, pp. 70–86.
12
Gilbert 2006.
13
Nisbett and Wilson 1977; Libet 1985; Wegner 2002; Wilson 2002; Wegner and Wheatley 1999.
 3 Varieties of Self-knowledge 69

the thoughts that we take to be the causes of our actions. Once more, the
scope of first-personal knowledge of our own mental states appears to be
limited but not to be vanishing.
Similar results have been achieved by studies concerning our moods,
such as pessimism, like the tendency to regard the future as bleak, and
emotions, such as anger. Gertler correctly points out that moods are
dispositional mental states and that emotions involve, besides a certain
feeling, with respect to which we do have first-personal access, certain
dispositional elements. This idea chimes well with the borderline concep-
tion of the emotions we put forward in Chap. 2 (§3), according to which
emotions cannot be reduced to either feelings or dispositions. It may well
be, then, that although we have privileged access to their phenomeno-
logical aspects, we do not have it to their dispositional elements.
Finally, if we do not have first-personal knowledge of dispositional
mental states, that entails that we are not authoritative, nor do we have
immediate and groundless access to a lot of propositional attitudes, such
as beliefs and desires, which are dispositional. Indeed, this is only to be
expected. Yet we do also have the impression of having first-personal access
to our beliefs, desires and intentions. Our view, presented in Chap. 2, that
we should distinguish between different kinds of propositional attitudes—
that is, those as dispositions and those as commitments—comes in handy
at this stage. For we may say that whereas we lack first-personal knowledge
of the former, we have it of the latter. How we do have it will have to be
explained, and we will consider various explanatory accounts in the fol-
lowing chapters. Yet, although this imposes a further limitation to first-
personal self-knowledge, it does not show that we lack it altogether. Rather,
it merely shows that first-personal self-knowledge is limited to occurrent
sensations, including those characteristic of at least basic emotions, percep-
tions, current thoughts and propositional attitudes as commitments.

3 Third-personal Self-knowledge
So far, we have examined the characteristic aspects of first-personal self-
knowledge—that is to say, those aspects, which set our knowledge of our
own mental states apart from other kinds of knowledge, such as knowl-
edge of truths about the external world, usually achieved through percep-
70 A. Coliva

tion, or about other people’s mental states, customarily obtained through

inference to the best explanation, starting from the observation of their
overt behaviour. The various limitations to first-personal self-knowledge
we considered, however, show that ample room for error is provided in
figuring out our own mental states. This can be explained by noticing that,
even in our own case, we often acquire knowledge of our mental states
in a third-personal way, as a result of observation of our behaviour and
inference to the best explanation, or else thanks to inference to the best
explanation starting from one’s first-personal knowledge of one’s ongoing
sensations, feelings, perceptions and propositional attitudes. We therefore
possess not only first-personal but also third-personal self-knowledge. In
general, that is, we are often self-knowing interpreters, as we may put it.
When third-personal self-knowledge is at stake, however, none of the
characteristic features of first-personal self-knowledge is present. There is
no groundlessness, for indeed our knowledge of our own mental states is
based either on observing our behaviour and inferring to its likely cause
or on our awareness of our inner phenomenology and to inferring to its
probable  causal explanation. Two avenues for error are open, though.
For we may mischaracterise our own behaviour and hence start out our
inferences with mistaken premises. Or we can be wrong in identifying a
certain mental state as the likely cause of our ongoing feelings or overt
behaviour. Thus, authority does not hold either. Nor does transparency,
as the occurrence of the relevant mental states is characteristically not
of a piece with one’s awareness of them or indeed with being in a posi-
tion to self-ascribe them. Indeed, to gain knowledge of them, we need to
engage in observation and inference and we may fail to perform them,
thus remaining blind to ourselves, as it were.
Still, there is an element of third-personal self-knowledge that sets it
apart from knowledge of other people’s mental states. Namely, some-
times the inference can start on the basis of other mental states one has
knowledge of in a first-personal way, such as one’s sensations, immediate
feelings, and propositional attitudes as commitments.14 This obviously
cannot be the case when we acquire knowledge of other people’s mental
states through inference to the best explanation. In that case, the inference

14
I was pleased to find a similar claim in Cassam 2014.
 3 Varieties of Self-knowledge 71

starts from observing their overt linguistic and non-linguistic behaviour.

Let us therefore consider some examples of third-personal self-knowledge.
Cases of inferences starting from the observation of one’s own behaviour
can be those that lead us to self-ascriptions of propositional attitudes as dis-
positions, whether of a Freudian kind or not. Consider a mother who reflects
on her behaviour towards her son and compares it with the one towards her
daughter. She can judge that she is more often benevolent towards the lat-
ter and therefore infer that she prefers her daughter to her son. Of course,
she may have gone wrong in judging that she is more munificent with her
daughter in the first place, as she may underestimate the number of times
in which she has been open-handed towards her son. Hence, her self-ascrip-
tion is neither groundless nor transparent or authoritative.
Sometimes, we figure out our complex emotions, such as love towards
someone, by reflecting on our behaviour and inner phenomenology in their
presence or at the prospect of some event involving them. This is the case
illustrated by the passage from Jane Austen’s Emma quoted in the second
chapter. Moreover, it is fair to say that it is a common experience to realise
that one is in love with someone else by reflecting on the kind of feelings
provoked by seeing them or by considering one’s hope to encounter them
on a given occasion or one’s careful dressing at the prospect of seeing them,
and so on. Obviously, we can go wrong in our self-interpretation. Our
behaviour and feelings may not be symptoms of love but of being posses-
sive and thus dreading the prospect that a good friend should inevitably
stop being so close to us as the result of getting married to another person.
Alternatively, we can mistake our behaviour and feelings as symptoms of
love, when they are in fact signs of our vanity and complacent attitude
towards those people who are usually kind to us (or to others in general).
Sometimes, in contrast, we are self-blind in the sense that despite showing
many of the characteristic symptoms of love, say, towards someone, we do
not realise that we are in love with them. That seems to have been the case
with Emma, before a sudden event made her realise her feelings for Mr.
Knightley. Such blindness can have different causes. Women often accuse
men of lacking introspection, because they seem to lack the propensity of
making sense of their overt behaviour and inner feelings in the way just
sketched. That may be due to a variety of factors both cultural and psy-
chological, like not paying attention to these cues. Some other times, there
72 A. Coliva

may be deeper unconscious reasons for that. For instance, if the person in
question is a colleague and one is generally convinced that intimate rela-
tions with one’s colleagues should be avoided at all costs in order not to
harm one’s career or public image, one may persistently ignore the evidence
that shows that one is indeed in love with a colleague.
We have also seen how alleged counterexamples to first-personal self-
knowledge drawing on recent findings in cognitive studies can in fact
be interpreted as interesting examples of third-personal knowledge of
our own mental states. Therefore, for instance, we think we have cer-
tain character traits, or that we will have certain emotional responses
in the light of alleged life-changing events, which we actually lack. The
interesting aspect of this failure at self-knowledge is that the reason why
we think we would have these mental dispositions is that we presum-
ably engage in imagining how we would behave in a given situation.
For whatever reason, which may or may not be biased, we think we
would behave in ways that would depend on having a certain mental
state. Yet, when faced with a real-life situation, we do not behave as we
imagined we would do. The kind of procedure followed in these cases,
in order to try to gain knowledge of our mental dispositions, is indeed
similar to the one we would apply if we were to figure out someone else’s
mental states by means of simulation and its correct outcome is obvi-
ously not guaranteed. It is therefore another example of third-personal
self-knowledge.
Another intermediate case, which does not immediately coincide with
knowledge of our own mental states either through inference to the best
explanation or through mental simulation, is the one in which we suddenly
notice an aspect of ourselves we had been previously blind or oblivious to or
indeed reconceptualise an aspect of ourselves by suddenly seeing it in a dif-
ferent way. We may all of a sudden realise that we are not nice to other peo-
ple—we do not actually care about them that much—but only act polite, or
we may suddenly see a character trait of ours not as a symptom of strength
but of weakness. Seeing-as, switches of aspect and noticing aspects are com-
plex phenomena even when they clearly concern perceivable objects and
properties and involve the operation of perceptual faculties.15 Obviously, in

15
Cf. Wittgenstein PI, II, xi. In Coliva 2012b, I consider the application of these notions to the use
of diagrams in geometrical proofs and provide the relevant bibliographical references.
 3 Varieties of Self-knowledge 73

the case of our own mental states, these notions find only mediated or indi-
rect employment, for no real perceptual faculty is involved and yet it does
not seem that we engage in any complex inferential thinking or in mental
simulation. Furthermore, the relevant psychological concepts are necessar-
ily needed, whereas, arguably, purely perceptual switches of aspects do not
necessarily require concepts (although they may and often do involve them).
We may say that we suddenly take a pattern of overt behaviour, and possibly
other inner aspects of our psychology, as instantiating this or that psycho-
logical property either because that pattern is manifest to us for the first
time or because the relevant concept is available to us at last and a pattern of
behaviour and inner elements we had possibly noticed before seems imme-
diately to fit the newly acquired concept. Sometimes, these new concep-
tualisations or reconceptualisations contradict previous ones and that gives
rise to a switch of aspect regarding ourselves and our personalities. Still, in
these cases, inference is involved at least in the following sense. While there
is no inference to the best explanation, the characteristic notes of the newly
acquired (or applied) concept guide the process of subsuming the observed
pattern of behaviour and inner elements under the concept. That may hap-
pen very rapidly, and so, phenomenologically, we may not even seem to be
engaging in inferences, yet epistemically that is what grounds our judgement
“I am F” (where F is a psychological concept). It is one possible case of self-
interpretation and it can go wrong because, after all, the pattern observed
may not fit the concept or we may have ignored other features which would,
if taken into account, lead to a different psychological self-ascription.
Another interesting case is the one in which we aquire self-knowledge
by means of someone else's testimony. For instance, a mother tells her
son he is pushy, or even that he is being pushy on a specific occasion.
He believes his mother and forms the belief, based on testimony, that he
is (being) pushy. This would count as a clear case of testimonial—and
therefore third-personal—self-knowledge. Testimonial self-knowledge is
also achieved in more institutionalized contexts in which, for instance, a
therapist makes a diagnosis about a subject's psychological situation and
the latter forms the corresponding belief.
What is important to notice at this stage is that third-personal self-
knowledge exhibits all the typical aspects of any substantive cognitive
achievement, which usually results in knowledge. We base our psychologi-
cal self-ascriptions on various kinds of evidence—both observational and
74 A. Coliva

testimonial—we engage in inferences and we deploy imagination to simu-

late a situation and predict our own actions, therefore determining their
psychological motivations. All these procedures admit for the possibility of
error and for the chance of not being deployed at all, thus resulting in a
form of self-blindness. Furthermore, the kind of abilities called upon can be
variously distributed among subjects, who are otherwise equal with respect
to cognitive functions, conceptual endowment and rationality. Moreover,
it can be finessed with experience, study and exercise. Third-personal self-
knowledge, therefore, is not any different from knowledge of truths about
physical objects around us or about other people’s mental states in these
respects. Yet it is special in one kind of evidence it may derive from—that
is, our first-personal knowledge of occurrent sensations, feelings, perceptual
experiences and propositional attitudes as commitments. Yet it is only for
third-personal self-knowledge that the usual traits of knowledge are clearly
present. For there is cognitive achievement, there is ample possibility of
error, or self-blindness, without resulting in failures at rationality or in lack
of concepts’ possession, and there is room for expertise and for improving
one’s skills. None of this, in contrast, seems to characterise first-personal self-
knowledge. As we shall see, for some theorists, this means that first-personal
self-knowledge is not, after all, real knowledge. Rather, it is a set of constitu-
tive claims, which can be variously grounded and redeemed. We explore this
route in Chap. 7–8. Before getting there, however, we need to consider the
most prominent attempts to account for first-personal self-knowledge as a
genuinely, yet clearly sui generis, epistemic phenomenon.

4 Summary
In this chapter, we have seen the characteristic traits of first-personal self-
knowledge (§1)—that is, groundlessness, transparency and authority—
and the different readings which can be offered of them. Assuming as a
methodological rule the least demanding reading of each of them, compat-
ible with the project of accounting for self-knowledge, we have argued that
it is a necessary and a priori aspect of first-personal self-knowledge, whose
failure would impair a subject’s conceptual mastery with respect to the rel-
evant concepts or her “thick” rationality and therefore her being rightly
considered a responsible agent. We have then defended the thesis that there
 3 Varieties of Self-knowledge 75

is first-personal self-knowledge from challenges coming from externalist

conceptions of the individuation of mental content by arguing that we
would retain first-personal self-knowledge of the kind of propositional atti-
tudes we enjoy (beliefs as opposed to desires and intentions, say) and of the
narrow aspects of mental content. We have also defended it from objec-
tions which come from cognitive science (§2). All the examples reviewed
show that we may fail to know dispositional (sometimes causal) elements of
our mental states. This, however, does not impugn the fact that regarding
the non-dispositional aspects of our mental life we can have first-personal
knowledge of them. In fact, the cases reviewed simply show that the scope
of first-personal self-knowledge is limited to certain classes of mental states
or to some aspects of them—namely, to those which are not dispositional
in character. For all these mental states or for the dispositional elements of
some of our mental states, the kind of knowledge we can have of them is
third-personal. In §3, we have reviewed the main features of third-personal
self-knowledge, which is based either on inference to the best explanation
starting from the observation of one’s behaviour or from one’s awareness of
one’s ongoing mental states (or from both) or on mental simulation or on
inferential conceptual deployment or on testimony. None of the distinctive
traits of first-personal self-knowledge—groundlessness, transparency and
authority—holds for third-personal self-knowledge, which then qualifies
as a kind of genuine cognitive achievement open to the possibilities of error
and mistake without impairment of one’s conceptual mastery and “thick”
rationality, which are characteristic of knowledge in other domains, such as
knowledge of the outer world and of other minds.

Bibliography
Boghossian, P. (1989). Content and self-knowledge. Philosophical Topics, 17, 5–26.
Burge, T. (2011). Self and self-understanding. The Dewey lectures (2007–2011),
The Journal of Philosophy, CVIII/6–7. Reprinted in Cognition through under-
standing (pp. 140–228). Oxford: Oxford University Press.
Cassam, Q. (2014). Self-knowledge for humans. Oxford: Oxford University Press.
Chalmers, D. (2004). Phenomenal concepts and the knowledge argument, In P.
Ludlow, Y. Nagasawa, & D. Stoljar (Eds.), There’s something about Mary:
Essays on phenomenal consciousness and Frank Jackson’s knowledge argument
(pp. 269–298). Cambridge, MA: MIT Press.
76 A. Coliva

Coliva, A. (2012b). Human diagrammatic reasoning and seeing-as. Synthese,

186(1), 121–148.
Evans, G. (1982). The varieties of reference. Oxford: Oxford University Press.
Gertler B. (2011a) Self-Knowledge (London/New York: Routledge).
Gilbert, D. (2006). Stumbling on happiness. New York: Knopf.
Libet, B. (1985). Unconscious cerebral initiative and the role of conscious will
in voluntary action. Behavioral and Brain Sciences, 8, 529–566.
McKinsey, M. (1991). Anti-individualism and privileged access. Analysis, 51,
9–16.
Nisbett, R., & Wilson, T. (1977). Telling more than we can know: Verbal reports
on mental processes. Psychological Review, 84, 231–259.
Nozick, R. (1981). Philosophical explanations. Cambridge MA: Harvard
University Press.
Parent, T. (2013). Externalism and self-knowledge. The Stanford Encyclopedia of
Philosophy (Summer 2013 Edition), Edward N. Zalta (Ed.), URL = http://
plato.stanford.edu/archives/sum2013/entries/self-knowledge-externalism/
Pritchard, D. (2005). Epistemic luck. Oxford: Oxford University Press.
Putnam, H. (1975). The meaning of ‘meaning’. Minnesota Studies in the
Philosophy of Science, 7, 131–193.
Russell, B. (1912/1959). The problems of philosophy. Oxford: Oxford
University Press.
Russell, B. (1921). The analysis of mind. London: George Allen & Unwin.
Ryle, G. (1949/2009). The concept of mind. London/New York: Routledge.
Schwitzgebel, E. (2008). The unreliability of naïve introspection. Philosophical
Review, 117, 245–273.
Shoemaker S. (1996b) “Self-Knowledge and Inner Sense. Lectures I-III”, in
Shoemaker S. (1996a), pp. 201–268.
Snowdon, P. (2012). How to think about phenomenal self-knowledge, in Coliva
A. (ed), 243–262.
Sosa, E. (2007). A virtue epistemology. Oxford: Oxford University Press.
Wegner, D. (2002). The illusion of conscious will. Cambridge MA: MIT Press.
Wegner, D., & Wheatley, T. (1999). Apparent mental causation: Sources of the
experience of will. American Psychologist, 54, 480–492.
Williamson, T. (2000). Knowledge and its limits. Oxford: Oxford University
Press.
Wilson, T. (2002). Strangers to ourselves. Discovering the adaptive unconscious.
Cambridge MA: MIT Press.
Wright, C. (1998). Self-knowledge. The Wittgensteinian legacy, in Wright C.,
Smith B. and Macdonald C. (eds), 15–45.
 4
Epistemically Robust Accounts

In this chapter, we consider epistemically robust accounts of self-

knowledge. We start with recent inner sense theories (§1), which have
been proposed by David Armstrong and William Lycan. Their differences
notwithstanding, they hold that self-knowledge is the product of a reli-
able cognitive mechanism that tracks first-order propositional attitudes
and produces the corresponding  second-order ones. The mechanism is
physically realised in our brains and is operative at the subpersonal level.
Indeed, according to Lycan, there is a real inner sense faculty, dedicated
to keeping track of first-order mental states through the operation of
attention. Although the model accounts for groundlessness, transparency
and authority, it severs the connection between self-knowledge and ratio-
nality and concepts’ possession. Hence, it is found to be problematical.
We then move on to Alison Gopnik’s theory-theory (§2). According to
this model, we know our minds in the same way we know others’—that
is, through inference to the best explanation, based on observation of
one’s overt behaviour and non-intentional mental states and the applica-
tion of a real theory of the mind subjects acquire around the age of four.
The model is variously criticised, mainly from a philosophical point view.
Its inadequacy to account for self-knowledge is revealed by seeing it as

© The Editor(s) (if applicable) and The Author(s) 2016 77

A. Coliva, The Varieties of Self-Knowledge,
DOI 10.1057/978-1-137-32613-3_4
78 A. Coliva

either collapsing into a form of crude behaviourism or as presupposing

the very knowledge it sets out to explain. A recent variant of inferential-
ism, proposed by Quassim Cassam, is also considered and found equally
wanting.
Finally, we examine simulation theories (§3), which have been vari-
ously proposed primarily by Alvin Goldman and Robert Gordon. The
main idea is that in order to know other minds we need to simulate other
subjects, rather than apply a theory, given the observation of their behav-
iour. Simulation theorists, moreover, argue that it is through simulation
that we gain knowledge of our hypothetical and future mental states.
They differ, however, in their account of the workings of simulation and
in their account of psychological concepts’ possession. Whereas Goldman
requires subjects to have knowledge of their own mental states, in order
to be able to simulate others, and thinks such knowledge is achieved
through introspection, by recognising and naming one’s ongoing phe-
nomenologically salient mental states, Gordon puts forward a radically
different and cognitively more parsimonious view, according to which
there is no need to have self-knowledge to simulate others. Furthermore,
he has a distinctive proposal about the acquisition of psychological con-
cepts. Even though there are important elements to Gordon’s proposal,
we cast doubts on several key aspects of it. It remains, however, that simu-
lation theorists are right in thinking that we often acquire knowledge
of our hypothetical or future mental states through simulation. Yet it is
shown that in these cases self-knowledge is in fact third-personal rather
than genuinely first-personal.

1 Inner Sense Theories: Armstrong

and Lycan
Armstrong (1968) has proposed a remarkably simple and appealing expla-
nation of self-knowledge. According to Armstrong, who is no friend of
metaphysical complications, there is just one kind of stuff—physical
stuff—and everything must be explained (or, at any rate, explainable) in
causal-nomological terms. According to Armstrong, self-knowledge is the
result of the operation of a reliable cognitive mechanism. That is to say, our
brains are so wired that whenever there is a first-order mental state, such as
 4 Epistemically Robust Accounts 79

my belief that there is a piece of paper in front of me right now, I am in a

certain brain state; then the operation of a suitable physical mechanism—
which is a kind of scanning mechanism of our first-order mental states—
brings about the occurrence of another physical state which corresponds to
a second-order mental state, that is, the belief that I believe that there is a
piece of paper in front of me right now.1 According to Lycan (1996), there
is a real inner faculty, which is dedicated to monitoring one’s own first-
order mental states, through the operation of attention. Since their views
are remarkably similar, we will consider only Armstrong’s in the following.
This model gets rid of the Cartesian idea that self-knowledge is a mat-
ter of inner observation: there is simply no observation going on here.
There is just a hard-wired mechanism which, given a certain brain state,
causally produces another one, without requiring the subject to be con-
scious of it and its workings. Since obviously there is no inference to the
best explanation either, groundlessness in its weak form is satisfied (see
Chap. 3, §1). It is satisfied also in its strong version, since the first-order
mental state does not play the role of evidence which speaks to the truth
of the second-order one—it is just its reliable cause.
Moreover, Armstrong’s model accounts for transparency even in its
strong form (see Chap. 2, §1): the mechanism gets into operation when-
ever there is the relevant first-order mental state and produces the corre-
sponding second-order belief. Notice, moreover, that this operation does
not require language possession, according to Armstrong. Hence, we are
immediately endowed with the relevant belief about our first-order men-
tal states, even though only creatures endowed with the relevant linguistic
competence could issue reports on their own mental states.2
In addition, the model explains authority at the very least in its weak
form if not also in its strong one (see Chap. 3, §1) because, accord-
ing to Armstrong’s anti-functionalist picture, the second-order mental

1
Armstrong focuses in particular on self-knowledge of intentional mental states. Although he is not
always clear on this point, his proposal is generally taken to be an explanation of how we can form
reliable (and therefore knowledgeable) beliefs about first-order mental states. Notice that if it were
otherwise—that is, if his theory accounted just for how we can be conscious of our first-order
mental states, in a way which does not involve  forming beliefs about them—then it should be
completed by how such consciousness could give rise to self-knowledge properly so regarded.
2
This, I take it, requires concepts’ possession, since one cannot entertain beliefs if one does not have
the concepts necessary to grasp the propositions which figure as their contents. Alas, Armstrong is
not always clear on this score.
80 A. Coliva

states will be reliably produced by the appropriate first-order ones, when

one has the latter (save, possibly, for cases of self-deception, on which
Armstrong is silent). Therefore, the second-order beliefs that are pro-
duced by means of its proper operation are going to be right. Moreover,
since they will be true and reliably produced, they will amount to knowl-
edge—on the reliabilist account of knowledge that will be the obvious
pendent of Armstrong’s position.
Finally, this model is scientifically acceptable for it relies only on physi-
cal stuff—the brain—and on causal relations that are subsumed under
physical laws. It is just a matter of time: neuroscientists will find out
where the mechanism is located and how exactly it works.
As simple and plausible as it may sound, this model is far from being
satisfactory. First of all, whenever a causal mechanism is involved, it may
break down. Hence, it would be a pure contingency that most of the
times in which subjects say they have a certain mental state, they do
have it and that most of the times in which they have it, they know
that they do. Thus, transparency and authority would hold only con-
tingently and a posteriori. However, it seemed that they were necessary
and a priori features of our first-personal knowledge of our own mental
states.3 Presumably, Armstrong will have to say that it was indeed an
illusion—grounded in the fact that people’s brains work so reliably that
any time there is a first-order mental state there is also the corresponding

3
This is indeed the driving thought behind Sydney Shoemaker’s objections to Armstrong’s model.
According to Shoemaker, who is a functionalist about mental states, each mental state is individu-
ated by its functional role—that is, its characteristic input and output. He then claims that, con-
trary to Armstrong’s causal model, there cannot be a given first-order mental state unless certain
behavioural outputs are in place. They do, however, require awareness of one’s first-order mental
states. For instance, since pain is individuated as the mental state that depends on physical injury
and gives rise to characteristic behaviour, involving the attempt to get rid of it, the belief (or the
ability to have the belief ) that one is in pain is necessary in order to be in pain. Therefore, there
cannot be self-blind creatures—that is to say, creatures who are capable of having first-order mental
states and yet do not have (or are in principle incapable of having) knowledge of them. Similarly,
beliefs as possessed by rational creatures ought to be revisable based on contrary evidence. This,
however, requires the ability to know that one has a given belief, which should be revised in light
of counter-evidence. Neither objection is decisive though. For Armstrong and Lycan could argue
that, on the one hand, pains are possible and can have a distinctive phenomenology even if one is
not aware of them qua pains. Infants and animals would indeed seem to be capable of feeling pain
without having knowledge of it. Similarly, a subject who lacked the relevant conceptual repertoire
could have a first-order belief and even change it on the basis of contrary evidence and yet not be
capable of having beliefs about one’s own beliefs or of answering the question “Do you believe that
P?”. We discuss Shoemaker’s position in Chap. 7 (§1).
 4 Epistemically Robust Accounts 81

second-order one. This view, however, is not without problems. Let us

review some of them.
We saw in Chap. 3 how groundlessness, transparency and authority
are constitutively tied to concepts’ possession and “thick” rationality. This
creates a problem for Armstrong’s position. To see it, consider a blind
subject, whose visual organs are impaired. Although she may not be able
just on the basis of sight to tell whether she is in front of a square object,
she could still have the concept of a square and of a square object and
be able to apply it on the basis of other sensory perceptions. Despite
being unable to see, she would still possess the relevant concepts and
form rational beliefs about objects in her surroundings. Not so in the
case of a subject whose inner scanning mechanism did not work properly.
Hence, it is doubtful that she would still possess the relevant psychologi-
cal concepts as she could not form beliefs about her own mental states. If
she then reverted to another method to form them (just as a blind person
will use some method other than sight to form beliefs about the shape
of objects in her surroundings), like inference to the best explanation,
she would no longer have first-personal self-knowledge. This will have a
bearing on her rationality, however, and on her counting as a responsible
agent, as we saw in Chap. 3. For she would know of her own motives for
action only after acting on them.4
Moreover, Armstrong’s model should be empirically testable (at least
in principle), since it is supposed to be scientifically amenable. Therefore,
suppose you want to find out the empirical correlation between a given
first-order mental state and the belief that one has it. First, you must
find out what neural states correspond to being in pain or to hoping that
peace will be reached in the Middle East. This, however, can be done on
the basis of purely behavioural—that is, physical—criteria only in a lim-
ited number of cases. It could be done for pain—assuming that if some-
one is physically injured, screams and moans they are in pain. Therefore,
when you see those symptoms, you look into the subject’s brain—with
appropriate instruments, of course—and find out what neural state she is
4
One could also toy with the idea that it is a mere contingency that the scanner mechanism, which
is a physical one, operates only on our first-order mental states. It should therefore be entirely
conceivable that we could be so hard-wired as to have knowledge of first-order mental states origi-
nated in someone else’s brain. There would then be a problem of specifying how exactly the model
can account for first- rather than third-personal knowledge of mental states.
82 A. Coliva

in. You then see what mental state she is in when she avows her pain and
hence has the belief that she is in pain. Finally, you test the correlation
between these two mental states to see whether it is reliable.
However, you cannot do this for the hope that peace be reached in the
Middle East: there are no merely behavioural symptoms that correspond
to hoping that peace be reached in the Middle East.5 Hence, in the case of
pain, one may individuate the neural configuration that realises it, inde-
pendently of the avowal of pain and then verify whether a subsequent
avowal is reliably caused by the neural realisation of pain. In the case of
the hope that peace be reached in the Middle East, in contrast, this can-
not be done, because, to repeat, we cannot have access to the first-order
mental state independently of the subject’s avowal that she has it. Hence,
Armstrong’s model, which promised to be scientifically amenable, seems
to fail to be empirically testable and to do so because it must presuppose
self-knowledge (and indeed in the form of an avowal) in order to indi-
viduate the relevant first-order mental states in the first place.6
Of course, one may want to protest that I am being unfair to
Armstrong because I am allowing too narrow a behavioural basis: there
is no reason why the relevant behaviour should be merely physical, as it
could comprise also linguistic items, such as the assertion of “Peace in
the Middle East would be good to have”. If this larger behavioural basis
is allowed—the train of thought would be—then one might find out
what neural state it corresponds to and see whether it is followed by the
appropriate psychological avowal—that is, “I hope peace will be reached
in the Middle East”. Conversely, one might test whether the avowal is
accompanied by the brain state one is in when one is enjoying the first-
order mental state. Let us suppose that by so doing we individuate two
different brain states—the brain state which would correspond to uttering
“Peace in the Middle East would be good to have” and the brain state

5
Or, at any rate, behavioural manifestations underdetermine the individuation of precisely that
mental state.
6
So, although there may be causal mechanisms that enable self-knowledge, as I presume there must
be causal mechanisms that enable thought in the first place, self-knowledge cannot be explained by
appealing only to them. Moreover, it may well be the case that the causal mechanisms that enable
self-knowledge are not anything like a dedicated faculty or scanner mechanism but the ones which
underwrite concepts’ possession and reasoning.
 4 Epistemically Robust Accounts 83

which would correspond to uttering “I hope peace will be reached in the

Middle East”. However, in this case, there is no guarantee that—even for
conceptually endowed subjects—the first-order brain state should always
(or mostly) be accompanied by the second-order one, as there is no guar-
antee whatever that the first utterance should be followed by the second.
Transparency would not be accounted for. The same objection would
apply to the case of sensations. There is no guarantee that one’s being in
pain will be accompanied by the relevant second-order mental state, as
there is no guarantee that one’s screams and cries will be accompanied by
the avowal “I am in pain”, let alone by “I believe I am in pain”.
Nor, for the same reason, is there any guarantee that the neurologi-
cal realisation of the psychological avowal should be accompanied by
the neurological realisation of the first-order mental state. For, on this
account, the two mental states are independent of each other and one
could conceivably have the relevant conceptual repertoire and routinely
avow mental states one does not have. Just as one may be blind and
yet capable of applying the relevant conceptual repertoire to physical
objects, on the basis either of the deliverances of other sensory organs
or of inference. Yet this seems problematical. We do not seem to be pre-
pared to acknowledge the possibility that one might have the concepts of
pain and belief, say, if one issued first-person psychological judgements,
which were normally out of kilter with the relevant  first-order mental
state and the concomitant pain- or belief-behaviour. Possession of psy-
chological concepts seems tied to the ability of applying them to one’s
own case. Alternatively, if one insisted that one would still retain the
relevant conceptual repertoire, perhaps because one would still be able
to apply psychological concepts to other people, the rationality of the
subject would seem to be impaired. For one’s overt behaviour and one’s
avowals would never be aligned. What, according to Armstrong’s model,
should be considered merely causal malfunctioning leading, in the usual
run of cases, to simple mistakes in judgement, would actually amount to
a failure at rationality. Armstrong’s account, therefore, seems incapable
of accounting for these intuitions, which are of a piece with the idea that
transparency and authority hold necessarily and a priori. Alternatively, if
the subject retained the ability correctly to ascribe mental states to one-
self, yet on the basis of inference to the best explanation, based on the
84 A. Coliva

observation of one’s own behaviour, she would have lost the ability of
having distinctively first-personal self-knowledge—the very ability which
Armstrong’s model was supposed to explain.
Finally, the test might also show that the two mental states have in
fact the same neural realisation. Then, ironically, the test would support
precisely constitutive positions. In particular, it would support those con-
stitutive accounts, such as Shoemaker’s (see Chap. 7, §1) that hold that
there is just one mental state—with its neural realisation—and that all
is needed to have knowledge of it is a normal degree of rationality, intel-
ligence and conceptual mastery.

2 Inferential Theories: Gopnik and Cassam

Inferential theorists are united in considering self-knowledge, in particu-
lar of our propositional attitudes, the result of inference. We will consider
two prominent versions of this model: Gopnik’s and Cassam’s.
Let us start with Gopnik’s (1983) well-known account of self-
knowledge, which aims to explain how we can know our propositional
attitudes only. In some ways, it is a development of the old-fashioned and,
by now, fallen-into-disrepute behaviourist approach insofar as it claims
that self-knowledge is wholly inferential and the basis for the inference
is (mainly) overt behaviour. Yet, in some ways, it also incorporates the
Cartesian point that self-knowledge is observational. The idea is that chil-
dren around the age of four come to possess a theory of the mind on the
basis of which they interpret their own behaviour as well as others’. The
application of this theory gives them knowledge of their own minds as
well as of others’. Hence, the idea is that they acquire knowledge of their
own mental states, such as their own desire to have an ice cream, reason-
ing as follows: “Since I am feeling hungry and I am going towards the
fridge where there is an ice cream, which will satisfy my appetite, I desire
an ice cream”.
Of course, Gopnik is well aware of the fact that it does not seem to us
to be doing any inference when we get to know our own mental states,
even less that we have to wait and see how we behave in order to find out
what we think. However, she thinks she can account for the distinctive
 4 Epistemically Robust Accounts 85

phenomenology of self-knowledge. Her idea is that much like a seasoned

scientist can actually see an electron in a cloud chamber, those who have
acquired the theory of mind will be able to apply it so rapidly that it will
naturally seem to them as if they were actually seeing their own mental
states. Groundlessness is therefore an illusion, brought about by the con-
stant employment of the theory, which eventually becomes automatic for
us. Transparency, in its weak version, which consists in being phenom-
enologically aware of one’s first-order mental states, is again an illusion,
for there is no distinctive phenomenology to propositional attitudes, but
it is an illusion Gopnik thinks she can account for, given the analogy with
the seasoned scientist just reviewed. Gopnik’s model accounts for trans-
parency in its strong version, since if one does have the relevant concepts,
one will be able to judge that one has the relevant propositional attitude.
Authority in its strong version, in contrast, is deemed a straightforward
illusion: there is no reason why, in principle, we could not be wrong
about our own mental states, just as we can be wrong about other people’s
mental states.7 Still, it is true that with practice and by being constantly
around ourselves we become very good at figuring out our own mental
states. That is why we are mostly right about them. So authority holds at
least in its weak version.
This model has been variously criticised. For instance, it has been held
to be inadequate, from a psychological point of view, as it would require
inferences, which are too complex for children of that age to perform.
Moreover, it has been criticised because it remains an open question where
the concepts children should employ could come from.8 Furthermore, it
has been deemed inadequate because it is not clear that we ever possess a
real theory about other minds or even our own. Nothing like nomologi-
cal generalisations can actually be provided in this domain. However, the
point of the criticism, which I will develop in the following, is to concede
that the model proposed by Gopnik could take care of these objections
and yet would remain philosophically unsound. In the process, the philo-
sophical problems of this model will be brought out.
7
Indeed, the experimental evidence on which Gopnik develops her model is precisely pointing to
the fact that before a certain age children just make a lot of mistakes both in the ascription of men-
tal states to others and to themselves.
8
See, in this connection, Dokic and Proust 2002, pp. vii–xxi.
86 A. Coliva

First, this whole idea of seeing our own mental states is a muddle.
Indeed, Descartes9 got the phenomenology of self-knowledge wrong: we
do not see our own mental states. The use of that verb is highly meta-
phorical. Hence, the analogy between the child and the seasoned scientist
is beside the point—simply, there is nothing like seeing which should be
accounted for in the first place.10
Secondly, even in the little sketch of reasoning I presented as an exem-
plification of what Gopnik has in mind, there is already a piece of self-
knowledge—that is, the subject’s knowledge of her feeling hungry. To be
fair, Gopnik allows for mental states other than propositional attitudes to
enter the inference and she does not regard this as a problem since she claims
to be merely offering an account of our knowledge of intentional mental
states such as beliefs. However, the point is that one needs knowledge of
one’s own intentional mental states too, to perform the relevant inference.
For, surely, one cannot infer that one desires to have an ice cream just from
observing one’s own physical behaviour—one’s moving towards the fridge,
say—since that behaviour would be compatible with thousands of different
explanations. Nor can one infer one’s desire to have an ice cream from that
behaviour together with one’s awareness of feeling hungry, because, again,
that would still be compatible with many different explanations. Therefore,
in addition, one seems to need the belief that there is an ice cream in the
fridge and that the ice cream is what will satisfy one’s appetite, or that one
feels like having the ice cream that one believes to be in the fridge. Still, in
this case, one will have to have knowledge of feeling like having exactly the
ice cream, which comes very close to already knowing one’s desire to have
it. For either the raw feeling is not phenomenologically distinguishable
from other ones and we need to conceptualise it as that very kind of feeling,

9
Or, at any rate, what goes by the name of the Cartesian model, whether or not it was actually held
by Descartes.
10
So I wholly agree with Moran (2001), who refers to the talk in terms of “inner vision” as a “mislead-
ing metaphor” (p. 13) and writes (p. 14): “While ‘representationalism’ is a controversial thesis about
the ordinary perception of objects in the world, on nobody’s view is the awareness of one’s headache
mediated by an appearance of the headache. And in the case of attitudes like belief, there is simply
nothing quasi-experiential in the offing to begin with. There is nothing it is like to have the belief that
Wagner died happy or to be introspectively aware that this is one’s belief, and that difference does not
sit well with the perceptual analogy”. Of course, Gopnik might recast her point by saying that just
like a seasoned scientist knows immediately—that is, without inference—that there in electron mov-
ing about, so does a subject know her own propositional attitudes. Yet this is not her official position,
and, even if it were, the following criticisms developed in the main text would apply.
 4 Epistemically Robust Accounts 87

or it is phenomenologically distinct but then, in order to play a role in the

inference, it must nevertheless be conceptualised for what it is11. After all,
raw feelings cannot, as such, enter inferences. But judging that one feels like
having an ice cream is dangerously close to judging that one desires to have
it.  Hence, Gopnik’s model is unsatisfactory, from a philosophical point of
view, either because it falls into a crude form of behaviourism or because it
risks presupposing knowledge of the very propositional attitudes it sets out
to explain or of something very close to them.
Cassam has recently defended a similar position, even though, surpris-
ingly, he does not pay much attention to Gopnik’s pioneering work. He
claims that circularity is fine as long as the circle is not too tight.12 But,
as just shown, the circle seems very tight indeed, for it requires knowing
one’s own feeling like having exactly the ice cream, which is dangerously
close to knowing one’s desire for it. Now, Cassam claims that even this
piece of self-knowledge, of what he would regard as an “internal prompt-
ing”, is inferential because it is interpretative. That is to say, raw feelings
do not manifest themselves as feelings of wanting an ice cream; they do
so only because we interpret them as such because of other aspects of
our self-knowledge and other contextual factors. But clearly we do not
seem to go through any reasoning when we are assailed by a sudden feel-
ing of having an ice cream. Surely, the feeling itself can be prompted by
contextual aspects of the situation: it may be hot, I may be passing by a
wonderful ice-cream shop and be assailed by such a sudden urge. Yet I
am assailed by it, I do not reason myself into it, and, having the requisite
concepts, I can immediately recognise it (some theorists would rather
say “voice it”)  as a feeling of wanting an ice cream. Of course, as we
shall see in the following, having these concepts may require inferential
competences. However, their application, in the relevant case, does not
seem any more inferential than the application of the concept red (even
if one held that that concept is inferentially individuated and
that, in order to possess it, subjects should be able to engage in
the relevant inferences), when one sees a red object in front of one.
Again, one may appeal to the fact that at a subpersonal level all cognitive

11
This view does not seem to me very plausible, though. For it seems difficult to distinguish feeeling
like having an ice cream from feeling like having a sorbet or an ice lolly merely on the basis of
phenomenological features.
12
See Cassam 2014, p. 169.
88 A. Coliva

abilities, including perception and concepts’ application, are inferential.13

Still, that would be neither here nor there with respect to disputes which
normally focus on the presence (or absence) of inference at an epistemo-
logical or at a personal psychological level. That is to say, in these debates,
the relavant inferences are either those that a subject consciously deploys
(or is able to deploy upon request) to arrive at a certain conclusion. Or
else, they are those inferences theorists would offer as part of a rational
reconstruction of the epistemic grounds on which the eventual judgement
rests. To emphasise, they are not the inferences which our best developed
cognitive science might posit at a subpersonal level, as part of the causal
process which allows subjects to perform certain cognitive tasks.
So, Gopnik’s theory seems to require personal inferences, which
involve a lot of complex cognitive abilities and risk presupposing knowl-
edge of the very same mental states whose self-knowledge the model
should account for. By contrast, Cassam’s theory qualifies as inferential,
also when self-knowledge of our inner promptings is concerned, only by
presupposing the existence of inferences at the subpersonal level, where
this has no bearing on the epistemological issue of whether the relevant
self-ascriptions are grounded in inferences. Furthermore, if it is agreed
that the model cannot provide a satisfactory inferentialist story regarding
self-knowledge of inner promptings, it will have to be supplemented by
other accounts of how we may have knowledge of mental states other
than standing propositional attitudes.14 Of course, this is no criticism
from a pluralist point of view, but it is certainly an unwelcome result
given Cassam’s monist ambitions—that is, his ambition of providing a
general account of how we know our own mental states.

3 Simulation Theories: Goldman

and Gordon
If the theory-theory tends to equate knowledge of other people’s and of
one’s own mental states by seeing the latter as an instance of the former,
simulation theories do the opposite. Knowing other minds is a matter of

13
See Cassam 2014, pp. 130, 138.
14
That is to say of non-occurrent propositional attitudes, a subject may nevertheless be aware of.
 4 Epistemically Robust Accounts 89

simulating them and hence of issuing judgements about someone else’s

mental states by projecting oneself into other people’s shoes to gain a
first-person perspective onto their minds.
Various brands of simulation theories have been proposed over the
years, and the main variants are due to Goldman and to Gordon. Jane
Heal, another supporter of simulation over theory-theory, has lately come
to a mixed view. However, we are not here interested in what simulation
theories have to say about knowledge of other people’s minds, but in
what they can offer in the way of understanding knowledge of ourselves
in hypothetical situations and regarding the nature of psychological con-
cepts, particularly concepts of propositional attitudes.
Concerning our knowledge of ourselves, Gordon is not primarily
interested in psychological self-ascriptions, but in our knowledge of our
immediate intentions like “I shall now write a letter”. In his view, we are
very successful at that because these intentions are the product of practi-
cal reasoning and these pronouncements can express the result of one’s
deliberations based on various factors (like remembering being asked by a
good student to write him a reference and wanting to act as a thoughtful
supervisor). The interesting issue, in his view, is that we can similarly make
predictions about our own behaviour in hypothetical situations or in situa-
tions temporally distant from the one we are in at present. When we make
these predictions, we simulate being in the future or hypothetical situa-
tion and, by re-centring ourselves in it—that is, in its salient aspects—, we
make a decision about what to do. Hence, if asked “What would you do
if you heard footsteps from the basement?”, I could answer “I shall call the
police”, following the simulative procedure just described.15
According to Gordon, one could not explain the reliability we have in
making these predictions if they were the result of applying a theory to
oneself, based on bridging laws between one’s beliefs and desires. For “I
don’t know enough about my beliefs and desires, and the laws would at
best yield only the typical effects of those states”.16 But, as he points out,

in real life we sometimes surprise ourselves with atypical responses (…).

Practical simulation imitates real life in this respect, giving us the capacity

15
See Gordon 1995, p. 62.
16
Gordon 1995, p. 63.
90 A. Coliva

to surprise ourselves before we confront the actual situation. If I pretend

realistically that there is an intruder in the house, I might find myself sur-
prisingly brave—or cowardly.17

Here, we are not yet being given an explanation of self-knowledge—

that is, of our ability to ascribe mental states to ourselves in hypotheti-
cal situations. Yet the simulating procedure just described gives us an
important clue about how we could have knowledge of our intentional
behaviour in hypothetical situations. It does not require knowledge of a
theory and it allows for the fact that, as human beings, we can be and
often are unpredictable and thus cannot be described in a law-like man-
ner. It remains to be seen what simulation theories have to say regarding
the possession of psychological concepts, which are necessary in order
to be able to make psychological self-ascriptions and therefore acquire
knowledge of our own minds.
It is in this respect that Goldman and Gordon are in stark disagree-
ment. According to Goldman, we learn to apply psychological concepts
in our own case, first, by learning introspectively to identify the men-
tal states we are in. Then we use this knowledge to simulate others and
extend the application of these concepts from our own case to the case
of other people. In fact, for Goldman, simulation requires knowledge of
one’s own mental states, which is then used to pass judgement on other
people’s mental states, after simulating them. This crudely introspection-
ist model inherits all the problems we have reviewed in Chap. 3 (§1) and,
in case it were developed along more materialist lines, the ones we have
raised against Armstrong’s account (this chapter, §1).
Gordon objects to the appeal to introspection as well as to the argu-
ment by analogy used by Goldman to move from first-personal to third-
personal applications of psychological concepts. He also objects to the fact
that simulation requires knowledge of our own mental states. Focusing
on concepts, if Gordon is right in rejecting introspectionism,18 this seems
prima facie to leave us with the alternative view, favoured by supporters of

17
Ibid.
18
Gordon does not expound his criticism of introspection but merely says that he does not find it
very plausible (see Gordon 1995, p. 59). Anyway, in Chap. 3 and in §1 of the present chapter, we
have already considered several objections to it.
 4 Epistemically Robust Accounts 91

the theory-theory, according to which concepts are inferentially individu-

ated. Hence, to possess the concept of belief, say, requires knowing the
following inferences, which constitute a bit of theory regarding beliefs:

1) If S believes that P, P may not be the case;

2) If P is the case, S may not believe it;
3) Ceteris paribus, if S believes that P and desires that Q, S will do φ;
4) If S does φ and desires Q, S will presumably believe that P;
5) If S believes that P, R may believe that not-P or not believe that P.19

There is, moreover, a continuity between the application of the con-

cept in the first- and in the third-personal case. The inferential account,
coupled with the theory-theory, would nicely explain that. For the same
kind of inferential knowledge is required in order to possess and apply
the concept of belief, let S be oneself or someone else.
According to Gordon, however, introspectionism and inferentialism
do not exhaust the alternatives regarding the nature and acquisition of
propositional attitude concepts. For, in his view and in a Wittgensteinian
expressivist spirit, we teach children to replace less sophisticated forms of
linguistic behaviour, by means of which they usually express their desires
for a banana, say, with pieces of more sophisticated linguistic behaviour
that attributes wants to them. As he puts it: “We commonly train children
to preface nouns like ‘banana’ … with ‘[I] want …’ under the appropriate
circumstances: for example, when they look longingly at a banana”.20 In
the case of belief, in contrast, we get knowledge of our beliefs not by intro-
specting but by putting into practice the following “ascent routine”.21 We
consider evidence for or against P and then preface P (or not-P) with
“I believe that”, or by saying “I do not believe either P or not-P˝, when
there is not enough evidence in favour of either.22 In simulating others, we
just project ourselves in their situation and issue first-person judgements

19
Of course, one may dispute that such a list captures the constitutive inferences of our concept of
belief or even that the right account of it is inferential. These possible objections are not relevant to
present purposes.
20
Gordon 1995, p. 59. We will come back to this procedure in Chap. 6.
21
Gordon 1995, p. 60.
22
We will come back to this procedure in Chap. 5, §3.
92 A. Coliva

regarding wants and beliefs following exactly these procedures. Indeed, in

Gordon’s view, when total projection takes place, even the use of ‘I’ is such
as to refer to the simulated subject and not to oneself.
Now, the important and perplexing thing to note is that, according
to Gordon, neither procedure actually requires or constitutes the posses-
sion of the relevant psychological concepts. For, in his view, psychological
utterances are just pieces of linguistic behaviour, involving no real posses-
sion of the relevant concepts. In the case of wanting or desiring, he writes
“such training obviously won’t give children mastery of the concept of
wanting or desiring, and it won’t even teach them that ‘I’ refers to the
speaker”.23 And in the case of belief, he insists that “such training would
not give children mastery of the concept of belief ”,24 because “merely to
train children to preface assertions with the formula, ‘I believe’, would
not enable them to distinguish self-ascription of belief from assertion
of fact”.25 He then concludes that children “are just parroting a formula
before saying what they really mean to say, namely that it’s raining”.26
As Goldman rightly points out, if these routines are not sufficient
to give children the relevant psychological concepts, then it is not clear
how they do acquire them.27 Introspectionism and inferentialism seem
to remain the only viable alternatives. Gordon, however, responds that:

such training, although not sufficient for the acquisition of the concept,
may be the most important step in the acquisition of the concept. (To be
specific, it links the name of a propositional attitude with the verbal expres-
sion of that attitude; and the child who can also expressively use ‘I want’
when speaking for another (that is, in the context of a simulation) comes
close, in my view, to have grasped the general concept of wanting.)28

In more recent writings, Gordon states:

such utterances, made without ascriptive intent, are in fact the way young
children first employ the linguistic forms of propositional attitude ascrip-
23
Gordon 1995, p. 59.
24
Gordon 1995, p. 61.
25
Ibid.
26
Ibid.
27
See Goldman 1993.
28
Gordon 1995, p. 65, fn. 7.
 4 Epistemically Robust Accounts 93

tion. For example, children first use “I want” merely as a way of getting
what they want. Yet the resulting utterance, were it construed as an ascrip-
tion, would generally be a true one. Thus, children are linguistically boot-
strapped from expression to reliable self-ascription. The same should hold
whether one is expressing one's own propositional attitude or vicariously
expressing another’s.29

Thus, the idea seems to be that first we learn to use “I want” and
“I believe” merely as alternative expressions of our wants and convic-
tions, that we may have expressed without introducing those expressions.
Then, by simulating others and by using the same routines we use in our
own case, we would get the semantic continuity between first-person and
third-person uses of those expressions.
Yet, if that is what Gordon has in mind, it seems too scant to give
us full-blown possession of the relevant concepts. For, by Gordon’s very
lights, the latter requires grasp of the distinction between facts and their
representation by subjects, when the two can dramatically come apart.
Furthermore, given Gordon’s specific conception of simulation as total
projection, where even the reference of ‘I’ is shifted from the simulator
to the simulated subject, possibly occurring at a stage when the former is
not even supposed to grasp that ‘I’ normally refers to oneself, simulation
would still be an exercise in first-personal, rather than in third-personal,
understanding. For I would be “attributing” the relevant mental states
to myself, even though my own self would have been totally projected
onto someone else. Therefore, Gordon does not seem to offer us a sound
explanation of the possession of propositional attitudes’ concepts, for the
central inferences which individuate the concept of belief in particular—
that is, (1) and (2) —and first-personal/third-personal semantic continu-
ities do not seem to be accounted for. Failing that, he does not propose
a viable account of self-knowledge, for which possession of the relevant
psychological concepts is necessary.30 It remains that simulation can be
an important tool, more effective and cognitively parsimonious than the
application of a theory to oneself, in order to form predictions about

29
Gordon 2009.
30
As we shall see in Chap. 6 (§2), it is possible to offer an essentially expressivist story as a viable
explanation of the possession of at least a rudimentary concept of belief, but further elements will
have to be built into the picture, besides the ones Gordon allows for.
94 A. Coliva

one’s own behaviour and, granted the possession of the relevant psycho-
logical concepts, a source of knowledge of our own minds in hypothetical
or future situations.
As we shall see in further chapters, however, Gordon’s views, as indebted
to Wittgenstein and Evans as they are, can be built upon to give a satisfac-
tory account of the possession of the relevant psychological concepts. So,
let us consider the epistemic characteristics of this model, assuming for
the sake of argument that it could be complemented by a sound account
of psychological concepts’ possession, alternative to the one which would
fit better with theory-theory accounts.
The account would respect groundlessness at least in its weak form,
for it would be anti-introspectionist and anti-inferentialist. It would
also respect transparency, given the ascent routine our self-ascriptions
would be based on and given the expressivist account of self-ascriptions
of non-intentional mental states provided. However, authority would
be impaired. Surely, simulation would give us more reliable predictions
about our future or hypothetical mental states than the application of a
theory. But they would still be predictions, which remain clearly fallible.
Notice, however, that that would not impair one’s rationality or concep-
tual abilities. A subject who got predictions wrong about her future or
hypothetical states of mind would not necessarily be irrational, nor would
she display lack of the relevant psychological concepts. She may just be
poor at performing the task, by being oblivious to some relevant aspects
of the simulated situation. Or else, in real life, she could actually react
differently from how she thought she would. That is enough to show that
the kind of simulation Gordon has in mind is, after all, a source of third-
personal self-knowledge, rather than of first-personal self-knowledge,
despite its inheritance of some features that characterise the latter.
In this connection, it is worth pointing out that also the theory-theory
could offer important elements to understand the way in which, at least
sometimes, we gain knowledge of our past and dispositional mental states,
in particular. For instance, one could reflect on one’s behaviour on a past
occasion and interpret the data by means of certain psychological “laws”
and therefore attribute to oneself this or that propositional attitude, emo-
tion or character trait, as a disposition. If the theory were in fact a piece
of Freudian theory, the very same procedure could yield knowledge of
 4 Epistemically Robust Accounts 95

one’s unconscious mental states. Clearly, one can be more or less reliable
in doing so and versed in this practice. Yet, obviously, one would by all
means get third-personal knowledge of one’s own mental states. Hence,
both theory-theory and simulation theories can offer important elements
to understand third-personal self-knowledge while failing, for different
reasons, to account for distinctively first-personal self-knowledge.

