Kant's Transcendental Psychology

!"#$%&'()%#*+)*+#,*'--.%-)/+%0-'*1%
 Kant's
Transcendental
Psychology
Patricia Kitcher

OXFORD UNIVERSITY PRESS

New York Oxford
 Oxford University Press
Oxford New York Toronto
Delhi Bombay Calcutta Madras Karachi
Kuala Lumpur Singapore Hong Kong Tokyo
Nairobi Dar es Salaam Cape Town
Melbourne Auckland Madrid
and associated companies in
Berlin Ibadan
Copyright © 1990 by Patricia Kitcher
Published by Oxford University Press, Inc.
198 Madison Avenue, New York, New York 10016-4314
First issued as an Oxford University Press paperback, 1993.
Oxford is a registered trademark of Oxford University Press
All rights reserved. No part of this publication may be reproduced,
stored in a retrieval system, or transmitted, in any form or by any means,
electronic, mechanical, photocopying, recording, or otherwise,
without the prior permission of Oxford University Press.
Library of Congress Cataloging-in-Publication Data
Kitcher, Patricia.
Kant's transcendental psychology
by Patricia Kitcher.
p. cm.
ISBN 0-19-505967-0
ISBN 0-19-508563-9 (pbk)
1. Cognition.
2. Kant, Immanuel, 1724-1804—Contributions in psychology.
3. Kant, Immanuel, 1724-1804 Kritik der reinen Vernunft.
I. Title. BF311.K57 1990 128'.092—dc20
89-78463

2 4 6 8 9 7 5 3
Printed in the United States of America
on acid-free paper
For P.S.K.
!"#$%&'()%#*+)*+#,*'--.%-)/+%0-'*1%
 Preface

Many people have helped me in the course of my study of Kant and in

writing this book. At Wellesley, Ingrid Stadler spent hours taking me
through the three Critiques, more hours, I fear, than I have spent teach-
ing any undergraduate. I also benefited from Richard Rorty's Kant
seminar at Princeton. Over the years my correspondence with Gary
Hatfield has been a source of inspiration and support. Hatfield's own
work and his responses to mine have encouraged me to believe that it
is not crazy to take Kant's psychology seriously and to try to understand
it in its own context.
Three people have been kind enough to read the entire manuscript
and to offer their criticisms: Henry Allison, Philip Kitcher, and Jonathan
Vogel. It was particularly kind of Professor Vogel to send detailed
criticisms, since we only met once. Henry Allison read drafts of each
of the chapters, which we then discussed at length. Anyone writing on
Kant would be extraordinarily fortunate to have such a critic and I hope
that I have used his expertise wisely. Philip Kitcher read more drafts of
these chapters than either he or I care to remember. I am also grateful
for Jill Buroker's comments on Chapters 2 and 3, many of which led
me to make substantial revisions. While we were colleagues at Min-
nesota, John Earman and I had several conversations about Leibniz.
More recently, I have been able to discuss Kant's predecessors with
Nicholas Jolley. Although I have learned from both, neither of these
fine scholars should be blamed for any mistakes in my interpretations
of Leibniz or Hume. I would also like to thank four research assistants,
Randy Wojtowicz, Warren Dow, and Valerie and Gary Hardcastle, for
their careful and prompt help, and the Committee on Research at the
University of California, San Diego, for funding them. I am particularly
viii PREFACE

grateful to the Hardcastles for organizing the bibliography and the

indices.
A book project incurs personal debts as well. I am grateful to my
father for spending many lunch hours in Sterling Memorial Library at
Yale hunting for obscure German books, and to both my parents for
supporting my work long before it was fashionable to have working
daughters. In the last two years of intensive work on the book, my sons,
Andrew and Charles, have never complained about my not being able
to do things for them; they only expressed the wish that I did not have
to work so hard. This book could not have been written without the
enormous patience and loving support of my husband, Philip Kitcher.
For a year he met the children virtually every day after school so that
I could have enough time to complete it.
Finally I would like to thank the editors of the Philosophical Review
for permitting me to use materials from several articles. Although I
rewrote the chapters completely, many of the ideas and arguments from
"Discovering the Forms of Intuition" (Philosophical Review XCVI,
1987: 205-248) and "Kant's Paralogisms" (Philosophical Review XCI,
1982: 515-547) recur in Chapters 2 and 7. The basic idea for Chapter
4, and indeed for the book, was presented in "Kant on Self-Identity"
(PhilosophicalReview XCI, 1982:41-70), although I have changed many
of the arguments as my ideas have changed. I am also grateful to J.-C.
Smith and to D. Reidel for permission to use material from section III
of "Kant's Dedicated Cognitivist System" (in J.-C. Smith, ed., Historical
Foundations of Cognitive Science, Dordrecht, Holland: D. Reidel, 1989,
pp. 189-209).
A Note on References: In the notes I give the reference for citations,
other than those from the Critique of Pure Reason, in the Akademie
Edition (Kant's gesammelte Schriften, edited by the Koniglichen Preus-
sischen Akademie der Wissenschaften, 29 vols., Berlin: Walter de Gruy-
ter and predecessors, 1902-), by citing the volume number and pages
after "AA," as well as the standard English source. For the Critique, I
use the standard "A" and "B" citations from Kemp Smith (Norman
Kemp Smith, trans., Immanuel Kant's Critique of Pure Reason, New
York: St. Martin's, 1968) in the text.

La Jolla, California P.W.K.

February 1990
 Contents

1. What Is Transcendental Psychology?, 3

The "Dark Side" of the Critique, 3
Countercurrents: Reinhold to Austin, 5
Kant Against "Psychology," 11
Transcendental Psychology, 14
In Defense of Transcendental Psychology, 21

2. The Science of Sensibility, 30

What the Transcendental Aesthetic Is About, 30
Early Modern Theories of Spatial Perception, 32
Kant's Analysis of Spatial Perception, 35
Intuition, Matter, and Form, 35
Pure Forms (and Nativism), 37
The Method of Isolation, 39
Distance, Extent, and Shape, 40
Touch: Leibniz Versus Berkeley, 41
Kant's Empirical Assumptions, 43
The Isolation Argument, 44
Two Arguments of the Metaphysical Exposition, 45
The Standard View, 45
The First Argument, 46
The Second Argument, 48
The Transcendental Exposition, 49
The Role of Geometry, 49
Geometry and the Space of Perception, 50
Parsons's Interpretation, 53
Kant's Results, 54
The Forms of Intuition and Contemporary Evidence, 55
Depth Perception, 55
Is the Space of Perception Euclidean?, 56
X CONTENTS

The Assumption of a Common "Outer Sense," 57

Is Spatial Perception Determinate?, 58
Discovering the Forms of Intuition, 59

3. Transcendental Psychology in the Transcendental

Deduction, 61
What (If Anything) Happens in the Deduction Chapter?, 61
Representing Objects, 65
The Problem, 65
The Views of Leibniz and Christian Wolff, 67
Empiricism and Sensationism, 67
Associationism, 69
A Priori Necessary Synthesis, 70
What Is an Object of Representations?, 70
Unity and Synthesis, 73
What Is Synthesis?, 74
The Law of Association, 77
Associationism and Apriority, 79
Representations and Concepts, 80
Kant's Defensible Results, 81
Synthesis and the Productive Imagination, 81
Robert Paul Wolff on Rules of Synthesis, 82
The "Problem" of Early Cognition, 83
Constructing Representations of Objects and the "Binding" Problem, 84
Making Judgments About Objects, 86
The Problem of Judgment, 86
The Synthesis of Intuitions, 88
The "One-Step" Deduction, 89
Constructing Judgments, 89
The Objective and Subjective Sides of the Deduction, 90

4. Replying to Hume's Heap, 91

Troubles with Apperception, 91
Avoiding the Subjective Deduction, 91
Apperception as the Cogito, 91
Strawson and the Self-Ascription Reading, 92
The "Logical" Reading of Apperception, 94
Two Mistaken Assumptions, 95
Hume, 95
Hume's Problem, 97
 Contents xi
Hume's Absence, 97
Kant's Knowledge of Hume's Position, 98
The Denial of Real Connection, 100
Synthesis and Apperception, 102
Connecting Cognitive States by Synthesis, 102
Transcendental Synthesis, 103
Apperception and Transcendental Synthesis, 104
Apperception, 105
Arguing for the Synthetic Unity of Apperception, 108
Apperception and Representation, 108
Judgments, 110
Kant's Functionalism, 111
Intuitions, 113
The Reply to Hume. 114

5. A Cognitive Criterion of Mental Unity, 117

Unity of Apperception as Mental Unity, 117
Synthetic Connection, 117
Connection and Connectibility, 118
The Plan of the Chapter, 120
Refining the Account of Synthetic Connection, 121
Is the Self the Combiner?, 122
Locke and Leibniz on Personal Identity, 123
The Issue, 123
Leibniz Versus Locke, 123
Moral Responsibility, 125
The Problem of Self-Consciousness, 126
Modern Mentalism, Wiggins, and Parfit, 128
Modern Mentalism, 128
Wiggins's Argument Against Mentalism, 130
Parfit's Denial of Personal Identity, 131
Objections Considered, 133
Is the Cognitive Criterion Too Weak?, 133
Is It Too Strong?, 134
Is It Too A Priori?, 135
Modularity, 137
Summary of the Account, 138
Apperception and Kant's System, 139
Too Many Selves, 139
The Ideality of Time, 140
xii CONTENTS

6. Perceiving Times and Spaces: The Cognitive Capacity

at the Center of the Deduction, 142
Cognitive Tasks, Apperception, and the Deduction of the
Categories, 142
Perception: The Eighteenth-Century Background, 147
The Standard View, 147
Intellectual Theories of Perception, 148
The Synthesis of Apprehension in A, 148
A99, 148
A119-20, 149
Examples from the Politz Lectures, 150
The Case for the Synthesis of Apprehension, 151
A99 Revisited, 152
A Role for Concepts in Perception in A, 153
The Need for Nonreproductive Synthesis, 153
Perceptual Recognition, 153
Concept Application, 153
26 in the B Deduction, 155
The Centrality of 26, 155
Perception as "Scanning an Image," 156
Perceiving Times and Spaces, 157
Differences Between the Editions, 158
P-Functions as Spatial and Temporal C-Functions, 160
Perceiving Objects by Perceiving Spatial and Temporal Arrays, 161
Kant's Long Argument, 162
Additional Considerations, 162
The Basic Argument, 162
Universal Applicability and Objective Validity, 163
The Argument from Apperception, 166
How the Argument Fails, 167
Defending the Long Argument Interpretation, 169
Some Advantages, 169
Henrich's Antipsychological Reading, 170
Allison's Apsychological Reading, 171
The Loss of Generality, 173
26 as Completing the Argument of the Metaphysical Deduction, 173
How Serious Is the Loss of Generality?, 174
Transcendental Psychology in the Second Analogy, 174
Guyer's Interpretation, 174
Guyer's Objection to a Psychological Reading, 177
 Contents xiii
Versus Guyer's Antipsychologism, 177
What Kant Has Shown, 178

7. The Limits of Transcendental Psychology, 181

Kant's Paralogisms, 181
Puzzles of the First Paralogism, 183
Understanding the First Paralogism, 187
Identity Through Time, 195
Leibniz and the Simplicity of the Soul, 198

8. Cognitive Constraints on Empirical Concepts, 205

Kant and Cognitive Science, 205
Do We Employ Necessary and Sufficient Conditions?, 207
Difficulties with the Classical View, 207
Kant on Concepts and Concept Application, 209
Empirical Warrant and the Open-Ended Character of Experience, 210
When Should We Codify Our Concepts?, 211
Implications for Necessary and Sufficient Conditions, 212
Further Implications, 213
Empiricism and "Original Sim," 214
Quinean Empiricism, 214
Current Directions, 216
Concepts and Reasoning, 217
The Task of Inference, 217
Analytic and Synthetic Approaches to Concepts, 217
How to Carve Nature at the Joints, 219
The Structure of Concepts, 221
Examples of Dependency Relations, 222
Implementing the Demand of Reason, 225
Theoretical and Experimental Implications, 227
Conclusion, 230

Notes, 231
Bibliography, 273
Index of Cited Passages, 283
General Index, 289
!"#$%&'()%#*+)*+#,*'--.%-)/+%0-'*1%
Kant's Transcendental Psychology
!"#$%&'()%#*+)*+#,*'--.%-)/+%0-'*1%
 1
What Is Transcendental
Psychology?

The "Dark Side" of the Critique

But it [the deduction of the categories] is also an essay in the imaginary

subject of transcendental psychology. . . . The theory . . . , like any essay
in transcendental psychology, is exposed to the ad hominem objection
that we can gain no empirical knowledge of its truth. . . .
Yet there is no doubt that this doctrine [about our cognitive faculties] is
incoherent in itself and masks, rather than explains, the real character of
[Kant's] inquiry. . . .1
P. F. Strawson opens his important essay on Kant's Critique of Pure
Reason with these brief denunciations of transcendental psychology in
order to justify a particular interpretive strategy. Although he is forth-
right that "[t]he idiom of the work is throughout a psychological idiom,"2
he will not read the book psychologically, but as an analytical argument.
He describes this strategy as one of "disentangling"3 the psychological
and the analytical sides of the Critique. In the discussions of the Aes-
thetic, Analytic, and Dialectic that follow, however, there is little di-
sentangling. Psychological topics are raised only to be dismissed. His
treatment of synthesis is typical: The hope is to "bypass" this doctrine
"altogether."4 In practice, the Bounds of Sense either ignores Kant's
frequent appeals to psychological processes and faculties or administers
a quick scolding about alleged lapses from philosophical sanity. The
many studies spawned by this deservedly influential book follow its
interpretive model. Such widespread acceptance disguises the radical
character of Strawson's approach to the Critique. Even a superficial
3
4 KANT'S TRANSCENDENTAL PSYCHOLOGY

reading of the text, particularly the central deduction of the categories,

reveals that, if interpreters do excise or ignore all the discussions
of cognitive processes and powers, then they will have very little left
to read.
Among twentieth-century scholars, Strawson's attitude toward tran-
scendental psychology is distinguished only by its explicitness and elo-
quence. Contempt for this aspect of the Critique has been almost
universal.5 In Kant's Analytic, whose simultaneous appearance with the
Bounds of Sense produced a revival of interest in Kant, Jonathan Ben-
nett also recoils from hints of "psychologism." So he "flatters" Kant by
crediting him with a Wittgensteinean, rather than a psychological, view
of concepts and of synthesis.6 Two recent books, which stand to inherit
the mantle of Bennett and Strawson as dominant influences on Anglo-
American Kant studies, continue the pattern. Although Henry Allison
and Paul Guyer offer dramatically opposed interpretations and evalu-
ations of the Critique, they are united on this point. Guyer refers to
Kant's "bouts of transcendental psychology"; and Allison labors to de-
psychologize central doctrines.7 Taking Dieter Henrich and Gerold
Prauss as representative, transcendental psychology has the same unen-
viable place in German scholarship that it has in the English-speaking
academy.8 Avoid it!
Norman Kemp Smith complained in 1918 that some scholars were
ignoring Kant's psychological doctrines: "No interpretation which ig-
nores or under-estimates this psychological or subjective aspect of
[Kant's] teaching can be admitted as adequate."9 Unfortunately, his
account of transcendental psychology made the doctrine so unattractive
that he undoubtedly lost rather than gained converts to its importance.
On his reading, Kant took the processes that contributed to cognition
to be unconscious and hypothetical, not attributable to the empirical
self but not rightly attributed to the noumenal self either. These were
his results after promising the reader that he would try, to the best of
his ability, to make sense of transcendental psychology, despite the
delicacy and difficulty of the task.10 Others must have been daunted by
this interpretive fiasco, but H. J. Paton took up his familiar role of
defending Kant against Kemp Smith's interpretations. The activities
described by transcendental psychology are not mysterious, but con-
scious; further, we may slip between the horns of the empirical-
noumenal dilemma by characterizing these activities not as psycholog-
ical, but as logical.11 Although Paton may be closer to a solution than
Kemp Smith, as W. H. Walsh points out, there is still a large problem:
 What Is Transcendental Psychology? 5
How can something be part of an actual process and yet be merely
logical?12
Since I propose to examine transcendental psychology, I need to dispel
the initial reaction that the topic is just not worthy. Scholars do not
avert their eyes from this side of the Critique merely because Strawson
has a wonderful way of characterizing positions he opposes or because
Kemp Smith and Paton failed to make much sense of Kant's doctrine.
(They left other work for later interpreters.) Powerful currents within
and without Kant scholarship have combined to keep transcendental
psychology out of the mainstream, beyond the pale of serious philo-
sophical discussion. The first task of this chapter is to chart those cur-
rents. From the metacritical efforts of the late eighteenth century to
contemporary behaviorism, I will argue that, although certain intellec-
tual trends have been influential in relegating transcendental psychology
to a lowly status, none of them actually furnishes cogent objections
to it. Rebutting objections is hardly sufficient to reopen this area of in-
quiry, however. I will also provide a positive account that reveals tran-
scendental psychology as a sane and even illuminating approach to
philosophy.

Countercurrents: Reinhold to Austin

From the beginning, readers recognized the psychological side of the

Critique. And, from the beginning, they were puzzled over the status
of its psychological claims. Karl Leonhard Reinhold (1758-1823), whose
Briefe tiber die kantische Philosophie had been instrumental in popu-
larizing Kant's views, concluded that the Critique was insufficiently crit-
ical in its use of key psychological concepts.13 He proposed to remedy
this defect by providing a critical foundation for the critical philosophy,
a metacritique.14 Anchoring Kant's assumed but insufficiently examined
psychological concepts, particularly the ubiquitous "representation"
(Vorstellung),15 required finding a self-evident first principle from which
they could be deduced. Since this principle could not be a formula, a
concept, or a definition, Reinhold inferred that it could only be a de-
scription of a self-revealing fact, yet one general enough to permit the
deduction of all the psychological machinery of the Critique. Descartes
supplied the fact: Only the fact of consciousness can survive skeptical
doubt. To make this starting place truly presuppositionless, no psycho-
logical or metaphysical theories can be assumed. This leaves only a
6 KANT'S TRANSCENDENTAL PSYCHOLOGY
description, or phenomenology, of what appears in consciousness to be
the first fact of philosophy.16
Reinhold convinced his student, Jakob Friedrich Fries (1773-1843),
that the Critique could not stand on its own. Fries's rescue effort pro-
ceeded in a somewhat different direction, however. He took the key to
lie in a proper understanding of the methodology of the Critique, which,
unfortunately, its author did not posses. What the Critique actually
presents are regressive arguments from empirical facts of consciousness
to various forms as their preconditions. Its arguments are a posteriori,
even though its conclusions are about a priori faculties. The facul-
ties are there all along, but we only learn about them through self-
observation, hence experience. Thus, Kant's protestations to the con-
trary, his philosophy rests on psychology. Worse still, it rests on casual
psychology. Fries's new critique would correct this defect by beginning
with a careful self-observation of mental processes.17
Reinhold's and Fries's efforts to vindicate Kant's psychology only
succeed in making the issue seem intractable. They raise what appear
to be very serious problems: What justifies introducing particular fac-
ulties? Mustn't arguments to the preconditions of experience presuppose
empirical facts and thus be a posteriori? But their metacritical solutions
are un-Kantian and profoundly unsatisfactory. Reinhold insists on a
Cartesian starting point, while Fries opts for premises based on self-
observation. To anyone who has appreciated Kant's insights into the
errors of Cartesianism, however, these suggestions are hopeless. Kant
never tires of repeating the point that we have no special access to our
minds that could furnish an invulnerable starting point for philosophy.18
In the Anthropology, he criticizes introspection as unreliable, unstable,
unnatural, and a potential route to lunacy!19 Reinhold and Fries offer
fairly direct psychological readings of Kant. However, both maintain
that the integrity of Kant's psychology must be underwritten by some
type of introspection. As the foibles of introspection were demonstrated
again and again during the nineteenth and twentieth centuries, this
approach would become less and less attractive. I shall argue that the
legitimacy of appealing to certain faculties and of using factual premisses
can be defended by less dubious means.
Despite these troubles, psychological readings of the Critique were
prevalent for the first 100 years of its existence.20 Reviewing the literature
on Kant's psychology in 1870, Jiirgen Bona Meyer set the stage by
quoting Kuno Fischer: "The question of whether the critique of reason
is supposed to be metaphysical or anthropological is a real problem,
unavoidable in the history of the development of German philosophy
 What Is Transcendental Psychology? 7

since Kant."21 Later, Bona Meyer staked out a familiar (but untenable)
position. Clearly, Kant deliberately used psychological foundations, so
he must agree that the a priori can be known through direct self-
observation.22
At this point two works appeared that enabled later interpreters to
avoid Kuno Fischer's fundamental question, by eliminating the psycho-
logical [anthropological] alternative. As with Reinhold and Fries, the
blows were delivered by men who described themselves as committed
Kantians. A few years before Bona Meyer's study, Hermann von Helm-
holtz completed his massive Physiological Optics (1856-1866). Its third
and final volume contained a sustained discussion of Kant. Like most
psychologists, Helmholtz took Kant to be a nativist about spatial per-
ception.23 Although he did not believe that there was decisive evidence
in favor of the empiricist position, he criticized nativism as unnecessarily
complex and nonexplanatory. He also objected to Kant's theories of
space and geometry, citing Riemann's work in geometry and Eugenio
Beltrami's descriptions of what it would be like to walk through a non-
Euclidean space.24 With these attacks, the best known of Kant's psy-
chological doctrines, the theory that Euclidean space is the form of outer
perception, was subjected to systematic and seemingly devastating crit-
icisms by one of the major scientists of the latter half of the nineteenth
century. The revolutionary developments in physics that followed ap-
parently sealed the case.
Kant's metaphysical claims about the forms of intuition and the overall
importance of the doctrine were certainly undercut by the discovery that
physical space is not Euclidean. Nevertheless, the psychological aspect
of Kant's position could still be true, as others have noted.25 He could
be right that Euclidean space is the form of human perception. As
Helmholtz candidly admitted, he could not make a conclusive case
against Kant's position. I return to the question of spatial perception in
Chapter 2. Further, whatever their associative links, Kant's other psy-
chological doctrines are relatively independent of the physics of space
and time.
Gottlob Frege administered the decisive blow to Kant's psychology.
In the Foundations of Arithmetic, Frege was explicit about the principles
to be used in conducting his [or this] work. His well-known first principle
is "always to separate sharply the psychological from the logical, the
subjective from the objective.... "26 Even more bluntly, he declared
that "psychology should not imagine that it can contribute anything
whatever to the foundations of arithmetic."27 Early readers had accused
Kant of psychologism. Reinhold complained that the entire epistemo-
8 KANT'S TRANSCENDENTAL PSYCHOLOGY
logical tradition had confused the strictly logical question about the
conditions of knowledge with the metaphysical question about the sub-
ject of knowledge.28 Johann Friedrich Herbart, who was at Jena with
Fries, took Kant's great failing to be the attempt to found philosophy
on psychology: "logic as a whole is an ethics of thought, but not a natural
history of the understanding."29 It was Frege, however, who made the
sin of psychologism unforgivable in a serious philosopher.
Others have demonstrated the dominant influence that Frege's con-
ception of philosophical investigations has exercised on the subsequent
development of philosophy.30 I only repeat the highlights, and then
consider two episodes that appear to be of particular relevance to Kant
scholarship. Before the twentieth century, most epistemology was psy-
chologistic. As Alvin Goldman notes, the historical literature is "replete
with descriptions and classifications of mental faculties and endowments,
processes and contents, acts and operations."31 Frege attacked psy-
chologism directly only in philosophy of mathematics and in logic. He
made these areas central to philosophy, however, through his influence
on Russell, Wittgenstein, and Carnap. The three founders of analytic
philosophy established logic, philosophy of mathematics—and Fregean
philosophy of language—as paradigms of the philosophical enterprise.32
This model has been so dominant that even recent philosophy of mind
has been largely antipsychologistic!33
Frege's influence on the course of Kant scholarship extends beyond
the general direction that he provided for analytic philosophy. In 1894
he wrote a stinging review of Edmund Husserl's recently published
Philosophic der Arithmetik: "When reading this book I came to rec-
ognize the devastations which have been brought about by the incursion
of psychology into logic "34 Duly chastised, Husserl turned from
psychologism with a vengeance. The Logical Investigations devotes sev-
eral chapters to cataloging its evils. Beyond the familiar criticism of
confusing the normative with the empirical, psychologism would make
the laws of logic merely probable and contingent, rather than neces-
sary.35 Like Strawson, Husserl sees two sides to Kant's philosophy, one
of which must be dismissed: "For even a transcendental psychology also
is a psychology."36 Thus, Frege's influence also appears to be responsible
for the strong antipsychologistic stance of phenomenology,37 which has
been a major force in the development of Continental philosophy. Kant
scholars trained in this tradition would be just as wary of psychologism
as their analytic counterparts—and equally aware that Kant had trans-
gressed the proper bounds of philosophy. I end this brief account of
Frege's role in the fall of Kant's psychology with a footnote. J. L. Austin
 What Is Transcendental Psychology? 9

was a towering figure in British philosophy. When his translation of Die

Grundlagen der Arithmetik appeared in 1950, it produced an important
revival of interest in Frege among the next generation of philosophers.38
One prominent member of that group was Strawson.
What is psychologism? I have postponed this essential question, be-
cause the answer is complex. "Psychologism" is a blanket condemnation
for importantly different types of arguments. In its central usage, "psy-
chologism" refers to the fallacy of trying to base normative principles
on factual premises. As in the citation from Herbart, it is often presented
by analogy with ethics (and later with the "naturalistic fallacy" in ethics).
At one extreme, there is what might be called "strong" psychologism
in logic: the attempt to establish the validity of logical principles by
appeal to facts of human psychology. Although such a move is clearly
fallacious, I find no evidence of this kind of psychologism in Kant. The
principles defended by appeal to transcendental psychology—most fa-
mously, "All alterations take place in conformity with the law of the
connection of cause and effect"—are not logical, but metaphysical.39
Reinhold's critique amounts to a different charge of psychologism.
Kant tries to extract conclusions that are necessary and certain from
psychological premises. As I argue later, however, although he fre-
quently describes his results as "certain," he really means that the rea-
soning is certain, and so are the conclusions—given the presumption of
experience for creatures with our mental capacities.40 Others have noted
that Kantian necessity is also relativized to worlds that could be expe-
rienced by creatures with our mental capacities.41 Thus, although the
arguments invoke psychological factors, the conclusions contain implicit
psychological qualifications, so there is no fallacy. Later, I argue a re-
lated point for the crucial notion of apriority.
Frege denied any relevance of psychology to philosophy of mathe-
matics or to logic. "Weak psychologism" is the view that psychological
facts may be important to philosophical or normative claims, even
though they cannot establish such claims. In logic, even weak psychol-
ogism seems inappropriate. Given Frege's influence, however, weak
psychologism was also banished from the rest of philosophy. This move
now appears extreme. For how can we hope to understand the nature
of thought or the limits of knowledge—or to prescribe methods for
improving our reasoning practices—without having some understanding
of the capacities that make cognition possible?
Kant's epistemology is clearly weakly psychologistic. It does not, for
that reason, rest on a fallacy of confusing the normative with the factual.
Goldman observes that it is only against the background of philosophy's
10 KANT'S TRANSCENDENTAL PSYCHOLOGY

recent past that it is even surprising to hold that epistemologists should

be concerned with the strengths and weaknesses of the cognitive pro-
cesses that underlie our capacity for knowledge.42 Since Kant's project
was to determine our ability to have certain types of knowledge, he
naturally began by considering the mental equipment that we standardly
have. Despite the venerable and almost reflexive character of the psy-
chologism charge against transcendental psychology, Kant is only guilty
of weak psychologism. And that is not automatically a sin. In later
discussions of specific topics, I argue for the relevance of claims of
transcendental psychology to philosophical issues.
Finally, the ideology of twentieth-century psychology has had highly
negative implications for the status of transcendental psychology. As-
suming that introspection was the only way to study mental processes,
J. B. Watson and other behaviorists convinced their colleagues that they
could write a psychology and "never use the terms consciousness, mental
state, mind, content,... imagery, and the like."43 If the self-conscious
methodology of behaviorism is granted, then Kant erred even in trying
to talk about cognitive processes. Further, his claims were often cast in
terms of "faculties," and faculty psychology has long been regarded as
bogus by psychologists. Given the views of their colleagues in psychol-
ogy, most Kant scholars concluded that transcendental psychology had
no value for philosophers, or for anybody else. W. H. Walsh is a re-
freshing exception. Refusing to join the chorus of detractors, he made
the sensible observation that a "faculty" is simply a short way of referring
to a set of powers or capacities, and hence perfectly legitimate.44 Fac-
ulties are still officially banned from psychology, but the journals contain
constant references to short-term memory systems, phonemic proces-
sors, face recognition modules, and the like. The rise in the fortunes of
cognitive psychology (and more recently cognitive science) has been
even more dramatic. Psychologists have realized that they cannot even
explain behavior without appealing to the cognitive processes that lie
behind it—and that such processes can be studied without resorting to
introspection.
Transcendental psychology runs counter to the empiricism, behav-
iorism, and most importantly, antipsychologism that have permeated
recent philosophy. As a result, it has been accused of resting on in-
trospection, being refuted by empiricist psychology, committing the fal-
lacy of psychologism, and, generally, depending on a methodology that
fails to meet the standards of either philosophy or psychology. In this
intellectual climate it is no wonder that Kant scholars believe that their
contributions to philosophy and its history will be greater, the less that
 What Is Transcendental Psychology? 11

they say about this doctrine. Individually, however, none of these

charges stands up, so their collective weight should not continue to block
serious discussion of transcendental psychology—as it has for the last
100 years.
As historians became convinced of the errors of transcendental psy-
chology, they tried to divorce the Critical philosophy from it. Granted
that there were many references to psychological faculties and processes;
there were also negative comments about the value of psychology. These
have been given prominence in order to plead the case that Kant saw,
at least dimly, that psychology could not play the role that he appeared
to give it in his own system. Before embarking on a study of transcen-
dental psychology, I will consider a final objection: Given Kant's own
reservations about psychology, this material should be dismissed.

Kant Against "Psychology"

In a sense, Kant's attitude toward psychology is clear. He claims that
Empirical Psychology is unable to contribute to a priori knowledge
(B152) and ought to be banished from metaphysics (A848/B876). He
devotes an entire chapter to deflating the pretensions of Rational Psy-
chology. In a well-known passage in the Metaphysical Foundations of
Natural Science, he argues that Empirical Psychology can never be a
proper (eigentlich) natural science, because it is not quantitative.45 He
continues by noting additional defects. Psychology cannot really be a
good experimental science either, because it is not possible to isolate
different thoughts. Further, observation through inner sense or intro-
spection alters the state of the object observed.46 (In the Anthropology,
he recommends external behavior as a better source of evidence for
anthropology.47)
Besides attacking psychology in general, Kant appears to undercut
his own psychological claims. In a frequently cited passage from the first
preface, he confesses that the subjective side of the deduction, which
is concerned with cognitive faculties of the understanding, is "somewhat
hypothetical in character" and inessential to his chief purpose (Aviii).48
Since references to faculties are muted in the second edition, these
hedges are often thought to indicate a recognition that he had erred in
casting his arguments in psychological form. He also makes several
remarks that seem to anticipate the antipsychologism of later philoso-
phy. Thus, he reminds logicians to observe the rule that "pure logic...
does not, as has sometimes been supposed, borrow anything from [em-
12 KANT'S TRANSCENDENTAL PSYCHOLOGY

pirical] psychology...[!]" (A54/B78). In explaining the novel aims and

procedures of a transcendental deduction, he draws a contrast with
Locke's efforts to discover the psychological origins of various concepts
(A86/B118-19). He characterizes the field of his study as "transcen-
dental logic."
Together these pieces seem to create a convincing picture of disillu-
sionment with transcendental psychology. That picture is badly out of
focus, however. We cannot understand any of Kant's comments on
psychology until we know how he saw the contrasting fields of psy-
chology and logic. The Anthropology offers a clear, but to modern
readers, shocking, answer to this question.
The lower cognitive power is characterized by the passivity of the inner
sense of sensations; the higher by the spontaneity of apperception—that
is, of pure consciousness of the activity that constitutes thinking—and
belongs to logic (a system of the rules of the understanding), just as the
former belongs to psychology (to a sum-total of all inner perceptions under
laws of nature) and establishes inner experience.49

Taking psychology first, its job is to seek laws governing perceptions

observed through introspection. Kant's usage closely follows the defi-
nition given by Christian Wolff in the prolegomena to Psychologia em-
pirica: "we experience that of which we are aware (cognoscimus) by
attending to our perceptions. Hence we come to know the subjects dealt
with In empirical psychology by attending to those occurences in our souls
of which we are conscious."50 Given Empirical Psychology's calling, to
search for regularities among items revealed to introspection, we can
appreciate Kant's dismal and prescient appraisal of the enterprise.
Wolff also championed the demonstrative science of Rational Psy-
chology. Rational Psychology begins with the principles of Empirical
Psychology, and then "gives the reason for whatever actually occurs in
the soul or can occur in it." Alternatively, through analysis and dem-
onstration, Rational Psychology deduces the essential properties of the
soul that contain the sufficient reason for what occurs in it.51 Although
Rational Psychology can fall into error—by giving a spurious reason—
Wolff did not think that this raised any practical difficulties. The dem-
onstration of why something occurs in the soul would be false, but the
fact of its occurrence would still be attested to by Empirical Psychology.52
Kant's opposition to Rational Psychology is well known and well ar-
gued.53 Although he offers specific criticisms of associationist psychology
(see Chapter 354), his general denunciations of psychology are aimed at
the two disciplines just presented.
 What Is Transcendental Psychology? 13
If psychologists would be surprised by the strange and cramped pur-
suits that Kant places under their banner, logicians would be equally dis-
mayed by his largesse about logic. The Anthropology's account of logic as
a system of the rules of the understanding is no slip of the pen. He begins
his logic lectures by explaining that "the understanding... is bound in its
acts to rules we can investigate."55 The topics in these lectures are even
more revealing: presentations, consciousness, empirical concepts as
springing from the senses, comparison, and abstraction.56 As Hatfield
observes, seventeenth- and eighteenth-century authors do not discuss
these matters to provide psychological backing for principles of formal
logic. They take the intellect or understanding to be the logical or ep-
istemic power, and their goal is to study that power itself.57
Kant's work is plainly within this tradition. At the beginning, he
explains that he is not offering a "critique of books and systems, but of
the faculty of reason in general, in respect of all knowledge after which
it may strive independently of all experience" (Axii). He begins the
Transcendental Logic by noting the two fundamental sources of knowl-
edge in our minds (A50/B74). He contrasts transcendental logic with
general logic: the former "also would go to the origin of cognitions of
objects, in so far as that origin cannot be attributed to objects" (A55-
56/B80, amended translation). Although his conclusions encompass
many issues, the central argumentative project of the Critique is the
examination of cognitive faculties—sensibility, understanding, imagi-
nation, and reason—to determine which aspects of our knowledge derive
from them, rather than from objects. None of his remarks about what
he called "psychology" disturbs that fundamental interpretive anchor.
What of Kant's prefatory retreat from the subjective side of the de-
duction? The passage is quite ambivalent:
[It is not essential, for]... the chief question is simply... what and how
much can the understanding and reason know apart from all experience?
not:—how is the faculty of thought itself possible? The latter is, as it were,
the search for the cause of a give effect, and to that extent is somewhat
hypothetical in character (though, as I shall show elsewhere, it is not really
so)... [Axvii].

Later I argue that transcendental psychology analyzes cognitive tasks

to determine the general specifications for a mind capable of performing
those tasks. That is how Kant is going to show that certain aspects of
our knowledge are grounded in our faculties: by showing that any faculty
that can perform the task at all must meet certain specifications and
that the knowledge produced by a faculty with those specifications will
14 KANT'S TRANSCENDENTAL PSYCHOLOGY
always include certain elements. Thus, highly abstract descriptions of
faculties are essential to his project.
His ambivalence has two sources. First, he is concerned that in his
efforts to provide abstract descriptions of features that minds must have
to be capable of knowledge, he may have fallen into giving explanations
of how the mind has those features. Faculty terminology makes con-
fusion between these importantly different projects quite easy. For a
faculty name could refer to the actual cause of certain mental processes
(The Memory), or it could just indicate the type of process required (a
memory, which might be produced by ten separate mechanisms). His
second worry is more provincial. Rational Psychology tried to find the
sufficient reason for various aspects of mental life, and Kant does not
want his project to be confused with what he regards as a disreputable
enterprise.58 His subsequent reference to offering opinions, where other
opinions are equally possible (Axvii), is probably an oblique criticism
of the Wolffians.59
Kant's reservations about the subjective deduction do not show that
he came to doubt transcendental psychology. His problem is how to
express his novel ideas without seeming to be offering hypothetical
causes. In the second edition the arguments are presented in a different
form, with fewer references to distinct faculties, and greater emphasis
on the types of processes required. And there is no apology.

Transcendental Psychology

Kant introduces a new, transcendental method into philosophy. He

explains:
I call all knowledge transcendental which is not concerned so much with
objects, but with our manner of knowing objects, insofar as that should
be possible a priori [A11-12/B25, amended translation].

what can alone be entitled transcendental is the knowledge that these

representations are not of empirical origin . . . [A56/B80-81].

Such [transcendental] proof does not show that the given concept... leads
directly to another concept.... The proof proceeds by showing that ex-
 What Is Transcendental Psychology? 15
perience itself, and therefore the object of experience, would be impos-
sible without a connection of this kind [A783/B811].
Putting these together, transcendental knowledge concerns how we know
objects. Its specific province comprises those features of cognition that
can be traced to a priori origins. Transcendental knowledge is established
by transcendental proofs, which show that experience would be impossi-
ble but for the a priori origins of certain features of our cognition.
The focus on a priori origins is clear in the text. Nevertheless it is
often missed. Besides a general antipathy to any hint of psychology,
scholars are thrown off by Kant's insistence that accounts of the empirical
origins of concepts or beliefs are without philosophical interest (A86/
B118-19). They infer that the general question of origins must also be
irrelevant. Even in the crucial passage where he contrasts his method
with Locke's empirical method, and a transcendental deduction with a
questio facti, however, he ties his project to origins. A deduction must
provide a "certificate of birth quite other than that of descent from
experiences" (A86-87/B119). Many other passages and an entire ap-
pendix, The Amphiboly, stress the importance of origins to transcen-
dental philosophy (e.g., A44/B61-62).
If empirical origins are irrelevant to philosophy, why are a priori
origins a central concern? Kant is explicit about the overarching goal
of his book. He wants to show that we can have knowledge of synthetic
a priori propositions. A priori origins are critical, because of their role
in the justification of a priori knowledge. At this point, we must come
to grips with the varying uses of "a priori" in the Critique.
"A priori" has three primary senses for Kant. A claim or judgment
is "a priori^" if it has a particular logical form: It is universal and
necessary. Kant presents these as mere marks of apriority (B3-4), but
it is simpler to take this to be one sense of the term. It is this sense of
"a prioriL" that is used in setting up the problem. He wants to defend
our claims to know propositions that are universal and necessary.
The second sense connects with Kant's interest in origins: A propo-
sition or concept is a priori0 if it includes elements that do not derive
from sensations (Bl). This may seem a needlessly convoluted way of
saying that these items are innate. Even this sense of "apriori0" should
not be confused with "innate," however. Kant clarifies his position in
the polemic against Eberhard:
The Critique admits absolutely no divinely implanted or innate represen-
tations ... There must, however, be a ground in the subject which makes
16 KANT'S TRANSCENDENTAL PSYCHOLOGY
it possible for these representations to originate in this and no other
manner This ground is at least innate.... w l

Theorists err in moving from evidence about nonempirical elements to

claims that beliefs, or concepts, or mechanisms61 are innate. For these
are disguised and unjustified hypotheses about the causes of nonempir-
ical elements in cognition. The expression "a priori0" enables Kant to
state only what is known: The element does not derive from the senses.
Kant's third and most complex sense of "a priori" is epistemological.
He provides an explicit definition. "In what follows ... we shall under-
stand by a priori^ knowledge, not knowledge independent of this or that
experience, but knowledge absolutely independent of all experience."
(B2-3). This definition raises two immediate problems. The memorable
opening sentence of the Introduction proclaims that all our knowledge
begins with experience. If so, then how can any knowledge be completely
independent of experience? Further, there appear to be just two types
of knowledge for Kant: a priori^ and empirical. The critical philosophy
is clearly not in the business of producing merely empirical knowledge.
Its results are intended to beaprioriK (e.g., Axv). As we have just seen,
however, transcendental proofs begin with experience, or with the pos-
sibility of experience.
Many have tried to resolve these conflicts by suggesting that Kant is
only ruling out particular experiences as a basis foraprioriK knowledge;
it need not be independent of experience in general. This seems correct,
but excessively vague. What is experience in general? There is a further
problem about the central notion of "experience." This term is used in
different senses in different arguments. At points, Kant construes having
experience as involving making judgments and bringing items under
concepts (e.g., A124-25). At other times, experiencing seems to require
nothing more than perceiving (e.g., B164-65). As many have noted,
the multiple ambiguity of "experience" raises the likelihood that some
arguments will be trivial or question begging. If the richer sense of
"experience" is assumed, it would hardly be news that experience re-
quires that items can be brought under concepts.
Both difficulties with "experience" can be resolved by employing a
more technical term. Kant's quarry is cognitive experience. Different
passages study the necessary conditions for the possibility of experience
in different senses of "experience," because there are many aspects to
cognitive experience that need to be considered. Or, expressing the
multiplicity more plainly, there are many types of cognitive task that
make up the totality of cognitive experience. "Experience in general"
 What Is Transcendental Psychology? 17

refers to the repertoire of cognitive tasks that we can perform. This is

the range of our experiences, independently of particular content. A
priori^ knowledge is knowledge that is established independently of
particular experiences, but it is tied to our cognitive capacities.
Kant's initial characterization of a priori^ knowledge is purely neg-
ative: It is established independently of experience. Much later, he
provides positive accounts. A priori^ knowledge of synthetic claims can
be produced in two quite different ways. Mathematical proofs can yield
a priori^ knowledge, because they are constructive (A713/B741). 62 The
second mode of producing synthetic a priori^ knowledge is via tran-
scendental proofs (A783/B810). In the Introduction, Kant explained the
problem with synthetic a priori claims. They assert that the predicate
concept is universally and necessarily connected with a subject concept,
even though the predicate concept is not contained in the subject con-
cept. In the case of a posteriori claims, no problem arises. The subject
and predicate concepts are connected by experience: an individual sees
something that is a body and that has weight. Since experience cannot
justify either universal or necessary claims, something else is needed as
the basis of the connection in synthetic a priori claims. Kant's well-
known answer is that the third thing that makes the connection is not
any particular experiences, or even all experiences taken together, but
the possibility of experience, that is, the possibility of our performing
any cognitive tasks at all.
The proof proceeds by showing that experience itself, and therefore the
object of experience, would be impossible without a connection of this
kind. Accordingly, the proof must also at the same time show the pos-
sibility of arriving synthetically and a priori at some knowledge of things
which was not contained in the concepts of them [A783/B811].

How this works will be made clear.

With some understanding of the notions of "a priori" in play, we can
consider the general problem of the synthetic a priori. A standard as-
sumption is that if Kant intends to prove synthetic a priori propositions,
then his premises must be synthetic a priori and his inferences must
preserve these properties. Since "a priori" is used in different senses,
this strategy would require us to determine whether all premises are "a
priori" in the same sense. Fortunately, this daunting project need not
be undertaken, because Kant does not conceive of himself as giving a
logical proof. He believes pure logic to be incapable of delivering the
results he seeks.63 So the rationale for confirming that the premises have
the same status as the conclusion misconstrues the nature of transcen-
18 KANT'S TRANSCENDENTAL PSYCHOLOGY
dental proof. Still, as we have seen, Kant does not take his arguments
to be empirical in the normal way. They are supposed to produce a
priori^ knowledge, but not because they are logical proofs of a priori^
claims.
Kant's general argumentative strategy can be framed in terms of these
three senses of "a priori." He will justify our ability to know certain a
priori^ propositions, by showing through an a priori^ argument, that
these propositions contain elements that are apriori0. The starting point
is experience, that is, various tasks that make up cognitive experience.
These tasks are analyzed in order to show that they require certain
elements that cannot be supplied by the senses.
Since cognition is a joint product of our cognitive faculties and in-
formation supplied by the senses, these elements must have their source
in our cognitive faculties themselves. They are a priori0. Such elements
include spatial properties of perceived objects and necessary connections
among events as we know them. Since they areapriori0, these elements
will be invariant across all objects as we perceive them and all events
as we know them. Thus, in the world of our perception and cognition,
or our "experience," as Kant puts it, certain claims will hold universally,
such as "all objects are side by side in space" and "all events have
causes." Further, these claims will be necessary, in an unusual sense of
"necessity." For Kant, something is necessary if it is true in all situations
that we can experience, constituted as we are.64 Because spatial prop-
erties are a priori0, were any object to be perceived by us, it would be
perceived as having a determinate position in space. Within the world
of our experience, objects are necessarily spatial.65
This is Kant's broad argument. Through reflecting on the sources of
various elements in cognition, we can see that we are justified in making
universal and necessary claims within the world of our cognitive expe-
rience, because, having followed his transcendental proof, we can see
that, within this realm, these claims are universal and necessary. Al-
ternatively, by considering a priori0 sources of elements of cognition in
a transcendental proof, claims that were merely a priori^ become a
priori^. His response to the quid juris on behalf of the apriori0 concepts
involved in such claims is, "We can do no other." Unless the a priori0
elements in these concepts were supplied by our cognitive faculties them-
selves, we could not perform even simple cognitive tasks; these concepts
are indispensably necessary for the possibility of any cognitive life for
us at all. Further, since these elements are apriori0, they will be invariant
features of human cognition. Locke's accounts of the occasioning causes
for empirical concepts have no sweeping implications for our cognitive
 What Is Transcendental Psychology? 19
66
lives. By contrast, transcendental investigations of the sources of
knowledge—transcendental psychology—disclose universal and neces-
sary features of human cognition.
Kant's transcendental proofs appeal to psychological factors to justify
a priori^ claims, but there is no fallacy of psychologism. He does not
argue from the a priori0 origin of spatial elements to the claim that all
objects are necessarily spatial, because he does not argue for the latter
claim at all. As he insists, he shows only that the proposition that all
things are side by side in space is valid, when limited to objects as
perceived by us (A27/B43). Still many find the transcendental defense
of the a priori^ status of claims of mathematics and of various meta-
physical principles profoundly unsatisfying.67 Whether formally falla-
cious or not, this is just not the sort of proof that is required. Husserl's
attitude is typical. Psychology is irrelevant, because we have insight into
the truth of logical [and so mathematical] laws.68 Although such alter-
natives seem no better than Kant's approach, these matters are beyond
the scope of my study. Regardless of its appropriateness to mathematics
or metaphysics, transcendental psychology seeks to determine the nec-
essary and universal elements of human cognition, and so is of great
interest to epistemology and to cognitive psychology as well, as I shall
argue.
Having located transcendental psychology in the general context of
the Critique—it is the new transcendental method—I will consider the
analyses of cognitive tasks in more detail. So far, my account has not
distinguished two importantly different types of arguments:
1. Arguments showing that we must have a certain type of faculty (i.e.,
something in the constitution of the mind that supplies a priori0 elements)
if knowledge, or a particular kind of knowledge, is to be possible; argu-
ments showing that a particular set of faculties is sufficient for knowledge,
or for a particular type of knowledge. (These analyses of the necessary
or sufficient conditions for knowledge, or for a particular type of knowl-
edge, may be called epistemic analyses.)
2. Arguments showing that, given an account of the cognitive tasks
that we perform that would be accepted by all parties to the dispute, it
follows that we possess one or several of the types of faculties noted in
arguments of type 1. (Because they are concerned with capacities that we
actually have, I call the second kind of study analyses of empirical
capacities.)
For example, Kant considers how we can know empirical laws of nature.
Part of the answer is that we must have a faculty of reason that is capable
of systematizing particular judgments, a faculty that introduces a sys-
20 KANT'S TRANSCENDENTAL PSYCHOLOGY

tematicity into our judgments not derived from experience. An analysis

of an empirical capacity then establishes that we have such a faculty,
thus showing that we have the ability to achieve such knowledge.69
(Whether we succeed in discovering laws of nature or bungle it is a
matter of the accidents of history.)
Discussions of Kantian epistemology focus on the first type of argu-
ment, or a depsychologized version of it, to the exclusion of the second.
It is critical to realize, however, that had Kant only offered epistemic
analyses, then his conclusions could only be conditional: We can have
X knowledge, i f . . . , or, only i f . . . .70 To take the most obvious example,
he would be utterly unable to show that we have knowledge that the
causal principle holds throughout the world of our experience.
Kant clarifies the importance of analyses of empirical capacities to his
theoretical philosophy in a passage lamenting the less fortunate position
of practical philosophy:
But human insight is at an end as soon as we arrive at fundamental powers
or faculties, for their possibility can in no way be understood and should
not be just arbitrarily imagined or assumed. Therefore, in the theoretical
use of reason only experience could justify their assumption. Such empirical
proof... is, however, denied to us with respect to the pure practical faculty
of reason [my emphasis].71

In the case of practical reasoning, it is question begging to assume that

we make bona fide moral judgments in order to argue that certain
conditions are necessary for this practice. For the skeptic's claim is that
our apparent moral judgments are no such thing.
By contrast, the epistemological skeptic grants that we carry out cer-
tain cognitive tasks. For example, in stating his position, he acknowl-
edges that we can have mental representations or that we can determine
that one of our mental states follows another in time. Should one of
Kant's opponents deny that we can perform a certain cognitive task—
as Hume famously denies any ability to attribute self-identity—then the
task cannot be used in arguments against him. Once a cognitive task is
granted, Kant argues that some important feature[s] of it reflects the
way we think, rather than the way the world is, by arguing that we
cannot account for the presence of that feature as an acquisition from
the world through our senses. In the phrase of his empiricist opponents,
that feature was never "in the senses." Thus, there must be something
in our mind, some faculty, that supplies it. A necessary or sufficient
condition for some cognitive achievement is met.
In contemporary terminology both epistemic analyses and analyses of
 What Is Transcendental Psychology? 21

empirical capacities are "functional analyses."72 They provide a func-

tional specification of the kind of processing, or faculty, required for a
given cognitive task; alternatively, they decompose a cognitive task into
its basic subtasks and so reveal that it involves elements that cannot be
supplied by the senses. There is a significant difference between the two
types of analyses, however. Epistemic analyses have a normative di-
mension. To return to the earlier example, Kant wants to determine
what is required for us to be empirically justified in raising a generali-
zation to the status of an empirical law of nature. One requirement is
that we have a systematizing faculty. Thus, normative and factual claims
comingle in transcendental psychology, but there is no fallacy of psy-
chologism. The normative claims are established by normative argu-
ments; appeals to actual cognitive tasks only establish the existence of
faculties, or other mental equipment, specified in the normative argu-
ment. Although Kant does not try to squeeze normative conclusions out
of factual premises, these premises are nonetheless crucial to his epis-
temological project. To borrow a well-known epistemic analysis from
Strawson, Kant does not wish to argue merely that we can reidentify
particulars only if we can operate with spatial relations or some analog
of them.73 He wants to show that we can always assign perceived objects
determinate spatial position.74

In Defense of Transcendental Psychology

Earlier I rebutted some general objections to an enterprise like tran-

scendental psychology. Having described Kant's project in detail, I turn
to objections to his particular version—and to my description of it. What
is transcendental psychology the psychology of?75 If we adhere to the
letter of Kant's ontology, then there are only two possible answers to
this question and both seem unacceptable. Transcendental psychology
is a study of the phenomenal self, and hence empirical. It is the study
of the unknowable noumenal self, and hence impossible. Following
Kemp Smith, Strawson elects the second option and proceeds to reveal
the incoherence of maintaining that a timeless unknown self is respon-
sible for important elements in cognition.76
Some aspects of transcendental idealism raise additional problems
that I consider much later, but the way out of this dilemma for tran-
scendental psychology itself is clear. Many commentators have noted
that the two official selves of the Critique are joined by a third self, who
turns out to be its central character, the "I" of apperception, or the
22 KANT'S TRANSCENDENTAL PSYCHOLOGY

thinking self.77 Transcendental psychology is the psychology of the think-

ing, or better, knowing, self.
If the phenomenal-noumenal distinction is exclusive and exhaustive,
then transcendental psychology must be about the phenomenal self, and
so empirical, for the straightforward reason that no positive doctrines
can be noumenal. Although it provides only a highly abstract, functional
description of a thinking self, the description is still positive. As we will
see in Chapter 5, Kant resists placing the thinking self in the phenomenal
realm for reasons having nothing to do with transcendental psychology.78
Given his own doctrine of noumena, and an exhaustive dichotomy,
however, the thinking self must be phenomenal. Hence, transcendental
psychology must be empirical, in this sense, even though it is very
different from what he regarded as "empirical psychology," and, as I
shall argue, quite different from contemporary empirical studies.
Another common complaint, familiar from the discussion of Reinhold
and Fries, centers on Kant's appeal to faculties. The plethora of faculty
names suggests a less attractive argumentative structure than the one I
present. Given a cognitive task, Kant invents or borrows a list of can-
didate faculties that might carry it out. He then deletes options until
one remains, which he claims to be necessary for the task.
In fact, only two alternatives matter for transcendental psychology:
All the salient features of a cognitive task derive from experience; some
feature[s] must be traced to the constitution of the mind itself. Kant
begins with widely shared assumptions about the range of cognitive tasks
that we can perform: We can perceive, imagine, judge. But he eschews
his contemporaries' mechanical and psychological speculations about
how the tasks are done. His own analyses of particular empirical ca-
pacities justify the positing of a faculty. Since the task requires nonem-
pirical elements, something, some faculty, must supply them. Different
names appear, because as he investigates different cognitive tasks, he
cannot attribute them to the same faculty without committing the fallacy
of double definition. It is a matter for empirical discovery whether the
same faculty enables us to represent objects and to judge them. He
could label the faculties "X," "Y," "Z," and so forth. Instead, he
enhances the familiarity and memorability of his doctrines by adapting
standard terminology, where apt: "sensibility" for the faculty concerned
with perception, "imagination" for the faculty that represents things
that are no longer present, "understanding" for the faculty of concepts,
"reason" for the faculty concerned with inference. In the first edition,
he adds new terms of his own, "the synthesis of recognition in a con-
cept," for example, to allow himself enough distinctions. Pace Reinhold
 What Is Transcendental Psychology? 23
and Fries, the danger in the method is not borrowing unexamined the-
ories but failing to posit enough distinct faculties or trying to chart the
relations among faculties encountered in different tasks.
I turn to the most likely objection to my account of Kant's psychology.
The enterprise I present as transcendental psychology is too empirical.
In the case of analyses of empirical capacities, experience contributes
two essential elements. Most obviously, it establishes the existence of
various cognitive activities, judging, representing objects, ordering
events in time, or whatever. Experience makes another vital contribu-
tion to both types of analysis, however. For what besides experience
can tell us what the senses can and cannot pick up from the environment?
And, in the absence of such considerations, Kant is in no position to
argue either that a faculty is necessary for a particular epistemic task or
that the existence of a task establishes the existence of a faculty. On my
reading, Kant has a reply to one of Strawson's opening charges against
transcendental psychology—the charge there can be no empirical evi-
dence for its truth. However, that may only reinforce the impression
that I transform transcendental psychology into a glorified empirical
psychology. I consider this objection in two parts. Is transcendental
psychology, as I describe it, too empirical to be attributable to Kant?
Is it too empirical to be philosophy?
The results of the Critical philosophy are supposed to be a prioriK,
necessary, and certain (e.g., Axv). Can analyses of cognitive tasks lead
to conclusions with these properties? For Kant, a priori^ knowledge is
absolutely independent of the content of all experiences but dependent
on facts about cognitive capacities. What sort of facts? Does he pre-
suppose only the existence of cognitive activities? Or can facts about
those activities, such as the kind of information available through the
senses, also be included? From our perspective, such information clearly
belongs to empirical psychology. At this point, we must recall the very
great differences between our conceptions of psychology and logic, and
Kant's. He locates the study of cognitive faculties in the domain of logic.
As already noted, the Transcendental Logic opens by observing that we
have two faculties, one for receiving impressions and a second for know-
ing objects. Presumably, the starting assumption about receptivity in-
cludes widely held beliefs about the limitations of the senses. In that
case, Kant would still regard his results as a prioriK, even though his
analyses draw on both types of facts about cognitive capacities, because
they only draw on facts about cognitive capacities.
Could Kant have regarded the results of such analyses as necessary?
As already noted, he employs an unusual sense of "necessity." Some-
24 KANT'S TRANSCENDENTAL PSYCHOLOGY
thing is necessary if it is true in all worlds that we can experience con-
stituted as we are. Given this conception, if his account of the features
of human cognition required for various cognitive tasks is true in the
actual world, then it is necessarily true.
Kant's claims for certainty are considerably more problematic. Certain
claims are immune from refutation. Beginning with assumptions about
cognitive capacities, it may seem that he could not hope for certain
results. As he explains in the Methodology, however, his notion of
"certain" is also relativized:
pure reason does, indeed, establish secure principles, not however directly
from concepts alone, but always only indirectly through relation of these
concepts to something altogether contingent, namely possible experience
[the range of our cognitive tasks or capacities]. When such experience is
. . . presupposed these principles are indeed apodeictically certain...
[A736-37/B764-65].