4 Summary
In this chapter, we have considered epistemically robust accounts of self-
knowledge. We have started with recent inner sense theories and we have
examined Armstrong’s proposal in particular. In his view, self-knowledge
is the product of a reliable cognitive mechanism that tracks first-order
propositional attitudes and produces the corresponding  second-order
ones. The mechanism is physically realised in our brains and is opera-
tive at the subpersonal level. We have claimed that although the model
accounts for groundlessness, transparency and authority, it severs the
connection between self-knowledge and rationality and concepts’ posses-
sion. Hence, it is deemed unsatisfactory.
We have then moved on to Gopnik’s theory-theory. According to this
model, we know our minds in the same way we know others’—that is,
through inference to the best explanation, based on observation of one’s
overt behaviour and non-intentional mental states and through the appli-
cation of a real theory of the mind subjects acquire around the age of four.
We have taken issue with the model from a merely philosophical point
view, even though it has been variously criticised also from a psychologi-
cal point of view. Moreover, we have done so by granting, albeit merely
for the sake of argument, that it could somehow satisfy or at least assuage
the requirements for distinctively first-personal self-knowledge—that is,
groundlessness, transparency and authority. We have therefore claimed
that it is inadequate from a philosophical point of view because it either
collapses into a crude form of behaviourism or presupposes the very self-
knowledge it aims to explain or something too close to it to avoid the
charge of vicious circularity. We have raised similar worries with respect
to Cassam’s recent defence of inferentialism as well as cast doubts on the
96 A. Coliva

idea that knowledge of the “internal promptings”, which would consti-

tute one’s evidence for the relevant inferences, could really be inferential.
Finally, we have turned to simulation theories, which have been pro-
posed mainly by Goldman and Gordon. They are first and foremost theo-
ries about our knowledge of other people’s minds. Accordingly, in order
to know them, we need to simulate other subjects, rather than apply
a theory to them, given the observation of their behaviour. Simulation
theorists, moreover, argue that it is also through simulation that we
gain knowledge of our hypothetical and future mental states. They dif-
fer, however, in their account of the workings of simulation and in their
account of psychological concepts’ possession. Goldman requires subjects
to have knowledge of their own mental states, in order to be able to simu-
late others, and thinks that such knowledge is achieved through intro-
spection, by recognising and naming one’s ongoing phenomenologically
salient mental states. Gordon holds a radically different and cognitively
more parsimonious view, according to which there is no need to have
self-knowledge to simulate others. Goldman’s theory is quickly dismissed
on the grounds that it inherits all problems that beset introspectionist
views regarding self-knowledge. Gordon’s version is therefore deemed
more promising, even if, in the end, it is argued that it fails to account
for the possession of the relevant psychological concepts. From an epis-
temic point of view, however, it is interesting because, while inheriting
some aspects of first-personal self-knowledge, it actually qualifies as a way
of gaining third-personal knowledge of our own future or hypothetical
mental states. Similarly, we have argued, theory-theories might be useful
to account for certain cases of third-personal self-knowledge—especially
when knowledge of our past or dispositional mental states is at stake—
while failing as accounts of first-personal self-knowledge.

Bibliography
Armstrong, D. (1968). A materialist theory of the mind. London/New York:
Routledge.
Cassam, Q. (2014). Self-knowledge for humans. Oxford: Oxford University Press.
Dokic, J., & Proust, J. (Eds.). (2002). Simulation and knowledge of action.
Amsterdam/Philadelphia: John Benjamins Publishing Company.
 4 Epistemically Robust Accounts 97

Goldman, A. (1993). The psychology of folk psychology. Behavioral and Brain

Sciences, 16, 15–28.
Gopnik, A. (1983). How we know our minds: The illusion of first-person
knowledge of intentionality. Behavioral and Brain Sciences, 16, 1–15.
Reprinted in Goldman, A. (ed) (1993). Readings in philosophy and cognitive
science. Cambridge MA: MIT Press.
Gordon, R. (1995). Simulation without introspection or inference from me to
you, in Davies M. and Stone T. (1995b) (eds), 53–67.
Gordon, R. (2009). Folk psychology and mental simulation, The Stanford ency-
clopedia of philosophy (Fall 2009 Edition), Edward N.  Zalta (Ed.), URL =
http://plato.stanford.edu/archives/fall2009/entries/folkpsych-simulation/
Lycan, W. (1996). Consciousness and experience. Cambridge MA: MIT Press.
Moran, R. (2001). Authority and estrangement. Princeton: Princeton University
Press.
 5
Epistemically Weak Accounts

In this chapter, we look at still epistemic accounts of self-knowledge,

which, however, are explicitly meant to avoid any form of robust cog-
nitive achievement on a subject’s part. They therefore devise ways of
trying to rescue the idea that first-personal self-knowledge is a form of
knowledge, which differs, however, from observational and inferential
knowledge. Yet they are united in dispensing with the idea that it may
be achieved through the operation of a scanning cognitive mechanism or
through inference to the best explanation starting with the observation of
one’s own behaviour or further non-intentional mental states.
The first proposal we consider is Christopher Peacocke’s rational inter-
nalism (§1). The distinctive aspect of this view is that the conscious occur-
rence of first-order mental states is meant to ground one’s psychological
self-ascriptions, and to rationalise them, without thereby requiring a sub-
ject to have observational or inferential knowledge of them. We put pres-
sure on the idea that there is a notion of what it means for a first-order
mental state to be conscious, which can play the role Peacocke’s theory
requires it to play. We also raise the worry that the account could fulfil
such a task only by reverting to some form of externalism. We then move

© The Editor(s) (if applicable) and The Author(s) 2016 99

A. Coliva, The Varieties of Self-Knowledge,
DOI 10.1057/978-1-137-32613-3_5
100 A. Coliva

on to Tyler Burge’s rational externalism and we raise several qualms with

respect to it, especially about the notion of entitlement it relies on (§2).
Then we turn to a different kind of proposal which gives centre stage
to Gareth Evans’s so-called “transparency method” (§3). According to that
method, in order to get knowledge of our own beliefs, we should not
direct our eyes inward, as it were, and try to introspect our own minds to
see whether we can find certain mental states which we would then self-
ascribe. Rather, we have to direct our gaze outward, to determine whether
it is the case that P. If it is, then we can simply prefix P with “I believe
that”, thereby making the relevant self-ascription. We look at two promi-
nent ways of developing Evans’s method: one, more epistemically oriented,
recently put forward by Jordi Fernández (§3.1), and another one, more
deliberative-oriented maintained by Richard Moran (§3.2). The result is
a kind of dilemma for Evans-inspired accounts. Either they go down the
epistemic route but then they are defective for the grounds for one’s psy-
chological self-ascriptions do not seem to be the same as the grounds for
believing P or they go down the deliberative route but they do not seem to
have anything to offer with respect to how the relevant psychological self-
ascriptions can amount to knowledge, contrary to their advertised aim.

1 Peacocke’s Rational Internalism1

According to Peacocke, “conscious thoughts and current attitudes (…)
can give (…) the thinker a reason for self-ascribing an attitude to the
content which occurs to the thinker, provided our thinker is conceptu-
ally equipped to make the self-ascription”.2 To illustrate: suppose you
have (1) an apparent memory that Italy was a monarchy before World
War II. That, according to Peacocke, gives you a (non-conclusive) reason
for (2) judging that Italy was a monarchy before World War II, which,
in turn, gives you a reason for (3) self-ascribing the belief that Italy was
a monarchy before World War II (provided you possess the first-person
1
The label is mine but in keeping with Peacocke’s pronouncement of being interested in defending
a  form of  rationalism; see Peacocke 2002, 2003. The  earlier account of  self-knowledge, offered
in Peacocke 1999, is indeed in line with Peacocke’s later work.
2
Peacocke 1999, p. 214.
 5 Epistemically Weak Accounts 101

concept, the concept of belief and those concepts that are necessary for
the specification of the content of your belief ). In Peacocke’s view, when
the self-ascription is formed in this way, it amounts to knowledge because
the second-order belief would be true—in virtue of the conceptual truth
that if one judges that p, one believes it—and justified by the correspond-
ing first-order mental state (namely, the judgement that P).3
In Peacocke’s view, this proposal avoids inferentialism because “to say
that (2) is the thinker’s reason for making the judgement in (3) is not to
say that he infers the self-ascription from a premise that he has made such
a first-order judgement”, together with the bridging principle that if one
judges that P, one believes it.4 Rather, the transition is “blind”—that is,
the subject makes a judgement that P and then self-ascribes the belief,
without inferring to it. Yet the transition counts as rational because it
can be rationalised by appealing to two aspects of it: the fact that the
first-order mental state is a conscious one and the fact that it conforms to
an a priori principle, which links judgement and belief. Peacocke’s pro-
posal also avoids observationalism—both in classic Cartesian and more
modern reliabilist versions—because to say that a first-order mental state
is conscious does not mean, according to Peacocke, that it is an object
of (quasi-perceptual) attention, but, rather, that it occupies a subject’s
attention: being in a certain intentional state contributes to a subject’s
phenomenology—to what things are like from his subjective point of
view—despite the fact that the state is not presented to him as an object.5
Furthermore, contrary to Armstrong’s reliabilist account, on Peacocke’s
view, a first-order conscious mental state is taken to be a subject’s own
reason for making the corresponding self-ascription and not merely its
cause.6 Finally, Peacocke’s model is not a version of constitutivism, because
he denies both the constitutivist’s ontological claim that first-order men-
tal states depend for their existence on the corresponding self-ascriptions
and the constitutivist’s anti-epistemic claim that subjects’ psychological

3
According to Peacocke, a judgement that p may not always occur, in fact, but it should always be
available to a subject, in order for his psychological self-ascription to be justified (cf. Peacocke
1999, pp. 222–223; 241–242).
4
Peacocke 1999, p. 214.
5
Peacocke 1999, pp. 205–209.
6
Peacocke 1999, pp. 224–225.
102 A. Coliva

self-ascriptions are not justified by their first-order mental states.7 In par-

ticular, for Peacocke, while it is a conceptual truth that if one judges that
p, one believes it and that it is part of the possession-conditions of the
concept of belief that one be disposed to judge that one believes that p if
one judges that p, this does not at all preclude the possibility that first-
order conscious mental states have independent existence of their self-
ascription nor, especially, that they can function as a subject’s own reasons
for the corresponding self-ascriptions.8 Hence, Peacocke’s account allows
for one’s psychological self-ascriptions to be grounded in conscious first-
order mental states. Yet it avoids both observationalism and inferential-
ism. Therefore, groundlessness, in his view, holds only in its weak form
(see Chap. 3, §1).
It is important to emphasise here that Peacocke’s proposal is offered as
a form of epistemologically internalist model: reasons are conscious states
of a thinker, which, moreover, are open to first-personal scrutiny. After
all, at least in some cases, a subject might realise that he has mistaken
either the nature or the content of his first-order mental states, as cases
of wishful thinking and of self-deception seem to show, and has thereby
been prompted to make an erroneous self-ascription. This is, of course,
compatible with the idea that conceptually endowed subjects are usually
authoritative with respect to their own mental states and that massive
error with respect to one’s ongoing mental states would impair a subject’s
rationality. So authority holds for him but only in its weak version (see
Chap. 3, §1).
What is precluded by Peacocke’s model, however, is arrival at the self-
ascription on the basis of self-conscious consideration of one’s own first-order
conscious mental states. For that would place the very self-knowledge to
be accounted for at the foundation of a purported account of how such
knowledge is grounded. What Peacocke requires is that first-order mental
states can, without being explicitly judged and thereby made the content
of a second-order state, somehow be salient to consciousness and taken
at face value in the formation of the relevant self-ascriptions and subse-
quently can be offered as reasons for them and, on occasion, scrutinised

7
Peacocke 1999, pp. 230–231.
8
Peacocke 1999, pp. 232–233.
 5 Epistemically Weak Accounts 103

and assessed. On his account, transparency in its strong form would hold
as well (see Chap. 3, §1). For these transitions from the mere awareness of
a given mental state to their self-ascription would be immediately and non-
inferentially made by conceptually endowed, attentive and lucid subjects.
Peacocke’s proposal thus relies crucially on the claim that first-order
mental states may be given to a subject in such a way as to function as rea-
sons, from his own point of view, for the corresponding self-ascriptions.
Hence, in order for the proposal to be implemented satisfactorily, he
needs a notion of a conscious mental state that can support the claim that
the corresponding self-ascription would be rationally justified merely by
its occurrence.
Concerning what it is for a mental state to be conscious, however, he
offers two, not obviously equivalent, proposals. One is that a mental state
is conscious if there is something it is like to have it.9 This suggestion
amounts to the idea that a mental state is conscious if there is something
subjectively distinctive about being in it. The other proposal is that a
mental state is conscious if it occupies our attention without being an
object of attention.10 This suggestion, in contrast, amounts to the idea
that a mental state is conscious if it is such that it can occupy our atten-
tion in the following sense: it is something whose nature, content, pre-
suppositions and theoretical and practical consequences we can attend to
without turning the state itself into an object of attention.11

9
Peacocke 1999, pp. 205–206.
10
Peacocke 1999, p. 207.
11
For instance, Peacocke (1999, pp. 209–210) writes: “I now attempt some further analysis of the
occupation of the attention by conscious thought. When you have a thought, it does not normally
come neat, unconnected with other thoughts and contents. Rather, in having a particular thought,
you often appreciate certain of its relations to other thoughts and contents. You have a thought, and you
may be aware that its content is a consequence, perhaps gratifying, perhaps alarming, of another conclu-
sion you have just reached; or you may be aware that its content is evidence for some hypothesis that you
have formulated; or that it is a counterexample to the hypothesis” (italics mine). “Now when you think
a particular thought, there is of course no intention in advance to think that particular thought.
But there can be an intention to think a thought which stands in a certain relation to other thoughts
or contents” (italics mine). This clearly suggests the idea that if a thought occupies one’s attention,
then, according to Peacocke, its content must be known to the subject who has it, as well as its
practical consequences—like causing one’s feeling alarmed or gratified—and theoretical ones—
such as its relation to other thoughts. As we shall presently see, however, this requires for Peacocke
too knowledge of the nature of one’s own mental states as well—that is, of their being beliefs as
opposed to wishes, imaginings, and so on.
104 A. Coliva

The second suggestion seems highly problematical. For, presumably,

one can attend to the nature, content, presuppositions and consequences
of one’s mental states only if one is aware of what kind of mental states
they are (judgements, desires, wishes, and so on), of what content they
have, and of how they can be related, as such, to one’s other mental states,
with their own respective contents. After all, one can attend to the pre-
suppositions and consequences of, for example, one’s judgement that Italy
was a monarchy before World War II, only if one is aware12 of the fact that
it is indeed a judgement and not just a supposition or a wish or a suspicion
that that was the case.13 Moreover, Peacocke maintains that if a mental
state is occupying one’s attention, one will probably react to it—feeling
gratified or alarmed and so on. However, one can feel alarmed or gratified
by a certain thought occurring to him only if one is aware of the kind of
thought it is. For instance, in the normal run of cases, I can feel gratified
by the belief, or even the certainty, that I will soon have a promotion, but
not by the corresponding desire or fear. (Of course, we may complicate
the phenomenology and suppose, for instance, that I might feel gratified
also by desiring such a coming event, if that showed, say, that I am recov-
ering from a state of dangerous apathy with respect to my work and my
career. Still, this would show that for one affectively to react to one’s men-
tal states, one should (at least on occasion) be aware also of their relations.
Yet this in turn would be possible only if one were aware of the kind of
mental states one has as well as of their content.) Thus, the evident prob-
lem with Peacocke’s suggestion is that the awareness of the kind and con-
tent of one’s own first-order mental state required if it is to occupy one’s
attention and inform further attention to its presuppositions and conse-
quences seems—on a quite natural reading of Peacocke’s own account of
it—to amount to nothing less than knowledge of that very mental state of
which, on his view, it is supposed to provide a rational basis! At the least,
the gap between them seems to be vanishingly small. Yet, unless it can

12
Actually, Peacocke talks in terms of “rational sensitivity” to one’s mental states. One way of
understanding this (rather vague) expression is in terms of “awareness”. However, for a different—
still problematical—interpretation, see the following paragraph.
13
Peacocke himself (1999, p. 216) makes the point: “Now the thinker who successfully reaches new
beliefs by inference has to be sensitive not only to the content of his initial beliefs. He has also to
be sensitive to the fact that his initial states are beliefs. He will not be forming beliefs by inference
from the contents of his desires, hopes, or daydreams”.
 5 Epistemically Weak Accounts 105

be widened, Peacocke’s account of self-knowledge would bluntly presup-

pose exactly what it should explain—and, indeed, not because some other
piece of self-knowledge would be presupposed, but, rather, because the
very instance of self-knowledge to be accounted for would be.14
One then may try to propose different glosses on the notion of a men-
tal state’s occupying one’s attention.15 According to a first, alternative
reading of it, a given mental state could occupy one’s attention if one
were attentively engaging in it, without thereby considering it as the par-
ticular mental state it is (say, a judgement) with the particular content it
has (that Italy was a monarchy before World War II, for example), pretty
much as one could be performing attentively a certain activity (like run-
ning) without considering it qua the particular activity it is. This gloss
on the notion of the occupation of the attention would not presuppose
explicit awareness of the very mental state whose knowledge is supposed
to be accounted for. Nonetheless, it could not be of avail to Peacocke
because his model requires that one’s first-order conscious mental states
be a subject’s own reasons for his corresponding self-ascriptions. However,
for psychological self-ascriptions to be rational responses to one’s own
occurrent conscious mental states, one must actually be aware of them
qua the particular mental states they are, with the contents they have.
To see this, consider the analogy with running: as much as one could
run attentively, without considering it qua the activity it is, if one were
to make a rational judgement about what one is attentively doing, one’s
action should be manifest to him as the particular action it is. Merely
engaging in it, as attentively as it might be, would not constitute a reason
for the corresponding judgement. Similarly, in the case of mental states,
their rational self-ascription seems to require switching—however shortly
that might be—from being attentively, yet non-reflectively, engaging in
them, to considering them qua the mental states they are, with the con-
tents they have.
14
Peacocke could reply that what is required is knowledge of one’s judgement but that what gets
self-attributed is a belief. Yet, first, a judgement is nothing but a mental state (or action). Hence,
knowledge of it would be a case of self-knowledge, which we are supposed to account for. Secondly,
if we applied Peacocke’s model to explain how this latter piece of knowledge could be possible in
turn, we would get, on the present proposal, that this very mental state should already be known to
a subject who self-ascribed it.
15
I think the following three paragraphs can take care of the kind of objections raised by McHugh
(2012) against my criticism of Peacocke’s proposal.
106 A. Coliva

Another possible way of trying to rescue Peacocke’s present proposal

would consist in maintaining that the kind of knowledge of one’s first-
order mental state that the occupation of the attention requires is merely
tacit or implicit or perhaps practical.16 One then might suggest that a sub-
ject may have the ability to have certain thoughts, act on them and reason
on their basis, thus implicitly knowing what kind of mental states they
are, without yet having the necessary conceptual repertoire to self-ascribe
them. This, however, is what implicit self-knowledge amounts to. Here a
problem arises: how is this implicit knowledge achieved? An answer to this
question is fundamental if implicit self-knowledge has to be the backbone
of an explanation of self-knowledge in general. For, otherwise, Peacocke’s
account would explain, at most, only how one could come to self-ascribe
mental states that are already implicitly known. No suggestion of how
implicit self-knowledge is achieved can be found in Peacocke’s work, to the
best of my knowledge.17 Furthermore, even supposing that an account of
implicit self-knowledge were forthcoming, how would the conceptualisa-
tion of implicitly known first-order mental states proceed? If it depended
on a grasp of the conceptual role of first-personal propositional attitudes—
in the case of belief, for instance, on grasp of the rule “If I judge that p, then
I should be disposed to self-ascribe the corresponding belief ”—would not
that after all require explicit knowledge of one’s first-order mental states?
Finally, one could try to revise Peacocke’s own account and propose
less demanding readings of the notion of a mental state’s occupying one’s
own attention. For instance, following Bonjour (2003), one might claim
that there is something like a “constitutive awareness” of the kind and
content of one’s intentional mental state which does not coincide with
one’s judging that one has that mental state with such and such a con-
tent.18 As appealing as it might be, this view seems irremediably unclear:
if the awareness of the kind and content of one’s mental state is implicit,

16
I would like to thank Barry Smith for bringing this point to my attention. McHugh dubs it
“practical” in his 2012 paper. See previous footnote.
17
McHugh 2012 in fact seems to take it as a brute datum about certain kinds of creature.
18
Bonjour (2003, p. 62): “[Awareness of the kind and content of one’s own occurrent mental state
is] not in any way apperceptive or reflective in character: [it does] not require or involve a distinct
second-order mental act with the propositional content that I have the belief in question. Instead,
[it is] partly constitutive of the first-level state of occurrent belief ”.
 5 Epistemically Weak Accounts 107

then the problems just reviewed will re-emerge. If, in contrast, it is not,
it is difficult to see how it can avoid collapsing into full-blooded knowl-
edge of one’s first-order mental state. For it may be true that when one
is consciously enjoying a first-order intentional state one does not actu-
ally have to judge that one has that mental state with that content.19
However, if that knowledge has to be somehow explicit, it seems that it
should be constitutive of having a conscious first-order mental state that
one should at least have the disposition to make that kind of judgement.
Yet, in this latter case, Bonjour’s proposal would be a mere notational
variant of more traditional constitutive accounts of self-knowledge,20 a
fact that would betray its whole point. Finally, if such constitutive aware-
ness is neither implicit knowledge nor constituted by the disposition to
judge that one has a given first-order mental state and merely consists in
the fact that the kind and content of that state are part of its phenom-
enology, then the proposal will turn into a version of Peacocke’s first,
phenomenological account of what it means for an intentional mental
state to be conscious and will be open to the criticisms we shall presently
level against it.
Let us now turn to Peacocke’s other, phenomenal gloss on what it
means for a mental state to be conscious.21 Here, too, problems arise. The
first one is that the work the proposal assigns to the phenomenology of
first-order intentional states now restricts it to the case of self-ascriptions
of occurrent mental states and rules it out for standing attitudes that lack
any phenomenological impact. Secondly, even in occurrent cases, the
capacity of the phenomenology of first-order mental states to carry the
resulting epistemological load falls into question as soon as one reflects
that, for example, occurrent desires, hopes or wishes with the same con-
tent need hardly be different from a strictly phenomenological point of
view yet will be required somehow to rationalise different self-ascriptions.

19
According to Bonjour (2003, pp. 65–68) that would either involve us in a vicious regress or never
provide an account of how a given mental state (first-order or otherwise) could ever be conscious.
20
Such as Bilgrami 2006, for instance. In his view, it is a necessary condition for having (a certain
class of ) first-order propositional attitudes that one knows them. There will be more on these
accounts in Chap. 7.
21
Further criticism of this aspect of Peacocke’s proposal can be found in Heal 2012; see also
Peacocke’s response in Peacocke 2012.
108 A. Coliva

Thirdly, it is widely agreed,22 indeed by Peacocke himself,23 as it seems,

that, even in cases where occurrent intentional states do have a distinctive
phenomenology, it remains that rational and justificatory relations hold
only between states with representational content in the first place. The
idea is that the kind of phenomenology that, for instance, the judgement
that today is a sunny day can have—such as the pleasant and relaxing
feeling this thought may produce, together with, perhaps, a strong sense
of confidence—will not, by itself, give one a reason to self-ascribe the cor-
responding belief. Rather, it is only if the first-order mental state presents
itself to one as the type of mental state it is—that is, a belief as opposed to
a wish or a desire—with the content it has—that today is a sunny day, as
opposed to anything else—that it can constitute a thinker’s own reason
for the corresponding self-ascription. For, otherwise, why should one be
more justified in making that self-ascription than in self-ascribing the
corresponding hope, say? To see this, consider cases of wishful think-
ing: certain contents may manifest themselves to a subject with such an
intensity and “colouring”, as it were, that, though being merely hopes,
can actually be taken for beliefs. Thus, their phenomenology would be
pretty much the same, yet only the self-ascription of the relevant hope
would be rational on their basis (not just correct). If so, however, it is

22
This point is vigorously maintained, for instance, in McDowell 1994 and Brewer 1999. For
opposite views, see, for instance, Jim Pryor 2005.
23
Peacocke (1999, p. 216) writes: “In cases of consciously based self-ascription of attitudes and
experiences, a thinker […] makes a transition not only from the content of some initial state, but
also makes it because the initial state is of a certain kind (…). In the case of consciously based self-
ascription, the distinction between those events which are occurrent attitudes of the right kind to
sustain the resulting judgement and those which are not is a distinction which is conceptualised by
the thinker”. All this clearly implies that the relevant transitions are made on the basis of how the
first-order mental states are represented to the subject and not just on the basis of their phenome-
nology and this seems to ensure their rationality from the subject’s own point of view. Hence,
although, as Pryor has brought to my attention, Peacocke 2001 maintains that sensations—devoid
of any representational content—can immediately justify one’s corresponding self-ascriptions, he
does not seem to be inclined to offer an analogous account of intentional mental states and of their
self-ascriptions.
Furthermore, for reasons of internal coherence with his earlier work, I think Peacocke should
acknowledge that only representational contents—let them be psychological or otherwise—can
serve as rationalisers of judgements and, in particular, of self-ascriptions of intentional mental
states. Indeed, it is only on such an assumption that one can understand why, instead of defending
the so-called “Myth of the Given” against McDowell’s attacks, he elaborated a notion of non-con-
ceptual yet fully representational content for experiences (see Peacocke 1992, Chap. 4).
 5 Epistemically Weak Accounts 109

unclear how any purely phenomenological account of what it means for

a mental state to be conscious can support the claim that a first-order
state can stand in a rational, justificatory relation to the corresponding
self-ascription.24 To repeat, if the phenomenology can be pretty much the
same and no other element is required to account for the rationality of a
given self-ascription, why should we think that only the self-ascription of
the relevant hope, and not of the corresponding belief, would be rational?
To recap: on the attention-occupancy account of what it means for a
mental state to be conscious, Peacocke’s proposal that a first-order mental
state may provide a reason for the corresponding self-ascription does not
discernibly escape presupposing the very piece of (implicit or explicit)
self-knowledge we should be accounting for. On the purely phenomeno-
logical account, in contrast, there is a real danger of not being provided
with anything that could serve as a reason for the corresponding self-
ascription. Therefore, Peacocke’s account runs the risk either of failing
to offer an independent reason for the self-ascription (or of presupposing
implicit knowledge of the very item of self-knowledge to be accounted
for) or of offering no reason at all.
Peacocke (2003) presents some ideas which may seem suited to
evade the dilemma just reviewed. Key amongst them is the claim that
there can be unmediated, non-inferential transitions, viz. movements of
thought—in our case, they would be from conscious first-order mental
states to second-order ones25 that are nevertheless rational, because they
are truth-conducive and a priori. By this, Peacocke means that there can be
movements of thought which are truth-conducive not because—or just

24
Indeed, I think that it is very unclear how proponents of such a view could solve the problem just
mentioned and known as the ‘arbitrariness problem’ (see Pryor 2005, pp. 192–193). For instance,
Pryor’s own attempt to solve it by appealing to a notion of mental events which are themselves logi-
cally structured is dubious both from a metaphysical point of view, as Achille Varzi has remarked
to me, and from an epistemological one. For an internalist needs a justifier that is given to the
subject and that can play a rationalising role for his self-ascription from his own point of view. The
fact that an event might be, unbeknownst to him, logically structured and suited, in principle, to
rationalise the transition from its occurrence to its self-ascription, is of no use to the development
of a sound internalist epistemology of psychological self-ascriptions.
25
I need to emphasise that Peacocke 2003 does not explicitly consider the case of self-knowledge.
Rather, I am freely extending views he develops with respect to the relation between perceptual
experiences and empirical beliefs to the case of transitions from first-order mental states to second-
order ones.
110 A. Coliva

because—they are underwritten by reliable physical mechanisms, but

because they conform to certain a priori principles. An example would
be the principle that if one judges that p, one will believe that p. Thus,
the truth of “I believe that p” will be guaranteed by such a principle.
According to Peacocke, the thoughts in which such transitions culminate
can be non-inferentially justified merely by the very occurrence of the rel-
evant anterior states from which the transition proceeds.26 Applying the
idea to our present concerns: when one’s psychological self-ascription—
“I believe that p”—is formed by moving from one’s first-order mental
state—a judgement that p—to it directly (that is, without the mediation
of any other mental state), one will have a non-inferential justification—
an entitlement in Peacocke’s terminology—for the self-ascription.27 This
proposal promises relief from the consciousness dilemma by enabling
us to deflect its second horn: it would allow first-order mental states to
count as justifiers after all, even if their nature and content were no part
of their phenomenology, just so long as there are suitable a priori prin-
ciples, which a subject need not be aware of, which could rationalise
the transitions he makes to the corresponding second-order states. So,
although someone’s judgement that p need not present itself as such to
the subject who makes it, it could still rationalise a self-ascription of the
corresponding belief, since the two states may be connected by the a
priori principle that if one judges that p, one will believe that p. It is well
beyond the scope of this note to try to evaluate the general standing,
let alone the detail, of the notion of entitlement Peacocke explores in
the relevant parts of The Realm of Reason. My remaining point concerns
purely what it may have to offer, in the best case, to bolster his earlier
Rationalist account of self-knowledge against the problems I have been
describing. For this purpose, it is crucial to recall the internalist aspira-
tions of the Rationalist account: its goal was to provide space for, at once,
the rationality, from the first-person perspective, of the transitions involved
and the possibility that a first-order mental state could be a subject’s own

26
Strictly speaking, as Peacocke himself notices (2003, p. 26), the results of these transitions would
be only “relatively a priori” since they would be justified by the occurrence of particular mental
states. Still, they would not be inferred from them.
27
See Peacocke 2003, p. 11.
 5 Epistemically Weak Accounts 111

accessible reason for his corresponding self-ascription28 so that, if needed,

he could scrutinise and assess it.
However, under the aegis of Peacocke’s notion of entitlement, the ratio-
nality, from the first-person perspective, of the relevant transitions need
not be visible. Visibility would require the movement from a first-order
to a second-order mental state to be informed by an appreciation of the
rational connection between those states—that is, by an appreciation of
the fact that one’s self-ascription is rational because it is arrived at by mov-
ing from the appropriate kind of first-order state.29 To illustrate with our,
by now, standard example: one’s self-ascription of a belief that Italy was
a monarchy before World War II would be rational, from a first-person
point of view, only if a subject was aware of the fact that it is formed by
transition from the occurrence of one’s first-order mental state of judging
that that was the case. However, such an awareness would presuppose the
very knowledge (implicit or explicit, as we saw before) of one’s first-order
mental state the model was supposed to dispense with. Hence, there is
actually no clear sense in which the refurbished proposal could qualify
as internalist: although the rationalisers would be internal and even con-
scious states of a subject, their rationalising role would not—and, actu-
ally, could not—30 be manifest to him. Hence, they would not (and could
not) play the role of a subject’s own reason for his judgement.

2 Burge’s Rational Externalism

A proposal, which actually embraces what we may call “rational exter-
nalism”, is Burge’s.31 In his view, we should distinguish two types of
warrant. Justifications, on the one hand, require a subject to be able to
articulate them as her reasons in favour of a given belief she happens to
have. Justifications, therefore, are internalist kinds of warrant. You have a

28
See Peacocke 2003, pp. 12, 101, 177–178.
29
Notice that I am not claiming that the rationalising principles should be self-consciously or even
tacitly employed by a subject, even less that their truth should be appreciated by him in order to
have an account of self-knowledge which would be acceptable by internalist lights.
30
On pain of falling back into either horn of the consciousness dilemma.
31
See, in particular, Burge 1996 and 2011.
112 A. Coliva

justification, for instance, when you are able to say that you believe that
there is a red wall in front of you because you see it. Similarly, you have a
justification for believing that the sum of the internal angles of a triangle
is 180° because … where the dots should be replaced with a proof of that
theorem. Entitlements, on the other hand, are warrants for one’s beliefs
that need not be articulable by a subject and yet can have a rationalising
role with respect to the beliefs held on their basis, because they impinge
on their likely truth (at least in normal conditions). Hence, for instance,
a child has a perceptual entitlement for the belief that there is a red wall
in front of her, thanks to the kind of perceptual experience she is hav-
ing. She has it, even if she is unable to give her reasons for her belief,
perhaps because she lacks the ability to conceptualise her experience.32
Entitlements are thus externalist kinds of warrant.33
Moving on to self-knowledge, roughly speaking, one’s first-order men-
tal states, according to Burge, play the role of entitlements with respect
to one’s psychological self-ascriptions. Yet he thinks that entitlements do
not derive from the reliability of an inner scanning mechanism or from
the reliable operation of any other cognitive mechanism or procedure.
Such reliability is necessary in order to possess an entitlement, but not
sufficient. Entitlements derive, for Burge, from certain a priori consid-
erations concerning constitutive aspects of what it means to be a critical
thinker or a reasoner as well as a subject to whom moral norms may
apply34—whence the title of “rational externalism” for his proposal.
According to Burge, critical thinking is deployed when we provide
proofs for theorems, think through a plan, construct a theory or engage
in a debate. What happens in these cases is that, for instance, we start
out with a belief that P. We find out or are presented with conclusive

32
This is a very simplified account of perceptual entitlements. In fact, several a priori conditions
characteristic of perception will have to be met too, for Burge. His proposal, therefore, differs from
crudely reliabilist versions of perceptual justification (or warrant, in Burge’s terminology). See
Burge 1993 and 2010. For an assessment, see Coliva 2012a.
33
In Burge 2011 (p. 189), it is pointed out that although one can have a perceptual entitlement and
yet lack knowledge of the world around one, if, for instance, one is in a sceptical scenario, one can-
not have an entitlement for one’s psychological self-ascriptions and yet be wrong about one’s own
mental states. Burge dubs this second kind of entitlement “immune to brute error”.
34
Being a reasoner and a subject to whom moral norms apply are, according to Burge 2011, con-
stitutive features of what it means to be a subject.
 5 Epistemically Weak Accounts 113

evidence against it and we realise that. We therefore immediately revise

our initial belief. In fact, we are rationally required to do so. If we
did not do it, we would be irrational. Such belief revision, however,
presupposes knowledge of one’s mental attitude towards P in the first
place as well as knowledge of the fact that contrary evidence to it, call
it Q, was not merely hypothesised but judged to be true. To be critical
reasoners, therefore, we need to know our own mental states. We do so
by having true beliefs about them, warranted, in the form of entitle-
ment, by the very occurrence of the relevant first-order propositional
attitudes.
In contrast to Peacocke’s proposal, Burge makes no requirement that
a subject be conscious of her first-order beliefs, in order to be justified in
self-ascribing them. I take it to mean that, according to Burge, there is no
need for one’s first-order mental states to be phenomenologically salient
in one or more of the ways discussed in connection with Peacocke’s pro-
posal. The transition from one’s first-order propositional attitudes to their
self-ascription may well be “blind” and yet rational since it is under-
written by the kinds of consideration just presented having to do with
the rational norms (and abilities), which are constitutive of being criti-
cal reasoners. To put it differently, first-order propositional attitudes, for
Burge, cannot be a subject’s reasons for the corresponding psychological
self-ascriptions.35
Let us now review how Burge’s proposal fares with respect to ground-
lessness, transparency and authority. Weak groundlessness—that is, the
absence of any substantial cognitive achievement based on observation or
inference—is certainly respected. Furthermore, since no previous aware-
ness of one’s first-order mental states is required in order for one to be in
a position knowledgeably to self-ascribe the relevant propositional atti-
tude, a stronger kind of groundlessness is guaranteed too. Yet it is not part
of Burge’s overall proposal adherence to an even stronger form of ground-
lessness, according to which one’s first-order, not necessarily conscious
mental states cannot function as epistemic grounds for one’s judgement.
For they can be its grounds, even if they are not a kind of conscious evi-
dence on which the judgement is based.

35
See Burge 2011, pp. 192–193.
114 A. Coliva

With respect to transparency, according to which (at least when concep-

tually endowed, attentive and lucid thinkers are concerned) the occurrence
of one’s first-order mental states is of a piece with one’s knowledge of them,
Burge’s proposal seems to fare well. For it appears to be a central aspect of it
that such a subject, insofar as she is also capable of critical reasoning, should
have knowledge and therefore be disposed to self-ascribe the relevant prop-
ositional attitudes, on the basis of which she conducts her reasoning.
Finally, authority is guaranteed too. Here, it is important to specify that
Burge starts out by considering cogito-like thoughts, like “I am thinking
that P˝, taken in the performative, not in the descriptive, way, which are
actually self-verifying.36 For, by thinking those second-order thoughts, one
also happens to think that P. Hence, one has the corresponding first-order
mental state. Cogito-like thoughts therefore enjoy the strongest form of
authority possible—that is, infallibility.37 Still, Burge does not confine the
proposal to that kind of psychological self-ascription. In other cases, he
allows for the possibility of error. The details of his proposal are quite com-
plex. The basic idea, however, is that memory preserves the content and
mode of one’s earlier and lower-level attitudinal state and therefore ensures
that, if one is cognitively well functioning, one’s psychological self-ascrip-
tion is correct and warranted. Error, in contrast, would signal a subject’s
cognitive malfunctioning. Hence, it would impair one’s ability to engage
in critical reasoning and therefore one’s rationality. It would therefore be
incompatible with the retention of a warrant for the relevant psychologi-
cal self-ascription. Thus, there cannot be false yet warranted psychological
self-ascriptions when first-personal self-knowledge is at stake.
Similar views are put forward in connection with forward-looking
propositional attitudes, like intentions, which are fundamental to practical
reasoning, as well as with respect to self-ascriptions of perceptions and
other non-attitudinal mental states involving perceptual contents.

36
In this case, “I am thinking/judging that P” would be similar to the explicit performative “I
promise you to φ”, whereby one is promising to φ as well as saying that one is so doing. Thus, by
judging “I am thinking/judging that P”, one would thereby think or judge that P while also judging
that one is so doing.
37
Burge (2011, pp. 210–211) claims that “I hereby judge that P” is an impure cogito-like case and
that error is possible. Yet, if it occurred, it would signal a malfunctioning in a subject’s cognitive
capacities and that would actually defeat one’s entitlement for the relevant self-ascription.
 5 Epistemically Weak Accounts 115

A slightly different proposal is made with respect to sensations, such as

pain and other phenomenologically salient mental states. Again, accord-
ing to Burge, it is constitutive of being a critical reasoner and a subject
of moral norms and evaluations to have knowledge of one’s phenomenal
states. Here, too, however, the capacities involved in making the relevant
self-ascriptions draw only on one’s phenomenal awareness of the sensa-
tion and on the preservation of that content in memory. These capacities
are, for Burge, constitutively tied to the possession of the phenomenal
concepts deployed in the relevant self-ascriptions. Once more, although
error can occur, it would be incompatible with retention of an entitle-
ment for one’s self-ascription and its presence would signal some kind of
cognitive malfunctioning.
Let us now turn to the evaluation of Burge’s proposal. I will not here
push any worry concerning Burge’s purely psychological notion of self
and his equation of selves with subjects capable of critical reasoning and
moral appraisal. I will confine myself merely to his proposal regarding
first-personal self-knowledge. A first problem arises in connection with
Burge’s proposal concerning non-cogito-like cases. For it is unclear what
it means that the mode and content of one’s attitudes are preserved in
memory. If it means that one retains a memory of one’s previous cogito-
like self-ascriptions, it is fine, but unsuitable as an account of psycho-
logical self-ascriptions of attitudes which are not entertained through
a cogito-like performance. If, in contrast, it means that, after having
formed a first-order attitude with a given content, one retains a memory
of both its constitutive aspects, which then licence a psychological self-
ascription, the model would presuppose prior awareness of one’s atti-
tudes. This, however, as we saw while discussing Peacocke’s views in the
previous section, is highly problematic. Such an awareness would have to
depend either on the dubious claim that the attitude itself is phenomeno-
logically salient or on the ability to identify it as the attitude it is. This,
however, would make the resulting account of self-knowledge circular.
For it would presuppose the very knowledge it was supposed to explain.
Alternatively, Burge could try to avoid the problems rehearsed in con-
nection with Peacocke’s account by denying that first-order conscious
mental states and episodes serve as independent bases for the correspond-
ing self-ascriptions. This move, however, would make the analogy with
116 A. Coliva

the perceptual case falter  and it would then become mysterious what
would ground the relevant psychological self-ascriptions. This problem
may escape notice if one focusses on cogito-like thoughts, where the
second-order mental state brings the first-order one into existence. Yet it
is a problem when one aims to account along these lines for other kinds
of psychological self-ascriptions.
Another problem emerges in connection with Burge’s account of our
knowledge of our own sensations. For he seems to presuppose an account
of phenomenal concepts as tags for phenomental states, which is highly
dubious, as we saw while presenting Wittgenstein’s remarks against the
possibility of a private language or indeed of concepts designed directly to
refer to one’s phenomenologically salient and distinct experiences.
Moving on to more general worries, if we leave aside cogito-like
thoughts and think of first-order propositional attitudes and their self-
ascriptions as separate existences, to be told that thanks to constitutive
principles concerning what it is to be critical reasoners we have an entitle-
ment—that is, a particular kind of externalist epistemic warrant for the
relevant psychological self-ascriptions—does not tell us anything about
how we actually move from the former to the latter. That is to say, it
may still be the case that a subject self-ascribes the relevant propositional
attitude thanks to the subpersonal, therefore unconscious, operation of
a reliable inner mechanism.38 Burge’s proposal would then be entirely
compatible, from a merely epistemic point of view, with crude forms of
reliabilism such as Armstrong’s. That is to say, Burge could insist that the
kind of rational norms involved in accounting for our entitlement to self-
knowledge are a priori. Yet it is not precluded to a reliabilist to endorse
this claim and still insist that it is one thing to explain how knowledge
of our own propositional attitudes comes about and quite another to
insist that it is a priori true that first-personal self-knowledge is neces-
sary in order to engage in critical reasoning and to be subject to moral
appraisal. The latter does not account for the former, and the former is
epistemically significant whereas the latter is not—no matter how illumi-
nating it may be about the importance of self-knowledge with respect to

38
Indeed, Burge 1996, p. 103, fn. 12 allows that his account is compatible with the presence of “a
causal mechanism that relates attitudes to judgments about them”.
 5 Epistemically Weak Accounts 117

the very possibility of being a critical reasoner and a subject capable of

moral appraisal.
Possibly in response to such a worry, Burge (2011, p. 219) claims that
“fulfilling epistemic norms governing self-knowledge thus entails exercis-
ing powers constitutive of critical reasoning. They are not norms that are
merely necessary to critical reasoning, but constitutively independent of
it”. I take this to mean that the norms of self-knowledge stem from those
for being a critical reasoner. Thus, if one is such a reasoner, one cannot
but have first-personal self-knowledge and, in its turn, that kind of self-
knowledge does not depend on the exercise of abilities exceeding those
needed to be a critical reasoner at all.
Hence, Burge’s vindication of our entitlement to self-knowledge
depends on ruling out the possibility of subjects who can perform belief
revision without having propositional knowledge of their first-order
mental attitudes. For it is only if subjects like that are shown to be impos-
sible that self-knowledge will be proven to be at least necessary for being
a critical reasoner. Yet, on reflection, there seems to be no bar to the pos-
sibility of conceiving of a subject who engages in belief revision and is
at least capable of doing so in connection with devising and executing a
practical plan and yet is unable to make the relevant psychological self-
ascriptions. This might be the case if she lacked the relevant conceptual
repertoire. Here is an example taken from Peacocke (1998, p. 276):

Suppose you come home, and see that no car is parked in the driveway. You
infer that your spouse is not home yet … Later, you may suddenly remem-
ber that your spouse mentioned in the morning that the breaks of the car
were faulty, and wonder whether she may have taken the car for repair. At
this point, you suspend your original belief that she is not home yet. For
you come to realise that the absence of the car is not necessarily good evi-
dence that she is not home. If the car is being repaired, she would have
returned by public transport. Then finally you may reach the belief that she
is home after all, given your next thought that she would not have taken
any risks with faulty breaks.

Such a subject would presumably count as a critical reasoner, even if he

were unable to form the corresponding self-ascriptions. Thus, it does not
118 A. Coliva

seem to be a constitutive aspect of being such a reasoner that one should

have propositional knowledge of one’s first-order propositional attitudes.39
It would, only if an argument were produced to the effect that one cannot
even have beliefs and intentions unless one had propositional knowledge
of them. That is to say, we would need an argument for holding that first-
order propositional attitudes are constituted by, or are at least asymmetri-
cally dependent on, the corresponding self-ascriptions. However, Burge is
strongly against such a claim.40 Of course, normal adult human beings,
endowed with the relevant conceptual repertoire, do have propositional
knowledge of their propositional attitudes. Yet, in order to redeem one’s
entitlement to self-knowledge the way Burge conceives of it, a subject of
the kind proposed should not be a conceivable alternative at all. If this is
correct, then it is also problematic to think that all and only the very same
capacities involved in critical reasoning suffice for self-knowledge.
The other strategy open to Burge, and indeed pursued in several of his writ-
ings on the topic, is to insist on a more demanding notion of what it means
to be a critical reasoner. If only creatures capable of self-conscious belief revi-
sion count as reasoners, it would be true that being a reasoner (in this more
demanding sense) requires having knowledge of one’s mental states and not
simply of their contents. For instance, if we consider a detective engaged in
finding the culprit who reasons on her case, she would presumably engage
in forms of reasoning like the following one: “There was some evidence in
favour of P, so I believed S was the culprit, I therefore looked for further evi-
dence, but I found out that S had a strong alibi; so I no longer believe he is
the culprit, and I will have to start my investigation from scratch”. Yet, if this
is the notion of being a critical reasoner at stake, it seems of little significance
in accounting for critical reasoning in general and it would be quite plati-
tudinous that self-knowledge is entailed by it. So the link between critical
reasoning and self-knowledge would be trivial and established on the basis of
an ad hoc notion of what it means to be a critical reasoner.41

39
A similar worry can be found in Peacocke 1996, Moran 2001 (pp. 109-113), Bar-On 2004 and
Cassam 2014.
40
See Burge 2011, p. 192.
41
Furthermore, it would be dubious that only creatures capable of such a robust form of critical
reasoning would count as selves and that, in turn, would make trouble for Burge’s vindication of
our entitlement to self-knowledge, which heavily depends on constitutive claims about what it
means to be a self.
 5 Epistemically Weak Accounts 119

3 Evans’s Transparency Method

Evans (1982, p. 225) famously claimed that,

In making a self-ascription of belief, one’s eyes are, so to speak, or occasion-

ally literally, directed outward-upon the world. If someone asks me “Do
you think there is going to be a third world war?”, I must attend, in answer-
ing him, to precisely the same outward phenomena as I would attend to if
I were answering the question “Will there be a third world war?”.

If the response to the question is in the affirmative, then I can prefix P

with “I believe that”. The interest of this method is that it is clearly anti-
Cartesian. No “look inside” at one’s own mental states is presupposed
in order to answer a question about what one believes. Hence, Evans’s
recipe for getting knowledge of our beliefs is often called the “transpar-
ency method” because it bypasses one’s inner world, as it were, to direct
one’s mind and attention to P itself and to the reasons for or against it.
Evans did not develop his suggestion in The Varieties of Reference and
he did not live long enough to take it up in further writings. However,
the transparency method has become of interest to several philosophers
after him42. We will look at two prominent attempts to develop it into
a viable account of first-personal self-knowledge. One of them is more
epistemically oriented and is due to Jordi Fernández; the other one is
more deliberative-oriented and is championed by Richard Moran. We
will consider them in turn.

3.1 Fernández’s Epistemic Account

Jordi Fernández’s Transparent Minds (2013) is a sustained attempt at devel-

oping Evans’s suggestion in an epistemic way, so as to make it suitable as an
account of how we have knowledge of our own propositional attitudes—not

42
Although Evans did not develop his proposal, he actually maintained his model for self-knowl-
edge of beliefs could be extended to other cases of first-personal knowledge of our own mental
states. Hence, he embraced monism. (See Evans 1982, p. 225).
120 A. Coliva

only beliefs but also desires. We will consider only the former for the objec-
tions we will raise against it could be pari passu repeated against the latter.
Fernández’s central thesis is that one’s psychological self-ascriptions are
based on whatever grounds support the corresponding first-order beliefs.
Therefore, the former are at least prima facie justified if one’s beliefs are
as well. Usually they are, if they follow the “production-of-belief prin-
ciple” (2013, p. 46, 55): if one apparently perceives/remembers/acquires
testimonial information/has an intellectual seeming that P, one comes to
believe P; and so does one if one believes Q and believes that Q entails P.
Fernández then addresses the objection that the bypass model is a bet-
ter explanation of how one generates or forms one’s beliefs rather than of
how we know pre-existing ones. Fernández concedes that many times that
is the case, but he contends that there is also a genuinely doxastic reading
of the question “Do you believe that P?” whose answer depends on the
application of the bypass model. He considers a case of a client who asks
his lawyer whether she believes him innocent. He writes (2013, p. 51):

Clearly he is not asking whether he is innocent. He must already know

that. So the deflationist reading is not available in this context. Yet, … the
lawyer [to answer the question]… will focus on those considerations that
would support or challenge the belief that her client is innocent”.

It is not easy to see how this could be a response to the original challenge.
For the lawyer in this case seems actually to be making up her mind
regarding the innocence of her client, by considering evidence for and
against it.
Nevertheless, Fernández then explains how the bypass model satisfies
groundlessness and authority. It does satisfy weak groundlessness because
we enjoy certain mental states (like experiences, memories, etc.) which
naturally and usually lead us to form the corresponding beliefs as well as
second-order self-ascriptions, without any need to observe our own overt
behaviour and without having to reason from the basing mental state to
the second-order self-ascription. By contrast, other people will be in a posi-
tion to ascribe a belief to us only by observing our overt behaviour and by
inferring to its likely cause. If, however, our self-ascriptions are not based
on outer observation and inference to the best explanation, they are not
open to the kinds of error that can beset third-personal ascriptions of belief.
 5 Epistemically Weak Accounts 121

Hence, our own access to our own beliefs is different and more secure
than whatever access other people might have with respect to them. For
this reason, our self-ascriptions of belief are at least weakly authoritative.
Fernández also claims that the bypass model is not causal, because the self-
ascription is not causally brought about by the first-order belief, but, rather,
it is grounded on the evidence on which that very belief is based. This
guarantees that also groundlessness in the stronger form is respected since
first-order mental states do not play the role of evidence for the relevant
self-ascriptions.43 Finally, transparency is guaranteed in its strong version
because provided one has the relevant concepts and the relevant first-order
mental states, one will immediately be in a position to self-ascribe them.
Two remarks are in order: first, it seems odd to say that one’s self-ascriptions
of belief should be justified by the evidence that prompts the first-order
belief itself. Intuitively, the self-ascription, if justified at all, should be justi-
fied by the corresponding first-order mental state. This would indeed match
the intuition that although we may have unjustified first-order beliefs, if
they were formed on inappropriate, scant or no grounds, we could never-
theless be justified in self-ascribing them, for the simple reason that we do
have them. To rescue this compelling intuition, Fernández has to say that
although the subject would have evidence but no justification for the first-
order belief, that very evidence would count as a justification for her self-
ascription. Now, besides the difficulty of distinguishing between evidence
and grounds in a satisfactory way, it is clear that this could be so only if we
considered the role of that evidence in giving rise to the first-order belief
which then would tend to correlate with one’s self-ascription of it. However,
if that is the case, in the end what “justifies” one in self-ascribing the belief is
precisely the first-order belief itself, not the evidence that prompts it.
Second, it is difficult to see how the bypass model would count as
non-causal. For, after all, it heavily depends on the fact that once we are
exposed to certain kinds of evidence, we tend to form the corresponding
first-order beliefs and, if conceptually equipped, we tend to form the rel-
evant psychological self-ascriptions. True, if Fernández’s story were right,
43
However, it may be objected, as Giorgio Volpe observed, that Fernández’s account violates a
particular form of groundlessness—namely, one which would ban that the relevant psychological
self-ascription be based on any kind of evidence or grounds whatsoever, even those which would
ground the first-order belief in P. I agree with this objection, but at least the restricted versions of
groundlessness we introduced in Chap. 3 would be respected.
122 A. Coliva

the causal basis of one’s psychological self-ascription would not be the

first-order belief but the grounds on which the latter is based, but the
eventual psychological self-ascription would be arrived at by means of a
causal process nonetheless.
Fernández considers two further objections to the bypass model: the
one from the absence of grounds and the one from the absence of belief.
The former has it that if one has lost the evidence on which a given first-
order belief of hers is based, this would seem to entail, on the bypass
model, that she could not self-ascribe that belief. The latter, in contrast,
points out that the bypass model has difficulties in accounting for our
own knowledge of lack of belief. The response to the former objection is
that as long as one seems to remember that P, no matter what evidence
originally prompted one’s belief that P, one has evidence for the self-
ascription of the belief that P. The answer to the latter, in contrast, is that
by reflecting on P one can find out that there are no sufficient grounds
to endorse it (or its negation) and therefore come to form the belief that
one does not believe that P.
As resourceful as they are, these replies seem to miss the point of the
objections, however. For the first objection hinges on the idea that it can-
not be the evidence that led one to believe that P that justifies one’s self-
ascription. The latter, in contrast, hinges on the fact that although there
are no sufficient grounds to believe that P, there is sufficient ground to
self-ascribe the absence of belief. Therefore, the grounds on which these
beliefs are based cannot be the same.
The preceding discussion seems enough to motivate us to look at a dif-
ferent way of developing Evans’s transparency method.44

3.2 Moran’s Deliberative Account

Richard Moran, in his influential Authority and Estrangement (2001),

proposes an account of self-knowledge whose key element is the idea
that we have first-personal self-knowledge when and only when we form
beliefs (and other propositional attitudes) as responsible agents—that is

44
For a discussion of further aspects of Fernández’s proposal, see Coliva 2014.
 5 Epistemically Weak Accounts 123

to say, when we do so on the basis of weighing reasons for and against P

and of deliberating, on the basis of such an assessment, that P is the case
(or is good to have) and therefore believe that P (want P or intend to φ).
Key to Moran’s account is therefore Evans’s method. For, in his view, it
is only when we can answer a question about whether we believe that P
by looking directly at the reasons for or against P that we would have
first-personal self-knowledge. In all other cases, we will have to consider
evidence of a psychological or a behavioural kind (or both) and will thus
have third-personal access to our minds.
When we apply Evans’s method, we would not be believing at will,
according to Moran. For reasons for or against P exert normative con-
trol on us.45 Still, we would be exercising our agency. For we would
be applying a procedure which allows us to answer the question “Do
you believe that P?” by in fact answering the question “Shall I believe
P?”, which is eminently a practical question, according to Moran. We
are making up our minds, by deciding, rationally, what to believe. This
is something we can do only for ourselves and not for others, whose
minds therefore can only be described. Moreover, making up our minds
contrasts sharply with the case in which we find ourselves saddled with
the belief that P or even when we find ourselves assailed by such a (pass-
ing) thought or else discover that previously unconscious belief of ours
through therapy. In the terminology introduced in Chap. 2 (§2), when
propositional attitudes as commitments are at stake, then we have first-
personal knowledge of them.46 In contrast, we do not have it when
propositional attitudes as dispositions are at stake, nor, apparently, do
we have it when we are considering other kinds of mental states that are
not in any sense the result of an exercise of our agency. Hence, Moran’s
account is not applicable to our knowledge of ongoing sensations, emo-
tions and perceptions.

45
In Moran 2012 (p. 235), he calls this form of agency “agency as responsiveness to reasons” as
opposed to “agency as production”. The latter is “at will”, the former is constrained by epistemic
reasons. It is a form of agency nonetheless.
46
Moran, however, would probably reject this rendition of his ideas because he wants to preserve a
unitary meaning to “belief ” (see Moran 2001, pp. 83–94). For reasons exposed in Chap. 2 (§2), I
think that we can explain why commitments and dispositions are two species of the same genus
“belief ”.
124 A. Coliva

It is worth pointing out that Moran, contrary to Fernández, is not so

much interested in Evans’s method as an epistemic procedure that, start-
ing from having a first-order propositional attitude, allows us to form
a belief about it without having to look inward, recognise that mental
state for what it is and then self-ascribe it. Rather, he is interested in
Evans’s method insofar as it captures the idea that we can answer ques-
tions about what we believe by deliberating what to believe just on the
basis of reasons for or against P.  The deliberative dimension, together
with the relevance of the reasons for or against P itself, is the element of
Evans’s method that Moran focuses upon. Another way to stress the same
point is to remark that Moran does not consider the ensuing psychologi-
cal judgement “I believe that P˝ as a description, but as an avowal—that
is to say, as a direct expression of one’s having formed the belief that P
through rational deliberation.
Let us then review how Moran’s account fares with respect to the con-
straints that any account of first-personal self-knowledge should meet.
Clearly, groundlessness in its weak version at least would be satisfied,
for the answer to the question “Do you believe P?” will not be based
on observing one’s own mental states or behaviour and on inferring
to their likely cause. Moreover, it is satisfied in its strong version too,
for the self-ascription is not based on one’s previous awareness of one’s
propositional attitudes as commitments. That is to say, our first-order
propositional attitudes as commitments do not play the role of evidence
for the corresponding self-ascriptions, in Moran’s account.47 Nor is the
relevant psychological self-ascription arrived at by inferring that because
one has judged that P is the case, one believes it.48 Transparency, indeed
in its strong version, would be guaranteed when propositional attitudes
as commitments are at stake, precisely because, for any conceptually
endowed creature, by application of Evans’s method, she would be in

47
Like Fernández’s account, Moran’s would not respect an unrestricted version of groundlessness
(see fn. 42). For our purposes, though, it is sufficient that it respects indeed both the weak and
strong version of it we have been discussing throughout this book.
48
Implausibly, Cassam 2014 raises this objection to theories that hinge on the application of the
transparency method. For a discussion, see Coliva 2015b. See also Boyle 2011a and 2015. However,
his gloss (Boyle 2011a) on Moran’s position as entailing tacit self-knowledge of one’s first-order
mental states, which needs be made explicit only through reflection, seems to betray the gist of
Moran’s non-evidential account.
 5 Epistemically Weak Accounts 125

a position to self-ascribe her propositional attitudes as commitments.

Authority holds as well because we may make a mistake with respect to
our assessment of the reasons in favour of P, but once we have formed
the belief that P, based on those reasons, and we have self-ascribed it, by
means of the application of Evans’s method, there is no room for mistake,
save for cases of self-deception, which Moran does not wish to explain
away (see Moran 2001, pp.  83–94). Hence, given Moran’s account of
self-knowledge, weak authority as we characterised it in Chap. 3 holds.49
Still, Moran does not endorse a constitutive view of first-personal self-
knowledge. For he does not wish to espouse the idea that one’s psychological
self-ascriptions determine or bring about one’s first-order propositional atti-
tudes as commitments. Nor does he think that to judge or assert “I believe
that P˝ is just a notational variant of asserting or judging “P˝. Similarly,
he does not seem to endorse the idea, which would be very congenial to
him, that groundlessness, transparency and authority hold a priori and are
constitutive aspects of being a responsible agent. In his view, some, how-
ever minimal, cognitive work must be done in order to make a psychologi-
cal self-ascription, even if the latter is not based on taking one’s first-order
beliefs as commitments as evidence for the corresponding self-ascription.
Having dispensed with constitutive, indeed non-epistemic, accounts of
first-personal self-knowledge, Moran does not offer any detailed account
of the epistemology of psychological self-ascriptions that satisfy ground-
lessness, transparency and authority, though. This has left several readers
in the dark regarding the epistemic credentials of his proposal.50 Indeed,
one might worry that Moran is more concerned with the phenomenon
of creating (at least some of ) our own mental states rather than with the
issue of how we can and do know them in ways that are systematically

49
It is difficult, though, to see how psychological self-ascriptions reached through the transparency
method could be wrong. For the subject will have the relevant judgement-dependent attitude, if
she has formed it by considering reasons in favor of its content. Still she may also exhibit a behvior
which at least partially goes against her psychological self-ascription. In that case, it would seem
that Moran’s account would be more consistent with treating cases of self-deception as due to the
presence of two conflicting mental states, only one of which is formed through deliberation and is
self-known by application of the transparency method. There will be more about this account of
self-deception in Chapter 7 (§§3-4).
50
Moran acknowledges this charge in his exchange with Sydney Shoemaker and Lucy O’Brien. See
Moran 2003, especially fn. 4 and pp. 409–417.
126 A. Coliva

different from the ones in which we would know of other people’s mental
states or of those aspects of our minds which are not of our making.51
Along these lines, one might even suggest that a better title for his book
would have been Authorship and Estrangement, leaving aside the issue of
how we can have any special knowledge of (some of ) our own mental
states and thus epistemic authority over them.
In response to this kind of worry, Moran has replied that he welcomes
the suggestion that his account should be completed by endorsing the
idea of an entitlement as a non-discursive kind of warrant.52 In his view,
having an entitlement “is not so much a matter of concept possession, as
a matter of possession of the relevant practical and cognitive abilities”.53
I suppose that what this means is that, in his view, entitlements in this
area would not be a priori arguments based on reflection on certain
concepts, such as the notion of being a responsible agent or a reasoner,
which would render one’s psychological self-ascriptions warranted, even
if one were not able to entertain them. Rather, according to Moran, it
seems that a subject will count as being entitled to her psychological
self-ascriptions just as long as she is actually able to make up her mind
on the basis of weighing reasons for or against P (or P would be good to
have), is able (in the case of desires and intentions as commitments) to
pursue certain courses of action for the time necessary to realise them,
and is able to apply the concepts which are necessary in order to make
the self-ascription.
Hence, a fair rendition of Moran’s account, I think, would be the
following one. Suppose someone asks us “Do you believe that P?”. We
can answer that question by deliberating what to believe on the basis of
assessing evidence for or against P directly. Hence, we can answer “yes”
(or “no”) to a question about our beliefs without having to rely on any
introspection and just by considering reasons for or against P itself. If
we then say or judge “I believe that P”, that should be considered an
avowal—that is, a direct expression—of our having deliberated that P is

51
This objection has been raised by several scholars (such as Bar-On 2004 and Shah and Velleman
2005). For a defence of Moran’s position, see Boyle 2009.
52
This is the gist of his response to Lucy O’Brien’s observation. See Moran 2003, p. 412.
53
Ibid.
 5 Epistemically Weak Accounts 127

the case. Still, it is an avowal for which we possess a non-discursive jus-

tification—that is, an entitlement—given the way it has been arrived at.
Despite its attractiveness, Moran’s account is wanting on several fronts.
First, because of its partiality. For Moran does not confine himself to
claims which, if correct, would account for first-personal self-knowledge
of only our commissive mental states and could be supplemented by an
account of how we can have first-personal knowledge of other mental
states of ours. Rather, he equates genuinely first-personal self-knowledge
with the capacity of deliberating and therefore of making up one’s mind.
But a lot of our mental states are known to us in a distinctively first-
personal way and yet they are not the result of a deliberation.54
Second, and more importantly, because by providing a non-epistemic
reading of Evans’s method it is not clear that he can actually account for
self-knowledge properly so regarded. The tension emerges clearly when
we consider that “I believe that P” would be an avowal, therefore a direct
expression of one’s just formed belief that P, based on considering evi-
dence for or against P itself. On a very natural reading of that claim, one
would be inclined to think that epistemically speaking, judging/asserting
“I believe that P” and “P”should be regarded as on a par, even though,
clearly, they would not be conceptually on the same footing. That is to say,
they should be considered alternative ways of giving expression to one’s
mind. It is only if one sees the psychological self-ascription as somehow
epistemically grounded on the corresponding first-order belief that the
two judgements (or assertions) would differ. This, however, is just what
Moran denies. His recently added-on talk in terms of entitlements does
not improve the situation. For, as we have just seen, he does not think
of entitlements as a priori justifications that would be there, even if they
were inaccessible to specific subjects. Rather, they would be possessed by
subjects just in virtue of having certain abilities. The epistemic abilities
which would underwrite the relevant psychological self-ascriptions are

54
Or at the very least he seems to hold that this kind of self-knowledge is more fundamental than
our self-knowledge of sensations, perceptions, and so on. (For a similar objection, see Finkelstein
2003.) Boyle 2009, who is sceptical of the claim that genuinely first-personal self-knowledge
should be equated with deliberative self-knowledge of one’s propositional attitudes, is however in
favour of the claim that it is more fundamental than any other kind of self-knowledge because it is
inherent to our notion of belief. The latter claim is specially defended in Boyle 2011b.
128 A. Coliva

precisely those that would lead a subject to deliberate whether to believe,

want or intend P. Of course, the conceptual capacities required in order
to issue the psychological self-ascription would be diverse, but that is not
the bone of contention for those who claim that there something epis-
temically significant and peculiar to psychological self-ascriptions. Hence,
it is not clear at all how there can be any epistemic difference between
judging/asserting “P” and “I believe that P”. For such a reason, there is
nothing in Moran’s account that can actually account for the fact that
the relevant psychological self-ascriptions are indeed the result of a spe-
cific kind of cognitive achievement which sets them apart from the mere
judgement/assertion that P (or that P is good to have).

4 Summary
In this chapter, we have looked at weak epistemic accounts of self-
knowledge. We have started with Peacocke’s rational internalism (§1),
whose characteristic claim is that the conscious occurrence of first-order
mental states grounds one’s psychological self-ascriptions and rationalises
them, without thereby requiring a subject to have observational or infer-
ential knowledge of them. We have shown that various possible glosses of
the notion of what it is for a first-order mental state to be conscious create
problems for Peacocke’s account, either because they would make it cir-
cular or because they would be unsuited to play a genuinely rationalising
role with respect to the relevant self-ascriptions.
We have then looked at Burge’s rational externalism (§2). We have
found it wanting because it relies on an idea of entitlement whose epis-
temic credentials are unclear. If it is modelled after the role of perceptual
experiences vis-à-vis judgements about the outer world, it risks reinstating
the idea that one’s first-order mental states should be conscious, to play a
warranting and rationalising role with respect to the relevant psychologi-
cal self-ascriptions. If it is construed as an a priori argument that should
link the ability to be critical reasoners with having knowledge of our own
mental states, it would actually be compatible with a reliabilist account
of how the latter is achieved. Moreover, on reflection, it is doubtful that
we cannot conceive of a critical reasoner who does not have propositional
 5 Epistemically Weak Accounts 129

knowledge of her own mental states. Hence, the required a priori connec-
tion between these notions is dubious. Or else, if it is not, it is because an
ad hoc notion of being a critical reasoner is being presupposed.
We have then moved on to a different kind of proposal which gives
centre stage to Evans’s so-called “transparency method” (§3). According
to that method, in order to get knowledge of our own beliefs, we should
not direct our eyes inward, to introspect our own minds to see whether we
can find certain mental states which we would then self-ascribe. Rather,
we have to direct our gaze outward, to determine whether it is the case
that P. If we find (what we consider) sufficient evidence for P, then we
can simply prefix P with “I believe that”, thereby making the relevant
self-ascription. We have considered two prominent ways of develop-
ing Evans’s method: one more epistemically oriented, recently put for-
ward by Jordi Fernández (§3.1), and another more deliberative-oriented
maintained by Richard Moran (§3.2). The former has been criticised for
implausibly claiming that the evidence which justifies one’s belief in P
would also justify one’s self-ascription of that belief and also because it
would have difficulties in accounting for knowledge of beliefs based on
no evidence in favour of their contents and for self-ascriptions of open-
mindedness with respect to P.  The latter, in contrast, has been found
wanting because it identifies first-personal self-knowledge with the ability
to deliberate and make up one’s mind, when intuitively we would like
to account also for the special way in which we do know our ongoing
sensations, emotions and perceptions. Moreover, we have argued that,
its merits notwithstanding, it does not have anything to offer in the way
of explaining how self-ascriptions of beliefs and other propositional atti-
tudes, which are the result of rational deliberation, would amount to
knowledge, despite its ambition to present an alterntive to non-epistemic
accounts of self-knowledge.
Given the difficulties which beset weak epistemic accounts of first-
personal self-knowledge, as well as the strong epistemic accounts we
reviewed in the previous chapter, the time is ripe for considering accounts
that somehow dispense with the idea that what goes by the name of first-
personal self-knowledge is actually the result of some peculiar cognitive
achievement and for reckoning with the fact that it is perhaps a misno-
mer to call it “knowledge” after all.
130 A. Coliva

Bibliography
Bar-On, D. (2004). Speaking my mind. Expression and self-knowledge. Oxford:
Oxford University Press.
Bilgrami, A. (2006). Self-knowledge and resentment. Cambridge, MA: Harvard
University Press.
Bonjour, L. (2003). A version of internalist foundationalism. In L. Bonjour &
E. Sosa (Eds.), Epistemic justification. Internalism vs. externalism, foundations
vs. virtues (pp. 61–68). Oxford: Blackwell.
Boyle, M. (2009). Two kinds of self-knowledge. Philosophy and Phenomenological
Research, 77(1), 133–164.
Boyle, M. (2011a). Transparent self-knowledge. Aristotelian Society Supplementary
Volume, 85(1), 223–241.
Boyle, M. (2011b). Making up your mind. Philosophers’ Imprint, 11(17), 1–24.
Boyle, M. (2015). Critical study of Quassim Cassam Self-knowledge for humans.
The European Journal of Philosophy, 23(2), 337–348.
Brewer, B. (1999). Perception and reason. Oxford: Oxford University Press.
Burge, T. (1993). Perceptual entitlement. Philosophy and Phenomenological
Research, 67, 503–548.
Burge, T. (1996). Our entitlement to self-knowledge. Proceedings of the
Aristotelian Society, 96, 91–116. Reprinted in P. Ludlow and N. Martin (eds.)
Externalism and Self-Knowledge, Stanford, CSLI Publications, 1998,
pp. 239–264.
Burge, T. (2010). Origins of objectivity. Oxford: Oxford University Press.
Burge, T. (2011). Self and self-understanding. The Dewey lectures (2007–2011),
The Journal of Philosophy, CVIII/6–7. Reprinted in Cognition through under-
standing (pp. 140–228). Oxford: Oxford University Press.
Cassam, Q. (2014). Self-knowledge for humans. Oxford: Oxford University Press.
Coliva, A. (2012a). Critical notice of Tyler Burge Origins of objectivity. Disputatio,
4(33), 515–530.
Coliva, A. (2014). Review of Jordi Fernández Transparent minds. Theoria, 81,
442–445.
Coliva, A. (2015b). Review of Quassim Cassam Self-knowledge for humans,
Analysis.
Evans, G. (1982). The varieties of reference. Oxford: Oxford University Press.
Fernández, J. (2013). Transparent minds. Oxford: Oxford University Press.
Finkelstein, D. (2003). Expression and the inner. Cambridge MA: Harvard
University Press.
 5 Epistemically Weak Accounts 131

Heal, J. (2012). Consciousness and self-awareness. In A. Coliva (Ed.), The self

and self-knowledge (pp. 123–138). Oxford: Oxford University Press.
McDowell, J. (1994). Mind and world. Cambridge MA: Harvard University
Press.
McHugh, C. (2012). Reasons and self-knowledge. In A. Coliva (Ed.), The self
and self-knowledge (pp. 139–163). Oxford: Oxford University Press.
Moran, R. (2001). Authority and estrangement. Princeton: Princeton University
Press.
Moran, R. (2003). Responses to O’Brien and Shoemaker. European Journal of
Philosophy, 11(3), 402–419.
Moran, R. (2012). Self-knowledge, transparency, and the forms of activity. In
D. Smithes & D. Stoljar (Eds.), Introspection and consciousness (pp. 211–236).
Oxford: Oxford University Press.
Peacocke, C. (1992). A study of concepts. Cambridge MA: MIT Press.
Peacocke, C. (1996) “Our entitlement to self-knowledge: entitlement, self-
knowledge and conceptual redeployment”, Proceedings of the Aristotelian
Society, 96, pp. 117-158. Reprinted in P. Ludlow and N. Martin (eds)
Externalism and Self-Knowledge, Stanford, CSLI Publications, 1998, pp.
265–303.
Peacocke, C. (1999). Being known. Oxford: Clarendon Press.
Peacocke, C. (2001). Does perception have a nonconceptual content? Journal of
Philosophy, XCVIII, 239–264.
Peacocke, C. (2002). Three principles of rationalism. European Journal of
Philosophy, X, 375–397.
Peacocke, C. (2003). The realm of reason. Oxford: Clarendon Press.
Peacocke, C. (2012). Conscious events and self-ascriptions: Comments on Heal
and O’Brien, in Coliva A. (ed), 180–188.
Pryor, J. (2005). There is immediate justification. In M. Steup & E. Sosa (Eds.),
Contemporary debates in epistemology (pp. 181–202). Oxford: Blackwell.
Shah, N., & Velleman, D. J. (2005). Doxastic deliberation. Philosophical Review,
114(4), 497–534.
 6
Expressivism About Self-knowledge

In this chapter, we consider expressivist accounts of self-knowledge. We

start with Wittgenstein’s views (§1), as mostly presented in the Philosophical
Investigations and in the Remarks on the Philosophy of Psychology, which are
at the origins of expressivism. We see how the central claim of expressiv-
ism is that psychological self-ascriptions are avowals of on one’s ongo-
ing mental states. Avowals directly express the relevant mental states and
do not describe or report on them. According to expressivists, avowals
replace more instinctive behavioural manifestations. We see how, in
Wittgenstein, these claims are coupled with the idea that avowals are
semantically discontinuous with respect to third-personal psychological
ascriptions and past-tense first-personal ones. In particular, avowals are
not truth-apt. Since, in Wittgenstein’s view, they are not the result of
any robust epistemic achievement, they do not express any knowledge—
let alone special knowledge—subjects have of their own mental states.
The use of “I know” in relation to one’s mental states serves, at most, a
grammatical function. That is to say, it signals the fact that, in virtue of
our linguistic practice, subjects are accorded a distinctive authority over
their own mental states. This authority does not have an epistemic origin,
though. Rather, it is built into the rules of our psychological language.

© The Editor(s) (if applicable) and The Author(s) 2016 133

A. Coliva, The Varieties of Self-Knowledge,
DOI 10.1057/978-1-137-32613-3_6
134 A. Coliva

We then move on to Dorit Bar-On’s (§2) more recent and comprehen-

sive defence of expressivism in the psychological domain. We see how she
renounces both the semantic and the deflationary epistemic aspects that
characterise Wittgenstein’s position, while subscribing to the expressivist
core claim that avowals are direct manifestations of our ongoing mental
states. While Bar-On’s arguments in favour of the semantic continuity of
avowals with third-personal and past-tense first-personal psychological
self-ascriptions, which amounts to seeing them as truth-apt, are a decisive
improvement over Wittgenstein’s position, her defence of the epistemic
credentials of avowals is found wanting.

1 At the Origins of Expressivism:
Wittgenstein
Wittgenstein is rightly considered at the origins of expressivism in the
philosophy of mind, especially for his annotations in the Philosophical
Investigations and in the Remarks on the Philosophy of Psychology (vol. 2).
As we shall see, he can also be seen at the origins of constitutive models,
at least to some extent. It is therefore appropriate to expound on his
views. His positive proposal, however, is better appreciated if it is seen
in connection with his criticism both of the introspective model and of
behaviourism.
On the one hand, Wittgenstein attacks introspectionism because it
would lead to a language of sensations and, more generally, of mental
states, which would be private. A private language, however, is, for him,
a category mistake. For the rules of this putative language would be just
“impressions of rules” (PI 259), hence no rules at all. Discussing the pos-
sibility of a subject’s keeping a diary of his sensations, named after a pro-
cedure of private ostensive definition, Wittgenstein writes: “whatever is
going to seem right to me is right. And that only means that here we can’t
talk about ‘right’” (PI 258). That is to say, if “S” is the name I give to a
sensation of mine, that no one else can have access to, its use over time
counts as right whenever I would think so, even when I may use it to
refer to a different kind of sensation or to nothing at all. This, however,
simply means that the distinction between correct and incorrect uses of
 6 Expressivism About Self-knowledge 135

that term has vanished. If that distinction is no longer in place, though,

there is no rule-governed use of the term. Thus, the word has no mean-
ing, given Wittgenstein’s normative conception of linguistic meaning.
Furthermore, the idea that we are acquainted with our sensations,
which are the referents of our psychological terms such as “pain”, “hun-
ger”, “thirst”, “joy”, and so on, would lead to solipsism—that is to say, to
the idea that one can know only one’s own mind and merely surmise what
goes on in other people’s minds, if indeed there are other people at all. For
one cannot know what another person is referring to when she uses those
very terms, not even by analogy with one’s own case, since their referents
are foreclosed to one. Indeed, one cannot even know whether another
person is referring to anything at all, by using those terms. Hence, one
cannot know if it is a person—that is, someone capable of having sensa-
tions (and other mental states)—rather than a creature who resembles us
in its looks and behaviour but who is not really enjoying any mental state.
On the other hand, Wittgenstein is critical of behaviourism, although
his expressivism might be taken to be a covert form of it. Crude behav-
iourism has it that there is no mental state, just behaviour. Although,
as we shall presently see, for Wittgenstein the language of sensations
replaces their natural and instinctive manifestations and is not consid-
ered directly to refer to those sensations, he is careful to stress that this is
not tantamount to denying the existence of sensations and further men-
tal states. What he does object to is simply the way in which, follow-
ing introspectionism, we are led to think of the language of sensations
after the name-object referential model, which immediately leads to the
endorsement—albeit implicitly—of the idea of a private language. As he
writes in a series of telling passages:

Suppose everyone had a box with something in it: we call it a “beetle”. No

one can look into anyone else’s box, and everyone says he knows what a
beetle is only by looking at his beetle.—Here it would be quite possible for
everyone to have something different in his box. One might even imagine
such a thing constantly changing.—But suppose the word “beetle” had a
use in these people’s language?—If so it would not be used as a name of a
thing. The thing in the box has no place in the language game at all; not
even as a something: for the box might even be empty.—No, one can “divide
through” by the thing in the box; it cancels out, whatever it is. That is to
136 A. Coliva

say: if we construe the grammar of the expression of sensation on the model

of “object and designation” the object drops out of consideration as irrele-
vant. (PI 293)

“Are you not really a behaviourist in disguise? Aren’t you at bottom really
saying that everything except human behaviour is a fiction?”—If I do speak
of a fiction, then it is of a grammatical fiction. (PI 307)

The origin of this “grammatical fiction”, for Wittgenstein, resides in

the fact that when we talk of mental processes and states, we leave their
nature undecided, yet fail to understand them because the analogy with
physical states and processes “falls to pieces”.

So we have to deny the yet uncomprehended process in the yet unexplored

medium. And now it looks as if we had denied mental processes. And natu-
rally we don’t want to deny them. (PI 308)

Again:

“So you are saying that the word ‘pain’ really means crying?”—On the
contrary: the verbal expression of pain replaces crying and does not describe
it. (PI 244)

Now, if the language of sensations does not work—in fact, cannot

work—on the basis of a name-object referential model, how does it work?
Wittgenstein raises and answers this very question in PI 244:

How do words refer to sensations?—There doesn’t seem to be any problem

here; don’t we talk about sensations every day, and give them names? But
how is the connection between the name and the thing named set up? This
question is the same as: how does a human being learn the meaning of the
names of sensations?—of the word ‘pain’, for example. Here is one possibil-
ity: words are connected with the primitive, the natural, expressions of the
sensation and used in their place. A child has hurt himself and he cries; and
then adults talk to him and teach him exclamations and, later, sentences.
They teach the child new pain-behaviour.
 6 Expressivism About Self-knowledge 137

Hence, according to Wittgenstein, subjects do have sensations, which

have primitive and natural expressions. For instance, in the case of pain,
it will be crying and moaning. Adults teach children to replace that prim-
itive behaviour with a linguistic one (or to accompany that behavior with
the appropriate linguistic expression), which becomes more and more
articulated, up to making use of an entire sentence like “I am in pain”,
after having gone through other less complex linguistic formulations, like
“Ouch!” and “Pain!”. Yet the function of that eventual sentence is not to
describe one’s inner state of mind nor, for that matter, to describe one’s
primitive pain-behaviour. Rather, it is to express and give voice to one’s
sensation, just as a cry is, in appropriate circumstances, the immediate
expression of one’s ongoing pain. “I am in pain”, therefore, is not the
linguistic manifestation of one’s belief to be in pain, reached through a
judgemental procedure in which one’s present sensation is recognised as
pain and named that way. “I am in pain”, for Wittgenstein, is just the
refined linguistic, yet by no means less epistemically immediate, expres-
sion of pain.
Several consequences follow from this alternative picture of our psy-
chological language. First, it can be apprehended and make sense only in
the context of  one’s instinctive behavioural manifestations. Hence, for
Wittgenstein:

“But doesn’t what you say come to this: that there is no pain, for example,
without pain-behaviour?”—It comes to this: only of a living human being and
what resembles (behaves like) a living human being can one say: it has sensa-
tions; it sees; is blind; hears; is deaf; is conscious or unconscious. (PI 281)

An “inner process” stands in need of outward criteria. (PI 580)

The human body is the best picture of the human soul. (PI II, iv, p. 178e)

The idea is that only of a living human being (and more generally of a
living animal being that behaves similarly to us), contra a human corpse
or a zombie, can we correctly say that it does have sensations and feelings
and can thus teach a subject to express them in linguistic terms.
138 A. Coliva

Secondly, “I am in pain” has, in the relevant context, an expressive func-

tion like “Ouch!” or “Damn it!”. That is, even if its surface grammar is that
of a descriptive sentence, which can be meaningfully denied and embed-
ded in suppositions (“Suppose I am in pain”) and therefore in conditional
statements (“If I am in pain, I go to see the doctor”), its function is not
descriptive1. Hence, either it is not up to semantic evaluation or, if it is,
it is only in a minimal sense. Wittgenstein is often considered one of the
inspirational sources of minimalism about truth. The idea, that is, that
there is no more to truth than its disquotational function (“Grass is green”
is true iff grass is green) and similar platitudes (if “Grass is green” is true,
“Grass is not green” is not true). Surely, Wittgenstein flirts with minimal-
ism from time to time, but it is not clear that he really endorsed it, at least
in connection with psychological avowals.2 Rather, as we shall immediately
see, he is more drawn towards a form of radical semantic contextualism.3
Thirdly, not all uses of the very same sentence are on a par. For, as we
have just seen, “I am in pain” is very often an avowal, hence an immediate
linguistic expression of one’s ongoing pain. Yet that very sentence can be
used differently when it is embedded in negation or in wider, especially
suppositional contexts. However, given Wittgenstein’s equation of mean-
ing and use (PI 43), that is tantamount to saying that, despite the identical
linguistic form, different occurrences of “I am in pain” can have different
meanings. Indeed, we explain their function, that is, their use, differently.
Since, for Wittgenstein, “the meaning of a word is what is explained by
the explanation of the meaning” (PI 560), that means that the very same
sentence can actually have an altogether different meaning depending on
the context of its utterance, not just a different function. In particular,
Wittgenstein distinguishes between the avowing, merely expressive func-
tion, and the descriptive one, which occurs when—in the context of a con-
ditional statement, for instance—we suppose being in pain and describe
what we would or would not do, just as we would do in the case of another
person. As we shall see in the following, for Wittgenstein it is important
1
Of course we can lie, but in that case the pretend verbal pain behavior, just like pretend bodily
pain behavior, would still retain an expressive function.
2
Pace Jacobsen 1996, which is more a development of Wittgenstein’s views to make them compat-
ible with minimalism about truth than a reliable historical reconstruction.
3
Contra what Jacobsen 1996 maintains.
 6 Expressivism About Self-knowledge 139

to realise the existence of these semantic asymmetries we are often oblivi-

ous to, given our tendency to think of language as working in a uniform
manner. In particular, it is important to realise that even psychological self-
ascriptions can depend on assuming a third-personal stance with respect to
oneself. Still, different occurrences of “I am in pain” in an avowing and in a
descriptive mode are related to one another and are not like “I went to the
bank” when used to refer to one’s going to the financial institute or to the
bank of the river. In both cases, the criteria for using that sentence depend
on characteristic forms of behaviour. In the one case, the sentence replaces
the instinctive behaviour elicited by an ongoing painful sensation; in the
other, it is used after a process of inference to the best explanation, based on
one’s reflection on one’s behaviour over time, which is not the one elicited
by an ongoing painful sensation, but still produced by distressing feelings
or moods. Moreover, when it comes to “I am in pain” and to “S is in pain”,
it should be kept in mind that one’s avowals of pain replace an instinctive
kind of behaviour, which is the same kind of behaviour that allows us to
say of some other person that she is in pain, even though we could never be
in a position to avow her pain. Although the similarity in human reactions
could be used to build some bridge between the use in the first-person and
in the third-person present of our psychological vocabulary, Wittgenstein
is very much attracted to a strong form of semantic contextualism,4 which
tends to deny the uniformity of meaning of “pain” across changes of person
and tense. So there is a family resemblance between “pain” when avowed
and “pain” when used to describe one’s own dispositions or someone else’s
pain, but no identity in meaning.
Fourthly, the use of “I know” in connection with one’s own psycho-
logical avowals is problematic. Here are some famous quotes:

In what sense are my sensations private?—Well, only I can know whether I

am really in pain; another person can only surmise it.—In one way this is

4
By strong contextualism, I have in mind what is nowadays called “semantic eliminativism”, a view
maintained by Wittgenstein and, more recently, by Charles Travis. I do not mean anything having to
do with forms of assessment sensitivity. Nor is the label “contextualism”, as used here, indebted to
Kaplan’s notion of context and to Grice’s distinction between semantics and pragmatics. Crudely put,
for Wittgenstein, the notion of context roughly coincides with that of language game and certainly is
not exhausted by the triad subject-place-time or any suitable extension of it. Moreover, given his equa-
tion of meaning and use, there is no room for a sharp division between semantics and pragmatics.
140 A. Coliva

wrong, and in another nonsense. If we are using the word “to know” as it
is normally used (and how else are we to use it?), then other people very
often know when I am in pain.—Yes, but all the same not with the cer-
tainty with which I know it myself!—It can’t be said of me at all (except
perhaps as a joke) that I know I am in pain. What is it supposed to mean—
except perhaps that I am in pain? Other people cannot be said to learn of
my sensations only from my behaviour, for I cannot be said to learn them.
I have them. The truth is: it makes sense to say about other people that they
doubt whether I am in pain; but not to say it about myself. (PI 246; cf. PI
II, xi, p. 221e)

I can know what someone else is thinking, not what I am thinking. It is

correct to say “I know what you are thinking”, and wrong to say “I know
what I am thinking”. (A whole cloud of philosophy condensed in a drop of
grammar). (PI II, xi, p. 222e)

Hence, according to Wittgenstein, other people can know my own mental

states. Thus, the typically solipsist claim “Only I can know my own pain”
is false. Others do too, even though not by analogy with what happens in
their own case, but through the kind of pain behaviour that I exhibit. In
fact, “I know I am in pain” is nonsensical, when “I am in pain” is an avowal,
and if “I know” is taken to express an epistemic relation between a subject
and a proposition, or even a fact. Why so? To become clear about that, it is
useful to recall the criteria for the correct and meaningful use of “I know”
Wittgenstein presents in On Certainty.5 Here is a quick summary of them:

1. One must have reasons for one’s knowledge attribution;

2. These reasons must be stronger than what they are supposed to ground;
3. There must be a method to find out whether what one claims to know
is the case;
4. The knowledge claim must be relevant;
5. It must make sense to say “I do not know”.

All these criteria would be violated in the case of “I know I am in pain”.

For what could one offer as a reason for “I know I am in pain” apart from

5
For an extended examination, see Coliva 2010, Chap. 2.
 6 Expressivism About Self-knowledge 141

just repeating that one is in pain? (1) would thus be flouted. One might
think that the reason is the sensation itself. Surely, Wittgenstein would
have objected to that, since, in his view, sensations as such cannot be rea-
sons for their ascriptions. As he famously wrote (PI 289):

“When I say ‘I am in pain’ I am at any rate justified before myself”.—What

does that mean? Does it mean: “If someone else could know what I am
calling ‘pain’, he would admit that I was using the word correctly”?

Clearly, the question must be answered in the negative, for him, and
in fact, he goes on to say:

To use a word without a justification does not mean to use it without

right [zu Unrecht]6.

Now, in the recent literature on Wittgenstein’s epistemology, this remark

is often at the basis of epistemic interpretations, which tend to attribute to
Wittgenstein an early endorsement of non-evidential warrants or, as they
are usually called nowadays, entitlements.7 As a matter of fact, however, to
have a right to use the expression “I am in pain”, in this context, is not to
be interpreted in an epistemic way. What Wittgenstein is saying, rather, is
that even if “I am in pain” is not based on having recognised one’s ongoing
mental state as pain, to use that sentence to avow and therefore to express
or exhibit one’s pain is correct. In other words, one’s use of that sentence
is correct even if it is not backed by any epistemic warrant or guarantee.
Let us suppose, however, that one’s sensations were one’s reasons for
the relevant self-ascriptions. In that case, they would certainly not be any
stronger than the very knowledge claims they are supposed to ground
(contra (2)). As to condition (3), clearly one does not find out whether
one is in pain (at least when “pain” is meant as meaning an occurrent
sensation and not a dispositional state one enjoys and self-attributes
through observation and inference to the best explanation). One is or

6
Hanoch Ben-YAmi has kindly pointed out to me that the correct translation would be “incor-
rectly”, rather than “without right”. This would further support my reading of Wittgenstein and
my criticism of epistemic interpretations.
7
See, in particular, Wright 2004 and Williams 2004a, b.
142 A. Coliva

has pain and gives voice to it. Furthermore, a claim like “I know I am
in pain” would—in the normal run of cases—be totally irrelevant (thus
flouting (4)) and indeed odd in the course of a conversation, up to the
point that, if no proper context for its occurrence could be provided,
one would start doubting whether a subject is in her right mind, or
would start thinking that perhaps she is trying to make a joke, and so on.
Finally, in the normal run of cases, it does not make sense to suppose that
one might not know whether one is in pain (contra (5)). Of course, there
can be cases of confused sensations—an itch which borders pain—but
the phenomenon of vagueness cannot support the idea that when we
experience sensations we may not know that we do. As to the possibility
of unconscious sensations, they would be a conceptual solecism in this
connection. Thus, for Wittgenstein, the use of “I know” in connection
with one’s avowals would not make sense.
Still, there is a grammatical sense of “I know” which would be appro-
priate in connection with one’s psychological avowals. Again, the exis-
tence of such a grammatical use of “I know”, for Wittgenstein, is clear
from his remarks in On Certainty. Two aspects of the grammatical use of
“I know” are particularly relevant with respect to avowals—namely, (1′)
the fact that “I know” would actually mean “A doubt is excluded” or “I
cannot be wrong” (OC 59) and (2′) that the “I” is unimportant (OC 58).
Here are the relevant passages in the Philosophical Investigations:

The truth is: it makes sense to say about other people that they doubt
whether I am in pain; but not to say it about myself. (PI 246)

If anyone said “I do not know if what I have got is a pain or something

else”, we should think something like this, he does not know what the
English word “pain” means; and we should explain it to him.—How?
Perhaps by means of gestures, or by pricking him with a pin and saying:
“See, that’s what pain is!” This explanation, like any other, he might under-
stand right, wrong, or not at all. And he will show which he does by his use
of the word, in this as in other cases.
That expression of doubt has no place in the language-game; but if we
cut out human behaviour, which is the expression of sensation, it looks as
if I might legitimately begin to doubt afresh. My temptation to say that one
might take a sensation for something other than what it is arises from this:
if I assume the abrogation of the normal language-game with the expres-
 6 Expressivism About Self-knowledge 143

sion of a sensation, I need a criterion of identity for the sensation [itself ];

and then the possibility of error also exists. (PI 288)

And with respect to (2′), the idea that “I˝ is not a referential expression
when used in psychological avowals, which is present in Wittgenstein
since his remarks in the Blue Book (pp.  66–67), makes its appearance
again in PI 404–411. Here are some passages:

In saying this [“I am in pain”] I don’t name any person. Just as I don’t name
any person when I groan with pain. Though someone else sees who is in
pain from the groaning. (PI 404)

It would be possible to imagine someone groaning out: “Someone is in

pain—I don’t know who!”—and our then hurrying to help him, the one
who groaned. (PI 407)

“But you aren’t in doubt whether it is you or someone else who has the
pain!”—The proposition “I don’t know whether I or someone else is in
pain” would be a logical product, and one of its factors would be: “I don’t
know whether I am in pain or not”—and that is not a significant proposi-
tion. (PI 408)

“I” is not the name of a person, nor “here” of a place, and “this is not a
name”. But they are connected with names. (PI 410)

To say that “I know” in connection with one’s sensations could, at most,

have a grammatical use, means to say that there is no inner epistemology,
really—a claim, we will see, at the heart of constitutive accounts nowa-
days more than of contemporary expressivist ones. Furthermore, it means
to say that it is a characteristic trait of our linguistic practice that sub-
jects’ avowals are not challenged unless there are reasons to doubt of their
sincerity. As Wittgenstein writes, although in the context of explaining
away, as we may put it, the epistemology of self-ascriptions of intentions:

“Only you can know if you had that intention.” One might tell someone
this when one was explaining the meaning of the word ‘intention’ to him.
For then it means: that is how we use it. (And here “know” means that the
expression of uncertainty is senseless). (PI 247)
144 A. Coliva

It is important to dwell a bit longer on what it means for a remark to be

grammatical, according to Wittgenstein. The notion of grammar, which is
subject to continuous development over the years, from the immediately
post-Tractarian works up to Wittgenstein’s last collection of notes, namely
On Certainty, is connected to the ones of analyticity, aprioricity and con-
ceivability. Early examples of grammatical sentences, for Wittgenstein, are
(i) “An object cannot be of two different colours all over its surface at the
same time” or (ii) “Patience is played alone”. The idea, on the one hand,
is that these are meaning constitutive propositions. Thus, it is constitutive
of what we mean by “patience” that it is played alone. If we changed the
rules of playing patience and allowed multiple players, the very meaning
of that word would alter. Similarly, for what we call “(physical) object”.
Now, grammar is “autonomous” for Wittgenstein. This means that it is
not answerable to reality. This is clear from Wittgenstein’s treatment of
grammatical propositions such as (i): we do not hold (i), because objects
cannot physically be of two different colours all over their surface at the
same time. Nor is it because the structure of our sensory experience is of
that kind and excludes the possibility of perceiving an object as having
two colours all over its surface. The order of explanation goes the other
way round, in his opinion: given our concepts, reality and experience are
constrained in the way (i) prescribes. Similarly, it is not because imagina-
tion fails us that we cannot conceive of an object as being of two colours
all over its surface at once (cf. PI 251). Rather, if by “object” we mean
what we do, we would not know where to start from in order to conceive
of a physical object as having two different colours all over its surface at
the same time. Hence, to say that “I know I am in pain” can at most be a
grammatical proposition means to say that it makes explicit a fact about
how we use the vocabulary of sensations or, equivalently for Wittgenstein,
a structural aspect of our conceptual scheme. Hence, in the normal run of
cases, subjects’ pronouncements over their ongoing sensations are taken at
face value. Moreover, of someone who sincerely expressed a doubt about
whether she is or is not having an ongoing sensation, we would not under-
stand what she might actually mean (bar vagueness). The impression of
those words as having nonetheless a meaning would be due to the projec-
tion of meaning onto them from their ordinary contexts of use, in which
they do have meaning. Or even, given the usual meaning of those words,
 6 Expressivism About Self-knowledge 145

we would not understand what that person may try to communicate, up

to the point of calling into question the fact that she means the same as we
do with her use of those words.
Still, in philosophy, we constantly run the risk of mistaking grammati-
cal propositions for empirical ones yet as having the remarkable property
of indubitability. That is how the ideas of metaphysical necessity and of
infallibility arise. When taken that way, however, Wittgenstein’s reaction
is to declare them nonsensical; when they are understood as being gram-
matical, instead, his reaction is to find them obvious. As he writes:

“This body has an extension”. To this we might reply: “Nonsense!”—but

are inclined to reply “Of course!”—Why is this? (PI 252)

The answer to this question is precisely that, underneath their super-

ficial identity, different tokens of the same sentence may actually be
employed in so different a way as to amount to nonsense in one case or
to empirical or indeed grammatical claims in other cases.
Let us now turn to Wittgenstein’s account of propositional attitudes
and of our knowledge of them. According to him, we should pay atten-
tion to the variety of mental states we can enjoy. In PI 574, he writes: “The
concepts of believing, expecting, hoping are less distantly related to one
another than they are to the concept of thinking”. The difference depends
on the fact that thinking is occurrent and has salient phenomenological
aspects to it. Believing, expecting and hoping, in contrast, have a disposi-
tional element to them, and there is an internal relation between these atti-
tudes and their contents. That is to say, it is constitutive of these attitudes
that they have a certain content and the content they have individuates
each of their tokens as the particular belief, desire or intention it is. This
means, for instance, that if some external cause stops me from desiring
that P, it does not mean that my original desire was a desire that had that
cause as its object. If someone punches me and causes me to stop wanting
to have an ice cream, it does not mean that I wanted to be punched. This
relates also to Wittgenstein’s discussion of actions (PI 611–648, cf. BlB
23–24). Intentional actions have reasons and reasons, for him, cannot be
their causes. In general, causes may be “detached” from actions without
turning them into different ones, whereas reasons cannot. Intentions, in
146 A. Coliva

contrast, are internally related to their contents. They are not the causes of
the actions that fulfil them (PI 632, cf. BlB 23–24). In his view, moreover,
intentional actions and willing do not stop short of the action willed (PI
622), although there may be cases in which trying and making an effort to
bring about the action are relevant. Nor are they individuated by specific
(bodily) feelings (PI 625), although they may be accompanied by them.
When it comes to our knowledge of intentions and actions (as well
as of beliefs and desires), it depends on the circumstances. In the usual
run of cases, we express, avow or voice our own intentions. In that case,
there is no real epistemology of the mental, for the reasons we rehearsed
in connection with our putative knowledge of our own sensations. That
is, the criteria for the empirical and therefore meaningful use of “I know”
are flouted. To go over them once more: we do not have independent
reasons for “I know I ψ that P” (where ψ is a propositional attitude verb
like intending, desiring, believing, hoping, willing, etc.)—that is, reasons
other than the very mental state itself (contra (1)). A fortiori, even if one
conceded that the mental state could be a reason for our knowledgeable
self-ascriptions, it would not be any stronger a reason than what it is sup-
posed to ground (contra (2)). We do not find out our intentions, beliefs,
and so on. We do have them and give immediate expression to them by
saying or thinking “I intend to φ” or “I /believe/desire that P”, or indeed
we form them through  those very self-ascriptions. We do not discover
them either through introspection or inference to the best explanation
given our behaviour or feelings, which may, at least on occasion, accom-
pany the occurrence of these mental states (contra (3)). To say “I know I
ψ that P” must be relevant, but this is not the case in ordinary circum-
stances. In particular, asserting it would not add anything to one’s simple
self-ascription of the relevant mental state (contra (4)). Finally, given the
conditions which are supposed to hold, it would not make sense to say
or judge “I do not know whether I ψ that P” (contra (5)). For that kind
of judgement could meaningfully occur only in a context in which I do
not have yet formed the (specific) intention (or the belief or the desire).
Now, if the expressivist story has to get purchase in relation to propo-
sitional attitudes like intending, believing and desiring, it will have to be
the case that the relevant self-ascriptions are learned as ways of substitut-
ing pieces of instinctive behaviour. The question arises, however, whether
 6 Expressivism About Self-knowledge 147

there is such a distinctive kind of instinctive behaviour, which would

manifest our intending, desiring or believing. Whereas with pain it seems
clear that there is a characteristic pre-linguistic manifestation, with prop-
ositional attitudes it is not that obvious. Here is Wittgenstein’s reply:

What is the natural expression of an intention?—Look at a cat when it

stalks a bird; or a beast when it wants to escape. (PI 647)

The idea then is that also when it comes to wanting, intending and desiring
there are characteristic primitive reactions. For instance, a child wants to
have a glass of water and tries to reach for it and the adult sees her doing so
and says “You want a glass of water. I’ll give it to you”. Again, a child puts
her coat on and stands by the door. The adult says “You want to go out.
We’ll go in a moment”. Little by little, the child is taught to replace her
primitive behaviour with pieces of linguistic vocabulary, up to the point
where she herself is in a position to say “I want a glass of water/to go out”.
The case of belief, however, is certainly more complicated. For there do
not seem to be specific primitive reactions which would signal one’s believing
that today it is sunny or that one’s favourite toy is soft and tender. In particu-
lar, even if the child reaches for her toy and holds it close to her face, we do
not normally teach her to replace that behaviour with a linguistic one that
makes explicit the attitude of belief. That is to say, we do not say “You believe
that your teddy bear (say) is tender”, even if, to us, the child’s behaviour man-
ifests that attitude. We simply voice the embedded content of the attitude. It
is perhaps for this reason that when Wittgenstein explicitly talks about self-
ascriptions of belief in the present tense, he takes them to be equivalent to the
assertion of their embedded content, just in a more tentative way. In PI II,
x, while addressing Moore’s paradox—“I believe it is raining, but it isn’t”—
Wittgenstein observes, first, that “I believe that” is “transparent” (in Evans’s
sense of the term). For “The expression ‘I believe that this is the case’ is used
like the assertion ‘This is the case’” (p. 190e). Perhaps, sometimes, as a “hesi-
tant assertion [of P]” (p. 192e). This leads him to hold that “to believe” has a
different meaning when used in the first-person present, as opposed to in the
first-person past, or within a suppositional context, or in the third person.
Here are some relevant passages, whose significance is better appreciated by
keeping in mind Wittgenstein’s equation of meaning and use (PI 43, 560):
148 A. Coliva

Moore’s paradox can be put like this: the expression “I believe that this is
the case” is used like the assertion “This is the case”; and yet the hypothesis
that I believe this is the case is not used like the hypothesis that this is the
case. (PI, II, x, p. 190e)

Similarly: the statement “I believe it is going to rain” has a meaning like,

that is to say a use like, “It’s going to rain”, but the meaning of “I believed
then that it was going to rain”, is not like that of “It did rain then”. (ibid.)

Of course, we do not find this claim intuitive: semantic continu-

ity across changes of tense and person seems obvious to us. Here is
Wittgenstein’s response:
“But surely ‘I believed’ must tell just the same thing in the past as ‘I believe’
in the present!”—Surely √-1 must mean just the same in relation to -1, as
√1 means in relation to 1! This means nothing at all. (ibid.)

Hence, the idea seems to be that since √1 admits of two possible solu-
tions—1 and −1—both of which belong to real numbers, whereas √−1
does not admit of a solution within the domain of reals, “√” means some-
thing different in the two cases. Although the solutions of the application
of the function are different depending on whether it is applied to 1 or
−1, and indeed in order to provide a solution to √−1 we need to extend
numbers to imaginary ones whereas no such extension is needed in order
to provide a solution to √1, that does not mean that the function itself
is ambiguous, even if its application requires something very different of
us in the two cases. Thus, the analogy provided by Wittgenstein to sup-
port a strong form of contextualism about meaning is dubious. That does
not ipso facto prove his semantic contextualism wrong, though. Yet it can
make us suspicious of it. As we shall see, this is a thesis contemporary
expressivists do not generally hold.
Notwithstanding Wittgenstein’s strong form of contextualism about
linguistic meaning, there is something else worth noticing—namely, the
different function the same words can have depending on context, con-
ceding that they may retain the same linguistic meaning. He writes:

This is how I think of it: Believing is a state of mind. It has duration; and
that independently of the duration of its expression in a sentence, for
 6 Expressivism About Self-knowledge 149

example. So it is a kind of disposition of the believing person. This is shown

me in the case of someone else by his behaviour; and by his words. And
under this head, by the expression “I believe …” as well as by the simple
assertion.—What about my own case: how do I myself recognise my own
disposition?—Here it will have been necessary for me to take notice of
myself as others do, to listen to myself talking, to be able to draw conclu-
sions from what I say! (PI II, x, pp. 191e–192e)

Does it make sense to ask “How do you know that you believe?”—and is
the answer: “I know it by introspection”? In some cases it will be possible to
say some such thing, in most not. (PI 587)

What Wittgenstein seems to suggest here is that “S believes that P” is used

to ascribe a dispositional state of mind and that “I believe that P” may, on
occasion, though quite rarely, be used to describe one’s own dispositions
and be applied on the basis of the same criteria that govern third-personal
ascriptions of belief. In the vast majority of cases, however, “my own relation
to my words is wholly different from other people’s. That different develop-
ment of that verb would have been possible, if only I could say ‘I seem to
believe’” (p. 192e, cf. RPP II, 3). So, in some unusual cases, we would know
of our beliefs as dispositions in a third-personal way—that is, by inferring
from the observation of our behaviour, as well as of our words, to their likely
mental cause. In those cases, Wittgenstein notices, “it would also be possible
for someone to say ‘It is raining and I don’t believe it’, or ‘It seems to me
that my ego believes this, but it isn’t true’”. Hence, in these cases, Moore’s
paradox would disappear. Still, in the vast majority of cases, “I believe that
P” is used in an expressive way and, for him, is tantamount to asserting “P”,
albeit in a tentative voice, so to speak. This, as we shall see in the Appendix,
is wrong, even though it is right to remark upon the variety of uses of the
very same words. In particular, we can see how Wittgenstein comes close to
the distinction we drew between propositional attitudes as dispositions and
as commitments with the attendant distinction between first-personal and
third-personal self-knowledge. What seems to preclude its full endorsement
is Wittgenstein’s distinction between propositional attitudes as dispositions
and as occurrent mental states, with a distinctive phenomenology and his
insistence on the different kind of speech act performed, which, given the
150 A. Coliva

equation between meaning and use, comes down to the insistence on the
different meaning the same words can have. Here are some relevant passages:

When I sat down on this chair, of course I believed it would bear me. I had
no thought of its possibly collapsing”. But: “In spite of everything that he
did, I held fast to the belief …”. Here there is thought, and perhaps a con-
stant struggle to renew an attitude. (PI 575)

Similarly,

We say “I am expecting him”, when we believe that he will come, though

his coming does not occupy our thoughts (…). But we also say “I am expect-
ing him” when it is supposed to mean: I am eagerly waiting for him. We
could imagine a language in which different verbs were constantly used in
these cases. And similarly more than one verb where we speak of “believ-
ing”, “hoping”, and so on. Perhaps the concepts of such a language would
be more suitable for understanding psychology than the concepts of our
language. (ibid.)

When someone says “I hope he’ll come”—is this a report about his state of
mind, or a manifestation of his hope?—I can, for example, say it to myself.
And surely I am not giving myself a report. It may be a sigh; but it need
not. If I tell someone “I can’t keep my mind on my work today; I keep on
thinking of his coming”—this will be called a description of my state of
mind. (PI 585)

In PI 586, Wittgenstein makes similar observations and then goes on

to draw a difference between the “exclamation” ‘I’m longing to see him!”
and a different use of the same words: “But I can utter the same words
as the result of self-observation, and then they might mean: ‘So, after
all that has happened, I am still longing to see him’”. In PI 587, talking
about love, he writes: “It makes sense to ask: ‘Do I really love her, or am I
only pretending to myself?’ and the process of introspection is the calling
up of memories; of imagined possible situations, and of the feelings that
one would have if …”. In PI 588, discussing the role of self-ascriptions of
intentions, Wittgenstein writes: “‘I am revolving the decision to go away
tomorrow’ (This may be called a description of a state of mind.) (…). I
 6 Expressivism About Self-knowledge 151

say at the end of a quarrel ‘All right! Then I leave tomorrow!’; I make a
decision”. References could be multiplied ad libitum. They all go in the
direction of distinguishing between different uses or functions the same
words can have and, at least in the case of self-ascriptions of intentions,
between the self-ascription of a disposition and of (something close to)
a commitment, with the attendant diagnosis of their underlying episte-
mology. None in the latter case, and third-personal in the former.
We have now examined at length Wittgenstein’s expressivism. We
can thus sum up how it fares with respect to an adequate theory of
self-knowledge. In the case of so-called first-personal self-knowledge, it
respects groundlessness, indeed in a strong form, since it not only does
steer away from the observational and the inferential account of self-
knowledge but does not consider occurrent mental states as evidence or
reasons for the corresponding self-ascriptions either.
It also respects transparency, for on this account, any mental state
which can be expressed by means of an avowal is a mental state a subject
is aware of, even if it is not named, described or reported by means of
the corresponding self-ascription and even if its verbal expression presup-
poses a blind substitution of a piece of instinctive and primitive behav-
iour with a characteristic linguistic expression, which is considered the
immediate expression of it.
Finally, authority is respected as well, although it is seen as a prod-
uct of our linguistic practice of according it to subjects. That is to say,
once the C-conditions are satisfied, then there is no room left for doubt
regarding the fact that a subject does have the mental state she actually
avows. Whether authority is satisfied in its weak or strong version in
Wittgenstein’s account is not clear, because he gives no explanation of
self-deception. Yet, at the very minimum, it holds in its weak version.
The problems with Wittgenstein’s expressivism, however, are mostly
related to the kind of semantics of avowals it provides for, such as the
endorsement of a strong form of semantic contextualism and the dif-
ficulty in accounting for the fact that, after all, even when uttered in the
avowing mode, these sentences have a syntactical structure that makes
them truth-apt. In connection with avowals of belief and other propo-
sitional attitudes, moreover, it is not clear what kind of pre-linguistic
behaviour they would substitute. Finally, their equation with the asser-
152 A. Coliva

tion of their embedded contents is highly problematic, as we have already

seen in connection with Evans’s point and we will further see in discuss-
ing Moore’s paradox (see Appendix).