Granted that we have experience, that is, granted broadly shared as-
sumptions about our cognitive capacities, Kant maintains that his con-
clusions about the necessary and universal features of cognition are
certain.
Even allowing these substantial assumptions, however, certainty
seems beyond his reach. The analyses themselves could be mistaken.
In the case of analyses of empirical capacities, the danger may be less
acute. These only look at the end points: What does a task involve?
What do the senses contribute? So there may not be much room for
error. Turning to epistemic analyses, the picture is even less comforting.
How are their normative elements to be preserved from refutation?
Enlightening as they are, can Kant's insights into the requirements of
particular kinds of knowledge never be supplanted?
I do not see how this problem can be overcome. Like many philos-
ophers before and since, Kant is overconfident about the accepted wis-
dom of his day and overconfident about his understanding of
epistemology, so he claims certainty for his results. Although transcen-
dental psychology, as I present it, cannot deliver certain results, this
does not imply any misreading of his method. No method is infallible,
so his claims for certainty will be wrong on any reading. As I argue
below, however, more recent methods for guaranteeing certain results
in philosophy fare no better.
Is transcendental psychology too empirical to be philosophy? If, like
Reinhold, we insist on a Cartesian starting point, then this is not phi-
 What Is Transcendental Psychology? 25

losophy. However, many arguments—including Kant's in the Paralog-

isms—demonstrate the impossibility of a presuppositionless beginning
to philosophical inquiry.79 He makes assumptions, but so must we all.
Transcendental psychology investigates the faculties required for the
performance of basic cognitive tasks. Is this glorified empirical psy-
chology? I adapt the notion of a cognitive task from Allen Newell and
Herbert Simon's idea of a task analysis. In Human Problem Solving
they muse on the question of whether their work is really psychology.
Since task analyses tell us more about the nature of the task tljan about
particular subjects who perform it, they acknowledge that the resistance
to the "psychology" label has some justification.80 The same is true for
Kant's work. He is totally uninterested in the actual physical or psy-
chological embodiments of particular mental processes; the only goal is
to explore the requirements of various cognitive tasks. In this respect
his work is centrally in epistemology and very different from empirical
psychology.
Nevertheless, there is a relation between transcendental psychology
and what we (but not Kant) call "empirical psychology." They are
different modes of addressing a common subject matter. Kant explains
the relation between transcendental studies and empirical studies for
the case of space. From the Transcendental Aesthetic, we know that
the spatial features of perceived objects derive from our own faculties.
Thus, we have transcendental knowledge of the fact that all objects that
we perceive are spatial. We can also learn empirically that various ob-
jects are spatial, however. "The distinction between the transcendental
and the empirical belongs therefore only to the critique of knowledge;
it does not concern the relation of that knowledge to its objects" (A56-
57/B80—81). The relation would be the same between transcendental
psychology and empirical psychology. We learn from transcendental
psychology that we must have a productive imagination, because some
cognitive tasks require a priori0 contributions from such a faculty. How-
ever, we could also learn from simple observation that people see more
than meets the eye.
Besides commonality of subject matter, there is a further relation
between transcendental psychology and empirical psychology. In de-
fense of applying the "psychology" label to their work, Newell and
Simon argue that highly abstract, normative analyses of the requirements
of knowledge are relevant to experimental work aimed at identifying
psychological mechanisms. The same point applies to transcendental
psychology. If we discover that certain cognitive tasks require the syn-
26 KANT'S TRANSCENDENTAL PSYCHOLOGY
thesis of diverse representations, then we have an abstract description
of mechanisms that can be sought and further described through em-
pirical investigations. Kant sees this relation clearly:
What I call applied logic... is a representation of the understanding and
of the rules of its necessary employment in concrete, that is, under the
accidental subjective conditions which may hinder or help its employment.
It treats of attention, and its impediments and consequences, of the source
of error, of the state of doubt, hesitation, and conviction, etc. [A54-55/
B78-79].
What he calls "applied logic" we call "empirical psychology." Tran-
scendental psychology determines the general specifications for a mind
capable of performing various cognitive tasks. It therefore guides and
constrains empirical psychology in its attempt to determine the "sub-
jective conditions" under which the mind actually performs those tasks
(A53/B77, A54-55/B78-79).81
Transcendental psychology is a kind of psychology and it has striking
affinities with empirical psychology: It makes empirical assumptions
about cognitive capacities; it has the same basic subject matter; it can
guide empirical research. On the other hand, in trying to determine
what the mind must contribute for various cognitive tasks to be possible,
its driving concerns are epistemological, and its abstract specifications
are far removed from paradigmatically empirical work. So it is also
epistemology and a branch of philosophy. We tend to place studies under
one discipline or another; but in reality they can belong to both. Nothing
of substance turns on whether Kant's new transcendental method is
called "transcendental psychology" or "transcendental logic." I use the
former label for two reasons. Given the contemporary meanings of
"psychology" and of "logic," it is somewhat less misleading. And in the
spirit of the Roundheads, this is the name under which his work has
been abused.
In pursuing Kant's transcendental psychology, I break with the dom-
inant tradition of interpretation in Anglo-American philosophy. As a
final piece of defense, I will explain why. What I call "epistemic anal-
yses" are often presented in a different guise. All agree that a central
project of the Critique is to investigate the conditions that are necessary
for the possibility of knowledge. This neutral description is then glossed
by suggesting that Kant is analyzing the concept of "objective knowl-
edge," "experience," or "objective experience" or that he is inves-
tigating the "logic" of various sorts of knowledge claims. In the
eighteenth-century sense of "logic," the last description is accurate, but
that is not the intended sense.
 What Is Transcendental Psychology? 27

These redescriptions bring the Critique into the mold of post-Fregean

philosophy. There are however, four substantial reasons for preserving
transcendental psychology and resisting analytic interpretations. I have
already given one: The classic objections to transcendental psychology
can be turned aside. The second will be argued throughout the book:
Transcendental psychology, as Kant conceives of it, has important con-
tributions to make to philosophy and psychology. A third reason comes
from the work of W. V. Quine. In a series of papers, culminating in
"Two Dogmas of Empiricism," Quine demonstrated two crucial theses
about analytic philosophy. If philosophy takes formal logic as its model,
then its results can attain the special status of logical theorems only by
employing analytic definitions of any nonlogical vocabulary. But the
required analytic definitions are impossible, because there is no such
thing as analytic truth—because the only candidate for analytic truth is
truth by convention, and nothing can be true "by convention."82 If
Quine's position is granted (and it would be inappropriate for me to
reargue it here), then we diminish the arguments of the Critique by
casting them in analytic form.
Finally, these glosses run counter to Kant's own attitude toward an-
alyticity, which is surprisingly close to Quine's.83 As commentators
sometimes lament, his introductory account of analyticity makes it a
boring psychological property: In analytic judgments, the predicate is
"thought in" the subject (A6/B10ff). Further, he denies that empirical
or a priori concepts can be defined84; only arbitrarily invented concepts
can be defined. In his own lifetime Kant resisted Fichte's and Eberhard's
attempts to replace his synthetic claims with analytic counterparts. His
"Open Letter" to Fichte complains that such a strategy reduces the pure
theory of science to logic, and no real object can be gotten out of pure
logic.85 The reply to Eberhard (written by Schultz, but approved by
Kant) is even more Quinean:
Let one place just so many marks in the concept of the subject that the
predicate, which he wishes to prove of the subject, can be derived from
its concept through the mere principle of contradiction. This trick does
not help him at all. For the Critique grants to him without dispute this
kind of analytic judgment. Then, however,... it asks: how did it come
about that you have placed so many different marks in this concept?86

Kant's novel approach to philosophical questions has also been given

an analytic interpretation. The new method of transcendental philoso-
phy is supposed to be the "transcendental argument." Transcendental
arguments are intended to refute skepticism of the potent Cartesian
28 KANT'S TRANSCENDENTAL PSYCHOLOGY
variety. They start by assuming only the experience of a self-identical
being. Their special characteristic is that they ensnare skeptics in self-
contradiction. Once a skeptic grants that he has experience, an analysis
of the concept of a possible experience shows that it is is part of the
meaning of "experience" that the proposition the skeptic doubts must
also be true.87
Despite its popularity, this interpretation of Kant's transcendental
method has serious weaknesses. The argument form itself has been
subjected to devastating criticisms that have yet to be satisfactorily an-
swered.88 It is also far from clear that Cartesian skepticism is a primary
target of the Critique.*9 Later, I argue that Kant does not and cannot
assume a subject of experience who is identical through time.90 The
preceding discussion implies additional shortcomings. Transcendental
arguments employ a notion of analytic truth that Kant rejects and Quine
refutes.
It seems time for another approach. No one doubts that a central
project of the Critique is to examine how the workings of our thought
processes are reflected in the knowledge claims we make. An essential
part of the complex doctrine of transcendental idealism is that what we
know is partly a reflection of our ways of knowing. The questions concern
the value of this study. So far I have only tried to dispel initial doubts.
In what follows, I will argue that transcendental psychology has impor-
tant contributions to make to contemporary discussions. Within philos-
ophy its positive doctrines form the basis for an excellent account of
mental unity (see Chapters 4 and 5); its negative teachings about the
limits of transcendental approaches provide needed discipline for con-
temporary speculations (see Chapter 7).
If we look beyond the bounds of Kant scholarship and philosophy,
the need to reconsider transcendental psychology is even clearer.91 As
already noted, the face of psychology has changed. Cognitive psychology
and cognitive science are now areas of intense and exciting work. If
transcendental psychology aims to determine what our faculties must
bring to cognition, then any lasting results will be significant contribu-
tions to cognitive science. Kant was not an interdisciplinary cognitive
scientist. He was totally uninterested in the embodiments of cognitive
processing; he had only a raconteur's interest in the effects of culture
on cognition; he specifically denied any interest in how children acquire
knowledge; and he could have had no understanding of the twentieth-
century discipline of computer science. His sole objective was to deter-
mine what our cognitive powers had to be like for them to be capable
of producing knowledge. Contributions to interdisciplinary projects
 What Is Transcendental Psychology? 29

need not be interdisciplinary, however. Because he was not distracted

by other issues, Kant was able to offer unusually rich analyses of the
task environment of various aspects of cognition. I argue later that these
analyses can provide insights and direction for contemporary research
in cognitive science (see Chapters 3, 5, and especially 8).
In neglecting transcendental psychology, scholars lose the opportunity
to contribute to current debates in philosophy, psychology, and cognitive
science. They also harden the Critique's reputation as an intellectual
tyrant. All must pay tribute to its greatness. Graduate students must
master enough of its doctrines to pass comprehensives; a well-rounded
intellectual must recognize its pervasive influence on Western thought.
Behind the scenes, however, those who know admit that it is impossible
to read and that its major sections are fatally flawed: The Aesthetic is
outmoded; the celebrated transcendental deduction is impenetrable by
finite understandings. Perhaps there are a few worthwhile negative les-
sons. If we attend to their psychological doctrines, however, the Aes-
thetic offers important insights about perception, the Analytic raises
serious issues about representing and judging, and the transcendental
deduction is a long but coherent argument. So I will argue in Chapters
2, 3, and 6. Not surprisingly, the book is also much easier to follow if
you read all of it.
Although I do not make the mistake of ignoring transcendental psy-
chology, I do not aspire to Kemp Smith's goal of an adequate treatment
of Kant's teaching. Given the range of topics and the sophisticated level
of the discussions, it is hard to see how any individual could offer a
complete account. My aim is only to begin the rehabilitation of the dark,
psychological side of Kant's work, so that it is available for contemporary
research. Even my treatment of transcendental psychology is far from
exhaustive, because I do not consider the Second and Third Critiques
and other relevant texts at all. Nor do I deal with all the discussions in
the First Critique. Many careful and illuminating studies of Kant's ep-
istemic analyses shine through their analytic garb. Rather than duplicate
efforts, I focus on the neglected analyses of empirical capacities. In a
well-known passage Descartes extols the virtues of a single architect for
a city. Although this may produce superficial elegance, surely the needs
of inhabitants are better met when many contribute their expertise.
 2
The Science of Sensibility

What the Transcendental Aesthetic Is About

"The science of all principles of sensibility a priori, I call transcendental

aesthetic" (A21/B35, amended translation). This introduction leaves no
doubt about the subject matter of the Transcendental Aesthetic. Its
doctrines concern sensibility, one of the two fundamental faculties re-
quired for knowledge, just as the doctrines of the Transcendental An-
alytic concern understanding, the second fundamental faculty. Every
beginning student of Kant is taught this fact, sometimes through appeal
to the analogy of colored spectacles. After class, however, Kantians do
not seem to accept the basic point about the Aesthetic that they dutifully
impart to students. Few recent discussions make any but fleeting ref-
erences to the Aesthetic's science of sensibility.
Scholars avoid this issue because they believe that Kant's theory of
sensibility is both unphilosophical and obviously flawed. For reasons
surveyed in Chapter 1, current opinion will reject any theory involving
a psychological faculty as unphilosophical and false to Kant's own prin-
ciples. The virtually universal consensus that the doctrine of the forms
of intuition is, in any case, hopeless was probably created by Hans
Vaihinger's lengthy summation of his predecessors' views,1 which was
made widely available to Anglo-American philosophers by Norman
Kemp Smith. Here is the charge (in Kemp Smith's words):
This distinction between matter and form is central in Kant's system
(because it is essential to his solution to the problem of the synthetic a
priori)... .
Kant proceeds to argue: (a) that the distinction is between two elements
of fundamentally different nature and origin. The matter is given a pos-
30
 The Science of Sensibility 31
teriori in sensation; the form . . . must lie ready a priori in the mind, (b)
Kant also argues that form, because of its separate origin, is capable of
being contemplated apart from all sensation. The above statements rest
upon the unexpressed assumption that sensations have no spatial attributes
of any kind. In themselves they have only intensive, not extensive, mag-
nitude. Kant assumes this without question and without the least attempt
at proof.2
For the parade case of space, the doctrine of the forms of intuition is
not only about a psychological faculty; it rests on an unsupported and
implausible psychological premise.
Although sensibility is the focal topic of the Aesthetic, it has been
easy to discuss this part of the Critique and still avoid faculty psychology,
by concentrating on other weighty matters: space, time, and mathe-
matics. Jaakko Hintikka has offered an interpretation of "intuition"
that even makes it possible to discuss this crucial concept of the Aesthetic
without dealing with sensibility.3 Although Hintikka's interpretation is
not widely accepted,4 the fact that it has a significant place in the lit-
erature is a measure of current eagerness to read the Aesthetic without
touching on any doctrines of transcendental aesthetics.
In this chapter I try to reverse the long-standing negative appraisal
of Kant's work on sensibility. Both Kant and his critics take the claim
that space is the form of outer intuition to be the test case. That will
also be my topic, partly because of precedent and partly because he has
a much stronger argument about space than about time, and the resulting
doctrine is less problematic.5 Far from being an embarrassing failure, I
will argue that the Critique's first effort in transcendental psychology—
the analysis of spatial perception—is largely successful and offers im-
portant insights into the phenomenon.
Spatial perception is a central issue in the development of Kant's
philosophy for a variety of reasons. From his first published work,
Thoughts on the True Estimation of Living Forces, he was fascinated
with the problem of why space has three dimensions and why it is
apparently impossible for us to imagine a space of more than three
dimensions.6 As many others have observed, he was vitally concerned
with the Newton-Leibniz debate over the nature of space [and time].
From his earliest writings, he was also deeply perplexed about the status
of mathematics.7 Does mathematics apply to nature, and if so, how?
Further, as I argue later, spatial perception was a prominent item on
the philosophical agenda. In reflecting on contemporary puzzles about
spatial perception, Kant realized that it must involve a priori^ elements.8
This critical insight provided what he believed to be correct solutions
32 KANT'S TRANSCENDENTAL PSYCHOLOGY
to the problems of the nature of space and the status of mathematics.
His analysis of the task of spatial perception thus became the prototype
for transcendental philosophy. By determining the a priori0 origins of
our perceptual representation of space, he argues that certain features
will be invariant throughout the world of our experience.
My focus is almost exclusively on the neglected topic of the tran-
scendental psychology of spatial perception. However, I believe that
Kant adopted the position that space is a form of outer intuition, because
it provided a single solution for pressing problems in all three areas: the
metaphysical status of space and time, the epistemological status of
mathematics, and contemporary debates about spatial perception. The
ability of one theory to solve important problems in three previously
unconnected areas is remarkably strong evidence for its truth. If we
ignore the transcendental psychology of perception, then it is impossible
to appreciate the strength of Kant's conviction that, whatever its peculiar
consequences, his theory had to be right. Since sensibility is the topic
of the Aesthetic, this practice also makes it extraordinarily difficult to
follow the cryptic argumentation of major passages. I shall argue that
abstracting from the context of perception also distorts our understand-
ing of the theory of geometry.9 The main reason for reconsidering Kant's
analysis of spatial perception is the one already given, however. Despite
more than 100 years of philosophical detractors, this account has been,
and continues to be, a significant contribution to the subject. To ap-
preciate the strengths of that analysis, we need to do something he does
not do in the text: Consider the problem of spatial perception as it came
to him.

Early Modern Theories of Spatial Perception

Kant's predecessors tried to explain a number of perceptual puzzles.

An important explanandum was our ability to perceive the third di-
mension. William Molyneux, the Dr. Molyneux of Locke's Essay, gave
the problem its classic formulation:
For distance of itself, is not to be perceived; for 'tis a line (or a length)
presented to our eye with its end toward us, which must therefore be only
a point, and that is invisible.
(Dioptrika Nova, 1692)'°
Descartes had tried to solve Molyneux's puzzle in La Dioptrique, pub-
lished 54 years earlier. The best-known feature of Descartes's account
 The Science of Sensibility 33

was the hypothesis of natural geometry. On this view, one way that we
know about distance is by the relation of the eyes to each other in
perception. Given the distance between the eyes and the angles between
the line connecting the eyes and the lines of sight to the object, we may
calculate the distance by a "sort of natural geometry" that enables us
to determine the height of the triangle formed by the two eyes and the
object.11
Descartes' hypothesis of natural geometry was widely influential. It
was also the target of one of the most important works in eighteenth-
century psychology, An Essay Towards a New Theory of Vision. Berke-
ley's attack on natural geometry was two-pronged. Negatively, he argued
against the "receiv'd Opinion" on the grounds that we are not aware
of the angles of the lines of sight, not to mention the interocular distance.
Then there is the problem of children and others ignorant of geometry.
Positively, he argued that there were various distance clues that we are
aware of and able to use, because experience has taught us the regular
correlation between these clues and tactile distance.
Berkeley vacillated about the ability of touch to inform us about
distance. In the Principles and the Three Dialogues, he adopted the
position that touch does not provide evidence for outness or outer ob-
jects. In Alciphron and The Theory of Vision or Visual Language: Vin-
dicated and Explained, he returned to his more popular position that
the objects of touch are without the mind. Nevertheless, this theory was
widely accepted, perhaps because of the dramatic empirical support that
he was able to cite in the latter work. In 1728 the Philosophical Trans-
actions of the Royal Society published William Cheselden's report of the
results of an operation that provided sight to a "young Gentleman" of
13 or 14. By Cheselden's account, the boy "when he first saw... was
so far from making any Judgment about Distances, that he thought all
Objects whatever touch'd his Eyes. . . . " I2 So much for the Rationalist
theory that we perceive distance by means of an innate geometry. Be-
sides Berkeley's own citation, the Cheselden case was widely reported
on the Continent. Like Washoe's "baby in my drink" today, the Che-
selden boy's testimony would have been known to most scientifically
literate people in the last two thirds of the eighteenth century.
To what extent was Kant aware of the controversy over spatial per-
ception? Ranged on one side of the debate we find Descartes, Barrow,
Malebranche, and Molyneux. Leibniz also jumps into the fray in his
critique of Locke's discussion of Molyneux's query about the cube and
the sphere. Twenty-four years before Cheselden's dramatic report, Leib-
niz points out the need for care in questioning the newly sighted, for
34 KANT'S TRANSCENDENTAL PSYCHOLOGY

they will be "dazzled and confused by the strangeness," and what they
"actually do on the spot" might be quite misleading.13 Lining up behind
Berkeley, more or less, are Voltaire, a converted Condillac, a host of
other French philosophers, including D'Alembert, Diderot, and Buffon,
Dr. Robert Smith, David Hartley, and Thomas Reid.14 Given the num-
ber and influence of the protagonists, it is hard to see how Kant could
have avoided at least a superficial acquaintance with the drama.
Turning to more specific points, Kant had a German translation of
Berkeley's Three Dialogues in his library.15 The publication date is 1781,
making its influence on the first edition of the Critique doubtful. How-
ever, he may well have seen an earlier translation by Johann Christian
Eschenbach, the first German translator of Berkeley. In the First Dia-
logue, Philonous raises Molyneux's difficulty about the perception of
distance. It is uncertain whether Kant could have had access to the Essay
on Vision itself. Jessop maintains that an edition of Alciphron with the
Essay appended to it was translated into German around 1756, but later
bibliographers have been unable to verify the claim.16
Eschenbach introduced Johann Nicolaus Tetens to British philosophy,
and in turn, Tetens's work was closely studied by Kant.17 Tetens ad-
dresses the problem of spatial perception in his Philosophische Versuche
uber die menschliche Natur und ihre Entwickelung, which appeared be-
tween the Inaugural Dissertation and the 1781 version of the Critique.18
He accepts a Humean principle that ideas must be abstracted from
experience.19 The idea of space cannot be abstracted from particular
alterations and impressions made by objects themselves, however.
Rather, it is abstracted from the act of seeing or feeling.20 When, for
example, I see that a tower is further from me than a tree, I move my
eyes, and that is felt as something present and absolute.21 So we have
an impression from which the idea may be abstracted. In essence, Tetens
accepts Berkeley's criterion for a solution—if information about our
eyes is to be used in computing distance, then it must be conscious—
but then claims a richer deliverance from introspection. Oddly, he main-
tains that his introspective account of spatial perception is exactly the
view offered in the Inaugural Dissertation.22
Vladimir Satura has collected all the published sources of Kant's re-
flections on psychology. Two sources have him referring explicitly to
the Cheselden case, and arguing that we acquire our concept of bodily
form from touch, because vision only presents us with flat images.23
These references occur in Starke's version of the lectures on anthro-
pology, usually referred to as the Menschenkunde, and Kowaleski's
edition of Graf Heinrich zu Dohna-Wundlacken's metaphysics lecture
 The Science of Sensibility 35

notes. Scholars have long debated the accuracy of these reports of Kant's
lectures. Nevertheless, in the former case at least, it seems reasonable
to believe that these issues were covered, because Starke's account is
compatible with the discussion in Anthropology from a Pragmatic Point
of View. In his own version of the lectures, Kant claims that we can
acquire the concept of bodily form only from touch. He also notes that
someone whose sight is restored by an operation must learn to see.24
The anthropological lectures were not published until 1798, but he began
lecturing on these subjects in the fall of 1772.
Neither Descartes's theory of natural geometry nor Berkeley's famous
account of vision is mentioned in the Inaugural Dissertation's discussion
of sensibility, in either version of the Transcendental Aesthetic, or in
the corresponding portion of the Prolegomena. (Had Kant cited these
theories, I would not have appealed to the lecture notes of a 14-year-
old count.) It is important not to be misled by Kant's silence, however.
He habitually fails to mention predecessors who defined the issues he
takes up in the Critique. The first edition's discussion of dogmatic ide-
alism does not refer to Berkeley either and Newton's name does not
appear in any Critical discussion of space or time. (See Chapter 4 for
further examples.25) Of course, there was no need to remind his con-
temporaries that this labyrinth arose from Newton's theories. I have not
belabored the historical evidence just to prove that Kant was aware of
the controversy over spatial perception. That is fairly obvious. What I
have tried to show is that he could reasonably assume his readers' fa-
miliarity with the problem, and with the theories of Descartes, Berkeley,
and Leibniz. Hence, he does not trouble to do what I have just done:
Set the stage for his solution to the problem.

Kant's Analysis of Spatial Perception

Intuition, Matter, and Form

Both the Inaugural Dissertation and the Critique prepare the way for
an account of spatial perception by analyzing the elements of perception
in general. The Dissertation offers a three-part analysis of sensory per-
ception: the sensory representation itself (which Kant fails to mention
explicitly), its matter, and its form. Sensations caused by the sensible
object are the matter of sensory representations. The form "arises ac-
cording as the various things which affect the senses are coordinated by
a certain natural law of the mind."26
36 KANT'S TRANSCENDENTAL PSYCHOLOGY
Just as the sensation which constitutes the matter of a sensual represen-
tation is evidence at least for the presence of something sensible, but in
respect of its quality is dependent upon the nature of the subject to the
extent that the latter is capable of modification by the object in question,
so also the form of the same representation is undoubtedly evidence of a
certain respect of relation in the sensa [sense impressions]. But properly
speaking it is not some adumbration or schema of the object, but only a
certain law implanted in the mind by which it co-ordinates for itself the
sensa which arise from the presence of the object.27

This passage uses "form" ambiguously to refer to both a property of

sensory representations and a law of the mind that produces sensory
representations out of sensa. To avoid confusion I will refer to the
putative law of the mind as "process form" and to the putative property
of the representation as "product form." The thesis is that as sensations
depend on both the objects and the sensory organ, sensory represen-
tations depend on both properties of the sensa and the mind's mode of
producing representations out of sensa.
The Critique presents the same tripartite analysis of perception, al-
though it uses four terms: "sensation," "intuition," "matter," and
"form." With the new terminology Kant is able to sharpen the contrast
between sense impressions and sensory representations.28 He introduces
the name "intuition" for sensory representations, that is, subjectively
available representations of external or internal objects produced by
outer or inner sense.29 Here the matter of sensory representations is not
sensations themselves, but something corresponding to sensations. Sen-
sations are the effects of objects on the faculty of sensibility (the sensory
organs taken collectively) (A19-20/B33-34); the matter of intuitions
comprises those features of representations that derive from sensations,
for example, color (A20,21/B34,35). Kant offers a justification for why
there must also be form:
That in which alone sensations can be ordered and posited in a certain
form cannot itself be sensation; and therefore while the matter of all
appearance is given to us a posteriori only, its form must lie ready for the
sensations a priori in the m i n d . . . [A20/B34, amended translation].
This is a very difficult passage. One implicit claim is that because
sensations and sensory representations or intuitions are distinct, some
process must produce intuitions out of sensations (even if it only pro-
duces representations whose qualities all derive from sensations). Kant's
discussion is predicated on the recognition that we do not standardly
perceive sensations, or the effects of objects on our sensory organs.30
 The Science of Sensibility 37

This point would have been familiar to his readers, partly because of
its role in solving the problem of inverted retinal images. (The problem
vanishes when we recognize that we do not perceive our retinal image;
we perceive via the image.) Thus, the first and most straightforward
claim in the passage:
(1) Given that sensations and intuitions are distinct, there must be some
process that produces intuitions out of sensa, even though that process
might yield intuitions with qualities that all derive from sensations.
Intermingled with it are three other separable points:
(2) There is a certain product form in [all] intuitions, namely a certain
relation of the elements or parts.
(3) If there is a certain product form in intuitions, then there must also
be a certain process form.
(4) Given that what assures the product form characteristic of intuitions
is not a sensation, but a process form, then it must "lie in the mind a
priori."
Plainly, (2), (3), and (4) require justification considerably beyond what
I have provided for (1). Further, even if (2) can be established, that
will not suffice to defend (3) or (4). In the succeeding paragraph, Kant
introduces a concept of a "pure" form. This notion is crucial in dem-
onstrating the connections among the preceding claims and in setting
up the analysis of spatial perception.

Pure Forms (and Nativism)

A representation is "pure" just in case "nothing is met in it that is due
to sensation [zur Empfindunggehort]" (A20/B34, amended translation).
A part, aspect, or feature of a representation may also be "pure" (A20/
B34). In the first edition Kant repeats the point that, ipso facto, a pure
form includes "no sensation (nothing empirical) in itself" (A29a,
amended translation). Thus, a feature or an entire representation is
pure if it does not derive from sensation, if it cannot be traced back to
sensations. He continues, "The pure form of sensible intuitions in gen-
eral, in which all the manifold of intuition is intuited in certain relations,
must be found in the mind a priori" (A20/B34). Since it has not yet
been established that there are any pure forms, this claim should be
viewed as a hypothetical: If there is a pure product form of all intuitions,
then it must be traced to a process form that lies in the mind a priori.
Chapter 1's discussion of Kant's multiple uses of "a priori" can clarify
38 KANT'S TRANSCENDENTAL PSYCHOLOGY

this claim. Since pure forms, if there be such, cannot be traced to

sensations, they are a priori0. That is, the pure product forms are a
priori0. Does this make pure process forms innate, even though they
are described as a priori4? Several pieces of evidence suggest that this
must be so. In the Inaugural Dissertation, Kant had claimed that the
form of sensory representations was determined by "stable and innate
laws [of the mind]."31 Further, in the polemic against Eberhard already
cited in Chapter 1, he writes:
The Critique admits absolutely no divinely implanted or innate represen-
tations There must, however, be a ground in the subject which makes
it possible for these representations to originate in this and no other
manner.... This ground is at least innate.... Only this first formal
ground, e.g., the possibility of a representation of space, is innate, not
the spatial representation itself [my underscoring].32

On the other hand, as noted earlier, Kant believes that in the case of
vision at least, we learn to see (i.e., to have perceptual representations
of objects in space). Further, he has no sympathy for lazy nativist
hypotheses.33
The way out of this interpretive impasse is to recognize that a
"ground" is not a "process." I believe that Kant's eschewing talk of
stable and innate laws in the Critique reflects a more sophisticated anal-
ysis. Suppose that the pure product form of human intuition could be
produced from sensa by any member of a set 5 of processes {P1; P2, P3,
...}. Although we must have some innate initial process to produce our
first intuitions, the course of those intuitions could affect the process
form, by substituting a different member from 5. Thus, the actual pro-
cesses that produce the pure product form of adult human intuition
might be neither stable nor innate, and Kant never claims either attribute
in either version of the Critique or in the Prolegomena. He is trying to
avoid unnecessary causal hypotheses. All he needs to argue is that if
there is a pure product form in human intuition, then some member of
S must be available to produce that form, although not necessarily the
same member at all times. The ground of any pure product form would
lie in our constitution, because perception is a conjoint product of the
sensations we receive and our ways of dealing with them—and a pure
form cannot be traced to sensations. However, that does not imply that
the process (or processes) that actually enables us to have spatial rep-
resentations is innate.
If we abstract from the differences among the members of S, then we
can describe them all indifferently as "pure process forms" of intuition.
And this is exactly how Kant's terminology works. In its fundamental
 The Science of Sensibility 39

meaning, the "pure form of intuition" refers to a feature of intuitions

themselves. Kant uses "form" in this sense to provide a functional char-
acterization of mental operations. In the latter contexts, "form" is ellipt-
ical for the "mental operation, whatever it is, if there is just one, or
whatever they are, if there are many, that produce[s] the pure form in
sensory representations" (see A20/B35, A22/B36, A26/B42, A27/B43,
A29). Hence, the ambiguity noted earlier between "product form" and
"process form" and "process form" is not vicious but principled. Kant's
second use of "pure form" is extremely abstract, because it is of no con-
cern to transcendental psychology which of many possibilities is realized
in us. Thus, we should resist the temptation to read the claim that pure
process forms lie in the mind a priori as a fancy way of saying that they are
innate. Kant means only what he says: pure forms—product and pro-
cess—are apriori0.
Kant also uses "a priori''' to describe the epistemological status of
claims produced by his transcendental method.34 Thus, fully explicated,
his claim is: We know a priori^ that if there is a pure or a priori0 product
form of intuitions, then there must be some pure or a priori0 product
form producer—a pure process form—that yields intuitions with that
feature from sense impressions. In this heavy psychological passage,
Kant is not indulging in speculation about psychological mechanisms for
which he has no evidence. He is offering a simple analysis of the task
of perception involving pure forms, and his analysis is correct. Further,
this analysis fills in the gaps in the four claims from his preceding par-
agraph (see above). Claim (3)—that if there is a product form in sensory
representations, then there must be a process form that accounts for
the presence of that feature—makes sense on the assumption that the
product form he has in mind is pure. Likewise, claim (4)—that this
process form must be a priori{0]—will follow, if the product form is pure.
The earlier paragraph is somewhat confusing because he is drawing on
common background assumptions about perception and anticipating his
analysis of pure forms at the same time. As already noted, standard
assumptions about the distinction between sense impressions and sen-
sory representations support (1); the later analysis shows the connections
among (2), (3), and (4). What remains to be supported is the crucial
assertion (2), that sensory representations have pure product forms.

The Method of Isolation

Kant proceeds at once to explain how to discover pure or a priori0
product forms. His famous method of isolation follows in part from the
definition of a pure form. One discovers pure product forms of intuition
40 KANT'S TRANSCENDENTAL PSYCHOLOGY

by eliminating those elements of a representation that are conceptual

and then by eliminating those elements that can be traced back to sense
impressions. For the purposes of this chapter, it will be unnecessary to
take up the vexed question of how to distinguish conceptual and sensory
representations.35 He explains the second isolation as follows: We take
away "what is due to sensation, impenetrability, hardness, color, etc."
(A20-21/B35, amended translation). In a reflection, he explains that we
are to take away anything that strikes the senses.36 The isolation yields
positive results: Extent and form (Ausdehnung and Gestalt) are features
of sensory representations that do not derive from sensations. The In-
augural Dissertation makes the related claim that "space itself cannot
be gotten out of the senses."37
Current readers should be puzzled by these pronouncements. Nothing
in the preceding discussions prepares us for the result and Kant offers no
hint of an argument in either the Aesthetic or the Dissertation passages.
Vaihinger, Kemp Smith, and many others try to fill out the enthymeme by
assuming that Kant adopted the implausible claim that spatial relations
(or perhaps all relations) are incapable of being sensed.38 (Kemp Smith's
version of this charge is cited at the beginning of the chapter.) There are
three serious objections to this move. The view it imputes to Kant is in-
consistent with the Dissertation claim that the form of the representation
does reflect an aspect or relation of the sensa. Further, it makes the the-
ory vulnerable to a fatal objection. If sensations can provide no relational
information at all, then whether we perceive something to be square, tri-
angular, round, or oval will be independent of our actual sensations.39 As
critics from Feder to Herbert to Vaihinger have pointed out, however,
this consequence of the implausible enthymematic premise that they at-
tribute to Kant is totally implausible. Finally, the logic of the passage
standardly cited in support of this interpretative stratagem is the reverse
of what is needed: "Since, however,... neither the intuition of space nor
that of time is to be met with in [sensation], its magnitude is not extensive
[i.e., measurable in units] but intensive [i.e., not measurable in units]"40
(A166/B208). The argument is not that space is not encountered through
sensation, because sensation conveys only intensive information, but
that since space and time are not met with in sensation, it [sensation]
conveys only intensive information.

Distance, Extent, and Shape

The contemporary debate over spatial perception enables us to make a
much more plausible interpolation that reveals the logic of Kant's rea-
 The Science of Sensibility 41

soning. All sides recognized that distance could not be registered in the
visual system. As Molyneux observed, distance is a line straight out from
the eye, and so cannot be seen. In the absence of distance information,
however, it is impossible to determine the size of perceived objects, for
objects of vastly different sizes will produce the same size images on
the retina (or fundament, in eighteenth-century terminology), if they
are at sufficiently different distances from the perceiver. Thus, without
distance information, it is impossible to determine the extent of an object
in any dimension (not merely the third). It is not even possible to
determine relative dimensions (e.g., that an object is wider than it is
tall, because if it is tilted toward or away from the viewer, its height
will be foreshortened on the retina). If vision cannot supply information
about any dimensions of perceived objects, however, then it can tell us
nothing about their shapes. The same point can be made slightly more
formally. Given the well-established fact that the organ of vision is a
nearly planar retina41 and the obvious geometrical fact that a two-
dimensional configuration could be the planar projection of indefinitely
many different three-dimensional objects, it follows that by itself the
retinal array can provide no information about shape or size at all. Planar
projections preserve some spatial properties of three-dimensional ob-
jects: left-right relations, up-down relations, and betweenness relations.
But they lose precisely the information that Kant attributes to pure or
a priori0 forms: extent (in any dimension) and form.42

Touch: Leibniz Versus Berkeley

Since the only serious candidates for sensing distance were vision and
touch, the only remaining possibility is that distance is registered in
tactile sensations. This option was, of course, the central tenet of Berke-
ley's Theory of Vision. It can be captured in three claims: (1) true visual
images are flat; (2) our ideas of distance derive from the sensations of
touch produced during motion; (3) we falsely believe that our visual
images are three-dimensional because, through past association, flat
visual images call up three-dimensional tactile images, which are then
mistaken for three-dimensional visual images.43
Berkeley's view was widely accepted44 and Kant undoubtedly had
knowledge of it. So he could not simply dismiss tactile sensations as a
source of distance information unless he was aware of compelling reasons
for believing that, whatever its popularity, the touch and motion hy-
pothesis had to be wrong. Further, he must have believed that his readers
were sufficiently familiar with these objections to accept his assumptions.
42 KANT'S TRANSCENDENTAL PSYCHOLOGY

The influential critic was Leibniz. In the New Essays on Human Un-
derstanding, he offered several reflections that tell against Berkeley's
position. Although this discussion is directed at Locke's negative answer
to Molyneux's question about the cube and the sphere (the Theory of
Vision did not appear until 5 years after Leibniz wrote the New Essays),
it makes two points that cut directly against Berkeley's claims for touch.45
First, Leibniz notes that people who are paralyzed could and must
learn geometry through sight. Since Berkeley's claim that touch is nec-
essary (and sufficient) for acquiring ideas of distance and magnitude
ascribes a crucial role to motion, the actual or assumed case of a par-
alyzed individual learning geometry would be a counterexample. Sec-
ond, Leibniz observes that geometry is usually learned by sight without
employing touch. In opposition to Berkeley's general Empiricism, this
discussion implies that our spatial ideas must have some source other
than the senses. An individual who is paralyzed could not acquire such
ideas from touch—nor, of course, could he acquire them from visual
images, since the starting premise of the debate over spatial perception
was that retinal images cannot supply three-dimensional spatial
information.46
Leibniz's general position is that it is essential to distinguish images
from exact ideas, which are composed of definitions. He uses the case
at hand to illustrate his point: the images of sight and touch are disparate,
but the individual who is blind and the individual who is paralyzed have
the same exact ideas of geometry.47 It is generally agreed that Kant read
the New Essays about 1769.48 One year later, in the Dissertation, he
introduced the theory that space is the form of outer sense; 1769 was
the year that "gave him great light."49 Seemingly, Leibniz's unintended
reply to Berkeley was a significant part of that illumination, for it pro-
vided the final piece of the problem of spatial perception that would be
solved by the form of intuition doctrine. Although composed around
1704, the New Essays were not published until 1765. They would have
been current in the minds of Kant's intended audiences in 1770, 1781,
and 1787.
Leibniz offered persuasive reasons for doubting that perceptual in-
formation about distance derives from touch. Vision and touch were
the only serious candidates for the sensory sources of information about
distance, however. And, as I have already argued, in the absence of
distance information, it is impossible to determine the Ausdehnung or
Gestalt of perceived objects. Hence, Kant asserts that extent and form
cannot be gotten out of the senses.50 Pace Vaihinger, Kemp Smith, their
sources and disciples, he had no need to help himself to an implausible
 The Science of Sensibility 43

premise about our inability to perceive spatial relations; his starting

assumption is a reasonable summation of the results of the debate over
spatial perception up to that point. Further, if the enthymematic premise
is that distance cannot be sensed, then he can maintain that the size
and shape of perceived objects are a pure product form of intuition,
without being saddled with the unacceptable consequence that whether
we perceive something to be a cube or a sphere is totally independent
of our sensations. Even though a two-dimensional retinal array provides
no depth or size or shape information, because it could be the planar
projection of different three-dimensional objects, any putative three-
dimensional interpretation of the two-dimensional array must be an
interpretation of that array. Thus, as the Inaugural Dissertation explains,
the sensory representation or intuition does reflect relations among the
sensa.
Kant seems incredibly (even annoyingly) cryptic to us, but there was
little need to remind his audience of the familiar points on which his
discussion turns. The inability of vision to register depth was the common
assumption that gave rise to the problem of spatial perception; the
implausibility of the touch hypothesis had just been discussed in the
recently published New Essays. What he is doing in this passage is
drawing out the obvious, but in his view momentous, consequences of
these well-known facts. Given these facts, and the fact that we are
capable of perceiving objects in spatial arrays, it follows that the spatial
properties of objects of perception are a pure, a prlorl0 product form
of intuition and that any possible theory of spatial perception must
appeal to some pure, a priori0 process form (or forms).

Kant's Empirical Assumptions

Having defended Kant against a venerable charge that he makes an
unwarranted psychological assumption, I should note that he does begin
his analysis of spatial perception with a different empirical assumption.
In the New Essays, Leibniz had argued that however disparate their
images might be, an individual who is blind and an individual who is
paralyzed employ the same ideas of geometry. Kant assumes that all
outer senses employ a common mode of spatial representation. This is
his initial description of our empirical capacity of spatial perception:
By means of outer sense, a property of our mind, we represent to ourselves
objects as outside us, and all without exception in space. In space their
shape, magnitude, and relation to one another are determined or deter-
minable [A22/B37].
44 KANT'S TRANSCENDENTAL PSYCHOLOGY

That is, spatial representations do not differ from one sensory mode to
the next. All outer sensory systems enable us to locate objects in the
same all-embracing spatial network.
This is a contingent matter, however. It is possible (although re-
markably inconvenient) for spatial locations assigned to objects in visual
perception to be incommensurable with those assigned through tactile
perception. In describing transcendental psychology, I noted that Kant
needs to analyze actual empirical capacities if he is going to argue that
we have any of the faculties required for objective knowledge, so it is
not a fault of his argument that he starts with empirical assumptions,
but a virtue. Leibniz offered a mathematical rationale for attributing
the same spatial ideas to individuals who are blind and to individuals
who are paralyzed. They must share the same geometry. Although Kant
undoubtedly concurs that the spatial representations produced through
touch cannot differ from those produced through vision, because ge-
ometry describes space and there is only one actual geometry (A165/
B206), he has not yet raised any considerations about geometry by A22/
B37 in the Aesthetic. Presumably, he feels justified in making this as-
sumption on the basis of the sort of commonsense empirical observations
and conjectures that Leibniz offers in the New Essays.51 Further, even
Berkeley, who argues for the opposite view, grants that, prima facie,
the ideas of vision and of touch appear to be the same.52 So Kant's
starting assumption is, and would have seemed, quite plausible. Never-
theless, if he is mistaken about our empirical capacity for spatial per-
ception, then his conclusion that there is one form of outer sense will
be in error. At the end of the chapter I consider contemporary research
that bears on the assumption that all outer intuitions represent objects
in the same spatial framework. I will also try to assess the soundness of
his second opening assumption (see A22/B37, cited earlier): Through
outer space sense, we are able to assign all objects a definite position
in space relative to other objects.

The Isolation Argument

I read the opening pages of the Transcendental Aesthetic as offering
the following argument. In light of his own and his predecessors' re-
flections on spatial perception, Kant takes it as given that shape and
size of perceived objects cannot be traced to vision or touch, and hence
not to sensation at all. These properties therefore indicate a pure or a
priori0 product form of intuition, a feature of intuitions that does not
derive from sense impressions. Since a pure product form requires a
 The Science of Sensibility 45

pure a priori0 process form, it follows that we have some faculty that
is necessary and sufficient for the production of full spatial intuitions
out of two-dimensional sensa. Thus, the regularity we observe in our
perceptions, that all perceived objects are located in space, derives from
our faculty of perception itself. We now understand the origin of spatial
perception, and so understand its nature. It is not a simple registering
of information about objects beyond ourselves, but a reflection of our
own ways of perceiving. Further, having followed Kant's analysis of the
task, we know a priori^ that, as long as we are able to perceive normally,
all objects that we perceive will be locatable in one all-embracing spatial
network. As we shall see in Chapter 6, this result of the Aesthetic
provides an essential premise for the argument for the special status of
the categories.
On my interpretation, the first few pages of the Aesthetic offer both
the basic outline of Kant's theory that space is the form of intuition and
the major perceptual arguments in favor of his position. This material
is presented with shocking brevity, because other vital issues need to
be addressed in the rest of the text: the resolution of the debate over
the nature of space and the defense of the a priori status of geometry.
Interwoven with discussions of these topics are further reflections on
the transcendental psychology of spatial perception. I will look at the
first two arguments of the Metaphysical Exposition53 and at the Tran-
scendental Exposition in order to fill out Kant's analysis of our capacity
for spatial perception and to illuminate these notoriously difficult texts
by considering them in relation to transcendental psychology.

Two Arguments of the Metaphysical Exposition

The Standard View

The usual approach to the Metaphysical Exposition takes the first two
arguments to be aimed at establishing the apriority of our representation
of space, the last two to be demonstrating its intuitive character. Al-
though the logic of this interpretation is obvious, commentators have
been hard pressed to find material in the first two arguments that can
serve as reasonable premises for the intended conclusion. If we reflect
on the controversy about spatial perception and on the solution offered
in the opening discussion of pure forms, then we can look at these
passages in a new way. Kant has already presented a major argument
for the claim that we know a priori^ that space is an a prlorl0 form of
46 KANT'S TRANSCENDENTAL PSYCHOLOGY

sensible intuition. What I will argue is that these subsequent discussions

draw on that result in disclosing unusual features of the representation
of space.

The First Argument

Kant begins with the claim that the representation of space is not an
empirical concept. This denial should be read in the same way as the
denial that Ronald Reagan is a black woman. Both descriptions are to
be rejected. The focus of his analysis is spatial perception. The expo-
sition is hampered by a terminological problem, however. Leibniz had
claimed that the space encountered in perception is not really in per-
ception but is an intellectual abstraction from perception. Thus, Kant
cannot straightforwardly focus on spatial perception, as opposed to the
conception of space, without begging the question against Leibniz. The
resulting discussion is somewhat contorted because he feels he must use
the neutral expression, the "representation" of space.
Here is the argument:
Space is not an empirical concept which has been derived from outer
experiences. For in order that certain sensations be referred to something
outside me (that is, to something in another region of space from that in
which I find myself), and similarly in order that I may be able to represent
them as outside and alongside one another, and accordingly as not only
different but as in different places, the representation of space must be
presupposed. The representation of space cannot, therefore, be empiri-
cally obtained from the relations of outer appearance. On the contrary,
this outer experience is itself possible at all only through that represen-
tation [A23/B38].
The argument seems to be the following. The concept of space is not
empirical, because I can abstract this notion from an experience only if
I represent that experience as spatial; but representing the experience
in spatial terms requires that I already have the representation of space.
Despite appearances, this cannot be the intended argument. For this
simple argument could be used, mutatis mutandis, to show that the
concept "horse" is not empirical either. The concept "horse" is not
empirical because I cannot extract this notion from any experience unless
I represent the experience in equine terms, and this requires that I
already have the representation "horse."
An obvious disanalogy with the "horse" example is the ubiquity of
spatial representation. Kant's claim is that in order to represent any-
thing at all as different from myself, I must represent it in a spatial
 The Science of Sensibility 47

location diverse from my own. Unlike more limited representations, I

could have no experience of any external object, or no experience of
any object as external, if I lacked a representation of space. Even if
we grant this point about our representation of the external world, it
is not clear that this line of argument can support the claim that
space is a pure process form of intuition. As numerous critics have
pointed out, the representation of space could be necessary for expe-
rience of external objects, and it could be acquired through our sen-
sing of external objects/4
If we bear the preceding discussion in mind, then we can find a more
plausible reconstruction of Kant's reasoning. The argument has the form
of an implicit reductio. Suppose that we acquire the concept of space
from outer experience. We can acquire the concept of space from outer
experience only if our outer senses can register spatial properties of
distance, size, and shape. As we have seen, however, everyone ac-
knowledges that these properties are not registered by vision. Further,
Berkeley's suggestion that touch provides perceptual information about
distance appears to be contradicted by actual or Gedanken experiments.
So the perceptual representation of spatial features must have some
source other than the senses. The space of perception is not an empirical
representation at all, because this representation was never in the
senses.55
Kant makes it difficult for readers of the Critique by stopping the
argument once he has gotten to the point of showing that an empirical
derivation of the concept of space requires a prior representation of
space. The parallel argument in the Dissertation is more explicit. He
shows the dependence on the prior representation and then notes that
things in space affect the senses, "[but] that the space itself cannot be
gotten out of the senses."56 Why is this crucial element left implicit in
the later work? Kant has already reminded his readers that spatial prop-
erties cannot be traced back to the senses in the discussion of pure form,
a discussion that has no equivalent in the Dissertation. The method of
isolation serves to raise this point to a methodological principle in the
construction of theories of mental functioning. Presumably, Kant did
not feel his readers to be in need of further reminders.
Despite the abbreviated presentation and the terminological difficulty
noted earlier, Kant's argument against an empirical account of the rep-
resentation of space is fairly straightforward and quite reasonable. Given
the well-established and well-known facts about spatial perception then
current, it follows that our representations of the space of perception
are not empirical in the sense required by the Empiricist Credo.
48 KANT'S TRANSCENDENTAL PSYCHOLOGY

The Second Argument

The second argument is notorious because Kant appears to be resting
his account of spatial perception on an experiment in imagination:
Space is a necessary a priori representation, which underlies all outer
intuitions. We can never represent to ourselves the absence of space,
though we can quite well think it as empty of objects. It must therefore
be [Er wird also] regarded as the condition of the possibility of appear-
ances, and not as a determination dependent upon them [A24/B39].
Kemp Smith makes the imaginative exercise reading seem inevitable by
the way he renders the beginning of the third sentence. He translates
"also" as "therefore," which strongly suggests that the sentence is a
conclusion that is supposed to follow from the preceding sentence. Fur-
ther, he adds a gratuitous "must" that has no warrant in the German.
In this context, however, "also" could well function as an adverb rather
than as a conjunction, in which case, it should be translated as "so,"
"in this way," or "in this manner." Kant's sentence should be rendered
something like: "It [space] is in this way regarded as the condition of
the possibility of appearances, and not as a determination dependent
upon them." Thus, I believe that the logical relations among Kant's
claims are not what the Kemp Smith translation implies. First, Kant
characterizes [the representation of] space as a necessary a priori rep-
resentation that underlies all outer intuitions. Next, he notes that we
can never represent the absence of space, though we can represent it
as empty of objects. And, in the crucial third sentence, he points out
that, in this way [i.e., as a necessary a priori representation, the para-
graph's topic], space is regarded as a condition of the possibility of
appearances, not as a determination dependent upon them.
Correct translation of this clause is essential for an understanding of
the logic of Kant's argument. Now let us consider his target. The phrase
"and not as a determination dependent upon them" is a decisive clue.
In opposition to Newton's view of absolute space, Leibniz maintained
that space was dependent upon objects. Further, although Leibniz's
concerns about space were mainly metaphysical, he did offer an account
of how we come to form the notion of space in the fifth letter to Clarke.
Our representation of space (and perhaps space itself) is an abstraction
from the relations among actual objects.57 In this passage, and others,
Kant appears to be criticizing this genetic account (see A25/B39 and
A40/B57).
He cannot just deny that the perception of space depends on objects
 The Science of Sensibility 49

without begging the question against Leibniz, so we need to consider

what argument he has available to support the conclusion. If we locate
the argument in the preceding sentence, then we are back to argument
from imagination. I suggest instead that the working premise is contained
in the first sentence: "Space is a necessary, apriori[0andK] representation
that underlies all outer intuitions." Kant is making the straightforward
point that the result he has just established in the isolation argument—
that our representation of space requires a pure process form of intui-
tion—shows that the Leibnizian doctrine that the representation of space
is obtained by abstraction from the actual relations among objects cannot
be maintained.58
Since the isolation argument establishes that space is an a priori0 form
of intuition, the first two arguments of the Metaphysical Exposition do
not need to demonstrate that space is an a priori0 representation.
Rather, they are polemical explorations of the implications of this point.
Specifically, in the first argument, Kant points out that an empirical
derivation of the concept of space is not possible. In the second, he
extracts the conclusion that Leibniz's theory that the representation of
space derives from objects is also incompatible with the facts he has
established. Thus, Kant argues that in light of his analysis of the task
of spatial perception, the two influential rivals to the forms of intuition
doctrine must be given up.

The Transcendental Exposition

The Role of Geometry

Kant elevates the discussion of geometry from one argument among five
in the A edition to an independent section in B. In the Prolegomena
the question of the status of mathematical propositions dominates the
account of the forms of sensibility. Thus, it has seemed to many that
Kant's interest in the status of Euclidean geometry (hereafter Geometry)
was the spur for his theory of space and, a fortiori, for his theory of
sensibility. In Kant and the Claims of Knowledge, Paul Guyer offers
impressive evidence to show that the a priori status of Geometry was
the driving force behind Transcendental Idealism itself.59 This issue is
extremely difficult. The applicability of Geometry to the empirical world
was unquestionably a central concern. Nevertheless, I believe that its
importance has been overestimated.
Even in the second edition, where Geometry is much more prominent,
50 KANT'S TRANSCENDENTAL PSYCHOLOGY

it enters the discussion only after Kant has argued that space is a form
of intuition. The order suggests that the implications for Geometry
confirm, rather than inspire, the theory of space. Further, Ted Hum-
phrey has pointed out that Kant did not develop the Critical version of
the analytic-synthetic distinction until some time after the completion
of the Dissertation..60 Without the distinction, he cannot formulate the
problem about the synthetic a priori status of Geometry. The Critique's
theory that space is the form of intuition is, however, essentially un-
changed from the doctrine originally presented in 1770. Again, this
suggests that the case of Geometry is important confirmation of the
theory of space rather than its raison d'etre. Further, as already noted,
Kant's unalterable faith in the forms of intuition is much more under-
standable on the assumption that he believed this one theory to solve
several distinct and pressing problems.
Geometry has been assigned a preeminent role in the Aesthetic partly
because twentieth-century commentators have deep interests in the sta-
tus of mathematics and none in the theory of sensibility. Most discussions
are only concerned with whether the type of psychological theory that
Kant offers could support his claims about the a priori^1 status of Ge-
ometry. The reverse dependency is never considered. Once we recognize
that the theory of geometry makes an essential contribution to the theory
of spatial perception, it is easier to give credence to the idea that three
separate issues are brought together in one theory.

Geometry and the Space of Perception

I focus on the lengthier, more self-conscious discussion about Geometry
in the 1787 edition, which begins with an explicit statement of Kant's
objectives:
I understand by a transcendental exposition the explanation of a concept,
as a principle from which the possibility of other a priori synthetic knowl-
edge can be understood. For this purpose it is required (1) that such
knowledge does really flow from the given concept, (2) that this knowledge
is possible only on the assumption of a given mode of explaining the
concept [B40].
The Transcendental Exposition is to show that the theory of spatial
perception—specifically, the doctrine that we know a priori^ that space
is the a priori0 form of outer sense—provides insight into the synthetic
a priori^ status of Geometrical propositions, and that without that theory
 The Science of Sensibility 51

it would be impossible to understand how Geometry could enjoy this

status.
Kant's well-known argument for the latter claim, that we could never
understand how the propositions of Geometry could be synthetic a
priori^ unless we accept his theory of the perception of space, proceeds
by elimination. If the truths of Geometry are synthetic, then they cannot
be established by conceptual analysis. These propositions are supposed
to be "apriori^" however; they are necessary and universal (B41, com-
pare A24). How could we ever know that all the objects we encounter
in perception invariably have certain spatial properties? There are only
two possibilities. Either we have foreknowledge of the actual objects to
be encountered or we know something about our perceptual capacities
and limitations. Since the former option requires us to have powers we
obviously lack, the only remaining possibility is that we have a priori^
knowledge of a priori0 elements of perception (A26-27/B42-44). Fur-
ther, Kant has just shown that we do know a priori^ that all the objects
we perceive have spatial properties.
The usual and reasonable view of this argument is that it does nothing
to establish the thesis of the forms of intuition. Even if it is clear that,
if we accept the synthetic a priori^ status of Geometry, then we must
accept the theory of sensibility, that does not matter, because we do
not accept the antecedent claim. The point I wish to make is that if,
like Kant, we accept the antecedent, then this argument does not simply
provide an independent proof of the thesis established earlier. To see
this, we must first recognize that the thesis that we know a priori^, that
space is a pure a priori0 form of intuition, will not explain how the
propositions of Geometry can be known to hold of all actual objects of
perception. Our a priori^ knowledge must be more specific. We must
know a priori^ that our perceptual system is so constituted that we always
perceive objects in Euclidean space.
Where do we acquire this additional information? Standing behind
this discussion is an assumption that connects the practice of Geometry
and the space of perception. Kant begins with the uncontroversial as-
sumption that we can prove theorems of Geometry. Then he assumes—
for reasons finally spelled out in the Methodology—that geometrical
proof requires the construction of figures in perceptual space. Although
this theory strikes modern readers as bizarre, Michael Friedman has
recently argued that the appeal to construction is an unsuccessful attempt
to fill a genuine gap in the geometrical proofs of the day.62 Given these
two assumptions, we can trace a connection between geometrical proof
52 KANT'S TRANSCENDENTAL PSYCHOLOGY
and the space of perception. Although never made explicit, Kant's rea-
soning must be something like this:
1. We can prove theorems in Geometry.
2. Geometrical proof requires the construction of figures in perceptual
space.
3. Therefore, Geometrical figures are constructible in the space of
perception.
4. If Geometrical figures are constructible in the space of perception,
then the properties of that space must be compatible with the properties
described in the fundamental propositions of Geometry.
5. Any geometry whose fundamental principles are consistent with the
fundamental principles of Euclid is Euclidean.
6. Therefore, Geometry provides a true description of the space of
perception [3, 4, and 5].
Had Kant made these assumptions, then he would believe that Ge-
ometry correctly describes the space of perception. He has just shown,
however, that we know a priori^ that the space of perception derives
from a pure, apriori0 process form of intuition. It follows that Geometry
is a true description of the pure form of intuition. Whether we can now
be said also to know a prioriK that the space of perception will always
be Euclidean depends on whether something like the argument just
sketched can provide a priori^ knowledge. Kant seems committed to
this view, because otherwise the claim that the forms of intuition—and
only they—can justify the a priori^ status of Geometry would break
down.
I will not explore this issue about Geometry, because my concern is
the doctrine of sensibility. The appeal to Geometry in the Transcen-
dental Exposition enables Kant to elaborate his theory of spatial per-
ception: Given the practice of geometrical proof, it follows that the
space of perception is correctly described by the fundamental proposi-
tions of Geometry. Geometry bears exactly the same relation to the
theory of spatial perception in the Dissertation. It describes the spatial
relations that constitute the form of sensible intuition.63 Thus, Geometry
fills what would be a serious gap in Kant's account of spatial perception
by telling us exactly what the spatial form of sensible intuition is. This
point is not emphasized in 1787, because he is highlighting the recently
formulated quandary about the status of mathematics. Nevertheless, we
cannot understand why he believes that the theory of sensibility is the
only possible explanation of the alleged synthetic a priori status of Ge-
ometry unless we recognize the role that Geometry plays in the theory
of spatial perception.
 The Science of Sensibility 53

Parson's Interpretation
Although this interpretation enjoys considerable textual support, it is
inconsistent with prevailing views about the relation between sensibility
and Geometry in Kant. In a well-known article, Charles Parsons main-
tains that the theory of intuition can explain the universal applicability
of Geometry to objects of outer intuition only if we refrain from at-
tributing anything to the form of intuition that "is not revealed in the
way objects present themselves to us in perception." Parsons's reason-
able assumption is that if Kant is going to vindicate the universal appli-
cability of Geometry to objects of perception by appealing to a form of
intuition, then that form must be manifest in the objects of perception.64
This implies that we could determine the Euclidean or non-Euclidean
features of perceived objects through perceptual inspection. As a num-
ber of critics have noted, however, we can do no such thing.65 The
difficulty can be illustrated using one of Kant's own examples. However
carefully we peer at them, we cannot tell whether two putative straight
lines through two points are in fact straight.
On my account, the connection between Geometry and the properties
of perceived objects is not direct, but mediated by several independent
assumptions. There is no suggestion that we come to understand that
space is a form of intuition by discerning Euclidean properties in the
objects we perceive. By the time we reach the Transcendental Expo-
sition, Kant has already established that the spatial properties of objects
imply that space is a form of intuition, with no appeal to Geometry. Fur-
ther, as we have seen,66 he assumes (following Newton and Leibniz 67)
that positions in space are perfectly determinate relative to one another.
Alternatively, he assumes that the space of perception is a system of
perfectly definite positions. Thus, before Geometry takes center stage,
it is already clear that the space of perception has a determinate character
that derives from a pure apriori0 process form. Against this background,
our ability to prove theorems of Geometry and the assumption about
the role of construction in proof enable Kant to fill out the theory of
sensible intuition. Since Geometry can be done (and if this practice
requires construction), it follows that Geometrical figures are construct-
ible in perceptual space, and hence that the fundamental propositions
of Euclid are compatible with the properties of the spatial representation
that is the form of intuition. Thus, it follows that Geometry provides a
true description of the form of outer sense. (It also follows that figures
that are not Geometrical will not be constructible in the space of
perception.)
54 KANT'S TRANSCENDENTAL PSYCHOLOGY

By recognizing the work that has already been done in investigating

spatial perception, we can avoid Parsons's assumption about perceptual
inspection. Further, this interpretation is flatly contradicted by Kant's
discussion of the very objection (and example) Parsons raises:
Empirical intuition is possible only by means of the pure intuition of space
and of time. What geometry asserts of pure intuition is therefore undeniably
valid of empirical intuition. The idle objections, that objects of the senses
may not conform to such rules of construction in space as that of the
infinite divisibility of lines or angles, must be given up [A165/B206, my
emphasis].
This passage (from the Axioms of Intuition) makes it plain that the
order of argumentation that I attribute to Kant is right. At least in the
Critique,6* he does not believe that we know the propositions of Ge-
ometry to be true on the basis of perceptually inspecting actual objects.
Rather, he believes that perceived objects have certain properties, be-
cause certain propositions of Geometry are true.

Kant's Results
Although the argument I sketch connecting geometry and sensibility
would allow Kant to carry out the projects of the Transcendental Ex-
position,it is unsound. Even if Friedman is right that the constructive
view of proof is motivated, it is still false. Moreover, the description of
constructive proof eventually offered in the Methodology is remarkably
obscure. My point is not to defend all the theses of the Transcendental
Aesthetic, but to show the reciprocal relations among them. The analysis
of spatial perception tells us that outer intuition has a certain a priori0
form. Kant answers a major question raised by that account—what
exactly is this form?—by appealing to our ability to prove Geometrical
theorems. Sensibility is not merely the handmaiden to Geometry; Ge-
ometry is part of the theory of sensibility. If these reciprocal influences
are suppressed, then we cannot follow the reasoning of the Transcen-
dental Exposition. Nor can we grasp the full significance of the forms
of intuition.
Both Kant's defense of the synthetic a priori^ status of Geometry and
his elaboration of his analysis of spatial perception fail, because they
depend on his unusual and erroneous view of geometrical proof. The
implications of this failure for the status of mathematics have been widely
studied. How does it affect the issue of spatial perception? It leaves him
in an odd position. He has shown that a representation of space is the
 The Science of Sensibility 55

a priori0 form of human intuition, but he has no way of further specifying

what that representation is.69 In a more positive light, he has uncovered
an important research problem. What, exactly, is the spatial form of
human perception?

The Forms of Intuition and Contemporary Evidence

Depth Perception
As noted in Chapter 1, subsequent developments in mathematics and
physics have badly undercut Kant's metaphysical claims about space,
time, and the forms of intuition. How should his transcendental psy-
chology of spatial perception be judged in light of contemporary views
in perceptual psychology? Without trying to offer a comprehensive re-
view, I will consider two issues that appear to undermine his position.
Then I will turn to recent work that bears on two of his empirical
assumptions: (1) the assumption that there is one spatial representation
common to the intuitions produced by all outer senses and (2) the
assumption (borrowed from Newton and Leibniz) that the representa-
tion of space is a system of positions in which all objects can be located
in definite positions relative to one another.
On my account, Kant theorized that the perceptual representation of
space is a pure a priori0 form of intuition in order to resolve a continuing
puzzle about how we perceive the third dimension, and so the spatial
properties of objects. A casual reading of the history of psychology
might suggest that depth perception has now been explained, and that
the explanation is totally different from what he imagined. In 1838
Charles Wheatstone discovered the importance of binocular disparity
in the stereoscopic perception of depth.70 It turns out that we use two
slightly different two-dimensional retinal images in achieving a three-
dimensional representation of the viewed scene.
Wheatstone's discovery adds an important new resource; however, it
does not eliminate the mystery of depth perception, for two reasons.
First, although we can assign depths to objects in some cases just by
taking advantage of binocular disparity, the explanation of how we do
this is far from complete. To get a sense of the outstanding problems,
consider the fact that before our visual system can utilize binocular
disparity, it must match items in the left and right retinal arrays. Infor-
mation in the two arrays underdetermines correct matches, but somehow
the system accomplishes this task and assigns depths.71 Second, binocular
56 KANT'S TRANSCENDENTAL PSYCHOLOGY
disparity is useless at long distances, and we can perceive depth mon-
ocularly at short and long distances. One hundred and fifty years after
Wheatstone wrote, both the character of our three-dimensional repre-
sentation of space and the puzzle of how we use monocular and binocular
information to achieve it are active areas in perceptual research.72

Is the Space of Perception Euclidean?

In the Mathematical Analysis of Binocular Vision, R. K. Luneberg offers
experimental evidence for the view that visual space is not Euclidean,
but Lobachevskian.73 On the other side, R. B. Angell cites experiments
that he interprets as showing that the geometry of two-dimensional
visibles (visual appearances) is neither Euclidean nor Lobachevskian,
but Riemannian. Angell explains the discrepancy between his findings
and Luneberg's on the grounds that they are looking at different phe-
nomena: Luneberg is interested in judgments about the relations among
perceived objects: he is interested in visual appearances.74 Finally, in a
recent issue of Science, a trio of psychologists offers evidence for Kant's
view. They report on experiments in which a very young blind or blind-
folded child is taken by various routes to different items laid out in a
room. The experimental question is whether the child can then find a
new route between two previously visited items. Subjects can accomplish
the task. Although this may show certain things about the way subjects
represent space (see below), it is hard to see how it provides any support
for the authors' claim that subjects probably use a Euclidean metric.75
As this very brief review suggests, the present status of Kant's hy-
pothesis about the Euclidean nature of perceptual space is unclear.
Luneberg's pioneering effort has sometimes been cited as a refutation,
but more work needs to be done to determine the properties of our
perceptual representation of space. Unless we reach the surprising result
that our ability to do Euclidean geometry is a major clue to our rep-
resentation of space, however, the final outcome of this work will not
alter present evaluations of Kant's account. He believed that Geometry
was the royal road to knowledge of perceptual space on the basis of an
obscure theory of proof. Even if the space of perception turns out to
be Euclidean, or mostly Euclidean, he had no evidence that it must be
so. The burden of my argument has been that the interesting and well-
argued part of Kant's analysis is the claim that some three-dimensional
spatial representation is a pure form of human intuition.
 The Science of Sensibility 57

The Assumption of a Common "Outer Sense"

I noted earlier that Kant's analysis of spatial perception rests on the
empirical assumption that one system of spatial representation is com-
mon to all outer senses. For the project of transcendental psychology
to succeed, he must analyze actual empirical capacities, so an empirical
basis is appropriate. But the capacity must be real. Although the as-
sumption of one spatial representation common to all outer senses agrees
with common sense, it is nonetheless a substantial assumption. Notice
that he is not claiming that we can use information from different senses
to figure out what objects are where, say, by assigning a higher degree
of reliability to touch. He is claiming that one type of spatial represen-
tation is the form of sensibility regardless of which outer sense supplies
the sense impressions. As noted, he may have felt justified in adopting
this assumption on the basis of Leibniz reflections in the New Essays.
However, recent work offers some dramatic evidence that his assump-
tion about our empirical capacity is correct.
Paul Bach-Y-Rita and his colleagues at the Pacific Medical Center
have been trying to develop a device that will enable blind people to
perceive their surroundings.76 It consists of a 20 x 20 matrix of solenoid
vibrators mounted on the back of a stationary dentist's chair, a camera
that can be manipulated by the subject, and a movie screen. When an
image is flashed on the screen, the camera transforms the video image
into a tactile image of vibrating and nonvibrating "pixels." Subjects can
use the hand-held camera to scan the image on the screen. With practice
they can recognize such objects as coffee cups and telephones and they
can recognize the arrangement of objects.
After quite a bit of training with the device, subjects were presented
with looming images. When the tactile image was suddenly magnified
by a quick turn of the zoom lever of the camera, subjects ducked their
heads as if something were coming toward them. One of the subjects
was a blind psychology professor, who had lectured for years on the
relationship between the visual angle subtended by an object and its
distance from the observer. He was genuinely amazed to experience this
phenomenon for himself. The conditions of these experiments are ob-
viously quite different from the standard conditions of tactile perception.
Nevertheless this work clearly suggests that our sense of touch is capable
of creating spatial representations strikingly similar to those achieved
through vision. More work needs to be done in this area, but for now,
Kant's assumption of a single perceptual representation of the spa-
58 KANT'S TRANSCENDENTAL PSYCHOLOGY
tial arrangement of objects for all outer senses seems well supported by
the data.