2 Bar-On’s Neo-expressivism
Bar-On, in her Speaking My Mind: Expression and Self-Knowledge (2004),8
has presented the best-developed expressivist account of self-knowledge
to date. Her view inherits the key idea of Wittgenstein’s expressivism—
that is, that psychological self-ascriptions are avowals, which express one’s
occurrent mental states and are pieces of linguistic behaviour that replace
more natural expressions of ongoing mental states. Yet she thinks this cen-
tral contention can be coupled with a semantic and an epistemic account
whereby the relevant self-ascriptions are truth-apt and are expressions of
privileged self-knowledge.
First, it is important to stress that the kind of more instinctive behav-
iour avowals substitute can be both non-verbal, as in the case of pain, and
verbal. Hence, “I want o” can be acquired by learning to use it instead of
one’s pointing towards an object o while pronouncing its name.9 More
importantly, in the case of self-ascriptions of propositional attitudes, like
belief, “I believe that P” will be acquired as a substitution for one’s assert-
ing that P, mostly in circumstances when one gets corrected by an adult
about the fact that P is not actually the case, even if it seems so to one.10
Second, Bar-On’s idea is that “I intend to φ” and “I believe that P”,
as avowals, express the relevant mental states but have a content which is
truth-evaluable. Consider an explicit performative like “I promise to take
you to the zoo”. It both expresses one’s promise and makes explicit the fact
that one has made such a promise. Hence, the resulting self-ascription is
truth-apt. In particular, it is true if the subject actually intends to take her
child, say, to the zoo and false otherwise. Similarly, according to Bar-On,
“I intend to φ” and “I believe that P” express one’s intention and belief,

8
Another prominent version of neo-expressivism can be found in Finkelstein 2003.
9
See Bar-On 2004, p. 288.
10
See Bar-On 2004, p. 294.
 6 Expressivism About Self-knowledge 153

respectively, and are true iff one actually intends to φ and believes that
P. We should then distinguish between the act of avowing one’s intention
or belief and the product of that act, which is a self-ascription that can be
either true or false.11 According to Bar-On, these self-ascriptions can also
express one’s judgement of intending to φ or of believing that P. Yet, even
in this case, the self-ascription is not epistemically based on the relevant
first-order mental state. That is, the mental state elicits one’s intentional
self-ascription and even one’s judgement of being in it and therefore is their
rational cause, but it is not a kind of evidence on which the judgement is
based.12 Notice, moreover, that contrary to explicit performatives “I intend
to φ” and “I believe that P” are not ways of bringing about the relevant
first-order mental states, according to Bar-On. Rather, they are expressive
of them. Hence, she refuses the idea that one’s psychological self-ascrip-
tions can play a constitutive role with respect to one’s first-order mental
states (save for cases of self-verifying judgements, like the ones we reviewed
in our presentation of Burge’s views in Chap. 5, §2). In this respect, she is
entirely with common sense in recognising the independent existence of
mental states from their self-ascription (save for self-verifying thoughts).
Still, recognising a subject’s act as an avowal—that is, as a self-expressive
act—entails “presuming her avowal to be true”.13 For, for an intentional
linguistic act to count as an avowal, it must be prompted by the mental
state avowed. In fact, according to Bar-On, the avowal literally shows the
relevant mental state and thus makes it visible or audible to one’s audience.14
Avowals are therefore characterised by a transparency-to-the-subject
and to-the-world condition. The latter is the analogue of Evans’s notion
of transparency. A subject does not have to look within herself to answer a
question about her belief that P (and other propositional attitudes) but at
the reasons for (or against) P (or its desirability, etc.). Similarly, if an audi-

11
I think this point is not sufficiently appreciated by Boyle (2009), who criticises Bar-On’s proposal
for not explaining the difference between merely parroting “I φ that P” (where φ has to be substi-
tuted with a propositional attitude verb) and actually representing oneself as φ-ing. Along Bar-On’s
lines, see also Finkelstein 2003, Chap. 4.
12
There will be more on how the mental state can play a rationalising role with respect to the self-
ascription in the following.
13
See Bar-On 2004, p. 317.
14
See Bar-On 2004, pp. 417, 422–423.
154 A. Coliva

ence wants to challenge a subject’s avowal, they will take issue with the
reasons for (or against) P itself, not with the fact that the subject has self-
ascribed that belief (or some other intentional mental state). The former,
in contrast, is the presumption of truth we have already reviewed together
with the more problematic claim (as we shall see) that avowals wear the
relevant mental states “on their sleeve”.15 Hence, avowals do not play the
role of mere symptoms of those mental states that supposedly cause them,
according to Bar-On. Thus, one’s audience does not have to infer from the
former to the latter through an inference to the best explanation. Rather,
an audience can directly perceive the mental state in a subject’s avowal.16
Therefore, mental states are “conditions the subjects are in, not states that are
in the subjects”, made directly manifest in their avowals.17
It is worth pointing out that, according to Bar-On, the presumption of
truth avowals enjoy can be overridden. Hence, she is in sharp disagreement
with more traditional expressivist views “that maintain that a subject’s avow-
ing a mental condition conceptually guarantees her being in the condition (so
the avowal must be true), as long as the subject is linguistically or conceptu-
ally competent and sincere”.18 Consider someone who is so scared of the
dentist that as soon as he approaches her with his instruments she screams
“I am in pain” or “It hurts”. Let us suppose that the patient is being fully
sincere. Yet she cannot be in pain, for she has not even been touched. Her
avowal is thus false. Still, according to Bar-On, it is not based on a misiden-
tification of one’s ongoing mental state as pain. The subject has not made
a mistake but has committed an “expressive failure”.19 As Bar-On puts it,

though [the subject] has successfully expressed pain, she has not succeeded
in expressing her pain. She could not have expressed her pain, since there
was no pain for her to express. The subject has used an expressive tool that
is referentially associated with one condition (…) to give vent to a different
condition.20

15
See Bar-On 2004, p. 315.
16
See Bar-On 2004, p. 423.
17
Bar-On 2004, p. 424.
18
Bar-On 2004, p. 325.
19
Bar-On 2004, p. 320.
20
Bar-On 2004, p. 323.
 6 Expressivism About Self-knowledge 155

Bar-On offers a similar analysis of self-deception. In her view, a subject

does not mistake one’s actual propositional attitude for a different one.
Rather, she avows a given propositional attitude, which, however, she is
not enjoying. Hence, she is guilty of an expressive failure, not of a recog-
nitional mistake.
Furthermore, when we utter negative avowals, like “I don’t want/like/
believe/think such and so”, we are speaking from complementary mental
states—that is to say, from the state of wanting/liking/believing/thinking
something else. Hence, the idea that avowals are immediately elicited
from some mental state or other is preserved, even when we utter a nega-
tive, yet true, self-ascription. Obviously, this is problematical, for there
seem to be cases in which the negative avowal is not based on enjoying
a different, incompatible mental state. In such a predicament, however,
according to Bar-On, we should consider the self-ascription a report
based on introspection and not an avowal.21
It must be noted that, according to Bar-On, we also engage in non-
evidential reportive avowals.22 We do so when we utter an avowal as a
result of being asked why we are doing something or what we are think-
ing of. These questions direct our attention towards some aspect or other
of our present psychological state and we issue the reportive avowal from
the avowed state. Given that, even reportive avowals enjoy a special secu-
rity; that is to say, they are presumed to be true unless they are overrid-
den by available evidence. Finally, we can engage in purely theoretical
self-reports based on applying to ourselves a third-personal point of view
to make sense of our observed behaviour. In that case, our psychological
self-ascriptions do not enjoy any special security. In particular, there is
no presumption that they should be true, since they are not elicited from
one’s being in the self-ascribed mental state.
Turning now to Bar-On’s account of how our avowals can be expres-
sions of distinctively first-personal self-knowledge, it should be registered
that she devotes an entire chapter to showing how expressivism would be

21
See Bar-On 2004, p. 335. This position is not very plausible. Suppose someone is being tortured
and screams “I don’t want this”. It would be weird to say that this is not an avowal but a judgement
based on introspection. Yet it would be equally weird to hold that it is based on some other positive
mental state like wanting something other than what one is being inflicted.
22
See Bar-On 2004, p. 301.
156 A. Coliva

compatible with numerous accounts of the epistemic credentials of our

psychological avowals, such as reliabilism and Burge-style entitlements.
Thus, Bar-On does not join Wittgenstein or other expressivists in denying
that there is a substantial epistemology of the mental. In her view, such a
deflationary stance depends on embracing a presupposition she rejects—
namely, what she labels the Distinct Epistemic Basis presupposition:

If I can be said to have privileged knowledge that I am in a certain state of

mind, then this knowledge must have some distinct epistemic basis; there
must be a special epistemic method or route (a ‘special way of knowing’)
that I use to obtain this knowledge.23

However, she thinks both reliabilism and Burge-style entitlements are

somewhat problematical, for the former does not account for the dis-
tinctively first-personal kind of knowledge we have of our own occurrent
mental states, nor for its special security. The latter, in contrast, deploys a
transcendental argument to the effect that self-knowledge is necessary to
engage in critical and practical reasoning but does not explain how that
knowledge comes about.24 Hence, she proposes a different source of our
entitlement to our psychological avowals, which—it should be remem-
bered—are avowing acts with a propositional content, which can itself be
the content of a judgement or a belief about one’s present state of mind.
According to such a view, the entitlement is a kind of warrant that does
not presuppose any robust cognitive achievement on a subject’s part (like
observing or inferring), nor does it require a subject to know of it or of how
it comes about. Furthermore, it is appropriately linked to the mental state
whose self-ascription is thereby warranted. For, in effect, it actually stems
from being in that very mental state—namely, the very mental state which
elicits one’s avowal. Hence, by being in a given mental state M and by
avowing it, if a subject believes that she is in M, she will also have an entitle-
ment for her belief to the effect that she is in M. Furthermore, her belief
will be true, for she is in that very mental state. Hence, she will know—and
know in virtue of being in M—that she is in M. Since only the subject her-
23
Bar-On 2004, p. 344.
24
For obvious reasons, Bar-On’s criticism targets Burge 1996 and does not consider Burge’s later
views on self-knowledge.
 6 Expressivism About Self-knowledge 157

self can be in M and therefore have the relevant entitlement, the account
does explain not only how she has knowledge of her mental states but also
why that knowledge is privileged—that is, available only to the subject
herself. To put it differently, the very truth-maker of one’s psychological
self-ascription is also what warrants it. Now, in the case of false but sincere
avowals, since they are not issued from being in M, they are not warranted.
It may seem so to the subject. Yet they are not. In this respect, expressive
entitlements are not like ordinary instances of justification or warrant that
may be present even if the beliefs eventually based on them are false.
In order to review Bar-On’s proposal, it is worth making use of our
usual template. Hence, we have seen that groundlessness is respected in
its weak form, for it is a tenet of the present account that avowals are
not based on observation or inference to the best explanation. It is also
respected in the more demanding sense that the first-order mental state
does not play the role of evidence for the corresponding self-ascription,
not even in the minimal sense invoked by Peacocke (see Chap. 5, §1).
However, as we have seen, that mental state functions as a non-evidential
warrant for the corresponding self-ascription and thus as its rational, as
opposed to merely brute, cause. Bar-On’s position on this score is, how-
ever, extremely problematic. Consider the case of false yet sincere avowals.
Take a subject who is so frightened of the dentist as to avow pain before
he even touches her mouth, and runs away. In such a predicament, we
would like to say that she did so—from her point of view—because she
thought she was in pain, even though—from our own point of view—she
will have done so out of fear rather than pain. Given Bar-On’s critique of
the idea that avowals are based on how things seem to one in one’s mental
arena, as it were, and her thesis that in such an event the self-ascription
of pain would be unwarranted, as well as false, it is left in the dark how
we could recognise some degree of rationality in the subject, at least from
her own point of view. To say that

we can still give sense to the intuition that there is a common element
shared by acts of avowing truly and acts of avowing falsely: both are acts of
speaking one’s mind whose characterisation as reasonable acts may require
mentioning the (possibly absent) self-ascribed state of mind25

25
Bar-On 2004, p. 395.
158 A. Coliva

does not help, since there is nothing, on the present account, in the sub-
ject’s mind which could warrant her false self-ascription of pain. Nor does
it help to say that “a false avowal may still be regarded as immune to errors
of misascription, as well as misidentification”.26 For, after all, it is a core
aspect of Bar-On’s proposal that avowals can be false and hence that the
presumption of truth they normally enjoy can be overridden. Finally, it
does not help to say that “even if we take an avowal to be false, and thus
take the subject to be wrong, we may still not take her to have gone wrong,
epistemically speaking”.27 For if the warrant is identical to the truth-maker
and the latter is absent, so is the former. Hence, Bar-On cannot conclude
that “false and true avowals may equally enjoy entitlement by default”.28
Furthermore, it is dubious that avowals are direct manifestations of
subjects’ mental states, which actually make the latter immediately per-
ceivable by one’s audience. To see how implausible it is, consider the fol-
lowing thought experiment. Suppose an alien arrives in a country where
there are human beings and zombies, who all behave alike both verbally
and non-verbally. First, it is not clear why an alien should see their pain,
as opposed to merely seeing their overt behaviour (let it be linguistic
or non-linguistic) and should learn to interpret it as expressive of the
relevant mental states. Second, how could he be said to see a zombie’s
pain, say, since, ex hypothesi, a zombie does not feel any pain? Yet, since
he would be behaving (either linguistically or non-linguistically) just as a
normal human being and in Bar-On’s account the latter would be directly
showing her pain, he should be showing his pain too.
Moving on to transparency, Bar-On is extremely cautious in her
remarks. She thinks it is entirely possible to be self-blind creatures—that
is to say, creatures who do have mental states, even conscious ones, and
yet cannot have knowledge of them. Surely, she does not wish to make
self-knowledge constitutive of having conscious mental states,29 nor is she
attracted to the view that, if possessed of the relevant concepts and cogni-
tively lucid and alert and if a subject has a given mental state, she will be
in a position to believe or judge that she does. Still, she is hospitable to the

26
Ibid.
27
Bar-On 2004, p. 396.
28
Ibid.
29
See Bar-On 2004, p. 406.
 6 Expressivism About Self-knowledge 159

idea that a subject endowed with the relevant concepts, while cognitively
lucid and alert, may direct her attention so as to pass judgement over her
ongoing mental states, like in the case of non-evidential reportive avowals.
Moreover, she thinks that given the presumption of truth our avowals
enjoy, authority is respected. Yet, given that she allows for cases of false
avowals and in particular of self-deception, as we have seen, authority, in
her view, holds only in its weak version. In this connection, however, it
is worth stressing that Bar-On’s defence of the idea that avowals may be
false yet sincere is highly problematic. For, if it is constitutive of avowals
that they are taken to be elicited from the mental state avowed, it should
be said that there was no real avowal in the first place, when there is no
corresponding mental state. Even more so if one holds, as Bar-On does,
that the mental state is not merely the cause—that is, the symptom—of
the mental state but somehow embodies it. For, in the case of false (yet
sincere) avowals, no mental state is present to elicit and be embodied
in one’s words. If there is none, then one’s words may resemble avow-
als, but they are not. They are, at best, an attempt at avowing a mental
state. Saying that the subject nevertheless succeeds in expressing a mental
state, though not her mental state, is very confusing. Surely, she did not
express someone else’s mental state. Moreover, it is true that we may take
the failed avowal of pain as a symptom of the subject’s fear of the dentist.
However, if we construe this as meaning that she has managed to express
a mental state of hers, though not the one actually avowed, this raises the
question of how come the subject has used a self-ascription of pain, say,
to give vent to a mental state of fear of which, clearly, she is not aware
as such. Thus, the idea that there is a mental state M that gives rise to a
self-ascription of a mental state N suggests that the subject has wrongly
recognised her own mental state—that she has after all mis-taken M for
N—contrary to what Bar-On wants to maintain.

3 Summary
In this chapter, we have analysed early and recent expressivist accounts of
self-knowledge. We have seen how they strongly reject the idea that our
knowledge of our own minds is based on a robust cognitive achievement
and how they hold, in contrast, that psychological avowals are acquired
160 A. Coliva

as alternative ways of giving immediate expression to one’s mind. We

have seen how, in Wittgenstein, this idea is coupled with very strong
views about the semantics of avowals, which are seen as discontinuous
with respect to third-personal psychological ascriptions and with past-
tense first-personal ones, in point both of truth-aptness and of meaning.
Furthermore, we have seen how, for Wittgenstein, denying that psycho-
logical avowals are the result of a robust cognitive achievement means
denying that there is self-knowledge properly so regarded. What goes
by that name, rather, are the ideas of groundlessness, transparency and
authority that are a by-product of “grammar”—that is to say, of how our
distinctive language games related to mentality are actually played and
thus of the rules that govern them.
We then moved on to presenting and discussing Bar-On’s recent ver-
sion of expressivism. We have found it superior to Wittgenstein’s in its
semantic respects. For, according to Bar-On, avowals are linguistic acts
with a semantic content, which is truth-apt, and there is no endorse-
ment of the idea that the very same mental terms may have altogether
different meanings when used in different contexts. We have taken issue,
however, with her further idea that avowals can indeed express a special
kind of knowledge subjects would have with respect to their own men-
tal states. For the very truth-maker of the self-ascription would serve as
its warrant, thus rendering false yet sincere avowals unwarranted from a
subject’s point of view and not merely false. Also Bar-On’s idea that there
can be genuinely false yet sincere avowals has been found wanting, for
it seemed a constitutive aspect of avowals that they were caused by, and
showed “on their linguistic sleeve”, the relevant mental state. Absent the
latter, the very identity of the linguistic performance as an avowal would
be impaired too.
For all these reasons, it appears that expressivism as an all-encompassing
theory about knowledge of all kinds of mental states we may enjoy is
problematical, despite having several attractive elements to it. We will
come back to them in the final chapter. Before doing so, however, we
have to examine some alternative views.
 6 Expressivism About Self-knowledge 161

Bibliography
Bar-On, D. (2004). Speaking my mind. Expression and self-knowledge. Oxford:
Oxford University Press.
Boyle, M. (2009). Two kinds of self-knowledge. Philosophy and Phenomenological
Research, 77(1), 133–164.
Burge, T. (1996). Our entitlement to self-knowledge. Proceedings of the
Aristotelian Society, 96, 1–26.
Coliva, A. (2010). Moore and Wittgenstein. Scepticism, certainty and common
sense. London: Palgrave Macmillan.
Finkelstein, D. (2003). Expression and the inner. Cambridge MA: Harvard
University Press.
Jacobsen, R. (1996). Wittgenstein on self-knowledge and self-expression.
Philosophical Quarterly, 46, 12–30.
Wright, C. (2004). Warrant for nothing (and foundations for free)? The
Aristotelian Society Supplementary Volume, 78, 167–212.
 7
Constitutive Theories

So far, we have seen that first-personal self-knowledge can be based

neither on observation nor on inference and that alternative ways of
accounting for it as the result of a sui generis kind of epistemic achieve-
ment are ultimately unsatisfactory. Because we do not have any other way
of knowing truths, we should conclude—with Crispin Wright and Paul
Boghossian1—that it is based on nothing. What this means is that so-
called self-knowledge is not a kind of epistemic achievement after all and
it is somehow a misnomer to call it “knowledge” if knowledge is under-
stood as the result of a, however minimal, epistemic endeavour which
relates a subject on the one hand and a true proposition or a fact on the
other, in such a way that the latter is somehow correctly and justifiably
apprehended by the former. Rather, what we call “self-knowledge”—that
is, the distinctive kind of authority we recognise to our fellow humans
and to ourselves over our mental states as well as the distinctively ground-
less and transparent way in which we are aware of them—is guaranteed to
hold a priori, as a matter of conceptual necessity. These claims are at the
heart of so-called “constitutive” positions.

1
See Wright 2001c, p. 310; Boghossian 1989, p. 5.

© The Editor(s) (if applicable) and The Author(s) 2016 163

A. Coliva, The Varieties of Self-Knowledge,
DOI 10.1057/978-1-137-32613-3_7
164 A. Coliva

These theories therefore inherit an aspect of Wittgenstein’s position

we reviewed in the previous chapter, according to which, if it does not
make sense to question one’s psychological self-ascriptions, at least in the
normal run of cases, and thus there is no room for not knowing them,
it means that they are not an expression of genuine knowledge either.
It simply means that the kind of security they enjoy does not depend
on the subject’s being in an epistemologically privileged position, which
makes her right about her own mental states. Rather, it belongs to the
“grammar”—that is to say, to the rules—of our language game of making
psychological self-ascriptions, that subjects are accorded authority over
their own mental states. Our language thus turns out to be confusing, for
we ordinarily talk of knowledge of our own mental states, in the relevant
cases. This prompts us to try to give an epistemic account of the kind of
special security psychological avowals enjoy, when in fact they are not
the result of any special way of knowing our first-order mental states and
their security is, in effect, a characteristic trait of our language games.
Constitutive theories build upon this insight, yet they come in various
fashions and while they all subscribe to the no-knowledge thesis, they dif-
fer with respect to the kind of metaphysics they subscribe to. For there are
two further metaphysical theses they may hold. The first one is that first-
and second-order mental states are not separate existences. The second
one is that, at least to some extent, psychological self-ascriptions actually
give rise to, or constitute, one’s first-order mental states. Hence, within the
constitutive camp, we can distinguish between weaker and stronger posi-
tions, depending on whether only the no-knowledge thesis is endorsed or
whether one or both of the metaphysical claims are upheld on top of it.
Since, nowadays, very few theorists would subscribe to the view that
we constitute sensations and perceptions or simple basic emotions by
judging that we are enjoying them, to facilitate the presentation and com-
parison among the various constitutive positions, we concentrate here on
what constitutive theorists have to say about our knowledge of proposi-
tional attitudes only. With these specifications in hand, we can actually
see how all constitutive theorists agree that the following (scheme of a
thesis) is an a priori conceptual truth.

Constitutive thesis: Given C, one believes/desires/intends that P/to φ iff

one believes (or judges) that one believes/desires/intends that P/to φ.
 7 Constitutive Theories 165

Constitutive theorists, however, diverge with respect to not only

the  metaphysical import of  constitutivism but also regarding  which
side of the constitutive thesis must be accorded priority over the other
one, if any; they also differ with respect to the rationale they provide
for the constitutive thesis and, in particular, for the C-conditions within
which the relevant biconditional is supposed to hold. Thus, according
to Sydney Shoemaker (§1), who subscribes to the no-separate-existence
claim, priority must be given to its left-to-right side and the C-conditions
must be characterised by reference to subjects who possess normal intel-
ligence and rationality and are endowed with the relevant psychologi-
cal concepts. According to Wright (§2), whose metaphysical views seem
congenial to the metaphysical claim that the relevant psychological self-
ascriptions bring about the corresponding first-order mental states but
who—arguably—actually inclines towards a form of irrealism, the right-
to-left side is the fundamental one and the C-conditions must refer to the
communal linguistic practice of making psychological avowals, which are
usually taken as authoritative. Finally, according to Akeel Bilgrami (§3),
who is silent on metaphysical issues, the two sides of the biconditional are
on a par and the C-conditions must make reference to the fact that the
mental states at issues are such that it makes sense to regard the subject
as responsible for them—that is, to be either blame- or praise-worthy
for them. We examine each of these positions and find them wanting
because, apart from several problems of detail, they do not really vin-
dicate the metaphysical claims at the heart of a significant constitutive
position. Moreover, they tend to show that self-knowledge is a neces-
sary condition for various kinds of activity, but they say nothing or very
little about it can actually come about. Yet, even if one does not think
there is a distinctive epistemology, unless one endorses Wittgenstein’s
anti-explanatory meta-philosophical pronouncements, one cannot leave
things at that. In particular, it is argued that an account is due of how
we acquire and canonically deploy the relevant psychological concepts,
which figure in psychological self-ascriptions that exhibit groundless-
ness, transparency and authority—that is to say, the features which first-
personal self-knowledge reduces to.
We then introduce a metaphysically robust brand of constitutivism
(§4), which is claimed to hold for only a very limited class of mental
states—namely, for those propositional attitudes as commitments we
166 A. Coliva

undertake by deliberating what to believe, desire, intend to do, and so on,

on the basis of evaluating (or at least of being able to evaluate) evidence in
favour of P/φ-ing or of its desirability or advisability. When these propo-
sitional attitudes are at stake and the subject is endowed with the relevant
psychological concepts, which are acquired “blindly”,2 both sides of the
biconditional hold as a matter of conceptual necessity, and, in particular,
the right-to-left side actually makes good the second metaphysical claim
at the heart of proper constitutive accounts. What this means is that adult
human beings actually have two ways of forming commitments. Namely, 
either by judging their contents or by directly self-ascribing them. In the
latter case, then, authority is secured in a much stronger way, since the
psychological self-ascription is actually self-verifying. Furthermore, the
distinctive account of how we acquire and canonically deploy the relevant
psychological concepts dovetails nicely with the first metaphysical claim
central to constitutive positions properly so regarded—namely, that first-
and second-order mental states are not separate existences.
Such a position is then defended against the objection that we may be
self-deceived and ascribe to ourselves propositional attitudes we in fact
lack. The key move, following Bilgrami’s lead, consists in denying that
self-deception is a case in which one goes wrong about one’s first-order
mental states. Rather, it consists in having two mutually inconsistent
propositional attitudes—one as a commitment and one as a disposi-
tion—which give rise to a subject’s somewhat irrational behaviour. Yet
one’s self-ascription of the commitment is actually correct, even if one
happens to behave in ways which run contrary to it, because of one’s
counter dispositions.
Nevertheless, it is clear that the proposed constitutive account holds
for only a very limited class of mental states—namely, those as commit-
ments. Therefore, it is necessary to integrate it with different explanations
of our knowledge of the other kinds of mental states we enjoy. We take
up this task in the final chapter.

2
That is to say, by being drilled to substitute their immediate avowal, “P”, “P would be good to
have”, “I will φ”, with the corresponding psychological one—that is, “I believe that P”, “I want/
desire that P”, “I intend to φ”.
 7 Constitutive Theories 167

1 The Left-to-Right Side of the Constitutive

Thesis: Shoemaker
Shoemaker has been one of the first and main proponents of a constitu-
tive thesis, which he developed in opposition to the observational account
of self-knowledge, both in its early—Cartesian—form and in its materi-
alist development, put forward by David Armstrong (see Chap. 4, §1).
Shoemaker’s interests emphasise and mainly concern the left-to-right side
of the constitutive thesis, according to which if one has a given mental
state M, in particular a propositional attitude, one believes that one does.
Of course, Shoemaker is well aware of the existence of unconscious men-
tal states, of self-deception and of the fact that only creatures endowed
with the relevant conceptual repertoire could be in a position to make the
relevant self-ascriptions. He therefore puts forward several constraints on
the C-conditions for the constitutive thesis. In particular, he maintains
that that thesis holds for a “rational agent”, meaning a “person with nor-
mal intelligence, rationality and conceptual capacity”.3 To establish the
thesis on a priori grounds, Shoemaker asks us to conceive of a self-blind
person—that is, someone who has no cognitive impairment and has the
relevant conceptual repertoire yet who, ex hypothesi, does not know “from
the inside”, her beliefs, desires, intentions and further mental states, such
as sensations and colour perceptions. She would be acting as a normal
subject does, yet she would not be able to answer directly the question
whether she believes, desires, intends, wants, sees or feels such-and-so.
To do so, she would have to look at her behaviour and infer that she did
what she did because she felt, saw, wanted, intended, desired and believed
thus-and-so. In such a case, according to Shoemaker, we should conclude
that “maybe the body in question houses two persons, or two ‘selves’ or
‘personalities’”.4 One personality would be the “agent” and the other one
the “agnostic”.5 Now, according to Shoemaker, this understanding of the
case does not show that the agent and the agnostic are blind to them-
selves. It shows only that, like in a case of knowledge of other minds, the

3
Shoemaker 1996a, p. 236.
4
Shoemaker 1996a, p. 234.
5
Ibid.
168 A. Coliva

agnostic has to infer to the mental states of the agent starting from the
observation of the latter’s (linguistic and non-linguistic) behaviour. If,
then, there are reasons for us to think that the agent and the agnostic are
one single subject because the agnostic’s beliefs and desires rationalise the
agent’s actions, then the agnostic too will be in a position to know that
his beliefs and desires rationalise the agent’s actions.

For it is the agnostic’s introspective knowledge of her own beliefs and

desires, as manifested in her utterances, that provides the basis of our rea-
soning; so, necessarily, the agnostic knows at least as much about her beliefs
and desires as we do. On the other hand, if she didn’t have this introspec-
tive knowledge, neither we nor she could even begin the enterprise of find-
ing out whether the agent’s actions are hers.6

Notice that this argument rests on the premise that it is only if the
agnostic makes the relevant psychological self-ascriptions that we can
start making sense of the agent’s behaviour as in fact identical to the
agnostic’s. This, however, is clearly not necessary. In fact, the agnostic
could just say things such as “It is raining. It is bad to get wet” while the
subject opens the umbrella. This would give us reasons to think that the
agnostic and the agent coincide because the relevant first-order mental
states ascribed on the basis of what they say and do, without the need of
any psychological self-ascription on their part, rationalise their behaviour
and give it the necessary unity to make us think that it is performed by
a single subject.
Of course, for the agnostic to know that the agent’s mental states coin-
cide with hers, she would have to have access to her own. But we could
imagine, in keeping with the conditions of this thought experiment, that
she could know of them inferentially. That is to say, she would know that
she said or judged “It is raining. It is bad to get wet” while opening the
umbrella, and she could then infer, by reflection on this kind of overt
linguistic behaviour, that she believes that it is raining and desires not to
get wet. Hence, nothing so far in the argument presented by Shoemaker
requires direct or immediate self-knowledge.

6
Shoemaker 1996a, p. 235.
 7 Constitutive Theories 169

More generally, however, it does not seem necessary, in order to engage

in practical syllogisms, belief revision and other forms of critical reason-
ing, that one knows one’s beliefs, desires and further propositional atti-
tudes. As we saw in connection with Burge’s proposal (see Chap. 5, §2),
it seems enough to be able to have them and to orchestrate them in the
appropriate manner. Again here is Peacocke’s (1998, p. 276) example:

Suppose you come home, and see that no car is parked in the driveway. You
infer that your spouse is not home yet… Later, you may suddenly remem-
ber that your spouse mentioned in the morning that the breaks of the car
were faulty, and wonder whether she may have taken the car for repair. At
this point, you suspend your original belief that she is not home yet. For
you come to realise that the absence of the car is not necessarily good evi-
dence that she is not home. If the car is being repaired, she would have
returned by public transport. Then finally you may reach the belief that she
is home after all, given your next thought that she would not have taken
any risks with faulty breaks.

Of course, we often do engage in critical reasoning by asking ourselves

whether our beliefs are correct and our desires and intentions appropri-
ate, and, in order to answer these questions, we consider whether it is the
case that P or whether Q would be good or worth having. If that is what
is meant by “critical reasoning”, then it is clear that it requires knowledge
of one’s propositional attitudes because it is built into this “thick” notion
of critical reasoning that it is a reflection on the propriety of one’s beliefs,
desires and intentions. This is how Shoemaker understands it, even if the
argument provided does not require such a robust form of reasoning.7 As
we have just seen, there is a more minimal, or “thin”, sense of “critical
reasoning” whereby we simply revise beliefs/desires/intentions and so on,
in light of countervailing evidence or further practical considerations. Or
we simply act in certain ways as a result of executing a piece of practical
reasoning, by appropriately orchestrating the relevant first-order mental
states, without any need of self-ascribing them. That is, given my first-
order belief that it is raining and my first-order desire not to get wet (plus
knowledge of the fact that I carry an umbrella and that an umbrella can
7
See Shoemaker 1996a, p. 240.
170 A. Coliva

protect me from the rain), I open the umbrella as soon as I get out of my
office and I do so without having to self-ascribe these beliefs and desires
to myself, but just by having them and by letting them shape my further
behaviour. Although “thick” critical reasoning is certainly interesting and
characteristic of a lot of adults’ psychological lives, it is quite clear that
it cannot be taken to be the basic notion in terms of which we could
advance theses about the rationality of human beings in general and, in
this context, as the origin of our self-knowledge.
Let us then focus on “thin” critical reasoning, which, in contrast, seems
to be more basic and constitutive of what it means to be a rational sub-
ject. Of course, being able to engage even in “thin” reasoning requires
some kind of sensitivity not just to the contents of one’s beliefs and desires
but also to the kind of mental states they are contents of. For it makes
a difference whether I believe it is raining or whether  I desire it, vis-à-
vis the practical reasoning that would eventually lead me to opening the
umbrella. Yet it does not seem incredible to suppose that a physical system
is so engineered that (other things being equal) it opens the umbrella only
when the content “It is raining” appears in her belief-box, after seeing the
rain fall down, and the content “Not get wet” appears in her desire-box.
One may object to this reconstruction by noticing that if this is indeed
a piece of reasoning the subject is doing at the personal, rather than at the
subpersonal, level, she should be aware of the fact that the former is the
content of a belief and the latter of a desire. Hence, she should have tacit
knowledge of the mental states she is enjoying. That poses the problem
of understanding what one means by “tacit” knowledge. According to
Shoemaker, a subject has tacit knowledge of her belief that P “just in case
it is as if A has an explicit belief in P”—that is, “A’s cognitive dispositions
are relevantly as if A has an explicit belief in P”.8

My claim is that to the extent that a subject is rational, and possessed of the
relevant concepts (most importantly, the concept of belief ), believing that
P brings with it the cognitive dispositions that an explicit belief that one
has that belief would bring, and so brings with it the at least tacit belief that
one has it.9
8
Shoemaker 1996a, p. 241.
9
Ibid.
 7 Constitutive Theories 171

There are two strands to this claim. The first one is that if one believes
that one believes that P, one will be disposed to assert “P”, to use P as a
premise in one’s reasoning and to hold P true. These are just the same dis-
positions one would have if one had the first-order belief that P. The sec-
ond sense, however, is that just by being a rational agent in Shoemaker’s
sense of the word, one will be disposed to make the relevant psychological
self-ascription. This is not a disposition one would have just by having
the first-order belief, however. In effect, Shoemaker’s gloss on the notion
of tacit knowledge comes down to this: if a subject is a rational agent,
she has tacit knowledge of her first-order belief that P, just in case she
behaves like someone who believes that P and is disposed to self-ascribe
that belief. Yet, to count as having that disposition, one should, at least
on occasion, make that kind of psychological self-ascription.
Hence, the argument used to establish the necessity of self-knowledge
for critical and practical reasoning need not depend on the strong claim
that these forms of inference necessarily involve a reflection on the sound-
ness of one’s propositional attitudes. It could allow, that is, for a “thin”
notion of inferring and reasoning in a critical way, or in accordance with
practical syllogisms, just as long these inferences are supposed to take
place at a personal, rather than at a subpersonal, level and tacit knowledge
of one’s attitudes is explained along the lines just reviewed—that is to say,
as consisting in being capable, at least on occasion, to self-ascribe the rele-
vant mental states. Notice, however, that tacit self-knowledge thus would
presuppose explicit self-knowledge. If this were the correct account of
tacit self-knowledge, it then would turn out that even being critical rea-
soners in a “thin” sense presupposes being capable of self-ascribing one’s
propositional attitudes. Shoemaker’s claim would be (at least condition-
ally) vindicated; a rational agent—that is, someone possessed of normal
intelligence, rationality and the relevant concepts—cannot be self-blind.
However, it is doubtful that this is the correct account of tacit knowl-
edge. In particular, it explains tacit knowledge in terms of explicit
knowledge and thus it  seems to deprive the former of any theoretical
significance. Furthermore, it cannot be used to explain (explicit) self-
knowledge, for ultimately it presupposes it. But if having tacit knowl-
edge of one’s first-order beliefs reduces to the first strand to that notion
(namely, to being disposed to using P as a premise in one’s reasoning, to
172 A. Coliva

holding it true and to asserting it in the appropriate circumstances), then

it would be built into the very notion of having a given first-order belief
that one would have to have—albeit tacit—knowledge of it. Having
a belief and having (tacit) knowledge of it would amount to the same
thing. This is a conception of the mind as (tacitly) transparent to itself
that Shoemaker associates with Cartesianism and aims to reject.10 Thus,
it cannot be his considered view on the matter. Hence, so far, we have not
found a way of vindicating Shoemaker’s claim, based on considerations
regarding the conditions for being critical reasoners at all, that the very
idea of a self-blind subject is a conceptual solecism. Of course, this is a
conditional result. For one could buy into this form of Cartesianism or
provide a sound account of tacit self-knowledge that, in turn, could show
that being critical reasoners in a thin sense requires knowledge of one’s
attitudes.
Alternatively, one may propose to take tacit knowledge as basic and
as deployed whenever we are able to orchestrate the inferences which
characterise “thin” critical reasoning. One then could hold that tacit self-
knowledge is a necessary condition of the latter and therefore claim to
have vindicated the idea that being critical reasoners in the thin sense
requires self-knowledge—albeit of a tacit variety. Now, although this
move is certainly available, it leaves too much in the dark. In particular,
it does not tell us where tacit knowledge of our mental states comes from
and it does not clarify whether it is ultimately built into the very notion
of propositional attitudes one is operating with that they should figure
in inferences so as to require tacit self-knowledge. Hence, it seems that if
at the basis of an account of self-knowledge we have to buy into such a
mysterious notion, we would not have made much progress.
Be that as it may, Shoemaker (1996a, pp.  239–240) famously con-
cludes by saying:

From an evolutionary perspective it would certainly be bizarre to suppose

that, having endowed creatures with everything necessary to give them a

10
See Shoemaker 1996a, p. 224. I suspect that Boyle’s account of tacit self-knowledge amounts just
to such a view. See, in particular, Boyle 2011a (cf. Boyle 2011b). Yet this account seems to me in
tension with Boyle’s position in Boyle 2009 (§5.2), where tacit self-knowledge seems to depend on
the ability, at least on occasion, to self-ascribe the relevant mental states.
 7 Constitutive Theories 173

certain very useful behavioural repertoire—namely that of creatures with

normal human intelligence, rationality, and conceptual capacity, plus the
ability to acquire first-order beliefs about the environment from sense per-
ception—Mother Nature went to the trouble of installing in them an addi-
tional mechanism, a faculty of Inner Sense, whose impact on behaviour is
completely redundant, since its behavioural effects are ones that would
occur anyhow as the result of the initial endowment.

As we have seen, however, the argument aiming at showing that self-

knowledge is necessary for “thin” critical reasoning is not satisfactory and
the idea that it is necessary for “thick” critical reasoning, though correct
as such, cannot carry much weight since it clearly leaves out too many
cases in which there would be self-knowledge and yet not this kind of
critical reasoning. Furthermore, even if an argument managed to estab-
lish the required connection between self-knowledge and “thin” critical
reasoning, it would simply show that the former is a necessary condition
of the latter, but it would not show how self-knowledge comes about.
What we need from a constitutive account is to show in detail how, sim-
ply by having the relevant first-order mental states and a normal degree of
intelligence, rationality and conceptual capacity, one would be in a posi-
tion to make the relevant self-ascriptions without having to do any sub-
stantial epistemic work. As we shall see in §4, this is doable, at least for a
certain class of propositional mental states. Yet Shoemaker’s account does
not fill in these fundamental details. Before turning to other constitutive
proposals, however, we need to tackle one last aspect of his proposal—
namely, his defence of the idea that one’s propositional attitudes and one’s
beliefs about them are not distinct existences. Here is the relevant pas-
sage, which is worth quoting in full:

Instead of the first-order state and the belief about it having different core
realisations but overlapping total realisations, it might be that they have the
same core realisation and that the total realisation of the first-order state is
a proper part of the total realisation of the first-person belief that one has
it. (…) What my earlier discussion suggested is that if one has an available
first-order belief, and has a certain degree of rationality, intelligence, and
conceptual capacity (…) then automatically one has the corresponding
second-order belief. If it is possible to have the available first-order belief
174 A. Coliva

without having the second-order belief, this is because it is possible to have

it without having that degree of rationality, intelligence and conceptual
capacity—which is perhaps the case with some lower animals. But on this
conception, all you have to add to the available first-order belief, in order
to get the second-order belief is the appropriate degree of intelligence, etc.
It is not that adding this pushes the creature into a new state, distinct from
any it was in before (…). It is rather that adding this enables the core-
realisation of the first-order belief to play a more encompassing role.11

Hence, at an ontological level, if it can be shown that forming beliefs

about one’s first-order propositional attitudes requires nothing more than
a normal degree of intelligence, rationality and conceptual capacity, there
is no need to posit that the relevant first-order and second-order men-
tal states are differently realised in one’s brain. One single physical state
could play double duty: it could motivate action—also the speech act
of asserting “P”—and enter inferences while also giving rise to “thick”
critical reasoning, when required, and to assertions such as “I believe that
P”, for creatures who had the relevant concepts (plus a normal degree of
rationality and intelligence).
This is just an application of Ockham’s razor, and, as such, it would be
welcome: why posit a rich ontology if a more parsimonious one would
do just as well? Yet it remains to vindicate the idea that self-knowledge
requires merely having the relevant first-order mental states, conceptual
capacities and a normal degree of intelligence and rationality.

2 The Right-to-Left Side of the Constitutive

Thesis: Wright
In a number of writings, Crispin Wright has defended a constitutive
position with respect to our knowledge of propositional attitudes.12
To appreciate the gist of the proposal, it is important to place it in its
11
Shoemaker 1996a, pp. 243–244. The idea then is that the canonical inferences constitutive of a
given mental state are also constitutive of the second-order one, while the latter can have a richer
causal impact.
12
See also Heal 2002, pp. 1–19 for similar views. In fact, Wright defends constitutivism also with
respect to what he calls “phenomenal avowals”—that is, self-ascriptions of sensations and experi-
ences. In his view, they are strongly authoritative. For to doubt them would mean to doubt either
 7 Constitutive Theories 175

proper context—namely, Wright’s interpretation of Wittgenstein’s rule-

following considerations, his later philosophy of mind and his quietist
metaphilosophical pronouncements. Wright agrees with Saul Kripke’s13
view according to which, for Wittgenstein, there is no pre-determined,
mind-independent fact one should grasp when one understands a given
symbol, like “+”, which would constitute its meaning, by fixing once and
for all the rule it would express, in such a way that certain applications
of it would count as correct and some others as incorrect. Such a fact
would be a piece of Platonic mythology that would make it mysterious
how minds like ours could actually grasp it. Nor is the meaning of “+”
to be given by a subject’s disposition to use it a certain way, for there
is nothing in the disposition that precludes the possibility that, from a
certain moment onwards, one uses “+” differently from what one has
done up to that moment. Similarly, the meaning of “+” cannot reside
in a subject’s interpretation of it, for, once again, nothing prevents the
possibility of interpreting that symbol a certain way up to a certain point
and then of using it differently afterwards. If these are the only options,
we should conclude with Kripke’s Wittgenstein that there is no fact at all
about what “+” means. We should therefore embrace a form of semantic
scepticism, which would find relief merely in a “sceptical solution” of the
kind proposed by Kripke himself. According to that solution, it is only
the agreement in the community to use “+” a certain way that determines
what it means, by establishing which rule it expresses and hence which
are to count as correct or incorrect applications of it.
Now, Wright objects both to Kripke’s interpretation of Wittgenstein’s
rule-following considerations and to Kripke’s sceptical solution. Yet this is
not the place to pursue these matters further. What is important to note,
however, is that, in order to avoid the previous trilemma and Kripke’s
solution, Wright puts forward an alternative response to the sceptical
paradox about meaning. This alternative response is, in fact, a statement
of his constitutive position, which he thinks encapsulates Wittgenstein’s
main insights into both the rule-following issue and self-knowledge.

the conceptual and cognitive competence or the sincerity of the subject. As we shall see, in contrast,
attitudinal avowals, for Wright, are only weakly authoritative.
13
See Kripke 1982.
176 A. Coliva

According to Wright’s Wittgenstein, meaning “+” a certain way is a

subject’s (semantic) intention. When one pronounces over one’s inten-
tions (semantic or otherwise), one does so by means of avowals, like “By
‘+’ I mean thus and so”. Avowals, for Wright, are groundless, transparent
and authoritative. They are groundless insofar as they are based neither
on observation of one’s mental states nor on inference to the best expla-
nation starting from pieces of overt behaviour. Moreover, they are such
that, if asked to justify them, one could not offer any independent rea-
sons for them. Hence, groundlessness holds for them in its strong version.
Avowals of intentions are transparent for, provided one has the relevant
conceptual repertoire and is sincere and cognitively lucid, one will imme-
diately be in a position to make them. Finally, they are authoritative for
we do not normally question them, unless we have positive reasons to do
so—that is to say, unless we have reasons to think a subject is being insin-
cere, conceptually incompetent, cognitively unfit or self-deceived. Hence,
avowals of intentions (as well as of other propositional attitudes) are only
weakly authoritative, according to Wright.14 Since, moreover, intentions
may be general, they may possess “potentially infinite content”15—that
is, they can extend to all possible uses of that particular symbol. Hence, if
challenged to justify the claim that I meant or mean “+” a certain way, if
my avowals of present and of past (semantic) intentions are authoritative
(provided there are no positive reasons to discount them), there simply is
no further issue about what I mean or meant by “+”.16
This is not the place to evaluate either the soundness of Wright’s read-
ing of Wittgenstein or the prospects of the proposal vis-à-vis the rule-
following problem.17 Let us therefore concentrate on the account as a
14
By contrast, phenomenal avowals are strongly authoritative, in his view. See Wright 2001e,
pp. 320–321, 2015, for to doubt them would mean to doubt either a subject’s sincerity, or her
conceptual and cognitive competence.
15
Wright 2001a, p. 126.
16
See Wright 2001b, p. 206.
17
Personally, I have doubts on both scores. For a different and, to my mind, more accurate reading
of Wittgenstein on the rule-following problem, see Hacker (1986). In regard to the prospects of the
proposed solution, it seems powerless to fend off the challenge raised by Kripke. For the issue was
whether by using “plus” like we ordinarily do up to “100 + 1”, say, there is anything which compels
the subject to go on the same way, rather than returning the answer “102”, from that point in the
series onwards. Now, the fact that she can authoritatively say that by “+” she means “plus” does not
prevent her from returning the answer “102”, when she reaches the relevant point in the series. For
 7 Constitutive Theories 177

proposal about self-knowledge of intentions and further propositional

attitudes.
Regarding the source of their authority, in a Wittgensteinian spirit,
Wright proposes the following:

[T]he authority standardly granted to a subject’s own beliefs or expressed

avowals, about his intentional states is a constitutive principle: something
which is not a consequence of the nature of those states, and of an associated
epistemologically privileged relation in which the subject stands to them,
but enters primitively (my emphasis) into the conditions of what a subject
believes, hopes and intends.18

Thus, the authority of attitudinal avowals is not due to a kind of privi-

leged epistemic access subjects have with respect to their own mental
states. For that would mean to fall prey to a Cartesian conception of
the mind Wittgenstein wanted to dismantle. Rather, their authority is
a feature of the “grammar”—in Wittgenstein’s sense of that term19—of
the psychological vocabulary here under consideration. Now, grammar
is autonomous for Wittgenstein; that is, it does not respond to any pre-
ordinate fact. Thus, the fact that our concept of intention (and of other
propositional attitudes) works a certain way—in particular, is such that
one’s pronouncements over one’s intentions normally go unchallenged
unless there are positive reasons not to do so—is to be explained neither
by reference to the nature of the mental state in question nor by reference
to a peculiar epistemology. Rather, it is simply to be acknowledged—
this is how the language game is played.20 Nothing is hidden when it
comes to the mental, and the appearance that things are otherwise and
that they call for explanation and theory is due to a misapprehension of
the constitutive aspects of our psychological concepts. As Wittgenstein
famously remarked, “a cloud of metaphysics”—and, we may add, “of

she may have meant “+” thus all along. For criticism of Wright’s so-called “Intention view” of rule-
following, see Boghossian 2012, Peacocke 2012 and Bilgrami 2012. Wright has recently aban-
doned this position. See Wright 2015.
18
Wright 2001d, p. 312.
19
The set of rules and criteria that constitute our concepts and determine the meaning of the words
used to express them.
20
See Wright 2001f, p. 372.
178 A. Coliva

epistemology”, when it comes to first-personal self-knowledge—“is con-

densed in a drop of grammar” (PI II, p. 222e). Thus, the only sensible
philosophical project to be pursued in this connection is one of proper
description of the salient features of the relevant concepts. This project
leads to an important finding—namely, that first-personal intentional
avowals are authoritative unless there are positive reasons to doubt them.
According to Wright, this entails that attitudinal avowals are extension-
determining. They do not track, in an epistemologically mysterious way,
a predetermined mental fact, but they determine or constitute the kind
of mental state at stake. In terms of the Euthyphro contrast,21 it is not
because S has the intention I that if S believes/asserts she has I, she is
right. Rather, it is because S believes/asserts that she has the intention I
that she does have it. This, however, holds only when certain conditions
C obtain—that is, S is sincere, cognitively lucid and alert and has the
relevant conceptual repertoire and there is no reason to think she might
be self-deceived. Hence, a subject’s “best”22 judgement about her inten-
tion determines the fact that she does have that intention.
It is important to stress that, as far as I can see, this is not to say that,
according to Wright, the self-ascription of the intention brings about that
very intention. If that were the case, the relevant self-ascriptions would
be self-verifying and this is something Wright does not seem keen to
endorse. He writes:

When possession of a certain intention is an aspect of a self-conception

that coheres well enough both internally and with the subject’s behaviour,
there is nothing else that makes it true that the intention is indeed
possessed.23

I take this passage to suggest that just as an attitudinal avowal does not
track a pre-existing mental state, it does not give rise to it either. Elsewhere,

21
Wright has exploited this idea in numerous writings and in different areas of discourse such as
taste, morality (although he is sceptical of the prospects of this strategy in that realm) and second-
ary qualities (colour in particular). See Wright 1992. Concerning propositional attitudes, the most
thorough discussion is in Wright 2001b.
22
Wright 2001b, p. 204 et infra.
23
Wright 2001b, p. 204.
 7 Constitutive Theories 179

Wright insists that the self-ascription suffices for having the intention, but
it is not necessary for it.24 Moreover, Wright (2001f, pp. 351–352), in the
course of discussing Burge’s account of self-knowledge (cf. Chap. 5, §2),
criticises the self-verifying thesis regarding attitudinal avowals for explain-
ing too much—that is, infallibility—when the relevant self-ascriptions
may be open to error, essentially due to self-deception. But he also criti-
cises the self-verifying view because, in his opinion, it is guilty of explain-
ing too little. For it would hold only for episodic thoughts, originating
in the relevant self-ascriptions. In contrast, Wright thinks that authority
extends to all kinds of attitudinal avowals, also to those which do not
exploit the characteristic feature of some episodic thoughts and indeed
also to avowals of past intentional states (cf. Wright 2001b, p. 206).
This, however, confers a strange metaphysical status to intentions (and
to other propositional attitudes as well). For intentions are not brought
about by the self-ascription itself. But they do not seem to exist already, at
least as such, either. For, otherwise, it would be difficult to maintain that
our “best” judgement about them—that is, a judgement made sincerely
when there is no reason to think a subject might be self-deceived—deter-
mines their extension. Wright’s position then seems to lean towards a
form of irrealism, of an instrumentalist fashion, which is usually associ-
ated with Dan Dennett’s (1987) idea of an intentional stance and with
Donald Davidson’s (1970) notion of anomalous monism. Intentions
(and other attitudinal mental states) are not metaphysically robust enti-
ties. They are just useful devices—possibly just concepts—employed to
interpret and make better sense of subjects’ behaviour. Indeed, accord-
ing to Wright’s view, the authority we grant to subjects’ avowals of their
intentions is a function of the fact that doing so proves useful in the
application of a mutual interpretational scheme. As he writes:

It is part of regarding human beings as persons, rational reflective agents,

that we are prepared to ascribe intentional states to them, to try to explain
and anticipate their behaviour in terms of the concepts of desire, belief,
decision and intention. And it is a fundamental anthropological fact about

24
Wright 2001d, pp. 313–314. This way, he fends off the objection that one would be caught in a
regress of beliefs whenever there is anything believed, and that un-self-conscious creatures would
not be able to believe anything.
180 A. Coliva

us that our initiation into the language in which these concepts feature
results in the capacity to be moved, who knows exactly how, to self-ascribe
states of the relevant sorts—and to do so in ways which not merely tend to
accord with the appraisals which others, similarly trained, can make of
what we do but which provide in general a far richer and more satisfying
framework for the interpretation and anticipation of our behaviour than
any at which they could arrive if all such self-ascriptions were discounted.
The roots of first-personal authority for the self-ascription of these states
reside not in cognitive achievement, based on cognitive privilege, but in
the success of the practices informed by this cooperative interpretational
scheme.25

This view is problematical for a number of reasons. For instance, we seem

to act on the basis of intentions, but, if instrumentalism is right, we are
caused to act a certain way by what goes on in us at the subpersonal,
physical level. Our impression of freely forming intentions, on the basis
of which we go on to act in certain ways, would be merely epiphenom-
enal. True, the intentional stance may not be reducible to physical laws
and there may be an unbridgeable gulf between the intentional vocabu-
lary we use to describe and make sense of ourselves (and others) and
the vocabulary of neuroscience. All the same, at bottom, there would be
no intentions and other propositional attitudes, just physical states and
causal laws. So the impression we have of forming intentions and that
doing so is up to us is just a well-crafted, perhaps indispensable—that is,
irrevocable—illusion.
If that is right, Wright’s proposal seems to be an over-reaction: not
only does it do without a substantial epistemology of the mind, it does
so ultimately because the very existence of the relevant mental states is
no longer a datum of the problem of explaining how such mental states
may be self-known.
Moreover, it is in tension with Wittgenstein’s repeated pronouncements
against behaviourism. As we saw in the last chapter (§1), he did not want
to be considered a behaviourist and he affirmed the existence of mental
states. Still, he thought it was wrong to think of them, in a Cartesian
fashion, as objects or entities, luminously presented to consciousness.
25
Wright 2001a, pp. 140–141.
 7 Constitutive Theories 181

Wittgenstein, therefore, was against a certain conception of the mind,

not against the mind—that is, the existence of mental states—altogether.
In later writings, not concerned with self-knowledge, Wright him-
self has put forward several criticisms of this form of irrealism26 and has
explored the possible consequences of another anti-realist view about
mental states—namely, a “minimalist” position, according to which
psychological self-ascriptions are truth-apt, as they can be meaningfully
negated, inserted in conditionals and so forth but do not play a robustly
representational role; there is nothing “out there” that they represent and
that makes them true. However, according to Wright, even minimalism is
wanting for reasons which should not deter us here. Still, in his opinion,
even if no extant anti-realist view about psychological discourse stands
up to careful scrutiny, this does not mean that realism about ordinary
psychology is thereby vindicated. For one thing, there may be alterna-
tive forms of anti-realism yet to be examined; for another, all forms of
psychological anti-realism examined so far seem to be willy-nilly commit-
ted to the robustness of psychological discourse. But that, according to
Wright, does not show that their commitment is sound.
This, however, leaves the status of mental states totally in the dark: we
seem committed to regarding mental states as real, but that very com-
mitment may turn out to be false. Once caught in this impasse, I think
it is wise to reconsider the options taken at each fork, which led to such
an unappealing result. Indeed, as we shall see in §4, some traditional
anti-realist elements, such as the expressive function of at least some
psychological self-ascriptions, the absence of a truth-tracking epistemol-
ogy regarding our own psychological self-ascriptions and the Euthyphro
side of the test for determining the truth of our psychological self-ascrip-
tions, can be reconciled with a realist view about mental states.
Be that as it may, Wright’s view about self-knowledge seems wrong on
matters of detail. For it is not clear that our self-ascriptions of past inten-
tions are on a par with avowals of ongoing intentions. For one thing,
they may be based on self-interpretation: you may figure out, by reflect-
ing on it, that your past behaviour was led by the overarching intention
of becoming a mother, for instance. For another, they may be based on

26
Wright 2002.
182 A. Coliva

remembering what one intended to do at some earlier time: you form the
intention of going to the kitchen to start preparing a meal. The phone
rings. Once you hang up, you ask yourself “What was I about to do?”
or “What did I want to do?” and then you suddenly remember that you
wanted to go to the kitchen. Hence, it seems that in either case the rele-
vant self-ascriptions will not be groundless. Nor are they transparent: just
by having had them at some earlier time, a subject—even when equipped
with the relevant concepts and being attentive and alert—is not thereby
able immediately to self-ascribe them at a later time. For one can have
difficulties in figuring out what led one to act a certain way, or one may
forget what one intended to do before receiving a phone call. Authority
is not guaranteed either. For, clearly, one can make mistakes in interpret-
ing oneself, and one can misremember what one was about to do, before
receiving the phone call. These occasional failures, however, contrary
to what Burge maintains (see Chap. 5, §2), do not impair our overall
cognitive well-functioning. Surely, the latter situation is often unverifi-
able, since we do not often speak out loud our ongoing intentions. But
suppose we were so trained that any time we formed an intention, we
immediately gave voice to it. Suppose, furthermore, that we lived in a
community where subjects were always heard by some other member of
the community and where their utterances were always recorded. If that
were so, the idea that we are authoritative with respect to our avowals of
past intentions would evaporate. For others would be just as well placed
to contradict us, based on their recollection (or on records stored some-
where) of what we had said.27
Moreover, as we shall see in the following section, there is an open
issue about the correct account of self-deception, such that it may actu-
ally turn out that, contrary to appearances, self-deception is no bar to
the possibility that avowals of intentions and other propositional mental
states be always authoritative (save for cases of insincerity and of concep-
tual misunderstanding).

27
Consider, moreover, the possible tension between an avowal of an ongoing mental state at t1 and
a different past one at t2. It is not clear, in Wright’s framework, to which one we should accord
priority and why.
 7 Constitutive Theories 183

Thus, to sum up, Wright’s proposal is wanting because it inclines towards

a problematic form of instrumentalism about intentions and other propo-
sitional attitudes; it is clearly wrong with respect to avowals about past
intentions (and other intentional mental states) and is hostage to a by-no-
means correct view of self-deception. Of course, the last challenge is condi-
tional upon proving Wright’s conception of self-deception wrong, and one
may try to rescue the proposal that at least avowals of present intentions
are authoritative and do so, without embracing the self-verifying view, by
saying that a subject’s pronouncement at least identifies the kind of mental
state she has by determining its nature—its being an intention rather than
a hope or a desire—and its content—P rather than Q. But, as interesting
as this refurbished proposal might be, it is not clearly Wright’s.

3 The Two Sides of the Constitutive Thesis:

Bilgrami
In his Self-Knowledge and Resentment (2006), Bilgrami has proposed a
new kind of constitutive position. Its novelty resides in, on the one hand,
assigning no priority to either side of the constitutive thesis (both need
explaining) and, on the other, in  explicating the C-conditions under
which the constitutive thesis is supposed to hold by reference to the idea
of “the irreducibly normative nature both of human agency and of the
intentional states of human agents” (Bilgrami 2012, p. 263).
According to Bilgrami, in order to redeem the left-to-right side of the
biconditional—that is, the one which accounts for transparency—we
need to take into account an important distinction between propositional
attitudes as dispositions and as commitments (cf. Chap. 2, §2). Bilgrami’s
characterisation of the latter is quite weak. He simply claims that beliefs
and desires as commitments are such that if one has them and one does
not live up to them, then one ought to try to do better to live up to them
(see Bilgrami 2012, pp. 265–266).28 Now, if we are considering beliefs

28
He also holds that one incurs various commitments related to the inferential role of the concepts
that specify the contents of one’s commitments. For instance, if I believe as a commitment that
there is a table in front of me, I also ought to believe that if I run hard into it, I will be injured.
184 A. Coliva

as commitments, they should be transparent by their very nature. For in

order to have them, one ought to know them; otherwise, one could not
try to do better in order to live up to them, when one fails to do so.
However, Bilgrami does not confine transparency to propositional atti-
tudes as commitments. For he thinks that also beliefs and desires as dispo-
sitions are transparent to their subjects when they “potentially go into the
production of (…) an action”, which is “free and accountable” (Bilgrami
2012, p. 267)—that is to say, an action with respect to which one can
have “justifiable reactive attitudes” (Bilgrami 2012, pp. 267–268). The
background against which Bilgrami’s claim becomes intelligible is Peter
Strawson’s account of the compatibility between determinism and free
will.29 According to that account, reactive attitudes—such as blame and
resentment—need to be appealed to in order to identify those actions
we regard as free (or non-coercive).30 According to Bilgrami, in order for
those reactive attitudes to be justifiable, one needs to presuppose that
the mental states that caused the relevant actions are self-known. Thus,
under such a proviso, transparency holds both for commitments and for
dispositions.
In contrast, regarding the right-to-left side of the constitutive thesis
(that is, the one of authority), Bilgrami claims that it holds for proposi-
tional attitudes as commitments only. These are, in his view, intrinsically
normative mental states. By that he means that norms of rationality are
not brought to bear on them just from the outside, as it were. Rather,
although Bilgrami does not go into details here, having a certain belief as
a commitment, say, requires withdrawing one’s assent to anything incom-
patible with it (cf. Bilgrami 2012, pp. 273–274). I take this view to be
entirely compatible with the one which we presented in Chap. 2 (§2) and
which we will avail ourselves to in the following to provide an account of
Moore’s paradox (cf. Appendix).

29
See Strawson 1962.
30
Free actions are those brought about by non-coercive causes (contra those which are brought
about by coercive ones). Strawson’s idea is that non-coercive actions can be individuated only by
bringing in normative considerations and, in particular, our reactive attitudes and the practice of
blaming and resenting they give rise to. To that picture, Bilgrami adds that our reactive attitudes
should be justifiable ones and, being so, they presuppose that the mental states in which the
resented and blamed actions originated be self-known. If they were not self-known, our reactions
would not be justified and the actions we unjustifiably react to should be considered as non-free.
 7 Constitutive Theories 185

Further to support the idea that propositional attitudes as commit-

ments are intrinsically normative mental states, Bilgrami proposes a
“pincer argument”. Its first arm is as follows. There is an open-question
argument, indebted to G. E. Moore’s one with respect to “good”, which
can be taken to bear on any attempt to give a definition of propositional
attitudes as commitments in terms of first-order dispositions. Suppose
you have such a definiens, which appeals to a given first-order disposition.
In that case, it would always be open to one to ask “I have such-and-
such a first-order disposition, but ought I to φ?”, just like, in the case of
any proposed definition of good in terms of some other non-normative
notion, it would always be open to one to ask “This action maximises
shared utility (say), but is it good?”. Hence, any attempt to reduce nor-
mative concepts to non-normative ones would face the open-question
challenge.
If a dispositionalist then claimed not to provide a definition of the
concept but just to identify a posteriori yet necessary identities, like water
= H2O or Hesperus = Phosphorus, then one could rationally deny that
beliefs (or desires) are identical to the first-order dispositions proposed,
just like one can rationally deny that water is H2O or that Hesperus is
identical to Phosphorus. According to Frege, as is well known, this is
because the senses associated with the terms figuring on opposite sides of
the identity are different, although those terms pick out the same refer-
ent. This, according to Bilgrami, entails an undesirable consequence—
namely, “if it is the sense (or meaning or definition) that is given in terms
of the naturalistic property, then it is precisely what the Moorean open
question consideration is once again effective against” (Bilgrami 2012,
p. 273). Hence, “a Moorean argument, supplemented by a Fregean argu-
ment, together construct a pincer effect against the naturalistic equation
of intentional states with dispositions” (ibid.).
The second half of this argument seems particularly weak, for although
one may rationally deny those identities, this would merely show that
the concepts involved are somewhat “incommensurable” because one of
them would be intrinsically normative whereas the other would not. Still,
if the identities are true, it would remain that the referents of those con-
cepts would be identical. Hence, propositional attitudes as commitments
would have been reduced to first-order dispositions.
186 A. Coliva

Personally, I think the half-baked thought according to which com-

mitments are intrinsically normative because, by their very nature, they
exclude assent to contents inconsistent or, more generally, incompatible
with them is more promising. However, I have to beg the reader to wait
to hear more about that until the Appendix.
Finally, once the existence of propositional attitudes as commitments
is established, Bilgrami considers the objection to authority based on
self-deception. Namely, how can we be authoritative with respect to our
mental states (as commitments) if, at least from time to time, we can be
wrong about them because of self-deception? His reply consists in put-
ting forward a specific account of self-deception, which actually explains
the objection away. For, in his view, self-deception is not a phenomenon
whereby one mistakenly attributes to oneself a given belief or desire.
Rather, when one is self-deceived, one has the relevant belief or desire as
a commitment, which would lead one to avow “P” or “P would be good
to have”, for instance. Yet one would have also a countervailing belief or
desire as a disposition, which is normally unconscious, that would lead
one to behave in ways that run contrary to one’s beliefs and desires as
commitments. Thus, in his view, authority holds in its strong version
when self-ascriptions of commitments are at stake.
Groundlessness, then, is part and parcel of the form of constitutivism
advocated by Bilgrami. In his view, that is, there is no genuinely epistemo-
logical problem about first-personal self-knowledge. As he writes: “if one
thought instead that self-knowledge, being knowledge after all, was just
another narrow epistemological theme, I don’t think we could account for
our intuitions about privileged access” (Bilgrami 2012, p. 277). Rather,
what goes by the name of “first-personal self-knowledge”, captured by
the two sides of the biconditional figuring in the constitutive thesis, is a
conceptual, necessary and a priori truth redeemed through the kind of
considerations appealed to thus far—namely, considerations that stress
the conceptual connections between freedom, and therefore responsible
agency, which in turn requires knowledge of our own mental states as well
as the ability to have intrinsically normative mental states. Says Bilgrami:

There is something honest, then, about those who refuse to grant anything
special to self-knowledge and view it as getting a causal account based on a
 7 Constitutive Theories 187

measurably more than usual reliable mechanism that will account for our
intuitions misleadingly expressed as ‘privileged access’. They see it as a nar-
row question in epistemology, they find the exceptions to be ubiquitous,
and they draw their conclusion that there is nothing radically set apart
about self-knowledge. Their conclusion is honestly drawn from their
framework. It is their framework that is wrong. Self-knowledge is unique
only if it is embedded in a much wider framework integrating very large
themes in philosophy such as the normative nature of intentionality and
agency that I have been expounding. (Bilgrami 2012, p. 277).

Now, as is evident, Bilgrami embraces the anti-epistemological strand

of constitutivism. However, he does not hold the metaphysical the-
sis that self-ascriptions of propositional attitudes actually constitute or
bring about the relevant first-order mental states. In this sense, he is not a
strong constitutivist. This, by itself, is no objection; it is only a remark to
better situate Bilgrami’s position within the constitutivists’ camp.
However, there is something problematic about his proposal—namely,
that it accords transparency also to dispositional propositional attitudes,
as long as they are the object of justifiable reactive attitudes. This claim,
however, sounds oxymoronic to some extent. For if a propositional atti-
tude is something that happens to one or that one finds oneself saddled
with, like a sudden urge to smoke a cigarette, in what sense can a reactive
attitude against it be legitimate? One cannot be blamed for that urge,
but, at most, for realising it and yet indulging in it. What is genuinely
blameworthy, that is, would be to behave in accord with that disposi-
tion, while having knowledge of one’s urge, and a further commitment
to preserve one’s health, say. If, in contrast, one thought that the very
urge could be legitimately blameworthy, then the distinction between
propositional attitudes as dispositions and commitments, based just on
an appeal to justifiable reactive attitudes, would founder. For they would
just be treated on a par vis-à-vis the justifiable reactive attitudes they
would elicit.
Consider, moreover, that according to Bilgrami, also Freudian mental
states would be dispositional ones. Notice, furthermore, that in his view
they are known in a third-personal way (cf. Bilgrami 2006, Appendix
1). If they could be the object of justifiable reactive attitudes, that would
188 A. Coliva

entail that they are transparently known to their subjects. But this is
clearly absurd. What, however, does preclude the possibility of having
justifiable reactive attitudes towards them if it is allowed that also dispo-
sitional mental states can be the object of such attitudes, as long as those
mental states enter rationalising explanations of subjects' behaviour?
As interesting as Bilgrami’s constitutive proposal appears to be, we
have seen that it is also problematic when it comes to the details of the
vindication of the left-to-right side of the constitutive thesis, based on the
“pincer argument”, and of the idea that propositional attitudes as com-
mitments are intrinsically normative mental states.

4 A Metaphysically Robust Kind

of Constitutivism: Coliva
In previous work,31 I have maintained a metaphysically robust kind of
constitutivism. Central to the proposal is the defence of both sides of the
constitutive thesis, which proceeds in two different steps. Let us consider
them in turn.

4.1 The First Half of the Constitutive Thesis:

Transparency

In order to redeem the left-to-right side, the strategy consists in pointing

out that our notion of an intentional mental state is not univocal. On
the one hand, as we saw in Chap. 2 (§2), there are intentional mental
states as dispositions: states that are attributed to a subject to make sense
of her observable behaviour, which she may not be aware of and which,
even if she were, would not be within her direct control. Yet, manifestly,
adult human beings have also a further kind of mental states—namely,
mental states that are within their control and for which they are ratio-
nally responsible. In the terminology introduced in Chap. 2 (§2), they
have intentional mental states as commitments. To remind the reader of

31
Coliva 2009, 2012d. The account presented here partially differs from the one presented in those
papers.
 7 Constitutive Theories 189

what that means, recall that, in my view, what is essential to propositional

attitudes as commitments is the following:

(i) they are the result of an action—the mental action of judging that
something is the case—on the subject’s part;
(ii) they are normatively constrained—that is, they must respond to the
principles governing theoretical and practical reasoning;
(iii) they are so constrained (also) from the subject’s point of view;
(iv) they are mental states for which the subject is held responsible.

A belief or a desire as a commitment cannot be formed unless (i) the

subject herself endorses their content upon considering evidence for or
against them and, therefore, unless she deliberates what to do with respect
to “P” (or “P would be good to have”)—that is, assenting to it or refusing
to assent to it or assenting to its negation. (ii) differentiates commitments
from mere drives and brute dispositions: if I believe that p as a commit-
ment, then I ought to do so on the basis of sufficient evidence and ought to
withhold from it in case sufficient contrary evidence came up. Obviously,
no such “oughts” hold for drives, like the urge to smoke a cigarette after
dinner; or for brute dispositions, such as the disposition to form a certain
thought upon hearing a given word, tune, and so on. (iii) points out that
these “oughts” must be appreciated by the subject herself and are not
merely added from a third-personal point of view. Finally, it is precisely
because beliefs and desires as commitments would meet conditions (i)-
(iii) that one would be held responsible for them. Thus, should one not
withdraw from one’s belief that P, or desire that Q, in the face of counter-
evidence for “P” or “Q would be good to have”, then one would incur
criticism for not doing so and, moreover, should accept to incur it since,
as a matter of fact, one would not have lived up to one’s commitments.
Hence, the C-conditions under which the biconditional holds should
be specified by reference to propositional attitudes as commitments, not
as dispositions, besides containing the usual reference to sincerity cum
conceptual and cognitive aptness.32

32
Hence my proposal sharply differs, on this score, from Bilgrami’s, who, as we saw in §3, main-
tains that transparency holds also for propositional attitudes as dispositions, provided they enter
the explanation of actions that are the object of justifiable reactive attitudes.
190 A. Coliva

If this is right, then we can easily defuse the usual—yet apparent—

counterexamples to the transparency side of the constitutive thesis. First
of all, it seems safe to hold that animals do not have mental states as a
result of judgement and of actively bringing evidence and practical con-
siderations to bear on what they think. Nor are they capable of being self-
critical or of accepting criticism if they did not live up to them. Hence,
whatever kind of mental states they can actually enjoy, they cannot have
mental states as commitments. Similarly, unconscious mental states are
not brought about by judgement. They are produced by experiences we
have had but are not formed by consciously assenting to certain contents.
Furthermore, one can be self-critical or accept criticism for them but that
requires knowledge of them, which can be achieved only through third-
personal means. Moreover, changing them in light of countervailing con-
siderations may well not be immediate or stable, may require effort and
is open to (partial) failure. Hence, they are not commitments. Thus, the
cases of animals’ and of unconscious mental states will not be counterex-
amples to the first half of the constitutive thesis, once it is appropriately
qualified—that is to say, if it is taken to hold only for mental states as com-
mitments. Yet we must still explain why having beliefs and desires as com-
mitments should entail that they are known to the subject who has them.
We have seen that commitments are such that they are brought about
by judgement based on evidence and that one should accept criticism
(or be self-critical) for not living up to them. The latter feature obviously
requires knowledge of them qua the mental states they are. For if one did
not know that one has a given belief as a commitment, say, one could not
be self-critical or accept criticism for not having lived up to it. However,
to say so means to establish merely a necessary condition for having com-
mitments—that is to say, it means simply to show that self-knowledge
is necessary for having propositional attitudes as commitments. As we
saw, self-knowledge is necessary for any (self-)critical assessment, even
the one concerning one’s unconscious mental states. Yet, in that case,
third-personal self-knowledge is needed. Thus, simply to stress that self-
knowledge is a necessary condition for this kind of self-critical appraisals
says nothing about its nature and provenance. Unless one embraces the
Wittgenstein anti-explanatory mantra, one cannot leave things at that.
What more is needed, though?
 7 Constitutive Theories 191

The missing ingredient is an account of conceptual mastery with

respect to the relevant psychological concepts. If Shoemaker’s claim is
to be vindicated—for commitments only, of course—that in order to
know one’s mental states one needs just normal rationality, intelligence
and conceptual mastery, the latter has to be accounted for. The allegedly
epistemological problem of self-knowledge thus becomes the problem of
explaining how we learn and canonically apply the relevant psychological
concepts. Yet we cannot say that we conceptualise our mental states as
commitments either by having these first-order mental states in view, as
it were, and labelling them as the mental states they are or by (always)33
self-consciously applying the rule (in the case of belief ) that if I judge
that P is the case on the basis of evidence, then I believe that P. For, in
the former case, we would be back to the Cartesian conception of men-
tal states and to the observational model of self-knowledge. In the latter
case, in contrast, we would presuppose the possession of other intentional
psychological concepts, such as the concept of judgement, which, argu-
ably, will have to be explained along the same lines as the possession of
the concept of belief (intention, desire, etc.). Moreover, the self-conscious
application of the introduction rule for the concept of belief (in this case)
would presuppose the possession of the latter concept too. Hence, the
explanation would be hopelessly circular. And, as we saw in the course of
our discussion of Peacocke’s and Shoemaker’s proposals, reverting to the
idea of a tacit form of self-knowledge would only raise the same kinds of
problem at one remove.34 Hence, it is crucial to come up with a differ-
ent account of what mastery of the concept of belief (in the first-person
present) consists in.
Here is a proposal. Take a subject who is able to judge that P, give
evidence in favour of it and withdraw from it if required and, therefore,
has the first-order belief that P based on judgement. Suppose you ask her
“Do you believe that P?” and she is unable to answer. You conclude that
she does not have the concept of belief. In that case, you would simply
train her to the use of that verb by drilling her into using the expression

33
This can sometimes (and perhaps often) be the case, but it cannot always be the case, because it
would presuppose self-knowledge.
34
See Chap. 5, §1 and this chapter, §1.
192 A. Coliva

“I believe that P”. You teach her to substitute one form of behaviour—one
kind of expression of her mind, that is, the outright assertion of “P”
accompanied by the ability to give reasons for it, which manifests her
first-order belief (based on judgement)—with another, that is, the asser-
tion of “I believe that P”. Similarly, take a subject who says “Peace in the
Middle East would be good to have” and is disposed to offer consider-
ations in its favour and withdraw from it if these considerations did seem
no longer compelling, but if asked “Do you hope/desire that peace be
reached in the Middle East?” did not know how to answer. Then, again
you drill her to use “I hope/desire that P” as an alternative expression of
her mind—that is, of her asserting “Peace in the Middle East would be
good to have” for this and that reason. Let me stress that it is absolutely
essential in order for the proposal to steer away from any observational
model that one be adamant that “I believe that P” or “I desire that Q”
is taught neither on the basis of evidence nor on the basis of the rule “If
you are disposed to judge thus-and-so, then you believe/desire that P/Q”.
Rather, “I believe that P” and “I desire that Q” are taught blindly, as alter-
native expressions of one’s mind: they are ingrained as alternative ways of
expressing one’s first-order beliefs and desires (based on judgement) other
than asserting that P or that Q would be good to have.35 Hence, in this
account, there is no inner epistemology, just a substitution of one form
of behaviour with another. But—and this is crucial—the kind of behav-
iour which would get replaced would already be quite rich. For, in order
to have beliefs (and other propositional attitudes) based on judgement,
a subject will already have to have the ability to differentiate between,
for instance, believing P and P’s being the case, by being sensitive to the
fact that her point of view may be challenged—thus responding with
reasons in favour of it—or indeed proved wrong—thus abandoning it. It
is only on the background of this already complex pattern of behaviour,
which, however, does not seem to require the concept of belief, as we saw
35
Indeed, this seems to me to be the right development of Wittgenstein’s idea that avowals substi-
tute behaviour. It is just that when we move from avowals of sensations to avowals of propositional
attitudes the behaviour we must take into account is not merely physical but also linguistic. As the
reader will recall, Dorit Bar-On makes a similar point, even though in the course of developing a
neo-expressivist account of self-knowledge (cf. Chap. 6, §2). Robert Gordon too seems to endorse
it, even though he does not think it suffices for possessing the relevant psychological concepts. We
criticized Gordon’s view in Chap. 4 (§3).
 7 Constitutive Theories 193

in the course of our discussion of Shoemaker (§1), that I think we can

maintain that “I believe that P” may be taught blindly. “I believe that P”
would then be taught as an alternative way of making the commitment
to P other than judging that P. But what “I believe that P” would make
explicit—to the subject herself and others—is the fact, which remains
only implicit in judging P, that that is just her own point of view among
other possible ones, which need not be correct. This would happen by
telling the subject, for instance, “See, you have said that P, but it is not
the case that P. You merely believe it”.
A consequence worth mentioning is that if one’s psychological self-
ascriptions are canonically issued as alternative expressions of one’s
first-order propositional attitudes as commitments and are not brought
about by observational or inferential cognitive procedures, then it
becomes quite obvious why, as Shoemaker has it, they are not separate
existences with respect to the relevant first-order mental states. Just like
an infant’s cry when she is injured does not have separate existence from
her pain, but is its individuating behavioural manifestation, similarly
avowing “I believe/desire/intend that P/to φ” becomes the articulate
verbal manifestation of one’s ongoing judgement-dependent proposi-
tional attitudes.
An important feature of the present account is that it tights first-person
and third-person uses of “to believe” together from the start. For it is only
by being taught by someone else to replace the direct expression of one’s
mind by means of asserting “P”, while being disposed already to retract
it, if shown wrong, with appropriate psychological self-ascriptions, that
one acquires the concept of belief. Once endowed with the capacity of
making explicit her belief that P as a commitment, the subject then can
articulate the conceptual role which individuates the concept of belief:
for she now can express the difference between believing that P and its
obtaining, both in her own case and in the third-personal one.
If this is right, whenever one is in a position to judge that P is the
case or that Q would be good to have, one will also and immediately be
in a position to avow (or to judge) that one believes/desires it. So, men-
tal states as commitments and conceptual mastery—acquired in the way
proposed—suffice, together, to give one knowledge of one’s own mental
states for free—that is, without any cognitive endeavour.
194 A. Coliva

4.2 Objections from Empirical Psychology

Coming from the psychology camp, various objections may be raised

against this proposal about the acquisition of psychological concepts and
the ability to apply them in one’s own case. To look at them will also
allow us to go back to the debate between theory-theorists and simula-
tion theorists, reviewed in Chap. 4 (§3), this time interpreted as a debate
about what is required in order to master the relevant concepts, in par-
ticular the concept of belief.36
According to theory-theorists, empirical evidence shows that children
take time to acquire the concept of belief and that that goes hand in hand
with the development of a theory of their own as well as of other minds.
This evidence would not sit well with the present proposal and would
rather favour an account of concepts’ possession, according to which to
possess a concept—and, in particular, the concept of belief—consists in
knowing its conceptual role.
In response it may be said that, quite apart from the conceptual prob-
lems that would pose, such as presupposing self-knowledge and the pos-
session of a lot of intentional concepts, what has been presented here is
not a psychological theory of concepts’ possession. After all, what I have
suggested is simply how someone who is already able to have first-order
judgement-dependent beliefs may come to acquire such a concept. It
may be that young infants simply do not qualify—they may not yet be
able to take rational considerations to bear onto them. However, I think
it is an entirely empirical issue whether the psychological data currently
at our disposal, like the age at which children pass the false-belief test in
their own case as well as in the case of other subjects, should be taken to
show that children take time to learn how to use “I believe” (and acquire
the corresponding concept) or should be taken to show that it takes time
for them to become capable of judgement-dependent beliefs, apt to result
in commitments once one has the relevant psychological concepts and
thus can reflect on them. For, as I understand it, children at that age
come to pass the false-belief test in their own case when they actually

36
This is indeed something which goes to the heart of the debate anyway, as Davies and Stone
(1995a) remark in their Introduction.
 7 Constitutive Theories 195

understand that their own point of view about the world (or that of other
subjects) may be wrong. This, I take it, is at least a necessary condition
for having beliefs as commitments. Furthermore, the ability to pass the
false-belief test in the case of others may be explained differently than by
appeal to the fact that children would possess a theory of other minds.
For it would be enough to explain their correct answers to suppose that
they issue them as if they themselves were in the other person’s shoes. So,
the ability to pass the false-belief test need not show that children pos-
sess a veritable theory of their own minds as well as of others’. In fact,
it may actually be taken to prove just that they are capable of first-order
beliefs based on judgement and to project themselves onto others and
therefore issue the correct answer to the false-belief test, without thereby
having any explicit knowledge of their own and other minds, which, in
my view, crucially depends on the possession of the relevant psychologi-
cal concepts.
The reason why the data at our disposal do not seem to tell us clearly
what is the case is that—quite understandably—the experiments have not
been designed to test the possibility I am advocating here. For, usually,
children will be exposed to talk in terms of belief when they are actually
in the process of acquiring the ability to have beliefs based on judgement.
This may well have confused the issues: we may have mistaken the fact
that it takes time for children to learn to have these first-order mental
states as a sign of the fact that it takes them time to acquire the concept
of belief. Furthermore, we may have imputed that difficulty to the fact
that mastery of that concept would depend on the acquisition of a theory
of one’s own as well as of others’ minds. A more telling test, then, would
be to look at children who have not been exposed to psychological talk
up to the age of 3 or 4; see whether, around that age, they pass the false-
belief test in their own case as well as in others’ (where, crucially, the test
should not be phrased in terms of beliefs); and then introduce them to
talk in terms of belief (and other propositional attitudes).37 If, at that

37
Should this test prove impossible, one might see whether there are languages which do not have
talk in terms of belief and other propositional attitudes. If speakers of those languages were in fact
capable of beliefs as commitments, as I think there is no reason to be sceptical of, we then could
test how long it would take them to acquire the ability to express their minds by self-ascribing
beliefs and other propositional attitudes in a different language which contained these devices.
196 A. Coliva

stage, it actually takes them a short amount of time to learn how to use
“I believe”, then we would have shown that the account of concepts’ pos-
session I have been proposing would, in fact, be compatible with human
psychological development.38
Furthermore, it has to be noted that the conceptual role of the concept
of belief is what theorists of concepts offer as an abstract individuation
of that concept that supervenes on a practice of its use which, however,
may come about in different ways. In particular, since according to the
present proposal the commitments undertaken by asserting “P”, as an
expression of one’s belief based on judgement, and by asserting “I believe
that P” would actually be the same (save for the fact that the latter would
make explicit what the former leaves implicit—namely, that the assertion
of “P” expresses one’s own point of view that need not be correct), it may
well be that the conceptual role of the concept of belief in the first-person
present specifies the rules for the use of that concept which, in practice,
may have been acquired by becoming able to have first-order beliefs based
on judgement first and then by being blindly drilled to express them by
prefacing one’s assertion of “P” with “I believe that”.
Assuming that what I have been proposing is along the right lines, it is
perhaps worth noticing that it is a consequence of the suggested account of
conceptual mastery that while it may be an open issue whether and to what
an extent language is necessary in order to have first-order propositional
attitudes based on judgement, on my view language is indeed necessary for

38
The apparently difficult case, for my own proposal, would in fact be constituted by autistic
patients affected by Asperger syndrome. Although they do not fail the false-belief test, they seem
not to have a theory of their own minds and to lack a theory of other people’s minds (Frith and
Happé 1999). But several things must be noted: 1) when we look at their reports, what they show
is that subjects affected by this syndrome have different kinds of experiences, particularly of speech,
and different sensations, if not an altogether lack of painful sensations (pp. 15–18). None of this
would show anything relevant with respect to their propositional attitudes and their knowledge of
them. 2) The only report which has a bearing on this issue (from Donna Williams (1994)) in fact
seems to imply that she did not have, as an autistic child, desires as commitments (she writes
(p. 15): “Autism had been there before I’d ever known a want of my own, so that my first ‘wants’
were copies of those seen in others (a lot of which came from TV)”). In such a case, it would not
be surprising that they would have to gain knowledge of their own minds in a third-personal way
and that this would require some kind of theory of other minds. Finally, all the data are based on
personal reports and, obviously, this would not have any bearing on the possibility of having, and
of having knowledge of one’s occurrent commitments. For much of what they say could actually be
due to forms of self-interpretation.
 7 Constitutive Theories 197

one’s knowledge of them. So I am committed to the view that only linguis-

tic creatures can have self-knowledge—that is to say, the ability to make
psychological self-ascriptions which paradigmatically exhibit groundless-
ness, transparency and authority.
In this connection, it is worth stressing that the psychological liter-
ature on self-knowledge in non-linguistic creatures is both difficult to
interpret and actually potentially irrelevant. For what has been tested par-
ticularly in chimpanzees is merely the ability to know others’ perceptions,
such as seeing, and not their propositional attitudes—let alone the highly
specialised class of propositional attitudes as commitments I have been
trying to make plausible so far. Moreover, these studies also show crucial
discrepancies. For instance, research conducted by Povinelli and his asso-
ciates denies that chimpanzees have knowledge of other subjects’ percep-
tions, whereas those conducted by Tomasello and his lab support the
opposite interpretation.39 So it seems safe to conclude that, at the present
stage of the inquiry, the empirical data currently at our disposal have, in
fact, no bearing on the issue of whether only linguistic creatures can have
knowledge of their own judgement-dependent propositional attitudes.

4.3 The Second Half of the Constitutive Thesis:

Authority

So far, we have seen that, in order to account for transparency, the

C-conditions figuring in the constitutive thesis must include reference
to a lucid and sincere subject who is capable of having propositional atti-
tudes as commitments and who is endowed with the relevant psychologi-
cal concepts, acquired blindly. Our problem now is: how can we account
for free, as it were, for the claim that when a sincere and conceptually
competent subject avows her own mental states, she has them? And even
before engaging in this task, what grounds would there be to accept that
any sincere psychological self-ascription made by a conceptually endowed
subject is correct? Aren’t cases of self-deception, however rare they might
be, just a clear counterexample to that half of the constitutive thesis?
39
See Povinelli and Vonk (2004); followed by an Appendix with replies to objections coming from
the other camp, at pp. 24–28. Tomasello et al. (2003a, b).
198 A. Coliva

One might, with Wright, add to the C-conditions that the subject
should not be self-deceived (or anyway that it is reasonable to assume
that she is not). But quite apart from sounding an ad hoc move, it seems
that the very possibility of self-deception would show that constitu-
tive accounts do not have much of a point: after all, how could one be
mistaken about one’s own immediately available mental states if not by
somehow going wrong in identifying them? Wouldn’t such room for error
be compatible only with non-constitutive accounts of self-knowledge?40
Hence, it would be good news if we could account for self-deception dif-
ferently, thereby showing that its existence would not constitute a threat
to constitutive accounts.
Here is where I agree with Bilgrami: self-deception is a case where a
subject self-ascribes a mental state and has it as a commitment yet also
has another, opposite mental state as a disposition. The irrationality is
brought about by the clash between one’s commitments and one’s own
unconscious dispositions. Thus, for instance (the example is mine), take
a jealous wife who openly and sincerely asserts with her friends that she
believes that her husband is totally faithful to her—and has all the rea-
sons in the world to do so—but, then, once at home, is often inquisitive,
searches his belongings, and so on. According to Bilgrami,41 what we
should say is that she believes as a commitment that her husband is faith-
ful—after all, she is prepared to assert it with friends and has all the rea-
sons to think so. Yet she also has the unconscious belief, as a disposition,
that he is unfaithful to her, which is operative in her inquisitive behav-
iour. So, she is self-deceived, in the sense that she sincerely avows a belief
and behaves in ways that run contrary to it. Yet it is not the case that she

40
Wright (2001b, p. 324) seems to me to underestimate the implications of allowing for cases of
self-deception, and so I think does Heal (2002, p. 276).
41
See Bilgrami (2006, pp. 140–157; 278–280); cf. also Stoneham 1998. It must be stressed, in
order to avoid confusions, that I am endorsing Bilgrami’s account of self-deception with respect to
those mental states one would not attribute to oneself on the basis of inference and observation of
one’s own behaviour and further available mental states. I think in the latter cases one could make
genuine mistakes and self-attribute mental states one does not really have. If, then, one were to
restrict self-deception, properly so conceived, only to these cases, as Wright suggested to me in
conversation, then the fact that one might go astray in self-interpreting oneself would not represent
a counterexample to the view that non-observational or immediate self-ascriptions of propositional
attitudes are not open to failures of authority. The authority of immediate attitudinal avowals then
would remain unchallenged.
 7 Constitutive Theories 199

has a false belief about her own beliefs. Rather, she has two, different
beliefs, both in kind and in content, that give rise to her distinctively
irrational behaviour.42
Here are some considerations in favour of Bilgrami’s account of self-
deception. First, I agree with Bilgrami that his account does well at
explaining cases of motivated self-deception—that is, cases in which self-
deception is the outcome of a conflict in the subject between, say, believ-
ing that P and believing that not-P. In these cases, one of the two mental
states gets suppressed while the other one is endorsed. Yet the former can
remain operative in shaping (at least part of ) the subject’s behaviour and
lead her to various forms of inconsistency. Secondly, and more generally,
what makes us say that a subject is self-deceived is a conflict between her
psychological self-ascriptions and (some other aspect of ) her behaviour.
Now, conflict is usually brought about by the fact that there are two oppo-
site parties (or more) at fight, neither of which need be wrong but each of
which may simply be responding to different motivations and concerns.
In the case at hand, it makes sense to think that whereas one part of the
subject’s personality is entirely confident and mature, the other is full of
insecurities, which lead her to be suspicious of the behaviour of those
around her. Of course, there may be reasons for both attitudes: on the
one hand, the fully open and trustworthy behaviour of the husband and,
on the other, a perhaps (well-)motivated sense of insecurity about one’s
own power to attract a person and to involve him in a stable relation-
ship. Finally, suppose the subject realises, either through self-analysis or
through the aid of a therapist, that she has such an unconscious belief
about her husband’s infidelity. Now, if it were just a matter of realising

42
One may object that there are also cases of “negative” self-deception—cases, that is, in which one
says “I do not believe that P˝ yet behaves in ways that are explainable only by attributing to them
the belief that P. Stretching the example slightly, but just because that would help make the point
more vividly, think of Pascal who would say “I do not believe God exists (nor that he does not)”
and yet would behave—or, at any rate, recommend to behave—as an irreprehensible Christian. In
this case, one would not be self-ascribing any belief. Hence, the only option seems to say that one
falsely believes that one does not believe that God exists (nor that he does not). But we can recast
the example in such a way that it ceases to be a counterexample to authority. For we could say that
the avowal is still the expression of the subject’s mental state—namely, of her commitment to not
using “God exists” (nor its negation) as a premise of her practical and theoretical reasoning, which
runs against the disposition to behave as a kosher Christian and thus use that belief as a premise of
her practical reasoning.
200 A. Coliva

her own mistake in categorising a state of mind of hers, she should simply
correct her psychological self-ascriptions. After all, when I get to know
that the wall I am looking at is not red, but white and lit by a red light, I
would immediately correct my belief—that is, I would substitute it with
the new one. But clearly this is not what would happen in the case we are
considering. For the subject would (or, at any rate, should) try to realign
her behaviour, motivated by the belief as a disposition that her husband
is unfaithful to her, with her commitment to his fidelity. Obviously, this
can take a lot of time and personal effort and indeed may never fully
succeed. For all these reasons, it seems to me that Bilgrami’s account of
self-deception is by far preferable to the traditional explanation of this
phenomenon in terms of simply mistaken psychological self-ascriptions.
As a result, self-deception is entirely compatible with the fact that a sub-
ject is authoritative with respect to her own mental states as long as it is
clear that the mental states she is authoritative about are merely those as
commitments.
Having dispensed with the counterexample to authority—indeed,
with what is usually regarded as the only counterexample to it—let me
turn to the problem of explaining why it holds. Recall that we are look-
ing for an account of authority that does not make it the result of any
epistemic achievement. For any epistemic achievement may, in principle,
go wrong and, in that case, there could be counterexamples to authority.
But we have just seen that there aren’t any.43 Indeed, there cannot be
any if we want to be serious about the fact that the biconditional holds
as a matter of conceptual necessity. So we should not just be happy with
the result that, as a matter of contingency, self-deception is not really a
threat to authority. For one thing, one might dissent with our account
of that phenomenon and still hold that it raises a problem for authority.
For another, there may be other phenomena which might, at least con-
ceivably, impugn it. Hence, a satisfactory account of authority must dis-
pense with the result-of-a-cognitive-achievement picture, tout court. For,
so long as psychological self-ascriptions are seen as reports on one’s own

43
This is not to say that one’s own avowals of one’s mental states as commitments are always correct
but only that they are open to a very limited form of error: they are incorrect either because of
conceptual incompetence or because of slips of the tongue.
 7 Constitutive Theories 201

mental states, the question arises of whether they are true or false. Unless
one is prepared to suppose that our cognitive faculties may be infallible,
one could not account for the claim that authority holds as a matter of
conceptual necessity.
One strategy may consist in maintaining that, since there are not
counterexamples to authority, any competent and sincere assertion of “I
believe that P” (or of “I desire/intend/wish/hope that P/to φ”) would
entail that one has the corresponding first-order belief (or other proposi-
tional attitude, as a commitment). Still, this would hardly be an explana-
tion of why authority holds, but, rather, a simple acknowledgement, or
a consequence, of the fact that it does.44 Again, it is only if one is in the
grip of a Wittgensteinian anti-explanatory mantra that one could leave
matters at that.
The explanation of authority takes two different routes, depending on
whether one forms the belief as a commitment that P through its very
self-ascription or on whether one just gives expression to it by means
of its self-ascription. Let us look at them in this order. In the former
case,45 the crucial point is that when “I believe/desire/intend/wish/hope
that P”—that is, the corresponding psychological self-ascriptions are
acquired along the lines developed so far, as ways of making the same
commitments as the ones undertaken by judging and asserting “P” or “P
is worth pursuing/having”, while having in view reasons in favour of P
(is worth pursuing/having)—, it then becomes possible to use “I believe/
desire/intend/wish/hope that P” in lieu of asserting (or judging) that P
(is worth pursuing or having), in order to form one’s first-order belief
or desire that P directly. The difference between forming the first-order
mental state by means of the second-order judgement, instead of forming
it by means of the first-order one, is just the fact that “I believe/desire/
intend/wish/hope that P” makes explicit what the first-order judgement
leaves implicit—namely, that that is just one’s own particular standpoint

44
As we have seen in §3, this would be Bilgrami’s strategy.
45
As we saw in Chap. 5 (§2), Burge accords pride of place to cogito-like performatives and consid-
ers the ones in which one self-ascribes other propositional attitudes (as commitments) impure,
since they may be affected by error if one were cognitively malfunctioning. In my view, cogito-like
cases are by far less interesting than the other ones, as they would merely bring about episodic
thoughts rather than propositional attitudes as commitments.
202 A. Coliva

on P (or its being worth pursuing or having). Hence, we can bring about
the relevant first-order belief and other propositional attitudes (as com-
mitments) either by judging that P is the case (or is worth pursuing or
having) or by judging “I believe/desire/intend/wish/hope that P”, while
having in view reasons in favour of P (is worth pursuing/having). For,
to repeat, given the role of the latter locutions (either in speech or in
thought) and of “P (is worth pursuing or having)”, I can commit myself
to P (is worth pursuing or having), thus bringing about the correspond-
ing first-order beliefs or other propositional attitudes, either by simply
judging the latter or by judging the former, thus simultaneously making
explicit my commitment to P’s being the case (or to its being worth pur-
suing/having).46 According to this model, there would be a sense in which
it is literally true that we make up or create our minds. Moreover, since
one’s self-ascription would bring into existence the relevant first-order
mental states, those judgements would necessarily be true—in fact, self-
verifyingly so.47 Furthermore, there would be no temptation to think that
one should have the first-order mental state in view first, in order to make
one’s judgement, which thus would result in knowledge. For, if there
is no mental state before making the relevant assertions or judgements,
then of course there is nothing to know, or be aware of, in the first place,
which should be tracked in judgement.
But not all self-ascriptions of propositional attitudes as commitments
have this performative nature of bringing into existence the very first-
order mental state they are about. A lot of them simply avow a first-order
mental state which is, nonetheless, our making, yet not through its very
46
Obviously, it would always be available to one to justify one’s judgement—“I believe/desire/
intend/wish/hope that P˝, say—ex post by appealing to the fact that one’s evidence allow(ed) one to
judge that P (is worth pursuing or having), thus giving rise to one’s belief/desire/intention/wish/
hope that P (as a commitment). The possibility of giving such a justification for one’s judgement “I
believe/desire/intend/wish/hope that P˝, however, should not obscure the fact that the commit-
ment to P (is worth pursuing or having) was actually made by judging “I believe/desire/intend/
wish/hope that P˝. My account of attitudinal avowals then explains why philosophers, most nota-
bly Wittgenstein in The Philosophical Investigations II, x, have been tempted to reduce “I believe
that P˝ to “P˝. Their mistake was due to failing to see that the contents of those judgements are
different, whereas their insight was to recognize that the commitments undertaken by making those
judgements (or the corresponding assertions) are virtually the same. See the Appendix for an
account of Moore’s paradox.
47
Also, the corresponding assertions would have the same effect, as long as sincerity is granted. For
an account of what this means, in the present context, there will be more in the following.
 7 Constitutive Theories 203

self-ascription, but by merely judging that P is the case or that Q is worth

pursuing. The previous account of transparency comes in handy at this
stage, for it allows us to avoid falling back into any epistemic account. For
the relevant psychological self-ascriptions are just ingrained alternative
ways of expressing those very first-order judgement-dependent proposi-
tional attitudes. Hence, they give expression to those first-order mental
states while making explicit that they are just a subject’s own point of
view onto the world or on what is worth doing. And just as there is no
issue that a cry might not express pain (assuming sincerity, of course),
similarly there is no issue that one’s psychological self-ascriptions elicited
by one’s judging that P is the case or that Q is worth pursuing may not
be correct. The only possibility of error would depend on a slip of the
tongue, but clearly such a case would not count as a mistaken identifica-
tion of one’s ongoing mental state.
Going back to the performative function of some occurrence of “I
believe/desire/intend that P/Q/to φ” before turning to a defence of such
a claim, let me address one possible objection to this strong form of con-
stitutivism. Some theorists have argued against it on the grounds that it
would entail the unreality of first-order mental states48: if such mental
states do not pre-exist their self-ascription, then they do not have real and
independent existence. Since this is implausible, constitutivism is doomed
from the start. In response, I think it should be stressed that my brand
of constitutivism entails only that mental states as commitments do not
necessarily have independent existence of the corresponding second-order
judgements. Hence, the kind of constitutivism I am advocating allows for
(first-order) mental states as dispositions to exist independently of the cor-
responding self-ascriptions. Moreover, it also allows for the conceivability
of the independent existence of (first-order) beliefs and desires based on
judgement, when they are merely brought about by judging that P (is
worth pursuing or having), by subjects who do not yet have the concep-
tual resources necessary to make the corresponding second-order judge-
ment. Furthermore, it is part and parcel of the position advocated here
48
See, for instance, Bar-On 2004, pp. 412–413. Notice how Bar-On finds support against this view
in general, from its implausibility in the case of those mental states “we share with non-human
animals and pre-cognitive children”. But judgement dependency should not, I think, be meant to
apply to the latter cases. See also Heal 2002, p. 286.
204 A. Coliva

that, under suitable conditions, the relevant psychological self-ascriptions

do bring about the corresponding first-order propositional attitudes as
commitments. Contrary to Wright’s proposal (see §2), then, there is no
danger of falling into any form of irrealism about them.
Hence, it has to be stressed that the fact that the existence of certain
beliefs and other propositional attitudes—those as commitments—is
taken to depend on judging I believe/desire/intend/wish/hope that P does
not make those mental states less real. The crucial point is that judgement
dependence is a claim about the provenance of first-order mental states,
not about their (un-)reality. What I have been urging is that there are
two kinds of judgements that can bring about the same result (that is,
a belief/desire/intention/wish/hope as a commitment)—namely, either
judgements that are outright about the world (or what is worth doing or
having) or judgements that make explicit the particular standpoint from
which the world is conceived to be thus-and-so.
It may sound surprising, if not altogether alarming, that in my view
first-order judgements and second-order ones play such an interlocking
role. But to dissipate the resistance to such a view, consider that any
judgement/assertion that P (is worth pursuing or having) is always made
by someone and hence is necessarily the expression of a subject’s point
of view, even when its subject matter is, as it were, the world. By being
drilled to the use of first-person present-tense psychological vocabulary,
subjects are simply endowed with the means to make that “grammatical”
fact explicit and to take on responsibility for it. Once they are conversant
with that practice, they can use second-order judgements or assertions
directly, as ways of forming the same commitments they would form by
making the corresponding first-order ones. It then seems to me that the
point of our psychological self-ascriptions is first and foremost to make
explicit to ourselves and others the fact that the world, broadly conceived,
is always described or assessed from a particular standpoint—one among
potentially many. Their further performative role is simply a result of
having being trained to take part in a linguistic practice where the same
commitments can be undertaken in two different ways. This, I take it, is
also the deep truth in Wittgenstein’s and Wright’s positions: psychologi-
cal avowals—and their equivalents in thought—are the result of being
trained to take part in a linguistic practice and have their main point in
 7 Constitutive Theories 205

making any participant aware of the fact that her specific point of view—
formed through an act of rational deliberation—is just one among other
possible ones.
When understood in the way proposed, a judgement (or a sincere
assertion49) such as “I believe/desire/intend/wish/hope that P” is like a
performative—namely, like “I promise to buy you an ice cream”, “I hereby
thee wed”, “I hereby name you so-and-so”, and so on: it makes a certain
thing happen, for it does create the first-order propositional attitude as a
commitment. Where this, to repeat, is possible precisely because judging
“I believe/desire/intend/wish/hope that P” becomes just an alternative
way of undertaking the same commitments one would make by judging
that P (is worth pursuing or having, once provided with the relevant psy-
chological concepts), save for the fact that the former kind of judgement
would also make explicit what the latter leaves implicit—namely, that the
judgement reflects just a subject’s own point of view.
Many have argued against such a view of psychological self-ascriptions
by maintaining that it would commit one to the implausible claim that
they would lack content and could not, therefore, be sensibly prefaced
by negations, or be embedded in suppositions, and otherwise wider con-
texts.50 But this objection, if sound at all,51 could be raised only in the case

49
Notice the sincerity condition placed upon second-order assertions, which cannot, however, be
carried over to judgements, since judgements are—when made—necessarily sincere. This seems to
me enough to dispel the worry that one may judge or assert “I believe that P˝ and yet not form the
corresponding first-order mental state as a commitment. Of course, what remains entirely possible
is that I do not act on the basis of such a commitment and thus fail to actualize the connected
disposition. But this is no objection to the view which is being proposed here.
50
The locus classicus is Geach 1965.
51
The counter is usually that they could retain minimal assertoric content as well as minimal truth.
Accordingly, it would suffice for minimal assertoric content that performatives can be embedded in
negations and suppositions and that they can undergo the usual tense transformations. Obviously
they can. For “I am not going to buy you an ice cream” or “Suppose I bought you an ice cream”
and “I did buy you an ice cream” are perfectly sensible things to say. Minimal assertoric content
then pairs with minimal truth—with the idea that it is enough to qualify as a truth predicate that
some platitudes—in particular, negation and the T-schema—are respected. It would be a further
issue whether these statements should be taken as descriptions or as expressions of commitments one is
(or is not) undertaking thereby. This strategy can be found in Hacker (1986, p. 90) as well as in
Wright (1992, p. 28) and in Jacobsen (1996). Hacker (1986, p. 298), however, denies that minimal
assertoric content would be compatible with truth evaluations. Be that as it may, when we consider
judgements (and assertions) of “P˝ (is worth pursuing/having), we could distinguish between the
content of the judgement (or of the assertion), which obviously is truth-assessable, and what is being
206 A. Coliva

of implicit performatives—that is, those which do not make explicit the

kind of commitment one is undertaking. For explicit performatives, like
“I promise to buy you an ice cream” and “I believe that P”, are speech-acts,
which can have more than one function at the time: they can make things
happen but they can also say what is being done by means of them.52 In
this latter sense, they would retain truth-evaluable content. For instance,
“I promise to buy you an ice cream” is both a way of making the prom-
ise of buying you an ice cream and of saying what I am doing—namely,
promising to buy you an ice cream. Of course, what I am saying could be
false, since I could be insincere. Similarly, “I believe/desire/intend/wish/
hope that P” would be both a way of forming the commitment that P (is
worth pursuing/having) and of saying what I am doing. Moreover, what
I am saying—that is, that I believe/desire/intend/wish/hope that P/to φ
(as a commitment)—could be false, since I could, in fact, not be making
that commitment and simply trying to fool you.53
However, whenever sincerity conditions are introduced, one might
suspect that, perhaps surreptitiously, reference is being made to the fact
that one’s assertions or judgements should track one’s pre-existing mental
states. Hence, the whole point of the proposed account would be pre-
empted, for its main contention is precisely that first-order mental states
can be brought into existence by one’s making the relevant psychologi-
cal judgements or assertions, leaving no room for the idea of tracking a

done by means of the act of judging it. It is the act of judging that P that brings about the corre-
sponding belief (or further propositional attitude) as a commitment, which can get expressed in
one’s assertion of “P (is worth pursuing/having)”, although the content of one’s assertion would
remain P (is worth pursuing/having). And when the psychological self-ascription functions merely
as an avowal, it expresses one’s commitment while also saying that one has it. Hence, an avowal too
can have minimal assertoric content. A similar distinction can be found in Jacobsen (1996, p. 26)
and in Bar-On (2004, pp. 251–264).
52
A similar point can be found in Heal (2002, pp. 282–288) but also in Jacobsen (1996, pp. 23–28).
Strangely enough, Jacobsen, who, officially, sets out to characterize an expressivist account of avow-
als, in fact ends up defending the claim they are performatives. See Jacobsen (1996, pp. 26–28).
Thus, I agree with much he says, although I would insist on the difference between expressivism
and constructivism: on the former, first-order mental states are already there and get simply
expressed by the relevant utterances; on the latter, in contrast, utterances (if sincere) bring about
first-order mental states, at least in some cases.
53
Similar considerations could be made for the case in which psychological self-ascriptions do not
have a performative function but merely an expressive one. For one indeed may try to fool one’s
interlocutor by means of fake expressions of the relevant first-order mental states.
 7 Constitutive Theories 207

pre-existing mental reality. But, in effect, this account of sincerity is not

compelling. For the sincerity condition, in the case of performatives,54
just amounts to one’s lack of the intention to fool one’s interlocutor and does
not consist in a correspondence between one’s pre-existing (first-order)
mental states and utterances (or judgements about such mental states).
By contrast, when I do wish to deceive my interlocutor, it is not the case
that I first check within myself whether I have the belief or the desire that
P, find out I do not, say, and then say the opposite. Rather, I utter the
performative sentence without respecting one of its felicity conditions—
since I have another mental state—that is, the intention to fool you. Thus,
although I utter a performative sentence, I do not thereby bring about
the corresponding first-order mental state.55 That is why “I believe/desire/
intend/wish/hope that P” can be performatives and yet, on certain occa-
sions, be false, since what would make them true has not actually been
brought about. (Conversely, once the sincerity condition is satisfied, “I
believe/desire/intend/wish/hope that P” is true because what would make
it true has been brought about by that very judgement or assertion.)
One frequent objection raised against performative accounts of psy-
chological self-ascriptions, as well as against expressivist ones, is that
they would introduce a difference in meaning between the first-person,
present-tense use of the relevant psychological verbs and their third-
personal use (as well as their first-personal non-present-tense use).56 Since
this seems absurd, one should reject such an account of psychological
avowals. However, I do not think this objection is compelling. For, if the
meaning of a word is what is offered as an explanation of its meaning,57
then we will offer just one kind of explanation for “believe”, “desire” and
so on. For instance, that to believe that P means to be disposed, ceteris
paribus, to use P as a premise in one’s practical and theoretical reasoning;

54
As well as in the case of avowals. See previous footnote.
55
Notice that here an asymmetry between assertions and judgements may arise. For the performa-
tive judgement cannot be overridden. So, one way of putting the point is that when I am insincere
I utter a performative sentence without making the corresponding judgement.
56
Cf. Geach 1965, p. 260. As we saw in Chap. 6 (§1), this was indeed Wittgenstein’s unpalatable
position.
57
See also Jacobsen (1996, p. 18) for a similar point, which can be traced back to Wittgenstein (Cf.
Philosophical Investigations, §560).
208 A. Coliva

that in order to believe that P one needs evidence in favour of P, that if

one believes that P, then, ceteris paribus, one will assert that P, and so on.
But nowhere did the proposed account suggest that in the first-person
case things should be any different. After all, on the present proposal,
all that is being suggested is that the commitment to P can be formed
by judging (or asserting) “I believe that P”, where it is part and parcel of
making such a commitment that one should have the kind of dispositions
just mentioned or, in the absence of them, to be self-critical or to accept
criticism for not living up to one’s commitments.58 Thus, allowing for
the performative nature of first-personal, present-tense avowals merely
entails that there are two ways in which a subject can come to have cer-
tain dispositions (or in fact ought to come to have them): either because
she is somehow finds herself saddled with them or, as in the case at hand,
because she actively brings them about—or tries to bring them about—
as an implementation of her own deliberations.
However, what this suggests, in its turn, is that there is another, per-
haps more important distinction to be drawn—that is, a distinction
between self- as well as other-directed ascriptions of beliefs and other
propositional attitudes as commitments and as dispositions. True, we can
bring about commitments only for ourselves. Yet we may nevertheless
ascribe this kind of mental state to other people. Suppose you are listen-
ing to a subject who asserts that P, gives a lot of evidence in its favour,
bets her own head on P, as it were, and so on. When you then report on
her by saying “S believes that P”, obviously what you would be correctly
attributing to her is a belief as a commitment. If, in contrast, you were
interpreting S’s behaviour by attributing to her a mental state she has
never avowed (and might never be in a position to avow), which, how-
ever, would be helpful to you to make sense of what S is doing, or if you
were engaging in deep analysis of a Freudian kind of her behaviour, then
you would be attributing to her a belief as a disposition.