Is Spatial Perception Determinate?

Kant makes a second, related assumption about our capacity for spatial
perception. He assumes that through perception, the shape, size, and
relative location of all objects are determined or determinable. Alter-
natively, he assumes that the spatial locations of all perceived objects
are comparable. This assumption stands behind his belief that there is
one form of outer sense. As we will see in Chapter 6, it is also critical
for his defense of the categories.771 discuss the assumption in two stages.
First, I look at the more general position that in assigning size and,
particularly,location to one object, we are tacitly fixing the spatial pa-
rameters of virtually every other object. Then I consider the stronger
claim that the spatial features of all objects are completely determinate
relative to our perceptual system.
In The Hippocampus as a Cognitive Map, psychologists John O'Keefe
and Lynn Nadel offer a defense of Kant's theory of spatial perception.78
They distinguish between two types of spatial representations. The spa-
tial position of an object can be represented in terms of a ROUTE. A
route is a set of instructions that explains how to reach the location of
the object by engaging in particular activities—turn left, go straight, and
so forth—at a sequence of landmarks. By contrast, one can represent
the spatial position of an object in terms of a MAP. A map is a set of
places that are systematically related to one another by a group of spatial
transformation rules.79 When we locate something by assigning it a place
on a map, we are implicitly locating it relative to all the other objects
in the region covered by the map. This is not the case with routes. The
Capitol and the Washington Monument could both be mentioned on a
route guide through Washington that would provide no information
about their relative location. Conversely, if both sites are located on a
map, then their relation to one another is completely determinate.
O'Keefe and Nadel's distinction provides a clear formulation of Kant's
assumption about our perceptual capacity: Our perceptual system rep-
resents the spatial location of objects in terms of maps. Our ability to
construct maps does not have to be established by careful psychological
experimentation. What O'Keefe and Nadel show is that many human
and animal perceptual tasks involve the use of spatial maps. (This more
limited conclusion is also supported by the experiments with blind tod-
dlers cited earlier.) These experiments suggest that Kant was right in
 The Science of Sensibility 59

his basic assumption about the way in which spatial perception works.
Our perceptual system appears to represent the spatial position of ob-
jects in a way that does locate objects relative to all others.
Even if our perceptual system represents spatial properties in maps,
it does not follow that the spatial properties of objects are perfectly
determinate with respect to our perceptual system. My point is a simple
one. Suppose that perception involves locating objects in a general sys-
tem that coordinates all positions. Still, we cannot assume that all the
particular mechanisms subserving this function will operate optimally.
Different mechanisms could operate under different conditions. There
is no reason to believe that the operations of these systems are mutually
exclusive or exhaustive. Thus, the spatial properties of a perceived object
might be indeterminate or ambiguous. Some recent work suggests that
this may be the case.80 However, even if these results are confirmed,
they will not necessarily undermine Kant's assumptions or claims. As
he realized, transcendental psychology involves only very abstract de-
scriptions of cognitive functioning. So long as the assumption that the
perceived spatial properties of objects are determinate is a reasonable
idealization, his claim for one form of outer sense will also be a reason-
able idealization, even if the particular subjective conditions (i.e., the
actual mechanisms) in which that form is realized fall somewhat short.

Discovering the Forms of Intuition

Kant's analysis of spatial perception has almost always been regarded

as a major contribution by perceptual psychologists. In part, this is
because he has been misread as a simple nativist. So his views were
highly influential in the nineteenth-century battles between nativism and
empiricism. Current research includes several programs that are Kantian
in name or spirit. Some of these also rest on oversimplifications of what
views really are. Despite these errors, however, psychologists have a
much more realistic assessment of Kant's contributions to perception
than philosophers. The most valuable and distinctive aspect of his work
is precisely that most repugnant to his twentieth-century commentators:
the thesis of transcendental psychology that there is a specific a priori0
form of human sensibility.
Although he resisted any specific nativist hypotheses, he recognized
that the basic spatial form of perceived objects can come only from the
human perceptual system itself. Spatial perception seems so familiar
that we tend not to think of it as a phenomenon at all. His analysis
60 KANT'S TRANSCENDENTAL PSYCHOLOGY
shows that it is not diaphanous, but has distinctive properties of its own.
On the basis of the practice of Euclidean geometry, he believed that he
knew what the essential properties of spatial perception were. That belief
was ill founded. It is a difficult experimental problem to determine the
properties of our spatial representations of objects. Kant did not discover
the a priori0 form of outer sense, as he thought; he discovered the
problem of the a priori0 form of outer sense. If psychologists can specify
the inherent properties of human spatial perception, then they will dis-
cover the a priori0 form of outer intuition.
Since this project should be carried out by psychologists, there may
be a temptation to think that it no longer has any philosophical signif-
icance. As he originally conceived of it, Kant's doctrine of the forms of
intuition had implications for philosophy of mathematics, the meta-
physics of space and time, metaphysics generally, and epistemology. It
is doubtful that the consequences of more accurate accounts will be as
numerous or as momentous. Nevertheless, they will inform us about
the limitations and propensities of our perceptual access to important
aspects of reality. Such theories are standardly believed to have impor-
tant epistemological implications. So, for example, Thomas Kuhn's
claims about the limitations of perception provided half of the support
for his revolution in philosophy of science (incommensurability of lan-
guage was the other half).81 We cannot hope to determine the philo-
sophical implications in advance of results and without substantive
philosophical theories of the role of perception in various epistemolog-
ical enterprises. Since the results will concern the bounds of sense,
however, they should have important implications about the nature and
limits of our knowledge.
 3
Transcendental Psychology in the
Transcendental Deduction

What (If Anything) Happens in the Deduction Chapter?

Rumor has it that when a well-known Harvard Professor taught the

Critique for the first time, he reached the second chapter of the Analytic
of Concepts—"The Deduction of the Pure Concepts of Understand-
ing"—and postponed classes for two weeks while trying to figure out
something to say. This chapter evokes the same incoherent mixture of
reverence and disdain among Kant scholars that the Critical philosophy
as a whole evokes among nonspecialists. It is widely regarded as (1) the
philosophical heart of the book; (2) difficult to the point of being un-
intelligible; and (3) utterly barren of plausible or interesting philosoph-
ical positions or arguments. Not surprisingly, commitments to (2) and
(3) have led more tough-minded Kantians to doubt the viability of (1).
Over the years, the value of this chapter has repeatedly been ques-
tioned, at least by Anglo-American scholars.1 Kemp Smith promulgated
Erich Adickes's and Hans Vaihinger's incredible explanation for its
alleged failings: the chapter was quickly pieced together from notes that
included doctrines that Kant had already [sic] abandoned, as well as
newly worked out positions.2 In attempting to defend against the "patch-
work" charge, Paton famously compared the difficulty of following its
reasoning to crossing the Great Arabian desert by foot. (To avoid con-
fusion, I will capitalize "Deduction" when referring to this chapter, and
use the lowercase to refer to the deduction as a whole.) Jonathan Bennett
cut through the exegetical niceties of the patchwork debate: "The De-
duction is not a patchwork, but a botch."3 As noted in Chapter 1,
61
62 KANT'S TRANSCENDENTAL PSYCHOLOGY
Strawson recommended that we skip the doctrine of synthesis, which is
a central topic of the chapter. His analysis of the deduction focused
instead on the Refutation of Idealism and on the Analogies, where he
saw anticipations of the antiskeptical "transcendental" arguments of the
1950s. Most recently, Guyer has dismissed the arguments of the De-
duction chapter, again finding solid argumentation only after reaching
the Principles chapter.4 Even as sympathetic an interpreter as Allison
concedes that the Deduction "fails to demonstrate how the categories
make experience possible, [and so seems to lose]... much, if not all,
of its philosophical significance."5
The suspicion that nothing of serious philosophical interest transpires
in the Deduction chapter is the result of two independent factors. First,
Kant's organization leads to mistaken expectations. In the "Clue to the
Discovery of All Pure Concepts of the Understanding," widely referred
to as the "Metaphysical Deduction," he introduces the list of categories.
At the beginning of the Deduction, he explains that the categories re-
quire a "deduction," or a proof of their legitimacy. Readers expect that
they are about to be given such proofs, but no proofs appear. Most of
the categories are not even mentioned, much less argued for, in the
chapter. Specific arguments about individual categories can be found
only in the Metaphysical Deduction and in the Principles. This suggests
that the Deduction plays only a supporting role in the main enterprise.
It must complete some unspecified, unfinished business from the Meta-
physical Deduction, or provide preliminaries for the arguments of the
Principles.
The Deduction chapter performs both these functions, as I argue in
Chapter 6.6 That is not its only significance, however, despite the mis-
leading implications of the sequence of Kant's discussion. His remarks
in the preface to the First Edition provide a more accurate picture of
its importance: This chapter cost him the greatest labor and is of un-
rivaled importance as an exploration of the faculty of understanding
(Axii). As the Transcendental Aesthetic examined sensibility to discover
a priori0 elements in cognition, the central purpose of the Deduction
is to reveal the apriori0 contributions of the understanding. It does this
by analyzing very basic empirical capacities (e.g., representing objects
and making judgments about objects) in order to determine what fac-
ulties we must possess and what processes we must carry out, given that
we are able to perform these tasks. Only when these broad analyses are
complete is Kant in a position to argue that certain concepts are indis-
pensable to, and invariant features of, our cognitive lives.
The second reason for the low status of the Deduction chapter is that
 Transcendental Psychology in the Transcendental Deduction 63

scholars avoid these analyses. Nowhere has the antipsychologistic ap-

proach to Kant had more damaging effects than on interpretations of
this chapter. Chapter 2 argued that we cannot appreciate the strength
of the Transcendental Aesthetic unless we consider the analysis of spatial
perception. Even in the absence of this material, however, scholars could
evaluate the theories of space, time, and mathematics. With the De-
duction chapter, there is nothing but transcendental psychology. Since
it is off limits, commentators can only sift through what remains to find
other threads either to recast as analytical arguments7 or to tie to some
argument about the categories in the Metaphysical Deduction or in the
Principles. As a result, there is little interpretive consensus and the
chapter appears confused and disjointed, a paradigm of futile Teutonic
obscurity.
Kant's early readers also had difficulty with the Deduction, but that
is hardly surprising, given the novelty and abstractness of the arguments
and, in some cases, a lack of desire to understand.8 There are also
substantive problems. In analyzing specific tasks, the neat dichotomy
between sensibility and understanding breaks down. Further, analyses
of several different tasks are used to establish the mind's contributions
to knowledge. Nevertheless, in this and the next three chapters, I will
argue that Kant has coherent lines of argument in both editions, ar-
guments that begin with insightful analyses of cognitive tasks and end
with claims to justify a special status for the categories. In the end (in
Chapter 6), it will turn out that even the best argument for the categories
fails, in part, because it contains a mistake in transcendental psychology.
That does not vitiate all that comes before, however. I will argue that
Kant's analyses of the tasks of representing objects and of judging them
demonstrate the necessity of various synthesizing activities of the mind.
They thus provide important evidence for the general thesis of tran-
scendental idealism that what we know is partly a function of our ways
of knowing. That is why Kant labored over this material and why the
chapter is the heart of the book. These analyses and the resulting doc-
trine of synthesis also enable him to provide an account of mental unity
that is both an effective reply to Hume's skepticism and, as I will argue,
significantly better than more recent accounts.
In this chapter I focus on two task analyses that are the centerpieces
of the objective side of the Deduction in the two editions: How are we
able to represent objects? How are we able to make judgments about
objects? Chapter 4 follows the central task analysis of the subjective
side of the Deduction: What must a subject be like to have represen-
tational states? This analysis and the earlier ones result in an account
64 KANT'S TRANSCENDENTAL PSYCHOLOGY

of mental unity that I explore and develop in Chapter 5. Finally, Chapter

6 brings these analyses and a new task analysis (How are we able to
perceive spatial and temporal arrays of objects?) together in one long
argument for a special status for the categories. Almost all my discus-
sions of the transcendental deduction deal with material from the De-
duction chapter. For reasons already given, interpretations of the
deduction often race lightly through this chapter and then concentrate
on the more compact arguments of the Principles, chiefly the argument
of the Second Analogy. My interpretation does the opposite. The de-
duction is not completed until the Principles, so I offer a brief account
of the Second Analogy at the end of Chapter 6. However, this is pre-
sented only as a coda to an interpretation that is firmly centered on the
Deduction chapter.
If we allow ourselves to consider Kant's psychological analyses, then
we can follow the argumentation of this chapter. Further, we can make
a more realistic assessment of its worth. When this material is ignored
and the chapter is seen only as an obscure lemma for a proof of the
categories in the Metaphysical Deduction or in the Principles, then a
mixed verdict is precluded. Either the proof (of the other section) works
or it does not. By attending to the details of Kant's analyses, we can
determine whether his laborious exploration of the faculty of under-
standing yielded any positive results. To what extent, if any, can he
justify his revolutionary claim that the mind itself is responsible for
important elements of cognition? Do his analyses provide insights of
enduring value into different aspects of cognition? Although this is the
best reason for reconsidering transcendental psychology, I will just add
that this approach also permits quick resolution of three classic in-
terpretive puzzles associated with the chapter: Is the argument "re-
gressive," because it merely analyzes the necessary conditions for
knowledge,9 in a strong sense of "knowledge," or "progressive" in ar-
guing from weak premisses to a strong conclusion? What are the relative
merits of the first and second versions of the Deduction chapter? What
is the relation between the "subjective" and "objective" sides of the
chapter, and is the former really necessary?
Kant analyzes a variety of cognitive tasks. Sometimes he makes strong
assumptions about our capacities; other times, quite minimal assump-
tions. So the movement of argument is neither consistently progressive
nor consistently regressive. This approach also provides a systematic
account of the differences between the first and second versions of the
Deduction chapter, in terms of the different cognitive tasks that they
analyze and the comparative successes of the analyses.
 Transcendental Psychology in the Transcendental Deduction 65

Kant's prefatory admission that only the objective Deduction is es-

sential to his purpose (Axvi) is often cited as a means of throwing out
transcendental psychology with the subjective Deduction, as I note in
Chapter 1. As I argue there, however, his ambivalence toward the
subjective Deduction rests on an understandable confusion between
providing a highly abstract description of capacities a mind must have
to carry out various cognitive tasks and speculating about the psycho-
logical mechanisms that realize those capacities.10 Even if the subjective
Deduction were eliminated, however, transcendental psychology would
remain. The objective Deduction aims to establish the legitimacy of the
categories by showing that we must employ these concepts if we are
going to be capable of having very basic cognitive relations to objects.
This project leads Kant to explore the prerequisites of cognition, which
inevitably involves discussion of the processes of cognition, whether or
not he assigns them to particular faculties. Thus, the objective and
subjective Deductions are two sides of one enterprise. On the objective
side, we start with different forms of object cognition and try to decom-
pose these tasks into their essential subtasks. The subjective Deduction
examines the same problem from the point of view of the subject: What
capacities must a subject have to enjoy various cognitive relations to
objects?

Representing Objects

The Problem
The objective side of the Deduction examines how in general (Kant's
ubiquitous "iiberhaupt") objective knowledge, or knowledge of ob-
jects, is possible for creatures with our basic cognitive constitution.
Some discussions offer epistemic analyses of the conditions required
for valid, intersubjective knowledge claims.11 Since these have been
widely discussed, my efforts are directed toward the more obviously
psychological analyses of empirical capacities. Among these, I focus
on the most basic and most central task analysis of each of the two
editions. These analyses start with very minimal—and actual—capaci-
ties and are fairly sound. So they offer the best defense of the claim
that the mind influences knowledge and the strongest foundation for
an argument that certain concepts are and must be invariant features
of cognition.
We can know an object only if we can represent it and the first edition
66 KANT'S TRANSCENDENTAL PSYCHOLOGY

highlights this most basic subtask: How is an object of representations

possible (A104ff)? To understand Kant's analysis, we must begin by
anchoring his fluid terminology. In German philosophy of this time, the
most general term for a mental state was "Vorstellung" which is usually
translated as "representation." However, Kant does not believe that all
Vorstellungen (representations) represent (see A320/B376). Although
this is no more inconsistent that the contemporary claim that atoms can
be split, it can be confusing. So I use "cognitive state" as the generic
term.12 There is an important complication, however. "Vorstellung" can
be used to indicate the contents of cognitive states. It exhibits the same
ambiguity as the English terms "representation" and "idea" (which is
one reason that, from a strictly translational perspective, "representa-
tion" is a very good rendering of "Vorstellung"). When the "content"
sense is indicated, I will translate "Vorstellung" as "the contents of a
cognitive state," and where the term is ambiguous, I will use "cognitive
state or its content." Although these renderings are inelegant, I believe
that they will permit a clearer discussion of substantive issues. They also
have the advantage of keeping the essential duality of "Vorstellung"
constantly before us.
In the Logic, Kant claims that there are two types of cognitive states
that "consciously refer to objects": intuitions and concepts. He gives
these referring states the common name "Erkenntnisse"13 which might
be translated as "cognitions" (but which Kemp Smith barbarously trans-
lates as "modes of knowledge"). Because it is more perspicuous, I will
describe these representational cognitive states as "representations."
And, again, where the context indicates, I will sometimes use "the
contents of representations." Where "Erkenntnis" is used as a mass
noun, I will render it as "cognition" or "knowledge." In this terminol-
ogy, one project of the first edition Deduction is to consider what is
required for a cognitive state to be a representation.14 In Brentano's
well-known terminology, Kant is trying to understand how a mental
state could be "intentional."
Like most topics Kant discusses, this issue arose in the context of
contemporary debates. As we saw in Chapter 2, however, he rarely
gives any space to presenting previous positions, even when they are
essential to an appreciation of his views. (We will see an even worse
example of this expositional shortcoming in Chapter 4.) Kant's analysis
of the task of representing objects is clearer when viewed against the
contemporary options that he rejects and (apparently) accepts. So before
turning to that analysis, I will consider where his predecessors and con-
temporaries stood on the issue.
 Transcendental Psychology in the Transcendental Deduction 67

The Views of Leibniz and Christian Wolff

Leibniz made "perception" a central notion in his metaphysics. Monads
are the basic constituents of reality and they alter by having different
perceptions. Monads teem with perceptions, including petite perceptions
of which they are unaware. Through these masses of perceptions, each
monad mirrors the entire universe. Although Leibniz is explicit about
the distinction between petites perceptions and conscious perceptions,
he says very little about the nature of perceptions.15 Four points seem
clear, however. First, Leibniz believes that mental states are in active
causal community with each other. Indeed, all previous mental states
have some effect on subsequent states.16 Second, given the mirroring
analogy, Leibniz must regard perceptions as representations. They mir-
ror or represent the universe. Tetens believed that the Leibniz-Wolff
concept of representation implied that every modification of the soul is
a representation.17 However (third), since monads do not causally in-
teract (they are "windowless"), perceptions must represent states of the
universe without being in causal contact with what they represent. Fi-
nally, perception involves unity. For the multifarious contents of a per-
ception are represented together in a simple substance.18
Christian Wolff offered a systematic presentation of Leibniz's widely
scattered philosophical writings. On the topic of representation, Wolff's
account was particularly neat. He distinguished the "powers" of the soul
from its various capacities. The soul's fundamental power is the vis
representiva, the power of representation. Allegedly, the unity of the
soul and the commonalities across its various expressions prove the
uniformity of this power. The task that remains is to show how each
mental capacity could be reduced to the vis representiva.19 Alternatively,
the task was to show how each capacity could be deduced from the
definition of the essence of the soul, the vis representiva.
For reasons that will be explored in detail in Chapters 7 and 8, Kant
could not accept this kind of account. The operations of actual capacities
could not be explained by appealing to definitions, even definitions
purporting to capture essences. From Kant's perspective, Leibniz and
Wolff make the question of representation fundamental, but they fail
to provide an adequate solution to it.

Empiricism and Sensationism

Turning from the Leibnizeans to the Lockeans, we find a diametrically
opposed account of representation. Mental states represent the objects
68 KANT'S TRANSCENDENTAL PSYCHOLOGY

that cause them. Objects cause impressions in the mind that resemble
the objects themselves, so the question of how they represent is an-
swered in the causal story. They represent objects by resembling them.
By Kant's time, this seemingly more plausible account had been badly
shaken. According to Locke himself, there was no resemblance between
property and idea in the case of secondary qualities. As we saw in the
discussion of Berkeley in Chapter 2, however, it was not clear to Locke's
successors that we directly perceive primary properties, like shape and
distance. Rather, we seem to perceive objects by perceiving their sen-
sible, that is to say, secondary qualities, which do not resemble anything
in the object itself. So how do our mental states represent objects?
In an illuminating article, Rolf George makes a persuasive case that
sensationism formed part of the background to Kant's epistemological
discussions.20 Sensationism is the position that the sensations produced
by objects impinging on the senses are nonreferential. George credits
Malebranche as the original discoverer of this position but suggests that
it may have come to Kant via Condillac. Kant must have been somewhat
familiar with Condillac's views, because Tetens discusses them.21 Fur-
ther, the distinction between nonreferential sensations and referential
perceptions was central to Reid's philosophy, which was quite influential
in Germany at the time Kant was writing.22 In describing sensations as
the ways we are affected by objects (A19-20/B34) and in denying that
sensations are objective representations, Kant seems to adopt the sen-
sationist position.
Given the sensationist premise, the problem of how cognitive states
represent becomes pressing, since mere causal interaction is insufficient.
Reid did not really offer a solution to the problem. His primary concern
was to make philosophers recognize the "commonsense" distinction
between sensation and perception, in order to blunt the force of skep-
ticism.23 George suggests that Kant's own approach may have been
influenced by Condillac's view that reference to objects is achieved by
various sensations being pulled together in a judgment.24 I believe that
something like Condillac's position that perception involves judgments,
or Reid's position that perception involves concepts, does shape the
discussion in the A Deduction. It also undermines the analysis, however.
As I argue below, the problem is that these assumptions were highly
controversial, so Kant had no right to make them in analyzing the task
of representing objects. Despite this miscue, however, many aspects of
his analysis depend only on well-established and much less controversial
assumptions, and these can be preserved.
 Transcendental Psychology in the Transcendental Deduction 69

Associationism
Finally, let's consider Hume. In the Inquiry Concerning Human Un-
derstanding (which Kant owned in German translation25), Hume ex-
amines our belief in objects. Although he casts the issue in terms of
belief, not reference, the discussion covers the relevant territory. How
do we come to believe that an object is present?
Whenever any object is presented to the memory or senses, it immediately,
by the force of custom, carries the imagination to conceive that object
which is usually conjoined to it; and this conception is attended with a
feeling or sentiment [namely]... a more vivid, lively, forcible, firm, steady
conception of an object than what the imagination alone is ever able to
attain.26
For example:
I hear at present a person's voice with whom I am acquainted.... This
impression of my senses immediately conveys my thought to the person,
together with all the surrounding objects. I paint them out to myself as
existing at present, with the same qualities and relations of which I for-
merly knew them possessed.27
Although resemblance enters into Hume's account, the crucial role
is played by the law of association. How, on the basis of sensation, can
we achieve reference to an object? Objects have multiple properties.
Each property causes a resembling impression in the mind. Since these
impressions are experienced together, the ideas produced by them are
bound together by the law of association. Again through the operation
of the law of association, when a fresh instance of one of these impres-
sions is in the senses, it brings these ideas to mind and conveys its own
inherent liveliness to them. So when we hear the voice of an acquain-
tance, we immediately have a lively conception of the person's face,
build, manner, and so forth. That is all there is to belief in an object.28
Hume expresses satisfaction with his account: "Here, then, is a kind of
pre-established harmony between the course of nature and the succes-
sion of our ideas."29 In nature, various properties and objects are linked
together. Thanks to the operations of custom, these properties and
objects are also connected in our thoughts.
But Hume knows that this simple story does not work. In his earlier
and more extended treatment in the Treatise, he recognized that the
law of association is inadequate to explain our beliefs in objects.30 The
properties of objects are constantly, if subtly, changing. Further, our
70 KANT'S TRANSCENDENTAL PSYCHOLOGY

impressions are in constant flux. How do such irregular impressions

produce stable clusters of ideas to be called up by a present impression?
Kant notes the same weakness in the law of association: "If. .. cognitive
states reproduced one another in any order, just as they happen to come
together, this would not lead to any determinate connection of them,
but only to accidental heaps... " (A121, my underscoring, amended
translation). To make up for this defect in the operation of association,
Hume assumes that the imagination (through various propensities) pro-
duces more uniformity in the connections among ideas than exists in
the connections among impressions.31 Here he is considerably less san-
guine about his account of belief in objects and commends carelessness
and inattention as good alternatives to delving into these perplexities.32
Representation was a central issue for the Leibnizeans and the Lock-
eans, and for the lesser-known school of sensationism. By the time Kant
was writing, two classic solutions to the problem, the Empiricists' re-
semblance assumption and Wolff's via representiva, were no longer vi-
able options. On the issue of causation, Leibniz's successors took the
reference of cognitive states to objects to involve a causal relation.33
This view was shared by Kant from his earliest writings and spelled out
clearly in a passage in the Reflexionen:
Leibniz takes all sensations (deriving from) certain objects for represen-
tations of them. But beings who are not the cause of the object through
their cognitive states [or their contents] must in the first instance be af-
fected in a certain way so that they can arrive at a representation of the
object's presence. Hence sensation must be the condition of outer cog-
nitive states [representations?] but not identical with i t . . . . Hence rep-
resentation is objective, sensation subjective.34
So causal connection is necessary for representation, but not all cognitive
states caused by objects represent.

A Priori Necessary Synthesis

What Is an Object of Representations?

Kant introduces his analysis of the task of representing objects by asking
what it means to talk about an "object of representations" (A104).
Although this task is, in some respects, very minimal, he does assume
that we are able to represent objects. So this part of the Deduction is
not intended to address highly skeptical opponents. His conclusions
 Transcendental Psychology in the Transcendental Deduction 71
should only be accepted by those who grant our capacity to represent
objects. Besides this essential empirical assumption, his analysis rests
on two widely shared background assumptions (and, as noted, on a
controversial assumption to be discussed further below). First, the prob-
lem of representing an object should be cast as the problem of how
various cognitive states can yield a representation of an object. This
assumption derives support from his theory of perception, which I dis-
cuss in Chapter 6. However, the underlying rationale rests on contem-
porary scientific views that were accepted by both the Leibnizean and
the Lockean traditions. Vision is the paradigm sense and our retinas
receive a constant and varied stream of stimulation.35 Hence, Hume
describes the mind as an inner theater that witnesses constant flux,36
and Leibniz refers to a myriad of petites perceptions. So the problem of
representing objects is really the problem of how we can represent
objects on the basis of a varied and fluctuating stream of cognitive states.
The second, related, well-entrenched background assumption is that
the simulacra theory—the view that we perceive objects by their giving
off simulacra of themselves which (somehow) migrate into our minds—
is false.37 Kant refers to this empirical fact explicitly in the Prolegomena:
For I can only know what is contained in the object in itself if it is present
and given to me. It is indeed even then inconceivable how the intuition
of a present thing should make me know this thing as it is in itself, as its
properties cannot migrate into my faculty of representation.38
Our senses do not take in whole objects or their properties through
some type of migration. Rather, we can derive information about objects
only by the effects of their various sensible properties on our various
sensory organs.
Kant begins his analysis with an apparently simple point. To represent
an object, a set of cognitive states must be consistent and coherent:
Since the contents of representations [Erkenntnisse] are to relate to an
object, they also agree and relate to one another in a necessary way, that
is, they must have that unity which makes up the concept of an object
[A104-5, my translation].

If one of my representations represents the desk before me as black and

another represents it as brown, then they cannot both be (accurate)
representations of the same object (at the same time). However, this is
merely a superficial reflection of a deeper point, as becomes clear in the
next paragraph. To yield a representation of an object, cognitive states
must be united in an overarching representation.
72 KANT'S TRANSCENDENTAL PSYCHOLOGY

Kant continues by noting that although unity is necessary for repre-

senting objects, we cannot derive it from objects. Objects are nothing
to us, because they correspond to cognitive states, and so should be
distinguished from them (A105). "It is clear that. .. we have only to do
with the diverse elements [Mannigfaltigen] of our cognitive states [Vor-
stellungen]..." (A105, amended translation). A similar point is given
prominent notice in B: "Combination . .. cannot be borrowed from . . .
[objects], and so, through perception, first taken up into the under-
standing" (B134). To represent an object, we must unify information
from multiple cognitive states from various sensory modalities. Since
the goal is to explain how we achieve a unified representation (in our
heads or minds), it is unavailing to appeal to objects outside us.
Although this point may seem painfully obvious, it is worth dwelling
on for a moment, since it is so often missed. So, for example, the
contemporary "ecological" school of perception founded by J. J. Gibson
makes exactly this error. Gibson was concerned to show that the en-
vironment is a richer source of information than many perceptual psy-
chologists had realized. In particular, he argued that the environment
contains a variety of high-level invariants, such as ratios and proportions,
that do not alter under local changes in optical stimulation. The rec-
ognition of higher-level environmental properties has been important
in recent work on perception. However, Gibson drew the wrong con-
clusion from his discoveries:
the function of the brain when looped with its perceptual organs is not
to decode signals, nor to interpret messages, nor to accept images. These
old analogies no longer apply. The function of the brain is not even to
organize the sensory input or to process the data, in modern terminology.
The perceptual systems, including the nerve centers at various levels up
to the brain, are ways of seeking and extracting information about the
environment from the flowing array of ambient energy.39
That is, Gibson inferred that once we have discovered invariances in
the environment, we have an account of our perception (and so, rep-
resentation) of them. But the fact that the environment contains invar-
iances does not (by itself) explain how we are able to represent them,
any more than the fact that various qualities are united in objects ex-
plains how we are able to produce representations in which these qual-
ities are united. Gibson's critics echo Kant: He has failed to explain
how we perceive invariances, because he has not explained how (by
what processes, by what means of decoding) we derive representations
of invariances from fleeting sensory information, which is all that we
receive. 4O
 Transcendental Psychology in the Transcendental Deduction 73

Unity and Synthesis

Since the object cannot impart its unity to cognitive states or their
contents, something else must unify them in a representation of an
object. Kant claims that the required unity can be created only if the
intuition (representation) "can be produced by such a function of syn-
thesis in accordance with a rule that makes possible the necessary a
priori reproduction of those diverse elements, and their union in a con-
cept" (A105, my translation). Far too much is packed into this sentence
and this doctrine. To begin to appreciate the strengths and weaknesses
of Kant's analysis, it is necessary to separate his points. This claim
contains four distinct subclaims:
1. The unity of intuitions requires a function of synthesis.
2. This function must be carried out in accordance with a rule.
3. This rule makes possible the necessary a priori reproduction of var-
ious contents.
4. Finally, this rule makes it possible for various contents to be united
in a concept.

Although none of these claims is exactly limpid, the third seems es-
pecially problematic. Why "necessary" and "a priori'"? Unless Kant is
being clumsily redundant, he cannot mean "a priori" in the sense in
which it is equivalent to "universal" and "necessary." If we look back
to his introductory remarks, and forward to his summation of his position
on representing objects, then it seems that the sense of "a priori" he
has in mind must be what I distinguish as "a priori0" in Chapter 1. We
are looking for "apriori0" conditions, which remain even "when every-
thing empirical is abstracted from experience" (A96). He concludes that
representing objects requires a "pure," "nonempirical" concept of a
transcendental object (A109). This concept cannot contain anything
definite but refers only to the necessary unity of an object representation,
which is "nothing but the necessary unity of consciousness, and therefore
also of the synthesis of the diverse elements, through a common function
of the mind, which combines it in one representation" (A109, amended
translation). That is, the unity of representations is not empirical, but
derives from the mind. Alternatively, the function of the mind that
produces unity cannot be traced back to sensations. Reading this later
explication back into A105, I think Kant's third claim is itself complex:
(3a) this rule makes possible the necessary reproduction of various
contents.
74 KANT'S TRANSCENDENTAL PSYCHOLOGY
(3b) [although not itself a priori0], the rule must involve apriori0 elements,
or relate to a priori0 elements, that enable it to produce unity.

What Is Synthesis?
Before we can make any progress with this complex analysis of repre-
senting objects, we need to understand the key technical term "syn-
thesis." Synthesis is an absolutely central notion of the Deduction. It is
mentioned more than 60 times in the A edition, and reappears in the
less overtly psychological B Deduction, both under its own name and
as "combination." As will be evident, it is not possible to depsychologize
this notion as it occurs in Kant's text. Thus, it is rarely the subject of
extended discussion, despite its great prominence.41 Much of this and
the next two chapters will deal with various aspects of the synthesis
doctrine. A fundamental difficulty in following the text of the Deduction
chapter is that this doctrine is quite complex, and Kant's discussions are
highly compressed. The complexities need to be explored seriatim. How-
ever, I will try to give some sense of the whole by beginning with a
summary presentation of a large chunk of the doctrine, even though
some claims will only receive needed detail and defense later.
Kant introduces the term "synthesis" with an explicit definition:
By synthesis, in its most general sense, I understand the act of adding
different cognitive states [or their contents] [Vorstellungen] to each other,
and of comprehending their diverse [elements] in a single representation
[Erkenntniss] [A77/B103, amended translation].
Again, the synthesis is "what gathers the elements of representation
together and unites them in a definite content" (A77/B103). As at A105,
Kant sometimes speaks of "functions of synthesis." His definition of
"function" suggests that the two notions are closely related: "By 'func-
tion' I mean the unity of the act of bringing various cognitive states [or
their contents] under one common one" (A68/B93, amended transla-
tion). This is why the depsychologizing project is futile. In these intro-
ductory passages, common to both Deductions, Kant defines "synthesis"
as an act performed on cognitive states.
A synthesis is an act, or to be more neutral, a process that produces
a representation, by adding or combining diverse elements contained in
different cognitive states in a further state that contains elements from
these states. The easiest way to think about syntheses may be to regard
them as processes that realize (mathematical) functions. Given a set of
input states, a synthesis produces a certain output state. Thus, Kant's
 Transcendental Psychology in the Transcendental Deduction 75

talk of rules of synthesis, or functions of synthesis, is, as it often seems

in the text, pleonastic. The domain of syntheses comprises cognitive
states (including representations); their range, cognitive states and rep-
resentations. Kant provides no account of the intrinsic nature of cog-
nitive states, representations, or syntheses. These matters are beyond
the interest and capabilities of transcendental psychology.42 Syntheses
might be processes that produce physical states from sets of physical
states, that produce immaterial states from sets of immaterial states, or
that produce symbols from symbolic inputs.
Kant regards representations as comprehending, or containing, the
diverse elements (Kemp Smith's "manifolds") of other cognitive states
and as possessing unity, precisely because they are produced from those
states by syntheses. That is, as I will argue in Chapter 4, he believes
that a representation represents something, in part because it was pro-
duced by sensory states that were themselves caused by the sensible
properties of the object striking the sense organs (and in part because
it can lead to further representations). In contemporary terminology,
Kant holds a "functionalist" theory of the content of representations.
What about the unity of representations? For Kant, as for Leibniz,
unity is a central feature of representations. As is clear in the Paral-
ogisms, however, he does not believe that the unity of representations
derives from their inherence in a simple substance. At one level, Kant's
position on unity is clear enough. A representation produced by syn-
thesis has unity, because the contents of various states are united in it.
This explanation is merely verbal, however. How are these contents
united? In what sense are they united? Kant's answer is, "Through
synthesis." The representation has unity, and the contents are united,
because the representation was produced through synthesis; it is, as he
says so often, a "synthetic unity."
One way to grasp Kant's point about the relation between unity and
synthesis is to consider the now familiar example of a Necker cube (see
Figure 3.1). What happens when we perceive figure (a)? We scan the
lines and vertices, and the relations among them, and on the basis of
that information, we interpret the figure as (b) or (c). What is peculiar
about this figure is that we are able to perform two processes, which
realize two different functions and so yield different outputs, on the
basis of the same stimulus values. Now consider the vertices that I have
marked 1 and 2. We can interpret these vertices as both lying on the
front plane of the solid figure, as in (b), or as both lying on the back
plane, as in (c). What cannot happen is that we interpret vertex 1 as
forward and vertex 2 as to the back. We cannot do this in the following
76 KANT S TRANSCENDENTAL PSYCHOLOGY

Figure 3.1 The Necker Cube (a) can be perceived as (b) or (c). (Drawn by
Philip Kitcher.)
sense: Such an interpretation will not produce a stable visual represen-
tation of a three-dimensional cube. When Kant describes the diverse
elements of different states as being united in a representation, or that
representation as possessing unity, what he means is that there is some
process that realizes a function that yields that representation as a stable
output,43 given the right cognitive states as inputs. (I use a visual image
only to illustrate the point. Kant takes it to apply to all states produced
by synthesis.) Representations produced through synthesis inevitably
involve unity. For if they lacked it, they would not exist (A105, B132).
With some understanding of the doctrine of synthesis, we can make
 Transcendental Psychology in the Transcendental Deduction 77

considerable sense of the first two subclaims of A105. Neither objects nor
their properties can migrate into our minds. We take in information
about objects through their effects on different senses at different times.
Because objects are outside the mind, we cannot explain how this infor-
mation is united in a representation by appealing to objects. The only
alternative is that some process within the mind—synthesis—carries out
the unification. So subclaim 1 is justified. Because a synthesis is a process
that takes some cognitive states as inputs, and yields others as outputs,
it can be described as a function, or as a rule (make Rl from inputs
CS1, CS2 and CS3). So subclaim 2 is also justified. For reasons that
will be clear later, I skip to subclaim 3b: Although not itself a priori0,
the rule of synthesis that is required to reproduce multiple contents
from cognitive states must involve or relate to a priori0 elements.
To see the significance of this aspect of Kant's analysis, it is helpful
to take stock of what has already been shown. Has he already dem-
onstrated that our capacity for representing objects can only be ex-
plained by reference to a priori0 elements? In a sense, he has.
Representation requires that the diverse elements of cognitive states be
combined. Since this combination takes place on items in the mind, it
must be carried out by an activity of the mind. So the mind itself supplies
this precondition for cognition. Still, this is not very exciting. One could
argue that the ability to be affected by sensible properties is also in the
mind and so demonstrates the mind's necessary contribution to cogni-
tion. A clearer way to cast the issue is in terms of Kant's technical notion
of a priority0: Can combination be traced back to the senses? Was the
combination or synthetic unity required for representing objects ever in
the senses?

The Law of Association

Empiricists thought it was. This was clear in the preceding discussion
of Hume. Different ideas become united through the law of association
when they are sensed together, or nearly together. At A105 Kant takes
a cryptic and backhanded, but nonetheless deliberate, slap at this po-
sition. The law of association does not explain how we represent objects;
the law is itself explained by the rules of synthesis that make reference
to objects possible: "[the synthesis must be governed b y ] . . . a rule [that]
. . . makes possible the necessary a priori reproduction of the diverse
elements" (A105, my translation).44 This allegedly fundamental law of
the mind is derivative.
The groundwork for this claim was laid in the discussion of the "Syn-
78 KANT'S TRANSCENDENTAL PSYCHOLOGY
thesis of Reproduction in Imagination." Kant accepts the law of
association:
cognitive states which have often followed or accompanied one another
finally become associated, and so are set in a relation whereby, even in
the absence of the object, one of these cognitive states can, in accordance
with a fixed rule, bring about a transition of the mind to the other [A100,
amended translation].

However, he notes that this law "presupposes that appearances are

themselves actually subject to such a rule, and that in the diverse ele-
ments of these cognitive states a coexistence or sequence takes place in
conformity with certain rules" (A100, amended translation).
He illustrates the point by a series of examples, beginning with the
oft-quoted case of cinnabar: "If cinnabar were sometimes red, some-
times black..." (A100). What he means is, "If we represented cinnabar
sometimes as red, sometimes as black... "45 This discussion has some
major problems, but the point Kant is making is a sound one. The law
of association can get a foothold only if, whenever we are in sensory
contact with cinnabar, for example, we construct a representation of it
as red from our various sensory states. For, as he observes, this law
presupposes that the contents of our representations display certain
patterns. The gentle force of association cannot connect representations
of "red" and "cinnabar" unless these representations occur together.
The crucial point, which is badly obscured by the use of "Vorstellung"
instead of "Erkenntnis," is that this law operates on representations of
objects or properties. When ideas with particular contents have regularly
been experienced together, they will tend to be associated. Without the
assumption that ideas have determinate, repeatable contents, the law
makes no sense. In that case, however, the law presupposes and cannot
explain our ability to construct representations of objects and properties.
There is a second difficulty with Kant's argument. Even if the law of
association operates on representations, Empiricists could reply that
something like this law explains how various cognitive states are united
in representations. When cognitive states occur together, they tend to
become associated, and through this association produce representations
of objects and properties.
This account of the prerequisites of association occurs in Kant's pre-
liminary review of the topics of the Deduction. Fortunately, both the
terminological and the substantive difficulty are removed in the formal
presentation of his position. The problem with association is the one I
noted in discussing Hume:
 Transcendental Psychology in the Transcendental Deduction 79
If cognitive states [Vorstellungen] reproduced one another in any order,
just as they happen to come together, this would not lead to any deter-
minate connection of them, but only to accidental heaps, and consequently
would not give rise to any representation [Erkenntnis] [A121, amended
translation].
The stream of cognitive states is too fluctuating and too varied to yield
representations by the simple mechanism of spatiotemporal contiguity.
Another way to see the difficulty is to recognize that the law of asso-
ciation is but one principle. It links cognitive states related by spati-
otemporal contiguity. So, for example, it would connect cognitive states
produced by observations of striking matches and flames, and cognitive
states produced by observations of different parts of telephones. Only
in the second case, however, do we unite those cognitive states in the
representation of an object. The law of association operates in the same
way in all cases, however, and so could not explain how we achieve
different types of representations. Kant concludes that we need a rule
that connects a cognitive state with another particular cognitive state
in preference to others (A121). Spatiotemporal contiguity is too pro-
miscuous.

Associationism and Apriority

This argument, which is somewhat mishandled at A100 and then pre-
sented more clearly at A121, is the basis of Kant's claim at A105 that
the rule for producing representations of objects must involve or relate
to a priori0 elements and of his later claim that object representation
requires a pure, nonempirical concept of an object. But has he really
demonstrated the need for an a priori0 contribution? His case for the
inadequacy of association seems perfectly sound. Spatiotemporal con-
tiguity is in the senses, so if it were an adequate foundation for our
representations of objects, then this matter would be purely empirical.
However, the fact that this sense-based aspect of mental life is inade-
quate does not show the impossibility of some other, more subtle pattern
of sensory stimulation being the basis of object representation. The law
of association had a preeminent place in post-Humean Empiricism.46
Against this background it is not surprising that Kant sometimes con-
cludes that a capacity requires a priorl0 elements, on a basis of an
argument showing that it cannot be explained by the law of association.
Suppose, however, that there are many, more subtle, sensory patterns,
whose presence leads our minds to form particular types of representa-
tions. Does this show that these aspects of cognition derive from the sen-
80 KANT'S TRANSCENDENTAL PSYCHOLOGY
ses or that they require a substantial contribution from the mind itself? If
the sensory patterns are quite remote from the eventual representation,
so that a great deal of processing is required to construct the representa-
tion, then we might be tempted to say the representation is a reflection of
the mind itself. Even in this case, however, there is a more obvious an-
swer. The exasperation of many scholars over nineteenth-century de-
bates that pitted nativism against empiricism was often justified. For, in
the kind of case just described, the obvious answer is "both." To under-
stand how we are able to construct representations of objects, for exam-
ple, we need to look at both the patterns of our sensory stimulations and
the ways the mind uses those patterns to construct representations.
Because Kant's argument for a priori0 elements is really an argument
against the law of association, it is not totally successful. He does not rule
out the possibility just described, so he cannot legitimately claim that our
representations of objects do not derive from the senses. On the other
hand, by arguing against association, he shows the need to consider how
the mind constructs representations on the basis of more subtle sensory
patterns. Our capacity for object representation is not data driven in any
simple or obvious way. So any explanation must appeal to more elaborate
mental construction than mere association, and this may reveal that the
resulting representations are underdetermined by the sensory data. They
may include elements that were never in the senses.
The discussion of subclaim 3b enables us to deal quickly with claim 3a:
This [apriori0] rule makes possible the necessary reproduction of various
contents. To produce a representation, we must connect the multiple
contents of cognitive states that occur at different times in one represen-
tation. So those contents must be available, in some sense, for them to be
connected. Kant describes this situation as requiring the "reproduction"
of the contents. Again, what he has actually shown is that some rule for
reproducing and connecting the diverse elements of cognitive states other
than the law of association is necessary for representing objects. He has
not shown that this rule is a necessary and universal feature of human cog-
nition, because it derives from apriori0 sources.

Representations and Concepts

I turn to the last piece of the analysis, subclaim 4: The [necessary and
a priori0 rule of synthesis that yields a representation] makes it possible
to unite the diverse elements in a concept. Kant supports this claim with
a simple assertion: "All representation [Erkenntnis] demands a concept"
(A106). Here is the controversial premise noted earlier. The task of
 Transcendental Psychology in the Transcendental Deduction 81
representing an object requires the use of concepts. George may well
be right that Kant gets this idea from Condillac. The problem is that he
has no right to make the assumption. If Tetens is a reliable witness,
then whether or not cognitive states could enable us to represent objects
in the absence of concepts was a disputed question:
Nevertheless the major issue in the dispute over the existence of mere
cognitive states is not yet decided.... Are there cognitive states in us that
are regarded as images and signs, sufficiently articulate, and sharply enough
separated from others in the imagination, so that they themselves and
through them, their objects, can be differentiated from others! ... Are they
and can they be fully prepared and apperceptible without being actually
apperceived at the same time? Or must they perhaps receive that material
clarity only first, through the same act through which they are actually
perceived and are actually used as pictures and signs? through the act in
which they are informed with consciousness, and become ideas [my
emphasis]?47
That is, it is an open question whether cognitive states can indicate
objects before they have been brought under intellectual representa-
tions. Since analyses of empirical capacities can only presuppose well-
established facts, this part of Kant's analysis rests on a mistake, even if
the claim turns out to be correct.48

Kant's Defensible Results

Synthesis and the Productive Imagination

The core of Kant's analysis is that we cannot represent objects at all
unless there is some process that can construct unified representations
on the basis of the multiple contents of cognitive states occurring at
different times through the mediation of different senses. Any possible
explanation of object cognition must include an account of this process
of connection or synthesis. Further, this process cannot be governed by
the unaided law of association. Synthesis must be carried out inside the
mind, so it requires some mental faculty that has the power of syn-
thesizing. Kant labels this faculty "imagination" (A120). More precisely,
because he believes that we have an empirical "reproductive" faculty
of imagination (that is governed by the empirical law of association),
he claims that we have a second, "productive" imagination, which is a
"transcendental," "apriori0" and "pure" faculty. "Productive" is used
to make a contrast with a faculty that (only) follows the law of associ-
82 KANT'S TRANSCENDENTAL PSYCHOLOGY
ation, so the first epithet is justified. So is the characterization of the
productive imagination as "transcendental," if that indicates only that
this faculty is necessary for knowledge. If it means that this faculty makes
an a priori0 contribution to cognition, then this claim, like those for
apriority0 and purity, has not been adequately supported.
Kant posits a mental faculty, but this is hardly wild psychological
speculation. His analysis invokes two unwarranted assumptions: Rep-
resentations must be able to be brought under concepts, and there is
but one rule for connecting the contents of cognitive states in a repre-
sentation, that associated with the concept of an object in general (the
transcendental object = x) (A109). Perhaps the latter assumption is not
what it appears. He may intend "the rule" associated with the concept
of a transcendental object only as a metarule that stands indifferently
for a variety of rules. He offers no argument that there must be a single
rule for representing objects. Still, the presentation is misleading and
the "transcendental object = x" appropriately disappears from the sec-
ond edition, even though he maintains his general position on the syn-
thetic unity required for representing objects (B137, B139). These
assumptions should not be granted. Even if they are rejected, however,
the central portion of the analysis remains intact. We can represent
objects only if the imagination has rules of synthesis for combining the
multiple contents of cognitive states in a unified representation.
Why has this reasonably clear, well-argued, and modest proposal been
denigrated so consistently? The main reason is the nearly universal
resistance to any appearance of psychological material in philosophy.
However, Kant's discussions of synthesis have also raised two specific
objections that need to be addressed. One arises because his arguments
for rules of synthesis sometimes involve a confusion of levels. The effects
of this confusion can be seen clearly in the most extensive discussion of
the theory of rule-governed synthesis available in English, Robert Paul
Wolffs unfashionably psychological commentary.49

Robert Paul Wolff on Rules of Synthesis

Wolff regards rule-governed synthesis as the centerpiece of Kant's work,
because it promises to solve the problem of unity among diversity. If
an activity is governed by a rule, then it has a coherence and complete-
ness (and hence a unity), despite the diversity of its parts. I do not agree
that Kant is trying to explain the unity of an activity (synthesis). As I
note earlier, the explanandum is the unity of a representation; the ex-
planans is that it was produced from diverse states by a rule-governed
 Transcendental Psychology in the Transcendental Deduction 83
synthesis. Wolff goes on to characterize a number of features of rule-
directed activities. He then observes, however, that synthesis must be
"preconscious."50 The problem is that it is very hard to see how rule-
directed activities, as he describes them, could be preconscious (or un-
conscious). Wolff tackles these matters explicity, but I suspect that many
readers have engaged in similar reflections. The upshot is that Kant's
position seems incoherent.
Although textual infelicities (which I note later) makes these in-
terpretive problems understandable, Kant's basic position on rule-
governed synthesis is clear. The syntheses that produce representations
are not conscious processes. He attributes them to the imagination,
which is explicitly characterized as a "blind, but indispensable function
of the soul" (A78/B103, my emphasis). Further, Kant has as clear an
understanding as anyone of the distinction between conscious and un-
conscious rules. He presents this very distinction in the Groundwork:
"Everything in nature works according to laws. Only a rational being
has the capacity of acting according to the conception of laws "51
Wolffs description of rule-governed activities strongly suggests that rule
followers act according to their conception of the rules. But rules govern
syntheses only as the law of gravity governs the movements of the
planets: Theorists can appeal to these rules to describe what is happen-
ing. Kant's text is confusing, because he makes unwitting shifts between
the perspective of the individual who is engaging in various mental
activities and that of the theorist who is describing those activities.52 So,
for example, he suggests that individuals can only represent a number
if they are conscious of its production from the synthesis of units (A103).
Wolff relies heavily on this passage, and does not take account of the
fact that in the very next paragraph, Kant doubts that we [as individuals]
are aware of any such thing, even though we [as theorists] must assume
that it takes place.53

The "Problem" of Early Cognition

The second problem about synthesis arises because it seems to many
that accounts of early cognition must be, like their subject, inarticulate.
So in trying to describe the prerequisites of various cognitive tasks,
including full conceptual thought, Kant and his heirs must lapse into
babble. Onora O'Neill addresses this worry in her unusually sympathetic
article on transcendental psychology.54 She offers a strategy for avoiding
the difficulty. Instead of trying to describe what life would be like without
indispensable concepts, Kantians can shift the burden to opponents. As
84 KANT'S TRANSCENDENTAL PSYCHOLOGY
the opponents try to describe a situation where these concepts are lack-
ing, Kantians could point out that they have fallen into incoherence.55
Although this tactic might be useful in some cases, I believe that it
misrepresents the thrust of Kant's psychological analyses. In analyzing
cognitive tasks, he looks at the ability, then looks at the known re-
sources, and then argues that we can only explain the ability by assuming
some additional resources (typically a faculty and/or apriori0 elements).
His criticism of opponents would not be that they are incoherent but
that they cannot explain the cognitive abilities that we have. Once we
see the form of these analyses, however, the reply to the charge of
inevitable inarticulateness is straightforward. Since Kant appeals to
known resources, his accounts of the early stages of cognition are no
more problematic than those of his peers; in particular, there is nothing
in his method that requires him to try to describe these "from the inside."
Kant's many appeals to rules and his efforts to describe the early
stages of cognition raise suspicions about the soundness of the doctrine
of synthesis. As I have just argued, however, neither of these issues
provides any justification for dismissing the account. Kant's confusions
about rules are only text deep. When he is being careful, he presents
these rules as unconscious; it is only when trying to explain cognition
that we must be cognizant of rules of synthesis.

Constructing Representations of Objects and the

"Binding" Problem
In Chapter 6 I consider how the analysis of the task of representing
objects contributes to an argument for a special status for the categories.
Since that demonstration is unsuccessful, if ingenious, any enduring
value of the analysis must be intrinsic. Kant's analysis is both partial
and preliminary, but it advances the subject. The Lockeans realized that
cognitive states can inform us about objects only if they are derived
from sensory contact with objects, but they saw little need for elaborate
mental processing. Connections among mental states could be explained
by appealing to the law of association. Leibniz and his followers appre-
ciated that representation involved the unity of diverse contents in a
single state. At one level, the doctrine of synthesis is a classic Kantian
"synthesis" of the insights of his predecessors. By appealing to synthesis,
he can explain how representations can have unity and yet be anchored
in sensory contact with objects. Although this point will be more fully
developed in the next chapter, synthesis also provides the basis for
Kant's plausible functionalist account of how cognitive states can rep-
 Transcendental Psychology in the Transcendental Deduction 85

resent at all. This is possible because representations are produced from

sensory states that are directly caused by objects themselves and because
they can interact with other representations to produce further repre-
sentations.
At another level, however, Kant's account is no mere synthesis of
previous views. In revealing the dependence of a simple task like rep-
resenting objects on synthesis, he gives a forceful demonstration of the
need to consider the constructive powers of the mind. His contempo-
raries accepted the facts on which his analysis rests. We have unified
representations of objects; we receive a constantly varied stream of
sensory stimulation from diverse senses. But they did not fully realize
the startling implications of those facts. If conscious life is a series of
unified representations, and sensory stimulation offers a diverse and
ever-changing flow of information, then the mind must constantly be
engaged in synthesizing. Hence, the Deduction's constant references to
synthesis.
With the renaissance of interest in cognition, this issue has reemerged,
sometimes under the name of the "binding" problem. Anne Treisman
discusses aspects of it in a recent article in Scientific American.56 How,
for example, can the color, shape, and luminance of an object all be
brought together in a representation of one object with those various
properties? Treisman's work has offered dramatic experimental confir-
mation of the theoretical argument for the existence of processes of
connection. We must engage in combining, because sometimes we do
it erroneously. By requiring subjects to take in a variety of items at a
glance, she and her colleagues induced them to perceive "illusory con-
junctions."57 When rapidly presented with a display of a purple E, an
orange S, and a green V in a row, for example, some subjects will claim
that they have seen an orange V. The illusion is so strong that subjects
often resist the suggestion that they have erred.
Like synthesis, the binding problem has a ubiquitous quality. For as
Patricia Churchland has observed, if we can figure out how the nervous
system integrates information from various sources, then it is not clear
that there would be anything interesting left to know about it.58 Current
work has focused on how information from different sensory modalities
can be combined and on how different features within a modality can
be combined. Kant emphasized the problem created by the temporal
dimension of cognition:

all our representations are subject to time... and in it they must be or-
dered, connected, and brought into a relation. This is a general obser-
86 KANT'S TRANSCENDENTAL PSYCHOLOGY
vation which must be understood as fundamental throughout what follows
[A99, amended translation].
Although there may be as many different binding problems as there are
features to be combined, the temporal issue seems fundamental. For it
is hard to see how any binding problem can be completely solved without
solving a problem of temporal integration; and if there is any uniform
solution to binding problems, then it seems plausible to believe that it
would involve temporal integration. Thus, in insisting that the temporal
character of cognitive life requires synthesis, and in ruling out the simple
solution of spatiotemporal contiguity, Kant offers an abstract analysis
of the task of representing objects that is still useful in conceptualizing
the problem to be solved.
Finally, although the analysis of representing objects does not dem-
onstrate the necessity of a priori0 elements, it makes this a serious
possibility. Even basic representations of objects need to be constructed.
Thus, the constitution of our minds may influence our knowledge in
very basic ways. Although this is a far weaker result than Kant wanted,
even a plausibility argument for the thesis that what we know reflects
our ways of knowing is an important result. What is hard to understand
about the recent history of philosophy is how this result could ever have
been regarded as irrelevant to epistemology and the philosophy of mind.

Making Judgments About Objects

The Problem of Judgment

The objective Deduction takes a very different form in the second edi-
tion. In a well-known footnote to the preface of The Metaphysical Foun-
dations of Natural Science (1786), Kant estimates that the deduction of
the categories could be carried out "almost by a single conclusion" from
an exact definition [or analysis] of the act of judgment.59 Accordingly,
the second edition Deduction focuses on the task of making judgments
about objects, although there are still important references to the re-
quirements of representing objects (e.g., B137). By "judgment" Kant
does not mean an expressed judgment; he uses the term ambiguously
between silent or expressed judgments and the representational cog-
nitive states that produce them. Like the analysis of representing objects,
this analysis is not directed against skeptical opponents. Paragraph 19,
where the discussion is centered, presupposes that we can make judg-
ments about objects.
 Transcendental Psychology in the Transcendental Deduction 87
Kant opens the discussion with an explicit criticism of previous ac-
counts. Logicians characterize judgment as "a relation between two
concepts" (B141).60 Of what this relation consists, however, they do not
inform us. To appreciate Kant's analysis of this capacity it is again helpful
to consider some contemporary alternatives. By Tetens's account, most
teachers of reasoning regard judgments as "nothing more than the com-
parison and recognition of identity and differences [between ideas]."61
This view was shared by Leibnizeans and Lockeans. Locke was explicit
that in judging we presume agreement or disagreement between ideas,
even if we do not see it clearly.52 The principle of sufficient reason entails
that the predicate of a judgment be related to the subject by difference
or by partial identity. Condillac concurs: "[In judging]... we compare
our ideas, the consciousness we have of them is the cause of our knowing
that they are the same... or else is the cause of our knowing that they
are not the same.... "63 Tetens points out that this widespread as-
sumption cannot be correct.
Since not every relation [between ideas or concepts in a judgment] consists
in identity or difference [or containment or lack of containment], the act
of judging cannot always be comparison.... There must be many more
judgments of different forms.... M
Besides comparison, the other obvious option for connecting ideas in
a judgment was the law of association. This alternative is very much on
Kant's mind. He sets up the problem by announcing that he will try to
explain how judgment—as opposed to mere association of ideas—is
possible (B141). Later, he observes that, in a judgment, we claim that
properties occur together in an object (e.g., "Bodies are heavy"). The
law of association only connects ideas through their co-occurrence in
the subject: "When I have the impression of a body, I also have the
impression of heaviness" (B142). Presumably, the problem with asso-
ciation is the one encountered before. It is too promiscuous. The con-
tents of representations may be connected by association, even when
the properties represented are not connected in objects.
Kant's analysis of the task of making judgments includes several neg-
ative theses. We make judgments about objects. Assume provisionally
that judgments involve a connection of concepts. This connection cannot
be a matter of conceptual connection (one concept being identical with
or contained in the other), for in judging that objects are heavy, "I
certainly do not say that the [contents of] these cognitive states belong
necessarily to each other..." (B142, my translation, original emphasis).
On the other hand, the connection of concepts in a judgment about
88 KANT'S TRANSCENDENTAL PSYCHOLOGY
objects cannot be explained by invoking the law of association, because
association is too promiscuous. Although Kant does not recur to this
point in analyzing judgment, I might add that because objects themselves
are outside our minds, there is no point in appealing to them to explain
how we connect concepts in a judgment. If connections established by
the law of association are too liberal and conceptual connections are
too strong, then how can concepts be connected in a judgment that
represents the properties they signify as connected in an object?

The Synthesis of Intuitions

It is easier to follow Kant's positive analysis if we recall the parallel
problem about representing objects. Since objects themselves could not
furnish the requisite unity for representations, we needed a surrogate:
a rule of synthesis for connecting the multiple contents of cognitive
states. The reasoning underlying the account of judgment is the same.
The connection of concepts in a judgment can only be explained by
positing some (nonassociative) rules of synthesis that do the connecting.
Looking back to Kant's earlier analysis of representing objects, we
can see that judgment actually requires two levels of synthesis. One
level is needed to produce representations from the fluctuating stream
of cognitive states. In the present discussion, he investigates how the
contents of given representations [Erkenntnisse] can be related in a
judgment.65 He poses the problem in terms reflecting the standard view:
How are two concepts related in a judgment? However, having shown
the difficulty in appealing to conceptual connections, he casts his solution
in terms of representations and intuitions and their contents. If the
contents represented in a judgment are going to relate to an object,
then they must be related to each other by a synthesis of intuitions
(B142). That is, the relation between contents that are represented in
a judgment can be grounded in sensory data only if there are rules of
synthesis that enable us to connect the contents of representations that
are themselves fairly directly related to objects (see also A69/B94).
Thus, judgments require a synthesis of intuitions [Erkenntnisse] (B142).
Kant's analysis of making judgments is hard to follow, because it is
intertwined with considerations about the requirements of apperception.
If those elements are teased out, then the analysis is quite straightfor-
ward. The representations in judgments are not related by identity,
difference, containment, or mere association; rather, they belong to one
another by virtue of a synthesis of intuitions. In a sense, there is one
relation between representations in a judgment: They are related by a
 Transcendental Psychology in the Transcendental Deduction 89

synthesis of intuitions. In another sense, however, the relation varies,

because it is produced by sensory data (by rules for synthesizing that
data into representations) and by rules for synthesizing representations
into judgments. That is why logicians cannot tell us of what these re-
lations consist.