58
Hence, I would not be too keen to endorse the kind of dilemma Wright thinks there is in the fact
that, on the one hand, avowals are authoritative and, on the other, they self-ascribe a disposition
whose obtaining is assessable from a third-personal point of view. See Wright (2001a). Jacobsen
(1996) and Heal (2002) too seem to be highly struck by this dilemma. The dilemma, however,
seems to me very much a function of a simplistic description of the situation. Hence, it calls more
for a dissolution—for an account of why it does not really stand—than for a positive solution.
 7 Constitutive Theories 209

Furthermore, not all uses of “believe” (or other propositional attitudi-

nal verbs) in the first-person present are performatives (or indeed avowals
expressive of one’s first-order mental states). For, sometimes, the same
judgement or assertion can be used as a simple description, like when
one finds out about one’s own beliefs or desires through a process of
self-interpretation. Conversely, both past and otherwise embedded uses
of “believe” (or of other propositional attitudinal verbs) in the first per-
son, though not themselves performatives—for they cannot bring about
a commitment—may nevertheless be an ascription of a commitment one
had in the past, based on the memory of having made it, or, for instance,
of a commitment one is supposing to be making.59
So, what we witness here is a variety of uses of self- as well as other-
directed psychological ascriptions: although it remains that we can bring
about commitments only in our own case, we can ascribe commitments
to both ourselves and others; moreover, we can ascribe to both ourselves
and others mental states as dispositions. The reason why it is so is simple:
each of us can deliberate only for herself, but we can, and obviously do,
see other people as deliberative agents—that is, we do know when they
are making commitments as opposed to when they are simply acting on
the basis of mental states which they may be saddled with but which are
not the result of any deliberation of theirs. The same, however, applies to
ourselves too, so we can report on previously made (and perhaps already
abandoned) commitments or engage in the supposition of undertaking
them; but we can also self-ascribe mental states as dispositions. What self-
interpretation and psychoanalysis help us do is to acquire that kind of
(third-personal) knowledge of the latter kind of mental states. What they
cannot do, however, is to turn us into deliberative agents with respect to
them, for no amount of theoretical knowledge about ourselves will, by
itself, ever transform the mental states we thereby become aware of into
commitments.60
Thus, what should be claimed is that we have two different notions
of belief and other propositional attitudes—those as commitments and
those as dispositions—that cut across the first-person/third-person divide.

59
There will be more about these cases in the next chapter (§1).
60
In fact, they may make action and deliberation even more difficult to attain.
210 A. Coliva

Indeed, we do explain their meanings differently, as we have seen before.

Nevertheless, the propensity to see them as two different species of the
same genus, instead of altogether different kinds of mental states (beliefs
and “shbeliefs”, say), and thus to talk, in both cases, of beliefs (speci-
fied “as commitments” or “as dispositions”, respectively) would then be
explainable by reference to the fact that both beliefs (and desires) as com-
mitments and as dispositions could be responses to evidence (although
only beliefs and desires as commitments would depend on actively assess-
ing it) and could have similar effects at least on our non-linguistic behav-
iour: in the case of either kind of belief, a disposition to behave on the
basis of P and, in either kind of desire or intention, a disposition to bring
about P. What is relevantly different is the way in which the respective
self-ascriptions are made. For when “I believe/desire/intend/wish/hope
that P” is judged or sincerely asserted to make a commitment, it is a
performative and brings about the corresponding first-order mental state.
But the same words or mental content can also be a report on one’s dis-
positions, known to oneself through observation and inference on one’s
own behaviour, hence in a third-personal way.61
A related point has to do with an oft-made observation that “the mark
of the mental” would be the first-personal, present-tense use of psycho-
logical verbs.62 Well, on the face of it, this remark is simply wrong. For
some, present-tense, psychological self-ascriptions are made in a third-
personal way, as we have seen, and are such that one is self-ascribing a
mental state as a disposition. By contrast, we have already noticed that
there can be third-personal, present-tense ascriptions of mental states
which, though not performatives, are nevertheless ascriptions of com-
mitments. What I think is distinctive about (most) adult human beings’
mentality, then, is both that we can make commitments and that we can

61
Notice, moreover, that they can also be used to report a commitment previously undertaken. In
such a case, they are based on remembering having made such a commitment and certainly not on
inspecting one’s own mind, as it were. Also, it may be possible to find out about one’s own disposi-
tions in a third-personal way and self-ascribe them and subsequently form a corresponding com-
mitment. All this would require two different mental actions, although the final self-ascriptions
may be identical in form. We dwell on both issues in the next chapter.
62
With the usual caveats having to do with difficulties in exegesis, Wittgenstein seems to have had
this view, as Jacobsen (1996, pp. 14–17) reminds us.
 7 Constitutive Theories 211

actually see others as capable of deliberative mental agency—that is, as sub-

jects who are capable of making commitments.
Finally, another objection often raised against performative accounts
of avowals is that “P” and “I believe that P” would turn out to have the
same content.63 This objection, however, is wrong because, obviously, the
truth conditions of the two sentences are different: “It is raining (at l at
t)” is true iff it is raining (at l at t), whereas “I believe it is raining (at l
at t)” is true iff I believe it is raining (at l at t). Clearly, these are quite
independent states of affairs—it may be raining (at l at t) and I may be
ignorant of it, or I may believe it is raining (at l at t) and be wrong. What,
however, would be identical in the two cases, according to the proposed
account of (as we can now say) some uses of “I believe that P”, are simply
the commitments one would undertake by judging or asserting either.64
This is why at least certain occurrences of “I believe that P, but it is not
the case that P” would be Moorean-paradoxical. For when “I believe that
P” is an expression of a commitment to P’s truth, it would be (at least)
irrational to commit oneself to P’s falsity, as well.65
The constitutive account of our knowledge of our own propositional
attitudes I have proposed, if correct at all, shows how constitutivism
worth its name will have to take a rather radical form. It will have to do so
if, while sensitive to the idea that there is no substantial epistemology, in
the case of self-knowledge, it is not prone to embrace an anti-explanatory
stance. This is evident in the account of concepts’ possession it provides
and in the kind of constructivist turn it takes to guarantee authority at
least in some cases. This makes constitutivism a viable explanation of
self-knowledge only for very specific and limited kinds of propositional
attitudes we can enjoy—those as commitments—and, connectedly, only
for specific kinds of self-ascriptions of propositional attitudes—those
which amount, in fact, either to performatives or to avowals of those very
63
This idea may be suggested by some of Geach’s observations (1965, p. 259). Wittgenstein is obvi-
ously considered the chief holder of this view, for his well-known remark that “I believe that P˝ is
just a tentative assertion of “P˝, which is key to his treatment of Moore’s paradox (Cf. (Wittgenstein
PI, II, xi, especially pp. 190–191). There will be more about Moore’s paradox in the Appendix.
64
In order to undertake these commitments, one obviously need not have the concept of a commit-
ment, nor should this claim be understood as implying that the content of “I believe that P˝, say, is
“I commit myself to P˝.
65
Cf. Appendix.
212 A. Coliva

first-order commitments. To my mind, this is no sign of irrelevance or

inadequacy, though. It would be only on a monist assumption. Yet what
this long and winding road has led us to see is that we do have a variety
not just of mental states but also of kinds of propositional attitudes as
well as of ways of knowing them and that such a variety can be adequately
accounted for only if we  embrace pluralism about self-knowledge.

5 Summary
In this chapter, we have seen several constitutive positions. They all sub-
scribe to the view that so-called self-knowledge is not the result of any
epistemic achievement—however minimal that might be. Rather, what
goes by that name is the constitutive thesis, according to which, given cer-
tain conditions C, one has a given propositional attitude iff one believes/
judges that one does. We have seen how Shoemaker (§1) is interested in
redeeming the left-to-right side of the thesis—that is, the one concern-
ing transparency. He does so by advocating that the C-conditions should
make reference to a subject endowed with rationality, intelligence and
the relevant psychological concepts. We have objected to Shoemaker’s
claim that self-knowledge is necessary for being a reasoner—that is, for
being able to engage in theoretical and practical reasoning and in belief
revision. In addition, we have maintained that even if it were necessary to
that end, that would just show that if one is such a reasoner, one would
have knowledge of one’s own mental states, but that would be silent on
the provenance of that knowledge. We have also seen how, ontologically,
for Shoemaker first- and second-order mental states do not have separate
existences.
We have then moved on to Wright’s account (§2), which is a vindica-
tion of the right-to-left direction of the constitutive thesis—that is to say,
the side of authority. We have seen how he embraces the Wittgensteinian
idea that this is just a feature of how our communal language game with
psychological avowals is played, which allows for undermining only
when there are positive reasons to think that a subject is self-deceived.
Thus, in order to secure the conceptual truth of the relevant side of the
constitutive thesis, the C-conditions must make reference to the absence
 7 Constitutive Theories 213

of self-deception, according to Wright. We have then pointed out that,

according to him, the relevant psychological self-ascriptions, though not
tracking any prior mental state, do not bring it about either. We have
then objected that this seems to entail an unsatisfactory form of psycho-
logical irrealism of an instrumentalist kind. Moreover, we have claimed
that Wright’s account is wanting on matters of detail, since it is based on
an objectionable account of self-deception and contains the problematic
suggestion that also avowals of past intentions are groundless, transparent
and authoritative.
We have then turned to Bilgrami’s account (§3), which aims to explain
both sides of the constitutive thesis. In his view, the C-conditions must
refer either to mental states as commitments, characterised as those
propositional attitudes one can reasonably be regarded as accountable
for, when authority is concerned, or to those propositional attitudes that
enter the explanation of actions that are the object of justifiable reactive
attitudes, when transparency is at stake. Furthermore, we have seen how,
for Bilgrami, self-deception is no bar to authority since, in his view, it
consists in having two opposite mental states, one as a disposition and
one as a commitment, where one retains one’s authority over the latter.
We have objected to Bilgrami’s inclusion of dispositional mental states
among the ones for which transparency holds as well as to his defence
of the idea, which we agree on, of intrinsically normative mental states.
Once more, we have also found the position wanting for it makes self-
knowledge a necessary condition for justifiable reactive attitudes, but it
does not say anything about its nature and source.
In general, the previous constitutive proposals are found wanting when
they refuse to explain how self-knowledge may come about and simply
state that it is a necessary condition for something else. Although this
attitude is understandable when placed in the context of Wittgensteinian
exegesis, it is less so when one abstracts from Wittgenstein’s meta-
philosophical pronouncements. Refusing to embrace his anti-explanatory
stance, moreover, need not push us back into an epistemic account of
self-knowledge. For, as we have shown in §4, the problem may just be
the one of explaining how we acquire and canonically apply first-personal
attitudinal concepts. Furthermore, to the exception of Shoemaker’s
position, the constitutive proposals reviewed so far are silent or, worse,
214 A. Coliva

problematic, when it comes to the metaphysical issue of the nature and

provenance of our propositional attitudes and of the relationship between
first-order mental states and their self-ascriptions.
Hence, we have proposed a metaphysically robust kind of constitutiv-
ism (§4), according to which both sides of the constitutive thesis need
explaining. The left-to-right one is taken to hold only for propositional
attitudes as commitments. For one cannot be self-critical or accept criti-
cism for propositional attitudes one is not knowledgeable about. Now,
propositional attitudes as commitments are constitutively tied to a sub-
ject's ability to be self-critical, or to accept criticism for them.  But it
is also explained how knowledge of one’s commitments—that is, their
self-ascription—is achieved. The key idea is that a subject is taught to use
first-person, present-tense attitudinal verbs “blindly”—that is to say, by
replacing the assertion/judgement of “P” or “P would be good to have”
with “I believe that P” and “I want/desire that P”, under appropriately
specified conditions. This way, it is also clarified in what sense first-order
mental states and their self-ascription may not be separate existences. Just
as a cry or an assertion of P may be simply the (behvioral at large) mani-
festation of one's on-going mental state, so can be one's psychological
self-ascription of the relevant sensation or of the relevant belief.
After defending the proposal from a number of objections which could
be raised against it, we have moved to the other side of the constitu-
tive thesis. We have agreed with Bilgrami’s account of self-deception and
offered further motivations in its favour. The relevance of that account
is that while it acknowledges the possibility of self-deception, it explains
it in such a way that that phenomenon ceases to be a counterexample to
authority. We have then explained why authority holds as a matter of con-
ceptual necessity. To that end, we have put forward a position, according
to which, once we are capable of making psychological self-ascriptions in
the way proposed, we can actually bring about first-order propositional
attitudes as commitments just by making those very self-ascriptions,
which are then self-verifyingly true. Or else, we may avow first-order
mental states, yet their immediate expression through the relevant self-
ascription is as secure as it can possibly be. In both cases, however, either
insincerity or slips of the tongue may occur. Hence, there is the possibil-
ity of false psychological self-ascriptions but not of having gone wrong in
identifying one’s first-order mental states. We have defended the proposal
 7 Constitutive Theories 215

from several possible objections but also pointed out that its scope is
limited only to commitments and to subjects endowed with the relevant
psychological concepts. It is therefore necessary to integrate it with dif-
ferent ones in order to account for our first-personal knowledge of other
mental states and, obviously, for third-personal knowledge of other kinds
of propositional attitudes and further mental states.

Bibliography
Bar-On, D. (2004). Speaking my mind. Expression and self-knowledge. Oxford:
Oxford University Press.
Bilgrami, A. (2006). Self-knowledge and resentment. Cambridge, MA: Harvard
University Press.
Bilgrami, A. (2012). The unique status of self-knowledge, in Coliva A. (2012),
263–278.
Boghossian, P. (1989). Content and self-knowledge. Philosophical Topics, 17,
5–26.
Boyle, M. (2009). Two kinds of self-knowledge. Philosophy and Phenomenological
Research, 77(1), 133–164.
Boyle, M. (2011a). Transparent self-knowledge. Aristotelian Society Supplementary
Volume, 85(1), 223–241.
Boyle, M. (2011b). Making up your mind. Philosophers’ Imprint, 11(17), 1–24.
Coliva, A. (2009). Self-knowledge and commitments. Synthese, 171(3),
365–375.
Coliva A. (2012c) (ed) The Self and Self-Knowledge (Oxford: Oxford University
Press).
Coliva, A. (2012d) “One variety of self-knowledge: constitutivism as construc-
tivism”, in A. Coliva (ed.) The Self and Self-Knowledge, Oxford, Oxford,
University Press, pp. 212–242.
Davidson, D. (1970). Mental events. In L.  Foster & J.  W. Swanson (Eds.),
Experience and theory. London: Duckworth. Reprinted in Essays on actions
and events (pp. 79–102). Oxford: Clarendon Press.
Dennett, D. (1987). The intentional stance. Cambridge MA: MIT Press.
Frith, U., & Happé, F. (1999). Theory of mind and self-consciousness: What is
it like to be autistic? Mind and Language, 14(1), 1–22.
Geach, P. (1965). Assertion. The Philosophical Review, 74, 449–465. Reprinted
in Rosemberg, J. F. & Travis, C. (eds) (1971). Reading in the philosophy of
language (pp. 250–261). Englewood: Prentice-Hall.
216 A. Coliva

Hacker, P. (1986). Insight and illusion: Themes from the philosophy of Wittgenstein.
Oxford: Clarendon Press.
Heal, J. (2002). First person authority, Proceedings of the Aristotelian Society, 102,
1–19. Reprinted in (2003). Mind, Reason and Imagination (pp. 273–288).
Cambridge: Cambridge University Press.
Jacobsen, R. (1996). Wittgenstein on self-knowledge and self-expression.
Philosophical Quarterly, 46, 12–30.
Kripke, S. (1982). Wittgenstein on rules and private language. Oxford: Blackwell.
Peacocke, C. (2012). Conscious events and self-ascriptions: Comments on heal
and O’Brien, in Coliva A. (ed), 180–188.
Povinelli, D. J., & Vonk, J. (2004). We don’t need a microscope to explore the
chimpanzee’s mind. Mind and Language, 19(1), 1–22.
Stoneham, T. (1998). On believing that I am thinking. Proceedings of the
Aristotelian Society, 98, 125–144.
Strawson, P. (1962). Freedom and resentment. Proceedings of the British Academy,
48, 1–25.
Tomasello, M., Call, J., & Hare, B. (2003a). Chimpanzees understand psycho-
logical states—The question is which ones and to what an extent. Trends in
Cognitive Science, 7, 153–156.
Tomasello, M., Call, J., & Hare, B. (2003b). Chimpanzees versus humans: It’s
not that simple. Trends in Cognitive Science, 7, 239–240.
Williams, D. (1994). Somebody somewhere. London: Corgi Books.
Wright, C. (1992). Truth and objectivity. Cambridge MA: Harvard University
Press.
Wright, C. (2002). What could antirealism about ordinary psychology possibly
be? The Philosophical Review, 111, 205–233.
Wright, C. (2015). The reality of privileged access. In S.  C. Goldberg (Ed.),
Externalism, self-knowledge and skepticism (pp.  49–74). Cambridge:
Cambridge University Press.
 8
Pluralism About Self-knowledge

In the previous chapter, we saw how a strong form of constitutivism, entail-

ing the metaphysical thesis that, at least on occasion, propositional atti-
tudes can be brought about, or can be constituted by their self-ascriptions,
has only limited purchase. For it can hold only when one actually deliber-
ates, through the self-ascription, to believe/want or intend such-and-so
(on the basis of reasons). This act is similar to the performative act of
promising to do something by judging/saying “I promise to do such-and-
so”. We also saw how we can grant that a subject who is capable of beliefs
(and other propositional attitudes) as commitments needs to know them,
in order to be self-critical or to accept criticism for not living up to them.
That, however, requires an explanation of how she can be in a position
to make correct self-ascriptions of the relevant attitudes. The problem of
self-knowledge, seen from that angle, ceases to be an epistemic one and
becomes, or identifies with, the problem of explaining how the relevant
concepts may be acquired and applied. Still, in addressing this problem,
one needs to avoid (surreptitiously) falling back into either the observa-
tional or the inferential model, according to which the relevant concepts
are acquired and applied either by having a given mental state in view and
by categorising it for what it is or by—tacitly or explicitly—applying the

© The Editor(s) (if applicable) and The Author(s) 2016 217

A. Coliva, The Varieties of Self-Knowledge,
DOI 10.1057/978-1-137-32613-3_8
218 A. Coliva

rule that individuates the relevant concept. For doing so will, in a large
amount of cases, presuppose knowledge of one’s own mental states already.
We have proposed—in line with some recent expressivist pronounce-
ments, which are otherwise silent on the particular kind of propositional
attitudes avowed—that the relevant concepts are acquired “blindly”, by
being drilled to use them to give alternative expression to one’s mind,
whenever one is already in a position to judge, based on reasons, that P
is the case or is worth having or doing, and therefore has the correspond-
ing judgement-dependent belief, desire or intention. The relatively late
appearance of psychological concepts—or the relatively late development
of simple tags into concepts, as Gordon would have it—in human beings’
cognitive repertoire thus is imputed to the fact that it takes time for them
to acquire the ability to have these kinds of propositional attitude, rather
than to the fact that it takes time for them to develop a theory of their own
and other minds. Sensitivity to reasons is—according to the picture we
have sketched—what takes time to develop. Once one has acquired such
an ability, psychological concepts come virtually for free. For, if one has
“blindly” acquired the relevant tags already, one needs only using them to
give expression to one’s judgement-dependent propositional attitudes. If,
in contrast, one does not yet have those tags, one simply needs to acquire
them—and do so “blindly”—by being taught to substitute one’s direct
expression of judgement-dependent belief, desire or intention with “I
believe/desire that P” or “I intend to φ”.
This kind of constitutivism, however, cannot be applied outside its
proper boundaries. For one thing, it cannot be of avail in connection
with avowals of sensations or basic emotions, for it does not seem right
that we do bring them about by self-ascribing them. If there is room for
some, much weaker form of constitutivism in this connection, as we see
in §2, there is none for the metaphysically robust variant I have been
proposing for propositional attitudes as commitments. For another, it
cannot be applied to our knowledge of dispositional mental states, let
them be propositional attitudes or the dispositional elements of complex
emotions, as is shown in §3. As we shall presently see (§1), however,
the strong form of constitutivism proposed at the end of the previous
chapter has further limitations when it comes to self-ascriptions of past
propositional attitudes as commitments too. A proper, necessarily plural-
ist account needs to take care of them too.
 8 Pluralism About Self-knowledge 219

1 Propositional Attitudes as Commitments:

The Limits of the Constitutive Account
We often engage in past self-ascriptions of propositional attitudes as
commitments. For instance, we judge and say that, at some point in
our life, we decided to do thus-and-so or came to believe thus-and-so.
Metaphysically robust constitutivism cannot be applied to these kinds
of self-ascriptions, though. For, as we saw while discussing Wright’s pro-
posal (Chap. 7, §2), we can remember our past commitments incorrectly,
even if this fact is often unverifiable since we do not usually avow them
when we form them. Given the particular—and by all means contin-
gent—structure of this practice of ours, it is true, as Wright has it, that
we are presumed to be right unless there are reasons to doubt of our self-
ascriptions, but this does not mean that it is a conceptual truth that if one
who has the relevant psychological concepts and is cognitively lucid and
rational self-ascribes a past propositional attitude, one really did have it.
Occasional mis-remembering—not to be confused with not remember-
ing anything at all—is like mis-seeing. For it does not impugn a subject’s
rationality and conceptual mastery. This poses the problem of explaining
our knowledge of our past mental states of a commissive nature. In order
to clarify the issue, we must make two preliminary remarks. First, we are
not here thinking of past psychological self-ascriptions based on ongoing
self-interpretations. Hence, we are not concerned with cases in which,
thinking of our past behaviour, we end up concluding, through reasoning
or even in a flash, as it were, that we believed, desired or intended to do
thus-and-so at some earlier time. These are simply cases of third-personal
self-knowledge of one’s past propositional attitudes as dispositions we
will take into account in §3. Second, we are not here thinking of the
relatively rare cases in which, after forgetting what one was about to do,
say, one in effect deliberates what to do anew through a self-ascription
of a past intention, which camouflages the newly formed intention, as it
were, so as to make it subjectively feel like a memory. Rather, we are con-
cerned with actual cases of self-ascriptions of commissive mental states
undertaken in the past.
Now, it must be registered that, under this rubric, there are actually
two distinct cases. The first one is when one actually remembers the
220 A. Coliva

actual deliberation one made. In this case, one remembers what went
on in one’s mind at a given moment in the past and makes a past self-
ascription based on that. In such a case, the self-ascription is based on
mnestic evidence and is akin to a judgement about one’s material sur-
roundings based on one’s perceptual evidence. In such a case, the self-
ascriptions is not groundless—not because one observes one’s past mental
state or because one infers to it starting from one’s behaviour and further
inner promptings to infer to its likely cause—but because one’s judge-
ment of having had that mental state is based on (mnestic) evidence that
stems directly from that mental event.1 In such a case, there is a pre-
sumption of truth—we usually remember our past, including our past
deliberations, correctly—but surely no conceptual guarantee that one’s
mnestic evidence is right and one’s past self-ascription correct.
This kind of case, though real, is not the norm, however. For it is often
the case that we do not actually have a recollection of the precise act of
deliberation we performed, in order to form a given propositional attitude
as a commitment. Yet we constantly engage in past self-ascriptions of the
relevant mental states. For instance, I have no recollection whatever of
when I decided to enrol for a first degree in philosophy. But I certainly
deliberated to that effect. Here again, the psychological self-ascription “I
decided to enrol for a first degree in philosophy” is actually based on mnes-
tic evidence, but not of one’s past mental state of deliberating to do such a
thing. The evidence has rather to do with one’s memories of some relevant
biographical element, like remembering having considered some options,
right after finishing high school, and having actually enrolled for a phi-
losophy degree when the time to enrol came, together with a recollection
of not being forced to do anything or of having enrolled going through
the motions. It is important to resist thinking of the self-ascription as a
case of inference to the best explanation. For, by judging that I did decide
to enrol for a first degree in philosophy, I am not just making sense of
who I am now by inferring that it is likely that at some point I deliber-
ated to do such a thing. I am sure I did make that decision, at some point

1
In order to avoid thinking of self-ascriptions of past commitments as based on having in view
one’s past mental states, it is useful to think of a subject who, having lost her mid- and long-term
memory, keeps a very accurate diary of her daily activities. She could still make past self-ascriptions,
yet in no sense would they be based on having in view one’s past mental states.
 8 Pluralism About Self-knowledge 221

in my life, but I do not have a recollection of it, just of its biographi-

cal antecedents and consequences, as it were. In this case too, therefore,
the self-ascription is not groundless—for it is based on mnestic evidence,
even though of a different kind from the one which, as we just saw, can
licence similar past self-ascriptions. Here again, there is a presumption of
truth, but no conceptual guarantee that one’s mnestic evidence is right
and one’s self-ascription correct. Hence, past self-ascriptions of commis-
sive propositional attitudes are not to be accounted for by means of a con-
stitutive account. Let me dwell a bit more on this issue. One may think
that even if one is not trying to make sense of one’s past actions through
formulating an explanatory hypothesis, one is nevertheless going through
an abductive inference. The point would thus be merely terminological.
That is to say, one might want to single out, among abductive inferences,
those which depend on formulating interesting and perhaps even con-
troversial hypotheses based on complex theories, from those which do
not. However, it seems rather that one’s recollections allow one to infer
that, having satisfied the characteristic notes of a given psychological con-
cept, one instantiated it in the past. The reasoning, therefore, would be
of a deductive—not of an abductive—kind. There is still room for error,
though, because one may be oblivious to some aspect(s) of the concept
or of one’s past and therefore either misapply the concept or erroneously
think one did instantiate it.
Moreover, the relevant self-ascriptions are neither transparent nor
authoritative. For, on the one hand, one will need to keep memory of the
relevant evidence. Hence, it does not suffice to have had the relevant men-
tal state, and be rational and possessed of the relevant concepts, to be in a
position to make past psychological self-ascriptions. On the other hand,
one can mis-remember the evidence on which one’s past self-ascription is
based. One can, that is, think of having decided to enrol for a first degree
in philosophy when in fact one was forced by one’s parents to do so and
one might have removed this fact and started believing one had made the
decision oneself for all sorts of personal reasons. Of course, in the normal
run of cases, we do remember things correctly. Hence, there is obviously
a presumption of truth. Yet it is by no means conceptually guaranteed.
Still, if the presumption of truth were systematically flouted (in par-
ticular, if it turned out that we systematically do not remember or
222 A. Coliva

mis-remember our past commitments), that would cast doubt on our

being rational and cognitively well-functioning subjects as well as on
the fact that we can actually be held responsible now for what we did
in the past. It thus goes together with our idea of a rational and cogni-
tively well-functioning human being, who is responsible for her actions
and commitments through time, that she does keep track of her past
(mental) actions. Hence, having memory-based knowledge of them, in
the sense specified so far, is actually constitutive of what it means to be
such a subject. Yet, to repeat, that very knowledge is based on evidence
and is not itself to be accounted for along constitutive lines.

2 Sensations, Basic Emotions, Perceptions

and Experiences: Constitutivism Meets
Expressivism
Let us now turn to those mental states of ours, which have an intrinsic phe-
nomenology. Paradigmatically, we will be concerned here with sensations
such as pain, perceptions like seeing or perceptual experiences like having
a visual experience of a dog, and basic emotions like fearing a dog in one’s
whereabouts.2 As we saw in Chap. 2, these mental states differ in that not all
of them have relational elements and correctness conditions. This will intro-
duce some complications, but for now it is more important to stress their
common features—that is to say, their involving conscious experiences.

2.1 Sensations

Let us start with pain. Clearly, a metaphysically robust constitutive

account cannot apply to it. For there is no room here for the idea that
by self-ascribing pain one might bring it about. Still, a weaker kind of
constitutivism can be defended with respect to it. Accordingly, it will be
a constitutive feature of being a subject capable of having these mental

2
I think similar considerations can be put forward for self-ascriptions of ongoing thoughts, mental
images and imagination.
 8 Pluralism About Self-knowledge 223

states and possessed of the relevant concepts, as well as of being ratio-

nal, intelligent and cognitively well-functioning, that the following thesis
holds:

Constitutive thesis: Given conditions C, S is in pain iff S judges “I am in

pain”.

According to constitutivism, that thesis is necessarily and a priori true.

It is important, in order to defend it, to qualify the C-conditions ade-
quately. Let us consider the two sides of the biconditional in their turn.
As we already saw in Chap. 3 (§1), the left-to-right side of the bicon-
ditional can hold only for conceptually endowed subjects. That immedi-
ately excludes animals and infants from the domain of creatures for whom
the constitutive thesis is supposed to hold. Still, it does not exclude them
from the domain of creatures who can feel pain. Furthermore, we should
include reference to the fact that a subject is lucid, attentive and alert.
For, as we saw in connection with Snowdon’s apparent counterexamples
in Chap. 3 (§1), if a subject were under the effect of drugs or were over-
loaded or emotionally distressed, she could be in pain and yet not be in a
position to self-ascribe it.
Having clarified the C-conditions within which the biconditional is
supposed to hold, it is important to be explicit about what concepts’ pos-
session involves. The allegedly epistemic problem of self-knowledge, once
again, becomes the problem of explaining how we can acquire and apply
the concept of pain, without surreptitiously falling back into either the
observational model—whereby we should have that mental state in view,
perhaps with its intrinsic quale, and learn how to label it—or the infer-
ential one—according to which we would learn to apply that concept by
taking into account our behaviour and then by inferring to its likely cause.
It will not come as a surprise that, in my view, we should think of
the concept of pain as being acquired “blindly”—that is to say, by being
drilled to use it in lieu of more instinctive expressions of pain, such as
moaning and crying. Indeed, it is in connection with pain that the expres-
sivist account was first proposed by Wittgenstein, as we saw in Chap. 6
(§1), and has its clearest application. Still, “I am in pain”, while having
an expressive function, also has truth-conditional content and, as we saw,
224 A. Coliva

can be false if pronounced insincerely; in addition, it can be meaning-

fully negated, inserted in conditionals and turned into the past. As we
shall presently see, moreover, if one or more of the C-conditions fail to
be satisfied, the ensuing judgement may well be false.
This way, it becomes clear how the first-order mental state and its self-
ascription need not have separate existence, just as Shoemaker maintains
(see Chap. 7, §1). For, just as we do think of instinctive pain behaviour as
having the same core realisation of the purely phenomenal state, being its
immediate manifestation, similarly we should think of the self-ascription
of pain as having the same core realisation as the purely phenomenal state,
since it is an alternative kind of pain behaviour. Obviously, the ability to
self-ascribe pain will allow that mental state to play a more encompassing
role in one’s cognitive life. Yet all it is required to that end is—as Shoemaker
has it—a normal degree of intelligence, rationality and concepts’ possession.
Let us now turn to the other side of the constitutive thesis, according
to which, given certain conditions C, if one judges oneself to be in pain,
one is. Again, it is important to clarify the C-Conditions within which
the constitutive thesis is supposed to hold. A subject must possess the
concept of pain, acquired in the way described previously. Furthermore,
she must be lucid, attentive and alert. For, as we saw in connection with
Snowdon’s challenge (Chap. 3, §1), if she were under the effect of drugs
or were emotionally distressed, affected by extreme fatigue, and so on, she
could say and judge that she is in pain and yet not be in pain. As we also
saw in that chapter, our self-ascriptions of pain may be more complex and
involve bodily location or comparative judgements. Furthermore, they
may be issued in borderline cases. In case of bodily location, as we shall
see, authority is not supposed to hold; but that exceeds the mere mastery
of the phenomenal concept of pain. So it is no real counterexample to the
constitutive thesis when applied to purely phenomenal mental states and
avowals. Similarly, comparative judgements involve further conceptual
resources beside the ones involved in simple pain avowal. Once more, the
possibility that our comparative judgements may be wrong is no counter-
example to the constitutive thesis applied to purely phenomenal mental
states and avowals. Indeed, Snowdon’s cases present us with subjects who
are emotionally impaired and wish to retract their judgements. This would
not so much show that their original judgements—assuming those were
 8 Pluralism About Self-knowledge 225

simple ascriptions of pain—were not authoritative, but, rather, that those

subjects are no longer cognitively lucid and this either affects their recol-
lections or makes them wish they might retract their previous judgements.
Finally, there can be borderline cases—pains which border or even merge
into itches or vice versa—where our conceptual apparatus is elastic enough
to allow for either predication or indeed for what looks like retraction but
isn’t. Once again, it is important to be clear that the relevant concepts are
not tags of phenomenally distinct sensations, but rather ways of giving
alternative expression to them, rather than merely moaning and crying,
in the case of pain, and of moaning and scratching in the case of itches.3
Let us now turn to groundlessness: it is a feature of constitutive accounts
that they do not see our self-ascriptions of pain, in this case, as the result
of any epistemic achievement—let it be observational or based on infer-
ence to the best explanation, starting from patterns of overt behaviour.
Weak groundlessness is therefore respected. But strong groundlessness is
too, in the following sense: while there is a sensation going on, it is not its
intrinsic feeling, or quale, which makes it the sensation it is and that epis-
temically grounds its self-ascription, as if it were the evidence on which
the latter is based. That would push us back into the arms of the Cartesian
conception of the mental that constitutive accounts are supposed to be an
antidote to. Rather, there is a sensation, whatever that might be and feel,
which normally prompts certain patterns of behaviour, like screaming and
moaning. We are then taught to replace those instinctive manifestations
of pain with a self-ascription of pain. That concept then individuates the
sensation as pain and it will find its proper (canonical) application when-
ever those conditions obtain. There is, therefore, a sense in which our

3
To be clear, I do not wish to deny that sensations such as pain, pleasure, hunger, thirst, coldness
and hotness have a distinctive phenomenology. What I do deny is that our concepts are tags we use
by having learnt to identify the relevant sensations while having in view their distinctive phenom-
enology. Rather, these sensations have distinctive behavioural manifestations and we are taught to
replace, or at least to accompany those behavioural manifestations with the relevant self-ascrip-
tions. In contrast, I do think that the phenomenological account would be a non-starter in the case
of propositional attitudes, for there does not seem to be anything phenomenologically salient in the
mere propositional attitude (let it be of belief, or desire, or intention, and so on). If there is some-
thing phenomenologically relevant is the kind of feeligs which can accompany those propositional
attitudes. Furthermore, it does not seem to me that phenomenology would help us distinguish
among the various fine-grained propositional attitudes we may enjoy. For instance, I simply don’t
see what could distinguish a hope from a wish at a merely phenomenological level.
226 A. Coliva

self-ascriptions constitute our sensations. It is not a metaphysical one, for

they do not create or bring about the sensation. Rather, they individuate
the sensation for what it is and different self-ascriptions, which make use
of different phenomenal concepts, individuate different sensations for
what they are. Yet, to repeat, it is only through their different instinctive
manifestations that we can acquire those concepts in the first place, by
learning to use them to give expression to those sensations. Among those
manifestations, there are also some which depend on feeling the sensa-
tion in some part or other of one’s body, such as touching one’s injured
limb or aching belly. Likewise, those manifestations have a duration and
undergo variations in time. All those manifestations are part of instinc-
tive pain-behaviour and allow us to acquire the relevant bodily concepts
of pain—like, toothache, stomach ache, and so on—as well as the ability
to grade pain. The application of those more specialised concepts then
individuates pain in a more fine-grained way. Evidently, however, it also
makes room for certain possibilities of error or at least of unclarity or
vagueness, since bodily felt pains can have diffused locations. Hence, for
instance, it is not always clear whether one is having toothache or gum
problems. Furthermore, their intensity may vary across time and thus it is
may not be obvious whether one is feeling medium or weak pain or when
the change occurred. These possibilities of error or lack of transparency,
however, do not impugn the constitutive thesis when it is taken to hold
for purely phenomenal mental states and their self-ascriptions. Nor does
it impugn it if one wished to extend it also to self-ascriptions of bodily
felt pains and comparative judgements, as long as one included in the
relevant C-conditions the absence of vagueness. This, however, is a move
which is not necessary and I am open to both suggestions. That is to say,
that we should include absence of vagueness in the C-conditions and
extend this weak constitutivism also to self-ascriptions of bodily located
sensations; or else, that we should retain the constitutive account only
for the self-ascription of the sensation while allowing for errors, based on
misleading evidence, or on a subject's epistemic shortcomings, regarding
the bodily location of her sensation.
Finally, let us move on to the grounds of the constitutive thesis.
Suppose that a subject, endowed with the concept of pain and quite pro-
ficient in applying it to herself and others, did so, in her own case, by
always observing her behaviour and by inferring to its likely cause. She
 8 Pluralism About Self-knowledge 227

would not know that she is in pain because she feels it, but because she
would see herself scream and moan and would infer that she must be
in pain. That would cast doubt on the fact that a subject of capable of
having sensations and other phenomenally salient mental states at all.
Presumably, people affected by congenital insensitivity to pain with anhi-
drosis (CIPA) exemplify such a possibility.
Furthermore, as we already remarked in Chap. 3 (§1), a subject who
had to infer to her painful sensation would strike us as someone who,
in some sense, is alienated from herself; that is, she would bear to her
mental states the kind of relation she bears to those of others. However,
what makes a mental state of ours, ours, is, besides its origin in our brain
activity, the fact that it is known to us in a way which differs from how we
know of other people’s mental states and in which no one else can know of
it. The failure of groundlessness would therefore result in a lack of “sense
of ownership” over the mental states one would be inferring to have.
Furthermore, if she had to find out through inference to the best
explanation whether she was in pain, when asked “How are you?”, she
would be entitled to respond “I do not know. Let me find out” and she
then would be entitled to preface whatever answer she might wish to give
with “Probably I am in pain” or “It is my hypothesis that I am in pain”.
These responses would actually strike us as a symptoms of mental illness
unless there were reasons to think that the subject is indeed affected by
CIPA.  They would strike us as reports on another subject, who is the
subject of that sensation and who would happen to live in one’s body.
Hence, groundlessness seems to be a necessary and a priori element of
what counts as being a subject of sensations—indeed, the subject of those
sensations and not merely the object whom one may observe as having
them—and of what counts as being a rational subject.
In order to redeem transparency, one might say that sensations are nec-
essarily conscious—that is, such that having them coincides with being
aware of them. That is to say, their transparency, on this reading, would
be a necessary and a priori aspect of what goes by the name of “sensation”.
Similar considerations may perhaps be put forward in the case of at least
those emotions that have a distinctive phenomenology to them, as we
shall presently see. As we saw in Chap. 2 (§1), we may distinguish between
sensory states, which need not be conscious—being merely registrations
of variations in proximal stimuli—and sensations properly so regarded,
228 A. Coliva

which are necessarily phenomenologically salient to their subjects. The

latter are also ways of registering variations in proximal stimuli, which, on
top of that, are conscious. Yet one may want to insist that, for instance,
the soldier who has been marching for days does feel pain, even if he is
not aware of it because his fatigue and stress are so conspicuous that he
is oblivious to them and just keeps walking. It is at this stage that the
C-conditions we have introduced in the characterisation of the constitu-
tive thesis are helpful to delimit those cases in which it makes sense to
think that our knowledge of our ongoing sensations meets the transpar-
ency requirement. In particular, in this case one might say that because
of extreme fatigue the soldier is no longer cognitively lucid, attentive and
alert. Still, if a subject were always unable, while feeling pain and being
attentive and alert, to judge that she is, this would cast doubt on the fact
that she does possess the relevant concepts. Alternatively, if we had reasons
to think that she does have those concepts, her inability to self-ascribe an
ongoing pain would cast doubt on her cognitive well-functioning to the
point of impairing the idea that we are dealing with a rational subject.
Finally, let us turn to redeeming authority. If one were proven wrong
in one’s self-ascriptions of pain, doubt would be cast upon one’s posses-
sion of that concept. Alternatively, if one wished to maintain that a mis-
taken subject (with respect to her own mental states) could still be said to
have the concept of pain, perhaps because she retains the ability to apply
it to other people’s mental states, her mistake would impair her rational-
ity. For she would avow pain while she would behave in ways which run
contrary to it. Hence, we could no longer make sense of her linguistic
and non-linguistic behaviour.
Groundlessness, transparency and authority, therefore, are constitu-
tively tied to what it means to possess psychological concepts such as
“pain” as well as to what it means for someone to be a rational subject
(in a “thick” sense, as we saw). Moreover, they are constitutively tied to
what it means to be a subject capable of enjoying sensations and indeed
to what it means to be the subject of one’s own sensations.
Let us now extend these considerations to the case of basic emotions
such as fear and to perceptions and perceptual experiences. As noted, these
cases are relevantly similar to the one of pain and other sensations, but
they also introduce an element of complexity which has to do with the fact
 8 Pluralism About Self-knowledge 229

that these mental states have relational objects and correctness conditions
(at least in the case of perceptions). For I am scared of the dog which is
approaching or I see the cat in front of me or even that there is a cat in front
of me, and I don't simply have a certain feeling or a visual appearance.

2.2 Basic Emotions

It is important to clarify how the acquisition of the relevant concepts and

their canonical applications are supposed to work. Take fear. Clearly, it
has instinctive behavioural manifestations. My 2-year-old girl, since she
was 1 year old, as soon as she saw a dog, even a small one, used to scream
and jump on her parents’ lap or shoulders. We always reacted by saying
things such as “Oh dear, you’re so scared. Don’t, it is just a small dog”.
Little by little, she learned to replace, or at leasy accompany, her instinc-
tive behaviour with a more culturally ingrained one. She now wants to
be physically protected, but she says (the Italian equivalent of ) “scared”
and she no longer screams or does so only rarely, when she is taken aback.
Presumably, she will soon be able to judge and utter the sentence “I am
scared” and use it to express her ongoing fear. Anything we have been
reviewing in the case of pain and of “I am in pain” will therefore apply
in this case too. The presence of intentional objects, in the case of emo-
tions, however, introduces limited room for error in the identification of
the relational object of one’s fear. A subject might mistake a cat for a dog
approaching her and avow her fear of a dog (intentional object), when in
fact there is none, or, at any rate, when the animal which is approaching
is not a dog (relational object). Her self-ascriptions are therefore authori-
tative only with respect to the self-ascription of fear and its intentional
object as long as she has the relevant concepts. Yet they are not authorita-
tive with respect to the individuation of their relational objects. Similarly,
transparency holds only for one’s fear and its intentional object as long
as one has the relevant concepts, but it does not necessarily hold with
respect to one fear’s relational object. Thus, one may not be in a position
immediately to self-ascribe fear of a dog if one is uncertain about the kind
of animal that is approaching.
Notice that both in the case of pain and of fear, once one has acquired
those concepts blindly, by learning to use them in lieu of instinctive
230 A. Coliva

behavioural manifestations of the corresponding mental states, one will

be in a position also to apply them inferentially in one’s own case (as
well as in the case of others). By reflecting on the fact that in the past
one behaved in certain characteristic ways, for instance, one will be in
a position to infer that one was scared or in pain; indeed, one then can
“predict”, as it were, how one will behave if scared or in pain. Focusing on
the case of emotions, however, we have also seen, in Chaps. 2 (§3) and 3
(§3), that there are complex emotions, with quite intricate dispositional
elements. The case of love is paradigmatic in many ways. One crucial
aspect of it is that it becomes evident, in that case, how the raw feel, as it
were, is not at all the relevant element, which licenses the application of
that very concept. I would not wish to extend the constitutive thesis to the
case of complex emotions (see §3), yet in that case too it is evident how
the observational model of self-knowledge would be as defective as the
idea that the concept of love, say, is a label or a tag for a specific kind of
raw feeling one would be immediately aware of. Furthermore, it is evident
in that case too how the self-ascription individuates the emotion for what
it is—love as opposed to being possessive, as we saw in our discussion of
Emma’s case. Yet the case of complex emotions differs from the case of
basic ones insofar as one needs to know the characteristic notes of the con-
cept in order to apply it to oneself (and others). They inferentially guide
one’s application to one’s own case and can give rise even to sudden (re-)
conceptualisations which will have an enlightening effect with respect to
one’s self-knowledge and self-conception. What distinguishes the applica-
tion of that concept to oneself from the application to other subjects is
that the cues which will have to be taken into account will often involve
further aspects of one’s own metal life, many of which will be the object
of genuinely first-personal self-knowledge. As we saw, one may sometimes
be blind to oneself insofar as one may not notice these cues or see that
they call for being subsumed under a given concept one indeed possesses;
or one may be oblivious to some other cues, which would indicate that
one is not in love with someone but merely possessive or that one is sim-
ply feeling affection while perhaps in need of emotional (and sometimes
interpersonal) security and stability. These self-ascriptions therefore would
not be either transparent or authoritative and certainly they would not be
groundless either. Indeed, as we shall see in §3, one may think of these
case as cases of (more or less motivated) self-deception.
 8 Pluralism About Self-knowledge 231

Yet going back to basic emotions, the eventual self-ascriptions of fear (and
of other basic emotions) will be groundless, transparent and authoritative.
These features, in turn, are constitutively tied to what it means to possess the
concept of fear (and of other basic emotions) as well as to what it means for
someone to be a rational subject (in a “thick” sense, as we saw). Moreover,
they are constitutively tied to what it means for a subject to be capable of
enjoying basic emotions and to be the subject of those very emotions.

2.3 Perceptions and Perceptual Experiences

Let  us now  consider perceptions and self-ascriptions of perceptual expe-

riences. Once again, it is crucial to clarify how the relevant concepts are
acquired and canonically applied. Here, it is interesting to note how they
can be taught in order to replace other forms of behaviour, like the excla-
mation “dog!” when one sees one. For usually other people around one
respond by saying things such as “Yes, you saw a dog” or “No, it looks like/
seems a dog, but it is a cat”. The concept of seeing, therefore, is not intro-
duced as a label for an experience but as an alternative way to give expres-
sion to a perception with a certain content. Since, as we saw in Chap. 2
(§1), perceptions have correctness conditions, they may be wrong. Hence,
when they are wrong, we are taught to back-track and to talk about how
things look or seem to us. Like in the case of emotions, one can be wrong
about the relational object of one’s perceptions, which may be different
from their intentional one or even non-existing; but one can also be wrong
about certain aspects of one’s merely perceptual experiences, like when one
is asked to count the number of dots in one’s after-image. Neither of these
cases impugns the fact that we are authoritative with respect to our self-
ascriptions of perceptions or perceptual experiences, though. For, in the
latter case, it is clear that counting is involved and that we may go wrong in
doing it, especially with respect to a kind of mental representation which is
not as stable as a physical one. Such a case, therefore, makes apparent the
need of a specification of the C-conditions, which are supposed to enter the
constitutive thesis. In particular, the C-conditions will have to exclude cases
of counting and similarly inferential operations. Likewise, the C-conditions
will have to contain further provisos to the effect that one is not subject to
massive deception (like the ones operative in sceptical scenarios) and that
232 A. Coliva

the object perceived is in clear view. If all these conditions obtain, together
with the usual ones, namely conceptual mastery, lucidity, attentiveness and
cognitive well-functioning, then if one sees a dog, one will immediately be
in a position to judge “I see a dog” and one’s self-ascription will be guar-
anteed of being correct. Similarly, if one has a red after-image, say, one will
be immediately in a position to judge “I am having an experience of red”
(or “I am appeared to redly”) and one’s self-ascription will be authoritative4.
To redeem the constitutive thesis in connection with the kinds of men-
tal state we are considering here, it is important to stress the strong link
between perception and action. Hence, since we have allowed for the
possibility of unconscious perceptions (see Chaps. 2 (§1) and 3 (§1)),
it is crucial to note that the actions we would perform on the basis of
unconscious perceptions would fall out of responsible agency. As we saw,
a subject affected by blind-sight does not know either that she is catch-
ing a ball or why. Hence, she cannot be held responsible for her action.
Groundlessness, transparency and authority with respect to perceptions
are therefore necessary and a priori elements only of conscious perceptions,
which, in turn, are constitutive elements of responsible outer actions.
Hence, to conclude, groundlessness, authority and transparency will
be necessary and a priori features constitutively tied to what it means to
possess the concept of seeing (and of other perceptual states), of having
perceptual experiences and conscious perceptions, and indeed of what
it means to be the subject of those very mental states. Finally, they are
constitutively tied to what it means for someone to be a responsible agent
and a rational subject (in a “thick” sense).

3 Propositional Attitudes as Dispositions

and Complex Emotions: Third-personal
Self-knowledge
We have already seen in Chap. 3 (§3) how, when it comes to disposi-
tional mental states, especially propositional attitudes as dispositions, our
knowledge of them is third-personal. That is to say, it is neither ground-

4
This account seems to me largely in keeping with the gist of Evans’s account of our knowledge of
our seemings, put forward in Evans (1982, chapter 7).
 8 Pluralism About Self-knowledge 233

less nor transparent or authoritative. It is not groundless, because it is

based on various kinds of evidence, as we shall presently review, and on a
variety of methods, involving abductive and inductive inferences, infer-
ential conceptual deployment, testimony and simulation. It is not trans-
parent either, because having the relevant mental states is clearly not of a
piece with being aware of them or with being in a position immediately
to self-ascribe them once one has the relevant concepts and is cognitively
lucid, attentive and alert. It is not authoritative either, because it is in
the nature of the methods employed to reach this kind of third-personal
self-knowledge that they can go wrong in a number of ways we will list in
the following. All this, however, is no surprise, given how propositional
attitudes as dispositions are individuated (see Chap. 2 §2).

(a) They are not the result of a conscious deliberation—that is, a judge-
ment, on a subject’s part, based on considering and, in particular, on
assessing (or even being able to assess) evidence in favour of P (or of P
is worth pursuing, it would be good if P happened, etc.);
(b) they are not within one’s direct control, being rather something one
finds oneself saddled with;
(c) hence, they are not something one will be held rationally responsible
for.

As we saw, some examples of mental states that will satisfy these condi-
tions are unconscious mental states of a Freudian kind, which, however, can
be operative in shaping a subject’s behaviour. As noted, one can acquire
knowledge of them through therapy, which is clearly a third-personal
means to get knowledge of oneself—a means, moreover, of an interpreta-
tive nature. That is to say, patterns of observed behaviour are subsumed
under a theory which explains them as due to certain unconscious mental
states. Two main sorts of error are possible—namely, either focusing on
irrelevant patterns of behaviour or mistakenly identifying their cause. A
distinctive trait of third-personal self-knowledge of these mental states
is that it is not sufficient, just by itself, to bring about a change in one’s
attitudes even if it can prompt one to try to change them in the future.
As already remarked in Chap. 2 (§2), deep-seated biases would be
another case in point, like gender preferences in offering certain kinds
of job preferably to male (or female) candidates. Again, one may act on
234 A. Coliva

the basis of such a bias—for instance, that when it comes to philosophy

men are subtler than women or, in general, that male employees are more
suitable to undertake responsibilities at work since they will not ask for
maternity leave and will not have to run a household. One could eventu-
ally recognise one’s bias by observing one’s past behaviour and seeing it
as explainable in terms of the relevant biased belief. Here, interpretation
is still called for but it is less theory-laden than in the case of Freudian
mental states. It requires just the possession of the relevant concepts and
their application, which may go wrong because one either focuses on
irrelevant patterns of behaviour or erroneously takes them to be due to
a certain biased belief. Once again, if one did correctly recognise one’s
biased belief, the first-order mental state, if still in place, would remain
dispositional, at least in the normal run of cases and could even continue
to shape one’s behaviour, even if one sincerely judged the opposite.
Moreover, we sometimes make past psychological self-ascriptions of
propositional attitudes as dispositions through acts of self-interpretation.
So, by reflecting on my past actions, I could find out I was moved by
certain desires as dispositions, which were neither of a Freudian kind
nor based on biases. Or else, in the case of dispositional beliefs based on
habit, the sudden surprise I am aware of when I see a black sheep in the
field can make me abductively realise that I had dispositionally believed
that all sheep (in that field) were white. Here again, the forms of error
reviewed so far would be possible.
Moreover, an interesting point about all interpretative knowledge is
that the promptings one has in one’s own case comprise not just pat-
terns of overt behaviour but also knowledge of one’s past or ongoing
feelings and emotions. As we noted in Chaps. 3 (§3) and 4 (§2), this
is an element which sets inferential—that is to say, abductive—third-
personal self-knowledge apart knowledge of other people’s mental states,
which otherwise is equally based on that kind of inference. For only in
one’s own case can one be prompted to make the relevant inferences on
the basis of first-personal self-knowledge of one’s own ongoing or past
feelings and emotions. This means, however, that, to the extent to which
it is possible to make certain kinds of mistake in recognising these feel-
ings and emotions, the ensuing self-ascriptions of propositional attitudes
 8 Pluralism About Self-knowledge 235

as dispositions, based on abductive inference starting from these inner

promptings, may be wrong.
As we noted in Chap. 3 (§3), as the converse of self-interpretation,
there may be mental states, which one can predict will assail one, in given
circumstances, which, however, will not be within one’s direct control.
Perhaps, owing to one’s long-lasting self-observations, one will know that
if one were to work in an unsupportive environment for a while, one
would start losing one’s self-confidence and believing that one’s work is
meaningless or of poor quality. The characteristic feature of these mental
states—in this case, the belief that unless one’s work gains some kind of
external recognition it is not worthy—is that one would seem to find one-
self saddled with them even if one were rationally able to find reasons that
should make one think differently. In these cases, the self-interpretation
could either be based on induction, and could clearly be wrong for one
may after all react differently in the new case from how one reacted in
the past, or it could be based on simulation. By placing oneself in the
imagined situation and by simulating some of its key elements, one may
indeed experience a sense of insecurity and then, on its basis, arrive at the
predictive self-ascription “I will believe my work is of poor quality”. In
such a case, third-personal self-knowledge would be based on simulation
and on first-personal knowledge of the kind of reactions one undergoes
when one simulates being in a certain situation, but it also comprises
some interpretative elements. For it takes interpretation of an abductive
kind to pass from the experienced sense of insecurity provoked by the
simulation to the individuation of its ensuing result—that is to say, a
belief as a disposition that one’s work is going to be of poor quality. Such
a complex cognitive endeavour can go wrong in a number of ways. It may
not be the case that one will have certain feelings and emotions when
actually placed in the unsupportive working environment; it may also be
the case that the experienced insecurity provoked in the simulation will
not lead one to believe that one’s work is of poor quality, or indeed one
may make a mistake in identifying the kind of feeling undergone dur-
ing the simulation as one of insecurity. It might, after all, be a different
kind of feeling. Indeed, the simulation can be useful to put one in touch
with one’s insecurities and fears in order to help one overcome them—for
236 A. Coliva

instance, by noticing how there is no strict correlation between working

in an unsupportive environment and producing work of poor quality.
Another intermediate case, noted in Chap. 3 (§3), which does not
immediately coincide with knowledge of our own mental states either
through explicit inference to the best explanation or through mental
simulation, is the one in which we suddenly notice an aspect of ourselves
we had been previously blind or oblivious to or indeed re-conceptualise
an aspect of ourselves by suddenly seeing it differently. We may all of a
sudden realise that we are not really nice to other people—we do not
actually care about them—but are only being polite, or we may sud-
denly see a character trait of ours—such as noticeable steadiness—not as
a symptom of strength but of weakness. As already remarked, seeing-as,
switches of aspect and noticing aspects are complex phenomena even
when they clearly concern perceivable objects and properties and involve
the operation of perceptual faculties. In the case of our own mental states,
these notions find only indirect employment, for no real perceptual fac-
ulty is involved and yet it does not seem that we engage in any complex
inferential thinking, of an abductive or inductive kind, nor that we do
engage in mental simulation, when (re-)conceptualisations of our char-
acter traits occur. Furthermore, the relevant psychological concepts are
necessary, while, arguably, purely perceptual switches of aspects do not
necessarily require concepts. We may say that we suddenly take a pat-
tern of overt behaviour, and possibly other inner aspects of our psychol-
ogy, as instantiating this or that psychological property either because
that pattern is manifest to us for the first time, and having the relevant
concept we subsume the former under the latter, or because the relevant
concept is available to us at last and a pattern of behaviour and inner ele-
ments we had possibly noticed before seems immediately to fit the newly
acquired concept. The newly acquired concept thus allows us to engage
in a kind of hermeneutical activity about ourselves. It allows us to notice
some aspects of our character and mentality we were oblivious to, prior to
acquiring that new concept.  Sometimes, these new conceptualisations or
re-conceptualisations may contradict previous ones, thus giving rise to a
switch of aspect regarding ourselves and our personalities and sometimes
we can alternate between these aspects thus seeing ourselves one way or
the other. Notice, moreover, that sometimes the patterns of behaviour
 8 Pluralism About Self-knowledge 237

and the internal cues  we attend to may remain identical but be ranked
or arranged differently, so as to result in a classic case of aspect-switching.
For, as is made familiar by Wittgenstein's discussion of the duck-rabbit
case, in typical aspect-switching the lines remain identical and yet we
see them differently. Still, in these cases, inference is involved at least in
the following sense. While there is no inference to the best explanation,
the characteristic notes of the newly acquired (or applied) concept guide
the process of subsuming the observed pattern of behaviour and inner
elements under the concept. That may happen very rapidly, and so, phe-
nomenologically, we may not even seem to be engaging in inferences, yet
epistemically that is what grounds our judgement “I am F” (where F is a
psychological concept). It is one possible case of self-interpretation and
it can go wrong because, after all, the pattern observed may not fit the
concept or we may have ignored other features which would, if taken into
account, lead to a different psychological self-ascription.
In other cases, our self-knowledge is based on testimonial evidence. If
a trustworthy person (or at least someone we have no reason to think is
conceptually incompetent or willing to deceive us) tells us we are arro-
gant, say, we may form that belief, based on her telling us so. Our third-
personal self-knowledge would thus crucially have a social dimension to
it, and it would not be confined to the highly peculiar case in which
we gain self-knowledge by consulting a professional therapist. Normal
chats with family and friends can actually be an important source of self-
knowledge. Indeed, even gossip, to the extent that one is willing to con-
sider it a way of gaining testimonial knowledge, can make us aware of
dispositional aspects of our character we might otherwise ignore. In par-
ticular, it can make us aware of the kind of reactions we tend to elicit in
other people and would thus make us aware of those dispositional aspects
of our character. Finally, it should be noted that testimony and gossip
really allow us to gain a third-personal perspective onto ourselves, in a
way in which the other forms of third-personal self-knowledge reviewed
so far will always depend on a first-personal application of third-personal
methods.
As already noted (see Chap. 3, §3), sometimes we figure out our com-
plex emotions, such as love towards someone, by reflecting on our behav-
iour and inner phenomenology in their presence or at the prospect of
238 A. Coliva

meeting or losing them. This is the case illustrated by the passage from Jane
Austen’s Emma quoted in the second chapter. Obviously, we can go wrong
in our self-interpretation, based on abductive reasoning. Our behaviour
and feelings may not be symptoms of love but of being possessive and of
dreading the prospect that a good friend will no longer be so close to us as
the result of getting married to another person. Alternatively, we can mis-
take our behaviour and feelings as symptoms of love, when they are in fact
signs of our vanity and complacent attitude towards those people who are
usually kind to us. Sometimes, in contrast, we are self-blind in the sense
that despite showing many of the characteristic symptoms of love, say,
towards someone, we do not realise that we are in love with them, even
if we do possess the concept of love and are generally capable of correctly
applying to it to other people. These forms of self-blindness can have dif-
ferent causes, which may be due to psychological or to cultural factors or
even to deeper unconscious causes.
There are also less dramatic cases, where we simply conceptualise our
behaviour and inner promptings in a mistaken way because we more or
less consciously ignore or downplay some other aspects of our behav-
iour and feelings. Hence, we arrive at a mistaken individuation of some
of our character traits. For instance, there is a difference between being
driven at work and being obsessed or even ruthless. Application of these
concepts to ourselves is not (necessarily) a case of abductive reasoning.
Quite often, it is just inferential conceptual application, where the char-
acteristic notes of the concept guide our psychological self-ascription
but where we can go astray by not paying (enough) attention to some
aspects of our behaviour and feelings which would actually call for a dif-
ferent categorisation. Similarly, we may not believe what a trustworthy
person tells us about our character or our attitude in a given situation. Of
course, there may be reasons why we do go so astray, or indeed why we
may not trust what we are told by another person. Clearly, it is not nice
to realise that one is obsessed or even ruthless and not merely driven.
Similarly, it is not nice to realise that one is an arrogant person.  Such
a realisation would presumably call for the endorsement of a different
conception of oneself and for a re-orientation of one’s behaviour. Both
outcomes would presumably pose problems of a psychological and even
of a moral nature. It is sometimes this kind of phenomenon which is
 8 Pluralism About Self-knowledge 239

used to identify cases of self-deception. Without engaging in a linguis-

tic dispute, let us therefore grant at least for the sake of argument that
these are cases of self-deception, albeit different from the ones reviewed
in Chap. 7 (§§3–4) and further considered in the Appendix, where there
are conflicting dispositions and commitments at work in one’s cognitive
life. Still, they are not counterexamples to first-person authority, because
they would arise in cases of inferential conceptual application, or of tes-
timonial self-knowledge, as we might call them, which are, by all means,
cases of third-personal self-knowledge. This should come as no surprise
since here we are dealing with complex psychological concepts, involving
a lot of dispositional elements, and, as we have just seen, we can go wrong
in their application for a variety of reasons, some of which would have to
do with possibly deep psychological motivations.
We have also seen, in Chap. 3 (§2), how alleged counterexamples to
first-personal self-knowledge that draw on recent findings in cognitive
studies can in fact be interpreted as examples of third-personal knowledge
of our own mental states. In particular, we think we have certain charac-
ter traits or that we will have certain emotional responses in the light of
alleged life-changing events, which we actually lack or will not have. The
interesting aspect of this failure at self-knowledge is that the reason why
we think we would have these mental dispositions is that we engage in
simulating how we would behave in a given situation. We may imagine
what we would do if we had the opportunity to give to charity or if we
won a large sum of money. For whatever reason, which may or may not
be biased or responsive to some deep psychological motivation, we think
we would do things which would show either that we are generous or that
we think would make us feel much happier than we do now. Still, when
the opportunity to give to charity arises or we actually win the lottery, we
either do not behave as we thought we would or quickly revert to feeling
as usual. The kind of procedure followed in these cases, in order to try to
gain knowledge of our mental dispositions and even future emotions, is
indeed similar to the one we would apply if we were to figure out some-
one else’s mental states by means of simulation and its correct outcome is
obviously not secure. There is in fact no guarantee that we will behave as
we imagined we would do or that our feelings and attitudes towards life
will permanently change in light of different living conditions.
240 A. Coliva

Third-personal self-knowledge, though different from knowl-

edge of other minds in that it can take inner feelings and emotions as
promptings, exhibits all the typical aspects of any substantive cognitive
achievement, which usually results in knowledge. We base our psycho-
logical self-ascriptions on various kinds of evidence (even of a testimonial
nature), we engage in inferences and we deploy imagination to simulate a
situation and predict our future mental states. Moreover, we inferentially
deploy concepts to make sense of ourselves and our behaviour. All these
procedures admit of the possibility of error and of not being deployed at
all, thus resulting in a form of self-blindness. Furthermore, the kind of
abilities called upon can be variously distributed across the population
and can be finessed with experience, study and exercise.

4 Summary
In this chapter, we have seen how the metaphysically robust constitutive
account, which we uphold for self-knowledge of propositional attitudes
as commitments, cannot be extended to our knowledge of other kinds of
mental states. In particular, in §1, we have seen that even when it comes
to past self-ascriptions of commitments, it will have to be integrated with
an evidential account. By that, we mean simply that the relevant self-
ascriptions are based on mnestic evidence and therefore are not groundless,
even though they are not based on the observation of one's past inten-
tion or on inference to the best explanation starting from the observation
of one's own behaviour. That opens up the possibility of forms of self-
blindness and of mistaken self-ascriptions. Yet there is a general presump-
tion that subjects do remember their past (mental) actions correctly and
that their self-ascriptions—arrived at by taking into account the relevant
mnestic evidence—are correct. It is that presumption that in turn is con-
stitutive of what it means to be a rational, well-functioning human being,
who can be held responsible for her (mental) actions through time. Yet
this does not mean that one’s knowledge of those past commitments can
be accounted for by means of a constitutive account of self-knowledge.
We have then moved on to our self-knowledge of sensations, basic
emotions, perceptions and perceptual experiences (§2). Here again, a
 8 Pluralism About Self-knowledge 241

metaphysically robust constitutive account cannot be applied, because

there is no room for the idea that we might bring about those phenome-
nologically mental states just by self-ascribing them. Yet there is room for
a weaker kind of constitutive account according to which it is necessarily
and a priori true that, given certain C-conditions, one is in one of these
mental states iff one self-ascribes them. We have carefully specified those
conditions, which will have to include concepts’ possession, attentiveness,
lucidity and rationality. We have then offered an expressivist account of
how the relevant concepts are acquired and (canonically) deployed, in
order to avoid either falling back into a Cartesian conception, whereby
the relevant concepts would be labels for phenomenologically salient and
sufficiently distinct feelings and sensations one would have to have in
view, or falling back into an inferentialist conception, whereby we would
have to infer to our mental states by considering their behavioural mani-
festations. The key idea here is that these mental states are phenomeno-
logically salient and have characteristic instinctive manifestations. The
relevant concepts then are acquired by being taught to substitute (or at
the very least accompany) those instinctive manifestations by means of
the relevant self-ascriptions.
Groundlessness, transparency and authority, then, have been redeemed
by showing how they are constitutively tied to what it means to be a sub-
ject capable of enjoying the relevant mental states and of enjoying them
as one’s own as well as being tied to what it means to possess the relevant
concepts, being rational and cognitively well-functioning and, in the case
of outer perceptions, being responsible agents for one’s actions.
Finally, in §3, we have turned to various cases of third-personal self-
knowledge having to do with one’s own dispositional mental states, let
them be propositional attitudes or complex emotions. We have seen
how in their connection we can apply a variety of methods, like infer-
ence to the best explanation involving psychoanalytic theories as well as
less theory-laden abductive explanations, testimony and simulation. An
aspect that sets third-personal self-knowledge apart from third-personal
knowledge simpliciter is the fact that the cues, which get subsumed under
a theory or which are elicited by the simulation, are often of a psychologi-
cal nature and known in a first-personal way. Finally, we have highlighted
the role of new conceptualisations or re-conceptualisations affected by
242 A. Coliva

being led by the characteristic notes of highly dispositional psychological

concepts in making us realise (or sometimes contrive) some of our deep-
seated character traits.
Hence, the label “self-knowledge” corresponds both to a variety of
mental states—which go from exercising one’s epistemic abilities in
order to arrive at knowledgeable psychological self-ascriptions, to simply
making psychological avowals which are a manifestation of one’s being
conceptually equipped, rational, intelligent, cognitively well-function-
ing and capable of enjoying a variety of mental states as one’s own, or
of one’s being a responsible agent—and to a variety of methods, when
third-personal self-knowledge is at stake. For, in that case, the kinds of
epistemic procedures we engage in, to arrive at our self-ascriptions, are,
as we have seen, many and diverse. It is indeed the gist of this book that
such a variety can be appropriately accounted for only within a pluralistic
framework, whose contours have herewith been delineated.
 Appendix: Moore’s Paradox

Moore’s paradox comes in two forms:

1. I believe that P, but it is not the case that P1; and

2. I do not believe that P, but it is the case that P.

The paradox arises from the fact that the conjuncts in (1) and (2) have
independent truth conditions and therefore can be both true (or false) at
once. If so, then, (1) and (2)—that is, these conjunctive propositions—
could be true and yet judging or asserting them would be so deeply absurd
as to result in self-defeat. We sense that a subject cannot really believe
that it is raining, say, while also judging that it is not, or that he cannot
disbelieve that it is raining while also judging that it is. Yet the nature
of Moore’s paradox remains elusive in many ways and several different
explanations of it have been proposed. Despite its elusiveness, study

1
Sometimes (1) is phrased differently—namely, (1*) “P and I believe that not-P”. Notice that (1*)
can be obtained from (1) through uniform substitution of P with not-P and commutation, pro-
vided that double negation elimination held. (1) (or 1*) is called the commissive form of the para-
dox and (2) the omissive one. See Green and Williams 2007, p. 5.

© The Editor(s) (if applicable) and The Author(s) 2016 243

A. Coliva, The Varieties of Self-Knowledge,
DOI 10.1057/978-1-137-32613-3
244 Appendix: Moore’s Paradox

of this paradox is considered to have extremely wide implications2 and

Richard Moran insists that Moore’s paradox is “an emblem for peculiari-
ties in the first person point of view, specifically how the possibilities for
thinking and talking about oneself are systematically different from the
possibilities of thinking and talking about other people”.3
In a recent paper,4 Tom Baldwin has argued that, in order to pro-
vide a satisfactory account of Moore’s paradox, we should bring in a
normative notion of belief—the notion of belief as commitment. For,
if we stick to purely functionalist accounts of belief, nothing would
prevent a subject from sensibly judging (or asserting) either (1) or
(2). I concur with Baldwin’s judgement and further motivations in its
favour will be offered in the sequel. Yet it will be argued that it is not
enough merely to introduce the notion of a belief as a commitment in
order to rescue Moore’s paradox. Rather, to avoid its disappearance—
that is, its being perfectly legitimate to judge (or assert) either (1) or
(2) at least on occasion—we should stick to a rather resolute notion
of commitment. The final sections of the paper will spell out such a
notion and will aim to establish this conclusion or, at least, to show
how, unless we are prepared to embrace it, we would be in danger of
actually losing Moore’s paradox. Given the high stakes at issue, such
a result would seem to entail rather unpleasant philosophical conse-
quences. Furthermore, it would become a total mystery why we intui-
tively find either (1) or (2) so defective. Hence, the loss of Moore’s
paradox would cause a severe blow both to our philosophical and to
our self-understanding.

2
Green and Williams 2007, pp. 3–4.
3
Moran 1997, p. 143.
4
Baldwin 2007, pp. 76–89. An analogous suggestion can be found in Millar 2004, p. 125. It will
be assumed that broadly functionalist accounts of beliefs and normative ones are the main options
on the table. Eliminativism, besides being problematical in its own right, would be a non-starter in
connection with Moore’s paradox, for it would actually condone “It is raining but I do not believe
it”, since, by its lights, there would be no beliefs. See Turri 2010, pp. 35–39. As will become appar-
ent, in some cases (1) and (2) can be coherently judged/asserted. This, however, does not mean that
there are never cases in which those judgements/assertions would result paradoxical. The task will
be to spell out the conditions in which judging or asserting either (1) or (2) would result in
self-defeat.
 Appendix: Moore’s Paradox 245

Moorean and Wittgensteinian Analyses

Two main kinds of analysis of the paradox have been proposed since
George Edward Moore first discovered it. They can be traced back to
Moore himself and to Wittgenstein. We therefore shall call them the
Moorean and the Wittgensteinian analysis, respectively. The details of
the developments they have been subject to over the years will not be
discussed. For, although these developments are interesting and insight-
ful, it is not clear that they will be able to eschew the problems that afflict
their respective predecessors.5 Nevertheless, the aim of this section is only
to highlight the main difficulties these analyses face and thereby provide
at least prima facie motivation to look for an alternative.
We will start by giving an outline of Moorean and Wittgensteinian
analyses and by exposing their shortcomings. It will be argued that the
major difficulty with the former is that it does not explain the oddity of
holding either (1) or (2) in thought.6 For it is committed to the view that
the paradox arises only because the assertion of either (1) or (2) violates
some pragmatic norm that governs that linguistic practice. By contrast,
the main problem with the latter is that it involves a loss of subject matter.
For, by reducing Moore’s paradox to an outright contradiction, it dis-
solves or explains away what needed explaining in the first place: how it
is possible for two independent propositions, once they are conjoined, to
give rise to self-defeating assertions or judgements.7
The proposal is to take these points quite seriously and hence to steer
away from any analysis of Moore’s paradox which either claims that, by

5
Moore 1942, pp. 540–543. See also Moore 1944, p. 204. Wittgenstein PI, II, xi pp. 190–192 and
1980, §§90–96. Among Moorean analyses, see Baldwin 1990 and Shoemaker 1995. Among
Wittgensteinian ones, see Collins 1996; Heal 1994; Lee 2001; Linville and Ring, 1991 and
Malcolm 1995.
6
Heal 1994, p. 6, Shoemaker 1995, p. 213, Moran 1997, p. 144 and Baldwin 2007, pp. 78–9
emphasise this point. Green and Williams 2007, p. 6 date it back to Sorensen 1988. Chan 2010
tries to make sense of the idea that there might be pragmatic paradoxes at the level of thought. Yet,
he concludes that, in the case of Moore’s paradox, this would nevertheless lead to an unsatisfactory
account.
7
Linville and Ring 1991 are committed to this claim. Heal 1994, p. 6, places as a requirement upon
any satisfactory analysis of Moore’s paradox that “the solution must identify a contradiction, or
something contradiction-like, in the Moorean claims”. Both requirements will be criticised at
length, for different reasons, in the text.
246 Appendix: Moore’s Paradox

application of whatever principle, the very content of (1) and (2) comes
down to a contradiction of the form “p and not-p” or appeals to some
pragmatic norm governing the speech act of assertion.
At the core of all Moorean analyses lie two main claims:

(a) Neither (1) nor (2) is a contradiction. For they are conjunctions of
propositions with different truth conditions. Hence, they can both
be true at the same time.
(b) Given (a), there is nothing wrong with (1) and (2) as such. Thus, there
is nothing wrong with believing or entertaining their respective con-
tents in thought. For one can perfectly well think that one has a false
belief—(1)—and that is ignorant with respect to P—(2). Rather, the
oddity of Moorean contents—what, in effect, entitles us to talk about
a paradox in their connection—is the fact that when one asserts either
(1) or (2), then one is doing something pragmatically inconsistent.

Assume for the sake of argument that the assertion of “P”—let it be

sincere or not—presents the speaker as believing that P. By then adding
in one breath, as it were, that one does not believe that P, one is present-
ing oneself, at one and the same time, as both believing and not believ-
ing that P. This, however, seems to be pragmatically inconsistent. For it
appears that one cannot consistently perform a speech act whose (non-
deductive) implication is that one believes what one is saying while at the
same time cancelling that implication by saying “But I do not believe it”.
However, this account is no longer available in the case of (1).8 For by
asserting “Not P” one presents oneself as believing that not-P is the case.
Yet by asserting “I believe that P” one presents oneself as believing the
opposite. Hence, by asserting (1) one presents oneself as believing con-
tradictory propositions. Yet where exactly is the pragmatic inconsistency
in presenting oneself as believing contradictory propositions? After all,
it seems possible even to form the intention of informing one’s audience
that in a process of self-scrutiny, for instance, one has found out that one
has contradictory beliefs (as we shall see in the following). Therefore, it is
not clear what pragmatic norm one would be violating by conveying the
thought that one has contradictory beliefs, if any at all.

8
Heal 1994, pp. 11–2.
 Appendix: Moore’s Paradox 247

Whether or not a Moorean theorist may be able to make a convincing

case for the view that to present oneself as believing contradictory propo-
sitions is a kind of pragmatic inconsistency, the main problem with the
Moorean analysis is as follows. If the paradox can be made sense of only
by appealing to pragmatic norms governing communication and speech
acts, this by itself shows that there should be no analogue of Moore’s
paradox at the level of thought. Still, judging (1) or (2) would indeed be
an odd thing to do. Yet, on the Moorean analysis, we are deprived of any
means to make sense of this very possibility. Hence, whether or not the
Moorean analysis is successful as an analysis of Moore-paradoxical asser-
tions, it does not have the resources to explain Moore-paradoxical judge-
ments. Any Moorean analysis, therefore, is at least wanting for a lack of
generality.
Wittgensteinian analyses of Moore’s paradox are united in holding the
following9:

(a’) Moore’s paradox is not really a paradox, because, in effect, it is an

outright contradiction. For asserting “I believe that P” is tantamount
to asserting “P” in a tentative voice, so to speak; and asserting “I do
not believe that P” is tantamount to asserting “not-P”. Hence, both
(1) and (2) come down to an assertion of “P and not-P”.
(b’) For this very reason, it is all right neither to assert nor to judge (1) or (2).

(b’) is the only feature which, at least prima facie, seems to speak in favour
of Wittgensteinian analyses. For (a’) has the well-known and unpalatable
consequence of depriving one of the means of explaining the following
platitudes.

Embedding: (1) and (2) can be embedded in wider contexts, like “Suppose
that I believe that P, but it is not the case that P” and “Suppose that it is the
case that P, but I do not believe it”. The possibility of embedding (1) and
(2) in suppositions, in turn, explains why they can occur as antecedents in
correct conditional statements.10

9
Moran 1997 denies that this is actually Wittgenstein’s position. This, however, is the vulgata and
constitutes the bulk of what I call “Wittgensteinian” analyses of Moore’s paradox.
10
Wittgenstein himself was aware of this. See in the following for a (failed) attempt at reconciling
this platitude with a Wittgensteinian account of Moore’s paradox.
248 Appendix: Moore’s Paradox

Past: (1) and (2) can be turned into the past. “I believed that P but it was
not the case that P” or “I did not believe that P but it was the case that P”
is perfectly fine.
Third-person: (1) and (2) can be put into the third person. “She believes
that P, but it is not the case that P” and “It is the case that P, but she does
not believe it” can both be used to make accurate statements.

However, no formal contradiction—that is, no contradiction of the form

“p and not-p”—can be the content of a coherent supposition,11 nor does
it turn into a non-contradiction once put into the third person or when
changed of tense. Yet, on the Wittgensteinian analysis, we are asked to
take (1) and (2) as formal contradictions while, at the same time, granting
the platitudes. How can these two claims be reconciled?
The Wittgensteinian strategy consists in claiming that the verb “to
believe” has a different meaning depending on whether it is used in the first
or in the third person and on whether it is used either in the present or in
the past. In particular, we should concede that no use of the verb “believe”
in the first-person, present, is such as to describe a subject’s belief.12
Yet this seems odd, for there is such a use. Consider “I do not know
who did it, but I believe it was the butler” said by a subject who has no
warrant for charging the butler with the mischief but is totally subjec-
tively convinced that he is the culprit. Now, if the speech act of making
an assertion is governed by the epistemic norm that one should have a
warrant for the content of one’s assertion (or even know it), then a subject
who asserts that sentence would precisely not be in a position to assert
“The butler did it”. Hence, a subject’s assertion of “I believe the butler
did it” cannot be taken as equivalent to the assertion of its embedded
content. Thus, it must be taken as a description of the subject’s doxastic
state—that is, of her conviction that the butler is the culprit, which can
be in place no matter whether she has a warrant for “The butler did it”

11
One might object that one can suppose anything. For instance, opponents of dialetheism suppose
that P and not-P in order to show that any consequence follows. However, exactly this case shows
that the supposition was not coherent in the first place. At any rate, formal contradictions at least
do not cease to be such when changed of tense or person.
12
By contrast, the meaning of “to believe” both in the first-person past and in the third person is
descriptive of a mental state, which one is either ascribing on the basis of behavioural criteria
(sometimes even in one’s own case) or on the basis of memory.
 Appendix: Moore’s Paradox 249

and, hence, independently of whether she would ever be in a position

warrantedly to assert it.13
When we turn to the use of the phrase “I do not believe that P”, the
case for not taking it to be equivalent to the assertion of “not-P” is even
more obvious.14 For it is not always the case that someone who is in a
position sincerely to assert “I do not believe that P” is also and automati-
cally in a position to assert “not-P”. If a subject has suspended judgement
on whether the butler is the culprit, because there is equal evidence for
and against him, her assertion of “I do not believe the butler did it” is
consistent with her epistemic situation in a way in which “The butler
did not do it” is not. Said otherwise, agnosticism with respect to P is
a coherent position. Hence, it is coherent for someone to assert “I do
not believe that P and I do not believe that not-P”. Yet, if we applied
the Wittgensteinian analysis to that assertion, it would turn it into an
assertion of “P and not-P” (provided double negation elimination held).
Agnosticism—allegedly a respectable position—thus would amount to
an outright contradiction.
Hence, while granting that sometimes the locution “I believe” is used
to weaken one’s assertion of “P” and that the locution “I do not believe”
is used to express the negation of the embedded content, some other
times those locutions are used to describe one’s belief or one’s absence
of belief. Therefore, even if Wittgensteinian analyses could apply to one
kind of use of “I (do not) believe”, they could not apply to the other.
Thus, Moore’s paradox could still arise for the latter use of “I (do not)
believe” and this is what needs explaining.15

13
de Almeida 2001, p. 38.
14
Heal 1994, p. 7, who favours a Wittgensteinian analysis of Moore’s paradox, acknowledges this
point, but, surprisingly, goes on to offer a uniform account of it, where both (1) and (2) come down
to contradictions of the form “P and not-P”. However, such a result can be obtained only by taking
“I do not believe that P” as equivalent to “I believe that not-P” and thus by taking its assertion as
equivalent to the assertion of “not-P”.
15
This point (see de Almeida 2001, p. 39) is not sufficiently appreciated by Heal (1994, pp. 20–24),
who puts forward the view that self-ascriptions of belief have a double role: that of describing one’s
attitude and that of expressing the belief in p. Only the latter aspect of the concept of belief would
lend itself to a Wittgensteinian treatment of Moore’s paradox. Yet, even if Wittgensteinian analyses
worked for certain instances of Moore-paradoxical sentences, they would not work for others (or,
in Heal’s account, even if they worked for the latter reading of Moore-paradoxical sentences, they
would not work for the former). Thus, Wittgensteinian analyses do not explain away Moore’s para-
dox and, when faced with genuine instances of it, are left with no means to address it.
250 Appendix: Moore’s Paradox

Finally, Wittgensteinian analyses, like Moorean ones, start by consider-

ing the assertion of (1) and (2). In particular, it is only because sometimes
the words “I (do not) believe” are used to weaken one’s assertion of “P”,
or to assert the negation of the embedded content, that Wittgensteinians
have the resources to claim that (1) and (2) come down to outright con-
tradictions. Moreover, it is only via considerations regarding the use of
those words that Wittgensteinians can claim that the content of Moore-
paradoxical sentences can be neither coherently asserted nor believed or
judged. However, for Wittgensteinians too, the problem arises of explain-
ing Moore-paradoxical judgements—that is, judgements that are not ver-
bally expressed (although they may be so expressed) whose content is given
by (1) and (2). In such a case, the characteristic appeal to the use of the
phrase “I (do not) believe”—at least in certain contexts—is blocked and
it is not easy to see what else they could appeal to in order to produce an
outright contradiction of the form “p and not-p” at the level of thought.16
Notice, in fact, that the so-called Evans’s point—according to which in
order to answer the question whether one believes that p one should put into
operation whatever procedure one has for answering the question whether
p—cannot be used in this connection.17 For, although it may well be true

16
This would be obvious if one held the principle that “what can be (coherently) believed con-
straints what can be (coherently) asserted” but not vice versa (Shoemaker 1995, p. 227, 1n holds the
quoted principle but does not explicitly commit himself to the negation of its converse. In contrast,
de Almeida 2001, p. 33 holds the whole principle). Wholehearted Wittgensteinians would proba-
bly reject this principle, but more recent proponents of the Wittgensteinian analysis of Moore’s
paradox, such as Jane Heal, do not explicitly reject it. Therefore, if one were to defend the
Wittgensteinian analysis of Moore’s paradox one had better either motivate the rejection of the
principle, or find a way of extending the Wittgensteinian analysis to the case of Moore-paradoxical
judgements. As we shall see, this is no trivial matter.
17
Evans 1982, pp. 225–226. As we saw in Chap. 5, §3, the idea is that self-ascriptions of beliefs are
not based on looking inwards and on finding out that one has a given mental state. Rather, they
draw on the same epistemic resources that would license the mere judgement (or assertion) that
P. While there is much to applaud in the attempt to avoid the Cartesian epistemology of the mind,
Evans’s point does not help the cause of Wittgensteinians, nor is very convincing as an account of
self-knowledge. For a sustained application of Evans’s point to self-knowledge, see Gallois 1996 and
Fernández 2013. For a critique, see Gertler 2011a, pp. 174–177, Coliva 2014 and Chap. 5, §3.
In response to the suggestion that one may use Evans’s point to explain Moore’s paradox indepen-
dently of developing it into a viable account of self-knowledge, it should be noted that it would be
an extremely implausible move to make. For what is under attack is exactly the distinctive claim
that would be used to explain the paradox—namely, that the very same grounds would support
both “P” and “I believe that P”.
 Appendix: Moore’s Paradox 251

that whenever I am in a position (warrantedly) to judge that p is the case, I

would also believe that p and would automatically be in a position to judge
that I believe that p—provided I have the relevant concepts—this would
not produce the intended result. For Evans’s point is a belief-introduction
rule—that is, a rule which licenses inferences from [P]18 to [I believe that
P]—while what is needed, in order to generate a contradiction of the form
[P and not-P], is a belief-elimination rule—that is, a rule which allows infer-
ences from [I believe that P] to [P]19—and thus allows one to eliminate the
belief-operator in the doxastic conjuncts occurring in (1) and (2).
It may be objected that, despite there being no contradiction of the
form [P and not-P], one cannot have a justification for [I believe that P]
as well as for [not-P]. For, given Evans’s point, one would have to have a
justification for [P] and also for [not-P]. As we shall see at length in the
following, however, one can indeed have justification for [I believe that
P] and also for [not-P]. This entails that Evans’s account of our knowl-
edge of our own beliefs as based on the very same grounds we may have
for their embedded contents is highly problematical and cannot provide,
as such, a satisfactory account of Moore’s paradox.
One then may try to apply Evans’s point to generate a contradiction
of the form [I believe that P and I do not believe that P]. This strategy
may be successful with respect to (2). For, if a subject with the relevant
conceptual repertoire is able (warrantedly) to judge that p, then she is
ipso facto able to form the belief that she believes that p. Hence, in (2), by
application of Evans’s point, one may derive a contradiction of the form
[I believe that P and I do not believe that P]. Yet the same strategy can-
not succeed with respect to (1). For, by application of Evans’s point, one
derives only [I believe that P and I believe that not-P]. This, however, is
not a contradiction or—as such—an absurd judgement.20

18
Square brackets will be used for propositional contents.
19
This is a difficult rule to motivate, since it would imply omniscience.
20
Heal 1994, p. 11. Williams 2007 provides a sustained defense of the application of Evans’s point
(see also his 2004). However, he makes use of two principles—that he calls “Evans’s Principle” and
“Analogue of Evans’s Principle” (at p. 95 and 101, respectively)—that are not clearly Evans’s and
relies on the problematical idea that Evans’s point can be developed into an adequate theory of self-
knowledge. For a discussion, see Vahid 2005; Brückner 2006, 2009. For a response, see Williams
2009.
252 Appendix: Moore’s Paradox

Let me stress one final point against Wittgensteinian analyses. If

Wittgensteinians are right about the use of the phrase “I (do not) believe”,
what one entertains in thought when one utters a Moore-paradoxical
sentence is the contradictory content [P and not-P]. Hence, either
there are no Moorean-paradoxical thoughts—just contradictions—or
Wittgensteinians too are forced to allow that there is a genuine occurrence,
in thought, of [I (do not) believe (that P)], which cannot be reduced to
an occurrence of [P] (or of [not-P]). Thus, despite initial appearances to
the contrary, Wittgensteinian analyses too might be wanting for a lack of
generality as well as either for failing to explain the platitudes or for doing
so only by denying that the verb “to believe” in the first-person present
can be used to describe a subject’s belief.
To conclude: Moorean and Wittgensteinian analyses face at least prima
facie serious difficulties. Although, no doubt, their supporters could try
to finesse them, the preceding criticisms offer enough motivation to
attempt to find an alternative explanation.

The Constraints on Any Feasible Account

of Moore’s Paradox
The previous analysis of both Moorean and Wittgensteinian accounts
helps us see what the constraints on any feasible explanation of Moore’s
paradox are. On the one hand, we do not want to lose our subject mat-
ter. That is to say, we do not want to lose sight of the idea that (1) and (2)
can be taken to be paradoxical. Hence, we cannot put forward an analysis
according to which the content of (1) and (2) is of the form [P and not-P],
for there is nothing paradoxical in formal contradictions. Moreover, avoid-
ing turning the paradox into a contradiction will allow us to respect the
platitudes while holding that the verb “to believe” can be used uniformly
across changes of person and tense to describe a subject’s mental state.
On the other hand, we do not want just an explanation of Moore-
paradoxical assertions but also of entertaining Moore’s paradox in
thought, either from which an explanation of the former can be derived
or to which it can be added, depending on one’s own theoretical prefer-
ences. Hence, the explanation of the paradox cannot proceed by appeal-
ing (just) to norms governing the pragmatics of communication.
 Appendix: Moore’s Paradox 253

Let us then take stock and put down clearly what we are looking for.
We want to be able to explain why holding in thought [I believe that P,
but it is not the case that P] or [I do not believe that P, but it is the case
that P] would be so absurd as to result in self-defeat. Yet we do not want
to do so by saying that those contents are equivalent to [P and not-P].

What Moore’s Paradox Isn’t About:

Jane’s Odd Case
In order to come to grips with Moore’s paradox, it is instructive to realise
that not all kinds of self-ascriptions of beliefs (in conjunction with the
relevant non-doxastic contents) would give rise to it. Consider the fol-
lowing story, which involves the self-ascription of a previously uncon-
scious belief.

Jane is married to Jim. They have been married for several years and have a
daughter. Jane is often at home, on her own, attending to domestic chores.
From time to time, she feels lonely and wishes that she had pursued her
own career. More often than not, however, she feels much rewarded by the
fact that her family is so serene. Indeed, when she meets with her friends,
who sometimes complain about their husbands, she cannot help remarking
that her life makes her happy and that her husband is adorable and com-
pletely trustworthy.
Still, it often happens that, while preparing for the laundry, Jane care-
fully searches Jim’s pockets. While tidying up his studio, she opens and
examines all the drawers. While dusting the furniture, she lingers on the
screen of his laptop, left open on the incoming messages.
One day, Ann, a psychoanalyst friend of Jane’s, approaches her and tells her
about Freud’s theories concerning the unconscious. Little by little, the deep
significance of a whole series of previously meaningless actions is disclosed
to Jane. Ashamedly, she realises that all that attention spent over the con-
tent of her husband’s pockets was a sign of her being insecure about him.
All that dusting the screen of his laptop, a symptom of her thinking that he
might have some intimate correspondence with another woman. Still, Jane
knew all too well that Jim had always been the most truthful of men. The
thought popped into her head: “I do believe that Jim is unfaithful to me,
but he is not”.
254 Appendix: Moore’s Paradox

The moral to draw is that if one ascribes to oneself a previously

unconscious belief, on the basis of the scrutiny of one’s own behav-
iour (together with some general background theory) and also holds the
negation of its embedded content—as is the case with Jane’s final asser-
tion—then Moore’s paradox is dissolved. In other words, in the context
envisaged, judging (or asserting) [I do believe that Jim is unfaithful to
me, but he is not] amounts to a correct description of one’s own mind.
This is not to say that Jane would not display some kind of irrational-
ity (cf. Chap. 7, §§3–4). The point, however, is that her judgement or
assertion would be perfectly legitimate though certainly depicting an
odd state of mind.21

What Moore’s Paradox Is About—First Pass

Paying attention to Jane’s odd case actually helps us put forward a diagnos-
tic suggestion and also identify the constraints that any adequate account
of Moore’s paradox must meet in order to save the paradox and hence be
an account of what it is supposed to explain. The diagnostic suggestion
is this: any satisfactory account of Moore’s paradox will have to unravel
the complexity of our concept of belief.22 For the mere self-ascription of
a belief (or of a lack of it) is not enough invariably to generate a paradox
(if accompanied by the negation of its embedded content or by the assent
to it).23 Here are the constraints.

21
Lee 2001, pp. 366–7 has a similar story. Remarks that are congenial to the gist of Jane’s story can
be found in Moran 1997, Martin 1998, Williams 2006, and Gallois 2007. Chan 2008 presents a
case of wishful thinking in order to make a similar point. Gertler contends, contra Gendler, that
assertions such as Jane’s are, in the relevant context, perfectly legitimate and such that the doxastic
conjunct expresses full and real belief. See Gertler 2011b, and Gendler 2008.
22
Other kinds of analysis of Moore’s paradox, which hinge on the idea of unravelling the complex-
ity of the notion of belief, are Heal’s and Baldwin’s. They stand opposed to other strategies, such as
Shoemaker’s, which try to derive an explanation of Moore’s paradox from an account of self-knowl-
edge. Self-knowledge is a necessary condition for giving rise to the paradox, but it is not sufficient
for it.
23
The reader will make the necessary adjustments to apply the same kind of claim to (2).
 Appendix: Moore’s Paradox 255

1. One’s first-order belief must be self-known.24

For there is nothing paradoxical in unconsciously holding a first-order

belief which contradicts other beliefs that one explicitly entertains. Jane,
before becoming aware of the fact that her actions were partly motivated by
the belief that her husband was unfaithful to her, would openly claim that
her husband was trustworthy. In such a case, one could say that Jane was self-
deceived. Now, on one very plausible account of self-deception,25 this would
mean that while Jane had the explicit belief that her husband was a faithful
man, she also had the unconscious belief that he was not.26 No doubt, Jane
would be irrational, in some sense,27 but there would be nothing paradoxical
in her situation. Hence, a necessary condition upon any genuine instance of
Moore’s paradox is that a given first-order belief be self-known.
The second necessary condition we should list is this:

2. The self-ascription of belief, which constitutes the doxastic conjunct of

either (1) or (2), is not merely the self-attribution of a disposition.28

For there is nothing paradoxical in finding out and, therefore, in self-

ascribing a given belief, which contradicts what one explicitly judges to
be the case, as long as that self-ascribed belief is, in fact, a mental dispo-
sition.29 Intentional mental states, like beliefs and desires, as dispositions

24
Gallois 2007 defends this point, although he prefers to say that the belief must be consciously
held. Williams 2006 talks of consciousness in this connection, but he is happy to endorse
Rosenthal’s higher-order account of consciousness, in which the latter comes down to
self-knowledge.
25
See Chap. 7 (§§3–4).
26
Self-deception would then be a case in which one’s (unconscious) dispositions are not in keeping
with one’s explicitly held commitments and this creates a conflict in one’s outward behaviour (con-
tra what Moran 1997 maintains).
27
As will become evident in the following, Jane’s irrationality is not the irrationality of knowingly
holding a contradiction or of undertaking mutually inconsistent commitments (if this is a genuine
possibility at all). Rather it is the irrationality of not aligning one’s dispositions with one’s
commitments.
28
Functionalism treats beliefs as dispositions, and therefore as lacking intrinsic normativity. Hence,
as noted in Heal 1994, pp. 12–20, it is bound to dissolve Moore’s paradox. A similar point can be
found in Baldwin 2007, p. 84.
29
Bilgrami 2006, chapters 4–5. Millar 2004 considers beliefs as intrinsically normative mental
states. For further discussion, see chapter 1 (§2) and chapter 6 (§§3–4).
256 Appendix: Moore’s Paradox

are typically manifested in a subject’s behaviour and need not be self-

known.30 More importantly, they are not intrinsically normative mental
states. They are not governed by intrinsic “oughts” a subject should be
in a position to appreciate in order to have the relevant mental state.
They merely mediate between certain inputs and behavioural outputs
(together with other mental states) and do not depend on a subject’s
(mental) agency—that is to say, on her ability, in the case of belief, to
gather and assess evidence in favour of [P] and to use [P] as a premise in
theoretical and practical inferences. Hence, one can find oneself saddled
with beliefs as dispositions while also reflectively criticising and distanc-
ing oneself from them.
Freudian unconscious beliefs like Jane’s with respect to her husband’s
infidelity are a case in point.31 Indeed, in that case, the self-ascription
could be a first step towards trying to put one’s beliefs in line with each
other. For Jane could reason as follows: [I do have the belief (as a disposi-
tion) that my husband is unfaithful. Yet my husband is faithful. Therefore,
I should stop having the belief (as a disposition) that he is unfaithful].
Accordingly, she could then undertake psychoanalytic therapy, which
may eventually help her to overcome her belief as a disposition that her
husband is unfaithful to her.
Hence, the suggestion is that the perceived absurdity of holding (1)
and (2) may come from:

3. Taking the self-ascription of belief, which constitutes the doxastic

conjunct of either (1) or (2), as the self-attribution of a commitment.

We have just seen that the self-ascription of a belief as a disposition

is not enough to give rise to Moore’s paradox. However, not only do we
have a notion of belief as a disposition, but we also have the notion of a
belief as a commitment. There are different views about what beliefs (and

30
When they are, they are usually known in a third-personal way. That is to say, by observation and
inference to the best explanation.
31
Freudian unconscious mental states do not exhaust the category of dispositional beliefs and
desires. There may be also beliefs and desires that shape one’s character traits. Moreover, insofar as
one is willing to grant beliefs and desires to animals and infants, their intentional mental states
would be dispositions.
 Appendix: Moore’s Paradox 257

other intentional mental states like desires and intentions) as commit-

ments amount to, in the literature.32 The largely agreed upon feature of
these accounts is that beliefs as commitments are attitudes of acceptance
of a given propositional content that are intrinsically normative. To use
Bilgrami’s helpful explanation:

In this latter normative usage, to desire something, to believe something, is to

think that one ought to do or think various things, those things that are
entailed by those desires and beliefs by the light of certain normative princi-
ples of inference (those codifying deductive rationality, decision-theoretic
rationality, perhaps inductive rationality and also perhaps some broader forms
of material inference having to do with the meanings of words as well).33

Typically, we form beliefs as commitments by considering and assessing

the evidence at our disposal in favour of a given propositional content.
Believing as a commitment is therefore the result of mental agency. It is
not merely something that happens to one, perhaps unconsciously, so
that, when brought to light, one will find oneself saddled with it. This is
not to say that we believe at will. It merely entails that we have to be men-
tal agents in order to form beliefs as commitments, by being able to con-
sider and assess the evidence that bears on the truth of [P] (and, of course,
we can go astray in doing that). Moreover, we do have an (often implicit)
understanding of the principles governing various forms of rationality
that normatively constrain our beliefs as commitments. These principles
concern their content (like, if I believe that it is raining, I will believe that
there are clouds). Yet they also concern what having that kind of proposi-
tional attitude involves. For example, if I believe that it is raining and I do
not want to get wet, I ought, ceteris paribus, to go out with an umbrella,
whereas if I merely so wished, I would not. Moreover, if it turns out that
there is no sufficient evidence in favour of [P], I ought to withdraw from
believing it. All that entails that a subject capable of a belief as a commit-
ment ought to form it on the basis of sufficient (still defeasible) evidence;
ought to use the proposition believed as a premise in one’s practical and

32
See in particular Moran 2001 and Millar 2004. Cf. Chap. 3 (§2) and Chap. 7 (§§3–4).
33
Bilgrami 2012, p. 213.
258 Appendix: Moore’s Paradox

theoretical reasoning; ought to provide (defeasible) reasons for it, if chal-

lenged; and ought to revise it, were sufficient contrary evidence to come
in. Subjects are therefore rationally responsible for their beliefs as com-
mitments. Should they fail to comply with these ‘oughts’, they ought to
be self-critical or accept to incur criticism.
The contrast between beliefs as commitments and dispositions, then,
could be illustrated by Jane’s situation. She finds herself with a belief
as a disposition that her husband is unfaithful to her. That disposition
shapes much of her behaviour and can have various causes. However, she
also has and avows her belief as a commitment, held on the basis of evi-
dence that she herself has assessed, that he is not. The latter belief exerts
normative force on her and, consequently, she ought to try to get rid of
her recognisably irrational disposition. In the end, she might not be able
to overcome it (completely). Still, if rational, she ought to recognise that
that is what is required of her, given her belief as a commitment that her
husband is a faithful man.
What Bilgrami’s quote leaves unspecified and is not clarified, to the
best of my knowledge, in the literature on commitments, is whether one
can knowingly and willingly hold inconsistent commitments and there-
fore bind oneself, no matter how irrationally that would turn out to be,
to opposite courses of actions (and, consequently, to accepting criticism,
were one not to behave accordingly), or whether no such inconsistent
commitments can be upheld. We will presently see the bearing of these
different takes on the notion of commitment onto the analysis of Moore’s
paradox. Yet, before moving on to that, let us consider the relationship
between beliefs as commitments and dispositions a bit more.
As Bilgrami’s quote made apparent, having a belief as a commitment
consists in knowingly and willingly binding oneself to those courses of
action that “are entailed by those desires and beliefs by the light of certain
normative principles of inference”. If one does not comply with them,
one will be held responsible for not doing so and will have to be self-
critical or accept criticism from others for it. Thus, to have a belief as
a commitment entails seeing oneself as having to implement a certain
behaviour (and accepting criticism for not “living up to one’s commit-
ments” should one fail to behave accordingly). Therefore, there is an
internal link between the content of one’s belief as a commitment and
 Appendix: Moore’s Paradox 259

the kind of actions that one ought to perform given that belief. However,
theorists disagree on whether, in order to count as really having the com-
mitment in question, one also has actually to comply with the course of
action it mandates, at least on occasion, or whether no such behaviour
is required in order to count as having the commitment.34 Those who
favour the former option will appeal to the idea that of a subject who,
for example, persistently professed “I ought to help the poor” and never
did anything to that effect, we would presumably be inclined to say that
she thinks she has the commitment to help the poor but actually she
does not. In contrast, those who favour the latter option would argue
that insofar as such a subject recognised what would be required of her
and were self-critical or accepted criticism for not complying with her
commitment, she would count as having it nonetheless. Clearly, the two
readings impose more or less stringent requirements with respect to the
conditions that have to be met in order to have beliefs as commitments
and propose a different configuration of the interplay between commit-
ments and dispositions. This issue is of great interest and significance; it is
not, however, of immediate relevance for the present discussion and I will
show why in due course.35 What is relevant, at this stage, is simply that
the internal link between the content of one’s belief as a commitment
and the course of action it mandates be recognised (yet not necessarily
implemented, on at least one possible understanding of the notion of a
belief as a commitment) by the subject who could not otherwise have
that belief as a commitment.
Let us now turn to the kind of analysis of Moore’s paradox that can be
offered if one held a “liberal” notion of commitment, which allows for the
possibility that a subject could knowingly and willingly undertake incom-
patible commitments. Accordingly, the analysis of Moore’s paradox would
be as follows.36 Being committed to [P], one ought to use it as a premise

34
Millar 2004 seems to maintain the former option, whereas Bilgrami (2006, 2012) defends the
latter.
35
For a defence of the latter interpretation, see Chap. 2, §2 and Chap. 7, §§3–4.
36
I surmise that Baldwin 2007 (p. 86) puts forward something along these lines, for he allows for
the possibility of finding “ourselves from time to time with inconsistent commitments”. What
would be absurd, on his analysis, would be “to make commitments whose inconsistency is obvious
in the very judgment itself ”.
260 Appendix: Moore’s Paradox

of one’s practical and theoretical reasoning. Hence, if Jane’s judgement

(or assertion) contained the self-ascription of a commitment—as opposed
to the self-ascription of a disposition—to holding that her husband is
unfaithful to her, this would give rise to a form of irrationality. For if the
doxastic conjunct self-ascribed such a commitment, then Jane ought to
use [My husband is unfaithful to me] as a premise of her reasoning.37 By
also assenting to its negation, however, she would commit herself to using
(knowingly and willingly) that content as a premise of her reasoning.
Thus, she would commit herself (knowingly and willingly) to reasoning
from contradictory premises. And this would be irrational.
Take now an instance of (2)—[I do not believe that P, but it is the case
that P]. One may take its first conjunct as a self-ascription of one’s open-
mindedness as to whether [P] is the case. Accordingly, it would describe
one’s commitment to using neither [P] nor [not-P] as a premise of one’s
reasoning. However, by holding the second conjunct, one would also
commit oneself (knowingly and willingly) to using [P] as such a premise.
And this, once again, would be irrational.
On the face of it, then, the notion of a belief as a commitment seems
to allow us to save the paradox because it helps us show that while (1) and
(2) do not amount to contradictions, holding them in thought would
be a manifestation of irrationality. The conclusion one should draw thus
would be this: any creature capable of beliefs as commitments ought not
to judge either (1) or (2)—when construed as satisfying conditions (I)–
(III)—on pain of irrationality.