The "One-Step" Deduction

In paragraph 20 Kant moves quickly from this analysis to an argument
for a special status for the categories. He has shown that judgments
must be constructed by rules or functions of synthesis. The Metaphysical
Deduction claimed that there are twelve forms of judgment and that
the categories are associated with those forms. Kant now asserts that
the categories are (or are associated with) the rules or functions
that produce judgments (B143). With these additions, the analysis of
judgment is able to provide a direct argument for the universal applica-
bility of the categories: Anything that we can know, or at least anything
that we can judge, must be amenable to synthesizing by functions as-
sociated with the categories.66
Although almost the one-step deduction that Kant promised, the ar-
gument is too quick. As many have noted, he does not adequately defend
the table of judgments or the alleged connections between these forms
and the categories.67 Further, nothing in the analysis of making judg-
ments establishes that the rules of synthesis are even a priori0. Again,
he has only shown that the law of association cannot serve as such
a rule.

Constructing Judgments
Although I accept the usual verdict that this analysis cannot serve as
the basis for a sound argument for the categories, I believe that it has
intrinsic merits. Most important, Kant reveals the inadequacy of popular
alternatives. Judgments about objects cannot merely be a matter of
relations between concepts or terms. This would include relations of
containment, and also syntactic relations among expressions (see B128-
29).68 A purely formal approach to judgment can never explain our
capacity for making judgments about objects. On the other hand, the
obvious failings of the law of association suggest that judgments cannot
be explained by appealing to simple patterns of sensory stimulation
either. So, like the analysis of representing objects, Kant's constructive
account of judgment requires us to consider more seriously how the
90 KANT'S TRANSCENDENTAL PSYCHOLOGY
mind might construct judgments on the basis of more subtle sensory
patterns. Like the previous account, it also opens up the possibility that
our mental apparatus affects cognition in very fundamental ways. It
even raises the possibility that the mind has a limited set of rules for
constructing judgments, so that there are, in fact, forms of judgment.
Thus, the analysis of making judgments about objects offers a plausibility
argument for some of Kant's key doctrines.

The Objective and Subjective Sides of the Deduction

Kant's discussions of synthesis and representing and judging objects are
far too cryptic, because they are overshadowed by a particular conse-
quence of his analyses. Once this discovery is announced, other issues
are not fully explored but are pushed to one side. In the second edition
the analysis of judgment does not occur until five sections into the
chapter, and is largely cast in terms of this other issue. I have not
discussed this result, partly to permit an exploration of other points,
and partly because I will be devoting Chapters 4 and 5 to an extended
study of it.
Both the analysis of representing objects and the analysis of judgment
demonstrate the need for synthesis. This is Kant's constant theme in
the Deduction chapter, presented in several variations in both editions.
For knowledge to be possible, we must synthesize cognitive states in
representations, representations in judgments, and combine judgments
with further judgments through reasoning (A303/B359ff). More gen-
erally, "Cognition [Erkenntnis] is a whole in which the contents of cog-
nitive states are compared and connected" (A97, my translation). It is
a theme with two parts. At all levels, the constructive syntheses required
for basic cognitive tasks reveal possibilities for mental structures to
influence cognition; and, at all levels, they place a requirement on the
subject. Any subject capable of performing these tasks must have cog-
nitive states related by synthesis. This is why the objective and subjective
Deductions are inseparable and can be distinguished only in terms of
focus. In this chapter, I have looked at cognitive tasks from the objective
side. Given that we can represent and judge objects, how in general is
this possible? The next two chapters explore the subjective side. What
must a subject be like simply by virtue of being able to have mental
representations?
 4
Replying to Hume's Heap

Troubles with Apperception

Avoiding the Subjective Deduction

When Kant focuses on the demands that cognitive tasks put on the
subject, he proclaims the existence of a "transcendental unity of ap-
perception". "Why not call it the transcendental diversity of appercep-
tion?", Richard Rorty once inquired.1 Rorty's irony reflects a common
reaction among less sympathetic readers of the Critique. With this doc-
trine Kant's tendency to sesquipedality becomes intolerable. The trou-
bles that contemporary readers have with apperception go far deeper
than matters of terminological style, however. This doctrine is Kant's
principal teaching about the necessary attributes of a being capable of
cognitive experience. As such, it is the centerpiece of the subjective
Deduction. For reasons now familiar, recent interpreters have sought
to downplay the subjective side of this chapter and to cleanse its central
claims of any taint of psychology. The doctrine of apperception presents
a formidable challenge to this strategy. Its centrality is beyond question
and its topic is a necessary attribute of a thinker. Seemingly, the only
way to protect Kant against the charge of psychologism is to maintain
that it is a nonpsychological, or innocuously psychological, claim about
thinkers. Three versions of this strategy are well represented in the
literature. As I will argue, all have serious weaknesses.

Apperception as the Cogito

The simplest strategy interprets die transzendentale Einheit (unity) der
Apperception as a German variation of Descartes' cogito.2 On this read-
91
92 KANT'S TRANSCENDENTAL PSYCHOLOGY
ing, the doctrine of apperception is the starting premise of the deduction
of the categories. Presumably, this interpretation is suggested by the
precise wording of the best-known statement of the doctrine: "It must
be possible for the I think to accompany all my cognitive states" (B131,
amended translation). It may also be favored, because the cogito is
perceived as the correct starting place for a philosophical argument.3 If
the doctrine of apperception asserted nothing more than the cogito, then
it would be a certain, even if psychological, premise. Despite these
advantages, there is an overwhelming objection to this assimilation.
Kant criticizes the cogito explicitly and at length in the Paralogisms
chapter.4

Strawson and the Self-Ascription Reading

Strawson introduced a second reading that distances apperception from
unwonted psychological connotations. Apperception is the doctrine that
subjects must be able to ascribe mental states to themselves.5 In Straw-
son's words:
Unity of consciousness to which a series of experiences belong implies,
then, the possibility of self-ascription... [that is] the possibility of con-
sciousness, on the part of the subject, of the numerical identity of that to
which those different experiences are by him ascribed.6
The idea that self-ascription is the core of the apperception thesis is
echoed by more recent British commentators:
T. E. Wilkerson: "Self-consciousness is more happily described as the
ability to identify one's own experiences as one's own."7
Ralph C. S. Walker: "an experience cannot be mine unless I am at least
potentially aware of it as such."8
Roger Scruton: "It [the unity of apperception] consists in my immediate
awareness that simultaneous experiences belong to me."9
The perceived virtue of this reading is that it allows an appealing
reconstruction of the deduction of the categories, in terms of the "logic"
of self-ascription. Again, apperception is the first premise and it asserts
that subjects must be able to self-ascribe mental states. In the Refutation
of Idealism and the Second Analogy, the argument can then be made
that self-ascription of mental states requires that we can make certain
kinds of judgments about objects.10 As Strawson puts it:
 Replying to Hume's Heap 93
The more fundamental point of the Kantian provisions is that the expe-
riences of such a subject [one who can self-ascribe] must themselves be
so conceptualized as to determine a distinction between the subjective
route of his experiences and the objective world through which it is a
route."
Although this reconstruction does not purport to vindicate the categories
per se, it is regarded as a good argument for a position quite like Kant's.
The argument is supposed to be good, because it has the form of a
transcendental argument.
In Chapter 1 I noted the intrinsic problems of transcendental argu-
ments and argued that there are systematic reasons for not casting Kant's
reasoning in this mold.12 Since the Strawsonian reconstruction of the
argument from apperception has enjoyed widespread acceptance, I add
three further objections to this key example. All are directed against
the reconstruction as an interpretation.13
First, a textual point. In Kemp Smith's translation, apperception is
presented in terms of ascribing experiences only once, at A122. "For it
is only because I ascribe all perceptions to one consciousness (original
apperception) that I can say of all perceptions that I am conscious of
them." "Ascribe" is Kemp Smith's translation of "zdhlen z«." However,
it would be at least as reasonable to render the phrase: "For it is only
because I classify all perceptions with one consciousness...," or, "For
it is only because I take all perceptions to belong to one consciousness.
..." When Kant presents the relation between cognitive states and a
single consciousness, he standardly uses gehoren ("belong") and its de-
rivatives,14 so the last rendering of A122 seems best. Without this one
piece of explicit textual support, the self-ascription reading must rely
on highly contentious texts, such as B132 and A116.
On this reconstruction, the Refutation of Idealism assumes a central
role, because that is the only place where Kant talks about cognitive
relations to objects being necessary to consciousness of one's own ex-
istence. Should the Refutation have such prominence? Since it was
added in the second edition, the first edition cannot be interpreted along
these lines. Further, there is no indication either in the prefatory foot-
note about the Refutation, in the second edition Deduction, or in the
Refutation itself that this section is needed to complete the deduction
of the categories. The more obvious hypothesis is that Kant was stung
by the comparison of his position to Berkeley's, and added this section
and a couple of other remarks to reinforce the explicit contrast that he
had drawn in the Prolegomena.,15 Berkeley denied the existence of the
94 KANT'S TRANSCENDENTAL PSYCHOLOGY

material world; while discussing the category of actuality, Kant inter-

rupts himself to offer a reply. The location itself suggests that the Re-
futation is not an integral part of the deduction, but only an aside, a
reply to critics.
Nathan Rostenreich raises a substantive objection to linking the Re-
futation and the doctrine of apperception.16 The Refutation is concerned
with the empirical consciousness of one's own existence—the recognition
of one's particular mental history (Bxl). Kant explicitly contrasts ap-
perception with the empirical unity of consciousness, however (B139-
40). In Chapter 5 I return to the issue of self-ascription and argue that
the ability of subjects to identify cognitive states as their own is only a
small and bungled part of the doctrine of apperception.17

The "Logical" Reading of Apperception

The most direct strategy for depsychologizing the doctrine of appercep-
tion is to present it as a logical, as opposed to psychological, thesis.
Wilkerson combines this strategy with the self-ascription reading: "The
unity of self-consciousness is ... a formal unity consisting simply of the
formal fact that experiences are mine."18 Allison claims that the act of
becoming aware of the identical self is "the form of the act of reflection
. . . it is nothing more than the 'logical act'... [original scare-quotes]."19
This approach seems to be supported by numerous texts (e.g., the I
think is a logical subject of thought [A350], or a formal subject of thought
[A105]. It is not always clear what is meant by "logical" or "formal" in
these readings. No one believes that it is a fact of formal logic that
mental states belong to a subject of thought.20 The force of "formal" is
carried by the implicit contrast with "psychological."21 Thus, this ap-
proach makes sense only on the presupposition of a sharp contrast
between what Kant meant by "logical" and what we mean by "psycho-
logical." As Chapter 1 made clear, however, this presupposition is false.
Kant believes that logic investigates the rules governing the understand-
ing. Transcendental logic "concerns itself with the laws of understanding
and reason solely insofar as they relate a priori to objects" (A57/B82).
In criticizing these well-known approaches to apperception, I do not
suggest that commentators have not been ingenious or that they have
not provided useful insights into Kant's position. My point is that they
have tied themselves to an interpretive strategy that is bound to be
inadequate. The doctrine of apperception is about the primary attribute
of thinkers that is necessary for cognition. As such, it is a thesis in
transcendental psychology and can only be understood in these terms.
 Replying to Hume's Heap 95

Besides this central difficulty, two other widespread mistakes have ham-
pered interpretations of apperception, one interpretive and one
historical.

Two Mistaken Assumptions

As noted, apperception is often regarded as the unargued first premise
of the transcendental deduction.22 This interpretation makes some sense,
if Kant's doctrine is assimilated to the cogito. Even then, however, it is
not very plausible. Could Kant have expected his readers simply to grant
a thesis that presupposed the soundness of his own highly technical
vocabulary? In any case, this assimilation is highly questionable in light
of Kant's later criticisms of the cogito.
The historical mistake supports the interpretive one. Historians have
taken Kant to be ignorant of Hume's attack on certain metaphysicians
[Descartes], who claim to espy a simply and continuing self in them-
selves. Later, I will present conclusive evidence that he was aware of
this critique. Thus, a fairly standard assumption about apperception,
that Kant adopted Descartes's starting point for philosophy without
benefit of Hume's insights, is exactly backward.

Hume
Given that Kant was aware of Hume's famous denial of mental unity,
he could not simply assert that mental life involves unity. He had too
much respect for Hume to ignore his criticisms. To maintain the unity
of a thinker, he had to demonstrate it. If we reconsider the task analyses
of the last chapter in this light, however, then they offer something close
to the needed demonstration. The analyses of representing and judging
objects establish two important facts about thinkers. Most obviously,
any thinker who shares our basic constitution, and so takes in infor-
mation through a stream of sensory impressions, must have a productive
imagination. They also demonstrate a slightly less obvious fact. To rep-
resent objects, for example, the productive imagination must combine
information from diverse cognitive states in a unified representation.
Some cognitive states must, therefore, depend on the existence of
others.23 The syntheses that result in judgments create further depend-
ency relations. Thus, pace Hume, there must be real connections among
some cognitive states.
To represent objects and to judge them, a thinker must synthesize
the contents of diverse cognitive states, thereby creating a "synthetic
96 KANT'S TRANSCENDENTAL PSYCHOLOGY

unity"—a unity produced by synthesis—among the states, as well as a

synthetic unity in the contents of the resulting state. This synthetic unity
is required for tasks that make up cognitive experience. Further, Kant
thinks that the rules guiding the syntheses are a priori0. This synthetic
unity should, therefore, be characterized as "transcendental," within
his system. From these analyses, Kant can extract an argument for a
"transcendental [or "synthetic"] unity of apperception," with a clear
anti-Humean import. (And although the labels may strike us as preten-
tiously technical, they have a point.)
Conceivably, this outcome of his analyses is a happy coincidence. It
seems far more reasonable, however, to assume that it was intended.
In which case, Kant's analyses of cognitive tasks have two objectives:
Reveal the influence of the mind's own constitution on cognition and
demonstrate the existence of real connections among the cognitive states
of a single thinker, thus replying to the most skeptical of Hume's po-
sitions. I will argue later that we can make a great deal of sense of
specific aspects of the apperception doctrine, if we take the Deduction
chapter to have both goals.
Although the analyses of representing and judging objects yield the
desired conclusion, they cannot provide direct arguments against Hume.
They begin by assuming that we can represent and judge objects "out-
side" us, and Hume denied this in "Of skepticism with regard to the
senses."24 To succeed in his antiskeptical argument, Kant must show
that the unity of a thinker is a necessary condition for some cognitive
task that even Hume will not deny. This leads him to consider another
cognitive capacity. How can thinkers have cognitive states that at least
represent some content to them! It is this analysis that leads Kant to
hold that the transcendental unity of apperception is an absolutely fun-
damental fact about any cognitive being.
Although the argument from apperception for the categories is un-
sound, the breach occurs at a later stage.25 In this chapter I argue that
the case for apperception is solid and constitutes an effective reply to
Hume. This result is nearly as significant for philosophy as a demon-
stration of the categories. For in establishing the necessary unity of
apperception, Kant defends a cognitive account of mental unity that is
superior both to contemporary rivals and to present-day accounts. When
the doctrine of apperception is approached as it must be—as a conclusion
established by transcendental psychology-—it emerges as a powerful so-
lution to the problem of mental unity.
In the next section, I document Kant's knowledge of Hume's position
and give a precise statement of it. With that available I shall argue that
 Replying to Hume's Heap 97

synthesis creates exactly the relation among cognitive states that Hume
denied. Finally, I present Kant's arguments for the claim that merely
being able to have cognitive states that represent something to us re-
quires that our states be connected by synthesis. This is Kant's reply to
Hume's skepticism about personal identity. In Chapter 5 I elaborate
and defend the resulting account of mental unity, thus supporting my
contention that this is a major, if largely unappreciated, contribution of
the Deduction chapter.

Hume's Problem

Hume's Absence
My interpretation of the subjective Deduction depends on the historical
claim that the doctrine of apperception is, in part, a reply to Hume's
skepticism about personal identity. Having criticized others, it is only
fair to point out that my own interpretation faces an obvious, seemingly
insurmountable, objection. Hume's name does not appear in any of the
passages where I claim that Kant is trying to refute the bundle theory
of the self. His account of personal identity is not mentioned even once
in the Critique*.
Although this evidence may seem conclusive, the picture changes
dramatically if we consider Kant's references to opponents systemati-
cally. He regularly fails to mention the protagonists in his discussions
by name: Leibniz is not mentioned in the Paralogisms chapter or in the
Aesthetic's discussion of space and time25; none of Kant's discussions
of space or time in the Critique refers to Newton; the Second Analogy
contains no reference to Hume; and, most surprisingly, in the first edi-
tion, Kant's interest in causation is not linked to Hume at all until the
Discipline of Pure Reason, 550 pages after his celebrated reply in the
Second Analogy. As I noted in Chapter 2, some of Kant's reticence
about his targets is probably due to the fact that they were obvious to
his readers. Whatever the reason, the habit of not placing his discussions
in their historical context has enormous potential for misleading later
scholars. Imagine trying to understand the Aesthetic's metaphysical
claims about space and time outside the context of the Newton-Leibniz
debate! In the case of Hume and apperception, that potential has been
fully realized.
Kant's omission has been compounded by an error of Norman Kemp
Smith's. One of Kemp Smith's most important contributions was the
98 KANT'S TRANSCENDENTAL PSYCHOLOGY
discovery that Kant would have been familiar with Hume's extensive
discussion of causation in the Treatise, through generous citations in the
German translation of James Beattie's Essay on the Nature and Im-
mutability of Truth.27 When discussing a passage from the Deduction,
however, he writes, "Kant teaches, in agreement with Hume, though,
as we may believe, independently of his direct influence, that there is no
single empirical state of the self which is constant throughout experi-
ence" (my emphasis).28 He supports this opinion with a footnote re-
minding his readers that personal identity is not discussed in the Inquiry
and claiming that it is not mentioned by Beattie. This is simply false,
however. As we will see, Beattie allows Hume to speak for himself in
laying out the theory of the self that he intends to ridicule. Robert Paul
Wolff has suggested that Kemp Smith makes this error because he
checked the sixth edition of Beattie's book, whereas the German trans-
lation was made from the first.29 This is not right either. In the sixth
edition, Beattie simply leaves out Hume's name, referring to him in-
stead, for example, as "our author," but still provides generous citations
on personal identity.30 However it came about, Kemp Smith's testimony
that Kant had no direct knowledge of Hume's position has led subse-
quent scholars away from the obvious hypothesis that the Deduction's
many references to the necessary unity of mental life were directed
against his predecessor's denial of unity.

Kant's Knowledge of Hume's Position

Kant knew about Hume's attack on the self and so did his readers.
Besides Beattie's long and popular (if silly) discussion, Hume's theory
was also referred to by J. C. Lossius (who probably knew it through
Beattie).31 Tetens attempted a reply of his own in his Philosophische
Versuche.32 His presentation suggests that his audience was familiar both
with Hume's denial of personal identity and with the criticisms of Reid
and Beattie. In Tetens's estimation the replies of Reid and Beattie are
"not incorrect, but unphilosophical." Unfortunately, his own solution
is both: There is an additional feeling that accompanies mental states
that Hume "overlooked."
At the time Kant was writing the Critique, Hume's views on personal
identity were known in Germany and were considered obviously false,
as they had been in his native Scotland. Still, the "refutations" of this
scandalous position were far from satisfactory. Did Kant believe that it
was important to discover the real error of Hume's ways? He clearly
 Replying to Hume's Heap 99

appreciated the insights behind Hume's skeptical positions and thought

that he could learn from them. Further, the denial of the self presents
an enormous intellectual challenge, because it is both obviously false
and very difficult to refute. Given these provocations, had Kant not
responded to Hume's challenge, scholars would need to consider why
he felt no obligation to do so.
Beattie's Essay on Truth provided Kant with the following excerpts
from Hume's attack:
The question concerning the substance of the soul is unintelligible...
What we call a mind is nothing but a heap or collection of different
perceptions (or objects) united together by certain relations and supposed,
though falsely, to be endowed with perfect simplicity and identity... If
anyone, upon serious and unprejudiced reflection, thinks he has a different
notion of himself, I must confess I can reason with him no longer. All I
can allow him is, that he may be in the right as well as I, and that we are
essentially different in this particular. He may perhaps perceive something
simple and continued, which he calls himself; though I am certain there
is no such principle in me. But setting aside some metaphysicians of this
kind, I may venture to affirm of the rest of mankind, that they are nothing
but a bundle or collection of different perceptions, which succeed each
other with inconceivable rapidity, and are in a perpetual flux and move-
ment. . . . There is properly no simplicity in the mind at one time, nor
identity in different (times), whatever natural propension we may have
to imagine that simplicity and identity.... They are successive perceptions
only that constitute the m i n d . . . ,33

Beattie omits some parts of Hume's position on the self. He fails to

mention the causal theory and Hume's despair in the appendix. 34 Still,
the excerpts contain three central and distinctive theses about the self,
two in opposition to the Cartesian view. As Hume notes, there is no
reason to believe that there is an entity in one that thinks; a fortiori,
there is no reason to believe that there is a simple, thinking self.
Like Kant, Hume accepts Locke's analogy between sensory percep-
tion and a faculty of "inner sense" through which we "perceive" the
contents of our minds. Although Locke's characterization of our ability
to monitor and report mental states is misleading, Hume was not misled
by it on this point. Whatever he expected inner sense to divulge, he
realized that it did not enable him to perceive a thinking self. His conces-
sion that another may perhaps perceive such a self is an obvious bit of
mockery. Since inner sense does not show us a self at all, it cannot
disclose the existence of a simple self. Kant agrees with Hume's reasons
100 KANT'S TRANSCENDENTAL PSYCHOLOGY

for denying the Cartesian account. In the Deduction, he points out, in

terms strikingly reminiscent of Hume's discussions, that inner sense does
not reveal a self:
The consciousness of the self, according to the determinations of our state
in inner perception, is merely empirical, and always changing. There can
be no permanent and continuing self in this flux of inner appearances...
[A107, my translation].

For the empirical consciousness, which accompanies different represen-

tations, is in itself diverse and without relation to the identity of the
subject.... [If inner sense were the only means of cognizing a self], I
should have as many-colored and diverse a self as I have representations
of which I am conscious to myself [B133-34],
Kant repeats this point in the Paralogisms chapter, twice in the first
edition and once in the second (A350, A381, B413). These passages
fairly shout allusions to Hume. A107 is the text that led Kemp Smith
to consider and reject any direct influence. In specifying the absence of
direct influence, he was presumably conceding indirect influence, al-
though this subtlety may have been lost on later scholars.
Hume's third claim about the self presents his positive conception. A
mind is a collection of different perceptions that are connected by certain
relations that lead us to mistake a succession of different things for one
enduring object. Although Beattie's citations do not say what those
relations are, Kant would have no trouble figuring it out. In the Inquiry,
Hume is explicit about the kinds of possible connections among mental
states: resemblance, [spatiotemporal] contiguity, and causation.35 Fur-
ther, he would know what these relations signify for Hume. Although
they might lead the mind to feel connections or identity across percep-
tions, they would not establish any real or necessary connections among
them.

The Denial of Real Connection

Before considering Kant's reply to this position, it will be helpful to
have a more precise characterization of it. The best place to find clar-
ification is the famous passage in the appendix to the Treatise, where
Hume admits that his account of personal identity is inconsistent. There
are two principles that he can neither renounce nor render consistent:
"that all our distinct perceptions are distinct existences, and that the
mind never perceives any real connection among distinct existences."36
As all readers quickly realize, these two principles are consistent, an
 Replying to Hume's Heap 101

issue to which I return later. For now I want to focus on the first prin-
ciple, for that is the key to his positive view. In the discussion, he elabo-
rates: "All perceptions are distinct. They are, therefore, distinguishable,
and separable, and may be conceived as separately existent, and may
exist separately, without any contradiction or absurdity."37 In de-
scribing minds as made up of distinct perceptions only, Hume is deny-
ing any relations of existential dependence.38 The mental states that
constitute a single mind may exist separately without contradiction or
absurdity; they do not depend upon each other for their existence or
properties.
Why does Hume despair over his treatment of personal identity? In
its more elaborate version, his second principle asserts: "If perceptions
are distinct existences, they form a whole only by being connected to-
gether. But no connections among distinct existences are ever discov-
erable by human understanding. We only feel a connexion... "39 He
could achieve local consistency by maintaining that all perceptions are
distinct and so have no real connections among them. He is obviously
unwilling to accept this solution, however. Why? Consider his most
famous explanation. We say that A's cause B's, because after A-
perceptions and B-perceptions have been constantly conjoined, we ex-
pect a B-perception after experiencing an A-perception. Unfortunately,
this statement of Hume's account ignores its hidden qualification. For
if every time that / see an A, someone else sees a B, then there will be
no expectation of a B, upon the presentation of an A. The qualification
is that the perceptions belong to a single mind. Hume's account of causal
thinking—like all his explanations of human thought—presuppose that
the thoughts in question belong to a single mind. Individual minds are
theoretical presuppositions of his own philosophy. Hence, it is not
enough to provide an account (even an implausible one) of how the
vulgar mistakenly come to believe in individual minds.40 He must find
some way of actually connecting diverse states in individual minds, but
real connections among distinct existences cannot be found. The incon-
sistency Hume rightly fears is not within his account of personal identity,
but between the demands of his general theory and the two principles
he cannot renounce.41
Since Beattie does not quote from the appendix, Kant may not have
had access to passages I have used to clarify Hume's position. Never-
theless, he knew enough about Hume's general views about causation
and existential inferences to see the implication of the bundle theory.
In the Inquiry, Hume is explicit that the existence of a cause without
the existence of its effect is consistent and conceivable.42 In his intro-
duction to the Analogies, Kant is explicit that the principles that he will
102 KANT'S TRANSCENDENTAL PSYCHOLOGY

try to justify in reply to Hume concern relations of "existence" between

objects (A179/B222). Thus, he would understand Hume's positive the-
ory as implying that the different perceptions that we think of as be-
longing to one mind can exist in total separation from one another. The
wonderful epithet "heap," which recurs in Kant's text,43 itself confirms
this reading. For Hume a person is a collection of intrinsically unrelated
perceptions.
Hume failed to resolve the inconsistency within his system, because
he believed that no connections could be discovered among distinct
existences. At 107, where Kant virtually repeats Hume's description of
the absence of a self in introspection, he also offers what appears to be
a diagnosis of his predecessor's methodological error:
The consciousness of the self, according to the determinations of our state
in inner perception, is merely empirical and always changing. There can
be no permanent and continuing self in this flux of inner appearances....
What has necessarily to be represented as numerically identical cannot be
thought as such through empirical data. To render such a transcendental
presupposition valid, there must be a condition which precedes all ex-
perience, and which makes experience itself possible [A107].
As implicit criticism, this is inaccurate. Although Hume needed to es-
tablish actual connections among perceptions, he did not need necessary
connections. Nevertheless, Kant's remarks provide a clear blueprint of
his own strategy. He will try to establish the unity of a mind by arguing
that we can explain how we perform various tasks that make up cognitive
experience only by assuming such unity. He has a skeptical opponent,
so insofar as his remarks are directed against Hume, he can only assume
very minimal cognitive tasks.

Synthesis and Apperception

Connecting Cognitive States by Synthesis

Hume maintained that perceptions were distinct and unconnected. The
Deduction chapter constantly refers to an act performed on cognitive
states, "synthesis" in the first edition, "combination" [Verbindung] or
"synthesis" in the second. Chapter 3 explored some aspects of the theory
of synthesis. Synthesis is a process that realizes a function that yields
representations as outputs from cognitive states or representations as
inputs. Here I consider what this theory implies about the connections
 Replying to Hume's Heap 103

among cognitive states. For convenience I repeat Kant's original pre-

sentation of synthesis:
By synthesis, in its most general sense, I understand the act of adding
different cognitive states [or their contents] to each other and of com-
prehending their diverse [elements] in a single representation [A77/B103,
my translation].
Synthesis is an act or process that leads to the diverse elements of
different states being comprehended [begreifen] in a single representa-
tion. This is an odd turn of phrase. Kant's point seems to be that the
elements of the earlier cognitive states are preserved or reflected in the
resulting state. Without those cognitive states and their particular ele-
ments, syntheses would have nothing to combine and preserve. It follows
that the later state exists and has a particular content, only because the
earlier states existed. Synthetic processes thus lead to a relation of ex-
istential dependence among cognitive states. If Kant can demonstrate
that syntheses, as just described, are required for cognitive tasks that
Hume must acknowledge, then he will have established precisely the
relation among cognitive states that his predecessor denied. Synthesis
also induces a second, weaker relation among the earlier states. They
are connectible by synthesis.
It might seem easier to note that "synthesis" appears to be a causal
process for Kant, and thus that earlier and later cognitive states would
be related as partial causes to effects. The problem with this terminology
is that establishing a causal relation among cognitive states would not
suffice to rebut Hume's denial of existential connection. For Kant has
not yet argued against Hume's analysis of causation. Thus, in the De-
duction he has to take a different tack. He needs to argue for a con-
nection among cognitive states that is stronger than Hume's, without
presupposing the Second Analogy's defense of a strong notion of
causation.

Transcendental Synthesis
Although Kant discusses empirical syntheses, his real interest lies in
transcendental syntheses. As we saw in Chapter 1, something is tran-
scendental if it is a prerequisite for cognitive experience and involves a
priori0 elements.44 In Chapter 3 I noted that Kant's arguments for a
priori0 elements in the syntheses required for representing and judging
objects fall short. They establish only the necessity of syntheses guided
by some rules other than the law of association. Insofar as the goal is
104 KANT'S TRANSCENDENTAL PSYCHOLOGY
only to demonstrate real connections among cognitive states, however,
he only needs to establish the first condition for transcendental status.
Syntheses are required for tasks that constitute cognitive experience.

Apperception and Transcendental Synthesis

Given what Kant means by "synthesis" and what he means by "tran-
scendental," it would be shocking if he began the transcendental de-
duction by assuming the existence of transcendental syntheses. He
virtually equates the doctrine of transcendental synthesis with the doc-
trine of apperception, however. At 118 he claims that a transcendental
synthesis of imagination is prior to and necessary for apperception and
at B134 that a priori synthesis is sufficient for it. "Synthetic unity of the
diverse [elements] of intuitions, as produced [hervorgebracht] a priori,
is thus the ground of the identity of apperception itself which precedes
all my determinate thought" (amended translation). Finally, he iden-
tifies the two at B135:
This amounts to saying, that I am conscious a priori of a necessary synthesis
of cognitive states, which is called the original synthetic unity of appercep-
tion, under which all my given cognitive states must stand, but under
which they must be brought through a synthesis [my translation, my
emphasis].

Two synthetic unities are required for cognition. In the last chapter,
we considered the synthetic unity of a representation that represents
an object. Here Kant maintains that there must be a synthetic unity
among cognitive states themselves. The unity of apperception, under
which all my cognitive states must stand, is created through synthesis.
Since the syntheses that create the unity of apperception are required
for cognition, and since the unity itself is required for cognition,
it is also a transcendental, or partly transcendental (see earlier),
unity.
In proclaiming a transcendental unity of apperception, then, Kant
asserts that a real connection among cognitive states, produced by a
priori0 syntheses, is a prerequisite for cognition. Clearly, this doctrine
could not be assumed at the beginning of an argument. Equally clearly,
it has nothing to do with the cogito.
So far, I have highlighted certain elements of this crucial passage—
synthesis and synthetic unity—while ignoring others. What of the claim
that I am "apriori conscious of a necessary synthesis of cognitive states,"
and what of the term "apperception" itself? "A priori conscious" sug-
 Replying to Hume's Heap 105
gests a special type of consciousness. It is hard to see what this could
be, however. The only awareness that we have of our states is through
inner sense. If Kant means to assert that we have another similar faculty,
through which we are aware of a necessary synthesis of cognitive states,
or a unity of apperception, then his position would be no better than
Tetens's suggestion that Hume simply overlooked a feeling that accom-
panies perceptions. In Kant's version, Hume would overlook (or lack?)
a special mode of self-awareness.
Henry Allison suggests that "a priori conscious" is just a "clumsy way
of referring t o . . . an awareness of something as necessarily the case."45
This seems right, in light of Kant's multiple uses of "c priori." If the
unity of apperception can be shown to be a necessary condition of
cognitive experience, then our knowledge of the doctrine would be a
priori^. For the argument for apperception would not depend on par-
ticular experiences, but only on the possibility of cognitive experience.
Further, what we know would have the logical form of a priori^ claims.46
In any world in which we can have cognitive experience, the unity of
apperception would obtain. It is a necessary truth, in Kant's sense of
"necessity."

Apperception
I turn to "apperception" itself. On my account, the "unity of apper-
ception" refers to the fact that cognitive states are connected to each
other through syntheses required for cognition. "Apperception" does
not indicate any awareness of a separate thing, a "self," or even that
different cognitive states belong to a separate thing, a "self." Rather,
they belong to the unity of apperception in being connected by syntheses
to each other.
This account of apperception is not intended to be exhaustive. Cer-
tainly one important part of Kant's doctrine that I have not mentioned
is the connection between apperception and the categories. He believes
that the unity of apperception is brought about by syntheses that are
guided by rules associated with the categories. I omit this aspect, because
it does not bear on the soundness of the reply to Hume and because I
do not believe that the arguments for categorically determined syntheses
succeed.
Robert Pippin has objected that an earlier version of this account of
apperception leaves out something else that is fundamental: the apparent
reflexivity of apperception. He believes that it is an essential part of the
Critical philosophy that "all human experience is ineliminably reflexive
106 KANT'S TRANSCENDENTAL PSYCHOLOGY

. . . because, according to Kant, whenever I am conscious of anything,

I also "apperceive" that it is / who am thusly conscious."47
One reason that it is difficult to give an adequate account of apper-
ception is that this term, and its close relative "self-consciousness," were
used by Kant's predecessors.48 So questions arise about whether his
usage implies any substantive agreements with earlier views. Certainly,
Christian Wolff took human consciousness to be inherently reflexive,
and he captured this claim in a doctrine of "apperception." According
to Wolff, all cognition would be reflexive, because it involves both
perception and "apperception," our ability to perceive our own per-
ceptions, and so ourselves.49 Despite the common terminology, Kant
appears to reject this view completely when he complains that systems
of psychology (presumably Wolffs) confuse inner sense and appercep-
tion (B153).
For textual reasons that I give later, I do not agree with Pippin's
strong claims for reflexivity. Still, there are clearly elements in Kant's
discussion that point in this direction. I suspect that these derive from
some intriguing ideas of Leibniz. In two of his best-known works, the
Principles of Nature and of Grace and the Monadology, Leibniz char-
acterizes apperception as follows:
[Apperception is]... consciousness or reflective knowledge of this inner
state [of a monad] itself and which is not given to all souls, or to any soul
all the time.

[Ajnimals [who can reason]... [have] souls [that] are capable of perform-
ing acts of reflection and of considering what is called T.

29. [I]t is the knowledge of necessary and eternal truths which distinguishes
us from simple animals, and gives us reason.

30. It is also by the knowledge of necessary truths and by their abstractions

that we rise to reflective acts, which enable us to think of what is called
I....50
Leibniz's offers four intertwined theses. Apperception is a reflective
consciousness of an inner state; it distinguishes rational animals from
all others; it is required for reasoning; it is through the reflective acts
involved in reasoning that we come to be aware of an I. The third claim
seems pivotal. A reflective consciousness of inner states is required for
 Replying to Hume's Heap 107

reasoning. Although the terminology is Baroque, the point seems

straightforward. Beyond whatever we do that is like animals, we also
have the ability to engage in conscious reasoning. We can consider what
our ideas imply and whether they are consistent, for example. This is
possible only because we can recognize our inner states as such, as
cognitive states that represent the world to be in various conditions.
Although much of the time (Leibniz guesses three quarters), we do not
consciously reason, and so are like beasts (and crude empiricists),51 when
we do, we must take note of our cognitive states as such and hence have
a reflective awareness of ourselves as I's, as subjects of thoughts.
Given Kant's interest in the prerequisites of cognition, he should have
been struck by this claim about reasoning. In the Anthropology, he
echoes Leibniz's view that it is apperception, the ability to say "I," that
separates us from the animals.52 Further, the preliminary discussion of
the topics of the A Deduction includes several claims about cognitive
tasks requiring us to be conscious of something: our states, ourselves,
our synthesizing (A103, A107, 108). On the other hand, the considered
accounts of cognitive tasks, in Section 3 of the A Deduction, and in the
B Deduction, explain cognition by reference to the synthesizing activities
of the imagination, which he acknowledges to be a blind (i.e., uncon-
scious) faculty. I think the problem may be that Kant recognizes that
there is some truth in Leibniz's case for the necessity of reflective aware-
ness, but he also recognizes with Hume that we are not, in fact, aware
of anything that we could call a self. And he is uncertain about how to
do justice to both points.
In any case, as Pippin acknowledges, in introducing apperception at
B132, Kant affirms that all my cognitive states must be capable of be-
longing to the unity of apperception, "even if I am not conscious of them
as such." Pippin suggests that this qualification indicates the sophisti-
cation of Kant's view. The claim is not that we are always conscious of
cognitive states as our own but that the possibility of this awareness is
"an inseparable component of what it is consciously to perceive, imagine,
remember, etc."53 But how does this happen? Various cognitive tasks
may require conscious attention, or the possibility of conscious atten-
tion.54 However, Kant is clear that such empirical consciousness can
never lead to the unity of apperception. The decisive text is B134:
For the empirical consciousness, which accompanies different cognitive
states, is in itself diverse and without relation to the identity of the subject.
That relation comes about, not simply through my accompanying each
cognitive state with consciousness, but only in so far as I conjoin the
contents of one cognitive state with those of another, and am conscious
108 KANT'S TRANSCENDENTAL PSYCHOLOGY
of the synthesis of them. Only in so far, therefore, as I can unify a diversity
of elements of given cognitive states in one consciousness, is it possible
for me to represent to myself the identity of the consciousness across
these cognitive states [amended translation, original emphasis; compare
A117a].

No individual cognitive acts can reveal the unity of apperception. This

unity only comes about through the syntheses that must be performed
on cognitive states for cognition to be possible and that create a synthetic
unity across the states. Further, we can only recognize that unity and
represent it to ourselves by recognizing these syntheses. Although Kant
may have been tempted to believe that certain cognitive activities require
reflective awareness of our own states, he sees clearly that individual
cognitive acts cannot lead us to justifiable claims about the unity of
apperception. Thus, the heart of his account of apperception is that the
unity of mental life is the outcome of synthesizing activities required for
cognition and can only be understood as such.

Arguing for the Synthetic Unity of Apperception

Apperception and Representation

There can be no doubt that Kant maintains that mental life involves a
certain unity, the unity of apperception. If he does not simply beg the
question against Hume's denial of unity—because he certainly knew of
it—then some of his many remarks about apperception must indicate
where he thinks his predecessor erred. All of Kant's analyses of cognitive
tasks establish the need to synthesize cognitive states. As noted, how-
ever, many also begin with nonskeptical assumptions. To answer Hume,
he must appeal to some fact about mental life that his predecessor takes
for granted.
Kant seems to offer just such an antiskeptical argument at the dramatic
moment when he proclaims the doctrine of apperception. The message
and the prominent treatment are the same in both editions. I begin with
the A version:
We are conscious a priori of the complete identity of the self in respect
of all cognitive states which can ever belong to our knowledge, as being
a necessary condition of the possibility of all cognitive states. For in me
they can represent something only in so far as they belong with all others
 Replying to Hume's Heap 109
to one consciousness, and therefore must be at least capable of being so
connected. [A116, amended translation, my underscoring].
Kant's fluid terminology raises a problem, since he claims that all cog-
nitive states [Vorstellungen] must represent, although he does not always
hold this.55 The problem is merely terminological, I believe. Just prior
to this passage he suggests that intuitions are nothing to us, unless they
can participate "directly or indirectly" in consciousness. Kant acknowl-
edges unconscious cognitive states, as well as intuitions and judgments.56
I believe that his point is that any cognitive state, properly so called,
must either represent something to a subject or contribute to such a
representation (and so participate indirectly in it). This is a property
that Hume could not deny. He clearly believes that both impressions
and ideas represent some content or other to subjects. If Kant is right
that any cognitive state can be representational only if it belongs to, or
could be connected to, a synthetically produced unity of apperception,
then Hume is wrong.
The B edition fastens on the same property, the ability of cognitive
states to represent:
It must be possible for the I think to accompany all my cognitive states;
for otherwise something would be represented in me which could not be
thought at all, and that is equivalent to saying that the cognitive state would
be impossible, or at least would be nothing to me... The thought that the
cognitive states given in intuition one and all belong to me, is therefore
equivalent to the thought that I unite them in one self-consciousness, and
[that] presupposes the possibility o f . . . synthesis... (B132-34).
Here Kant's point emerges in two steps: If cognitive states did not belong
to one consciousness, then they would not represent; cognitive states
belong to one consciousness in that they are connected or are connectible
by synthesis.
Obviously, these are not full-dress arguments, but only cryptic point-
ers about what is wrong with opting for disconnected heaps of cognitive
states. Still, Kant must have believed that these remarks were supported
by his various discussions of the need for synthesis, since it is not obvious
that representation requires synthetic connection. For an argument
against Hume, he would need to cover two cases, impressions and ideas.
(I will use Kant's "intuitions" and "judgments," because Hume's ter-
minology embodies the mistake that ideas are just like impressions,
except fainter.) Putting the issue in terms of cognitive tasks, Kant would
need to show that the capacity to have intuitions and judgments that at
110 KANT'S TRANSCENDENTAL PSYCHOLOGY
least represent something to a subject require syntheses that connect
intuitions and judgments to other cognitive states.

Judgments
The clearest discussion of the issue concerns judgments and occurs as
Kant prepares the reader in the A Deduction:
Without the consciousness that what we think is the same as what we
thought a moment before, all reproduction in the series of cognitive states
would be useless. For it would in its present state be a new cognitive state
which would not in any way belong to the act whereby it was gradually
generated, and the diverse elements would never form a whole, because
the unity would be lacking that only consciousness can provide to it. If,
in counting, I forget that the units, which now hover before me, have
been added to one another in succession, I should never know that a total
is being produced through this successive addition of unit to unit, and so
would remain ignorant of the number [A103, amended translation, my
emphasis].
Although this discussion runs several points together (which I disentan-
gle later), it brings out a crucial consideration. A thought or judgment
can have a particular content only if it has been generated from particular
types of intuitions.
To see the problem Kant is addressing, consider what it is about a
cognitive state that would enable it to have the content "five things" or
"there are five things." As was clear in Chapter 3, resemblance is not
an option at this point. Not only do judgments not resemble intuitions,
but intuitions do not resemble objects. So a judgment cannot have the
content "there are five things" by somehow resembling five things. A
judgment achieves content in being generated from a particular set of
intuitions. But for its mode of generation, it would not have this content.
Kant's claim is not merely that intuitions are used to generate, produce,
or cause judgments, but that if particular intuitions were not involved
in the generation of a judgment, then it would not have its particular
content, and so would not be the judgment that it is. So the judgment
depends for its existence on the intuitions that are used to generate it.
I have noted that Kant's introductory account of synthesis suggests that
diverse elements of earlier states are "comprehended"—somehow pre-
served and united—in the resulting state, so that syntheses create re-
lations of existential dependence. Here we get a less metaphorical
account of this process. It is not that syntheses gather up elements and
comprehend them as one might gather apples and place them together
 Replying to Hume's Heap 111
in a basket. Rather, the diverse elements of earlier states are preserved
in the later state, because the content of the later state is, in part,
constituted by its generation from those particular types of intuitions.
This point is very close to the first half of Kant's famous summation
of his position: "Without sensibility no object would be given to us,
without understanding no object would be thought. Thoughts without
content are empty, intuitions without concepts are blind" (A51/B75).
Since sensibility provides intuitions, Kant is asserting that thoughts
would lack content without intuitions. Given his belief that thoughts are
generated from intuitions, there is only a small step to the position that
thoughts would lack content unless they were generated from intuitions.
The point is somewhat obscured in the passage, because Kant suggests
that individual cognizers can complete the ascent to judgment only by
recognizing how their states were generated! Although he retreats from
the claim that we are conscious of the generation of a cognitive state in
the very next paragraph, it complicates the discussion. As I note in
Chapter 3, the vacillation in this passage57 (first he claims we watch the
synthesis, then that we are just aware of the outcome, and then that we
must assume that we are aware of the generation) may be a result of
confusing levels. Since, when theorizing about judgment, we can explain
how judgments have particular contents only by reference to their mode
of generation, Kant mistakenly assumes that in judging itself, we must
take note of the generation of our state. Leibniz's idea that some types
of cognition require a reflective awareness of our state may also con-
tribute to the confusion.
Despite these complications, Kant's analysis of how judgments can
represent makes a powerful case that they must be related by synthetic
connections (and so stronger than Humean connections) to other cog-
nitive states. Here he does not start with objects and ask how we rep-
resent them; he starts with judgments and asks how they can represent.
This enables him to show that the representational character of judg-
ments cannot simply be assumed. It must be explained, and the only
viable explanation involves their mode of generation, hence syntheses
and synthetic connection to other states.

Kant's Functionalism
In this crucial passage Kant suggests an account of the representational
content of judgments that is like that defended by contemporary func-
tionalists.58 Functionalists take cognitive states to have particular con-
tents in virtue of their causal relations to stimuli, responses, and other
112 KANT'S TRANSCENDENTAL PSYCHOLOGY

cognitive states. (One important disanalogy that I consider later con-

cerns the importance of behavior to content.) Could Kant have held a
view that is so close to contemporary wisdom? Besides the texts cited,
three external pieces of evidence support this interpretation. The first
was noted in Chapter 3. By the time Kant was writing, causal connection
was widely held to be essential to representation, and he always sup-
ported this view.59 However, he also believed that judgments are not
in an immediate [causal] relation to objects but that they are only related
through intuitions (A68/B93). Hence, it would be natural for him to
believe that the representational character of judgments could only arise
through their dependence on intuitions.
Second, Kant's position is not very different from the account that
he had been reading in Tetens:

The reference of cognitive states to other preceding modifications [of the

mind] is their essential characteristic....

And everything that we call a representation [Vorstellung] of anything

arises from such modifications of our being which refer to other preceding
alterations....

The analogy of representations [Vorstellungen] with alterations of the soul,

of which they are what remains behind, makes them capable of being
signs and images of them. They correspond to them.... The reference of
representations to preceding modifications is the general analogy between
cause and effect.60

Tetens seems to maintain a hybrid between a resemblance view and a

causal view. Cognitive states can represent, because they are caused by
modifications of the soul [which are themselves caused by objects],
because there is a general analogy between causes and effects. Kant
departs from this position by dropping the vestige of resemblance con-
tained in the final clause.
Finally, a somewhat speculative consideration. Functionalism has
been widely adopted, because philosophers have come to realize that
"nothing is intrinsically a representation of anything," in Daniel Den-
nett's phrase.61 Once this point is fully appreciated, some relational
account of content seems inevitable. At the time Kant was writing, the
hopelessness of theories of intrinsic representation was clearly in evi-
dence. Representation by resemblance was thoroughly discredited, as
was Wolff's idea of a vis representiva. Under these circumstances it would
not be surprising for him to adopt a relational view.
 Replying to Hume's Heap 113

Intuitions
What about intuitions? The crucial contention that cognitive states can
represent only if they are connected by synthesis in a unity of apper-
ception explicitly includes intuitions in the A edition (A116). Presum-
ably, this is also the import of "intuitions without concepts are blind."
Once we consider the problem of how judgments could have particular
contents, the solution of their method of generation seems straightfor-
ward. With intuitions, it is hard to see what the problem is. Kant in-
troduces the notion of "intuition" by claiming that it is the representation
[Erkenntniss] through which we are in immediate relation to objects
(A19/B33). This is not quite right, as the discussion of "objects of
representation" in Chapter 3 shows.62 Intuitions must themselves be
constructed from cognitive states that are the result of immediate, but
fleeting and ever-changing, sensory contact with objects. Still, the fluid
terminology raises no substantive problems. Intuitions are in relatively
immediate contact with objects compared to judgments. Given this fairly
direct contact, it might seem that intuitions have a particular content,
because they are regularly produced by the presence of particular ob-
jects. To see why this relation is insufficient to account for the repre-
sentational quality of intuitions, consider a popular example from
philosophy of psychology. The height of the mercury in a thermometer
varies regularly with the temperature, but it does not represent the
temperature—at least, it does not represent the temperature to the
thermometer.
The question about intuitions is how they can be (or function as)
representational for a subject. At B132 Kant asserts that a cognitive
state that could not be connected with others in one consciousness would
be impossible or at least would be "nothing to me." As noted earlier,
the same locution is used at A116—intuitions are nothing to us unless
they are at least capable of being connected in the unity of apperception.
Kant is quite insistent on this point, and mentions it in several other
passages (A117a, A120). These remarks are probably aimed at the Leib-
nizean doctrine of petites perceptions, which had received prominent
treatment in the recently published New Essays. They also contribute
to a case against Hume, however.
Leibniz offered some good reasons for believing in unconscious per-
ceptions. For example, the argument about perceiving the roar of the
sea. People standing on the shore hear the roar by hearing the noise of
each wave, but these cannot be separately distinguished.63 Kant defends
unconscious ideas in the Anthropology, by repeating this reasoning for
114 KANT'S TRANSCENDENTAL PSYCHOLOGY
the example of seeing a man in a far-off meadow.64 In these circum-
stances he says that we are not conscious of the ideas, but mediately
conscious of them. We can know that we have them by inference. The
roar of the ocean is, however, only the tip of the iceberg of Leibniz's
doctrine of petites perceptions. By appealing to the fact the mind always
thinks, the principle that nature never takes leaps, the identity of in-
discernibles, and preestablished harmony, he argues that each individ-
ual's petites perceptions reflect the entire universe.65
This bloated metaphysical doctrine appears to be the subject of Kant's
repeated denials. Unconscious, inert "perceptions" would represent
nothing to their subjects, because they have no effects in cognitive life.66
They would be no more to their subjects than levels of mercury are to
thermometers. Despite the soundness of this basic point, Kant's specific
claim is too strong. Intuitions can represent, can be something to us,
only if they can be reported in conscious judgments or if they can have
an indirect influence on other cognitive states. (The second clause also
applies to unconscious cognitive states. These must contribute to states
that do represent.) He overlooks, or perhaps rejects, a possible alter-
native, a perennial danger in transcendental psychology. If a cognitive
state covaries with an external stimulus and produces appropriate be-
havior, then there is some (perhaps weak) sense in which it represents
that object to its subject.
Kant considers this possibility explicitly in a well-known letter to
Marcus Herz, written in 1789.
[Unless the categorial condition for the unity of apperception is met] I
would not even be able to know that I have sense data; consequently for
me, as a knowing being, they would be absolutely nothing. They could
still (I imagine myself to be an animal) carry on their play in an orderly
fashion, as cognitive states connected according to empirical laws of as-
sociation, and thus even have an influence on my feeling and desire,
without my being aware of them.... This might be so without my knowing
the slightest thing thereby, not even what my own condition is.67
Even if cognitive states did influence behavior, they would still make
no contribution to cognition, unless they could be synthesized with other
cognitive states. Although this is an interesting point, it does not show
that intuitions could not represent at all, unless they are synthesized in
judgments.

The Reply to Hume

If sound, the considerations Kant invokes to banish hordes of petites
perceptions would also enable him to conclude that Humean impressions
 Replying to Hume's Heap 115
must belong to the synthetic unity of apperception. For if they did not,
they could not be representational, or even contribute to representa-
tions, and so could not be cognitive states at all. Together with the
analysis of judgments, this would establish his general claim that to
represent something, or to contribute to a representation, cognitive
states must belong to a synthetic unity of apperception. Although the
reasoning is not up to the mark, it comes close enough. The case for
judgments is sound and the reflections about intuitions show that totally
unconnected Humean impressions could not be representational. Be-
cause they depend for their representational character either on behavior
or on synthetic connection with judgments, they cannot exist separately.
No cognitive state, properly so called, can exist separately and in
isolation.
Through his analyses of how intuitions and judgments can represent
at all, Kant is able to meet Hume's challenge about mental unity in a
very satisfying way. Hume despaired because he had two principles (all
perceptions are distinct, the mind never perceives any real connection
among distinct existences) that he could not reconcile with his own
psychological account, which presupposes a single mind. Rather than
simply favor his own theory, he conceded defeat. Kant's reflections
about the prerequisites of representation provide an ideal way out. Since
the two principles are also in conflict with something that no one in this
tradition can deny—cognitive states are representational, or contribute
to representation—they must give way.
The considerations about representation are only part of Kant's over-
all demonstration of the unity of apperception. Every cognitive task that
he analyzes yields the same result. Cognition requires syntheses and
syntheses create a unity of apperception, in the sense of actual depend-
ence relations among cognitive states. They also presuppose another
kind of unity—the ability of cognitive states to be synthesized with each
other. This is the central result of the subjective Deduction: All aspects
of cognition require a synthetic unity of apperception. Hence, any phi-
losopher who reflects on the necessary conditions for knowledge must
recognize the unity of apperception. For students of cognition, if not
for cognitive agents, every cognitive state must be apperceptive.
In the Deduction Kant insists over and over, and then over again, on
the unity of apperception, because someone who mattered denied it.
He also claims that many important results can be derived from the
doctrine of the unity of apperception. This is a legitimate premise for
further arguments, only because, at the very moment when he announces
the unity of apperception, he explains how denying mental unity involves
a serious mistake about the nature of cognitive states. In Chapter 6 I
116 KANT'S TRANSCENDENTAL PSYCHOLOGY
will return to the central argument of the Deduction and follow its
development from the doctrine of apperception and the analyses of
cognitive tasks. As promised, the next chapter develops and extends
Kant's views about synthetic connection and the unity of apperception
as a plausible account of the unity of a mind. As I will argue, his analyses
of the prerequisites of cognition yield a deep understanding of the
grounds of mental unity.
 5
A Cognitive Criterion of
Mental Unity

Unity of Apperception as Mental Unity

Synthetic Connection
Through his analyses of the prerequisites of cognition, Kant discovers
a connection that links cognitive states. Even the most minimal cognitive
task—the capacity to have cognitive states that represent something to
the subject—demands a synthesis of states (or their diverse elements)
in further states. But acts of synthesis both create and presuppose re-
lations among cognitive states. Synthesis creates a relation of depen-
dence. The resulting state depends for its content, and so for its existence
as a particular cognitive state, on the existence of the earlier states.
Synthesis also presupposes that the earlier states are synthesizable, that
their diverse elements can be synthesized in a resulting state.
To make discussion clearer, I shall describe the state that results from
synthesis as a "synthetic product" and the states that are used in the
production of that state as "synthetic progenitors." We can then define
a general relation of connection by synthesis between any two cognitive
states: CS! and CS2 are connected by synthesis just in case they are
related as synthetic product and synthetic progenitor, or they are co-
progenitors of some further state. Although this may seem cumbersome,
Kant's basic idea is simple. Synthetic connection is a relation of con-
tentual connection. Synthetic products are contentually dependent on
synthetic progenitors. The elements of coprogenitors are combined in
further states. Kant's way of putting this appears needlessly abstract:
117
118 KANT'S TRANSCENDENTAL PSYCHOLOGY

Diverse elements [Mannigfaltigen] of progenitors are comprehended

[begreifen] in the product state (A77/B103). Contemporary philosophers
and cognitive scientists would say instead that information in states that
are closer to the sensory periphery is processed and so transformed into
useful perceptual or conceptual information. The superficial clarity of
our term "information" is quite misleading, however, for it suggests
both that lower-level states are, in fact, representational, and that there
are no particular problems in understanding how they can be so. Even
if "Mannigfaltigen" (particularly in Kemp Smith's rendering, "mani-
folds") is annoyingly elusive, it has the advantage of only implying what
Kant knows, and what we know. Sensory contact with objects induces
a vast and ever-changing array of changes in our internal states, which
we can somehow use to create useful perceptual and conceptual rep-
resentations, whose contents depend on those changes.

Connection and Connectibility

Both the A and B Deductions present the unity of apperception in terms
of the connection and connectibility of cognitive states by synthesis.
In A:
All cognitive states which can ever belong to our knowledge... [must]
belong with all others to one consciousness, and therefore must be at least
capable of being connected in it... .
This synthetic unity [of apperception] presupposes or includes a syn-
thesis. . . . [This] productive synthesis of the imagination can take place a
priori1;... [it is] prior to apperception and is the ground of the possibility
of all knowledge [A116-18, amended translation].
Cognitive states belong to a unity of apperception, because they have
been connected to each other by synthesis. However, Kant adds an
important qualification: Cognitive states must be at least capable of being
connected (with all others) in one consciousness. So possible connection
seems sufficient.
This vacillation (does belonging to the unity of apperception require
actual or possible synthetic connection?) results from the reciprocal
relations between synthetic connection and apperception. The end of
the cited passage is quite clear. Syntheses precede apperception, making
it and knowledge possible. On the other hand, Kant introduced his topic
as follows: to investigate "the inner ground of the connection of cognitive
states . . . in one knowledge . . . , we must begin with apperception"
(A116, amended translation). This suggests the reverse dependency.
 A Cognitive Criterion of Mental Unity 119
Synthetic connection of cognitive states and knowledge depend on the
unity of apperception.
This "chicken and egg" problem is not deeply worrying, however. It
arises because, as noted, synthesis both creates one kind of connection
among cognitive states, contentual connection, and presupposes an-
other, the connectibility of progenitor states. To see Kant's solution,
consider whether a set of states that were connectible by synthesis but
that involved no actual connections would belong to a unity of apper-
ception. Kant's answer is clearly no. The unity of apperception (and
cognition) can only be generated by actual syntheses (A118; compare
B134, B135). Once connections are established among cognitive states,
once we have a functioning thinker, then other cognitive states that are
connectible to those already connected would also belong to the same
thinker. For a set of states to belong to one consciousness, each must
at least be capable of being connected [by synthesis] "with all others"
(A116).
In sum, cognitive states belong to one consciousness just in case they
are connectible by synthesis with a set of states already connected by
synthesis, and all such states are connectible with each other. The B
version defends the same view:
As my cognitive states (even if I am not conscious of them as such) they
must conform to the condition under which alone they can stand together
in one universal self-consciousness, because otherwise they would not
without exception belong to m e . . . .
The thought that the cognitive states given in intuition one and all belong
to me, says after all, no more than that I unite them in one self-
consciousness, or at least can unite them in it; and although this thought
is not itself the consciousness of the synthesis of the cognitive states, it
presupposes the possibility of that synthesis. That is, only because I can
comprehend their diverse elements in one consciousness do I call these
one and all my cognitive states [B132-34, amended translation, original
emphasis].
To belong to the unity of apperception cognitive states must be able to
belong to one universal self-consciousness; that is, they must all be able
to be synthesized with each other. At the end of the section, Kant
reaffirms that cognitive states must be brought to the unity of apper-
ception by means of a synthesis (B135). So again, a unity of apperception
is created when cognitive states are connected through syntheses; a state
belongs to a given consciousness if it can be synthesized with cognitive
states already connected by synthesis.
120 KANT'S TRANSCENDENTAL PSYCHOLOGY
The doctrine of apperception is plainly an abstract account of the
unity of a mind. It is almost never considered in this light, however.2
Two factors appear to be responsible for the neglect. One is the failure
to realize that Kant's claims about a unity of apperception were made
against the background of Hume's denial of mental unity. Second, Kant's
account of mental unity highlights the operation of synthesis and "syn-
thesis" is taken to be too psychological to be worth pursuing in philo-
sophical discussions. When the Human background and the abstract
psychological operation of synthesis are left out, however, there is little
left to the doctrine. This may explain why it has been assimilated to the
cogito or regarded as a rather empty analytic claim: All my states are
mine.

The Plan of the Chapter

In this chapter I argue that we continue to ignore this account of mental
unity to our own disadvantage. The question of the grounds of mental
unity is pressing in both philosophy and the cognitive sciences. Philos-
ophers have venerable moral and metaphysical interests in understand-
ing the unity of a mind. More recently, psychologists and cognitive
scientists have realized that the apparent unity of consciousness is some-
thing that needs to be explained, and that it is rather hard to do so.
Kant's cognitive analyses offer a fresh and promising approach. Starting
with the most basic fact about a mind—its capacity for knowledge—he
argues that the states of any mind, like ours in achieving knowledge on
the basis of a vast and fleeting array of sensory stimulations, must be
connected or connectible by synthesis. Whatever other factors may or
may not unite cognitive states in a mind, synthetic connection is fun-
damental and invariably present. Failing this relation, our minds would
not be minds at all. Since it emerges from attempts to understand cog-
nition, this is a cognitive, as opposed to a bodily, moral, or legal account
of mental unity.
As with other results in transcendental psychology, the cognitive ac-
count of mental unity is highly abstract. In the remainder of this section,
I will try to clarify Kant's position by making explicit an important point
that he leaves implicit and by explaining why he cannot assert something
he occasionally appears to assert. The resulting account is still very
preliminary. Nevertheless, I will demonstrate its strengths in the next
three sections: first, by comparing it to contemporary rivals, the theories
of Locke and Leibniz; second, by comparing it to current theories of
mental unity; and finally, by considering some probable objections.
 A Cognitive Criterion of Mental Unity 121

Although this chapter is mainly laudatory, I conclude my examination

of apperception by addressing an awkward question. What is the relation
between this doctrine and Kant's other views about the self? One reason
scholars have avoided Kant's ideas about the self is that he has too many
of them. The self has so many roles to play in his system that the doctrine
of apperception threatens to become entangled with other concerns
about the self. I suggest a way of preserving this sensible account, but
only at the expense of other views that he would not happily sacrifice.
So the last section is deliberately revisionist.

Refining the Account of Synthetic Connection

Synthesis is, I have argued, Kant's solution to Hume's problem about
how cognitive states are connected to each other, and so united in a
mind. Synthesis involves relations of contentual connection: the ele-
ments of earlier states being combined in later states, the contents of
later states depending on the elements of earlier states. If synthesis only
involves contentual connection, however, then it may be too liberal as
a criterion of mental unity. If Harry screams, "Smoke!", then I will
have a cognitive state whose content depends on the content of one of
Harry's states. We are never of one mind, however, except figuratively.
Kant is unconcerned by this type of case, because his account of synthesis
involves a restriction that is not obvious,3 although it is an immediate
consequence of his general view of the sources of knowledge. Harry's
cognitive state can affect one of mine only by causing a sensation in me
via outer sense. Kant's implicit restriction is that synthesis operates
without benefit of transmission through outer sense. That is, two cog-
nitive states, CSj and CS2, can only be synthetic progenitors of a resulting
state CSR, if the contents of CSR depend on the elements of CS, and
CS2, without the mediation of some additional state, CS3 (an outer
sensation). Either or both of CS, and CS2 may be outer sensations. But
if they are coprogenitors of some third state, then its content cannot
depend on elements contained in CS, or CS2 courtesy of some additional
outer sensation. This (implicit) restriction is violated by the putative
counterexample. If this restriction on connection by synthesis is borne
in mind, then Kant's position is still that cognitive states belong to the
same consciousness if they are connected or connectible by synthesis.
By viewing the doctrine of apperception in light of Hume's criticisms
of simple selves, we can recognize a familiar pattern. Although Kant
opposes the skeptical conclusion, his position preserves a number of the
skeptic's insights.4 Persons are not mere heaps of perceptions, but we
122 KANT'S TRANSCENDENTAL PSYCHOLOGY
are no more than contentually interconnected systems of cognitive states
(at least as far as we can ever know). Selves are not substances, nor are
they anything that cognitive states are connected to. These negative
aspects of the apperception doctrine are not fully explored until the
Paralogisms chapter (the subject of Chapter 7). Even in the Deduction,
however, it is reasonably clear that cognitive states belong to the unity
of apperception, not by virtue of belonging to something else, but be-
cause they stand, or can stand, in relations of synthetic connection with
each other.

Is the Self the Combiner?

This point is repeated in a number of texts already cited (especially
B133-34, A118). Nevertheless, significant counterindications in the De-
duction need to be addressed. At B159 Kant declares, "I exist as an
intelligence which is conscious solely of its power of combination...."
Further, in many discussions the I is presented as what does the com-
bining: "I unite them," "I combine them." Taken at face value, these
expressions imply that a thinker is not a contentually interconnected
system of states, but that which connects cognitive states. This inter-
pretation is given further support by the fact that Kant links spontaneity
and apperception quite closely.5 So it is tempting to interpret the self,
or consciousness, or apperception as something like the power or source
of spontaneity itself.
While tempting, this approach does not lead to a coherent position.
The self cannot be identified with acts of spontaneity, since these are
distinct events. It could only be the agent that performs these acts. But
acts or processes of synthesis could not be performed by agents. They
are unconscious activities within agents that enable them to have cog-
nitive capacities required for agency. In Daniel Dennett's useful ter-
minology, they are "subpersonal" processes, not acts performed by
persons.6
As we saw in the preceding chapters, Kant is not always clear on this
point. Nevertheless he is clear, and explicit, that we have no way of
identifying the empirical consciousness that accompanies [or synthesizes]
different cognitive states (B133). Further, were selves identified as pow-
ers or faculties of spontaneity, then self-identity would be a matter of
faculty identity, and Kant would need an account of faculty identity,
which he does not present. This omission is unsurprising, since it is hard
to see how faculty identity could rest in anything but the identity of a
substance, and he explicitly denies that self-identity requires the identity
 A Cognitive Criterion of Mental Unity 123

of a substance.7 Thinking selves are not merely systems of cognitive

states, because some faculty must always be present to synthesize states.
These faculties are crucial in creating the unity of apperception. The
identity of consciousness is, however, a matter of the connections that
are or can be created; for the reasons just given, it cannot reside in the
identity of what connects. When Kant says that cognitive states belong
to a self only because I combine them, he is giving a synoptic version
of a more complex doctrine: Cognitive states belong to the unity of
apperception only because some faculty in whatever material or im-
material form in which those cognitive states are currently realized or
preserved creates synthetic connections among them.

Locke and Leibniz on Personal Identity

The Issue
The topic of this chapter has several names: "mental unity," "self iden-
tity," and "personal identity." Although the last expression is the most
common, I usually cast the discussion in terms of mental unity. As John
Perry has argued, the term "identity" has been misleading.8 For the
question of personal identity is not about the identity of two things.
There is no question of a gallant young officer being identical with an
aging general; the two have different properties, and so fail to meet the
uncontroversial criterion of the indiscernibility of identicals. The issue
of personal identity is about individuation. What relation between the
two different temporal stages of a person, the gallant young officer and
the aging general, makes them stages of the same individual? This ter-
minological confusion has produced substantive confusion. It led Ber-
nard Williams to claim that the memory criterion for personal identity
had to be rejected, because it could come into conflict with the logical
principle of the transitivity of identity.9 Since it avoids confusion and is
more perspicuous, I follow Perry in describing the issue in terms of
finding the "unity" relation for minds or persons. What relation unites
diverse states in one mind? For reasons that will be apparent later,
Kant's theory is more appropriately described as one of "mental" rather
than "personal" unity.

Leibniz Versus Locke

Kant was familiar with Locke's well-known views on the unity of con-
sciousness. In the Third Paralogism he echoes Locke's reason for con-
124 KANT'S TRANSCENDENTAL PSYCHOLOGY

eluding that continuity of the thinker does not require continuity of

substance. For all we know, the contents of cognitive states can be
transmitted from one thinking substance to another (A362-64a).10 Al-
though he follows Locke up to a point, he does not adopt the memory
criterion, but offers the deeper relation of synthetic connection. As we
have seen, Kant rarely engages in explicit critical discussions of com-
peting views. I shall suggest later that he makes a glancing objection to
Locke, in the course of laying out his own position. In this case it is
clear why he had no need to belabor the shortcomings Locke's view.
Leibniz provides extensive citations from Locke's discussion and a com-
pletely adequate criticism of the memory criterion in the New Essays.^1
Leibniz reiterates the standard eighteenth-century objection: Memory
does not make someone the same individual through time. His reasons
are interesting, however, and shed considerable light on the background
to Kant's reflections.
within each substance there is a perfect bond between the future and the
past, which is what creates the identity of the individual. Memory is not
necessary for this, however, it is sometimes not even possible, because of
the multitude of past and present impressions which jointly contribute to
our present thoughts; for I believe that each of a man's thoughts has some
effect, if only a confused one, or leaves some trace which mingles with the
thoughts which follow it. One may forget many things, but one could also
retrieve them, much later, if one were brought back to them in the right
way [my emphasis].12
The memory criterion is too crude. As has often been noted, actual
memory is too strong a requirement. Leibniz's more original point is
that memory is also too crude in failing to appreciate mental dynamics.
Like Hume, Locke pictures mental life as a sequence of separate states.
Aside from some states containing memories of earlier states, there is
no commerce among them. Leibniz observes that this is unfaithful to
the phenomena. Earlier experiences affect later thoughts in many ways
other than simply permitting recall of the experience itself. These in-
teractions reveal continuity of mental life as much as memory does.
Leibniz goes further and conjectures that since perceptions and thoughts
must leave some trace in the mental apparatus to be recalled, each
induces a permanent change that affects all subsequent activity.
Kant would accept many of Leibniz's points. Actual memory is an
unrealistically strong criterion, and memory is, in any case, too super-
ficial. Memory rests on synthesis, for the state to be remembered can
represent some state of affairs only through the operations of synthesis.
As Leibniz notes, there are many other influences of earlier states on
 A Cognitive Criterion of Mental Unity 125

later ones, and these also rest on synthesis. So, for example, an asso-
ciative connection across states would depend on the existence of syn-
thetic connections, as we saw in Chapter 3.13 Memory is only one
manifestation of the relations that are possible among cognitive states
through the basic operation of synthesis.14 In fixing on memory, Locke
has failed to uncover the deeper ties that unite a mind, so Kant wisely
rejects memory in favor of a cognitive criterion.
Although Kant would be sympathetic to Leibniz's criticisms, he could
not agree with the positive doctrine. Nothing but metaphysical specu-
lation supports the idea of a "perfect bond between the future and the
past... which creates the identity of the individual." Kant replaces this
metaphysical bond with an argument from transcendental psychology.
We can account for the many relations among cognitive states and for
the presence of very basic cognitive capacities only by assuming synthetic
connections among cognitive states. Since it rests on argument by in-
ference to the best (or better, only) explanation, rather than on meta-
physical assumptions, Kant's account is clearly superior to Leibniz's.

Moral Responsibility
Although Kant's cognitive theory is deeper and better grounded than
either of its contemporary rivals, there is another side to the story. For
all its problems, memory would have an important virtue as the unity
relation for persons. Locke casts his discussion in terms of this virtue,
and Leibniz recognized it in the Discourse on Metaphysics (which an-
tedates Locke's Essay).15 "Person" is, as Locke puts it, a "forensic"
term that is used in the assignment of moral and legal praise and blame.
With the memory criterion, a person would be responsible only for those
acts where his own conscience can accuse him. Further, moral respon-
sibility requires that the agent be conscious of his actions. What Locke
wants to claim is that continuity of consciousness—the very capacity
that enables us to be responsible—is what makes the same person.16
Even though consciousness of one's actions is essential to being a person,
it does not follow that the continuity of a person depends on continuity
of some power of consciousness. As we saw earlier, this approach leads
to a substantival account—or lapses into obscurity. Locke does not see
how we can assume that sameness of person requires sameness of sub-
stance in our present state of ignorance. He tries to avoid obscurity by
assimilating sameness of consciousness to memory. But the two are not
the same and memory is too strong a criterion. So none of this works,
and the potential virtue of a memory criterion remains unrealized.
126 KANT'S TRANSCENDENTAL PSYCHOLOGY

Still, the issue of finding a unity relation for persons that is adequate
to their moral dimension exerts tremendous pressure on Kant's discus-
sion. Both Locke and Leibniz (in the Discourse) try to force the capacity
that is necessary to being a person, self-consciousness, into the role of
the unity relation for persons. Kant's term "apperception" and his def-
inition of "person" in terms of the capacity for self-consciousness in the
Third Paralogism are clear indications of the influence of his predeces-
sors. Thus, it is hardly surprising that he makes occasional references
to "self-consciousness" in describing the synthetic connections that unite
states in a self. The problem is that this solution is merely verbal. Even
though he recognizes the importance of self-consciousness to person-
hood, he has no idea of how we are conscious of our experiences or our
cognitive states as our own.

The Problem of Self-Consciousness

Kant's bewilderment is understandable if we reflect on his historical
position. Descartes believed that we are directly aware of a continuing
mental substance; Locke urged that the same consciousness accompanies
all our mental states; in the Discourse, Leibniz claimed that we have a
memory or consciousness of immediately past happenings that gives us
knowledge of the I.17 When pressed, however, these theories of self-
consciousness and self-identity collapse. Besides begging the question
in favor of the immateriality of the soul, Descartes gives no indications
of how we tell that we are the same mental substance.18 Locke's position
becomes obscure, once the assimilation of consciousness to memory is
seen to be erroneous. Leibniz recognized the shortcomings of memory
and reduced its role in the New Essays to informing us (usually reliably)
about our past; he now takes self-identity to be constituted by the meta-
physical continuity of a substance.19 Kant saw the failures of his pred-
ecessors. He criticizes Descartes at length in the Paralogisms. And he
explicitly dismisses the "same consciousness " idea of Locke and the
early Leibniz (without mentioning either by name) at B133: "the em-
pirical consciousness that accompanies different cognitive states, is in
itself diverse and without relation to the identity of the subject." This
leaves him with a problem, because self-consciousness is essential to
our status as persons, and he has no way of accounting for it. His
half-hearted solution is that we recognize a cognitive state as our
own, because we have introspected the synthesizing activity that pro-
duced it! (A108) Kant is driven to this idea by the need to explain self-
consciousness within his theory of mental unity—synthetic connections
 A Cognitive Criterion of Mental Unity 127

among cognitive states—and encouraged to think that it is true by Te-

tens' psychology.20 Still, as we have seen, he cannot quite bring himself
to accept this alleged help from introspection, and he denies synthesis
watching almost as often as he puts it forward.21 Still, it is unfortunate
that he ever mentioned it, since synthesis watching has provided a perfect
vehicle for ridiculing the psychology of the Critique.
Apperception is not primarily a theory about the consciousness of
mental states as belonging to ourselves. As noted, at the end of B132,
Kant declares that "all my cognitive states (even if I am not immediately
[gleich] conscious of them as such) must conform to the conditions that
are necessary for them to belong together in one universal self-
consciousness . . . " (my translation, my underscoring). Those conditions
are that they be able to be synthesized with each other (B133). Self-
ascription has seemed central for two reasons. Commentators have rec-
ognized the Lockean-Leibnizean vocabulary, "apperception," "self-
consciousness," without always appreciating how muddled Locke and
Leibniz were about self-identity and self-consciousness, and a fortiori
without recognizing that Kant saw these muddles and tried to avoid
them. The second reason has been the tendency to read Kant in light
of Descartes' agenda for philosophy.22 Since Descartes assumed only
the cogito, his successors must play by the same rules. So the deduction
of the categories must proceed by assuming only that we can self-ascribe
mental states. In Chapter 6 we shall see that the categories enter the
account, because (Kant argues) cognitive states are connectible with
each other only if they can also be subsumed under the categories.
Synthetic connection and so the categories are crucial for the entire
range of our cognitive capacities, however, and not just for our ability
to ascribe states to ourselves. Further, the categories and synthetic
connection are not sufficient for self-ascription. So there is no espe-
cially close connection between apperception, the categories, and self-
ascription, despite the current popularity of understanding the
deduction in relation to these issues. Kant has no serious views about
how the distinctive abilities of persons relate to the continuity of a mind
with certain capacities. These abilities are simply grafted onto his ac-
count of mental unity in terms of synthetic connection. For this reason
his theory is more accurately described as one of "mental" rather than
"personal" unity.
Although Kant fails to advance our understanding of self-ascription,23
this is not a major objection to his theory of mental unity. As already
noted, the fact that an ability is crucial to being a person does not imply
that continuity of the faculty that underlies that ability is necessary for
128 KANT'S TRANSCENDENTAL PSYCHOLOGY

continuity of the person. This failing might also seem critical for a
different reason. John Perry has argued that there is a persistent con-
fusion in the personal identity literature between a metaphysical ques-
tion ("What relation holds between two mental states when they are
states of the same mind?") and epistemological questions ("How do we
tell when two states belong to the same mind?" "How does the subject
tell when past states are his?").24 This confusion is due in part to the
use of the term "criterion," which has both an epistemological and
metaphysical sense. Kant is also a victim of confusion. He conflates the
theorist's problem of specifying and defending a view of the connections
that unite cognitive states in individual minds (and so the metaphysical
problem, although he would not like that term), with agents' problems
in knowing which cognitive states are their own. Hence, the recurring
references to synthesis watching. Since these issues are separable, his
inability to explain how we come to recognize our own states does not
compromise his defense of synthesis as the ground of mental unity.

Modern Mentalism, Wiggins, and Parfit

Modern Mentalism
Kant's cognitive criterion is, I have argued, a sound reply to Hume, and
a far better account of mental unity than those provided by Locke or
Leibniz. I turn to contemporary theories. The last 30 years have pro-
duced a number of important studies on personal identity. Current work
may be divided into three positions: modern mentalism, which has been
the dominant trend, and the important dissenting views of David Wiggins
and Derek Parfit.25 Wiggins opposes mentalist criteria on the grounds
that they yield incorrect answers in "puzzle" cases; Parfit's disagreement
is more fundamental. He denies that there is any deep fact about per-
sonal identity about which criteria can be correct or incorrect.
Many recent writers, including Paul Grice, Anthony Quinton, Terr-
ence Penelhum, Sydney Shoemaker, John Perry, and David Lewis26
(among others), have argued for broadly mentalistic accounts of mental
unity. Although these accounts differ in details, the fundamental idea
is that the unity relation for minds includes some combination of con-
tinuity of memory, belief, and desire. Oddly, the contemporary impetus
for this consensus conies from Sydney Shoemaker's classic retelling of
Locke's tale of the prince and the cobbler.27 In the original version, the
two are said to switch souls, whatever that is supposed to amount to.
 A Cognitive Criterion of Mental Unity 129

Shoemaker made Locke's point immediately clear and compelling for

modern audiences by moving Mr. Brown's brain to Mr. Robinson's body
and then inquiring who the resulting person ("Brownson," to preserve
neutrality) is, Brown or Robinson. As Shoemaker argued, because
Brownson would believe that he was Brown, be devoted to Brown's
wife and children, be able to recall details of Brown's life, display
Brown's personality traits, likes, dislikes, mannerisms, and the like (or
at least those that don't depend on having a certain kind of body), and
because these psychological affinities could hardly be dismissed as co-
incidental, there is a strong inclination to say that he is Brown. That is,
there is strong reason to believe that the unity relation for persons
involves not bodily continuity but mental continuity.
Mental continuity is fairly similar to Kant's criterion of synthetic con-
nection, although it includes emotional as well as cognitive factors.
Nevertheless, many of the variations and amplifications of Shoemaker's
argument that have been offered over the years also provide support
for the cognitive criterion. Still, there are two salient differences between
Kant's view and contemporary accounts. One concerns how we under-
stand mental continuity; the other, how we understand the project of
finding the unity relation for minds. In both cases I will argue for the
surprising thesis that the older view is the better.
Modern mentalists take memory, belief, and desire continuity to in-
volve causal connections among mental states. So this aspect of mental
continuity is like Kant's relation of synthesis, except that synthesis only
concerns cognitive states.28 Current accounts differ in stressing the sim-
ilarity of a person's beliefs and desires across time.29 What justifies the
claim that mental unity requires similarity of beliefs and desires, in
addition to causal connection, however? No one knows how much sim-
ilarity in belief and desire people have over a lifetime. Perhaps there is
no similarity of belief and desire between the child's mind and that of
the adult, even though one evolved from the other.30 Causal dependence
might produce a large amount of similarity, so Kant's position is com-
patible with the idea that different stages of a person are psychologically
similar. His criterion does not prejudge the issue, however, and is for
that reason preferable to many contemporary accounts.
The second disagreement concerns the nature of the project. Kant is
trying to figure out what persons must be like—in particular, what con-
nections must exist among their cognitive states—for them to have the
cognitive capacities we recognize them to have. Recent studies have
often borne the stamp of linguistic philosophy. John Perry express the
prevailing understanding in the introduction to his collection, Personal
130 KANT'S TRANSCENDENTAL PSYCHOLOGY

Identity: "In studying the problem of personal identity, we are learning

about our own concept of a person, trying to articulate and analyze
knowledge we in a sense already have."31 Thus, Kant and our contem-
poraries seem to be addressing different questions. What relations must
obtain among cognitive states for persons to have certain capacities?
What relations must obtain among mental states for us to employ our
concept "same person"?32
In fact, these questions are intimately related and Kant's is the more
fundamental. Despite the frequent appeals to "what we would say"
about bizarre cases in the literature, contemporary work is not a pro-
tracted exegesis of the English expression "same person." The question
of personal identity concerns not the word "person" but the classification
'person', and so, what persons are like.33 To address this question, we
must first consider why we have two classifications, 'person' and 'human
being', that apply to pretty much the same organisms.34 Presumably,
the answer is that we use the classification 'person' to explain and predict
current behaviors, and the beliefs and desires that underlie them, on
the basis of past beliefs and desires. We need an additional classification,
beyond the biological one, because, tragically, some human beings never
exhibit the psychological regularities on which such predictions and ex-
planations are based, or only exhibit them for part of their lives. In
addition, "person" is used to indicate the presence of cognitive capacities
that make human beings moral agents. Given the point of the classifi-
cation—to indicate psychological regularities and capacities—we can
see, on reflection, that when we regard individuals as persons, we must
regard their mental states as connected by mental continuity. Puzzle
cases are one means of carrying out such reflections. When presented
with extraordinary cases, we realize that unless we regard the mental
states that are united by mental continuity as states of continuing per-
sons, then no individual in the case would have the properties and
capacities that are essential to being a person. So we say "same person"
where there is mental continuity. What we say is, however, a reflection
of what we believe about the traits and capacities distinctive of persons,
and about the relations that must unite states for an individual to have
those traits and capacities. That is, what we say depends on the answer
we give to Kant's question. Given the traits and abilities we acknowledge
persons to have, what relations must unite their cognitive states?

Wiggins's Argument Against Mentalism

Kant's cognitive criterion can receive additional support from arguments
for modern mentalism; equally, it is threatened by critics of this position.
 A Cognitive Criterion of Mental Unity 131
David Wiggins has argued that all mentalist criteria are fatally flawed,
because they would permit "branching" persons in unusual circumstan-
ces. For example, if a brain were duplicated or split by commisurotomy
and the two organs were implanted in different bodies, then mental
continuity criteria (including Kant's) would imply that each of the re-
sulting individuals was the same person as the original individual. So
someone might branch into two streams, while remaining the same
person.35
To avoid this result, Wiggins argues that we must use a hybrid criterion
that involves mental and physical factors. He suggests that "person" be
understood as a nonbiological qualification of "animal". Thus, there
might be different kinds of persons—human-persons, dolphin-person,
and so forth. The unity relation for person could then ride piggyback
on the unity relation for the particular kind of animal. If we have a
human being who has and retains the extra qualities that make him a
person, then he will continue to be the same person by virtue of con-
tinuing to be the same human being. The pure mentalist criterion would
be replaced by a hybrid: person capacities, plus the unity relation for
the relevant animal.
How does this proposal meet the difficulty, however? The same wor-
ries that have been raised about splitting persons could be raised about
splitting human beings. If science fiction is to be our guide,then we can
clone human beings, and so end up with branching persons cum human
beings. We do not seem to have made any progress. The salient response
to such fanciful considerations is that (almost) given by Lewis Carroll.
If pigs could fly, then we would use a different set of classifications for
describing the world. (Oddly, Wiggins offered a similar response himself
in an earlier discussion of the problem.36) More formally, Kant's cri-
terion can be adequately defended by noting that these considerations
are not relevant. As long as the problem of personal identity is under-
stood in terms of examining the limits of our concept 'same person',
then it might be necessary to probe those limits by asking what we would
say in extraordinary cases. As I have argued, however, this is not the
best approach and it is not Kant's approach. He wants to consider how
the cognitive states of people, constituted as we are, must be related to
ensure the functioning of basic cognitive capacities.

Parfit's Denial of Personal Identity

Derek Parfit has argued,in a series of writings, that we should no longer
think in terms of personal identity at all.37 His point is that even when
later mental states are psychologically connected to earlier ones, the
132 KANT'S TRANSCENDENTAL PSYCHOLOGY
similarities in views and values between the two person-stages may be
too feeble to justify a claim of identity and all that that entails. Thus,
he would reject the cognitive criterion. The ability of two cognitive states
to be connected by synthesis would not be sufficient to make them states
of a continuing person. On Parfit's view, we are wrong to fasten the
yes-or-no question of identity; what matters are the psychological con-
nections themselves, and these can hold to a greater or lesser degree.
The disagreement between Kant and Parfit seems to be verbal. Parfit
would agree with Kant that earlier states can and do influence the con-
tents of later ones. Further, he thinks that these relations of influence
are what is important. In cases where the psychological traits of a human
being have changed markedly over time, however, he would withhold
the label "same person."
Parfit would regard the difference not as verbal but as substantive,
for two sorts of reasons. First, he makes the common error noted earlier.
He construes personal identity as a matter of identity, not individua-
tion.38 If the issue is cast in terms of individuation, then his concerns
are eased considerably. Since individuals can alter with time, in talking
about a later stage of an individual, we are not attributing sameness in
the face of obvious differences. Parfit's second reason, or cluster of
reasons, focuses on the implications of saying "same person." These
range from legal punishment in finding the perpetrator of a long past
crime to terror in recognizing one's own nonimminent mortality.39 If we
are absolutists about identity, then, Parfit reasons, we may be too harsh
in reacting to distant misdeeds, and too frightened in contemplating the
death of someone who may not be very like us now, either physically
or mentally. Such worries are irrelevant to Kant's doctrine. He does not
construe mental unity as involving the continuity of a mystical Cartesian
ego, but in terms of a relation of contentual connection across states.
Since contentual connection need not imply great psychological simi-
larity, h would be perfectly consistent with his analysis in the First
Critique to appeal to facts about human psychology in setting up
appropriate social institutions,and in developing reasonable personal
expectations. Given such research, we might not punish someone for a
past misdeed, even though we acknowledge that we are confronting the
same mind. Locke himself realized that although "person" is a forensic
term, our legal practices do not invariably accord with our views about
the continuity of persons.40
Although Kant is an absolutist, he has nothing to fear from Parfit's
recent critique of absolutism. His position does not rest on metaphysical
error, nor need it lead to practical error. Conversely, the Critical in-
 A Cognitive Criterion of Mental Unity 133
vestigations into the unity of apperception raise serious doubts about
Parfit's claim that there is no fact of the matter about personal identity.
Kant advocates the cognitive criterion for theoretical reasons: Synthetic
connection is essential for basic cognitive capacities, so cognitive states
are parts of one mind if and only if they can be synthetically connected.41
Synthetic connection is a deep fact about mental life; it underlies mental
capacities that enable us to be persons. Further, it sets up a particularly
intimate relationship between past and present mental states. For the
past states that can be synthesized with present states exhaust our inner
cognitive resources for dealing with present problems. Given this the-
oretical justification for using 'same mind', and the absence of untoward
practical consequences, Parfit's basic contention that judgments of per-
sonal identity must be arbitrary (and can be pernicious) seems less
compelling.42'43

Objections Considered

Is the Cognitive Criterion Too Weak?

I conclude my argument for introducing apperception into contemporary
dialogues about personal identity and mental unity by examining three
likely objections. The doctrine that cognitive states belong to one mind
if and only if they can be synthetically connected to each other offers a
criterion of mental unity that is too weak, too strong, or too "a priori."
The criterion may seem too weak, because comentality is defined
in terms of which cognitive states can be synthesized with others.
Can under what circumstances? If we are willing to do enough tink-
ering, strange things can happen. So, for example, we might carefully
remove a small piece of Brown's brain and place it in a correspond-
ing location in Robinson's head, so that the cognitive states whose
memory traces were in that bit of gray matter might be synthesized
with Robinson's current states. The criterion is too weak, because
Brown's and Robinson's (and everybody else's) states can be
synthesized.
This implication is avoided by making explicit what the previous
formulation leaves implicit: can under fairly normal circumstances,
or, in Kant's preferred terminology, can constituted as we are. He
does not know how to add any details, however. The Critique's cen-
tral negative teaching about the thinking self is that we cannot deter-
mine the material or immaterial medium of thinking or information
134 KANT'S TRANSCENDENTAL PSYCHOLOGY
storage merely by analyzing the capacities required for cognition.44
The amended criterion may appear circular, because "normal" can
only be understood by reference to the way(s) that states of one
mind can be connected by synthesis. Cognitive states would belong to
the same mind if and only if they can be synthesized in the way(s)
that cognitive states that are states of one mind are synthesized. The
circularity is not vicious, however, because Kant is not trying to pro-
vide an epistemological criterion for telling which states belong to
one mind. He is addressing the theoretical or metaphysical question
of what the unity of a mind consists in. Given the capacities that we
attribute to minds, he argues, the states of individual minds must be
bound together by relations of contentual connection, however those
connections are normally brought about. This account is very prelimi-
nary, but not circular.

Is It Too Strong?
From a different perspective, the cognitive criterion may appear too
strong. Cognitive states are comental if and only if they can be syn-
thetically connected, but that implies that each of a person's cognitive
states is synthesizable with all the rest. Kant does not shrink from this
implication, but stresses it: "cognitive states [must] belong with all others
to one consciousness, and therefore must be at least capable of being
so connected" (A116, amended translation my emphasis). At other
points he refers to apperception as "all-comprehensive" (A123),
"thorough-going"(B133, A112), and "universal" (B132). Although this
is a stringent requirement, Kant has good reasons for insisting upon it,
and a strong motive to do so: Leibniz's wild proliferation of petites
perceptions.*5
Our mental life is spread out in time (A98-99). Some states are si-
multaneous, but the vast majority occur at different times. We may
think of a mind as a series of cognitive states, each with a temporal
parameter. To see the force of Kant's position, consider going through
the states in their temporal order. Suppose we are at state 492. The
cognitive criterion requires that for an earlier cognitive state to be com-
ental with 492, it must be possible for elements from that state to be
combined with elements in 492 to form a subsequent state. This could
occur directly or indirectly, if elements from the earlier state had been
combined in an intermediate state that could in turn be synthetically
combined with 492.46 Suppose, however, that this relation fails for some
state, 49. Kant's point is that if state 49 can make no epistemic contri-
 A Cognitive Criterion of Mental Unity 135

bution at all to a present state, then it is for the individual qua knower
as good as nothing (A116, B132). There is no point in assigning it to
the same mind as 492, because that assignment cannot explain any
cognitive attributes.
We may now turn to the states that follow 492. State 492 will be
comental with these states if and only if elements from 492 can be di-
rectly or indirectly combined with elements of each one to form
some subsequent state. Again, let us suppose that this relation fails
for some later state, 5118. Kant's point would be the same. We go
forward to state 5118. If state 492 can have no epistemic impact on
the now present state, then the states are nothing to each other. Fi-
nally, consider simultaneous states, 492 and 492a. If the existence of
492a is epistemically irrelevant to 492, then it is, Kant claims, mis-
leading to place them in the same mind. Such an assignment is com-
pletely idle in accounting for the cognitive attributes of any mind.
Thus, for any arbitrary state, that state will be comental with only
those earlier, later, or simultaneous states with which it is connected
or connectible by synthesis.

Is It Too A Priori!
Although Kant has a rationale for his stringent criterion, it may lead
to a different charge—the familiar charge that his theory is exces-
sively a priori (i.e., too far removed from actual or potential empiri-
cal data). In evaluating this charge, it will be helpful to have a less
abstract version of the doctrine. Kant regarded it as an open ques-
tion whether thinking is carried out by an immaterial or material sub-
stance or by a series of such substances. However, he presumably
believed that earlier mental states influence later states, because they
leave a trace in some medium or other. As we saw earlier, Leibniz
speculated that every perception leaves a permanent trace.47 Tetens
drones on incessantly about the sptiren (traces) left by previous men-
tal states. So presumably Kant thinks that a cognitive state has a po-
tential influence on later states through some sort of material or
immaterial trace.48 Given this somewhat more concrete model, we
can evaluate the charge of excessive apriority by imagining future sci-
entific discoveries that might threaten his doctrine.
Three possible results seem problematic. First, suppose it turns out
that not all cognitive states leave permanent traces. In fact, a number
of psychologists have speculated that information taken in by the senses
only reaches long-term memory through rehearsal (e.g., repeating the
136 KANT'S TRANSCENDENTAL PSYCHOLOGY

phone number you just heard several times) or through "deep process-
ing" (i.e., by being integrated with information that is already in long-
term memory).49 Consider the moment when someone hears a phone
number. Since that number might be rehearsed or might simply fit into
previous information, that intuition could be synthesized with previous
mental states. Thus, it would be comental with those states. Suppose,
however, that no further processing does occur and that no trace is laid
up. Two days later, that intuition will not be connectible by synthesis
with previous cognitive states.
Although this discovery would refute Leibniz's metaphysical specu-
lation, Kant does not prejudge the issue. His theory could accommodate
this development in either of two ways. First, it would be reasonable
to deny that ephemeral states that fail to leave traces are cognitive states.
In that case, they would not be included in the scope of his definition
of synthetic connection. Such states might have become functioning parts
of minds, but failed to and so should not be regarded as cognitive states.
A second response would be to relativize comentality to a time. Two
states would belong to the same mind, at a time, if, at that time, they
could be connected by synthesis. The set of cognitive states that belong
to the unity of apperception is constantly changing (by expansion)
through time. If there are traceless states, then that set would constantly
expand and contract. Since both options are reasonable, the discovery
that presumed cognitive states do not always leave traces would not
require us to reject the cognitive criterion.
Second, suppose that we discover that after 25 years, 38 percent of
brain traces disappear. Taking the set of traces laid down in human
brains in 1965, only 62 percent remain today. We may consider a par-
ticular cognitive state, 49, whose trace has vanished. Let us assume
further that although 49 could have been synthesized with other states,
it was inert. It has no trace and no derivatives. Now consider state 492
that occurred in 1970. For a long period of time these states were com-
ental, but they are no longer. Further, 492 was comental with 49 and
is comental with the current state, 5118, but 49 and 5118 are never
comental. Although this may seem perplexing, the cognitive account
can be defended by repeating the second response to the question of
traceless mental states. The set of states that belong to the unity of
apperception expands and contracts over time. Thus, states that are
comental at some times need not be comental at all times. This is both
logically sound and a natural response to the alleged discovery, inde-
pendently of prior theoretical commitments. For how else would we
 A Cognitive Criterion of Mental Unity 137

interpret the discovery of impermanent traces except in terms of tem-

porary constituents of minds or brains?

Modularity
Although the cognitive criterion could handle both these discoveries,
a third possible development raises serious questions. In recent years
there has been much discussion of the modular structure of the
mind. 50 Modularity means that the mind is composed of a group of
independent systems for dealing with different kinds of data. A cen-
tral tenet of the modularity hypothesis is that different modules do
not share information back and forth. In Zenon Pylyshyn's phrase,
they are " informationally encapsulated."51 Seemingly, the establish-
ment of the modularity thesis would refute Kant's view that the
states of one mind can all enter into relations of synthesis. But the
situation is quite complex. First, some versions of modularity are
compatible with Kant's position. For example, Jerry Fodor suggests
that sensory systems are modular, but they feed information forward
into a central processing system.52 Information about stereoscopic
depth might be processed by a module in the visual system but then
forwarded to the central processing system, where it can be com-
bined with information from other modules. Since states within that
module will have contributed information to the state that is for-
warded, they will be connectible by synthesis with states from other
modules. Still, other possible versions of modularity would be more
problematic. Suppose that different modules affect behavior directly,
without pooling their information at any stage. How should this situ-
ation be characterized? Is one body being directed by different
minds? Both Plato and Freud present this type of picture (although
neither should be thought of as a protomodularity theorist). Al-
though this option seems possible, in advance of knowing specific
details, it is hardly forced.
This objection reveals both the strength and weakness of Kant's
account. He develops the doctrine of apperception by carrying out a
transcendental psychology of cognition. If the problem is cast in
those terms, then I believe that his analysis is sound. Synthetic con-
nection is necessary for basic cognitive capacities, and therefore the
foundation of a necessary unity of apperception. Insofar as we con-
sider the mind exclusively in terms of its cognitive abilities, there is
no reason to classify states that can have no cognitive bearing on
138 KANT'S TRANSCENDENTAL PSYCHOLOGY

one another as belonging to the same mind. So Kant concludes that

such states would be "nothing to me." What the modularity objec-
tion shows is that this may be the wrong way to conceptualize the
problem. Perhaps a mind should not be considered exclusively in
terms of its cognitive abilities; perhaps it should be regarded as that
which directs the behavior of a particular body; or perhaps it should
be considered as a source of interests and intentions.
On the other hand, there is value in Kant's single-mindedness. Be-
cause he focuses exclusively on cognition, he is able to offer a rich
analysis that reveals that many cognitive tasks require certain connec-
tions across the states of a mind. Since a mind is engaged in cognition,
it is extremely important to have an account that explores the unity of
consciousness from this direction, even if that is not the only possible
or useful direction. Kant's account may well need to be augmented by
noncognitive factors, and so revised. This scenario hardly upholds the
charge of excessive apriority, however. It shows only that his analysis
is a function of his way of conceptualizing the problem—just as anyone's
analysis will be.

Summary of the Account

Chapter 4 and the first four sections of this chapter present an inter-
pretation of the transcendental unity of apperception that diverges
from the accepted wisdom on every major count. On my reading, ap-
perception is neither the first premise of the transcendental deduc-
tion, nor a version of the cogito, nor about the self-ascription of
cognitive states, nor about self-awareness, nor (in the pejorative
sense) a metaphysical doctrine. On the contrary, it is an adequate re-
ply to Hume's skepticism about personal identity that has much of
Hume in it. It presents a theoretical account of the unity of a mind
that is grounded in the attempt to understand how basic cognitive ca-
pacities are possible. As such, it goes far beyond the frequently cited
work of Locke and Leibniz on personal identity and can also illumi-
nate contemporary debates. This may seem too good to be true, but
considerable internal and external evidence reveals that the doctrine
I attribute to Kant is a central theme of the Deduction chapter.
Were that chapter read on its own, apperception would probably
have a much less negative reputation. But, of course, it never is.
When considered in light of Kant's other remarks about the self, and
in light of another feature of transcendental idealism, both the doc-
trine of apperception and transcendental psychology generally begin
 A Cognitive Criterion of Mental Unity 139

to look like transcendental doubletalk. In the last section, I confront

these problems and suggest how this valuable work may be pre-
served. As noted earlier, this calls for some fairly revisionary steps.

Apperception and Kant's System

Too Many Selves

I discussed the problem of "too many selves" briefly in Chapter I.53
Officially, the Critique maintains that there are two selves or that the
self may be viewed from two perspectives.54 From one perspective, it
is understood as "phenomenal" or "empirical," "passive," subject to
natural laws, and hence unfit to be the object of moral criticism. Ac-
cording to the other, the self is "noumenal," completely unknown and
unknowable, but morally evaluable (Bxxvii-xxix). As noted, the prob-
lem is that it is not clear how the I of apperception can be fitted into
this scheme. Kant is staggeringly ambivalent on this fundamental point,
and in the end he refuses to categorize it as either phenomenal or
noumenal:
The I think expresses an indeterminate empirical intuition, i.e., percep-
tion. Something real that is given, given indeed to thought in general, and
so not as appearance [phenomenon], nor as thing in itself (noumenon),
but as something which actually exists, and which in the proposition I
think is denoted as such . . . [B422-23n, my underscoring].
It is not hard to guess what is driving Kant to a position that is
incoherent in his own system. As we saw in Chapter 4, he occasionally
entertained hopes of using the spontaneity of apperception to make a
case for transcendental freedom. Although that project is abandoned
in the Critique, the possibility of transcendental freedom is still supposed
to be established by the phenomenal-noumenal distinction. The prob-
lem is that apperception falls on the wrong side of this distinction, and
so threatens to undermine its point.
Given the impossibility of noumenal knowledge, the doctrine of ap-
perception must present a phenomenal, if highly abstract, aspect of the
self. If Kant is right that anything of which we can have knowledge must
be governed by causal laws, then synthetic connection will be lawful.
Imagination, understanding, and reason will all have their own laws.
The I that thinks will be phenomenal and causally determined. This I
is, however, too close to us. It is the I with which we identify. What
140 KANT'S TRANSCENDENTAL PSYCHOLOGY

possible comfort—or even interest—could we have in knowing that some

noumenal self is free and potentially immortal, when the self with which
we identify, the thinker, is phenomenal? I believe that Kant refuses to
acknowledge the phenomenal character of the doctrine of apperception
to avoid this devastating implication. I see no coherent alternative,
however. The doctrine of apperception can only be phenomenal, and
so we might as well admit this fact,55 whatever the fallout for doctrines
in the later Critiques.

The Ideality of Time

Within the context of Kant's system, a second, major problem for the
doctrine of apperception, and transcendental psychology generally, is
the ideality of time thesis. Strawson gives the essential outlines of this
issue in his argument for the fundamental incoherence of transcendental
idealism.56 The various activities that are described in the Deduction's
account of how the mind influences (or might influence) what we know
can only be understood temporally. They are processes and so take
time.57 According to the Aesthetic, however, the mind's activities pro-
duce time. So they cannot take place in time.
I agree with Strawson that trying to understand the activities described
by transcendental psychology atemporally requires "traversing the limits
of intelligibility."581 also agree that central tenets of Kant's philosophy
produce intractable inconsistency.59 Not surprisingly, I demur on the
point that the incoherence is largely the fault of transcendental
psychology.60
Transcendental psychology is consistent with itself. There is no conflict
between the transcendental psychology of the Deduction and that of
the Aesthetic. The psychology of perception in the Aesthetic establishes,
for example, that we can perceive space only because our senses can
construct three-dimensional percepts from two-dimensional retinal data.
This hardly implies that space is not three-dimensional, however. It
shows only that we cannot infer that space is three-dimensional from
what we perceive (without further argument). Kant's arguments that
space and time are forms of intuition do not establish that they are not
real; and they are certainly not what convinced him that they are not
real. The reasoning runs in the opposite direction. He was convinced
of the unreality of space and time on metaphysical and scientific grounds.
In the Antinomies, he presents the metaphysical considerations. Typi-
cally, the obvious scientific background is left implicit. The lengthy,
brilliant, but fundamentally inconclusive debates over the nature of
space and time convinced Kant, I believe, that science (i.e., Newtonian
 A Cognitive Criterion of Mental Unity 141

physics) could not assign any determinate character to space or time.61

Since any real thing must have a determinate character, however (A39/
B56), space and time are not real, as far as science is concerned. Given
this conviction, and the contemporary puzzles about how we perceive
space, Kant then argued that the space we perceive only reflects the
way we perceive.
In one guise, the point that I have been arguing has been familiar
since the nineteenth century.62 It does not follow from the fact that space
and time are forms of intuition that they are not also real.63 Even though
the character of our perceptions is not evidence for time, time could be
shown to be real on theoretical grounds. Its reality could be established
if a well-supported theory required us to assume the passage of time.
Kant nearly considers this type of argument. Intelligent people unani-
mously object to the ideality of time thesis on the grounds that alterations
are real (A36-37/B53-54). At this point, what is needed is some good
theoretical explanation that requires alterations. Kant's own theory of
synthetic connection would do, but maddeningly he does not pursue the
objection in this or any other fruitful direction. Instead, he turns it into
a straw man. Intelligent people are not confident about alteration on
theoretical grounds, but because of the changes in their cognitive states.
But our perceptions only reflect the forms of intuition, so the objection
fizzles. Again, I suspect that Kant is unwilling to reintroduce theoretical
arguments about time, because the most sophisticated and fundamental
arguments (from physics) had already proved inconclusive.
In any case, Kant's transcendental psychology does not suffer from
internal inconsistency.64 The inconsistency is between the theory
of apperception and transcendental psychology generally and the doc-
trine of the ideality of time. If time is not real, then the accounts of
the identify of a mind through time, and of the cognitive processes
that enable us to have knowledge, are incoherent. Conversely, if the
theories of synthetic processing and apperception are true, then time
is real.
Under these circumstances I see no choice but to reject the meta-
physical claim, which is, in any case, independently problematic. This
is a drastic move within Kant's system. Still, it does not mean a total
rejection of transcendental idealism. For in keeping transcendental psy-
chology, we also keep the source of all the arguments that our knowledge
is influenced by the structure of our minds. And that doctrine is at least
as central to transcendental idealism as the theses of the ideality of space
and time. Although more devout Kantians will undoubtedly find my
proposals extreme, I see no other way to preserve Kant's important
work on apperception and the prerequisites of cognition generally.
 6
Perceiving Times and Spaces:
The Cognitive Capacity
at the Center of the Deduction

Cognitive Tasks, Apperception,

and the Deduction of the Categories

Chapters 3 and 4 followed Kant's explorations of three cognitive tasks:

How can we represent objects?
How can we make judgments about objects?
How can cognitive states be representational for a subject?
Chapters 4 and 5 showed how these analyses led to the development of
a persuasive account of mental unity, the unity of apperception. This
material constitutes a substantial part of the objective and subjective
sides of the Deduction chapter. What is missing from the account so far
is what is missing for most of the chapter itself: the connection between
these task analyses and apperception, and Kant's goal of establishing a
special status for the 12 categorial concepts. That is the project for this
chapter. I will use these task analyses and the account of apperception,
plus a more specific analysis—the analysis of perceiving spatial and
temporal arrays—to show how Kant's transcendental psychology enables
him to construct one long argument for the universal applicability and
indispensability of the categories, an argument that stretches from the
Aesthetic through the Principles.
In the prefatory material to the Deduction chapter, Kant describes
his goal. The deduction of the categories must show, "how subjective
conditions of thought can have objective validity, that is, can furnish
142
 Cognitive Capacity at the Center of the Deduction 143

conditions of the possibility of all cognition of objects" (A90/B122,

amended translation). Since the subjective conditions of thought are the
categories, the goal is to show how the categories furnish conditions for
the possibility of all cognition of objects.
Those conditions are explored in the objective deduction. Among
other factors, knowledge of objects is possible only if we can represent
objects and make judgments about them. As we saw in Chapter 3,1 we
can represent objects on the basis of our constantly changing stream of
sensory inputs only if we have a productive imagination that can syn-
thesize representations from cognitive states. At this point, we need to
be explicit about a requirement that was only implicit in the earlier
discussion. Cognizers must have rules of synthesis through which actual
cognitive states can be combined to form representations of objects.
More formally, individuals can achieve representations of objects only
if one of their functions of synthesis is defined over the cognitive states
that objects produce in them. Here is the problem. The states produced
would seem to depend on the particular objects encountered. How,
then, could we ever be sure that we can have knowledge of all, or any,
of the objects we will encounter? Kant believes that the categories solve
this problem by providing universal functions of synthesis. Had he a
sound argument for this view, then he would be able to show that the
categories supply a crucial condition for the possibility of cognition.2
His first attempt at this line of argument, in the A edition, is quite
unpersuasive, however. As noted,3 he assumes that the functions that
permit the construction of representations of objects also permit the
representations to be brought under a concept (A105). He then assumes
that the categories supply the universal functions of synthesis that permit
thought (and concept application) (Alll). 4 So he may conclude that
the categories make it possible to construct representations of objects
and so supply a critical condition for the possibility of knowledge.
The objective side of the B deduction focuses on judgment. Knowledge
of objects requires being able to make judgments about objects. In turn,
as we have seen,5 judgment presupposes a synthesis that constructs rep-
resentations of objects and then requires a further synthesis of those rep-
resentations. Kant believed that judgment provided a more direct
argument for the special status of the categories. But this argument is
much too quick. He takes as given the assertion of the Metaphysical De-
duction that the synthetic functions that permit judgment are functions
associated with the categories (B143). Thus, once again, he can claim that
the categories supply a condition for the possibility of knowledge. He has
yet to provide a convincing argument for this view, however.
144 KANT'S TRANSCENDENTAL PSYCHOLOGY
Kant's reliance on the Metaphysical Deduction's inadequately sup-
ported identification of the basic forms of judgment with the categories
distorts the argument of the Deduction chapter. To show that the cat-
egories have objective validity (i.e., furnish the conditions for the pos-
sibility of all cognition of objects), he would have to show that they
must be involved in all our perceptions and thoughts or judgments of
objects. Given this identification, however, he feels no obligation to
explain how the categories make judgments possible. Instead, all his
efforts to justify the categories are concentrated on explaining the "by
no means obvious," fact that "objects of sensible intuition... must con-
form to the conditions which the understanding requires f o r . . . thought
[i.e., the categories]" (A90/B123). Only when he returns to the consid-
eration of individual categories in the Principles can we entertain any
hopes of finding arguments that thoughts or judgments are also subject
to the "conditions required for thought."
Kant does not try to establish a special status for the categories just
by examining the conditions required for cognition of objects in general.
Either because he realized the lameness of the arguments just presented
or because he wanted to convince skeptical opponents and so could not
assume that we have knowledge, or for both reasons, he also offers an
argument from apperception. In this case, the argument is to show that
the categories are indispensable, because they furnish conditions that
make apperception possible. This is one point on which all commentators
agree. Somewhere in the Deduction chapter is an argument from ap-
perception, and it is supposed to bear much of the burden of establishing
the special status of the categories. This clue only helps us to understand
the structure of the deduction, however, if we have a clear understanding
of the doctrine of apperception. In the preceding chapters I have argued
that apperception does not concern the ascription of cognitive states; it
is a theory about what must be true of cognitive states for them to be
states of one mind.6 They must be connected or connectible by synthesis
with one another. It is further confirmation of this interpretation of
apperception that it enables us to see quite clearly how the argument
from apperception for the categories is supposed to go.
If the unity of apperception is a matter of synthetic connections across
cognitive states, then it is fairly obvious what conditions are necessary for
the possibility of apperception. Apperception is possible only if a tho-
roughgoing connection of cognitive states by synthesis is possible. Once
again, there is a potential problem. Cognitive states will be connectible by
synthesis only if there are functions of synthesis that are defined over
them, indeed, defined over any pair of them (since all states that belong to
 Cognitive Capacity at the Center of the Deduction 145

the unity of apperception must be connectible with each other). But how
could we ever be sure that rules of synthesis are available to produce the
unity of apperception? This problem is simplified for Kant, because he has
already handled the case of thoughts or judgments. These are subject to
the rules of thought, which reflect or are associated with the categories.
Once judgments are constructed, they can be combined in further judg-
ments by the logical connectives.7 For example, the elements of "I am
cold" and "The cat is on the mat" can be combined into a further judg-
ment, "I am cold and the cat is on the mat." So when he raises the issue of
the possibility of apperception, he casts the problem in terms of intuition:
For the diverse cognitive states that are given in a certain intuition would
not be one and all my cognitive states if they did not all belong to one
self-consciousness. As my mental states... they must conform to the con-
dition under which alone they can stand together in one universal self-
consciousness ... [B132, amended translation; see also A122].

The problem of the possibility of apperception thus comes down to the

problem of how we can ever know that all our actual intuitions are
connectible by synthesis with each other. Again, the argument Kant
wants to make is that this condition on the possibility of apperception
can be fulfilled by the categories. They provide universal functions of
synthesis that can guarantee mental unity (A125).
The task analyses of the objective Deduction and the subjective Deduc-
tion's doctrine of apperception thus prepare the ground for similar argu-
ments for a special status for the categories. Representing objects and
judging objects required functions of synthesis that can be applied to cog-
nitive states produced by whatever our senses present us with; the possi-
bility of apperception requires functions of synthesis that allow us to unite
all our intuitions in a single self-consciousness. In both cases, these condi-
tions can be supplied by functions of synthesis associated with the cate-
gories. Or so Kant wants to argue. Although the requirements for
objective knowledge and apperception can both be satisfied by the cate-
gories , the assumptions behind these arguments are importantly different.
In one case the possibility of cognition is assumed, and it is taken to involve
perceiving objects and bringing them under concepts. This is clear as Kant
sums up this side of the argument in the A Deduction:
Actual [cognitive] experience, which is constituted by apprehension, as-
sociation (reproduction), and finally recognition of appearances, contains
in the last and highest of these merely empirical elements of experience,
certain concepts which render possible the formal unity of experience,
146 KANT'S TRANSCENDENTAL PSYCHOLOGY

and therewith all objective validity (truth) of empirical knowledge...

[A125].8
This is a very rich notion of cognitive experience, and the argument
that the categories render experience possible threatens to become vac-
uous. Since experience involves thinking about objects, it must be
subject to rules of thought which are allegedly associated with the cat-
egories. By contrast, the argument from apperception begins with the
most minimal cognitive task: the ability to have cognitive states that
represent some content to the subject. From this minimal assumption,
Kant argues for the necessary unity of apperception, and then from
apperception to the special status of the categories. The categories are
required for the possibility of apperception, and that is required even
to have cognitive states that are representational for the subject.
As already indicated, Kant makes several attempts to argue from the
necessity of syntheses to a special status for the categories, attempts that
fall short of their goal. At H 26 of the B Deduction, he tries again, by ana-
lyzing yet another cognitive task, our ability to perceive times and places,
to perceive A following B, or C above and to the left of D. This analysis al-
lows Kant to forge a link between the Aesthetic's doctrine of the forms of
intuition and the arguments about our ability to assign temporal and spa-
tial position to objects and events in the Principles, and thus to construct
one long argument to show that the categories can furnish conditions for
the possibility of knowledge and for the possibility of apperception.
In the end, I will argue that Kant's analysis—and so his best argument
for the categories—fails. Like his contemporaries and our own, he has
great difficulty understanding the relation between perception and cog-
nition, and so makes an error in transcendental psychology. Still, this
analysis offers valuable insights into the tensions and issues that must be
dealt with by any adequate theory of perception or cognition. At the end
of the chapter, I examine yet another cognitive task, the Second Analo-
gy's investigation of our ability to perceive temporal order. I consider this
task, because (as I shall argue) the best version of the deduction of the
categories is not completed until the argument of the Principles, and this
is the critical example. In this case, I argue again that the analysis pro-
vides important insights. Although both these analyses are independently
valuable, the central purpose of this chapter is exegetical. I try to show
that by taking a psychological approach to apperception and by consid-
ering the analyses of various cognitive tasks, especially the analysis of
perceiving times and spaces, we can see Kant as offering a fairly clean line
of argument for a special status for the categories. Even though the de-
 Cognitive Capacity at the Center of the Deduction 147

duction contains mistakes, it brings together a number of important con-

siderations about cognition in a powerful and coherent case for a special
set of concepts. The deduction is neither a patchwork nor a botch, but a
bold and clever attempt to carry out a sensible and potentially very im-
portant project: Given the cognitive capacities that we have and the re-
sources that we bring to them, are there any necessary and invariant
features of human cognition? This is why, after 200 years, the deduction
of the categories is still a fascinating and important piece of work.
Paragraph 26 of the B Deduction is about perception. Kant begins by
drawing attention to the fact that by "synthesis of apprehension I under-
stand the combination of the manifold in an empirical intuition, whereby
perception, that is, empirical consciousness of the intuition (as appear-
ance) is possible" (B160, my underscoring). He concludes the main body
of the section with the observation that, "All synthesis, therefore, even
that which renders perception possible, is subject to the categories"
(B161, my underscoring). This paragraph is difficult to follow, because,
although Kant tells us that his topic is perception, he does not inform us
about his own or his contemporaries' assumptions about perception.
("Empirical consciousness of the intuition as appearance" merely re-
places one difficult concept with four difficult concepts.) In the next three
sections, I fill in four key pieces of the background. "Perception: The
Eighteenth-Century Background" briefly presents some widespread sci-
entific and philosophical assumptions about perception. In "The Synthe-
sis of Apprehension in A," I consider the A edition's argument for
constructivism in perception, specifically, for the synthesis of apprehen-
sion. "A Role for Concepts in Perception in A" follows Kant's explora-
tion of the relation between concepts and perception in perceptual
recognition. With this material in hand, I return to the crucial arguments
of 126. Once we have seen where Kant (and others) have been, it will be
much easier to see where he is going, and why.

Perception: The Eighteenth-Century Background

The Standard View

The standard eighteenth-century view of perception included the fol-
lowing elements. As always, the central case was vision. We see by the
light reflected from objects causing images on the fundament or retina
of the eye. As William Porterfield put the point in his widely read A
Treatise on the Eye, we do not, however, "see any pictures painted on
148 KANT'S TRANSCENDENTAL PSYCHOLOGY
the retina."9 From the retina, nerves travel up through the optic chiasma
to the thalamus. Many believed that because of the order of the nerves,
when sensations are finally produced in the sensorium—when the neural
signals somehow become mental and conscious—the images are iso-
morphic to retinal images. These elements of the standard view were
all presented in Johann Gehler's encyclopedic Physikalisches Worter-
buch, oder Versuch einer Erklarung der vornehmsten Begriffe und Kunst-
worter der Naturlehre, which was being prepared as Kant wrote the
editions of the Critique.10

Intellectual Theories of Perception

As noted in Chapter 3, Thomas Reid's influential theory proposed an im-
portant distinction between sensations and perceptions. According to
Reid, we do not usually take note of our sensations, but overlook them.
Kant seems to endorse this view in the Anthropology, when he claims
that, if we are more conscious of the organ being affected than of the ref-
erence to an external object, then cognitive states that belong to outer
sense become inner cognitive states.'' Reid was also well known for main-
taining that perception involves a conception of the object and a belief in
its present existence.12 Since he first proposed it, the belief account of per-
ception has attracted many psychologists and philosophers, including
Kant's sometime opponent, Johann August Eberhard.13 Oddly, it has a
much stronger etymological warrant in German than in English, for the
literal meaning of "Wahrnehmung" is "taking [for] true."
Although Reid goes further in proposing that perception actually in-
volves belief, both Locke and Condillac had taken perception to have an
intellectual component. Locke described perception as "the first opera-
tion of all our intellectual faculties.. .. "14 Condillac echoed the senti-
ments: "The perception or the impression caused in the mind by the
agitation of the senses, is the first operation of the understanding."15
On the other hand, if Tetens is a reliable witness, then the question of
whether there was a recognitional-intellectual-conceptual element in
all representations that allow us to differentiate objects was a major and
disputed one.16

The Synthesis of Apprehension in A

A99
Turning to Kant's own views, 11 26 focuses on the "synthesis of appre-
hension that makes perception possible." Because this synthesis is not
 Cognitive Capacity at the Center of the Deduction 149
mentioned previously in the B Deduction, it is hard to see how we can
understand the phrase except by looking at the first edition, where it
occurs twice. The first reference comes as Kant prepares the reader for
the transcendental deduction. He tries to explain the need for a synthesis
of apprehension. Although our senses supply us with diverse elements,
these cannot be represented as a diversity and contained in a single
cognitive state without a synthesis of apprehension (A99). He gives the
example of constructing a representation of space and notes that, be-
cause the contents of this representation include a priori0 elements, the
synthesis of apprehension must be pure and a priori0.
Kant's reasoning in this passage is very dark. He claims, enigmatically,
that insofar as it is contained in a single moment of time, a moment
that can be distinguished from the moments it follows and precedes, a
cognitive state [or its contents] must be an absolute unity. At this point,
it is important—indeed essential—to recall where this is leading. Kant
is preparing us for the discussion of perception that occurs at A119-
20.17 The single representation that we are being asked to consider is a
single perception.

A119-20
The second discussion of the synthesis of apprehension in A is vastly
clearer. This synthesis is needed to produce an image from the sensory
data (A120). Kant repeats the point, with some elaboration, in a
footnote:
That imagination is a necessary ingredient in perception itself, no psy-
chologist has yet realised.... This is due partly to the fact that that faculty
has been limited to reproduction, partly to the belief that the senses not
only supply impressions but also combine them so as to generate images
of objects. For that purpose, something more than the mere receptivity
of impressions is required, namely, a function for the synthesis of them
[A120a; first sentence, my translation].
Why have psychologists failed to see the crucial role of imagination in
perception? One explanation may be that they subscribe to the standard
view presented in Gehler's Worterbuch. What we see corresponds to—
indeed is isomorphic to—the retinal image, either because of the ar-
rangement of nerves or for some other reason. Kant's objection to this
position is not that it is false, but that it is much less than the entire
account. Although they differed over the possibility of unconscious per-
ception, Locke, Leibniz, Reid, Condillac, and Tetens—and Kant—all
acknowledged that innumerable retinal images never reach perceptual
150 KANT'S TRANSCENDENTAL PSYCHOLOGY

awareness. Hume misstated the case in claiming that (conscious) mental

life is nothing but a stream of perceptions in constant and incredibly
rapid flux; but his point is correct for unconscious retinal images. Under
these circumstances the claim that mental representations replicate the
topology of retinal images provides only a small piece of the account of
how we have stable and coherent visual images.
A common gesture in the direction of a solution was to suggest that
we attend to some mental representations and not to others. Condillac,
Reid, and Tetens all appeal to attention18; in the Anthropology, Kant
characterizes the power of apprehending given ideas to produce an
intuition as attentio (attention).19 This notion cannot bear much weight,
however. It is only a place holder to indicate some means of selecting
some elements of some cognitive states and discarding others. But se-
lecting and discarding is not enough, and this is Kant's point. If a stable
visual image of a scene is to be produced, then elements from the selected
representations must be combined.

Examples from the Politz Lectures

The student lecture notes published by Politz20 contain a remarkably
clear discussion of this issue, complete with three helpful examples:
My mind is always occupied forming itself an image of the diverse elements
by running through them. For example, when I see a city, my mind forms
itself an image of the object that it has before it by running through the
diverse elements. When later, a person comes into a room that is loaded
with pictures and adornments, he cannot make himself an image of it,
because the mind cannot run through it. He does not know where to begin
in order to make himself a likeness of the object. It has been reported
that when a stranger comes into St. Peter's in Rome, he becomes totally
confused because of the multiplicity of the splendors. The cause is: his
mind cannot run through the diverse elements to make a likeness of them.
The imitative faculty is the formative faculty of intuition. The mind must
employ many observations in order to make itself a likeness of an object,
by representing [for] itself the object from every other side. For example,
a city appears differently in the morning and in the evening. There are
thus many appearances of a thing according to different sides and points
of view. From all these appearances, the mind must make itself a likeness,
by taking them all together.2'
That is, we achieve perceptual images by the formative imagination
selecting and combining the data of sense. The Politz lectures must be
used with care. Few professors would want student notes to be taken
 Cognitive Capacity at the Center of the Deduction 151
22
as definitive of their views, Kant included. It seems reasonable to use
this material to illustrate doctrines that appear in the published writings,
however. Besides the discussion at A120 and A120a, the formative
(bildende) synthesis itself appears in the Critique in a highly relevant
context: "the formative synthesis through which we construct a triangle
in imagination is precisely the same as that which we exercise in the
apprehension of an appearance..." (A224/B271, my emphasis).23 That
is, the synthesis of apprehension is a formative synthesis.

The Case for the Synthesis of Apprehension

Kant's defense of the synthesis of apprehension, and so of a construc-
tivist view of perception, rests on three solid points. First, the task to
be explained is how we are able to achieve stable visual images on the
basis of an incredibly rapid flux of retinal images. This assumes only
what all will grant: We have stable visual images; the retinal data are
remarkably fluid. Second, insofar as the senses are understood as re-
cipients of information from the ambient environment, the creation of
visual images cannot be attributed to the senses. Descartes had intro-
duced the notion of a "common sense" (which was also taken up by
Locke) where data from the various senses were to be combined to
produce a common representation of the object. This doctrine may be
part of (or much of) the explanation of why psychologists have failed
to appreciate the need for imagination in perception. Kant's objection
to it is straightforward. If sensibility is characterized as the faculty that
receives information, then, in the absence of empirical investigation, it
cannot also be characterized as the faculty that combines information.
Finally, insofar as imagination is understood merely as a faculty for
reproducing the data of sense, it is not up to the task. Reproducing the
order of sense would simply reproduce the flux from the retina. Since
copying and reproducing are inadequate to the demands of the task,
Kant reasons that we must have a constructive faculty. The perceptual
image represents together aspects of the object that coexist in it but
that were not sensed together. As he notes explicitly in B, the imagi-
nation is commonly characterized as the faculty of representing in in-
tuition an object that is not present (B151). So he assigns the
construction to a constructive or productive imagination.
This analysis of perception is a special case of problem of constructing
a representation from diverse cognitive states that we considered in
Chapter 3. And it shares strengths and weaknesses of the general anal-
ysis. Although Kant defends the need for a synthesis of apprehension,
152 KANT'S TRANSCENDENTAL PSYCHOLOGY

he does not justify the claim that it is a priori0. Again, all he shows is
that the law of association is inadequate. If he could support the assertion
at A99 that the synthesis of apprehension is responsible for the con-
struction of representations of space (or time), then he would have a
more reasonable claim that it is a priori0.