What Moore’s Paradox Is About—Second Pass

However, this is a somewhat weak result: people are often irrational and
so, if we allow that they can have inconsistent beliefs as commitments, it
seems that one could really judge (1) or (2). That is to say, it seems that one
could really be in the mental state of committing oneself to [P], say, while
also knowingly and willingly denying that [P] is the case (and, mutatis

37
Obviously, she would also commit herself to using [I believe that P] as a premise of her
reasoning.
 Appendix: Moore’s Paradox 261

mutandis, given the omissive form of Moore’s paradox, one could then be
open-minded with respect to [P], while also endorsing it). In particular,
it does not seem impossible to reason from contradictory premises which
are actually knowingly and willingly upheld and not merely supposed,
like in the case of (1), or to reason from genuine open-mindedness with
respect while also actually holding that P, in the case of (2). It is just that
these doxastic situations would—unluckily—be irrational ones. On such
a view, rational norms would be external to one’s beliefs as commitments
and would be brought to bear on them just as norms of etiquette can be
brought to bear on one’s way of sipping tea or holding a cup.38 That is to
say, as post hoc normative evaluations of doxastic situations (or actions)
which would have been instantiated (or performed), nevertheless. Yet,
if it were possible for a subject to endorse incompatible commitments,
what would then prevent one from judging and even from asserting that
one did? Of course such a situation would be comparatively rare—or so
one would hope. However, if it is a possibility, the corresponding judge-
ment and assertion should be too. The absurdity perceived in hearing
either (1) or (2) would be just due to the fact that we are inductively
unfamiliar with open admissions of irrationality, which, however, on the
proposed conception of commitment, would be entirely possible.
Yet, if this were the explanation of why we find the assertion of (1)
or (2) absurd, then to label Moorean contents “paradoxical” would be
an overstatement. There are plenty of assertions that strike us as weird
because—supposing they are true and sincere—they are at odds with
our experience, but this would not justify calling them “paradoxical” (for
example, “He is not breathing but he is alive”). Furthermore, it was a
datum of the problem that when we hear Moorean sentences we perceive
some kind of real impossibility—a form of self-defeat—despite the lack of
a contradiction. Whereas, so far, our analysis has revealed no impossibil-
ity, just the fact that if those judgements were actually made, they would
correspond to an irrational state of mind.
In order to justify the idea that there is a real impossibility here and
hence a genuine paradox, we should understand the notion of a belief as

38
I would like to thank Margaret Gilbert for raising an objection that made me realize more clearly
what seems to me to be wanting in the first proposed account of Moore’s paradox.
262 Appendix: Moore’s Paradox

a commitment in a more demanding way and, in particular, as entailing a

negative answer to the question whether it is possible knowingly and will-
ingly to hold incompatible commitments. Accordingly, one could not (logi-
cally) have the belief as a commitment that p is the case and knowingly and
willingly assent to its negation, for this would actually undo one’s previous
commitment. To see why this more demanding notion could have traction,
consider the internal link between the content of one’s beliefs as commit-
ments and the courses of action, mandated by them, which ought to be rec-
ognised by a subject in order to have those very beliefs as commitments. Let
me stress a point made in the previous section—namely, that having a belief
as a commitment entails seeing oneself as bound to the kind of action man-
dated by that commitment, although, on certain readings of commitments,
one may not be able to implement it. For our purposes, it is important to
realise that having a belief as a commitment entails (at least) the relevant
second-order disposition of being self-critical or of accepting criticism from
others if one did not behave as mandated by one’s belief as a commitment.
Take a case in which the belief is part of practical reasoning. Suppose, there-
fore, that a subject desires not to get wet if (and only if) it rains. Suppose,
further, that the only possible action, in order not to get wet, is to open the
umbrella that the subject happens to carry with her and that there are no
countervailing considerations. Now, if she has the belief as a commitment
that it is raining, she ought to see herself as bound to opening the umbrella.
If she has the belief as a commitment that it is not raining, she ought to see
herself as bound to not opening the umbrella. Finally, if she is open-minded
with respect to whether it is raining, she ought to see herself as bound neither
to opening the umbrella nor to keeping it close. The three kinds of action,
which are respectively internally linked to each kind of belief as a commit-
ment a subject might have, however, are mutually exclusive. Since, in order
to have a belief as a commitment, a subject will have to recognise the courses
of action it internally mandates, given that they are mutually exclusive, she
will not be in a position (logically) to bound herself to any two of them at
once. Thus, no matter what she could think of herself as doing, she could not
actually have incompatible beliefs as commitments.39 To stress, this is a point

39
A similar conclusion can be reached by considering the ‘oughts’ as input that should be appreci-
ated by a subject in order to have the relevant beliefs as commitments—namely, that one ought to
 Appendix: Moore’s Paradox 263

about logic and the internal link between a belief as a commitment and the
kind of action it mandates and one ought to see oneself as bound to, in order
to count as having the commitment. Of course, since we are open to the
possibility that a subject may have a commitment while behaving contrary
to what it mandates, we have to be open to the possibility that she might act
contrary to what the belief (as a commitment) that it is raining mandates,
in the kind of scenario just envisaged. However, in such an event, so long as
she did recognise that she has not “lived up” to her commitment, she would
count as having it in the first instance, nonetheless. Notice, once again, that
even if she behaved contrary to her commitment and therefore did not have
the relevant first-order disposition, she should have a related second-order
one, which depends precisely, as required, on seeing herself as bound to act
in certain ways. Hence, if she did not act accordingly, she should be self-
critical or accept criticism from others for not doing so. What could not be
done, therefore, on such a less demanding notion of a commitment, is not to
have a belief as a commitment and act in ways that run contrary to it. Rather,
it is to have the former and think it may be all right to act in ways that run
contrary to it. Yet, to stress, the constitutive connection that is necessary in
order to come to grips with Moore’s paradox—that is, the one concerning
the content of one’s commitments and the kind of actions mandated by
them one should see oneself as bound to—would be respected nonethe-
less. That is to say, even if one might fail to bring these actions about. Since
the courses of action mandated by inconsistent commitments are mutually
exclusive, one can’t have the belief as a commitment that P while also know-
ingly and willingly assenting to its negation (or be open-minded with respect
to [P] while knowingly and willingly assenting to it or to its negation), even
if one actually behaved in ways that run contrary to [P].
By contrast, as we have seen with Jane’s case, one can perfectly well
have the belief as a disposition that p while also knowingly and willingly
assenting to its negation. Hence, one might be saddled with beliefs (as

believe that P only if there is sufficient evidence in its favour. So one could self-ascribe the belief as
a commitment that P only if one had sufficient evidence in favour of [P]. Yet one cannot have suf-
ficient evidence for [P] and also and at the same time for its negation. Hence, one cannot possibly
have the inconsistent commitments entailed by (1). Similarly, one can be committed to open-
mindedness with respect to [P] only if one had neither sufficient evidence in favour of [P] nor for
its negation. Yet one cannot have, at once, no sufficient evidence for [P] and [not-P] and sufficient
evidence for of [P]. Hence, one cannot possibly have the inconsistent commitments entailed by (2).
264 Appendix: Moore’s Paradox

dispositions) that one does not endorse reflectively, act on them and even
overtly recognise the kind of situation one is in. Yet it seems that, on
such a more demanding notion of a belief as a commitment, one cannot
even have the belief as a commitment that P if one also assented—and
thereby committed oneself—to its negation. The assent to the negation
thus would undo one’s previous commitment to [P].
Therefore, given (1), if the doxastic conjunct were the self-ascription
of a commitment, understood in the way proposed, that commitment
could so much as exist only as long as no known and willing assent to
the negation of its content were in place. Hence, if the doxastic conjunct
is a self-ascription of a genuine commitment, no assent to its negation is
possible. Conversely, if such an assent is in place, then there cannot be
any commitment to [P] in the first place.
Take now an instance of (2)—[I do not believe that P, but it is the
case that P]—where the first conjunct is taken as a self-ascription of a
commitment, understood in the way proposed. By judging that P is the
case, one would undo one’s previous commitment to open-mindedness
with respect to [P]—that is, one’s commitment to using neither [P] nor
[not-P] as a premise of one’s reasoning. Hence, again, if the doxastic con-
junct is an ascription of a genuine commitment to open-mindedness, no
known and willing assent to its embedded content is possible. Yet, if such
an assent is in place, there cannot be any genuine commitment to open-
mindedness in the first place.
Therefore, the moral to draw is that any creature capable of beliefs as
commitments could not possibly judge either (1) or (2), despite the fact
that their conjuncts have independent truth conditions, when (1) and
(2) are construed as satisfying conditions (I)–(III). For to do so would
actually be self-defeating.
Finally, the absurdity of asserting either (1) and (2) can be explained
as follows. Usually, though not invariably (as we saw, for instance, with
Jane’s case), we assert the contents of our beliefs as commitments.40 As
we have seen, however, if (1) and (2) are asserted as involving beliefs as

40
This does not require taking issue with any of the various proposals about the norm of the asser-
tion available in the growing literature on that topic. For the claim is that we usually assert the
content of our commitments, not that this is the norm of correct assertions.
 Appendix: Moore’s Paradox 265

commitments, they would not simply be a manifestation of irrationality

but actually would be self-defeating in the sense of expressing an impos-
sible cognitive situation—that is, the situation of endorsing incompatible
commitments.

An Objection
The previous account of Moore’s paradox gives rise to an immediate
objection, for it seems that sometimes we find ourselves with inconsistent
commitments. Consider, for instance, the case of a man with two families
who at lunchtime says to his first family “I will go on holiday with you[family
1]” and at dinnertime says to his second family “I will go on holiday with
you[family 2]” and thereby undertakes inconsistent commitments.41
Let us analyse this case in more detail.42 There seem to be only four
options. First, he means to compatibilise by dividing his holidays
between the two families (or by taking them all to the same place at the
same time). Second, he has changed his mind between lunch and din-
ner and will go on holiday with the second family only. Third, he has
forgotten the promise made over lunch and therefore is not knowingly
and willingly undertaking two incompatible commitments.43 Finally, by
willingly trying to bind himself to knowingly incompatible courses of
41
Another possibility is to have beliefs as commitments that, though not incompatible as such, in
given circumstances, may impose incompatible courses of action. Consider the belief as a commit-
ment that a supervisor should help her students and the other belief as a commitment that one
should give an academic position to the best candidate. Obviously, there is no incompatibility
between the two commitments. Hence, one can knowingly and willingly assent to both. If, how-
ever, the situation arises when there is a conflict, one will have to give priority one commitment
over the other.
42
Even if, as Marcello Fiocco pointed out to me, this case is not one of allegedly inconsistent beliefs
as commitments, I think it is useful for illustrative purposes, since beliefs as commitments are spe-
cies of commitments.
43
Cases like Frege’s puzzle where one can assent to “Hesperus is a planet” while denying or not
assenting to “Phosphorous is a planet” are cases in which the beliefs are incompatible but this is not
known to the subject, who precisely ignores that Hesperus is identical to Phosphorus. Hence, it is
important to qualify the claim here: it is only for knowingly incompatible beliefs as commitments
that one cannot (knowingly and willingly) have both of them at once. Another way to see the same
point is that to judge “I believe that Hesperus is a planet, but Phosphorus is not” (for a subject who
did not know that Hesperus is Phosphorus) would not be Moorean-paradoxical. I would like to
thank John Hawthorne for raising an objection that made me think more about these cases.
266 Appendix: Moore’s Paradox

action, he keeps alternating between the two, up to the point of actually

losing the capacity of committing himself to either. Yet there is no single
overall situation in which he is really knowingly and willingly committed
to both at once.
This point may escape notice because we have what we might call
a “social” or “third-personal” notion of commitment, such that peo-
ple who have heard his pronouncements will actually take him to have
bound himself on both occasions and then entered a state of holding
incompatible commitments. Indeed, he can even apply such a third-per-
sonal point of view to himself and regard himself as having committed
himself to both courses of action. However, he has not, at a “personal”
or even “first-personal” level. For he holds compatible commitments
(first option) or he has broken the promise made to one of his families
(second option) and committed to taking on holiday just family2 or
he has forgotten his earlier promise (third option). In the latter two
cases, he may respectively be accused either of not being trust-worthy
or of being absent-minded. Yet, in this latter case, he cannot be accused
of having personal (or first-personal) incompatible commitments. This
can be appreciated by noticing that by recalling the promise made over
lunch, he will have to deliberate again what to do. For, if he did not
deliberate again, he would not know what to do at all. If one finds one-
self confronting the question “What shall I do?”, that shows that one
does not have any first-personal commitment, even if one can recognise
that from a third-personal point of view one would be required to do
both things.
Note that I do not wish to oppose such a “social” or “third-personal”
notion of a commitment. The point of the last two sections of this
Appendix, however, has been to press the need for the recognition of what
we may call a “personal” or “first-personal” notion of a commitment—
that is to say, the notion of a particular kind of belief (with a characteristic
course of action one would have to see oneself as bound to, even if one
might not live up to it), which one can enter only by willingly and know-
ingly assenting to a given content and which is incompatible with (at
least) believing the opposite. Hence, first, we cannot just find ourselves
with first-personal commitments, incurred out of habit or out of com-
pliance with social expectations. Rather, first-personal commitments are
 Appendix: Moore’s Paradox 267

always the result of a conscious deliberation and therefore of an exercise

of agency on a subject’s part. Even the humdrum case of believing that it
is raining because one sees the rain fall outside one’s window will be the
result of consciously considering that perceptual evidence and of deeming
it sufficient to warrant one’s belief.
Second, for all we have seen in the previous two sections, a person may
still have a given belief as a commitment while not behaving in keeping
with it, so long as she were prepared to be self-critical or to accept criti-
cism for her dissonant behaviour. Hence, in the case of the man with two
families, he could still have the commitment to going on holiday with the
first family, while actually ending up going on holiday with the other one
(or simply saying so to the second family), so long as he were prepared
to be self-critical or to accept criticism for so doing. Notice, in this case,
that his first-order dispositions, which would run contrary to his com-
mitment, would give rise to courses of action, which would be carried out
somehow “against one’s own will and judgement” as it were, perhaps out
of fear or habit. If not, if the man had in fact willingly taken the second
family on holiday (or actually committed himself to doing so), we then
would be back to the second possibility envisaged above—that is, the one
in which he would have changed his mind and relinquished the promise
made over lunch to his first family.
If this account is on the right track, we can actually see why Moore’s
paradox is indeed so perplexing: not because it describes a situation,
which, as irrational as it may be, can actually happen. For, in that case,
one should just acknowledge that there will be occasions that would
license the judgement (and the assertion) of either (1) or (2), as weird as
they might be. Down that route, however, one would no longer have the
means to maintain that Moorean contents are paradoxical. They would
just seem weird because they would depict comparatively infrequent and
unusual states of affairs. Rather, Moore’s paradox is indeed a genuine and
deep paradox because, despite not containing any contradictory content,
judging or asserting (1) or (2) would express an impossible cognitive situ-
ation. That is to say, the situation of endorsing incompatible (personal)
commitments. If this is right, then, we will have the means to under-
stand why, as Moran remarked, Moore’s paradox is indeed “an emblem
for peculiarities in the first-person point of view”.
 Bibliography

Armstrong, D. (1968). A materialist theory of the mind. London/New York:

Routledge.
Baldwin, T. (1990). G. E. Moore. London/New York: Routledge.
Baldwin, T. (2007). The normative character of belief. In M. Green & J. N.
Williams (Eds.), Moore’s paradox. New essays on belief, rationality and the first
person (pp. 76–89). Oxford: Clarendon Press.
Bar-On, D. (2004). Speaking my mind. Expression and self-knowledge. Oxford:
Oxford University Press.
Bilgrami, A. (2006). Self-knowledge and resentment. Cambridge, MA: Harvard
University Press.
Bilgrami, A. (2012). The unique status of self-knowledge, in Coliva A. (2012),
263–278.
Boghossian, P. (1989). Content and self-knowledge. Philosophical Topics, 17,
5–26.
Bonjour, L. (2003). A version of internalist foundationalism. In L. Bonjour &
E. Sosa (Eds.), Epistemic justification. Internalism vs. externalism, foundations
vs. virtues (pp. 61–68). Oxford: Blackwell.
Boyle, M. (2009). Two kinds of self-knowledge. Philosophy and Phenomenological
Research, 77(1), 133–164.
Boyle, M. (2011a). Transparent self-knowledge. Aristotelian Society Supplementary
Volume, 85(1), 223–241.

© The Editor(s) (if applicable) and The Author(s) 2016 269

A. Coliva, The Varieties of Self-Knowledge,
DOI 10.1057/978-1-137-32613-3
270 Bibliography

Boyle, M. (2011b). Making up your mind. Philosophers’ Imprint, 11(17), 1–24.

Boyle, M. (2015). Critical study of Quassim Cassam Self-knowledge for humans.
The European Journal of Philosophy, 23(2), 337–348.
Brady, M. (2013). Emotional insight. Oxford: Oxford University Press.
Brewer, B. (1999). Perception and reason. Oxford: Oxford University Press.
Brückner, A. (2006). Justification and Moore’s paradox. Analysis, 66(3),
264–266.
Brückner, A. (2009). More on justification and Moore’s paradox. Analysis, 69(3),
497–499.
Burge, T. (1993). Perceptual entitlement. Philosophy and Phenomenological
Research, 67, 503–548.
Burge, T. (1996). Our entitlement to self-knowledge. Proceedings of the
Aristotelian Society, 96, 91–116. Reprinted in P. Ludlow and N. Martin
(Eds.). (1998). Externalism and self-knowledge (pp. 239–264). Stanford: CSLI
Publications.
Burge, T. (2010). Origins of objectivity. Oxford: Oxford University Press.
Burge, T. (2011). Self and self-understanding. The Dewey lectures (2007–2011),
The Journal of Philosophy, CVIII/6–7. Reprinted in Cognition through under-
standing (pp. 140–228). Oxford: Oxford University Press.
Byrne, A. (2011a). Transparency, belief, intention, Proceedings of the Aristotelian
Society Supplementary Volume, 85(1), 201–221.
Byrne, A. (2011b). Knowing that I am thinking, in Hatzimoysis A. (ed), pp.
105–124.
Cassam, Q. (2014). Self-knowledge for humans. Oxford: Oxford University Press.
Chan, T. (2008). Belief, assertion and Moore’s paradox. Philosophical Studies,
139(3), 395–414.
Chan, T. (2010). Moore’s paradox is not just another pragmatic paradox.
Synthese, 173(3), 211–229.
Claparède, E. (1928). Emotions and feelings. In M. L. Reynert (Ed.), Feelings
and emotions: The Wittenberg symposium (pp. 124–139). Worchester: Clark
University Press.
Coliva, A. (2006). I Concetti. Roma: Carocci.
Coliva, A. (2009). Self-knowledge and commitments. Synthese, 171(3),
365–375.
Coliva, A. (2010). Moore and Wittgenstein. Scepticism, certainty and common
sense. London: Palgrave Macmillan.
Coliva, A. (2012a). Critical notice of Tyler Burge Origins of objectivity. Disputatio,
4(33), 515–530.
 Bibliography 271

Coliva, A. (2012b). Human diagrammatic reasoning and seeing-as. Synthese,

186(1), 121–148.
Coliva, A. (Ed.). (2012c). The self and self-knowledge. Oxford: Oxford University
Press.
Coliva, A. (2014). Review of Jordi Fernández Transparent minds. Theoria, 81,
442–445.
Coliva, A. (2015a). Extended rationality. A hinge epistemology. London: Palgrave
Macmillan.
Coliva, A. (2015b). Review of Quassim Cassam Self-knowledge for humans.
Analysis. doi: 10.1093/analys/anv078.
Collins, A. W. (1996). Moore’s paradox and epistemic risk. Philosophical
Quarterly, 46(184), 308–319.
Davidson, D. (1970). Mental events. In L. Foster & J. W. Swanson (Eds.),
Experience and theory. London: Duckworth. Reprinted in Essays on actions
and events (pp. 79–102). Oxford: Clarendon Press.
Davies, M. (1998). Externalism, architecturalism and epistemic warrant, in
Wright C., Smith B. and Macdonald C. (Eds.), 321–361.
Davies, M. (2004). Epistemic entitlement, warrant transmission and easy
knowledge. The Aristotelian Society Supplementary Volume, 78, 213–245.
Davies, M., & Stone, T. (Eds.). (1995a). Folk psychology: The theory of mind
debate. Oxford: Blackwell.
Davies, M., & Stone, T. (Eds.). (1995b). Mental simulation: Evaluations and
applications. Oxford: Blackwell.
de Almeida, C. (2001). What Moore’s paradox is about. Philosophy and
Phenomenological Research, 62(1), 33–58.
Dennett, D. (1987). The intentional stance. Cambridge, MA: MIT Press.
Deonna, J., & Teroni, F. (2012). The emotions. A philosophical introduction.
London/New York: Routledge.
Dokic, J., & Proust, J. (Eds.). (2002). Simulation and knowledge of action.
Amsterdam/Philadelphia: John Benjamins Publishing Company.
Dummett, M. (1996). Origins of analytic philosophy. Cambridge, MA: Harvard
University Press.
Evans, G. (1982). The varieties of reference. Oxford: Oxford University Press.
Fernández, J. (2013). Transparent minds. Oxford: Oxford University Press.
Finkelstein, D. (2003). Expression and the inner. Cambridge, MA: Harvard
University Press.
Frith, U., & Happé, F. (1999). Theory of mind and self-consciousness: What is
it like to be autistic? Mind and Language, 14(1), 1–22.
272 Bibliography

Gallois, A. (1996). The world without, the mind within. Cambridge: Cambridge
University Press.
Gallois, A. (2007). Consciousness, reasons and Moore’s paradox. In M. Green
& J. N. Williams (Eds.), Moore’s paradox. New essays on belief, rationality and
the first person (pp. 165–188). Oxford: Clarendon Press.
Geach, P. (1965). Assertion. The Philosophical Review, 74, 449–465. Reprinted
in Rosemberg, J. F. & Travis, C. (Eds.) (1971). Reading in the philosophy of
language (pp. 250–261). Englewood: Prentice-Hall.
Gendler, T. (2007). Self-deception as pretense. Philosophical Perspectives, 21,
231–258.
Gendler, T. (2008). Alief and belief. Journal of Philosophy, 105(10), 634–663.
Gertler, B. (2011a). Self-knowledge. London/New York: Routledge.
Gertler, B. (2011b). Self-knowledge and the transparency of belief, in
Hatzimoysis A. (ed), 125–145.
Gilbert, D. (2006). Stumbling on happiness. New York: Knopf.
Goldman, A. (1993). The psychology of folk psychology. Behavioral and Brain
Sciences, 16, 15–28.
Gopnik, A. (1983). How we know our minds: The illusion of first-person
knowledge of intentionality. Behavioral and Brain Sciences, 16, 1–15.
Reprinted in Goldman, A. (ed) (1993). Readings in philosophy and cognitive
science. Cambridge, MA: MIT Press.
Gordon, R. (1995). Simulation without introspection or inference from me to
you, in Davies M. and Stone T. (1995b) (Eds.), 53–67.
Gordon, R. (2007). Ascent routines for propositional attitudes. Synthese, 159,
151–165.
Gordon, R. (2009). Folk psychology and mental simulation, The Stanford ency-
clopedia of philosophy (Fall 2009 Edition), Edward N. Zalta (Ed.), URL =
http://plato.stanford.edu/archives/fall2009/entries/folkpsych-simulation/
Green, M., & Williams, J. N. (Eds.). (2007). Moore’s paradox. New essays on
belief, rationality and the first person. Oxford: Clarendon Press.
Hacker, P. (1986). Insight and illusion: Themes from the philosophy of Wittgenstein.
Oxford: Clarendon Press.
Hatzimoysis, A. (Ed.). (2011). Self-knowledge. Oxford: Oxford University Press.
Heal, J. (1994). Moore’s paradox: A Wittgensteinian approach. Mind, 103(409),
5–24.
Heal, J. (2002). First person authority, Proceedings of the Aristotelian Society,
102(1), 1–19. Reprinted in (2003) Mind, reason and imagination (pp. 273–
288). Cambridge: Cambridge University Press.
 Bibliography 273

Heal, J. (2012). Consciousness and self-awareness, in Coliva A. (Ed.),

123–138.
Hume, D. (1978). A treatise of human nature. Oxfod: Oxford University Press.
Jacobsen, R. (1996). Wittgenstein on self-knowledge and self-expression.
Philosophical Quarterly, 46, 12–30.
James, W. (1950). The principles of psychology. New York: Dover.
Kripke, S. (1982). Wittgenstein on rules and private language. Oxford: Blackwell.
Lee, B. D. (2001). Moore’s paradox and self-ascribed belief. Erkenntnis, 55(3),
359–370.
Libet, B. (1985). Unconscious cerebral initiative and the role of conscious will
in voluntary action. Behavioral and Brain Sciences, 8, 529–566.
Linville, K., & Ring, M. (1991). Moore’s paradox revisited. Synthese, 87(2),
295–309.
Lycan, W. (1996). Consciousness and experience. Cambridge, MA: MIT Press.
Malcolm, N. (1995). Disintangling Moore’s paradox. In R. Egidi (Ed.),
Wittgenstein: Mind and language (pp. 195–205). Dordrecht: Kluwer.
Martin, M. (1998). An eye directed outward, in Wright C., Smith B. and
Macdonald C. (Eds.), 99–121.
McDowell, J. (1994). Mind and world. Cambridge, MA: Harvard University
Press.
McDowell, J. (1998). Response to Crispin Wright, in Wright C., Smith B. and
Macdonald C. (Eds.), 47–62.
McHugh, C. (2012). Reasons and self-knowledge, in Coliva A. (Ed.),
139–163.
McKinsey, M. (1991). Anti-individualism and privileged access. Analysis, 51,
9–16.
Millar, A. (2004). Understanding people. Normativity and rationalizing explana-
tion. Oxford: Oxford University Press.
Moore, G. E. (1942). Reply to my critics. In P. Schilpp (Ed.), The philosophy of
G. E. Moore (pp. 535–677). La Salle Illinois: Open Court.
Moore, G. E. (1944). Russell’s theory of descriptions. In P. Schilpp (Ed.), The
philosophy of Bertrand Russell (pp. 175–225). La Salle Illinois: Open Court.
Moran, R. (1997). Discovery, resolution and undoing. European Journal of
Philosophy, 5(2), 141–161.
Moran, R. (2001). Authority and estrangement. Princeton: Princeton University
Press.
Moran, R. (2003). Responses to O’Brien and Shoemaker. European Journal of
Philosophy, 11(3), 402–419.
274 Bibliography

Moran, R. (2012). Self-knowledge, transparency, and the forms of activity. In

D. Smithes & D. Stoljar (Eds.), Introspection and consciousness (pp. 211–236).
Oxford: Oxford University Press.
Nisbett, R., & Wilson, T. (1977). Telling more than we can know: Verbal reports
on mental processes. Psychological Review, 84, 231–259.
Nozick, R. (1981). Philosophical explanations. Cambridge, MA: Harvard
University Press.
Nussbaum, M. (2001). Upheavals of thought. The intelligence of emotions.
Cambridge: Cambridge University Press.
O’Brien, L. (2003). Moran on agency and self-knowledge. The European Journal
of Philosophy, 11(3), 375–390.
O’Brien, L. (2007). Self-knowing agents. Oxford: Oxford University Press.
Parent, T. (2013). Externalism and self-knowledge. The Stanford encyclopedia of
philosophy (Summer 2013 Edition), Edward N. Zalta (Ed.), URL = http://
plato.stanford.edu/archives/sum2013/entries/self-knowledge-externalism/
Peacocke, C. (1992). A study of concepts. Cambridge, MA: MIT Press.
Peacocke, C. (1999). Being known. Oxford: Clarendon Press.
Peacocke, C. (2001). Does perception have a nonconceptual content? Journal of
Philosophy, XCVIII, 239–264.
Peacocke, C. (2002). Three principles of rationalism. European Journal of
Philosophy, X, 375–397.
Peacocke, C. (2003). The realm of reason. Oxford: Clarendon Press.
Peacocke, C. (2012). Conscious events and self-ascriptions: Comments on Heal
and O’Brien, in Coliva A. (Ed.), 180–188.
Povinelli, D. J., & Vonk, J. (2004). We don’t need a microscope to explore the
chimpanzee’s mind. Mind and Language, 19(1), 1–22.
Pritchard, D. (2005). Epistemic luck. Oxford: Oxford University Press.
Pryor, J. (2004). What’s wrong with Moore’s argument? Philosophical Issues, 14,
349–378.
Pryor, J. (2005). There is immediate justification. In M. Steup & E. Sosa (Eds.),
Contemporary debates in epistemology (pp. 181–202). Oxford: Blackwell.
Putnam, H. (1975). The meaning of ‘meaning’. Minnesota Studies in the
Philosophy of Science, 7, 131–193.
Scanlon, T. (1998). What we owe to each other. Cambridge, MA: Harvard
University Press.
Schopenhauer, A. (1839/1999). Essay on freedom of the will. Cambridge:
Cambridge University Press.
Schwitzgebel, E. (2008). The unreliability of naïve introspection. Philosophical
Review, 117, 245–273.
 Bibliography 275

Shah, N., & Velleman, D. J. (2005). Doxastic deliberation. Philosophical Review,

114(4), 497–534.
Shoemaker, S. (1968). Self-reference and self-awareness. Journal of Philosophy,
65, 555–578.
Shoemaker, S. (1986). Introspection and the self. Midwest Studies in Philosophy,
10, 101–120. Reprinted in Shoemaker S. (1996a), 3–24.
Shoemaker, S. (1988). On knowing one’s mind. Philosophical Perspectives, 2,
183–209. Reprinted in Shoemaker S. (1996a), 25–49.
Shoemaker, S. (1990). First person access. Philosophical Perspectives, 4, 187–214.
Reprinted in Shoemaker S. (1996a), 50–73.
Shoemaker, S. (1995). Moore’s paradox and self-knowledge. Philosophical
Studies, 77(2–3), 211–228. Reprinted (with small variations) in Shoemaker
S. (1996a), pp. XX–XX.
Shoemaker, S. (1996a). The first person perspective and other essays. Cambridge:
Cambridge University Press.
Shoemaker, S. (1996b). Self-knowledge and inner sense. Lectures I-III, in
Shoemaker S. (1996a), 201–268.
Shoemaker, S. (2003). Moran on self-knowledge. European Journal of Philosophy,
11(3), 391–400.
Snowdon, P. (2012). How to think about phenomenal self-knowledge, in Coliva
A. (Ed.), 243–262.
Sorensen, R. (1988). Blindspots. Oxford: Oxford University Press.
Sosa, E. (2007). A virtue epistemology. Oxford: Oxford University Press.
Stoneham, T. (1998). On believing that I am thinking. Proceedings of the
Aristotelian Society, 98, 125–144.
Strawson, P. (1962). Freedom and resentment. Proceedings of the British Academy,
48, 1–25.
Tanney, J. (1996). A constructivist picture of self-knowledge. Philosophy, 71,
405–422.
Tomasello, M., Call, J., & Hare, B. (2003a). Chimpanzees understand psycho-
logical states—The question is which ones and to what an extent. Trends in
Cognitive Science, 7, 153–156.
Tomasello, M., Call, J., & Hare, B. (2003b). Chimpanzees versus humans: It’s
not that simple. Trends in Cognitive Science, 7, 239–240.
Turri, J. (2010). Refutation by elimination. Analysis, 70(1), 35–39.
Vahid, H. (2005). Moore’s paradox and Evans’ principle: A reply to Williams.
Analysis, 65(4), 337–341.
Wegner, D. (2002). The illusion of conscious will. Cambridge, MA: MIT Press.
276 Bibliography

Wegner, D., & Wheatley, T. (1999). Apparent mental causation: Sources of the
experience of will. American Psychologist, 54, 480–492.
Williams, D. (1994). Somebody somewhere. London: Corgi Books.
Williams, J. N. (2004a). Moore’s paradoxes, Evans’s principle and self-knowl-
edge. Analysis, 64(4), 348–353.
Williams, M. (2004b). Wittgenstein, truth and certainty. In M. Kölbel & B.
Weiss (Eds.), Wittgenstein’s lasting significance (pp. 249–284). London/New
York: Routledge.
Williams, J. N. (2006). Moore’s paradoxes and conscious belief. Philosophical
Studies, 127(3), 383–414.
Williams, J. N. (2007). Moore’s paradox, Evans’ principle, and iterated beliefs.
In M. Green & J. N. Williams (Eds.), Moore’s paradox. New essays on belief,
rationality and the first person (pp. 90–113). Oxford: Clarendon Press.
Williams, J. N. (2009). Justifying circumstances and Moore-paradoxical beliefs:
A response to Brückner. Analysis, 69(3), 490–496.
Williamson, T. (2000). Knowledge and its limits. Oxford: Oxford University
Press.
Wilson, T. (2002). Strangers to ourselves. Discovering the adaptive unconscious.
Cambridge, MA: MIT Press.
Wimmer, H., & Perner, J. (1983). Beliefs about beliefs. Representation and
constraining function of wrong beliefs in young children’s understanding of
deception. Cognition, 13, 103–128.
Wittgenstein, L. (1980). Remarks on the philosophy of psychology (Vol. 1). Oxford:
Blackwell.
Wright, C. (1985). Facts and certainty. Proceedings of the British Academy, 71,
429–472.
Wright, C. (1986). Does philosophical investigations §§258–60 suggest a cogent
argument against private language? In P. Pettit and J. McDowell (Eds.),
Subject, thought and context. Oxford: Oxford University Press. Reprinted in
Wright C. (2001), 223–290. (Referred to as Wright 2001c).
Wright, C. (1987). On making up one’s mind: Wittgenstein on intention. In P.
Weingartner & G. Shurz (Eds.), Logic, philosophy of science and epistemology:
Proceedings of the XIth international Wittgenstein symposium. Vienna: Holder-
Pickler-Tempsky. Reprinted in Wright C. (2001), 116–142. (Referred to as
Wright 2001a).
Wright, C. (1989a). Wittgenstein’s rule-following considerations and the central
project of theoretical linguistics. In A. George (Ed.), Reflections on Chomsky.
Oxford: Blackwell. Reprinted in Wright C. (2001), 170–213. (Referred to as
Wright 2001b).
 Bibliography 277

Wright, C. (1989b). Wittgenstein’s later philosophy of mind: Sensation, privacy

and intentions. Journal of Philosophy, 86, 222–234. Reprinted in Wright C.
(2001), 291–318. (Referred to as Wright 2001d).
Wright, C. (1992). Truth and objectivity. Cambridge, MA: Harvard University
Press.
Wright, C. (1996a). The problem of self-knowledge (I), in Wright C. (2001),
319–344. (Referred to as Wright 2001e).
Wright, C. (1996b). The problem of self-knowledge (II), in Wright C. (2001),
345–373. (Referred to as Wright 2001f ).
Wright, C. (1998). Self-knowledge. The Wittgensteinian legacy, in Wright C.,
Smith B. and Macdonald C. (Eds.), 15–45.
Wright, C. (2001). Rails to infinity. Cambridge, MA: Harvard University Press.
Wright, C. (2002). What could antirealism about ordinary psychology possibly
be? The Philosophical Review, 111, 205–233.
Wright, C. (2004). Warrant for nothing (and foundations for free)? The
Aristotelian Society Supplementary Volume, 78, 167–212.
Wright, C. (2015). The reality of privileged access. In S. C. Goldberg (Ed.),
Externalism, self-knowledge and skepticism (pp. 49–74). Cambridge:
Cambridge University Press.
Wright, C., Smith, B., & Macdonald, C. (Eds.). (1998). Knowing our own
minds. Oxford: Clarendon Press.
 Index

A argument
affective forecasting, 6, 8, 14, 16, 68
agency, 18, 62, 123, 123n45, 183,
186, 187, 232, 267
mental, 211, 256, 257
agent, 57, 66, 167, 168, 171, 179,
183, 209, 257
responsible, 57, 74, 81, 122, 125,
126, 232, 241, 242
alienation, 56
anti-individualism, 21–2
a posteriori, 80, 185
a priori,, 6, 10, 12, 13, 17, 19n, 51,
52n2, 56–8, 60–2, 66, 74,
80, 83, 101, 109, 110,
110n26, 112, 112n32, 116,
125–9, 163, 164, 167, 186,
223, 227, 232, 241
Fregean, 185
Moorean, 185
open-question, 185
pincer, 185, 188
why questions, 44, 45
Armstrong, David, 7, 77–84, 90, 95,
101, 116, 167
assertion, 82, 92, 127, 128, 147,
148, 151–2, 174, 192,
196, 201, 202, 202n46–7,
204, 205, 205n49, 205n51,
206, 206n51, 207, 207n55,
209, 211n63, 214,
244n4, 245–9, 249n14,
250, 250n17, 252, 254,
254n21, 260, 261,
264n40, 267

Note: Page number followed by ‘n’ refers to footnotes.

© The Editor(s) (if applicable) and The Author(s) 2016 279

A. Coliva, The Varieties of Self-Knowledge,
DOI 10.1057/978-1-137-32613-3
280 Index

attention, 9, 39, 43, 59, 60, 71, 77, 217, 244, 257–62, 262n,
79, 87, 101, 103, 103n11, 264, 265n41–3
104–6, 106n16, 108n23, as dispositions, 5, 17, 19, 29, 31,
109, 119, 145, 155, 159, 37, 69, 149, 186, 198, 200,
238, 253, 254 208, 210, 234, 235, 255,
Audi, Robert, 42 255n28, 256, 256n31, 258,
Austen, Jane, 29, 53, 71, 238 263, 264
authority judgement-dependent, 194, 218
strong, 67 perceptual, 24, 25
weak, 125 Bilgrami, Akeel, 12, 13, 27n5,
avowals 31n11, 32n12, 32n14,
extension determining, 178 107n20, 165, 166, 183–8,
negative, 155 189n32, 198, 198n41, 199,
reportive, 155, 159 200, 201n44, 213, 214,
avowing 255n29, 257, 258
descriptive function of, 138–9 blind-sight, 61, 62, 232
expressive function of, 138 Boghossian, Paul, 67, 163
awareness BonJour Laurence, 106, 106n18,
conscious, 105 107, 107n19
phenomenal, 115 Boyle, Mathew, 3n2, 124n48,
127n54, 153n11, 172n10
Brady, Michael, 43, 44
B Brandom, Robert, 31
background assumptions, 43 Brewer, Bill, 108
Baldwin, Tom, 244, 254n22, Brüeckner, Anthony, 251n20
255n28, 259n36 Burge, Tyler, 6, 9–11, 20–5, 20n1,
Bar-on, Dorit, 6, 10, 134, 192n35 25n3, 26, 66n9, 100,
behavior 111–18, 128, 153, 156n24,
linguistic, 17, 66, 71, 91, 92, 152, 169, 179, 182, 201n45
158, 168, 228 Byrne, Alex, 9, 9n7
non linguistic, 66, 71, 158, 168,
210, 228
behaviourism, 78, 87, 95, 134, 135, C
180 Cartesian picture, 53
belief(s) Cassam, Quassim, 4n6, 8, 78, 84–8,
as commitments, 36, 37, 60, 125, 95, 124n48
183–4, 186, 189, 190, 195, cause(s)
195n, 202, 204, 208–10, of actions, 146

of mental state(s), 1
subpersonal, 180
unconscious, 68, 238
C-conditions, 12, 16, 17, 60, 61,
64–6, 151, 165, 167, 183,
189, 197, 198, 212, 213,
223, 224, 226, 228, 231,
241
Chan, Timothy, 245n6, 254n21
character, 6, 16, 30n9, 75, 106n18,
236–8
Claparède, Edouard, 40
cognitive science, 6, 51, 67–9, 75, 88
Coliva, Annalisa, 42, 43n, 188n
Collins, Arthur W., 245n5
commitment(s)
incompatible, 259, 261, 262,
265, 266
personal (or first-personal), 38,
38n20, 266, 267
social (or third-personal), 38, 266
concept(s)
acquisition of, 92, 229
blind, 166, 197, 218, 223
dispositional, 239, 242
individuation of, 75, 196
inferential, 87, 91
nature of, 8, 11, 89
psychological, 6, 8, 11–14, 17,
51, 53, 54n, 61, 62, 66, 73,
81, 83, 89, 90, 92–4, 96,
165, 166, 177, 191, 192n,
194, 195, 197, 205, 212,
215, 218, 219, 221, 228,
236, 237, 239, 242
Congenital Insensitivity to Pain with
Anhidrosis (CIPA), 227
constitutive
Index 281
theories, 163–215
metaphysically robust, 12, 165,
179, 188–212, 214, 219,
240, 241
thesis, 12, 164–74, 183–93,
197–214, 223–32
constructivism, 206n52
content
externalism, 6, 67–9
narrow, 7, 52, 67
perceptual
conceptual, 20, 24–5
non conceptual, 47
wide, 7, 67
contextualism, 138, 139, 139n, 148,
151
contradiction, 245–8, 245n7,
248n11, 249–52, 249n13,
255n27, 260, 261
critical reasoning/reasoner, 6, 66n,
113–18, 118n41, 128, 129,
169–74
D
Davidson, Donald, 179
Davies, Martin, 43n, 194n
de Almeida, Claudio, 249n15,
250n16
deliberation, 1, 9, 28, 32–4, 36, 48,
58, 89, 124, 125n49, 127,
129, 205, 208, 209,
209n60, 220, 233, 267
Dennett, Dan, 179
Deonna, Julien, 40, 44
description, 124, 150, 178, 205,
208n58, 209, 248, 254
de Sousa, Ronald, 42
282 Index
dispositions
first-order, 31, 37, 185, 263, 267
second-order, 38, 262
distinct existences, 173
Dokic, Jérôme, 85n8
Dummett, Michael, 29n6
E Fernández, Jordi, 89, 100, 119–22,
eliminativism, 244n4
embedding, 247
emotions
basic, 6, 11, 14–16, 18, 39, 40,
52, 59, 61, 69, 164, 218,
222–32, 240
borderline view of, 19, 46–7
as evaluative judgments, 5, 43, 44
as felt bodily attitudes, 5, 40–1
as perceptions of evaluative
properties, 41–6
as sensations, 39
entitlement(s), 11, 100, 110–13,
115–18, 126–8, 131, 141,
156, 158
expressive, 11, 157
epistemic achievement, 9, 58, 62, 66,
133, 163, 200, 212, 225
Euthyphro contrast, 178
Evans, Gareth, 8, 58, 94, 100, 119,
119n, 250n17
Evans’s point, 8152, 250, 250n17,
251, 251n20
evidence
behavioural, 28, 123
mnestic, 14, 220–1, 240
observational, 73
expressivism, 11, 15, 133–60, 222
neo-, 152–9
externalism
about content, 6, 67–9
epistemological, 116
rational, 100, 111–18, 128
F
family resemblance, 47, 139
124, 124n47, 129
Finkelstein, David, 127n, 152n,
153n11
Frege, Gottlob, 185, 265n43
Frith, Uta, 196n
functionalism, 255n28
G
Gallois, André, 89, 250n17, 254n21,
255n24
Geach, Peter, 205n50, 207n56,
211n63
Gendler, Tamar, 254n21
Gertler, Brie, 53n6, 68, 69, 254n21
Gilbert, Daniel, 68n12
Goldman, Alvin, 8, 78, 88–96
Gopnik, Alison, 3n4, 7, 77, 84–8, 95
Gordon, Robert, 8, 78, 88–96,
192n, 218
grammatical, 15, 133, 142–5, 204
fiction, 136
Green, Mitchell, 243n, 244n2,
245n6
Grice, Paul, 139n
groundlessness
strong, 113, 121, 225
weak, 56, 79, 94, 113, 120, 124,
157, 225

H 150, 167, 169, 176–83,
Hacker, Paul, 176n17, 205n51
Happé, Francesca, 196n38
Heal, Jane, 11–12, 12n8, 89,
107n21, 174n12, 198n40,
203n48, 206n52, 208n,
245n5–7, 249n14–15,
250n16, 251n20, 254n22,
255n28
Hume, David, 39, 68
I
imagination, 43, 74, 144, 222n, 240
individuation, 75, 82n5, 196, 229,
235, 238
infallibility, 114, 145, 179
inference
to the best explanation, 5, 13, 14,
55, 57, 70, 72, 73, 75, 77,
79, 83, 95, 99, 120, 139,
141, 146, 154, 157, 176,
220, 225, 227,
236, 237, 240, 241,
256n30
deductive, 257
inductive, 233, 236, 257
inferential theories, 84–8
inner observation, 79
inner promptings, 8, 14, 88, 220,
235, 238
inner sense theories, 11, 77–84, 95
insincerity, 182, 214
instrumentalism, 180, 183
intention, 1, 2, 5, 19, 28, 29n6,
31–4, 56, 57, 60, 62, 68,
69, 72, 89, 103n11, 114,
118, 126, 143, 145, 146,
Index 283
207, 210, 213, 218, 219,
240, 246, 255, 257
past, 179, 181–3, 213, 240
internalism
about content, 102, 259, 262
epistemological, 102, 109n24
rational, 99–111
irrationality, 34, 198, 254, 255n27,
260, 261, 265
irrealism, 165, 179, 181, 204, 213
J
Jacobsen, Rockney, 205n51,
206n51–2, 207n57,
208n58, 210n62
James, William, 39
judgement, 5, 9, 10, 13, 14, 28,
32n12, 34–6, 38n20,
40–3, 43n21, 44–6, 64,
73, 83, 88–91, 101n3,
104, 105n14, 107, 108,
108n23, 110, 111, 113,
124, 127, 128, 146, 153,
155n21, 156, 159, 178,
179, 190–2, 195–6, 201,
202, 202n46, 203, 203n,
204, 205, 205n49, 206,
207, 207n55, 209, 214,
220, 224–6, 233, 237,
244, 245, 247, 249,
250n17, 251, 254, 260,
261, 267
justifications, 43–5, 55, 110–12,
112n32, 121, 127, 129,
141, 157, 202n46, 251.
See also reasons
284 Index
K dispositional, 11, 14, 65, 69, 94,
Kaplan, David, 139n
knowledge
empirical, 52
implicit, 106, 107, 109
inferential, 8, 91, 99, 128
observational, 52, 99, 128
practical, 156
tacit, 170, 171
Kripke, Saul, 175, 176n17
L other, 75, 78, 85, 88, 167, 194,
Lee, Byeong, 245n5, 254n21
Levi, Isaac, 31n11
Libet, Benjamin, 68n13
Linville, Kent, 245n5, 245n7
Lycan, William, 7, 77–84
M moods, 68, 69, 139
Malcolm, Norman, 245n5
Martin, Michael, 254n21
McDowell, John, 108n22,
108n23
McHugh, Conor, 105n15, 106n16,
106n17
McKinsey, Michael, 67
meaning, 6, 31, 53, 54n, 123n46,
135–9, 139n, 141, 143–5,
147–9, 159, 160, 167, 175,
176, 177n19, 185, 207,
210, 248, 248n12
mechanism
causal, 80, 82n6, 116n
subpersonal, 116
mental states
commissive, 127, 219
96, 188, 213, 218, 232, 241
first-order, 34, 190, 219
Freudian, 60, 187, 234
judgement-sensitive, 27, 27n5, 31
occurrent, 106n18, 107, 149, 151
second-order, 11, 79–80, 83, 111,
116, 164, 166, 174, 212
Millar, Alan, 244n4, 255n29,
257n32, 259n34
mind(s)
one’s own, 2, 3, 135, 210n61, 254
195, 196n, 218, 240
minimal
assertoric content, 205n51,
206n51
truth, 205n51
Monism, 4, 119n42, 179
monistic prejudice, 3
Moore, G. E., 185, 245, 245n5
Moore’s paradox, 16, 17, 32n13,
38n19, 147, 149, 152,
184, 202n46, 211n63,
243–67
Moran, Richard, 9, 27n5, 37,
86n10, 100, 119, 122, 123,
123n45–6, 124, 124n48,
125, 125n49–50, 126–9,
244, 245n6, 247n9,
254n21, 267
N
Nisbett, Richard, 68n13
Nozick, Robert, 52n1
Nussbaum, Martha, 40

O platitudes, 138, 205n51, 247, 248, 252
object(s)
intentional, 47, 229
relational, 229, 231
objectivity, 20, 22–3
O’Brien, Lucy, 125n50, 126n52
P reasoning, 32, 89, 114, 156, 169,
Parent, Ted, 67n10
Peacocke, Christopher, 9, 10, 24n,
42, 99–111, 113, 115, 117,
118n39, 157, 169, 177n17,
191
perception(s), 1, 2, 3n2, 4–6, 9, 11,
14–16, 18–26, 38, 39,
41–7, 52, 55, 59–62, 64,
69, 70, 81, 88, 112n32,
114, 123, 127n, 129
perceptual
attributives, 23–5
content
conceptual, 24, 100
non conceptual, 108n23
experience, 14, 18, 43, 43n21, 44,
74, 109n25, 112, 128, 222,
228, 231–2, 240
representation, 21–4, 25n3
performative(s), 114, 114n36, 152,
201n45, 202, 204, 205,
205n51, 206, 206n52,
207–11, 217
function, 203, 206n53
phenomenology, 1, 2, 9, 14, 15, 47,
55, 56, 59–61, 70, 71, 80n,
85, 86, 101, 104, 107,
108n23, 109, 149, 222,
225n3, 227, 237
Index 285
pluralism
about self-knowledge, 3, 4, 13,
212, 217–42
of methods, 2
of states, 4
Povinelli, Daniel, 197, 197n
practical
171, 189, 199n, 212, 262
syllogisms, 33, 169, 171
pragmatic inconsistency, 246, 247
Pritchard, Duncan, 52n1
private language, 53, 53n6, 116
argument, 134
propositional attitude(s)
commissive (or as commitments),
31–8, 221
dispositional (or as dispositions),
13, 28, 31, 32n13, 149,
183, 187, 189n32, 208,
218, 219, 232–40
Proust, Joëlle, 85n8
Pryor, James, 108n23, 109n24
Putnam, Hilary, 67
Q
Quale (qualia), 26, 223, 225
R
rationality
thick, 18, 74, 75
thin, 10
realization
core, 173, 174, 224
total, 173
286 Index
reasons, 6, 9, 28, 30, 34–7, 40, 41,
43–5, 57n8, 59, 61, 62, 72,
95, 102, 103, 105, 108n23,
111–13, 118, 119, 123–5,
123n45–6, 125n49, 126,
140, 141, 143, 145–6, 151,
153, 156n24, 160, 168,
176–8, 181, 192, 198–202,
212, 217–19, 227, 228,
235, 238, 239, 258. See also
justifications
reliabilism, 7, 116, 156
representationalism, 86n10
Rey, Georges, 37
Ring, Merrill, 245n5, 245n7
rule-following, 175, 176n17
rules, 107, 133, 134, 144, 160, 164,
196
Ryle, Gilbert, 53n4–5, 58
S 116, 117, 118n41, 123,
Scanlon, Thomas, 27n5, 31n10
scanning mechanism, 8, 11, 79, 81,
112
Schopenhauer, Arthur, 30n
Schwitzgebel, Eric, 64
seeing-as, 42, 72, 236
Self-ascriptions
of propositional attitudes
as commitments, 31–8, 221
as dispositions, 13, 28, 31,
32n13, 149, 183, 187,
189n32, 208, 218, 219,
232–40
psychological, 2–6, 11–14, 16,
17, 18n11, 54, 55, 59,
65–8, 73, 89, 90, 99–102,
105, 109n24, 110, 112–17,
120–8, 133, 134, 138, 152,
153, 155, 157, 160, 164–6,
168, 171, 181, 193, 197,
199–205, 206n53, 207,
209, 210, 213, 214,
219–21, 234, 237, 238,
240, 242
self-blind person, 167
self-deception, 41, 200, 255
motivated, 199
self-interpretation, 29, 30, 32n14,
65, 71, 182, 196n, 209,
219, 234, 235, 237, 238
self-knowledge
constitutive account(s) of
robust, 219, 222, 240, 241
weak, 241
epistemically weak account(s) of,
8, 99, 100n1, 102, 105–7,
109, 109n25, 112, 114,
125, 127n, 129
expressivist account of, 152
first personal, 52–66
inferential model of, 125, 217
neo-expressivist account of, 192n
observational model of, 192, 193,
223, 230
third personal, 2, 4, 8, 16, 29, 51,
69–75, 95, 96, 149, 190,
219, 232–42
sensations, 1, 3n2, 4–6, 9–11,
14–16, 18–26, 38, 39,
45–7, 54–61, 63, 64, 70,
74, 83, 108n23, 115, 116,
123, 129, 134–7, 139–45,
164, 167, 174n12, 192n,
 Index 287

196n, 214, 218, 222–9, theory-theory, 7, 8, 88, 89, 91, 94

240, 241
bodily, 14, 26
sense of ownership, 227
sensory states, 21, 25, 61, 227
Shah, Nishi, 126n51
shbeliefs, 210
Shoemaker, Sydney, 6, 11, 12, 17,
57n7, 58, 80n, 84, 165,
167–74, 174n11
simulation-theories, 88–96, 235
simulative methods, 14
sincerity conditions, 205n49, 206,
207
slips of the tongue, 16, 200n, 214
Snowdon, Paul, 59, 63, 223, 224
solipsism, 53, 135
Sorensen, Roy, 245n6
Sosa, Ernest, 52n1
Spelke, Elisabeth, 25n
Stoneham, Tom, 198n41
Stone, Tony, 194n
Strawson, Peter, 184, 184n30
subpersonal
level, 22, 68, 77, 87, 88, 95, 170,
171
mechanism, 77, 116
switch of aspect, 73, 236
symptom, 46, 54, 57, 64, 71, 72, 81,
82, 154, 159, 236, 238, 253
thought(s)
cogito-like, 114, 116
self-verifying, 153
Tomasello, Michael, 197, 197n
Transparency
mehod
deliberative account, 124,
125
epistemic account, 121, 122,
124, 125
strong, 60, 61, 79, 85, 103,
124
-to-the-subject condition, 153
-to-the-world condition, 153
weak, 59
truth
minimalism about, 138, 138n2
presumption of, 154,158,, 220,
221
Turri, John, 244n4
U
unconscious, 5, 28, 29, 29n7,
32n14, 60–2, 68, 72, 95,
116, 123, 137, 142, 167,
186, 190, 198, 199, 232,
233, 238, 253–6, 257n31

V
T Varzi, Achille, 109n24
Tanney, Julia, 30n Velleman, J. David, 126n51
Tappolet, Christine, 42 veridical, 21–3, 46
Teroni, Fabrice, 40, 44 Volpe, Giorgio, 30n, 121n
testimony, 14, 35, 73, 75, 233, 237, voluntarism, 33, 34
241 Vonk, Jennifer, 197n
288 Index
W Wittgenstein, Ludwig, 10, 15, 53,
warrant. See also justifications
externalist, 112, 116
internalist, 111
Wegner, Daniel, 68n13
Wheatley, Thalia, 68n13
Williams, Donna, 196n
Williams, John, 243n, 244n, 245n6,
251n20, 254n21
Williamson, Timothy, 52n1
Wilson, Timothy, 68n13
53n6, 54n, 72n, 91, 94,
116, 133–51, 156, 160,
165, 175–7, 180, 181, 190,
192n35, 201, 202n46,
207n56, 210n62, 211n63,
212, 213, 223, 227, 245–52
Wright, Crispin, 11, 12, 12n8, 30n9,
43n21, 163, 165, 174–83,
198, 198n40–1, 204, 205n51,
208n58, 212, 213, 219