A99 Revisited
This interpretation of A120 and A120a receives some additional con-
firmation from the fact that it allows us to make some sense of the earlier
discussion at A99. Kant's claim that a synthesis of apprehension is nec-
essary if diverse elements are to be contained in a single state repre-
senting that diversity is given considerable support by the later
discussion. By reflecting on the implications of that discussion, I believe
that we can also begin to see what Kant is driving at when he claims
that insofar as it is contained in a single moment, a cognitive state
contains an absolute unity. The point is easier to grasp by starting with
an example of perceiving a temporally extended event. Suppose we are
listening to the first few measures of "F«r Elise." We do not merely
hear the notes, we hear the melody. For this to happen, our awareness
of individual notes must enable us to construct a state that represents
the relations among those notes. The playing of the melody occurs in
time, and we take in auditory information about it in time, but to hear
the melody, we must bring that temporally disparate information to-
gether at some point.24 That is, we must represent information that is
temporally disparate in two senses—we receive it at different times, and
it informs us about different times—in a single moment. Although not
all objects of perception are spread out in time, our perceiving of them
is, because perceptual images are constructed from diverse sensory data.
That is the point made forcefully by A120 and A120a. Even though we
perceive very quickly, data are received at slightly different moments.
For us to form a perceptual image of the object, diverse elements taken
in at different times must be represented together in some single mo-
ment. We might have a perceptual image that endures for some time.
However, that time will be made up of moments, each one of which
contains information gathered at diverse times. Given Kant's construc-
tive account of perception, it follows that cognition is possible only if
our cognitive lives are made up of such moments. I believe that this is
at least part of what he is trying to get at in stressing the unity of
momentary cognitive states, and the need to differentiate such moments
from each other.
 Cognitive Capacity at the Center of the Deduction 153

A Role for Concepts in Perception in A

The Need for Nonproductive Synthesis

The first edition thus provides a clear and convincing defense of the
need for a constructive synthesis in perception. This synthesis is attrib-
uted to a productive imagination, as opposed to the reproductive, which
merely follows the law of association. But what does the productive
imagination follow? As noted in Chapter 3, in the A edition, Kant
assumes that there is some intellectual element involved in any repre-
sentation or perception.251 objected that this point was too controversial
to be assumed. The A edition makes another connection between con-
cepts and perception, however, and this connection seems sound. In
perceptions that lead to recognition in a concept, the imagination follows
rules of synthesis associated with concepts.

Perceptual Recognition
As he prepares the reader for the A deduction, Kant examines the
"synthesis of recognition in a concept." His point is that for perception
to lead to recognition—subsuming the perceived item under a concept—
there must be a synthesis of the diverse elements gathered by sense,
and that synthesis must be associated with the concept (A103). The
discussion of this passage is rather convoluted, and in the end, he only
manages to link the imagination to the faculty of concepts, the under-
standing, through an intermediary discussion of apperception. We can
appreciate his point about the role of concepts in perceptual recognition
more easily by turning to the discussion of concept application in the
Schematism.

Concept Application
Kant believes that concepts are associated with schemata, or rules for
producing different kinds of images. We apply concepts to presented
objects by noting in some unconscious way that the imagination followed
the same procedure in constructing a present image that it followed in
previous cases (A140/B179-80). That is, concept application involves a
comparison of procedures for constructing representations, rather than
a comparison of images themselves. How is this possible?26 We have a
finite store of concepts and associated schemata with which the proce-
dure for generating an image of a current scene may be compared. If
154 KANT'S TRANSCENDENTAL PSYCHOLOGY

the imagination tried all possible ways of selecting, discarding, and com-
bining the diverse elements provided by the senses in order to find a
match with one of the stored schemata, the process would be endless.
There are too many possible permutations and combinations. To yield
matches, one of three possibilities must be realized. (1) The selecting
and combining is directed from the top down, so that the imagination
tries various procedures, taken from our repertoire of stored schemata.
(2) Because it operates according to innate or acquired principles that
are well coordinated with our conceptual repertoire, the imagination
automatically uses procedures to generate images image that permit
matches. That is, there is something like a preestablished harmony be-
tween the faculties, so that the imagination uses procedures that permit
matches with stored schemata. (3) There is some combination of top-
down direction, and coordination between the faculties.
Although I have cast the issue in terms of Kant's specific theory of
concept application, the range of options is the same on more standard
theories, just as long as perception is taken to be constructive. Concept
application requires us to compare either a current representation or
something like a procedure for generating that representation, with
stored representations of concepts, whether they be schemata, images,
or whatever. So for concept application to be possible, the imagination
must generate something—its own procedure or a product of its pro-
cedures—that can be compared with stored representations of concepts.
In which case, the processes that subserve perception must be subor-
dinate to or coordinate with the processes involved in concept appli-
cation. Otherwise, the perceptual processes might synthesize endlessly
without ever permitting the application of a concept.
The crucial point is not the particular theory of concept application,
but the argument that perception is constructive. Without this assump-
tion the argument collapses, because the entire system could be data
driven. We would not select among sensory representations, but merely
forward them to some location in the mind-brain where they are com-
pared with stored copies of previous sensory representations. In these
circumstances neither subordination nor coordination would be re-
quired. Once perception is shown to be constructive, however, it follows
that, in concept application, the imagination and the understanding must
be related by subordination or coordination. The discussion in K 24 of
the B edition suggests that Kant leans toward subordination, but a
passage from the Anthropology employs a coordination model.27 For
the purpose of understanding the argument in H 26, we need not choose
among these options. Whether through coordination, subordination, or
 Cognitive Capacity at the Center of the Deduction 155

some combination of the two, in concept application, the imagination

is able to produce representations or procedures that match stored rep-
resentations of concepts. I will refer to the syntheses of the imagination
that produce these representations as "C-functions," because they sub-
serve concept application. Since we can apply concepts to objects
through perception, we know that the imagination can construct per-
ceptions on the basis of sensory data through the use of C-functions.
With some understanding of current assumptions about perception,
particularly Kant's own views about the necessity of a synthesis of ap-
perception and the role of concepts in perceptual recognition, I believe
that we can make considerable sense of the dense argumentation in
1126.

H 26 in the B Deduction

The Centrality of H 26
Kant is explicit about his goals for this section:
We have now to explain the possibility of knowing a priori, by means of
categories, whatever objects may present themselves to our senses, not
indeed in respect of the form of their intuition, but in respect of the laws
of their combination.. .. For unless the categories discharged this func-
tion, there could be no explaining why everything that can be presented
to our senses must be subject to laws which have their origin a priori in
the understanding alone [B 159-60, original emphasis].

That is, this section is going to explain the by no means obvious fact
that objects of sensible intuition must conform to the categories. But
this was a major goal that Kant announced right at the beginning of the
Deduction (compare A90/B123)! Presumably, he thinks that it has not
yet been attained. The final section of the chapter (f 27) merely reca-
pitulates the argument and relates the conclusion to other possible so-
lutions to this set of problems. Clearly, 126 is intended to make a crucial
contribution to the deduction of the categories.
In a classic study, Dieter Henrich argued for the importance of this
section from a slightly different perspective.28 Henrich drew attention
to a striking textual puzzle. At both H 20 and f 26, Kant announces that
he has now completed the deduction of the categories. The inconsistency
is mitigated somewhat by the immediate observation in H 21 that he has
really only begun the deduction; more work remains to be done. On
156 KANT'S TRANSCENDENTAL PSYCHOLOGY
the basis of these comments, Henrich proclaims that any acceptable
interpretation of the deduction must explain what is new in H 26, or
between H 20 and 11 26, which permits the completion of the deduction.
He calls this interpretive criterion the problem of the "two-steps-in-one-
proof." Although this constraint on interpretations seems persuasive, I
shall disagree with his own solution to the problem later.

Perception as "Scanning an Image"

After introductory remarks about his goals for the section, Kant presents
his broad topic: the synthesis of apprehension required for perception.
As noted, he then characterizes perception as the "empirical conscious-
ness of the intuition (as appearance)" (B160). Of course, Leibniz had
denied that perceptions are always conscious, but Kant's usage does not
appear to beg any substantive questions about the existence of petites
perceptions. As we have seen, Kant allows for the existence of uncon-
scious cognitive states (Vorstellungen).29 He simply reserves the term
"perception" for states of which we are aware.30 Although this section
does not offer further explicit characterizations, some examples are
appended to the main discussion that clarify Kant's understanding of
perception. In particular, the example of perceiving a house gives a
fairly clear sense of how he conceives of the central case of visual
perception:
When, for instance, by apprehension31 of the diverse elements of a house
I make the empirical intuition of it into a perception, the necessary unity
of space and of outer sensible intuition in general lies at the basis of my
apprehension, and I draw as it were the outline of the house in conformity
with this synthetic unity of the diverse elements in space [B162, amended
translation, my underscoring].
The passage I underscore suggests that Kant takes perception to involve
something like scanning the contours and boundaries of perceived ob-
jects. We perceive when we draw a line around the objects boundaries
and other salient geometrical features with our mind's eye. If we cannot
do this much, then it is not clear that we are even differentiating an
object from its background, so this is quite a minimal sense of perceiving.
With some trepidation, I will describe this model of perception in terms
of "scanning an image." "Scanning an image" is an abbreviation for
"scanning the visible contours of an object represented in an image."
(We don't see—a fortiori, we don't scan—our perceptual images
themselves.)
Given this model of perception, we can sharpen the characterization
 Cognitive Capacity at the Center of the Deduction 157

of Kant's goals. He will try to explain how the categories apply to all
objects that we can perceive, in the minimal sense of being able to scan
images of them. Or to return to another formulation (to which he returns
at the end of the section): The goal is to establish the objective validity
of the categories by showing how they furnish conditions for the pos-
sibility of all cognition of objects—even this very minimal cognitive task
of scanning the image of an object.32 Since it is not obvious how the
categories might contribute to the possibility of scanning an image, and
so why this type of perception might be governed by them, Kant has
set himself a challenging project.

Perceiving Times and Spaces

Kant argues that the categories are needed for, and involved in, the
perception of all objects in space and time—even in this minimal sense—
by arguing that they are needed for, and involved in, perceiving times
and spaces. This essential lemma is established by analyzing the task of
perceiving times and spaces. Space and time are not merely the forms
of intuition, but themselves intuitions. "[They are also represented a
priori] as themselves intuitions which contain a diversity of elements and
therefore are represented with the unity of this diversity" (B160). We
perceive spatial and temporal arrays. We perceive A before B and C
above and to the left of D. How are we able to do this? This discussion
draws on a point that Kant adds to the B edition:
The mere form of outer sensible intuition, space, is not yet cognition; it
supplies only the diverse elements of a priori intuition for a possible
cognition. To know anything in space (for instance, a line), I must draw
it. and thus synthetically bring into being a determinate combination of
the given diverse elements . . . ; and it is through this unity of consciousness
that an object (a determinate space) is first known [B137-38, amended
translation, my underscoring].
The form of outer intuition guarantees that whatever two-dimensional
data we take in will be represented in a spatial grid. In and of itself,
however, the process form does not produce a representation of any
particular (or determinate) spatial region. To perceive spatial [and tem-
poral] regions, we must perceive (or imagine) objects and events in
spatial [and temporal] arrays.
Kant repeats this point, and expands upon it, in a notoriously con-
voluted note that appears in the middle of 1 26.
Space, represented as object (as we are required to do in geometry33),
contains more than mere form of intuition; it also contains combination
158 KANT'S TRANSCENDENTAL PSYCHOLOGY

of the diverse elements, given according to the form of sensibility, in a

cognitive state whose contents are intuitive, so that the form of intuition
gives only the diverse elements, the formal intuition gives unity of the
contents of a cognitive state. In the Aesthetic I have treated this unity as
belonging merely to sensibility, simply in order to emphasize that it pre-
cedes any concept, although, as a matter of fact, it presupposes a synthesis
which does not belong to the senses but through which all concepts of
space and time first become possible. For since by its means (in that
understanding determines the sensibility) space and time are first given
as intuitions, the unity of this a priori intuition belongs to space and time
and not to the concept of the understanding [B160a; amended translation,
my underscoring].

That is, again, we have representations of particular or determinate

spaces and times, and these cannot be accounted for by the existence
of forms of intuition alone. (Or at least, not by what we have so far
learned about the forms of intuition.) They presuppose a synthesis that
makes determinate or unified representations possible. What is so con-
fusing is that Kant first attributes the unified representation to the senses,
then he notes that it requires a synthesis that belongs to the understand-
ing, and finally he claims again that the unity belongs to the senses!

Differences Between the Editions

It is clear from the texts preceding and following the note that the
synthesis at issue is the synthesis of apprehension. To begin to get a
handle on this text, we must recall Kant's earlier discussions of this
synthesis. As noted, the first edition provides a solid argument for the
need for a constructive synthesis in perception, which is attributed to
the productive imagination. The dramatic opening of the B Deduction
signals a change. He proclaims that all combining or synthesizing must
be attributed to the understanding (B130). At the time of the first edi-
tion, he had realized that merely reproducing the data of sense in the
order in which they came would never lead to stable and coherent
images, but he did not elaborate (A120-21, A120a). Between editions
he must have seen the implications of this discussion more clearly. We
need a faculty of imagination even to achieve the level of perception,
and if a reproductive imagination is inadequate, a faculty of imagination
that is unguided would be worse. So he assumes that perceptual
syntheses must be carried out by the imagination guided by the under-
standing, or just by the understanding itself. Let us call the syntheses
required for perception "P-functions." In the preceding section we saw
 Cognitive Capacity at the Center of the Deduction 159

that Kant also had a solid argument that perceptual recognition (in a
concept) required the use of C-functions.
The second important change is that Kant finally makes up his mind
on the issue of one versus two sets of synthesizing functions. In a well-
known passage common to both editions, he had claimed:
The same function which gives unity to the contents of various cognitive
states in a judgment also gives unity to the mere synthesis of the contents
of various cognitive states in an intuition; and this unity, in its most general
expression, we entitle the pure concept of the understanding [A79/B105,
amended translation].

Since Kant never supplies any backing for this claim in the first edition,
it has the look of an argument by fiat and does not play a prominent
role in the deduction. Further, he explicitly rejects the claim in the
Prolegomena, where he distinguishes between judgments of experience
and judgments of perception.34 "The latter require no pure concept of
the understanding, but only the [psychological connection of percep-
tions in a thinking subject."35 By 1787, Kant has obviously rejected this
view. All combination is carried out by the understanding [which is
governed by the categories]. The question is: Why does he change his
mind?
As noted, what is totally new in the B Deduction is the discussion of
perceiving spatial and temporal arrays, and the point that the forms of
intuition do not, by themselves, supply any cognition of determinate
objects, including spaces and times. I believe that we can gain insight
into Kant's new position by backing up and considering why it would
be natural to hold that perception does not involve the use of concepts,
a fortiori, not the use of a priori0 concepts. In the A edition Kant had
been tempted by the idea that all representing or perceiving involves
the use of a concept, as we have seen. Consider an obvious counter-
example, looking at a crumpled newspaper. You can look at the paper
and scan its contours and yet have no concept for its shape. Since
perception is constructive, however, you must have some function or
functions that select and combine sensory data to produce a stable image.
Notice, however, that as you perceive the paper, as you scan its image,
you scan edges, curves, straight pieces, folds, and other local geometrical
features. This is, I believe, the crucial point. Concepts for the local
spatial features of the object can provide the needed synthetic functions
(even in the absence of any concept for the whole object), and these
appear to be always available to do so, regardless of what concepts you
have or have not acquired. For Kant has already argued that spatial
160 KANT'S TRANSCENDENTAL PSYCHOLOGY
concepts are not acquired, because space was never in the senses.36 If
you can perceive at all, then you can construct perceptual images of
whatever objects you confront by constructing their surface geometry
in Euclidean space, and seemingly concepts for governing these syn-
theses are always available to you.

P-Functions as Spatial and Temporal C-Functions

Kant makes the crucial identification—of the syntheses that permit con-
cept application with the syntheses that produce intuitions that contain
a unity of diverse elements—for the case of spatial and temporal con-
cepts in the note to B160: "although, as a matter of fact, it [the unity
of the intuitive contents of representations of spaces and times] presup-
poses a synthesis which does not belong to the senses, but through which
all concepts of space and time first become possible." That is, the
syntheses that permit us to perceive spatial arrays, for example, in the
minimal sense of being able to scan images of them, are the syntheses
that permit us to apply spatial concepts to the arrays.
Although this claim is reasonably clear, what about the apparent
vacillation about whether the unity that characterizes the contents of
intuitions is conceptual or not? I start with the sudden shift back to
sense in the last sentence: This unity really belongs not to concepts but
to space and time. In H 24, to which we are referred at the end of the
note, Kant states that intuitions produced by imagination belong to
sensibility, because all intuitions belong to sensibility (B152). The point
of this final sentence is merely to stress that, despite the use of C-
functions in constructing spatial and temporal arrays, what are con-
structed are perceptions or conscious intuitions.
How can Kant claim that the P-functions involved in constructing
images of spatial and temporal arrays both presuppose syntheses asso-
ciated with spatial and temporal concepts and yet precede all concepts?
Although this claim is perplexing, we can make some sense of it by
reflecting on the special status of spatial concepts. Whatever modifica-
tions experience might make in the process forms of intuitions, we do
not acquire spatial and temporal concepts from experience, because
space and time are never in the senses. Thus, I believe that he is claiming
that our perceiving of determinate spatial and temporal arrays precedes
all concepts, as concepts are normally understood. It precedes all ac-
quired concepts. Finally, Kant presumably attributes the unity of spatial
and temporal arrays to the senses in the Aesthetic, because he has not
 Cognitive Capacity at the Center of the Deduction 161
yet presented the constructive account of perception, without which the
question of the source of unity in perceptions does not arise.
How are we able to perceive spatial and temporal arrays? The forms
of intuition themselves are inadequate. To perceive a spatial array, for
example, we must perceive something arrayed in space. And when we
perceive something arrayed in space, we can construct the perception
of that object by employing syntheses associated with our a priori0
spatial [or temporal] concepts, concepts like 'straight', 'curved', 'above',
'to the left of, 'larger than', 'congruent with'. This simple task thus
requires three different factors: forms of intuition, objects to supply the
diverse elements in cognitive states to be united in perceptions, and
syntheses associated with a priori0 spatial concepts to do the uniting.

Perceiving Objects by Perceiving Spatial and Temporal Arrays

Despite its apparent simplicity, this task has immense theoretical im-
portance, in part, because of the reciprocity between perceiving spatial
and temporal arrays and perceiving objects arrayed in space and time.
We can only perceive spatial arrays, for example, by perceiving objects
arrayed in space, but conversely, whenever we perceive an object arrayed
in space, we also perceive a spatial array, and so can synthesize the
perception by employing a priori0 spatial concepts. I take it that this is
what Kant means when he says in the sentence right after the one to
which the note is appended:
Thus unity of the synthesis of the diverse elements, without [in space] or
within us [in time], and consequently also a combination to which every-
thing that is to be represented as determined in space or time must con-
form, is given a priori as the condition of the synthesis of all
apprehension—not indeed in, but with these intuitions [B160-61, amended
translation].
That is, the perception of a determinate object in space is given along
with the perception of a determinate spatial array.
We can perceive something without knowing what it is. So we need
not have a concept for everything that we perceive, but only for those
things that we recognize in a concept through perception. On the other
hand, everything that we perceive is a spatial and/or temporal array.
And, Kant argues, in perceiving spatial and temporal arrays, we use
syntheses that make application of spatiotemporal concepts possible. So
even in cases where we do not recognize the perceived item in a concept,
our perceptual synthesis is still guided by rules associated with spatial
162 KANT'S TRANSCENDENTAL PSYCHOLOGY
and temporal concepts. Further, since space and time are the forms of
intuition, everything that we perceive is subject to them. Thus, even in
cases where we recognize an item in a (nonspatial and nontemporal)
concept, the rules governing the synthesis of the perception are still
subject to rules associated with spatial and temporal concepts (B160).
It follows that all perceptual syntheses are guided by rules associated
with spatial and temporal, and perhaps further concepts. After this
intricate discussion and the illustration of its general point in the case
of perceiving a house, Kant can finally, [and presumably happily] con-
clude: "It is one and the same spontaneity, which, in one case, under
the title of imagination, and in the other case, under the title of un-
derstanding, brings combination in the diverse elements of intuition,"
(B161n, amended translation).

Kant's Long Argument

Additional Considerations
How does the analysis of perceiving spatial and temporal arrays provide
an argument for the categories? So far, Kant claims only that such
perceptions are constructed by syntheses that permit application of spa-
tial and temporal concepts. The categories are not mentioned. The
centrality of this ability depends on two further considerations. Since
space and time are the forms of intuition, all our perceptions are of
objects and events in spatial and temporal arrays. Second, in the Prin-
ciples, Kant will argue that applying spatial and temporal concepts re-
quires applying the categories, so that any perception that can be
subsumed under spatial and temporal concepts can also be subsumed
under at least one of the categories.37 Thus, whatever syntheses permit
application of spatiotemporal concepts also permit application of the
categories.

The Basic Argument

With these two additional considerations, Kant's analysis of perceiving
spaces and times becomes part of a long argument for the universal
applicability of the categories to everything that we can perceive. Since
this reasoning is both critical to an understanding of the structure of the
transcendental deduction and somewhat complex, I will display it:
 Cognitive Capacity at the Center of the Deduction 163
*1. Perception of spatial and temporal arrays requires that a C-function,
which would permit application of spatial and temporal concepts to per-
ceived objects and events, synthesize the diverse elements of intuitive
cognitive states into an image. [Kant's analysis of perception and B160n.]
2. Space and time are the forms of intuition, so all our perceptions are
of objects and events in spatial and temporal arrays. [Transcendental
Aesthetic]
3. In all our perceptions, we produce stable images by synthesizing data
according to C-functions that would permit application of spatial and
temporal concepts. [1 & 2]
4. We can apply spatial and temporal concepts to objects and events
only if we can also apply the categories to them. [The Principles]
5. Thus, whatever syntheses permit application of spatiotemporal con-
cepts also permit application of the categories. [4]
6. Thus, the categories apply to everything that we can perceive, in the
minimal sense of being able to scan an image. [3 & 5]
Although this argument is not set out formally in H 26, all its major
pieces are presented. (*1) occurs in the passage just cited from B160n;
Kant reminds us that space and time are the forms of intuition, and he
infers that the syntheses that govern spaces and times also apply to
everything that is represented as determined in space and time (B160,
B161) (2); and, most importantly, he appends discussions of the house
and freezing water examples that offer previews of the arguments about
quantity and causation that are to come in the Principles (B162) (fore-
shadowing 4). Finally, he draws the conclusion (6): "All synthesis, there-
fore, even that which renders perception possible, is subject to the
categories..." (B161).

Universal Applicability and Objective Validity

If sound, this argument would establish the universal applicability of
the categories. It would also show that a priori0 elements must be
involved in very basic cognitive tasks. This argument is a significant
improvement over Kant's previous attempts to argue for the special
status of the categories, both in the A edition and in earlier sections of
the B Deduction. The discussion in A suffered from two related prob-
lems. First, the doctrine of the three distinct syntheses suggested that
perception requires only sense and imagination; concepts enter only
when we reach the level of recognition and knowledge. Kant avoided
this implication by assuming that perception, or any cognition, involves
concepts or some intellectual component. So, presumably, perceiving a
house involved the concept 'house'. This view had prominent supporters,
164 KANT'S TRANSCENDENTAL PSYCHOLOGY

including Reid and Locke. As we have seen, however, it could not be

assumed as an uncontroversial premise. In H 26, Kant makes the much
more modest assumption that perception only involves scanning an im-
age. He then argues that even this requires an intellectual component,
but a much more general component that is common to all cases and
always available, syntheses associated with spatial and temporal con-
cepts. Second, the A Deduction claimed that we do not reach the level
of experience until all three syntheses, including the conceptual, have
organized the data of sense. In relying on this rich notion of cognitive
experience, Kant was in serious danger of trivializing his conclusion.
Again, the modest assumption involved in H 26 avoids this problem,
because it is far from obvious that the categories, or any rules of thought,
should be related to the construction of perceptual images.
The argument of H 26 also has the advantage of making a direct link
between the categories and what we perceive. A number of passages in
both editions try to link objects of intuition to the categories through
the necessity of apperception. (In fact, this move also appears at B161.)
The serious drawback of all such arguments is that they rest on the
unsupported contention that the categories are, in fact, required for the
unity of apperception, or that they are, in fact, the laws of thought or
are associated with the laws of thought. This is the great advance of
H 26, and so the "second step" in the argument for the categories. Prior
to this section, Kant had argued that all cognitive states, judgments and
intuitions, must be able to belong to the unity of apperception and that
the categories somehow made this possible. Hence, all judgments and
intuitions are made possible by the categories, and so are subject to
them. As Kant wisely observes in K 21, this can only be a beginning of
a deduction of the categories. With the argument sketched in 126, there
is some hope of actually demonstrating an important role for the cat-
egories in cognition.
This was another major goal of the deduction of the categories. The
categories were going to be shown to be objectively valid, by demon-
strating that they supply conditions that are necessary for the possibility
of any cognition of objects at all. Kant extends the argument of f 26 to
include a demonstration of objective validity. After noting that the
categories must be involved in perception, he continues: "and since
experience is knowledge by means of connected perceptions, the cate-
gories are conditions of the possibility of experience, and are therefore
valid a priori for all objects of experience" (B161). In A the objective
Deduction had shown that knowledge of objects is possible only if we
have rules of synthesis that enable us to construct representations from
 Cognitive Capacity at the Center of the Deduction 165
the cognitive states that objects produce in us. Here the point is put
more perspicuously, in terms of perception. Knowledge requires the
construction of perceptions. Whatever sensory data we take in, we know
that, thanks to the forms of intuition, those data will be converted into
intuitions involving temporal and spatial properties. By the argument
just given, however, the syntheses that enable us to construct perceptions
of spatiotemporal arrays are also the syntheses that allow us to apply
the categories. Thus, so long as we are able to apply the categories, we
will be able to construct perceptions on the basis of whatever sensory
data we take in.
To see the force of the claim for objective validity, we need to recall
that Kant believes that the rules of synthesis for constructing either
representations or perceptions must involve a priori0 elements. The data
of sense are too numerous and too jumbled to supply any useful prin-
ciples themselves. Given this assumption (which as I argued in Chapter
3 and above is inadequately supported), Kant's argument for objective
validity can be represented as a continuation of the previous chain of
reasoning:
7. Knowledge of objects is possible only if we can find some rule of
synthesis to unite cognitive states in a representation or perception. [Ob-
jective Deduction in A]
8. All our intuitions involve temporal and spatial properties. [Tran-
scendental Aesthetic]
9. The syntheses that enable us to apply the categories enable us to
construct perceptions of spatial and temporal arrays. [1 & 5]
10. Thus, so long as we can apply the categories, we can construct
perceptions, whatever the data of intuition. [8 & 9]
11. Since the rules needed for the synthesis of representations or per-
ceptions cannot be supplied by sense, and we have no other a priori0
candidates, the syntheses associated with the categories make perception
possible. Hence, they make knowledge possible, and are objectively val-
id. [10]
While Kant does not return to the issue of judgment at this point, it
is easy to see how to construct a parallel continuation that begins with
the result of the objective Deduction in B. Knowledge requires judg-
ments, and judgments require that cognitive states be synthesized in
representations (which will themselves by synthesized in judgments). So
we could continue the displayed argument:
12. For knowledge to be possible, cognitive states must be united by
syntheses in representations that will themselves be synthesized in judg-
ments. [Objective Deduction in B]
166 KANT'S TRANSCENDENTAL PSYCHOLOGY

13. Regardless of what other information they contain, all intuitions

include spatial and temporal properties, and so can be synthesized in
perceptions by functions that are associated with the categories. [6 above]
14. Thus, so long as we can apply the categories, this requirement for
knowledge will be met. [12 & 13]
15. Since the rules needed for the synthesis of representations cannot
be supplied by sense, and we have no other a priori0 candidates, the
syntheses associated with the categories supply a necessary precondition
for judgment. Further, since the syntheses that permit the construction
of judgments from representations are associated with the categories, the
categories supply a second condition that is necessary for judgment. And
since judgment is necessary for knowledge, the categories make knowl-
edge possible and are objectively valid. [14]

As already noted,38 the assumption about the connection between the

syntheses involved in judgments and the categories is inadequately de-
fended, as is the claim that the rules of synthesis must be a priori0.
Nevertheless, if these points are granted for a moment, then we can see
why Kant believes that the categories make cognition possible. They
make perception possible and they make judgment possible. Further,
the line of reasoning just presented applies to all knowers, constituted
as we are and having space and time as the form of their sensibility.
Hence, the categories would also be objectively valid in the sense that
they are uniform across subjects. And the intersubjective validity of the
categories would not be a contingent matter, like the (more or less)
uniform perception of secondary qualities. Rather, it would be estab-
lished on the basis of a transcendental psychology of cognition itself.

The Argument from Apperception

Arguments that try to establish the objective validity of the categories
by showing how they make knowledge possible beg the question against
the skeptic, since they presume that knowledge is possible. To the skep-
tic, Kant claims, again and again, that the categories are required for
the possibility of apperception. Although he does not return to the theme
of apperception at the end of the B deduction, it is fairly clear how the
argument sketched at f 26 can be extended to demonstrate that the
categories supply a necessary condition for apperception. The subjective
Deduction has shown that cognitive state can belong to the unity of self-
consciousness only if they are connected or connectible by synthesis with
each other. These states include thoughts and perceptions, and for rea-
sons noted earlier, the potentially problematic cases are perceptions.
 Cognitive Capacity at the Center of the Deduction 167

To show the relation between the categories and the possibility of ap-
perception, the argument may be continued as follows:
16. Apperception is possible only if all my perceptions are connected
or connectible by synthesis with each other. [Subjective Deduction]
17. Regardless of what other information they contain, the contents of
all my perceptions can be subsumed under the categorial concepts and
spatial or temporal concepts. [6]
18. Once cognitive states have been united in representations (percep-
tions or concepts), they can be united in judgments. [Analysis of
Judgment]
19. However, judgments can be combined in further judgments by the
logical connectives. [Once we have judgmental representations, they are
subject to the laws of logic.39]
20. Hence, a thoroughgoing unity of apperception is possible, all my
perceptions are connectible by synthesis with each other, and all my other
cognitive states, because they are synthetic progenitors of conceptual
states that can be directly synthesized with each other. [Account of
Synthesis40]
21. Through the categories, all perceptions and judgments can belong
to a thoroughgoing unity of apperception. Since the rules needed for the
synthesis of cognitive states that is necessary for apperception cannot be
supplied by sense, and there are no other a priori0 candidates, the
syntheses associated with the categories supply a necessary condition for
the possibility of apperception. [16 & 20]
But for the categories, mental life would not be unified; rather, "ap-
pearances might crowd in upon the soul... and would be for us as good
as nothing (Alll); much might arise in empirical consciousness . . . in a
state of separation, and without belonging to a consciousness of myself
. . . " (A122). Hence, the categories would be necessary for the unity
of apperception, which is, in turn, necessary for us even to have cognitive
states that represent something to a subject. So even the skeptic must
concede the indispensability of the categories.

How the Argument Fails

This is, I believe, Kant's best argument for the various claims that
he seeks to establish about the categories. They are universally ap-
plicable and objectively valid; they make apperception possible. It
achieves startling results from minimal premises. If sound, it would
show that we cannot even have perceptual images unless we can, for
example, apply the categories of cause and effect to the objects and
events we perceive! This result is so startling that we might wonder
168 KANT'S TRANSCENDENTAL PSYCHOLOGY

whether Kant could have believed that he had an argument to sup-

port it. However, the argument derives its force from two principal
sources: the analyses of cognitive tasks, including the doctrine of the
forms of intuition, and the arguments of the Principles, in particular
the careful reasoning of the Analogies. As we have seen in Chapters
2-5, and will see later, these are surprisingly strong anchors. This
long argument unites key aspects of Kant's position in a single argu-
ment and places the burden of establishing the categories on its
strongest parts. It also rests on the best available scientific or protos-
cientific discussions of its central topics.
For all that, the argument fails. Even if the arguments in the Principles
were completely successful, the central link given in If 26 is too weak
for Kant's purposes. That link is the claim that in perceiving spatial and
temporal arrays, the syntheses that enable us to construct perceptions
also permit the application of spatial and temporal concepts, that is, the
P-functions are the C-functions. Although there are strong reasons to
believe that this is true for cases where perception leads to conception,
those are not the relevant cases. Kant is trying to rule out the possibility
that we might have perceptions that could not be conceptualized. For
all he has shown, P-functions that permit concept application could be
but a fraction of the functions that subserve perception.41
There are two problems with the argument. Even if space and
time are not in the senses, so that our spatial and temporal concepts
are not acquired in any straightforward sense, it does not follow that
such concepts must be available to us in advance of any [other] con-
cepts. Further, even when we clearly have spatial concepts, for ex-
ample, and can recognize instances through the use of C-functions,
it still does not follow that whenever we perceive a spatial array, we
connect its diverse elements via the same rule of synthesis that we
use when we apply a concept to the array. Kant's general project
was to reveal the preconditions of knowledge, so he takes percep-
tions culminating in recognition in a concept as paradigmatic, rather
than, for example, perceptions guiding movement. Further, his anal-
ysis of perception shows that we must construct perceptual images
by using some nonarbitrary functions of synthesis. Further still, his
analysis of the necessity of coordination or subordination between
the faculties in cases of applying concepts shows that there must be
C-functions that can construct perceptual images. Finally, the view
that perception involves concepts, or some intellectual element, was
widely held. Under these circumstances, Kant's analysis of perceiv-
ing times and spaces would seem extremely plausible. Nevertheless,
 Cognitive Capacity at the Center of the Deduction 169

for his purposes, he has picked up the wrong end of the stick. Even
in 11 26, he errs in regarding perceptions that lead to conception as
central cases. When he generalizes to all perception (*1), he ends
up begging the question he wants to address: Could there be per-
ceptions that cannot be synthesized with other cognitive states?
Two hundred years after the B Deduction, the verdict is still out on
the relation between perception and cognition. Since we apply concepts
to objects on the basis of perception, perception must be coordinated
with or subordinate to conception in some instances. Further, perception
is at an entirely different, and apparently more cognitive level, than
mere sensation. These facts have led many—Thomas Reid, Hermann
Helmholtz, Jerome Bruner, Richard Gregory, and contemporary com-
puter modelers,42 among others—to believe that concepts, or even belief
and inference, are involved in constructing perceptions. Kant may well
be right that the same processes underlie perception and concept ap-
plication. He has no argument that this must be so, however. He has
not shown that this is a necessary feature of any adequate theory of
perception.

Defending the Long Argument Interpretation

Some Advantages
On my interpretation, Kant's best argument for the categories has sub-
jective and objective elements; it is both regressive or analytic (because
it presupposes some cognitive capacities) and progressive or synthetic
(because it argues from the presupposition of fairly minimal capacities
to very strong claims about the categories); and it employs a rich and
suitably weak notion of [cognitive] experience at different points, and
so does not rest on an equivocation on the term "experience." Although
significant improvements are made in the second version, much of the
analysis that justifies Kant's position is provided only in A. As promised,
it is also a fairly clean line of argument. Many themes criss-cross and
interrelate, but they all revolve around the central issue of the prereq-
uisites for the various cognitive tasks that make up experience. For
obvious reasons, I cannot show that an equally coherent argument can-
not be constructed if the transcendental psychology is omitted. I will try
to illustrate the virtues of using this material, however, by contrasting
my interpretation with two influential accounts that also focus on 1 26,
those of Dieter Henrich and Henry Allison.
170 KANT'S TRANSCENDENTAL PSYCHOLOGY

Henrich's Antipsychological Reading

I have already presented Henrich's textual argument for the two-steps-
in-one-proof constraint on interpretations of the deduction.43 Plausible
readings must explain how H 26 is supposed to complete the proof of
the categories that is only partially completed by H 20. His own proposal
is straightforward: By 11 20 Kant has shown that intuitions, insofar as
they are given in a unified intuition [In Finer Anschauung], are subject
to logical functions of judgment (i.e., the categories). At f 26 he intro-
duces the factual claim that we have unified representations of space
and time. When this fact is combined with the conclusion established
in the Aesthetic, that our representations of space and time "include
everything that can be present to our senses," it follows that "every
given manifold without exception is subject to the categories."44
The most obvious objection to Henrich's reading of f 26 is that it
gives no explanation of the fact that, in apposition to the main discussion,
Kant offers previews of the arguments about homogeneous parts and
about causation that are to come in the Principles. Why does Kant feel
that it is necessary or appropriate to foreshadow those discussion here,
unless the argument of this section is to be completed in the Principles?
Further, five pages later, Kant says explicitly that the case for the cat-
egories will be completed in the Principles:
How they [the categories] make experience possible, and what are the
principles of the possibility of experience that they supply in their appli-
cation to appearances, will be shown more fully in the following chapter
on the transcendental employment of the faculty of judgment [B167].
Further, Henrich does not deal adequately with Kant's specific de-
scription in H 21 of what he is going to show in f 26:
In what follows (cf. f 26) it will be shown, from the way [aus der Art] in
which the empirical intuition is given in sensibility, that its unity is no
other than that which the category (according to 1 20) prescribes to the
diverse elements of a given intuition in general [B144-45, amended
translation].
Here Kant seems to be echoing his summary of the first edition De-
duction chapter:
This is all that we were called upon to establish in the transcendental
deduction of the categories, namely, to render comprehensible this re-
lation of understanding to sensibility, and, by means of sensibility, to all
objects of experience [A128].
 Cognitive Capacity at the Center of the Deduction 171

Henrich acknowledges that, besides establishing the categories, the de-

duction is supposed to clarify the relation of understanding to sensibility.
His proposal is that Kant carries out this mission by first showing that
the categories govern all unification by the understanding and then not-
ing that space and time are unified intuitions.
[In this way] we find in the second edition . . . a proof of the validity of
the categories which is at one and the same time an explanation of the
possibility of their relation to sensibility, a proof which avoids taking up
the problems of an analysis of the cognitive faculties [my emphasis].45
That is, Henrich tries to deal with Kant's interest in relating understand-
ing to sensibility apsychologically. In U 24, however (to which the reader
is referred in II 26), it is clear that Kant proposes to show how the
understanding relates to sensibility, via a third faculty, the imagination
(B151-52).
Henrich's antipsychological reading also directs attention away from
the synthesis of apprehension, which Kant highlights. On my interpre-
tation, the doctrine of the synthesis of apprehension embodies Kant's
recognition that perceptions must be constructed, and hence can be
described as unified. Without this material, what is the basis of the crucial
claim that our intuitions of space and time are unified? In the Aesthetic,
in the third metaphysical exposition of space (in B) and the fourth of
time, Kant claims that we represent to ourselves one space and one time
(B39, A31-32/B47). However, his point in those passages is that par-
ticular spaces or times are regarded as parts of one space and one time.
The Aesthetic provides no support for the claim that our perceptions
of determinate spatial and temporal arrays are unified, but that is the
topic of 1 26.

Allison's Apsychological Reading

I turn to Allison's careful study of the same passage.46 He takes the
second step of the deduction, which culminates at 11 26, to involve two
projects: Kant needs to show how the categories relate to actual em-
pirical intuitions and how they make experience possible. The first proj-
ect succeeds, according to Allison, because the synthesis of imagination
that governs the synthesis of apprehension is in turn governed by the
conditions required by the unity of apperception, and those conditions
are the categories.47 As already noted, the problem with any account
that links the objects of perception to the categories through appercep-
tion as an intermediary is that it must simply take Kant's word that the
172 KANT'S TRANSCENDENTAL PSYCHOLOGY
categories are required for the unity of apperception. However, the real
contrast I wish to draw with Allison's account centers on the second
project.
Allison reluctantly concedes that Kant has not shown that [a fortiori,
how] the categories make experience possible. I believe that even Kant's
best argument contains mistakes and so does not establish a special
status for the categories. However, on my interpretation, it is clear why
Kant thought that the categories supplied necessary conditions for var-
ious aspects of cognitive experience, and hence why he thought that he
could carry out a deduction that established their objective validity.
Allison regards the first part of the Deduction as an analytic argument.
It considers what is implicitly contained in the concept of 'objective
experience' and in the concept of a 'subject of knowledge'. The second
part of the argument is synthetic and tries to show that the categories
make experience possible.48 If the first part of the Deduction is merely
an analytic argument, however, then it could never make a contribution
to the second, synthetic step, as Kant would be the first to point out.
Further, although Allison considers the examples of perceiving a house
and the freezing of water, he believes that Kant's discussions are com-
pletely muddled. The categories are supposed to be necessary conditions
for distinguishing mere subjective perceptions from the objective order
of things and events; they are "objectifying conditions." In these pas-
sages, however, Kant seems to be saying that the categories are needed
just to perceive a house or the freezing of water.49
In fairly sharp contrast to Allison's account, I take the entire De-
duction to be a synthetic argument. Further, I take it—and not just the
Principles—to offer arguments about how the categories make cognitive
experience possible. Kant assumes that we can perform various cognitive
tasks and then argues that these tasks require rule-governed syntheses.
In this way, he is constantly trying to show how various aspects of
cognitive experience may require special rules of synthesis that are con-
nected with the categories. Paragraph 26 is a major step forward, be-
cause it shows that we need syntheses even to perceive spatial and
temporal arrays, a point that is clearly illustrated in the examples. Pre-
sumably, that is why he concludes the section by returning to the claim
that the categories make [cognitive] experience possible. Further aspects
of cognitive experience, including the ability to perceive an objective
time order, are explored in the Principles. On my reading, however, it
is clear how the Deduction chapter itself is supposed to advance the
project of showing that, and how, the categories make cognitive ex-
perience possible.
 Cognitive Capacity at the Center of the Deduction 173

The Loss of Generality

I conclude my defense of this interpretation by considering an obvious
objection. Even if sound, the long argument could not establish Kant's
general thesis that the categories are valid for every understanding,
regardless of mode of sensibility (A88/B121, B144). This point is correct,
for the argument depends on assuming that space and time are the forms
of intuition. Still, I believe that it is his best argument. As noted in
Chapter 3, individual categories are not discussed in the Deduction
chapter, but only in the Metaphysical Deduction and in the Principles.
So any argument for the 12 specific categorial concepts must draw from
one of these sections. Kant does not seem to have any reasonable way
of linking the forms of judgment to the categories, however. On the
other hand, the arguments of the Principles make explicit and essential
use of the spatiotemporal character of human experience. So any inter-
pretation that does not wish to rely on the Metaphysical Deduction must
forfeit generality, so this objection does not raise a special problem
for mine.

11 26 as Completing the Argument of the

Metaphysical Deduction
I should add that although I believe that the long argument is the best
Kant has, there are obviously many elements in the Deduction and in
f 26 that look back to the Metaphysical Deduction—most notably its
opening statement. This passage is the juncture of two very different
arguments. As I have shown, the analysis of perceiving spatial and
temporal arrays is the centerpiece of a long argument that anticipates
the results of the Principles. However, it also provides an essential
premise for an argument that draws on the Metaphysical Deduction. If
Kant assumes the results of the Metaphysical Deduction, then he has a
very compact argument for the universal applicability of the categories:
1. Since space and time are the forms of intuition, all our perceptions
are of objects and events in spatial and temporal arrays. [Transcendental
Aesthetic]
2. We can perceive spatial and temporal arrays only by synthesizing
[hence unifying] images from sensory data, [f 26 and analysis of task of
perception]
3. All synthesis is carried out by the understanding. [Opening of B
Deduction]
174 KANT'S TRANSCENDENTAL PSYCHOLOGY
4. The functions by which the understanding combines information are
associated with the categories. [Metaphysical Deduction]
5. Therefore, one of the categories can be applied to the contents of
all our perceptions and the categories make experience possible.
In defending the long argument, I do not mean to suggest that it is the
only argument that Kant offers, or even that it is the only argument
that he offers at 11 26.

How Serious Is the Loss of Generality?

Finally, how important is the loss of generality in the long argument?
As Walsh notes, "[Kant]... retains the idea of the pure category for
theoretical purposes . . . , but it plays no real part in his account of human
cognition."50 Paul Guyer rightly observes that the category for Kant is
causation. His earliest theory of experience consisted of a sketch of the
Analogies of Experience, and this is the most carefully argued section
of the Critique.51 Further, as DeVleeschauwer has pointed out, Kant
was committed to the categories, prior to the discovery of the Clue.52
If the arguments of the Principles are sound, particularly if the Analogies
succeed in establishing the necessity of the concept of cause for human
spatiotemporal experience, then Kant would have no reason to mourn
the loss of a more general conclusion.
I will conclude my discussion of the deduction of the categories and
my defense of the long argument interpretation with a brief account of
the Second Analogy. Although it has not been approached this way, I
will argue that the reasoning in this section is an obvious—and outstand-
ing—example of transcendental psychology. By presenting Kant's
closely reasoned analysis of the prerequisites for assigning temporal
position to objects and events, I will also support my claim that the long
argument is the best he has.

Transcendental Psychology in the Second Analogy

Guyer's Interpretation
The Second Analogy has been, dissected and reconstructed more than
any other section of the Critique. On the basis of a comprehensive review
of the secondary literature and of Kant's early writings, Paul Guyer has
offered an interpretation of the basic structure of the argument that
will, I believe, prove to be definitive. Guyer is determined to shield this
 Cognitive Capacity at the Center of the Deduction 175

reasoning from any taint of psychology, however. Contrary to his in-

tentions, I will use his interpretation to present the task analysis that is
at the center of the Second Analogy. Then I will explain why Guyer's
belief that a psychological approach must end in circularity is unfounded.
Hume demonstrated that the necessity involved in the concept of
causation cannot be based on experience. The Second Analogy is to
demonstrate that the causal concept is nevertheless objectively valid. It
is objectively valid, because it makes knowledge of objects possible.
Among other conditions, knowledge of objects is possible only if we
are able to unite the elements of various cognitive states in perceptions.
The argument of f 26 showed that we can form perceptions of all spatial
and temporal arrays, and hence of all objects and events arrayed in
space and time, by using syntheses that enable us to apply spatial and
temporal concepts. The project of the Analogies is to show that we can
apply temporal concepts to objects—we can assign them determinate
places in time—only if we can apply the concepts of substance, cause,
and community to them. Hence, the syntheses that enable us to apply
temporal concepts are the same syntheses that enable us to apply certain
categories. Since these syntheses are allegedly necessary conditions for
the possibility of cognition, so are the categories. Although I accept
Guyer's view that a fully adequate treatment needs to consider all the
Analogies, I will only deal with the argument of the Second Analogy,
since that is sufficient to make my point.
This section focuses on the cognitive task of determining temporal
position. To recognize any alteration in a substance, we must recognize
an objective succession, one condition of a substance succeeding an-
other. Guyer delineates the basic structure of Kant's reasoning.53 The
starting assumption is that time itself cannot be perceived. We cannot
perceive empty time, but only states of affairs in temporal arrays (cf.
B138). Even in the latter case, however, we cannot directly see when
a state of affairs obtained. As Guyer notes, our perceptions of scenes
do not come replete with a time clock in the corner to inform us of the
time of the action. The problem arises because our perceptions are
successive—all of them. Whether we are perceiving the diverse, but
coexisting parts, of a house or the movement of a ship downstream, our
perceptions are successive. For this reason, we could not consciously or
unconsciously appeal to succession in our perceptions to determine suc-
cession in the object. So how can we recognize that one condition of
an object occurred after another?
Kant reasons as follows (using "A*" and "B*" to indicate our per-
ceptions, and plain letters for the objects of the perceptions, rather than
176 KANT'S TRANSCENDENTAL PSYCHOLOGY

Kant's ambiguous notation): "in a happening .. . B can be apprehended

only after A; the perception A* cannot follow upon B* but only precede
it" (A192/B237). But how could we ever know that B* has to follow
A*? Nothing in the individual cognitive state qua cognitive state can
provide that information. Hence, we can only recognize objective suc-
cession—and that B* had to follow A*—if something in the contents of
A* and B* informs us that B had to follow A. What could that be except
our recognition that A is a particular state of affairs and that states of
affairs of this type are invariably followed by states of affairs of the type
exemplified by B? That is, we can recognize that B* had to follow A*,
if and only if we have a rule stating that B's have to follow A's. This is
a very important point and Kant makes it twice in three pages:
In our case I must therefore derive the subjective succession of appre-
hension from the objective succession of appearances, since the former is
otherwise entirely undetermined and distinguishes no appearance from
any other. The former [succession] alone proves nothing about the con-
nection in the manifold, because it is entirely arbitrary. The latter [suc-
cession] will therefore consist in the order of the manifold of appearance
according to which the apprehension of the one [state] (that which hap-
pens) follows on that of the other (which precedes), according to a rule.
Only by that means can I be justified in saying of the appearance itself,
and not merely of my apprehension, that in it a succession is to be found,
which means as much as that I cannot order the apprehension otherwise
than in this very succession [A193/B238, Guyer's translation].
It is therefore only in respect of a rule according to which appearances
[objects] in their succession, that is, as they occur, are determined by the
preceding state that I make my subjective synthesis (of apprehension)
objective, and it is only under this presupposition that the experience itself
of something that happens is possible [A195/B240, Guyer's translation
and emphasis].

The celebrated reply to Hume is that if we did not apply a concept

of 'cause' (that includes necessary succession) to states of affairs, then
we would not be able to determine temporal relations.54 This is Guyer's
account of the basic logic of Kant's argument. So far, nothing precludes
interpreting the Second Analogy as a task analysis in transcendental
psychology, as I have done. Further, this reasoning can be added to the
arguments considered earlier to show the universality of the causal prin-
ciple. We can apply temporal concepts to objects if and only if we can
also apply the concepts of 'cause' and 'effect' to them. Thus, causal
reasoning applies to all states of affairs that we can recognize to stand
 Cognitive Capacity at the Center of the Deduction 177

in temporal relations. And, by the long argument, that includes all the
objects and events that we perceive.

Guyer's Objection to a Psychological Reading

It is the next step in Guyer's analysis that allegedly rules out a psycho-
logical reading. We cannot recognize temporal relations unless we can
appeal to causal rules. But where do we get particular causal rules?
Guyer believes that Kant must assume that we establish them induc-
tively, just as Hume claims.55 I agree. He then infers that under these
circumstances we cannot regard Kant's "theory of judgment—or time-
determination . .. [on] a psychological model of the generation of beliefs
but [must regard it] as an epistemological model of the confirmation of
beliefs."56 On a psychological reading, Kant would allegedly be caught
in a vicious circle. We cannot recognize temporal relations unless we
appeal to causal rules, but how could we establish causal rules unless
we can recognize temporal relations? To avoid this difficulty, Guyer
maintains that Kant's point is really about justification. Only causal
reasoning can justify our temporal claims.

Versus Guyer's Antipsychologism

I have three objections to Guyer's antipsychologism. First, there are
numerous passages where it is clear that Kant's point is that we could
not make temporal assignments in the absence of a causal interpretation,
and not merely that we could not justify such assignments—for example
(B234), "In other words, the objective relation of appearances that follow
upon one another is not to be determined through mere perception."
In fact, the first part of the text Guyer cites in support of the justificatory
interpretation makes this point:
In our case I must therefore derive the subjective succession of appre-
hension from the objective succession of appearances, since the former is
otherwise entirely undetermined and distinguishes no appearance from
any other. The former [succession] alone proves nothing about the con-
nection in the manifold, because it is entirely arbitrary [A193/B238, Guy-
er's translation, my underscoring].
Second, Guyer's interpretation gives Kant a weak reply to Hume.
Hume would have an easy counter: claims about temporal order, like
most of our putative knowledge claims, are not justified. Guyer tries to
escape this difficulty by suggesting that the argument of the Second
178 KANT'S TRANSCENDENTAL PSYCHOLOGY

Analogy is not completed until the Refutation of Idealism. If the Refu-

tation shows that we cannot be justified in assigning temporal location
to our own cognitive states unless we are justified in assigning temporal
position to "outer" events, then he thinks the skeptic is beaten. I don't
see why. Why shouldn't Hume simply admit that we are not justified in
this either? Further, the Refutation of Idealism was not added until the
second edition and there is no reference to it in the B version of the
Second Analogy. Hence, it seems somewhat far-fetched to think that
Kant saw it as completing this very important argument.57
Finally, and for my purposes, most importantly, Guyer's rejection of
a psychological interpretation depends entirely on his equation of tran-
scendental psychology with developmental psychology. As I argued in
Chapter 1, Kant was extremely interested in the question of origins.
Transcendental psychology is an attempt to determine the necessary
conditions for cognitive tasks—especially any conditions that require a
priori0 elements. And, of course, 'cause' is the a priori0 concept. In-
terestingly, Kant does not argue that 'cause' cannot be extracted from
sensory experience in the Second Analogy. That result was established
independently by Hume and is presupposed. Rather, Kant examines
the task of assigning position in time and argues that we can only make
such judgments if we can make judgments of causal relations. Presum-
ably, he would also claim the converse dependency. For he is trying to
show that the very syntheses that enable us to apply temporal concepts
also enable us to apply causal concepts. That is why the concept of cause
is objectively valid—because allegedly these syntheses are required for
any cognition (perceptual or conceptual) of objects.
There is no circularity in this analysis, however. Kant is interested in
origins, not development. The claim is not that we must employ causal
concepts before we can order states of affairs in time but that we can
only do one by doing the other, and that 'cause' cannot be gotten out
of the senses. Although this analysis would put constraints on devel-
opmental psychology, it is not itself developmental. From the point of
view of transcendental psychology, how exactly we develop the capac-
ities and stocks of rules required for cognition is not an interesting
question.

What Kant Has Shown

In the Second Analogy, Kant discovers a truly puzzling fact about human
cognition.58 States of affairs do not wear temporal locations on their
sleeves and all our perceptions are successive, so how do we tell when
 Cognitive Capacity at the Center of the Deduction 179

we are observing succession in the world? Kant's transcendental psy-

chology is sound in laying out the problem, but is he right in the claim
he most wants to establish, that the recognition of temporal position
depends on interpreting the world causally? Here, as in the case of the
Euclidean nature of spatial perception, his positive suggestion is less
compelling than his basic account of the task. Although we must have
some way of making temporal determinations, "the" way may be an
aggregate of many ways.
There is an obvious logic to Kant's suggestion. We need a mark or
symptom of succession in time, and since effects follow their causes,
establishing causal relations would do the job. However, there appear
to be other low-level and high-level ways to accomplish the same task.
At the lower level, the discovery of motion detectors suggests that we
need not appeal to the contents of states at all. Given a particular
arrangement of neurons, it is possible for a slightly higher-level neuron,
M, to fire if and only if a stimulus registers sequentially on an array
of adjacent receptor level neurons, n^,n2,.. . ,n(. Very roughly, some
neural assemblies in the visual system operate in accordance with the
rule "M iff «2 a given interval after n,, n3 the same interval after n2, and
so forth."59 With such assemblies we could tell, for example, that a light
moved from left to right in our visual field, and so that it occupied one
position after another.
We also appear to make temporal judgments by tacitly appealing to
ordering principles that involve content but not causal relations. In a
fascinating series of experiments, Fodor, Bever, and Garrett showed
that grammatical considerations can affect temporal judgments. Subjects
listened to tapes of sentences that were interrupted by "clicks" (short
bursts of noise) at various locations in the sentence. For our purposes,
two interesting results emerged. Subjects are surprisingly inaccurate
about when the click occurred in the sentence. Consider the sentence
"That he was happy" was evident from the way he smiled." According
to the grammatical theory Fodor et al. were trying to test, the caret is
located at the major constituent break in this sentence. Now if the click
is placed right here, subjects will locate it correctly. For other objective
placements, subjects still tend to locate it at this boundary.60 Although
grammatical structure does not yield accurate temporal information (at
least in this experimental design), it does affect temporal judgment.
Interestingly, although the "click" experiment suggests that Kant's
claims for the ubiquity of causal interpretation are too strong, it gives
him a reply to a standard and potentially devastating objection to his
whole line of reasoning in the Second Analogy. Arthur Melnick offers
180 KANT'S TRANSCENDENTAL PSYCHOLOGY
a recent version. He suggests that we can be directly aware of succession,
because two states of affairs could succeed each other in the "specious"
present.61 Kant wrote over 100 years before the specious present became
part of psychological landscape.62 Thus, it might appear that his argu-
ment rests on ignorance of this phenomenon. Some of the force of this
criticism can be blunted by recalling that the specious present is so named
because it is specious. As I noted earlier, we can recognize melodies by
constructing a single representation of the relations among the individual
notes. However, we cannot literally hear the familiar E-D*-E-D*-E-
B-D-C-A opening of "Fur Elise" all together in one moment, even
though it seems as if we do. Introspection—as Kant was well aware—
is not a reliable guide to psychology. The click experiment provides
useful support for this reply. We hear sentences in the specious present,
but that does not mean that even in these brief time intervals, we operate
as a tape recorder. We do not simply replicate the data of sense. On
the contrary, what this experiment shows is that we construct our per-
ception of the sentence and the click.
The surprising claim of the Second Analogy—that we assign temporal
position to states of affairs only by interpreting them causally—is almost
certainly false. Still, Kant's careful reasoning has uncovered a major
research area: What are the means by which we recognize temporal
order? Given their limited range of operation, motion detectors cannot
be the whole story. The click experiment suggests that if we ever do
find answers that would enable us to complete the argument of the
Second Analogy correctly, they may be as surprising as Kant's own
premature solution. Finally, although Kant badly overrated the gener-
ality of his solution, it will undoubtedly be among the answers to his
problem.
 7
The Limits of Transcendental
Psychology

Kant's Paralogisms

Like the Critique as a whole, transcendental psychology offers both

positive and negative doctrines. To this point, I have considered only
the positive results of Kant's explorations of cognition. He also drew
important negative conclusions about what philosophy could not tell us
about the mind in the "Paralogisms of Pure Reason."
The Paralogisms chapter is standardly regarded as an extremely suc-
cessful critique of Kant's predecessors' pretensions to pneumatology.
Robert J. Richards's summation is typical:
[Kant's] incisive exposition of the paralogisms of rational psychology—
the uncritical, a priori deductions asserting the ego to be a substance, a
simple entity, an enduring personality, and related in specific ways to the
external world—slashed the roots of Wolffian rational psychology and
withered its derivative claims about the soul's immateriality, spirituality,
and immortality.'
Although this account is not incorrect, it is importantly incomplete.
There can be no question that the criticisms of this chapter are directed
against Rational Psychology. Kant says so, even though he does not
name names, or not enough names. Only Descartes is mentioned ex-
plicitly, but, as Margaret Wilson has argued, Leibniz was undoubtedly
another target.2 So was Christian Wolff. The discipline of Rational Psy-
chology that was formally introduced in Wolff's Psychologia rationalis
appears to be the focus of Kant's objections. As noted in Chapter 1,
Rational Psychology begins with propositions from Empirical Psychol-
181
182 KANT'S TRANSCENDENTAL PSYCHOLOGY

ogy. Through analysis and demonstration, it then deduces the essential

properties of the soul, from which its more superficial properties flow.3
One of the constant themes of the Paralogisms is that no amount of
analysis or demonstration can establish the type of synthetic doctrines
that Rational Psychology puts forward (e.g., the soul is a simple sub-
stance). In the simpler and more compact 1787 version, virtually the
entire critique is compressed into this one point: If we are dealing with
identical propositions, then no substantive or synthetic claims about the
soul can be extracted from them.
Criticizing the mistakes of others is only one purpose of the Paral-
ogisms, however. As Wilfrid Sellars, among others, has observed, the
I that thinks that is the topic of this chapter is the Deduction's I of
apperception (see, e.g., A400).4 Although this identification is often
acknowledged, the chapter has not been read in light of the transcen-
dental psychology of the Deduction.5 Rather, it is usually offered as
evidence of a sensible antipsychological approach to philosophy. This
is partly because Kant contrasts his "logical," "formal" account of ap-
perception with the claims of Rational Psychology (e.g., A350). As we
have seen, however, these labels do not mean what twentieth-century
readers take them to mean. Transcendental logic is no branch of formal
logic, as we know it, but a study of the sources and prerequisites of
cognition.
In this chapter I argue that we can only make sense of the details of
Kant's criticisms if we see the Paralogisms as a continuation of the
transcendental psychology of the subjective Deduction. The earlier text
offered the positive case for a unity of apperception; here Kant strives
to clarify what the doctrine of apperception does not say, by explaining
the inherent limitations of the analyses that transcendental philosophy
can provide. He undercuts the discipline of Rational Psychology, partly
by exposing its own fallacious reasoning and partly by reflecting on the
sorts of conclusions about the self that philosophy is and is not capable
of establishing.
This is why the Paralogisms of Pure Reason has more than historical
interest. It is not just an elegant methodological critique of a now long-
forgotten discipline. By examining the implications of his own work on
apperception, Kant discovered that there are important limitations to
any purely philosophical study of the mind. These limitations apply
equally to current efforts. I argue later that the claims of major con-
temporary philosophers (Thomas Nagel, Zeno Vendler, and John Searle
will be my examples) involve the very confusions and overstatements
that Kant warned against. Thus, I will argue that negative teachings of
 The Limits of Transcendental Psychology 183
transcendental psychology still have an important role to play in curbing
illegitimate philosophical speculation about the mind.
The Paralogisms chapter is the only section of the Dialectic that was
rewritten for the second edition of the Critique. I believe that Kant
abbreviated his presentation for two reasons. As I indicate later, the B
Deduction incorporated many negative points about apperception that
had been dealt with in the first version of the Paralogisms. With this
material already covered, Kant could present a much simpler line of
criticism: The Rational Psychologists err in trying to extract substantive
conclusions from analytic premises. (Even in the B version, however,
we cannot understand the opening discussion of the limitations of self-
knowledge without looking back to the transcendental psychology of
the Deduction.) Because it offers richer discussions, my interpretation
will focus on the original version. Specifically, I will consider the first
three Paralogisms in A, which concern the substance, simplicity, and
persistence of the soul. The Fourth Paralogism in A is a defense of
transcendental idealism that was reworked and relocated in the second
edition. Its place in B is occupied by cursory reflections about materi-
alism that I treat only in passing.
Like the chapter itself, I take the First Paralogism as a model for the
rest. I argue that we cannot follow Kant's discussion of this paralogism
unless we recognize that it is not just someone else's mistaken argument;
it is a possible misreading of the doctrine of apperception itself. Kant's
critique is simultaneously a criticism of the methods of Rational Psy-
chology, a caveat about possible misunderstandings of apperception,
and a lesson in the limits of transcendental psychology.

Puzzles of the First Paralogism

Kant's official account of his critique of Rational Psychology is not very

helpful. Paralogisms are formally invalid arguments. Both editions claim
that the First Paralogism exemplifies the fallacy found in all the Paral-
ogisms, ambiguous middle (A402, B411):
That, which is represented in the contents of cognitive states6 as the
absolute subject of our judgments and cannot therefore be employed as
determination of another thing is substance.
I, as a thinking being, am the absolute subject of all my possible judg-
ments, and this content representing me cannot be employed as predicate
of any other thing.
184 KANT'S TRANSCENDENTAL PSYCHOLOGY
Therefore, I, as a thinking being (soul), am substance [A348, amended
translation; cf. B410-11].

In the first edition, this criticism is elaborated by the suggestion that

terms are used "transcendentally" in the major premise and "empiri-
cally" in the minor premise and conclusion (A402-3). We have already
encountered Kant's primary meaning of "transcendental" Something is
transcendental if it concerns our manner of knowing objects, in partic-
ular, the nonempirical origins of cognition.7 An alternative meaning for
"transcendental" is introduced at the beginning of the Dialectic, how-
ever, and that would seem more appropriate in this context. A category
is employed "transcendentally" if it is employed beyond the limits of
experience, or without heed to the limits of experience (A296/B352-
53). (For ease in reference, I will tag the first sense "transcendental,"
and the second, "transcendental-") Kant's criticism appears to be that
the category ["substance" and related terms] is used "transcendentally2"
in the major premise, because it is used independently of conditions
that would enable us to tell whether objects that we encounter fall under
the concept (see A349). By contrast, it is used "empirically" if used in
conjunction with a "schema" that enables us to tell which objects should
be classified as substances. At a number of places Kant says that the
"pure" category is just the empirical category without its schema (A241-
42, A242-43/B300-301, A248/B305). Or, reversing the point, he regards
the "schematized" category as including a (needed) further specification
of the "pure" category. Thus, empirical substances would be a subset
of transcendental substances, empirical subjects a subset of transcen-
denta!2 subjects, and so forth. The First Paralogism is invalid, because
although all transcendental subjects might be transcendenta!2 sub-
stances, that does not guarantee that a particular empirical subject also
belongs to the subset of empirical substances.
The criticism of ambiguous middle through a confusion of empirical
and transcendental concepts is superficially clear. As we probe more
deeply, however, Kant's position becomes murky. In the second edition,
he elaborates the charge of ambiguous middle by claiming that the minor
premise involves a peculiar use of terminology (B411a). Whether the
major premise employs transcendental concepts and the minor involves
empirical concepts or vice versa, the argument would still be invalid by
virtue of ambiguous middle. Nevertheless, the unheralded reversal on
how the error occurs is puzzling.
Besides the confusion about how exactly the transcendental—empirical
mix-up occurs, the diagnosis of ambiguous middle is unsatisfactory for
 The Limits of Transcendental Psychology 185

another reason. It provides no justification for Kant's claim that the

arguments of the Dialectic are "natural," "inevitable," and can entrap
"even the wisest of men" (A298/B355, A339/B397). To fulfill the stated
mission of the Dialectic, and even to justify his own interest in these
arguments, he needs to make clear why these particular arguments are
so attractive and so treacherous.
Thus far we have only a confusing account of why we draw the con-
clusion given the premises, namely, because we fail to notice or under-
stand the ambiguous terminology. Why would anyone (or everyone)
accept the premises? Kant regards the major premise as compelling I
believe, because it merely states a standard definition of "substance."
He mentions the definition several times prior to the Paralogisms chap-
ter, without providing any explanation or defense (B149, B288, A241/
B300). This definition also occurs in the writings of Leibniz, also without
explanation or defense, suggesting that it was commonly accepted at
the time.8 Thus, presumably, we are all supposed to grant the major
premise, because it is merely a definition. (The same is true for the
major premise of the Third Paralogism.)
What about the minor premise? The crucial interpretive fact about
this chapter is that Kant endorses, on some reading, the minor premises
(and conclusions) of all three Paralogisms (A349, A350, A354, A356,
A363, A365). If we can figure out the basis of his support for these
claims, then that should tell us why he regards these arguments as
compelling and important. Presumably, it will also shed some light on
the empirical or transcendental status of the minor premises.
In the A edition, Kant endorses the First Paralogism's minor premise
in an unequivocal but rather confusing statement: "Now in all our
thought the I is the subject (in which thoughts inhere only as determi-
nations) [and this I cannot be employed as the determination of another
thing]" (A349, my parentheses and brackets).9 This is an odd passage,
because the minor premise is an odd claim. In the first phrase, it is not
clear whether Kant is describing the I as a subject, or "I" as the subject
of thoughts or sentences. The latter reading is suggested by the wording
of the paralogism itself.10 The phrase I put in parentheses appears to
be part of a claim that thoughts (themselves) are modifications of, or
belong to, an I (itself). Finally, the bracketed clause may also conflate
talk of representations and talk of things.
After A349 Kant reiterates the point that different thoughts belong
to a common I four times at A350: "I [is] the common subject in which
it [all thought] inheres," "The I is indeed in all thoughts," "[the I] is
the constant logical subject of thought... ," "in it [consciousness] all
186 KANT'S TRANSCENDENTAL PSYCHOLOGY
our perceptions must be found." Despite the barrage, commentators
have tended to ignore this point, for a reason suggested by Jonathan
Bennett. The claim that all my judgments are mine is true but trivial.11
This dismissal is encouraged by remarks in the B Deduction and in the
B version of the Paralogisms chapter that the doctrine of the I think is
"identical," "analytic." If any proposition is analytic, "all my judgments
are mine" is analytic. However, it is important to look closely at Kant's
statements about the analytic or synthetic status of the apperception
principle. Wolff and his followers took Rational Psychology to be a
demonstrative science. Thus, for them, its propositions are analytic.
When Kant emphasizes analytic formulations of apperception, he makes
the point that no substantive claims about the thinker can be extracted
from analytic propositions (see B135, B407). That is, these remarks are
intended as explicit or implicit criticism of the methods of Rational
Psychology.
In the Deduction and in the Paralogisms, Kant's own position on the
synthetic nature of the doctrine of apperception is clear. He explicitly
characterizes his principle of apperception as "synthetic" at A117a and
at B134a. He even indulges in a bit of word play, presumably at the
expense of the Wolffians: "the analytic unity of apperception is possible
only under the presupposition of a certain synthetic unity" (B133). Fur-
ther, if we examine the texts of A349 and A350, what he endorses is
not the minor premise itself, "I am the subject of my judgments," but
a related synthetic claim, "different judgments belong to a common
subject." If the latter claim is true, then the former will not only be
true (by identity of subject and predicate), but true of actual subjects
and judgments.
Counting the minor premise itself, Kant repeats the point that dif-
ferent thoughts, judgments, or perceptions belong to a common subject
five times in the space of two pages. Why? This is hardly a point that
he needs to press against the Rational Psychologists. After all, they
believe that all perceptions belong to a simple, numerically identical
soul that endures throughout life and considerably beyond. I do not see
how these repeated claims can be understood, except as reminders of
his own anti-Humean argument for the unity of apperception.
We have encountered five textual puzzles in Kant's discussion of
the First Paralogism. What exactly is the transcendental-empirical
confusion that is supposed to be the root of the paralogism? Why
does he take this argument to be compelling? Why does he support,
and believe all must support, its minor premise? Why does he con-
flate I's with the representation "I"? Why does he insist over and
 The Limits of Transcendental Psychology 187

over again that different cognitive states belong to a common sub-

ject? The relatively obvious solution to the last puzzle will also yield
solutions to the other four. If we go back and reflect on Kant's reac-
tion to Hume's denial of mental unity, two key points emerge that
will enable us to follow the twists and turns of this text. First, he is
in complete agreement with his predecessor about the failure of in-
trospection to divulge a continuing self. This point is made clearly in
both editions of the Deduction—and reiterated in the Paralogisms
chapter, twice in A and once in B (A107, B133-34, A350, A381,
B413). Second, Kant believes that by analyzing the prerequisites of
cognition, he has demonstrated that diverse cognitive states must
stand in real, synthetic connection with each other, and so belong to
the same consciousness. That is why he thinks the minor premise is
compelling. Its central claim, that all possible judgments belong to
one I, is the conclusion of his own argument in the subjective
Deduction.

Understanding the First Paralogism

To follow Kant's reasoning about the First Paralogism, we need to

disentangle the complex relations to his predecessors in these passages.
Like Descartes, he believes that any cognitive state, a fortiori any judg-
ment, must belong to a self, which we may call the "subject of the
judgment." Unlike Descartes, he does not conceive of this self as a
simple substance, but only as a system of synthetically connected or
connectible states. And unlike Descartes, he believes in this self, because
he has analyzed the prerequisites for various cognitive tasks. Against
the background of the subjective Deduction, we can understand how
Kant interprets the first clause of the minor premise and why he accepts
it: "I, as a thinking being, am the absolute subject of all my possible
judgments, and this content representing me cannot be employed as
predicate of any other thing" (A348, amended translation). It means
that any possible judgment must be regarded as belonging to an I
[to me or to another], that is, to a synthetically connected system of
states, and he supports this claim on the basis of his own argument
against Hume.
If we look back to prominent statements of the unity of apperception
thesis in the Deduction, then we can also clear up the representation-
object confusion (the "I"-I confusion) in these texts. The first edition
Deduction makes the claim that all judgments [actually, all cognitive
188 KANT'S TRANSCENDENTAL PSYCHOLOGY

states] must belong to a self in the material mode (A107, A116, A122,
etc.). In the second edition this claim is expressed by saying that the
representation "I think" can be attached to all my judgments (B131-
32). This manner of expression may court confusion, but the claims are
materially equivalent: If and only if for any judgment J, it must belong
to some subject, then it must be possible (for someone) to construct a
true sentence "I think (that) J."12 That is, given that all judgments must
belong to some subject, it must be possible for some "I think" to be
the subject of every judgment. Presumably, Kant adopted the second
formulation in the later edition, because it enables him to provide a
partial explanation of why we mistakenly believe that the self is a sub-
stance. One criticism of the First Paralogism is that we confuse the fact
that the representation "I think" could be an invariant feature of all
judgments with the notion that the self is perpetually intuitable (A350).
The second clause of the minor premise is more difficult, because
Kant never says why the "I" cannot occur as a predicate. Further, he
points out that any concept can occur in either the subject or the pred-
icate position in a sentence (cf. A349, B128-29, A242-43/B300-301).
One obvious possibility is that he simply throws in the second clause,
because of its occurrence in the standardly accepted definition of "sub-
stance" that occurs as the major premise. (Another, complementary,
possibility is that he uses this definition because it presents substances
as somehow basic and he wants to build at least a verbal bridge between
the Paralogisms chapter and his introductory remarks about reason's
need for ultimates.) In any case, all Kant is really serious about in this
premise is the claim that we must attribute all judgments to a self. This
is the claim that is repeatedly asserted in his discussion of the
Paralogism.13
How do the Rational Psychologists err? Kant's ultimate diagnosis
of the error is the same in both editions: "From all this it is evident
that Rational Psychology owes its origin simply to misunderstanding.
The Unity of consciousness . . . is here mistaken for the intuition of sub-
ject as object, and the category of substance is then applied to it" (B241-
22; cf. A402). Given the background of Kant's response to Hume, we
can understand one way that he thinks Rational Psychologists have been
led astray. As Hume noted, we do not have an intuition of the self. On
the other hand, in trying to understand our cognitive capacities, we must
recognize that cognitive states all belong to one consciousness. Hence,
contrary to Hume, we must talk about a continuing consciousness or an
I. We use the representation "I." The problem is that Rational Psy-
chologists recognize that we use the representation "I" without clearly
 The Limits of Transcendental Psychology 189
understanding the proper foundation of this usage. So they assume that
it rests on the deliverances of inner sense. Since there is, in fact, no
inner intuition of self, we do not perceive the self as an attribute, a
complex, or a series. The error arises because they expect to find an
intuition of the self and so mistake the absence of any intuition for the
intuition of something with remarkable properties: "in what we entitle
soul everything is in continual flux and there is nothing abiding except
(if we must so express ourselves) the I, which is simple solely because
its representation has no content, and therefore no diverse elements,
and for this reason seems to represent, or (to use a more correct word)
denote, a simple object" (A381-82, amended translation).
Who perpetrates this error of Rational Psychology? Seemingly not
Leibniz. He claims that we think of ourselves as being "of substance,
of the simple... of the immaterial" "by the knowledge of necessary
truths and by their abstractions [that enable us to rise to] reflective acts,
which enable us to think of what is called 7."14 Without considering
exactly what these reflective acts are supposed to be, they are clearly
not simple introspectings. Conversely, Descartes seems to be guilty of
something very like the error that Kant describes in the following pas-
sage: "when I consider the mind . . . I cannot distinguish in myself any
parts, but apprehend myself to be clearly one and entire. . . ,"15 This
is also a direct criticism of the basic assumption of Wolff's empirical
psychology. In every act of consciousness we are aware of ourselves, as
well as of the object of our consciousness.16
This general criticism of Rational Psychology reveals an important
source of confusion about the mind. We have various beliefs about the
mind and its states. The Cartesian picture, which represents us as directly
aware of thoughts and thinking, provides a seductive explanation for
the source of those beliefs that leads us to overestimate our epistemic
situation with respect to the mind. We think that we know more than
we do about the mind and we think of our knowledge as resting on
direct evidence. The explanation is seductive because the Cartesian
"source of evidence" about the mind can never provide any counter-
evidence to our preconceived notions, because it presents no diverse
elements of intuition [data] at all. Kant tries to prevent these confusions
by repeating and extending Hume's original objection to Descartes: We
have no knowledge of the thing that thinks "through awareness or
through reasoning" (A355, my translation).17
Although this criticism strikes at the heart of the Empirical Psychology
on which Rational Psychology rests, it bears no relation to the official
account of what goes wrong: a confusion of empirical and transcendental
190 KANT'S TRANSCENDENTAL PSYCHOLOGY
claims. To understand this diagnosis, we must again look back to the
Deduction's argument for the unity of apperception. As noted, Kant
uses "transcendental" in two quite different senses. The appropriate
sense for understanding the Paralogisms would appear to be that intro-
duced at the beginning of the Dialectic. A concept is used "transcenden-
tal!)^" if it is used without its schema. However, Kant's argument that
diverse cognitive states must belong to the unity of apperception—the
argument that stands behind his support for the minor premise—reveals
that this unity is a prerequisite for cognition. That is, his argument
establishes a transcendental, status for the principle of apperception.
Hence, in the Paralogisms, he legitimately refers to the subject of judg-
ment as "transcendental," (A346/B404, A340; see also B411 and
B411a). As he points out, the difficulty is that in order to infer that the
self is an empirical substance, the minor premise would have to be
asserted on the basis of empirical intuition, presumably the perceiving
through inner sense of a permanent self (A403-404), and it lacks this
kind of empirical support; it is a transcendental, claim. That is why in
proffering the minor premise as support for the conclusion, "this syl-
logism . . . puts forward the constant [transcendental] logical subject of
thought as being knowledge of the real subject.. . [and so palms] off
upon us what is a mere pretense of new insight" (A350). Since one way
to describe this misstep would be as a result of confusing transcendental,
and empirical claims, Kant's ambiguous usage provides a verbal victory
for the doctrine that all metaphysical errors rest on empirical-
transcendenta!2 confusions.
Since the diagnosis of a transcendental-empirical slide is itself the
product of a terminological slide, what is the real error in this rea-
soning? What mistake does one make in confounding empirical and
transcendental, claims? Kant's analysis of this error is critical to un-
derstanding transcendental psychology, because it clarifies the limits
of the enterprise. It is also important to philosophy of mind more
generally, because it reveals that certain kinds of inferences are in-
valid. In the Deduction Kant engages in transcendental psychology to
determine the prerequisites of cognition; on the subjective side, to
determine what properties thinking beings must have in order to per-
form cognitive tasks. One result is that the states of thinking beings
must be synthetically connected or connectible. The mistake comes in
believing that this type of analysis of the abstract properties or facul-
ties required for cognition provides information about what sort of
thing a thinking being is.
 The Limits of Transcendental Psychology 191
The analysis, then, of the consciousness of myself in thought in general,
yields nothing whatsoever towards the knowledge of myself as object.
The [transcendental] logical exposition of thought in general has been
mistaken for a metaphysical determination of the object [B409].
The point is made even more clearly at A398:
If anyone propounds to me the question, 'What is the constitution of a
thing which thinks?' I have no a priori knowledge wherewith to reply.
For the answer has to be synthetic—an analytic answer will perhaps explain
what is meant by thought, but beyond this cannot yield any knowledge
of that upon which this thought depends for its possibility... intuition
[would be] required; and owing to the highly general character of the
problem, intuition has been left entirely out of account. Similarly no one
can answer in all its generality the question, 'What must a thing be, to
be movable?'

Transcendental psychology is an abstract study; it addresses only very

general questions. What faculties are required for representation, judg-
ment, and other tasks involved in cognition? Despite the importance of
these analyses, Kant realizes clearly that they furnish no serious clues
about what the soul is like.18 This discovery came as a surprise: "Sus-
picion is thus thrown on the view, which at first seemed to me so plau-
sible, that we can form judgments about the nature of the thinking being,
and can do so from concepts alone" (A399). "From concepts alone"
does not mean that he ever believed that we could determine what the
soul is like through what we now call "conceptual analysis," that is,
either by figuring out what the ordinary person means by "thought" or
by giving necessary and sufficient conditions for the (correct) application
of the concept.19 Rather, Kant once believed that we could figure out
what a thinking thing was like by determining what characteristics were
necessary for a thinking thing (or thought) to be possible.20 He now
realizes that this is a mistake. To determine what something is made
of, its constitution, we need intuition (or observation). Understanding
the abstract subtasks that must be performed in cognitive tasks cannot
substitute for observation in determining constitution. This surprising
realization is an important insight. With the aid of reflecting on the
differences between the hardware and software in computers, it has
reemerged in recent years as the principle that function does not de-
termine form.21
When we turn to the Second Paralogism, I shall argue that Leibniz
seems to make the false step from a highly abstract description to a
192 KANT'S TRANSCENDENTAL PSYCHOLOGY
claim about constitution just described. With this Paralogism, however,
the argument for imputing diverse judgments or perceptions to a
transcendental, subject that stands behind the acceptability of the minor
premise is Kant's defense of transcendental synthesis in the Deduction
(see A350). There appears to be nothing in the doctrines of Rational
Psychology that could provide any legitimate support for the claim.
Further, although the announced purpose of the chapter is to study the
fallacies that flow from the doctrine of the I think, that is, from the
cogito, this model Paralogism is plainly not the cogito. Kant mentions
the cogito in the Second Paralogism (A355), but only discusses it, or
Wolff's syllogistic version of it, in a footnote to the B version.22
Thus, the First Paralogism is best understood as Kant's own potential
paralogism. Presumably, he came to understand the general problem
of moving from certain kinds of abstract descriptions to substantive
descriptions by reflecting on his own efforts in transcendental psychology
and then noting similarities between these analyses and some of the
work of his predecessors, in particular, Leibniz. In any case, many of
his general warnings about moving from highly abstract analyses to
substantive claims about the self, and his particular objections in the
First Paralogism, are more easily understood as caveats about his own
position than as criticisms of his predecessors. One error that the First
Paralogism warns us against is inferring from the fact that we must regard
different judgments as belonging to a common subject to the position
that "the soul is substance," unless the latter "does not carry us a single
step further," that is, unless we intend the conclusion of the Paralogism
to assert no more than the minor premise itself (A350; see also A365).
The view that this is a caution about his own doctrines is further con-
firmed by the fact that similar warnings appear in the B Deduction
itself—for example, "in the synthetic original unity of apperception, I
am conscious of myself, not as I appear to myself, nor as I am in myself,
but only that I a m . . . " (B157), and "The I think expresses the act of
determining my existence. Existence is already given thereby, but the
mode in which I am to determine this existence . . . is not thereby given"
(B157-58a).
There is a further reason behind Kant's conviction that the First Paral-
ogism is rooted in transcendental-empirical confusions. As noted ear-
lier, he changes his mind about how to classify the minor premise. This
vacillation reflects genuine perplexity. We have already seen why he
suggests a transcendental! status for this claim in the second edition
(B411a) and why he sometimes describes the subject of judgment as
"transcendental," in the first edition. Yet, in the second edition, he also
 The Limits of Transcendental Psychology 193

describes the claim as "empirical," on the grounds that it imputes ex-

istence (B157a, B422a). He has difficulty expressing this point because
he lacks the crucial concepts 'theoretical entity' and 'functional descrip-
tion,' but he gets the idea across:
The I think expresses an indeterminate empirical intuition, i.e. perception.
[It] signifies only something real that is given, given indeed to thought in
general, and so not as appearance, nor as thing in itself (noumenon), but
as something which actually exists... [B423a].23

Thus, Kant himself regards the minor premise as both empirical and
transcendental,. Small wonder that he thinks it is extraordinarily easy
to fall into thinking that this premise is "empirical," in the sense that
it rests on awareness of an I, or that he thinks this argument has a
wickedly ambiguous middle term.
Tying these points together, this is Kant's analysis of the First Paralo-
gism. The argument is compelling, because we must accept the major
premise (because it is a definition), and we must accept the minor prem-
ise on the basis of his own argument for the unity of apperception in
the Deduction. The Deduction establishes that basic cognitive tasks
require that diverse states be synthetically connected in a single con-
sciousness. Hence, the minor premise is a transcendentall claim, "I" is
a transcendental, representation, Fs are transcendental, subjects. Con-
fusion sets in because we do not know how to handle the minor premise,
and so fall into transcendental-empirical confusions. These can occur
in two different ways. We may not recognize the actual support for the
claim at all, and so assume that it is an empirical claim resting on
intuition, since it imputes existence. We then expect to find an intuition
of the self and the absence of such an intuition leads us to peculiar views
about what inner sense is disclosing. Alternatively, we may recognize
that the minor premise derives its support from analyses of the abstract
properties of the mind required for basic cognitive tasks, but not un-
derstand the limitations of these analyses. So we assume that they can
support descriptions of the nature of a mind. Both these confusions lead
us to believe that we can move beyond the minor premise and assert
that the self is a substance. In fact, this conclusion can be maintained
only if it merely restates the minor premise interpreted to mean that
there is a transcendental, subject of judgment.
The mistakes involved in this reasoning about the self are hardly
jejune. Rational psychologists—and their modern counterparts—
err, because they fail to grasp two surprising limitations on our self-
knowledge. Even though we can provide abstract descriptions of nee-
194 KANT'S TRANSCENDENTAL PSYCHOLOGY
essary features of minds for various cognitive tasks, we cannot move
from this level of description to a description of the stuff of which
thinking things are made. In "Leibniz and the Simplicity of the Soul"
I will consider a modern example of this failing.
The other limitation seems even harder to accept. Despite its "prox-
imity" to us, we have no direct acquaintance with the thing that thinks.
In believing that we do, philosophers overestimate what we know about
the self. Often this overestimation is disguised by the apparently negative
character of the claims. So, for example, Zeno Vendler writes:

My claims, therefore, that it is the same "I" that underlies transference

[imagining yourself to be another], and the same "I" that could be in
another state [that is, the state of being another mind], have to be under-
stood entirely via negativa: not distinct, not different. Thus, the unity of
this "I" is not like the unity of material things... for all these things can
be named and identified in many ways. Thus we are driven back to the
desperate analogy of the "prime matter." . . . One more instance of the
great strain under which our concepts labour in these matters.24

Although Vendler sees himself as accepting the negative teachings of

the Paralogisms, I doubt that Kant would agree. Our knowledge of the
self does not rise to the level of knowing that it is an ineffable something
that remains the same in transference. The point against Descartes works
equally well against Vendler: "the I, which is simple [ineffable] solely
because its representation has no content, and therefore no diverse
elements, and for this reason seems to represent, or (to use a more
correct word) denote, a simple [ineffable] object" (A381-82).
Thomas Nagel also implies that the lesson of the Paralogisms is merely
that the self "cannot be defined as a kind of object, either physical or
non-physical, but must be understood as the same subjective conscious-
ness."25 Later, he asserts that, "The concept of the self. .. implies only
that if it refers at all, it must refer to something essentially subjective,
often identifiable nonobservationally in the first person and observa-
tionally in the third."26 The essential subjectivity of the subject of
thoughts is the puzzle Nagel wishes to examine in The View from No-
where. Given Kant's analysis, however, this puzzle seems to be built
out of confusions. The claim that the subjective self is identifiable "non-
observationally" simply substitutes a mysterious process—nonobser-
vational "observation"—for introspection. The eighteenth-century
doctrine of inner sense is preferable, because we understand it well
enough to recognize with Hume and Kant that it does not disclose a
self, subjective or otherwise. Again, Kant's central objection would be
 The Limits of Transcendental Psychology 195

that our knowledge of the self does not rise to the level of recognizing
it to be subjective or mysterious. Philosophers attribute remarkable
properties to it, because there is no intuition of something unremark-
able—because there is no intuition of anything at all. Our belief in the
self has a totally different foundation.

Identity Through Time

Because the First and Third Paralogisms have many similarities, I skip
to the Third Paralogism: "That which is conscious of the numerical
identity of itself at different times is in so far a person. Now the soul is
conscious, etc. Therefore it is a person"(A361). The topic of the Third
Paralogism is the identity of the self through time. Like Hume, Kant
construes identity as requiring a permanent element that persists
throughout the changes in attributes (A361—62). He also uses that which
is permanently available to perception, or that which abides while other
things change, as the schema of substance, however (A143/B183, A349).
Thus, the First and Third Paralogisms appear to cover exactly the same
territory, the permanence of the I.
These arguments are also similar in that their major premises are
acceptable, because they merely state commonly accepted definitions.
Here the major premise is the definition of "person" employed by Leib-
niz and Locke.27 Again, the central interpretive problem is to determine
why anyone, or everyone, would accept the minor premise. Despite the
problems commentators have had with this passage, the logic of Kant's
argument is fairly clear, if we bear in mind his own earlier results.28 The
subjective Deduction showed that cognitive states must belong to a
synthetically connected system of diverse states, an I. Thus, it showed
that we must recognize that a present cognitive state belongs to a system
that includes cognitive states occurring at different times. We must be
conscious of [or, better, cognizant of] the identity of the I at different
times.
So far, the support for the minor premise is exactly the same as that
for the minor premise in the First Paralogism. What is new in this
Paralogism is the injection of a second distinctive Kantian doctrine: time
as the form of inner sense. This is explicit in the text:
all time is merely the form of inner sense. Consequently, I refer each and
all of my successive determinations to the numerically identical self, and
do so throughout time, that is, in the form of the inner intuition of myself
[A362, my emphasis].
196 KANT'S TRANSCENDENTAL PSYCHOLOGY
How does the addition of this doctrine bolster, or appear to bolster, the
case for a permanent I? If the results of the Aesthetic are combined
with those of the Deduction, then we construct an argument for the
minor premise, "Now the soul is conscious of the identity of itself at
different times." In fact, we can construct an argument for the stronger
claim that Kant takes to be the core of the Paralogism: The soul is
conscious of the identity of itself at all times ("in the whole time in
which I am conscious of myself... I am to be found as numerically
identical in all this time") [A362].
This is the argument. Time is the form of inner sense. For me, there-
fore, moments of time exist when and only when I am aware of a
cognitive state through inner sense. Whenever I am conscious of a cog-
nitive state, however, I must attribute that state to an I. Thus, from my
own point of view, I am conscious of, or better, cognizant of an I, at
all moments of time. That is to say, I am cognizant of the unbroken
continuity of myself. In Kant's words:
all time is merely the form of inner sense. Consequently, I refer each and
all of my successive determinations [cognitive states] to the numerically
identical self, and do so throughout time, that is, in the form of inner
sense.... This being so, the personality of the soul has to be regarded
not as inferred but as a completely identical proposition of self-conscious-
ness in time; and this, indeed, is why it is valid a priori... [A362].
Thus, to the consternation and confusion of some of his readers,29 Kant
maintains that [permanence] and identity are "necessarily bound up with
my consciousness" (A363) and that "we must necessarily judge that we
are one and the same throughout the whole time of which we are con-
scious" (A364). In a certain sense, the proposition that the thinker exists
continuously is true, and a priori0, because it follows from what Kant
takes to be the a priori0, form of inner sense, and the a priori0 unity
of apperception.30
Despite the overlap between the First and Third Paralogisms, the
reasoning turns on very different considerations. In the First Paralogism,
the argument tries to move from the soul as that to which judgments
belong, and hence the subject of judgments, to the soul as substance;
here there is a direct attempt to establish substantiality, by appealing
to unbroken continuity and permanence. What is the error in this rea-
soning? It does not concern the self, but time.31 Although it is true we
must attribute all cognitive states to an I, and thus, that we must be
cognizant of an I throughout all the times that we are conscious (A364),
we cannot infer that the I exists at all times, simpliciter. To dramatize
 The Limits of Transcendental Psychology 197

the impossibility of regarding these reflections as establishing an objec-

tively (i.e., intersubjectively) valid claim about identity and continuity,
Kant invites us to consider our own "permanence" from the point of
view of an outside observer. Whenever he attributes a cognitive state
to the person he observes, he must regard that state as belonging to a
continuing I for reasons that are by now familiar (A363). He will not
infer the unbroken continuity or permanence of this, however, for the
simple reason that there will be times when he does not, or cannot,
attribute a conscious state to the person at all.32 So this I will not be a
permanent element in his experience (A362-63). Thus, the claim that
I exist at every moment in time is not intersubjectively valid.
If the conclusion of the Paralogism, "I am a person" is construed to
mean "I exist at every moment in time," then it cannot be validly inferred
from considerations about time and apperception. As in the case of the
First Paralogism, however, the conclusion is assertible if it merely re-
states the minor premise claim that I am cognizant of the identity of my
consciousness at all different times at which I am conscious of anything
through inner sense. More bluntly, this conclusion is assertible if it is
interpreted as materially equivalent to the doctrine of apperception plus
the doctrine that time is the form of inner sense (A365). After noting
that uninterrupted continuity cannot be established in the way just de-
scribed, Kant reiterates his claim that substantiality [and permanence]
cannot be established by inferring the constitution of the soul from the
formal requirements for thought (A363-65). In a footnote, he argues,
via an analogy with the momentum of billiard balls, that a present self
may bear all the necessary connections of continuity of thought and
memory to earlier selves even though there is no continuity of substance
among them (A363-64a).
To whom is this Paralogism compelling? The issue of whether thinking
beings think continuously or exist continuously was hotly debated among
Descartes, Locke, and Leibniz. Recently, R. I G. Hughes has suggested
that this Paralogism offers a criticism of the "notion of time available
to a Cartesian thinking being."33 I do not see how this reading can be
squared with the direct references to Kant's own doctrines of the form
of inner sense that I highlighted earlier however. Descartes has no
special views about time that lead him to infer that thinking beings think
at all times. He is led into the view that we think continuously, because
thought is the denning attribute of minds; hence, if minds cease to think,
they would cease to be. I suspect that Kant deliberately scouts the
reasoning in the Third Paralogism to prevent his views about time and
apperception from being co-opted into the tedious metaphysical debate
198 KANT'S TRANSCENDENTAL PSYCHOLOGY

about whether souls always think. In any case, the support he offers for
the minor premise would only convince someone who accepts his views
about apperception and inner sense. Even more than the First Paral-
ogism, this argument is Kant's potential paralogism. His warnings about
the illusions of Rational Psychology are caveats about his own doctrine
of apperception: This theory and the doctrine of the ideality of time
only appear to provide knowledge about the mode in which a thinking
thing exists.

Leibniz and the Simplicity of the Soul

That, the action of which can never be regarded as the concurrence of

several things acting, is simple.
Now the soul or the thinking I is such a being. Therefore, etc. [A351].
The Second Paralogism deviates from the pattern of the First and Third
in significant respects. The major premise does not state a widely ac-
cepted definition of "simple." "Simple" had a standard definition, "with-
out parts," but the major premise makes a different claim. Further, this
argument seems to derive from one of Kant's Rationalist predecessors.
Margaret Wilson argues persuasively that a similar line of reasoning can
be found in the writings of Leibniz.34 Her hypothesis about the source
of the argument is strongly confirmed by the fact that Kant repeats the
central line of reasoning of the Second Paralogism in discussing Leibniz's
and Wolff's views in On the Progress of Metaphysics.35
Although I will disagree with Wilson's interpretation of Kant's eval-
uation of Second Paralogism, her account of its topic seems correct. On
her view, Leibniz argued against materialism through the following re-
ductio. If the thinking thing were a body or a machine, then it would
have parts. Imagine enlarging such a thinking machine, so that we could
enter it, as we can a mill. Leibniz continues:
If we did this, we should find nothing within but parts which push upon
each other; we should never see anything which would explain a percep-
tion. So it is in the simple substance, and not in the composite substance
or machine, that perception must be sought.36
As Wilson observes, this argument is hopelessly cryptic as it stands. She
offers a very plausible interpolation, however: the feature of thought
that could find no correlate in a machine is the unity that binds the
different elements into one thought.37 What is not entirely clear from
 The Limits of Transcendental Psychology 199
the passages she cites is the direction of Leibniz's argument. Does he
argue that thoughts must have a special unity because they belong to
the I which has a true unity (and whose unity is established through
quite general metaphysical considerations)?38 Or does he try to prove
the unity of the I from the unity of thought?
Leibniz seems to argue in both directions. An argument from the
simplicity of the soul to the unity of thought holds no interest for Kant,
however, since he would reject the metaphysical supports for the sim-
plicity claim. The argument that he considers tries to move from the
unity of particular thoughts or representations to the simplicity of the
thinking thing, or from the unity that characterizes thinking to the sim-
plicity of the thinker. This is the move that would be sanctioned by the
major premise. Presumably, he supplies this premise to bring out the
hidden assumption in this reasoning. He does not and cannot endorse
the major premise. He cannot endorse it, because it is flatly inconsistent
with his general claim that the [transcendental] logical exposition of
thought in general can yield nothing whatsoever toward the knowledge
of myself as object (B409). In the text of this Paralogism, he does not
endorse it, even when explaining how we reach the conclusion. We do
not conclude that "I am simple" through an inference at all. Rather, "I
am simple" is a direct expression of apperception (A355; see later).
Nevertheless, he does accept two pieces of this Paralogism—the minor
premise and the conclusion. Since we have no intuition of the self, the
representation "I think" cannot be associated with any diverse elements
of intuition at all; a fortiori, it cannot be associated with complex con-
tents. The representation "I think" can be regarded as designating some-
thing simple (i.e., without parts) in the purely negative sense that it
does not designate complex sensory data (A355, A356). Thus, Kant
grants the conclusion that the soul is simple, if "simple" is so understood
that the conclusion merely expresses his own view about the absence of
any intuition of the I:
The proposition, 'I am simple', must be regarded as an immediate expres-
sion of apperception.... 'I am simple' means nothing more than that this
content of cognitive states, I, does not contain in itself the least diversity
of elements, and that it is absolute (although merely logical) unity [A354-
55, amended translation].

This point is familiar from the First Paralogism. His rationale for the
minor premise introduces an important new line of thought. Like the
First and Third Paralogisms, the key to understanding this discussion is
his support of the minor premise: "The soul, or the thinking I, is such
200 KANT'S TRANSCENDENTAL PSYCHOLOGY

that, its action can never be regarded as the concurrence of several

things acting." More simply, thinking can never be regarded as the
concurrence of several things acting. He supports this premise in passage
strikingly like William James's better-known discussion of the topic:
For suppose it be the composite that thinks: then every part of it would
be a part of the thought, and only all of them taken together would contain
the whole thought. But this cannot consistently be maintained. For con-
tents of cognitive states (for instance, the single words of a verse), dis-
tributed among different beings, never make up a whole thought...
[A352, amended translation].39

Kant's presentation almost makes the argument appear trivial. One

thought cannot be realized in different beings, because then there would
be no one being who had the whole thought. His first sentence rules
out this reading, however, by conceding that the collection of different
beings would contain the whole thought. His point must be that a col-
lection of different beings is just not the right sort of thing to realize a
complex thought. Unfortunately, the last sentence simply states that
representations distributed among different beings cannot make up a
thought without explaining why this is so.
Wilson reads Kant as subscribing to the view that the elements of a
thought must stand in the appropriate "conceptual connection."40 By
this I assume that she means that the elements of the thought must stand
in the appropriate syntactic and semantic relations to one another. So,
for example, the verb not only must agree with the subject, but must
denote an activity that is possible for the subject. As already noted,
"Vorstellung" is ambiguous between the content of a cognitive state and
the cognitive state itself. In terms of this distinction, Wilson interprets
the claim about the "indivisible unity of a Vorstellung" (A355) to refer
only to the unity of its content, and not to its unity qua cognitive state.
She thinks that Kant criticizes the Second Paralogism by rejecting the
datum of the "true unity of consciousness" of cognitive states.41
This interpretive issue is difficult to settle within the confines of the
Second Paralogism, because Kant is very cryptic. Even in these passages,
however, he clearly identifies the I think with his own doctrine of ap-
perception (as in the preceding citation about simplicity). And since
apperception is a doctrine about the unity of consciousness, it is doubtful
that this is what he means to deny.
Further, the most prominent passage in the text of the Paralogism
echoes a number of Deduction themes about the unity of representa-
tions, synthesis, and the unity of consciousness.
 The Limits of Transcendental Psychology 201
The so-called nervus probandi of this argument lies in the proposition,
that if many contents of cognitive states are to form a single thought, they
must be contained in the absolute unity of the thinking subject [A352,
amended translation].
As we have seen, Kant claimed at A99 that although intuition sup-
plies us with diverse elements, these cannot be represented as a di-
versity and contained in a single cognitive state without a synthesis.
Further, at A116 and B132, he maintained that unless cognitive
states can be synthetically connected to each other, and so belong to
a unified consciousness, they could not represent anything. And, fi-
nally, at B137, he concluded that "all unification of cognitive states
or their contents demands unity of consciousness in the synthesis of
them" (see also A102, A103, A108, A113, A123-24, B131a). Given
these propositions, it follows that a thought with a complex content
must be produced by a synthesis of cognitive states and that these
states must all belong to a unity of apperception. This point is per-
fectly illustrated in the case of a verse. The speaker must produce a
verse sequentially, and the listener must hear it that way. For either
to comprehend it, however, the different pieces must be combined in
a single representational state occurring at a single moment in time.
And the content of that state must be the synthetic product of the
contents of the states containing the elements.
Thus, Kant does not deny the unity of consciousness. He has, how-
ever, a particular understanding of it: his own theory of the synthetic
unity apperception. Again, he accepts the minor premise, insofar as it
expresses the unity of apperception. Thinking cannot be regarded as
the concurrence of several distinct, unconnected things acting. Thought
is possible only where synthetic connection allows the different parts of
a thought to be united in a single thought, or representational state. He
expresses this point as the view that the "subjective I can never b e . . .
divided and distributed" (A354). For otherwise, we could not say, "I
think (the diverse elements in a single cognitive state)" (A354, my
translation). Since the subjective I comprises the synthetic connections
across cognitive states, it cannot be divided. For were these connections
severed, there could be no content in a thought.
What is the error perpetrated by this Paralogism? Once again the
crucial mistake comes in the attempt to move from abstract descriptions
of the necessary conditions for thinking, to claims about the constitution
of the self: "This much, then, is certain, that through the I, I always
entertain the thought of an absolute, but [transcendental] logical, unity
of the subject (simplicity). It does not, however, follow that I thereby
202 KANT'S TRANSCENDENTAL PSYCHOLOGY

know the actual simplicity of my subject" (A356; see also B408). Kant
spells out exactly what is wrong with the attempted inference from an
abstract analysis to a characterization of the specific nature of the soul:
For the unity of the thought, which consists of many cognitive states or
their contents, is collective, and as far as mere concepts can show, may
relate just as well to the collective unity of different substances acting
together... [A353, amended translation].
Although thinking requires synthetic connections among the temporally
distinct states that contain parts of the thought, this does not imply that
only simple beings can have thoughts. All it implies is that a collection
of different beings can have a thought only if their states are synthetically
or contentually interconnected. This may rule out some materialist ac-
counts of thought—those that provide no explanation of synthetic con-
nection.42 However, this abstract description provides no serious clues
about the kinds of physical or nonphysical systems that might realize
the unity of thought.
Just in case Rational Psychologists try to resist this criticism, Kant
drives the point home in the second edition. It is not merely fallacious
to try to adjudicate among possible characterizations of the soul by
appealing to abstract analyses, it is potentially dangerous to the cause.
If Rational Psychologists are permitted to argue for the simplicity and
immateriality of the soul by claiming that they do not see how a material
substance could realize the unity of thought, then materialists would be
free to employ the same strategy to "establish" the opposite conclusion.
Since the latter do not understand how an immaterial substance could
realize the unity of thought, they may claim that the soul is material
(B418a).
There are important differences between Kant's First and Third Paral-
ogisms and the Second Paralogism. Only in this Paralogism does he
appear to carry out the stated purpose of the chapter, namely, to criticize
arguments of his Rationalist predecessors. Further, this may be a case
where an argument of one of his predecessors makes a positive contri-
bution to the doctrine of the unity of apperception. Nevertheless, the
evaluation of this argument turns on exactly the same considerations as
his evaluation of the First Paralogism. Both arguments are related to
genuine insights about the lack of intuition of the self and about the
necessary unity of thinking things. These insights are extremely slippery,
however, and can lead to totally unwarranted claims about thinking
selves. We misconstrue the absence of an intuition of the I as an intuition
of something with remarkable properties, and we falsely assume that
 The Limits of Transcendental Psychology 203

we can move from abstract analyses to descriptions of the nature of

thinking things.
In discussing the First Paralogism, I noted that major contemporary
philosophers still mistake the absence of an intuition of the I for the
intuition of something mysterious. Kant's second important caveat—
that we cannot move from the sort of abstract analyses provided by
transcendental psychology to claims about the intrinsic properties of
thinking things—can also provide needed discipline to contemporary
discussions. In a well-known article John Searle makes the following
claim:
It is not because I am the instantiation of a computer program that I am
able to understand English and have other forms of intentionality... but
as far as we know it is because I am a certain sort of organism with a
certain biological (i.e., chemical and physical) structure, and this structure
under certain conditions is causally capable of producing perception, ac-
tion, understanding, learning, and other intentional phenomena.43

In this passage Searle indulges in exactly the sort of argument from

ignorance that Kant warns Rational Psychologists against. Thinking ex-
hibits intentionality; that is, our thoughts are about things. It is not at
all clear how a computer can have states that exhibit intentionality.
Since people do have intentional states and since obviously they think
with their brains, then brains must be so constituted (have a particular
physical and chemical structure) as to be capable of having intentional
states.
If we substitute "unity" for "intentionality" and "immaterial soul"
for "brain," this is exactly the Rational Psychologists' argument. It is
not clear how machines can have states that exhibit unity. Since people
do have unified mental states, and since mental states are obviously
states of their immaterial souls, immaterial souls are so constituted as
to be able to have unified states. Kant's critique is equally apt. If Searle
is permitted to run this argument, then we should also permit imma-
terialists to point out that we do not currently understand how brains
can have intentional states, and then to draw their own conclusions. The
problem with both arguments is that, although we have certain abstract
descriptions of thoughts (they must be unified, they must be intentional),
we do not have the slightest idea what kinds of things might be able to
instantiate those properties. Proponents of such arguments disguise this
fact, by announcing that souls possess the requisite unity or that brains
possess the requisite intentionality. But these bold claims do not rest
on an understanding of how anything could possess such properties;
204 KANT'S TRANSCENDENTAL PSYCHOLOGY
they are supported entirely by the antecedent conviction that it is brains
or immaterial souls that do the thinking. Searle's conviction that brains
think is certainly more plausible than Leibniz's speculations about mo-
nads. As Kant's analysis shows, however, the arguments of these clever
philosophers are equally fallacious.44
 8
Cognitive Constraints
on Empirical Concepts

Kant and Cognitive Science

At the beginning I claimed that denying the transcendental psychol-

ogy of the Critique has two harmful effects. Kant scholars must try
to interpret a difficult text without engaging one of its central
themes. By this point I hope to have shown that taking transcenden-
tal psychology seriously does not lead to embarrassment. Rather, it
allows us to make better sense of the Aesthetic's doctrine that space
is the form of outer sense, better sense of the complex role of the
principle of apperception in the Deduction, better sense of the argu-
ment for the indispensability of the categories, and better sense of
the critique of Rational Psychology.
I have been less systematic in illustrating the second disadvantage.
By ignoring transcendental psychology, historians forego the opportu-
nity to contribute to contemporary debates. Chapters 4 and 5 show how
the theory of mental unity that emerges from Kant's analyses of cognitive
tasks provides an attractive solution to a philosophical problem of great
current interest: personal identity. But I have not shown, in any detail,
how this work might contribute to current struggles to understand cog-
nition. I conclude by looking at a very important debate in cognitive
science. This will provide an extended example of how the Critique's
analyses of the psychological faculties required for cognition can inform
contemporary work. An opportunity is being lost.
Kant was uncertain about the status of a science of the higher faculties,
what we would call a "cognitive psychology" but he called "applied
205
206 KANT'S TRANSCENDENTAL PSYCHOLOGY

logic" (see A54/B78). In the Logic, he claimed that such a science must
be possible, because like everything else in nature, the exercise of our
own faculties must operate by laws.1 The Critique maintained that ap-
plied logic could not aspire to be a demonstrated science, because it
must appeal to empirical principles (A55/B79). And in The Metaphysical
Foundations of Natural Science, he seemed to deny that any science of
the soul could be a proper, mathematical science.2
Nevertheless, Kant realized that general logic, and so transcendental
logic (or transcendental psychology), bore an important relation to a
possible cognitive psychology or applied logic. He suggests two some-
what different models in the Critique. General logic describes the way
in which the understanding itself operates; but in characterizing the way
people actually think, we must factor in the "accidental subjective con-
ditions which may hinder or help its application" (A55/B79). Just before,
he had presented applied logic as characterizing the "rules of the em-
ployment of understanding under the subjective empirical conditions
dealt with by psychology" (A53/B77; see also B140). On the former
model, the normal operation of the understanding is interfered with by
other faculties; in the latter, the ideally characterized faculty of under-
standing is realized by faculties that only approximate its ideal func-
tioning, regardless of the influence of other factors. Either way, Kant
clearly recognized that his work must have a bearing on empirical re-
search, even while he doubted the ultimate status of that research.3
Transcendental psychology offers an idealization of cognitive function-
ing; as such, it can provide guidance about the sorts of mental equipment
that empirical researchers need to look for.
Concepts have become an important topic in cognitive science.
Eleanor Rosch's discovery of the apparent "prototype" structure of
concepts inspired many follow-up studies aimed at determining how
concepts are represented in us.4 Concepts have also been prominent in
the developmental literature. How do children's concepts differ from
those of adults? What mental equipment is implied by children's uses
of concepts?5 Attempts to simulate knowledge representation in com-
puters have also led theorists to reconsider the nature of concepts.
Although the issue of concepts is somewhat amorphous, it can be
summed up in two (amorphous) questions: What kinds of concepts do
we (or children) use? What kinds of mental representations and/or
mental faculties underlie our ability to use the kinds of concepts that
we use?
I begin by considering a more limited question, which emerged from
Rosch's early work: Do people apply concepts by representing and using
 Cognitive Constraints on Empirical Concepts 207

necessary and sufficient conditions for membership in a particular cat-

egory? Although this is obviously an empirical question, the experi-
mental results are fairly confusing. I will argue that Kant's reflections
on what concepts must be like for knowledge to be possible can provide
needed theoretical insight into the issue. Then I take up the broader
question of how concepts are realized in us. We cannot turn to Kant
and expect to discover that the true theory of concepts has been lying,
undetected, in our midst for 200 years. My goals are much more modest
(and, I hope, realistic). Kant approached concepts in terms of his general
project of transcendental psychology: What must concepts be like, and
what faculties must underlie them, for us to be capable of knowledge
at all? I argue that his analyses of concepts, and of the faculty of reason,
offer a novel suggestion about the kind of mental equipment that un-
derlies concept use. This suggestion has four substantial virtues: It ex-
plains a fair bit of experimental data, it is relatively simple, it has clear
implications for the direction of empirical research, and it is principled—
that is, it is clear, given the way concepts are supposed to function, why
they should be supported by this type of mental equipment. Lest I raise
expectations too high, I should repeat that Kant's analyses only point
in a particular direction. I develop those insights in some obvious ways
and show their bearing on current attempts to understand the psychology
of concepts.

Do We Employ Necessary and Sufficient Conditions?

Difficulties with the Classical View

The "classical" view of concepts6 assumes that the items in the extension
of a concept all share a manageably small set of salient features. These
features are listed in the concept, or mental representation, and that is
how items are grouped together under a concept.7 When an item is
presented, the possessor of a concept runs through the list. If the item
has all the features listed for the concept, then it is subsumed under the
concept. Wittgenstein attacked this notion from the extensional side. In
a well-known discussion, he argued there are no common properties
that all games, for example, have in common.8 Rosch's work undermined
the psychological assumptions of the necessary and sufficient conditions
model of concepts. If this story were correct, then for each item pre-
sented, a concept user would run through a list. The same list for the
same concept. What Rosch found, contrary to this model, is that dif-
208 KANT'S TRANSCENDENTAL PSYCHOLOGY
ferent items are treated differently. Some items, "typical" instances of
the concept, are classified immediately. Further, other items seem to
be classified under the concept, depending on their similarity to typical
items, rather than through the use of a checklist. This led her to propose
that the mental representation of concepts is not a list but a "prototype,"
that is, a typical example, some characteristic features, and an explicit
list of exceptions (e.g., "Penguins are birds, even though they cannot
fly").
Edward Smith and Douglas Medin offer a lengthy review of the lit-
erature spawned by Rosch's findings. They conclude with the obser-
vation that although the balance of the empirical evidence argues against
the presence of necessary and sufficient conditions, it does not really
establish one view as correct and rule out all the others.9 Medin has
also noted that the necessary and sufficient conditions model has sys-
tematic weaknesses, beyond empirical tests of classification (not all of
which are disconfirming).10 No one, experts included, seems able to list
defining features for concepts. Further, there appear to be unclear cases:
Is a radio a piece of furniture? If we apply concepts by employing
necessary and sufficient conditions, then all cases should be decidable."
On the other hand, prototype theories, or more generally, probabilistic
theories, also have problems. (Probabilistic views take concepts to be
organized around features that are typical of members, but not defining.)
Probabilistic theories imply that the only information that is stored in
mental representations concerns central tendencies. However, empirical
evidence suggests that the size of a classification, the variability of ex-
amples, and correlations among features are also used in classifying
instances.12
In sum, although the necessary and sufficient conditions view has
serious difficulties, the empirical results are not decisive. Smith and
Medin note that "there will likely be no crucial experiments or analyses
that will establish one view of concepts as correct and rule out others
irrevocably." They also predict that elements of the classical view will
appear in any complete theory of concepts.131 believe that Kant's anal-
yses of empirical concepts shed unanticipated light on this issue. For he
offers principled reasons for believing that the necessary and sufficient
conditions model cannot be correct, if empirical concepts are going to
fill their role in the acquisition of knowledge. Thus, he shows that the
classical view cannot be correct for the normal case. Besides its popu-
larity among philosophers and psychologists, a recent study by Mc-
Namara and Sternberg reveals that most lay people also hold a necessary
 Cognitive Constraints on Empirical Concepts 209

and sufficient conditions model (perhaps explaining why it is the classical

view).14 Given the strong intuitive appeal of this view, and the incon-
clusive empirical evidence, it is important to show why it must ultimately
be rejected. In addition, Kant's discussion provides positive clues about
what an adequate model of concepts might look like.

Kant on Concepts and Concept Application

Kant's general position on concepts is that they are rules. These rules
permit us to unite separately given perceptual materials under one label:
as regards its form, a concept is always something universal which serves
as a rule. The concept of body, for instance, serves as a rule for our
representation [Erkenntniss] of outer [objects], by unifying the diverse
elements which are thought in i t . . . . Thus the concept of body requires
in the perception of something before us, the content [of a cognitive state],
extension, and with this, impenetrability, form, and so f o r t h . . . [A106,
amended translation].

Despite the misleading implications of the example, Kant realizes that

we cannot explain how we apply the concept 'body', for example, by
noting that we first apply the concepts 'extension', 'impenetrability', and
so on, because we would be involved in an infinite regress (cf. A133/
B172).
In Chapter 615 I briefly described Kant's hypothesis about how we
apply concepts to objects. In opposition to the image-based accounts of
his predecessors, he argued that we do it through the use of schemata—
not images but rules for constructing images. These rules indicate the
sequence of operations that the perceptual system goes through in pro-
ducing images of instances of the concept (A141/B180; see also A103-
4, A77-79/B103-4). My chief concern is the question of necessary and
sufficient conditions, and so the fixed or changing nature of the rules
that enable us to apply concepts. Still, in passing, I will note two ad-
vantages of the schemata hypothesis. It avoids the regress problem,
because it explains how concepts are applied on the basis of perceptual,
not conceptual information. Second, it is plausible to think that what
matters in conceptualization is how perceptual data are treated by the
cognitive system, and not what they are like in themselves. Thus, the
key to applying concepts to objects might well be the perceptual pro-
cesses that occur when encountering instances.16
210 KANT'S TRANSCENDENTAL PSYCHOLOGY

Empirical Warrant and the Open-Ended Character

of Experience
Kant's support for schemata does not have any direct implications for
the necessary and sufficient conditions issue. Rules for producing images
could be treated as necessary and sufficient conditions as easily as feature
lists.17 In the Methodology and the Logic, however, he explains why all
definitional approaches must fail for empirical concepts. He often re-
peats the point that synthesis must precede analysis (B130, B133). We
cannot analyze what features are contained in our concepts before we
have synthesized diverse elements into a conceptual whole, before
we have taken the diversity provided by sense and united it in a rep-
resentation. Once we have done this for a number of instances, and
abstracted a rule for producing images, however—once we have a func-
tioning concept—it seems that we could draw up lists of features that
are crucial to different concepts by working backward from the instances.
The problem is that this cannot be done for empirical concepts: "Since
the synthesis of empirical concepts is not arbitrary but empirical and as
such can never be complete (for in experience ever new characteristics
of the concept can be discovered), empirical concepts cannot be de-
fined."18 Kant's simple point is that since experience is open-ended, we
could never work from a complete set of recognized instances.
The deeper point is that if empirical concepts are to play a role in
the generation of warranted belief, then they cannot be defined. Kant
agrees with his Empiricist predecessors that an empirical concept is one
that has been acquired from experience. We have the concept 'dog',
because, having encountered a number of dogs, we have abstracted,
among other things, a schema for generating perceptual images of dogs.
The crucial epistemological point about empirical concepts is that there
can be no question about their legitimate applicability to the world.19
Since we abstract similarities in image processing, for example, from
our dealings with instances, we cannot be deluding ourselves into be-
lieving that a class of similar objects exists (see A84/B116). This is not
to say that empirical concepts cannot be altered or even discarded.20 As
we encounter more and more instances, our grouping of perceptual
processes together as similar may change in subtle or dramatic ways.
Further, we may analyze the perceptual and other features that seem
common to instances, and discover that some concepts group objects
together that share only superficial or theoretically unimportant simi-
larities. These corrections to our original conceptual acquisitions are
made on the basis of (current) total evidence. For Kant empirical con-
 Cognitive Constraints on Empirical Concepts 211

cepts are acquired, refined, rejected, or maintained on the basis of

experience. That is why they are legitimate: They are warranted by the
very processes that produce and shape them. Since empirical concepts
derive their epistemological warrant by being malleable by experience
and since experience is open-ended, they cannot be denned.

When Should We Codify Our Concepts?

To see Kant's point consider when it might be appropriate for our
conceptual faculty to stop abstracting similarities from instances and
codify those similarities in definitions for future use in classifying new
instances. Since we continue to have experience, there can be no good
moment to stop. For whenever the process stops, it limits the evidence,
or at least favors some evidence over other evidence. Suppose that the
conceptual faculty shifts from the acquisition mode to the codifying
mode on an individual's twenty-first birthday. If the definition is used
strictly, then similar cases will be denied membership in the exten-
sion of the concept. Since those cases are potential evidence for the
adequacy of the classifications, however, the individual's system of
concepts would no longer be warranted on the basis of the total evi-
dence. The system of "empirical" concepts would lose its experien-
tial warrant.
To avoid this problem, let us assume that the conceptual faculty cod-
ifies concepts at 21 years but that other faculties can override the def-
initions, under certain circumstances. Standardly, definitions are used
to classify new instances. There is another faculty, however, that is able
to engage in conscious reflection on new cases and to use systematic
considerations about theoretically important similarities to remove old
concepts and replace them with more satisfactory concepts. This two-
faculty, or "liberal," definition model will avoid the epistemic error of
limiting the evidence. Further, this model seems to reflect the practice
of scientists in dealing with critical concepts in their disciplines. If we
probe a bit deeper, however, there are still epistemic problems with the
liberal definition model. First, we should note that it is evidentially
biased toward old cases. New cases can get a hearing, but they must
prove their importance as data, whereas during the acquisition of the
concept, any case was freely admitted to the data base. Favoring old
evidence has some epistemic virtues in the context of organized scientific
research. If a datum or a theory has been abroad in the scientific com-
munity for some time, then there is reason to hope that it has withstood
attempts to deny it. This is not the case for data that is old only in the
212 KANT'S TRANSCENDENTAL PSYCHOLOGY

lifetime of a particular individual, however. Yet, in the model, this is

the type of old data that is given pride of place.
The second epistemic shortcoming of the liberal definition model is
that new data are not admitted automatically, but only on the basis of
conscious reflection. Conscious reflection about the adequacy of our
concepts is epistemically virtuous, as Kant would be the first to stress.
For the perceptual similarities that tend to lead us to group instances
together may turn out to be theoretically unimportant. The demands
put on conscious reflection by this model seem too high, however. To
prevent our concepts from gradually slipping out of alignment with the
data experience provides, we would have to exercise constant vigilance
over our "automatic" classification system. If empirical concepts are to
be reflective of reality, because of the way they are produced, it is far
better to have a process that allows for automatic continuous updating.
Since we need such a process to explain how we manage to acquire
concepts at all, Kant assumed that it continues to mold our concepts to
fit the data that we continue to experience.

Implications for Necessary and Sufficient Conditions

Kant's analysis of what empirical concepts must be like for them to yield
knowledge has clear implications for the debate over necessary and
sufficient conditions. Although the particular spurs to study concepts
are varied, in general, concepts are important to cognitive science, be-
cause an adequate theory of concepts is necessary for a complete ex-
planation of feats of cognition. Further, the following assumptions about
concepts are ubiquitous among cognitive scientists21: Most (if not all)
concepts are acquired through experience, most concepts are reflective
of the experiences that give rise to them, concepts can be used to provide
more or less accurate descriptions of reality—and so contribute to cog-
nition—because they are derived from our encounters with reality. Since
these views are ubiquitous, they are almost never expressed. Still, I
believe that the vast majority of cognitive scientists would be shocked
if one of their number denied any of these claims, for more than a small
or isolated set of concepts. But if (empirical) concepts are reflective of
reality, because of the ways they are produced, then we should not
expect them to be represented by necessary and sufficient conditions.
A definition model implies rigidity in the face of new experience, but
the basic theoretical assumption about concepts is that they are molded
by experience. And, as Kant observes, in experience, ever new char-
 Cognitive Constraints on Empirical Concepts 213

acteristics of concepts may be discovered. An empirical concept realized

by necessary and sufficient conditions is not quite like a soluble fish,22
but it would not be very serviceable for cognition.
I should note that it is quite consistent with Kant's analysis that some
of our concepts appear to follow the necessary and sufficient conditions
model. This is to be expected if a concept has very few instances, if few
instances are encountered, or if there is very little variability across
instances. Concept users would appear to be applying necessary and
sufficient conditions. Further, Kant is explicit that if a science is ever
completed (or believed to be so), then it could have definitions. Hence,
the common view that there are definitions in various branches of math-
ematics would be compatible with this aspect of his philosophy. Thus,
Kant's analysis suggests that the data will be somewhat mixed, even
though necessary and sufficient conditions would be an inappropriate
way to represent empirical concepts, for the normal case.

Further Implications
Kant's attack on definitions has two further implications for work on
concepts. First, concept learning would turn out to be something of a
fiction. Insofar as this process is standardly understood as taking place
over a relatively fixed span of time, before which subjects lack the
concept and after which they have it, concept learning does not take
place. To meet the requirements for empirically warranted belief, con-
cept learning must be, in the educational cliche, a lifelong experience.
The second implication is even more radical. If cognition is possible
only because our conceptual repertoire is adaptive to the changing evi-
dence of the environment, then cognitive scientists may not want to
appeal to concepts at all, but rather to something like conceptual worms
that continually evolve through time. The current confusion about con-
cepts may be caused, in part, by a mistaken attempt to provide a static
account of a phenomenon whose essence is the ability to change. Kant's
analyses of the prerequisites for cognition reveal that to be the inherent
and theoretically fatal weakness of the necessary and sufficient condi-
tions model. Whatever its intuitive appeal, it presents concepts as com-
pleted, rigid mental representations. Once the open-ended character of
experience is fully appreciated, it becomes apparent that this model is
inconsistent with the basic theoretical assumption about empirical con-
cepts—they are central to cognition because they are molded by
experience.
214 KANT'S TRANSCENDENTAL PSYCHOLOGY

Empiricism and "Original Sim"

Quinean Empiricism
If concepts are not represented in us by a list of necessary and sufficient
features for category membership, what mental mechanisms do account
for our ability to classify under concepts? Many discussions begin with
the framework offered by Quine in "Natural Kinds."23 In the beginning,
classification occurs via perceptual similarity. Our sensory systems have
innate quality spaces that lead us to experience certain stimulations as
similar. Paradigmatically, our visual system has an innate standard of
similarity for colors, so that, for example, different shades of yellow
will strike us as quite similar. Quine hastens to add that this appeal to
innate structures in no way abandons the tenets of Empiricism. We can
measure quality spaces empirically, by determining the "generaliza-
tions" made by animals conditioned to respond to stimuli of certain
colors. And since quality spaces are required for all learning, they must
be innate.24 Further, Quine's Empiricist scruples are clearly evident in
his account of the second stage of concept acquisition. We start with a
color-slanted quality space, but science enables us to rise above it:
He [man] has risen above it by developing modified systems of kinds,
hence modified similarity standards for scientific purposes. By the trial-
and-error process of theorizing he has regrouped things into new kinds
which prove to lend themselves to many inductions better than tne old
[my emphasis].25

Not surprisingly, this account posits very little mental equipment un-
derlying the use of concepts.26 Innate quality spaces enable children to
get started; through the undefined capacity for ingenuity and trial-and-
error learning that produces scientific theories, the race acquires more
useful categories, and teaches them to novices.27
The "original sim" account, as Frank Keil labels it, has been criticized
in several recent empirical studies.28 Contrary to Quine's speculation,
children—even very young children—do not appear to group objects
together just on the basis of qualitative similarity. Perhaps the most
dramatic demonstrations are Gelman and Markman's studies with 3-
and 4-year-olds.29 The design of the study pitted appearance against
category membership. (An example of the test materials is illustrated
in Figure 8.1.) Having been told that the target (e.g., the cat) had a
particular property, children were asked which other items would have
the same property: items in the same category but with different ap-
 Cognitive Constraints on Empirical Concepts 215

Figure 8.1 One set of pictures used by Gelman and Markman to test category
membership versus similar appearance. (Reproduced from Child Development
58 (1987): 1532-41, by permission.)

pearances or items with similar appearances but in different categories.

Both 3- and 4-year-olds infer that the same property will be present
significantly more for the same category-different appearance items than
for different category-same appearance items. Further, this pattern held
for each of the 10 examples used in the study.
These results suggest that there is never a time when perceptual sim-
ilarity is the sole mechanism that leads children to group objects to-
gether.30 Gelman and Markman's data are consistent with findings of
other researchers. Susan Carey has argued that children's use of concepts
is tied to theories.31 Although they disagree over important details,
Frank Keil's work also implies that children's concepts are supported
by implicit theories.32 This might suggest that Quine was right about the
two elements in concept acquisition—similarity plus instruction in sci-
ence—but wrong about how early instruction begins.33
However, the second part of the similarity-corrected-by-science hy-
216 KANT'S TRANSCENDENTAL PSYCHOLOGY

pothesis is also at odds with recent empirical work. Douglas Medin and
his colleagues have observed that, in experiments on categories,
"[pjeople act as if things (e.g., objects) have essences or underlying
natures that make them the thing that they are."34 This would explain
why some properties are treated as more important for category mem-
bership than others. Further, it might account for the intuitive appeal
of the classical position. Medin notes, however, that although positing
an "essentialist heuristic" explains much of the data on concepts, it is
bad metaphysics [and bad science]. If concept users rely on an essentialist
heuristic, it is not because they have internalized the wisdom of science.

Current Directions
Without pretending to summarize the entire literature on concepts, let
me bring together some leading ideas about the factors underlying con-
cept use. Perceptual similarity is important, even though it can be over-
ruled by other considerations. Knowledge about the world, couched in
explicit or implicit theories, also seems crucial. Although the theoretical
knowledge underlying concept use is sometimes available to subjects,
often it is not. And in the case of children, it is unreasonable to attribute
explicit theories of physics or biology.35 Perhaps concept users employ
an essentialist heuristic that enables them to form coherent concepts.
In general, concepts appear to have a great deal more structure than
had previously been realized. Hence, it is implausible to represent them
simply as lists of correlated features.36 The features important to clas-
sification under a concept are themselves interrelated. Further, some
seem more fundamental, and some occur at different levels of abstrac-
tion or specificity. Both the essentialist heuristic and theory-driven
models are attractive, because they offer ways of explaining the relations
among features in a concept.37
Medin also favors the essentialist heuristic because it suggests how
similarity-based and knowledged-based mechanisms might be inte-
grated. Assuming that perceptual similarity is a clue to deeper similar-
ities, then similarity together with the essentialist heuristic could point
the way toward classifications that can later be rooted in knowledge.38
Medin's goal is a coherent account of concepts that is not itself just a
list of factors: similarity + theories + central features + different levels
of features + heuristics (e.g., essentialism). I shall argue that Kant's
reflections about how concepts and theories must be structured for
knowledge to be possible provide a coherent framework for understand-
ing these empirical results. Negatively, Kant argued against definitions
 Cognitive Constraints on Empirical Concepts 217

for empirical concepts. But he also believed that the demands of knowl-
edge had positive implications for the structure of concepts.

Concepts and Reasoning

The Task of Inference

The fundamental cognitive fact about concepts is that they are service-
able only if they so partition objects that we can engage in reasonably
successful inductions. This is clear in the concept literature, where dis-
cussions often focus on inferences.39 Approximately 300 pages into the
Critique of Pure Reason, Kant turns to the topic of reason, the faculty
that enables us to make inferences. Previously, we have considered his
analyses of relatively basic cognitive tasks. Here his quarry is quite
sophisticated: What kind of faculty of reason must we have for it to be
capable of producing knowledge through inference?
Contemporary discussions recognize the importance of inference to
knowledge but rarely consider the problem in its full generality. Instead,
they tend to concentrate on experimental validation of particular models
of inference. By contrast, Kant directs all his energies to the abstract
analysis of the task itself. He pursues his inquiries into the requirements
for making inferences in the Ideal of Reason, the Regulative Use of the
Ideas of Reason, and the introductions to the Critique of Judgment.
This capacity turns out to have surprising implications for the structure
of our concepts, and even for the structure of nature itself—or for any
nature that could be known by us through our powers of intuition,
understanding, and reason.

Analytic and Synthetic Approaches to Concepts

Kant can formulate the problem of the adequacy of concepts for in-
ductive inference with particular clarity and starkness, because he un-
derstands representations, concepts, and judgments synthetically. We
synthesize concepts and judgments on the basis of materials supplied
by experience. Why should there be any connection from one repre-
sentation or concept or judgment to another? Each is constructed on
the basis of data supplied by different objects. So how is even simple
instance induction possible? How can we know, for example, that the
dog we now confront is unlikely to attack us unless threatened? The
218 KANT'S TRANSCENDENTAL PSYCHOLOGY
difficulty of this problem is missed if concepts and judgments are under-
stood analytically.
This is a common failing of Leibniz and many contemporary cognitive
scientists. For Leibniz, all judgments are explicative: The predicate of
the judgment merely displays what is implicitly contained in the subject
concept. Judgments are made by comparing representations or percep-
tions to determine if the predicate concept is contained in the subject
concept. Now let us consider something apparently very different: the
standard paradigm for concept learning experiments. Here subjects or
computer programs are "taught" a new concept on basis of "experi-
ence." So a person or program is given a new term—"flurg" (to borrow
Jerry Fodor's example40—and set of instances and foils. In the case of
the computer program, the salient features of the examples are de-
scribed, and the program has to figure out which features must be present
or absent for the label to be correctly applied.41 Similarly, human sub-
jects are told which of the examples are "flurg" or "nonflurg," and they
must determine the criterial properties for this concept, say being green
or square. To see the affinities with Leibniz's view, consider what would
happen after this type of concept learning. A new flurg is encountered.
If it is correctly labeled a "flurg," it must be either green or square. So,
from the fact that it falls under the concept 'flurg', we may infer that
it, like other flurgs, is either green or square. But this inference is
vacuous. For on this model of concept learning and concept possession,
unless subjects recognized it to be either green or square, they would
not label it a "flurg." Although this position is not as strong as Leibniz's
claim that all judgments are analytic, it does seem to follow that what
is inferred about objects, through the concepts they fall under, are only
properties implicit in the concept.
In a note to the First Introduction to the Critique of Judgment, Kant
points out that empirical cognitions
cannot be abstracted analytically, from a simple comparison of perceptions
(as is commonly supposed), because the union of two distinct perceptions
by means of the concept of an object (so as to yield knowledge of it) is
a synthesis.42
To produce knowledge a judgment must be a function of properties that
are combined in the object.43 It can never arise merely by comparing
one concept with another. When the synthetic character of concepts and
judgments is appreciated, however, the problem of instance induction
becomes acute. We apply the label "X" to an object and observe its
properties. Then we consider a new X and presume that it will share
 Cognitive Constraints on Empirical Concepts 219

these properties, not because they are implicit in the concept 'X', but
simply because it is an X. How is this kind of inference, so basic to
knowledge, possible? What must concepts be like for it to be possible?

How to Carve Nature at the Joints

The standard answer is that concepts must "carve nature at the joints."
They must group things together under one label that share deep sim-
ilarities. Then our inductions will be successful. Unfortunately, like the
proverbial market advice to buy low and sell high, this bromide provides
no clues about how to find a substantive answer. Kant tries to provide
a real answer by considering the requirements of inference itself. In-
ferences require principles. To infer that the dog before us will not
attack if left alone, we must tacitly or explicitly appeal to a higher
principle. Previously encountered animals have behaved in this way, so
we assume that our judgments about them, and the judgment we are
about to make about the dog, fall under a higher principle: In general,
animals do not attack unless threatened. In the absence of some such
principle, we would not be making an inference, but merely a hopeful
guess. The appeal to principles raises the puzzle about concept-based
inference in a different guise:
But that objects in themselves, the very nature of things, should stand
under principles, and should be determined according to mere concepts,
is a demand which, if not impossible, is at least quite contrary to common
sense [A301-32/B358].
Despite the misleading hint of "objects in themselves," Kant does
not solve this problem by appeal to the phenomenal—noumenal distinc-
tion, but in quite a novel fashion. He considers the demands that reason's
need for higher principles puts on concepts and on nature itself. Nature
must have a certain kind of structure: Its properties must form a hier-
archy. "For we can conclude from the universal to the particular, only
in so far as universal properties are ascribed to things as being the
foundation upon which the particular properties rest." (A652/B680, my
emphasis). That is, inference will be possible only if it is mediated by
principles indicating that some properties are the grounds or foundation
of other properties.
Although Kant's position is clear, the argument for it seems vulnerable
to straightforward counterexample. Seemingly, we can and do make
inferences about new instances merely on the basis of correlations be-
tween properties—even when there is no relation of ground and de-
220 KANT'S TRANSCENDENTAL PSYCHOLOGY
pendence between the properties. To borrow a well-known example
from Hempel, the presence of Koplick spots inside the mouth enables
a physician to infer that the patient will soon run a fever, although there
is no presumption that the spots are the ground of the fever.44 When
we reflect on such counterexamples, however, we can see the deep point
behind Kant's claim. Although there is no dependency relation between
the fever and the Koplick spots, the spots can be used to infer the fever
only because we presume that they are indicative of some further con-
dition, which is the ground of both fever and spots. Kant's statement
of his analysis is, as often, excessively cryptic. What he means is that
inference must always be mediated by a principle, which is such that
either the principle itself, or some further principle that supports it,
appeals to properties that are related as ground and dependent.
There remain only two possible classes of counterexample. Through
exhaustive enumeration, we know that everything that has the property
X also has the property Y. In such cases, the inference is not inductive
or ampliative, however, because we would already know of the present
instance that it has both properties. The second possibility is that some
inferences are mediated by principles stating correlations that are taken
to be brute, even in the absence of exhaustive enumeration. "Brute" is
meant to signify that we have no reason to believe that anything mediates
the correlation between the two properties. The 1980 election provided
a classic example. It was noted that from 1820 through 1960, every
president elected at the even 20 years died in office. Kant's argument
is that this kind of pattern cannot license an inference about new cases,
because it provides no justification for the inference. Ex hypothesi,
previous correlations are not taken to be evidence of any connection
between the properties, so nothing but blind chance could be offered
in support of the inference. That is, nothing supports it.
Putting the matter positively, instance induction requires the use of
a principle that is taken to indicate a relation of ground and dependence
among some properties. This does not mean that any inference so me-
diated produces knowledge. Given the popularity of astrology columns,
many people seem to believe that there is a connection between the
position of the stars and the course of human events. People who use
the stars to make inferences are not justified, even though they take
whatever correlations they are able to find as signs of connection. Em-
ploying principles that are regarded as indicative of some relation of
dependency among properties is a minimal condition. Without it, Kant
maintains, we are not engaging in the task of inferring from previous
experience at all.
 Cognitive Constraints on Empirical Concepts 221
He expresses this situation by saying that reason puts a "demand" on
nature. It must have not joints, but hierarchical relations, with a few
basic properties supporting less fundamental properties, which in turn
support even more derivative properties. This "demand" does not es-
tablish any metaphysical claims about the actual structure of nature. As
Henry Allison has argued in detail, Kant would be the last philosopher
to infer from the fact that nature must be hierarchical for us to achieve
knowledge, that all nature is hierarchically arranged.45 His idea is,
rather, that we must approach nature in the expectation that it has a
hierarchical structure. Given the faculties we bring to the task of cog-
nition, that is the only way that we could acquire knowledge. This
approach to nature puts a demand on empirical concepts. "[W]e must
seek for a certain systematic unity of all possible empirical concepts, in
so far as they can be deduced from higher and more general concepts
. . . " (A652/B680). We must employ concepts that are related as genera
and species.

The Structure of Concepts

Although the relation of genera and species is often mentioned as a
criterion for the adequacy of empirical concepts, Kant never fully elab-
orates. Still, we can fill in some details from the passages in the Reg-
ulative Employment of the Ideas just cited and from two discussions in
the Logic. The Logic notes that when concepts are related as genus and
species, the higher concept contains the "cognitive ground of the
lower."46 This, I take it, is how the relation of genus and species between
concepts reflects the relation of ground and dependence among prop-
erties in nature. The genus concept indicates the presence of properties
that are the foundation of the properties indicated by the species con-
cept. Notice that the relation of genus and species between concepts is
much stronger than the frequently noted relation of superordination and
subordination. "Furniture" is a superordinate concept with respect to
"chair," but it could not be considered the genus. In describing a chair
as a piece of furniture, we do not indicate any properties that underlie
(or could be used to explain) its chair properties.
The Logic also offers a discussion of the structure of individual con-
cepts.47 Concepts have both an external and an internal [cognitive] use.
In the external use, concepts are used to distinguish objects. For this
purpose the characteristics associated with a concept must be sufficient
to distinguish one class of things from another.48 Concepts also have an
internal use, however, which "consists in derivation in order to cognize
222 KANT'S TRANSCENDENTAL PSYCHOLOGY

the thing itself through characteristics as grounds of its cognition."49

The characteristics can be coordinate with each other, or constitute a
[partial] series of subordinate characteristics, which ascend to ever
higher genera and/or descend to ever lower species.501 take Kant's point
about derivation to be that, when we do have ascending series of sub-
ordinate characteristics for the coordinate characteristics, we have char-
acteristics that indicate the relatively more fundamental properties that
ground the coordinate characteristics. Hence, we have a concept that
contains within itself information about the sorts of properties that un-
derlie more superficial properties associated with the concept.
If characteristics are sufficient for coordination and subordination,
then Kant describes the complex of characteristics as the "essence" of
the concept. He hastily adds that this is not a real essence, however. It
only indicates dependency relations among properties known to be as-
sociated with the concept. To get the real essence, we would have to
know everything whatsoever that belongs to the existence of the objects51
and, for reasons we have already seen, this is impossible. Thus, "es-
sence" is something of a misnomer. This complex of characteristics is
not fixed and does not provide necessary and sufficient conditions for
application of the concept.52
The relation of genera and species thus holds between concepts and
between characteristics [concepts] within individual concepts. This dual-
ity does not indicate inconsistency, but a substantive assumption. The
"external" relations of a concept to other concepts are mirrored in the
internal structure of the concept. Concepts are adequate to the demands
of inference only if they are related as genera and species, and the
internal structure of a concept mirrors its external relations. Although
the inference is never drawn, it follows that cognitively adequate indi-
vidual concepts must have a complex inner structure. They cannot be
lists of unrelated features. The mental representation of an empirical
concept must have a rich structure that captures dependency relations
among the different levels of properties associated with the concept.
Thus, the concept of an X must be an incomplete theory of X's.

Examples of Dependency Relations

Kant explicates the relation of genera and species in terms of dependency
relations. The genus concept indicates properties that are the founda-
tions of properties associated with the species concept. But what are
the dependency relations? He can provide no exhaustive specification,
for the obvious reason. Until the "end of science," no one can describe
 Cognitive Constraints on Empirical Concepts 223
all the various ways in which properties depend on more fundamental
properties. Still, despite his normal disdain for examples, we are given
several illustrations from eighteenth-century science. He considers the
hypothesis that all faculties of the human mind stem from a single fun-
damental power (A649/B677). These faculties have so many different
effects that this hypothesis would not even have arisen, but for the fact
that reason
requires [us t o ] . . . reduce, so far as may be possible, this seeming diver-
sity, by comparing these with one another and detecting their hidden
identity. We have to enquire whether imagination combined with con-
sciousness may not be the same thing as memory, wit, power of discrim-
ination, and perhaps even identical with understanding and reason [A649/
B677].

In a marginal note to the First Introduction to the Critique of Judgment,

he asks rhetorically whether Linnaeus could have hoped to construct a
system of nature had he assumed that each thing had an individual
essence, rather than being capable of belonging to a class, "which could
be brought under genus-species concepts."53
In both these examples, part of Kant's goal seems to be to provide
empirical confirmation for the thesis he has already defended through
his analysis of the task of inference. For inference to be possible, we
must seek principles that relate properties as ground and dependent—
and here are two examples where we see this demand of reason in action.
In the published discussion of psychology, he elaborates how the ground-
ing is supposed to work: Derivative powers will be resolved into more
fundamental powers, until all the powers can perhaps be understood as
one power operating under different conditions. That is, the relation of
ground and dependent is the relation of constituents to the whole. A
complex faculty has the effects that it does, because it is made up of
more basic faculties. The same relation is highlighted in Kant's more
extended discussion from chemistry. We do not encounter instances of
pure earth, pure air, and so forth. Nevertheless, in chemistry, the chem-
ical interactions of bodies are explained by reducing all kinds of matter
to more basic substances (earths, salts and inflammable substances,
water and air). Again, properties of wholes will be a function of the
properties of their microscopic constituents (A646/B674ff).
The Metaphysical Foundations of Natural Science offers a further il-
lustration, although not in the context of a discussion of genera and
species. This is Kant's well-known argument that solidity is not a fun-
damental power of matter. Solidity is supposed to account for the fact
224 KANT'S TRANSCENDENTAL PSYCHOLOGY
that objects can keep other objects from entering their current spatial
position. This is possible, he maintains, only if all parts of matter have
repulsive forces.54 Kant's belief that dependency relations include the
dependence of observable properties on properties of, or forces gov-
erning, constituents is hardly surprising in the historical context of trying
to extend Newton's ideas to other domains of physical science.
For reasons we have already seen, a complete list of dependency
relations cannot be given. Instead, we have some useful pointers. Con-
cepts display the systematic unity that will permit a hierarchy of genera
and species when a genus concept indicates (1) properties of, or (2)
forces governing, constituents that explain the powers or properties of
the phenomenon indicated by the species concept. This, then, is the
"demand" that reason puts on empirical concepts. To be adequate to
cognition, they must exemplify this type of systematic interconnection.
It is tempting to think that reason has no more ability to enforce this
demand than its "demand" for structure in nature. Kant encourages this
view by describing the systematizing demand as "subjective" and as "an
ideal." However, it is subjective only in the sense that it does not permit
the inference to objective hierarchies in nature (e.g., A645/B673); it is
an ideal because it aims to bring as much unity to concepts as possible,
without knowing how much is possible (A647/B675). The demand of
reason is not subjective in the sense of being merely apparent or arbi-
trary, and it is not ideal in the sense of being merely normative. It is
psychologically real.55 "Reason... has a real use, since it contains within
itself the source of certain concepts and principles [viz., seek systematic
unity in concepts]" (A299/B355). For Kant it is not merely a normative
doctrine that sound inferential practice requires appeal to principles that
indicate a relation of ground and dependence. Our inferential successes
and the development of scientific theories are evidence that we have a
faculty that seeks concepts that are systematically interrelated.
The false impression that the demand of reason is not psychologically
real is reinforced by the fact that Kant offers no suggestions about how
such a demand might be effected. Here he stands by his prefatory pledge
not to speculate about how something that is necessary for knowledge
actually comes about (Axvii). For my purposes, however, this issue is
critical. Although Kant believes in the psychological reality of the de-
mand for system on the basis of successful inferences and scientific
practice, the case needs to be strengthened. In the absence of any hints
about how such a demand could be implemented, its psychological real-
ity will seem less plausible. The next section tries to fill this gap, by
speculating about heuristics that might carry out reason's demand for
 Cognitive Constraints on Empirical Concepts 225

system. Reflecting on possible heuristics also serves a second purpose.

It enables us to see some of the implications of Kant's analysis of em-
pirically adequate concepts for current research on concepts.

Implementing the Demand of Reason

Reason demands that we use concepts that have a systematic unity. Or,
less metaphorically, we can acquire knowledge through inferences only
if we use such concepts. As already noted, however, empirically ade-
quate concepts are also based on experience. They are constructed and
modified on the basis of encounters with objects. So the concepts that
we acquire through experience must be subjected to further testing to
determine if they can be part of a unified system of concepts. However—
and this is the hard part—the testing must precede anything like a
complete scientific understanding of the dependency relations among
the properties of these objects. That is, we must try to figure out if a
concept can fit into a systematic description of the hierarchical structure
of nature, without knowing what that structure is (if it has one). Since
the end state is not known, it is, to say the least, highly unlikely that
we can find an algorithm to determine the systematic potential of con-
cepts. Instead, the hope is to find heuristics that are somewhat reliable
in leading us toward more systematic concepts.
Heuristics can be conscious or unconscious. Kant's discussion of gen-
era and species is intended, in part, to be an epistemic recommendation.
In the practice of science, we should deliberately set out to find theories
with hierarchical structure. However, he describes the demand for sys-
tem as "natural tendency" of reason (A643/B670). This suggests that
he believes that we are prone to look for system, even without trying.
The heuristics I consider could have conscious and unconscious versions.
Once science has become a conscious activity, it is very easy to see how
individuals could pursue the goal of systematic unity in concepts. They
could just follow Kant's advice! The earlier stage seems more problem-
atic. Without knowing anything about nature, including our own cog-
nitive powers, how can individuals or the race fasten on concepts that
are sufficient to get scientific theorizing started? How can they do this
in the face of the enormous variety of forms in nature (e.g., A651/
B679)? To solve the problem of early conceptualization, we need to find
heuristics that can be unconscious.
The best and most obvious heuristics for finding concepts that might
fit into a complete hierarchical system are those that favor concepts that
226 KANT'S TRANSCENDENTAL PSYCHOLOGY

exemplify at least a small piece of this pattern. For example, we might

favor concepts that indicate a cluster of correlated attributes. (Here I
use "attribute" as the generic term for property, power, or relation.)
So if items that bark also regularly engage in a distinctive form of
locomotion and also regularly behave toward people in a special way,
then we group these items under a concept. Such clusters are good
evidence of an underlying attribute or attributes on which members of
the cluster depend. In the early stages of conceptualization, this heuristic
could be unconscious, however. Leaving off the conscious rationale, the
heuristic is just: Fasten on concepts that indicate clusters of correlated
attributes.56 A slightly different aspect of the pattern would be captured
by a heuristic that favored concepts that indicate a dependency relation
among attributes. So if items get bigger because they eat, then they
should also be grouped together under a concept.
As Medin noted in defense of the "essentialist" heuristic, superficial
similarity can often be a clue to similarity of more fundamental attri-
butes. In fact, perceptual similarity can be understood as a special case
of the first heuristic. If we are prone to assign a concept to items that
share clusters of correlated attributes, then we will also classify objects
together that share a number of perceptually similar features. Thus, if
Kant is right, we would expect perceptual similarity to play an important
role in conceptualization, as it does. But this would not be the role that
Quine envisioned for "original similarity." The role is not original, but
derivative. Perceptual similarity would only be used in categorization,
because it can be indicative of a rich structure of dependency relations
among the attributes indicated by the concept. It would have no inde-
pendent status, but would merely be one of a bag of tricks for detecting
complex interrelations among properties. Both Quine and a contem-
porary Kantian will correctly predict that subjects will group things
together that look alike. The latter will add, however, that this heuristic
will be overriden where other indicators suggest that dependency re-
lations are to be found elsewhere. I take Gelman and Markman's results
that, in the face of richer sources of evidence (but not scientific theories),
even 3-year-olds will abandon perceptual similarity to be some confir-
mation of the Kantian position.
Barbara Landau, Linda Smith, and Susan Jones have recently pub-
lished some tantalizing results about the importance of shape in learning
lexical items.57 Adults, 2-year-olds, and 3-year-olds favor shape over
size and texture in grouping items together that "go together" and in
extending novel count nouns to new instances. The bias is greatest in
adults, and increases between 2 and 3 years of age. These authors spec-
 Cognitive Constraints on Empirical Concepts 227
ulate that shape bias develops in the process of learning new words
between the ages of 2 and 3, because "the earliest learned words, basic
level concrete nouns, do in fact largely divide the world according to
shape." Since it is useful here, children generalize to later word learning
tasks. There are several obvious objections to this hypothesis. Two-
year-olds already know many basic nouns.58 More importantly, it is not
clear how the hypothesis explains the stronger bias of adults, who are
no longer learning basic nouns, but abstract ones. I offer a different
hypothesis. Grouping objects together by shape is a heuristic that serves
the systematizing demand of reason. Shape would be a useful heuristic
for two different reasons. Many other attributes of objects depend on
their shapes. To take a relevant example, balls can be rolled, because
they are round. So shape can be viewed as a special case of the second
heuristic. Shape can also be understood as a cluster of more local geo-
metric properties of objects, however. For example, the distinctive
shapes of cat ears, faces, and paws are part of the overall feline figure;
but they can easily be regarded as separate but correlated attributes.
That is, using shape exemplifies first heuristic. Thus, similarity in shape
would be a good predictor of classifications that will have a rich structure
of dependency relations among the attributes of the objects in the clas-
sification, and so a good predictor of concepts with high systematic
potential.
The preceding arguments are not intended to show that Kant's claims
for the systematizing demands of reason offer the best explanation of
the empirical data. In this section I have only tried to establish the
plausibility of regarding this systematizing demand as psychologically
real. Given these four heuristics—correlated attributes, attributes re-
lated by dependency, perceptual similarity, and shape—we can see how
the demand of reason could be implemented. Extracting cognitively
adequate concepts from the abundance of nature, in the absence of
scientific understanding, is not easy, but seemingly we can get started
without the assistance of any magical abilities. In the final section I try
to show that Kant's idea of a "demand" for system provides greater
theoretical insight into recent empirical results than the available con-
temporary proposals.

Theoretical and Experimental Implications

As already noted, recent interest in concepts has produced a plethora

of experimental results. Perceptual similarity plays an important role in
228 KANT'S TRANSCENDENTAL PSYCHOLOGY
classifying items under concepts, but not the dominant role sometimes
attributed to it. Adults' and even young children's use of concepts seems
to imply the presence of implicit or explicit theoretical knowledge. In
general, concepts seem to have a great deal of structure. They indicate
attributes that are interrelated and that exhibit different levels of ab-
stractness or specificity. Several hypotheses about the mental equipment
underlying concept use have been proposed to explain these results, the
most popular being the theory theory. Concepts are supported by explicit
or implicit theories.
As Murphy and Medin observe, however, to explain a feature of
concepts by saying that concepts are backed by theories is not very
satisfactory, unless one can say what a theory is.59 Further, it is clear
that the implicit theories attributed to children on this model are not
very like scientific theories, so a question arises about how minimal they
can be and still be theories. Frank Keil, who defends the importance
of theoretical knowledge, has also noted that there appear to be theory-
independent principles, such as "properties have purposes" and "prop-
erties reflect essences," which guide us in classifying objects together
under a concept.60 Finally, this approach raises an obvious question that
is not very easy to answer: How many implicit theories are there?
Medin's proposal of an essentialist heuristic is simpler, because one
heuristic is applied across a variety of domains. Essentialism has vices
of its own, however. Kant notes that a belief in individual essences
would thwart efforts to develop a system of concepts based on genera
and species that is necessary to concept-based inference.61 Inference
only works when there are similarities across individuals. Even if es-
sences are located in kinds rather than in individuals, essentialism would
still block higher-level inferences across kinds. Thus, an essentialist heu-
ristic would be counterproductive in trying to develop a set of concepts
that permits the largest number of inferences about new instances. It is
bad metaphysics, as Medin points out, and a bad inferential strategy.62
Kant employs a very different perspective to answer the two funda-
mental questions about concepts: What kinds of concepts do we have?
What mental equipment underlies our ability to use such concepts? For
concepts to be adequate to cognition, in particular to the demands of
inference, they must have a complex inner structure that indicates de-
pendency relations among attributes. Given the richness of forms in
nature, we can fasten on adequate concepts only if we actively seek
concepts with high systematic potential, only if we "approach nature in
order to be taught by it" (Bxiii) by "projecting" a "systematic unity"
 Cognitive Constraints on Empirical Concepts 229

on nature (A647/B675). Hence, we must have psychological mecha-

nisms—the "demand" of "reason" is merely a place holder—that enable
us to seek adequate concepts. On Kant's model, concepts require at
least the following mental equipment. Individual concepts are repre-
sented by complexly interrelated sets of attributes, which amount to
partial theories of the dependency relations among objects, properties,
and forces; and the whole process is supported by various heuristics.
We can fill in some further details by recalling his objections to denning
empirical concepts. The representations underlying concepts are con-
stantly updated by the application of conscious and unconscious system-
atizing heuristics to the individual's expanding body of data.
This model provides a coherent framework for understanding a variety
of empirical results. As noted earlier, both the theory theory and the
essentialist heuristic are appealing, because they seem to provide some
account of the surprising amount of structure that has been discovered
in concepts. This is a virtue shared by Kant's model. Concepts must
have significant internal structure to be cognitively adequate, and we
have various heuristics for acquiring such concepts. Thus, the model
retrodicts the discovery of structure in concepts. It is simpler and more
plausible than the theory model, however. We do not need a number
of implicit theories. More importantly, perhaps, on this picture, we do
not attribute theories to children and lay adults. The heuristics subserving
the demand of reason suggest, instead, that we will adopt concepts that
indicate some dependency relations among attributes. Concepts would
be backed only by theory fragments. Besides the problems with essen-
tialism already noted, Kant's model has the advantage of providing a
principled explanation for the empirical results. Given his analysis of
what concepts must be like for inference to work at all, we can under-
stand why concepts should be expected to have the structure that they
do. As noted in passing, the model also offers a unified explanation for
the use of perceptual similarity, shape, and [partial] theories in deter-
mining classifications.
In addition to theoretical insight, Kant's analysis of what cognitively
adequate concepts must be like provides a clear directive for research
on concepts. If we are Quineans, then we will try to understand the
processes of concept formation by investigating innate quality spaces.
Theory theorists will try to uncover the implicit theories that stand
behind conceptual practices. An essentialist will look not for theories
but for essences. But if we take Kant seriously, then we should try to
discover general-purpose heuristics that lead to potentially systematic
230 KANT'S TRANSCENDENTAL PSYCHOLOGY

concepts. What are they? How good are they? Do they show a clear
developmental pattern? Do they, as Kant believed, lead to metaphysical
excesses in the pursuit of complete scientific understanding?63

Conclusion

I have presented only one example of how Kant's attempts to understand

the prerequisites of cognition can contribute to current debates, but I
think it is a convincing one. Through reflecting on what concepts must
be like to serve cognition, Kant develops a model of empirical concepts
that can give needed order to chaotic empirical results. In retrospect, I
hope that it is no longer surprising to hold that a great theoretician who
devoted enormous intellectual efforts to determining what cognition
requires of our mental faculties might have had some ideas that are
useful to cognitive science. The history of philosophy can be a philo-
sophical discipline only if the ideas of the past were so powerful that
they can be a continuing source of insight to later theorists. In denying
transcendental psychology, we run the risk of turning an intricate and
vital study into a museum piece. The great book is much more re-
warding—and makes much better sense—when it is read intact.
 NOTES

Chapter 1
1. P. F. Strawson, The Bounds of Sense (London: Methuen, 1966). The
citations are from pages 32 and 16.
2. Ibid., p. 19.
3. Ibid., p. 16.
4. Ibid., p. 96. After presenting interpretations of various Kantian doctrines
without recourse to the psychological aspects of the arguments, Strawson gives
an interpretation and criticism of transcendental psychology in Part Four. I
discuss this argument at the end of Chapter 5.
5. Three important exceptions are Robert Paul Wolff, Kant's Theory of
Mental Activity (Cambridge, Mass.: Harvard University Press, 1963), Onora
O'Neil, "Transcendental Synthesis and Developmental Psychology," Kant-Stu-
dien 75 (1984): 149-67, and Ralf Meerbote, "Kant's Functionalism," in J. C.
Smith, ed., Historical Foundations of Cognitve Science (Dordrecht, Holland:
Reidel, 1989). One piece of transcendental psychology that has received some
attention is Kant's theory of imagination. For references and a very interesting
recent discussion, see J. M. Young, "Kant's View of Imagination," Kant-Studien
79 (1988): 140-64.
6. Jonathan Bennett, Kant's Analytic (Cambridge: Cambridge University
Press, 1966), pp. 56, lllff. I discuss another aspect of Bennett's antipsychol-
ogism in Chapter 6, note 26.
7. Paul Guyer, Kant and the Claims of Knowledge (New York: Cambridge
University Press, 1987), p. 374. Henry Allison, Kant's Transcendental Idealism:
An Interpretation and Defense (New Haven: Yale University Press, 1983). I
discuss one of these efforts at depsychologizing in "The 'Logical' Reading of
Apperception" section in Chapter 4.
8. I discuss Henrich's attempt to reconstruct the deduction of the categories
while eschewing psychology in "Henrich's Antipsychological Reading" in Chap-
ter 6.
9. Norman Kemp Smith, A Commentary to Kant's "Critique of Pure Reason"
(New York: Humanities Press, 1962), p. 51. Cohen, Caird, and Riehl were the
offenders.

231
232 NOTES
10. Ibid. For the pledge, see p. 270; the account follows. I consider the issue
of the noumenal versus the phenomenal self below and again at the end of
Chapter 5.
11. H. J. Paton, Kant's Metaphysics of Experience, 2 vols. (New York: Hu-
manities Press, 1965), 572ff.
12. W. H. Walsh, "Philosophy and Psychology in Kant's Critique," Kant-
Studien 56 (1966): 186-98, p. 191.
13. My discussion of the early reaction to Kant's psychology draws on four
sources: Frederick Beiser's The Fate of Reason (Cambridge, Mass.: Harvard
University Press, 1987); Gary Hatfield's forthcoming The Natural and the Nor-
mative: Theories of Perception from Kant to Helmholtz (Cambridge, Mass.: MIT
Press); Jurgen Bona Meyer, Kant's Psychologic (Berlin: Wilhelm Hertz, 1870);
and several articles by David E. Leary, "The Psychology of Jakob Friedrich
Fries" (1773-1843): Its Context, Nature, and Historical Significance, Storia e
Critica delta Psiscologia 3, (1982): 217-48, "The Philosophical Development of
the Conception of Psychology in Germany, 1780-1850," Journal of the History
of the Behavioral Sciences XIV (1978): 113-21, "Immanuel Kant and the De-
velopment of Modern Psychology," in William R. Woodward and Mitchell G.
Ash, The Problematic Science: Psychology in Nineteenth Century Thought (New
York: Praeger, 1982), pp. 17-42.
14. Johann Georg Hamann also offered a well-known metacritique, but one
that was intended to overthrow rather than shore up the Critical philosophy.
See Beiser's discussion (op. cit.), Chapter 1.
15. In Chapter 3, "The Probem," I provide a clarificatory regimentation of
this concept.
16. Here I draw on Beiser's detailed account of Reinhold's position. Op. cit.,
Chapter 8, especially pp. 227, 237, 241-55.
17. For further details, see Leary's article on Fries, op. cit.
18. I discuss his arguments in "Apperception as the Cogito" in Chapter 4 and
in "Understanding the First Paralogism in Chapter 7."
19. Immanuel Kant, Anthropology from a Pragmatic Point of View, translated
by Mary J. Gregor (The Hague: Martinus Nijhoff, 1974), pp. 14-15 (AA
VII.-133), 39 (AA VII:161).
20. Hatfield offers this convenient approximation. Op. cit., p. 169.
21. Bona Meyer, op. cit., quoted on p. 5.
22. Ibid., pp. 122, 129.
23. See the discussion of "innate" versus "a priori" below. I discuss Kant's
views on spatial perception at length in Chapter 2.
24. For a detailed account of Helmholtz's position with respect to Kant, see
Hatfield, op. cit., Chapter 5.
25. Strawson discusses this possibility in Part Five of The Bounds of Sense,
op. cit. See also James Hopkins, "Visual Geometry," Philosophical Review 82
(1973): 3-34.
26. Dr. G. Frege, The Foundations of Arithmetic, translated by J. L. Austin
(Oxford: Basil Blackwell, 1968), p. xe.
27. Ibid., p. vie.
 Notes 233
28. Beiser, op. cit, p. 249.
29. Quoted in Hatfield, op. cit., Chapter 4, note 16.
30. See Michael Dummett, Frege: Philosophy of Language (New York: Har-
per & Row, 1973), Chapter 19, Richard Rorty, Philosophy and the Mirror of
Nature (Princeton, N.J.: Princeton University Press, 1979), and, especially, Hans
Sluga, Gottlob Frege (Boston: Routledge & Kegan Paul, 1980), Introduction.
31. Alvin I. Goldman, Epistemology and Cognition (Cambridge, Mass.: Har-
vard University Press, 1986), p. 6.
32. See Sluga, loc. cit., and Goldman, loc. cit.
33. The most obvious example is Gilbert Ryle's The Concept of Mind (New
York: Barnes and Noble, 1949). However, much philosophy of mind of the
1950s, 1960s, 1970s, and even 1980s was antipsychologistic. For two clear ex-
amples, see Donald Davidson's "Mental Events," reprinted in Essays on Actions
and Events (Oxford: Clarendon Press, 1980), Zeno Vendler, The Matter of
Minds (Oxford: Clarendon Press, 1984).
34. Cited in Sluga, op. cit., p. 39.
35. Edmund Husserl, Logical Investigations, vol. I., translated by J. N. Find-
lay (New York: Humanities Press, 1970), pp. 92, 98.
36. Ibid., p. 172.
37. Although Husserl resists acknowledging Frege's influence, it seems very
clear. See Sluga's discussion, op. cit., p. 40.
38. Michael Dummett made this interesting observation in a conversation
with Philip Kitcher.
39. Kant's Wolffian critics maintained that the principle of sufficient reason
is a logical principle. Since the principle of the second analogy can be regarded
as Kant's analog for sufficient reason (see Guyer, op. cit., pp. 238, 242), one
might try to build a case for psychologism in logic. Given that Kant certainly
does not regard his principle as logical—and given that it clearly is not—this
case can be rebutted, however. See Beiser, op. cit., pp. 196-201, and Henry
Allison, The Kant-Eberhard Controversy (Baltimore: The Johns Hopkins Uni-
versity Press, 1973), pp. 95-96.
40. See below, the "Transcendental Psychology" section.
41. See Philip Kitcher, "Kant and the Foundations of Mathematics," Phil-
osophical Review 84 (1975): 23-50. For further discussion, see the "Transcen-
dental Psychology" section.
42. Goldman, op. cit., Introduction. For a useful discussion of the psychol-
ogism issue against the background of recent work in cognitive science, see also
Adrian Cussins, "Varieties of Psychologism," Synthese 70 (1987): 123-54.
43. J. B. Watson, "Psychology as the Behaviorist Views It," reprinted in A
Source Book in the History of Psychology, edited by Richard J. Hernstein and
Edmund G. Boring (Cambridge, Mass.: Harvard University Press, 1965),
pp. 513-14.
44. Walsh, op. cit., p. 38.
45. Metaphysical Foundations of Natural Science, translated by James El-
lington (Indianapolis: Bobbs-Merrill, 1970) (AA IV:471).
46. Loc. cit. For an interesting account of what happened to psychology in
234 NOTES
the wake of Kant's criticisms, see Leary, "The Philosophical Development of
the Conception of Psychology in Germany," 1780-1850, op. cit.
47. Anthropology, op. cit., p. 4 (AA VII: 121).
48. Immanuel Kant's Critique of Pure Reason, translated by Norman Kemp
Smith (New York: St. Martin's, 1965). I shall standardly use Kemp Smith's
translation, except where I think it distorts Kant's meaning. In those cases, I
will indicate that I have amended this translation, or retranslated the passage
myself.
49. Anthropology, op. cit., p. 21 (AA VII:140-41).
50. Cited in Robert J. Richards, "Christian Wolff's Prolegomena to Empirical
and Rational Psychology: Translation and Commentary," Proceedings of the
American Philosophical Society 124 (1980): 227-39, (p. 231).
51. Ibid., see p. 236.
52. Ibid., p. 237.
53. In Chapter 7,1 argue that many of Kant's putative objections to Rational
Psychology are also caveats about his own transcendental method.
54. Chapter 3, "The Law of Association," "Association and Apriority."
55. Immanuel Kant Logic, translated by Robert S. Hartman and Wolfgang
Schwartz (New York: Bobbs-Merrill, 1974), p. 13 (AA IX:10).
56. Ibid., pp. 37, 97, 100 (AA IX:33, 92, 94).
57. Hatfield, op. cit., pp. 18-19. Robert Pippin also notes that Kant's mean-
ing of "logic" is quite different from contemporary usage. See Kant's Theory
of Form (New Haven: Yale University Press, 1982), p. 153 note. In The Art of
Thinking (widely known as the "Port Royal Logic"), Arnauld explains that
"Logic consists in reflecting on these natural operations [conceiving, judging,
reasoning, ordering]... [this helps us to reason better, to correct and explain
defects in the mind's operation and]... we become better aware of the nature
of the mind by reflecting on its operations." [Antoine Arnauld, The Art of
Thinking, translated by James Dickoff and Patricia James (Indianapolis: Bobbs-
Merrill, 1964), pp. 29-30.]
58. In Chapter 7 I argue that part of the motivation for the Paralogisms
chapter is to distinguish transcendental psychology from Rational Psychology.
59. See the discussion of Wolffs method of philosophical hypotheses in Rich-
ard J. Blackwell, "Christian Wolff's Doctrine of the Soul," Journal of the History
of Ideas XXII (1961): 339-54, pp. 349ff.
60. Allison, The Kant-Eberhard Controversy, op. cit., pp. 135-36 (AA
VIII:221-23). I'm grateful to Allison for drawing my attention to this passage.
61. I provide a specific example of the distinction between an a priori0 element
and an innate mechanism in Chapter 2, "Pure Forms."
62. I discuss Kant's constructive notion of proof briefly in Chapter 2, "Ge-
ometry and the Space of Perception."
63. See A56 and the citation from Kant's letter to Fichte below.
64. Philip Kitcher defends this analysis of Kantian necessity in "Kant and the
Foundations of Mathematics," op. cit.
65. Paul Guyer has recently suggested a different understanding of "neces-
sity" for Kant. I consider Guyer's position in note 63 of Chapter 5.
 Notes 235
66. This is Kant's evaluation. Locke thought that by showing empirical origins
for concepts, he could demonstrate the flexibility of the human mind to new
ideas.
67. I have presented only some aspects of Kant's account of the a priori status
of mathematics. Besides claims about the a priori0 status of spatial properties,
Kant also has a particular view of construction in mathematical proof. See,
Chapter 2, "Geometry and the Space of Perception," and Philip Kitcher, "Kant
and the Foundations of Mathematics," op. cit.
68. Husserl, op. cit., p. 100.
69. This epistemic analysis is defended by Philip Kitcher in "Kant's Philos-
ophy of Science," in Allen W. Wood, ed., Self and Nature in Kant's Philosophy
(Ithaca: Cornell University Press, 1984), pp. 185-215. I present the analysis of
the relevant empirical capacity in Chapter 8, "Concepts and Reasoning."
70. Because he does not recognize analyses of empirical capacities, Guyer
believes that either Kant's arguments are all conditional or they must presuppose
that a noumenal mind imposes structure on nature. See, Guyer, op. cit., for
example, pp. 53ff., and Paul Guyer, "Kant on Apperception and A Priori Syn-
thesis," American Philosophical Quarterly 17 (1980): 205-12. However, also see
the discussion of transcendental proof at the end of his book, pp. 417-18.
71. Immanuel Kant, Critique of Practical Reason, translated by Lewis White
Beck (Indianapolis: Bobbs-Merrill, 1956), p. 48 (AA V:47).
72. Daniel Dennett suggests the kinship between Kant and contemporary
cognitive science on this point in "Artificial Intelligence as Philosophy and as
Psychology," in Brainstorms (Cambridge, Mass.: Bradford Books, 1978),
pp. 109-26. For a further account of the strategy of functional decomposition,
see Robert Cummins, "Functional Analysis," Journal of Philosophy 72 (1975):
741-60.
73. P. F. Strawson, the Sounds chapter of Individuals (London: Methuen,
1959).
74. I argue this point in "Discovering the Forms of Intuition," Philosoph-
ical Review XCVI (1987): 205-48 and later in Chapter 2, "The Isolation
Argument."
75. Gary Hatfield offers this helpful way of summarizing one line of objection.
Op. cit., p. 133. Later I agree with his answer: Transcendental psychology is
the psychology of the knowing mind.
76. I consider Strawson's position in Chapter 5, "Apperception and Kant's
System."
77. Perhaps the most prominent advocate of this reading is Wilfrid Sellars.
See "Metaphysics and the Concept of a Person," in Karel Lambert, ed., The
Logical Way of Doing Things (New Haven: Yale University Press, 1969,
pp. 219-32, and, especially, "this I or he or it (the thing) which thinks," Sellar's
Presidential Address to the Eastern Division of the APA in 1970. I try to
disentangle some of the relations among these three selves in "Kant's Real
Self," in Allen W. Wood, ed., op. cit., pp. 113-47, and Chapter 5, "Apper-
ception and Kant's System."
78. See Chapter 5, "Apperception and Kant's System."
236 NOTES
79. See Chapter 7, "Understanding the First Paralogism."
80. Allen Newell and Herbert Simon, Human Problem Solving (Englewood
Cliffs, N.J.: Prentice-Hall, 1972), p. 53.
81. These discussions focus on the relation between general (and so tran-
scendental) and applied logic, but as I have already noted, given our contem-
porary understanding of "logic" and "psychology," it is clearer to make Kant's
point in terms of the relation between transcendental and applied, or empirical,
psychology. In Chapter 8 I show in detail how constraints from transcendental
psychology can guide empirical research.
82. W. V. Quine, "Two Dogmas of Empiricism," in From a Logical Point
of View (New York: Harper, 1961), pp. 20-46. See also "Truth by Convention"
and "Carnap on Logical Truth," both in The Ways of Paradox (New York:
Random House, 1966). For a very clear presentation of Quine's views, see
Gilbert Harman, "Quine on Meaning and Existence I," Review of Metaphysics
XXI (1967): 124-51. Quine's attacks on analyticity have not gone unchallenged.
Strawson and H. P. Grice tried to defend the notion of analyticity in "In Defense
of a Dogma," Philosophical Review LXV (1956): 141-58.
83. The similarity is argued in greater detail by Philip Kitcher in "How Kant
Almost Wrote 'Two Dogmas of Empiricism,' " in J. N. Mohanty and Robert
W. Shahan, eds., Essays on Kant's Critique of Pure Reason (Norman, Okla.:
University of Oklahoma Press, 1982), pp. 217-49.
84. I discuss his reasons for these claims at some length in Chapter 8.
85. Arnulf Zweig, Kant Philosophical Correspondence 1759-99 (Chicago:
University of Chicago Press, 1967), p. 9.253 (AA XII:370). My attention was
drawn to this letter by Derk Pereboom's interesting discussion in "Kant's Notion
of the A Priori and the Transcendental" ms.
86. Cited in Allison, The Kant-Eberhard Controversy, op. cit., p. 175. Al-
lison suggests that Kant approved the letter, p. 13.
87. This description follows the model of transcendental arguments given by
Ralph C. S. Walker, in Kant (London: Routledge & Kegan Paul, 1982), Chapter
II, and discussed by Anthony L. Brueckner in "Transcendental Arguments I,"
Nous 17 (1983): 551-75.
88. See Barry Stroud, "Transcendental Arguments," Journal of Philosophy
65 (1968): 241-56, and Judith Jarvis Thomson, "Private Languages," American
Philosophical Quarterly 1 (1964): 20-31. Although there have been many ar-
guments back and forth in the literature, the criticisms originally made in these
papers still seem completely cogent.
89. Margaret Wilson notes that Kant's target is Humean and not Cartesian
skepticism in "Kant and the Refutations of Subjectivism," in L. W. Beck, Kant's
Theory of Knowledge (Dordrecht: Reidel, 1974), pp. 208-17.1 also argue against
Descartes being a primary target in "Kant's Patchy Epistemology," Pacific Phil-
osophical Quarterly 68 (1987): 306-16.
90. Chapter 4, "Troubles with Apperception."
91. The relevance of Kant's work to current research in cognitive science and
cognitive psychology has been noted by scholars in these fields. See, for example,
 Notes 237
Colin Martindale, "Can We Construct Kantian Mental Machines?" The Journal
of Mind and Behavior 8 (1987): 261-68.

Chapter 2

1. Hans Vaihinger, Commentar zu Kants Kritik der reinen Vernunft, vol. II,
(New York: Garland, 1970), pp. 33-263.
2. Norman Kemp Smith, A Commentary to Kant's "Critique of Pure Rea-
son," (New York: Humanities Press, 1962), pp. 85-86.
3. In "On Kant's Notion of Intuition (Anschauung)" in T. Penelhum and
J. Macintosh, The First Critique: Reflections on Kant's Critique of Pure Reason
(Belmont, Calif.: Wadsworth, 1969), pp. 38-53. Hintikka reads "intuition" as
marking the singularity of a representation, and not its relation to sensibility.
He offers this interpretation despite the fact that Kant's introductory discussion
of "intuition" links it firmly to sensibility: When the mind is affected in a certain
way by objects, the faculty of sensibility supplies us with intuitions (A19/B34).
For criticisms of Hintikka's position, see Allison, Kant's Transcendental Ide-
alism, op. cit., p. 67, and Manley Thomson, "Singular Terms and Intuitions in
Kant's Epistemology," Review of Metaphysics 26 (1972): 314-43.
4. Robert Howell seems to presuppose Hintikka's position in "Intuition,
Synthesis, and Individuation in the Critique of Pure Reason," in Nous (1973):
207-32. For critical discussions, see previous note.
5. Kant does not appeal to an "isolation" argument to establish his thesis
about time in the Aesthetic. As I argue later, however, this is the heart of his
case for his theory of space as the form of outer intuition. See also the special
problems raised by the theory of time discussed in Chapter 5, "The Ideality of
Time."
6. John Handyside, Kant's Inaugural Dissertation and Early Writings on
Space (La Salle, 111.: Open Court, 1929), pp. 3-15; see pp. 11-12 (AA I:
24-25).
7. Ibid., pp. 13-14 (AA 1:139-40).
8. For an account of Kant's various senses of "a priori," see Chapter 1,
"Transcendental Psychology,"
9. In "Apperception and Kant's System" in Chapter 5 I discuss the relation
between Kant's theory of the forms of intuition and his metaphysical views
about space and time.
10. Quoted in Nicholas Pastore, Selective History of Theories of Visual Per-
ception, 1650-1950 (Oxford: Oxford University Press, 1971), p. 68.
11. See Richard J. Herrnstein and Edwin G. Boring, A Source Book in the
History of Psychology (Cambridge, Mass.: Harvard University Press, 1968),
pp. 113-17.
12. Cited in Pastore, op. cit., p. 414.
13. G. W. Leibniz, New Essays on Human Understanding, abridged ed.,
238 NOTES
translated and edited by Peter Remnant and Jonathan Bennett (Cambridge:
Cambridge University Press, 1982), p. 138.
14. See Pastore, op. cit., the relevant selections in Herrnstein and Boring,
op. cit., and T. E. Jessop, A Bibliography of George Berkeley (The Hague:
Martinus Nijhoff, 1973), Introduction.
15. Arthur Warda, Immanuel Kants Bucher (Berlin: Verlag von Martin Bres-
laner, 1922), p. 46.
16. The situation is somewhat cloudy because Jessop never saw the translation
he lists. The early editions of Alciphron all had the Essay on Vision as an
appendix. In 1734 a French translation was made, which included the Essay.
According to Jessop, a German translation was made from the English and the
French. Further, as Geoffrey Keynes points out in A Bibliography of George
Berkeley (Oxford: Oxford University Press, 1976), Jessop does not tell us where
to find any copies of this edition. The question of Kant's knowledge of Berkeley's
writings has often been raised. For a recent discussion, see G. J. Mattey, "Kant's
Conception of Berkeley's Idealism," Kant-Studien 74 (1983): 161-75.
17. According to Hamann, Tetens book lay open on Kant's desk as he wrote
the Critique. See Jurgen Bona Meyer, op. cit., p. 56.
18. Leipzig, 1777, republished by the Kantgesellschaft (Verlag, 1913). Page
references are to the later edition.
19. Ibid., Introduction, p. 36.
20. Ibid., pp. 42ff.
21. Ibid., p. 194. Cf. also p. 271.
22. Ibid., pp. 42-43ff.
23. Kants Erkenntnispsychologie, Kant-Studien Ergangungshefte 101 (1971),
pp. 1-176, pp. 90-91. Satura offers a useful discussion of the history of these
sources and the controversies surrounding them, pp. 4-24.
24. Gregor, op. cit., p. 49, (AA VII:173).
25. See "Hume's Absence" section.
26. In Kant: Selected Pre-Critical Writings and Correspondence with Beck,
edited by G. B. Kerford and D. E. Walford (New York: Barnes and Noble,
1968), p. 55 (AA 11:392-93).
27. Ibid. (AA 11:393).
28. The Dissertation account is somewhat confusing, because he describes
"sensations" both as the effects of objects on the sensory organs and as the
matter of sensory representations.
29. For further discussion of Kant's use of the term "intuition" see Chapter
3, "Representing Objects."
30. Kant makes this point explicitly in his anthropology lectures; see Gregor,
op. cit., p. 35 (AA VII:156-57).
31. Kerford and Walford, op. cit., p. 55 (AA 11:393).
32. Henry E. Allison, The Kant-Eberhard Controversy, op. cit., pp. 135-36
(AA VIH:221-23).
33. Kerford and Walford, op. cit., pp. 73-74 (AA 11:906).
34. See Chapter 1. "Transcendental Psychology."
 Notes 239

35. I will not discuss the third and fourth arguments of the Metaphysical
Exposition where this distinction is crucial, because I do not think the tran-
scendental psychology of spatial perception is particularly prominent in those
passages. The problem of distinguishing between and relating conception and
perception will be prominent in Chapter 6.
36. Reflexion 3958 (AA XVII:366).
37. Compare Kerford and Walford, op. cit., p. 68, but see AA 11:402.1 am
grateful to my former colleague Jasper Hopkins for retranslating this passage
for me. Hopkins pointed out that while the standard translations use a technical
term like "abstracted" or "derived," Kant's Latin expression "hauriri" is non-
technical. (It would be used in such ordinary contexts as drawing water out of
a well.)
38. See Vaihinger, op. cit., pp. 71-88. Robert Pippin's interesting discussion
of forms, op. cit., does not investigate the theory of spatial perception, because
he accepts the common view that Kant just assumes that sensation involves
"formless material," p. 56.
39. Vaihinger, op. cit., p. 180.
40. Kemp Smith actually refers the reader to B207 (op. cit., p. 86), but he
must mean B208. Here I offer what I take to be fairly standard readings of
Kant's terms "intensive" and "extensive." In providing these glosses I do not
try to resolve any serious questions about Kant's putative distinction between
the two types of magnitude. My point is that Vaihinger and his followers have
inverted the logic of Kant's reasoning at A166/B208, whatever meanings are
assigned to "intensive" and "extensive."
41. See Pastore, op. cit., pp. 6-10.
42. My analysis places great weight on our inability to sense the third di-
mension. Henry Allison has objected to me that Kant does not stress the three-
dimensional character of space and that many of his examples of geometrical
properties (e.g., at most one straight line through two points) are drawn from
planar geometry. Although this is true, I do not believe the examples undercut
my interpretation. Kant refers often enough to the "space [not planes] of the
geometers" and to its three-dimensional character [e.g., B41, Prolegomena,
op. cit., pp. 28-29 (AA IV:284-85), and Dissertation, op. cit., p. 60]. Further,
the standard mathematical view was that geometry describes space and that
planar geometry abstracts from full three-dimensional space. Presumably Kant
expected his readers to understand his position against this background. I am
also grateful to Lome Falkenstein, who sent me a paper critical of my original
discussion of these issues in "Discovering the Forms of Intuition," op. cit. Like
Allison, Falkenstein objects to the weight I place on the third dimension. I hope
the present discussion clarifies my original, excessively cryptic presentation.
43. The Works of George Berkeley Bishop ofCloyne, vol. I, edited by A. A.
Luce and T. E. Jessop (London and Edinburgh: Nelson, 1967), pp. 186-89.
44. The view that touch informs vision about extent and distance has enjoyed
widespread popularity at different times. Recent work suggests that it is, how-
ever, false. In a survey of the literature, Eleanor Gibson and Elizabeth Spelke
240 NOTES
draw the following conclusions: "There does, thus, seem to be a primitive eye-
arm coordination in the newborn that is adapted to the three-dimensional layout
of objects well before grasping and manipulation occur" (p. 7, my emphasis).
"There was little active manipulation before 24 weeks.... Even the youngest
age group differentiated the object from its picture, however... "(p. 8). See
Gibson and Spelke, "The Development of Perception," in John H. Flavell and
Ellen M. Markman, eds., Cognitive Development (Handbook of Child Psy-
chology, vol. 3, Paul H. Mussen, general ed. (New York: Wiley, 1983), pp.
1-76).
45. Remnant and Bennett, op. cit., p. 137. In this discussion, I do not con-
sider Kant's well-known philosophical objections to Berkeley's theory of space.
From the 'perspective of his own system, Kant chastises Berkeley for being a
transcendental realist about space and for not recognizing its a priori character,
among other complaints. At this point in the Transcendental Aesthetic, how-
ever, Kant is in no position to argue from the truth of his own system to the
shortcomings of Berkeley's. He must first find reasonably uncontroversial prem-
ises to establish his own theory. Margaret Wilson offers an informative discussion
of some of Kant's philosophical objections to Berkeley in "Kant and 'The
Dogmatic Idealism of Berkeley,' " Journal of the History of Philosophy (1971):
459-75.
46. Leibniz describes retinal images as "paintings (as it were) that he forms
at the back of his eyes." Remnant and Bennett, op. cit., p. 138.
47. Ibid., p. 137, Leibniz's emphasis.
48. See, for example, Remnant and Bennett, op. cit., p. xiii.
49. AA. XVIII:69.
50. Satura reads Kant as claiming that the third dimension is registered by
touch (op. cit., p. 90). The crucial evidence is the following passage about bodily
form from the Anthropology:
[Touch]... by touching the surface of a solid body we can find out what
shape it has.... Without this sense organ we should be unable to form any
concept at all of the shape of a body. So the other two senses of this first
class must be referred originally to its perceptions, if they are to provide
experiential knowledge [Gregor, op. cit., pp. 33-34 (AA VH:155)].
Satura's reading of this passage is quite plausible. Further, the passage suggests
that Kant might reject Leibniz's claims about paralyzed individuals learning
geometry that I present him as accepting in the text. This is a difficult exegetical
puzzle. Although Satura's gloss is plausible if the passage is read in isolation,
it is flatly inconsistent with Kant's claims about space and form in the Dissertation
and the Critique. Whatever reason one gives for Kant's denial that our perceptual
information about extent and form derives from the senses, there is no question
that he does deny this. Faced with contradiction, the natural assumption would
be that the major works offer a more precise account of Kant's considered
views.
The reading that yields contradiction is not forced on us, however. Kant does
 Notes 241
not say that touch yields sensations [Empfindungen] or sensa of space or form,
although he has this precise vocabulary available; what he claims is that touch
involves perception [Wahrnehmung] of form and is necessary for the concept of
bodily form. It is not surprising that Kant would hold this view, since most of
his predecessors, Leibniz included (see Remnant and Bennett, op. cit., pp. 77,
122-24), believed that touch was important to empirical knowledge and that it
had a special role in our acquiring the idea of solidity.
51. Kant has a well-known argument that there is only one space, the fourth
metaphysical exposition in A (third in B): We can represent to ourselves only
one space. However, I do not see how the considerations he raises there, that
spaces are limitations of space, provide any argument that different sense could
not furnish different, incommensurable spaces.
52. Luce and Jessop, op. cit., vol. I, pp. 187-88.
53. See note 35.
54. See Allison's discussion of this issue in Kant's Transcendental Idenlism,
op. cit., pp. 82-86.
55. It is sometimes argued that Berkeley and perhaps other Empiricists held
that sensations were completely aspatial. If this interpretation is accepted, then
Kant would have an easier but less interesting argument against them. See Gary
Hatfield and William Epstein, "The Sensory Core and the Medieval Foundations
of Early Modern Perceptual Theories," 75/5 70 (1979): 363-84.
56. AA 11:402. See note 37.
57. Gottfried Wilhelm Leibniz, Philosophical Papers and Letters, edited by
Leroy E. Loemker (Amsterdam: D. Reidel, 1976), pp. 703-704. Kant may also
have been thinking of Christian Wolff's closely related view. For an account of
Wolff's position and its differences from Leibniz's, see Beck, Early German
Philosophy (Cambridge, Mass.: Harvard University Press, 1969), pp. 268-70.
58. Kant may also be noting that Leibniz's own position in the correspondence
with Clarke suggests that our representation of space involves a priori0 elements.
Leibniz claims that we perceive objects in various positions relative to one
another. We then abstract from the objects and think of the positions themselves,
filling in the currently unoccupied places in the perception, to reach the intel-
lectual idea of space as a system of positions for actual and possible objects.
Thus, Leibniz seems committed to the view that the creative subject is respon-
sible for elements in our representation of space. So Kant's point may also be
that it is inconsistent for Leibniz to characterize [the representation of] space
as a product of the creative activity of the subject and then to claim to have
shown that it depends on actual objects encountered in perception.
I do not wish to downplay the difficulties in interpreting Leibniz, but I think
my interpretation of the fifth letter to Clarke is fairly widely held. See John
Earman, "Was Leibniz a Relationist?" in Midwest Studies, vol. IV, Peter French,
Ted Uehling, and Howard Wettstein, eds. (Minneapolis: University of Min-
nesota Press, 1979); Martial Gueroult, "Space, Point, and Void in Leibniz's
Philosophy," Leibniz: Critical and Interpretive Essays, edited by Michael Hooker
(Minneapolis: University of Minnesota Press, 1982); Robert McRae, Leibniz:
242 NOTES
Perception, Apperception and Thought (Toronto: University of Toronto Press,
1976); Jill Buroker, Space and Incongruence (Amsterdam: D. Reidel, 1981).
59. Guyer, op. cit., p. 357ff. Guyer presents a number of telling citations
from the Reflexionen and from student lecture notes (Metaphysik Mongrovius,
vol. 29 of AA). Let me give just one example:
That synthetic a priori propositions are possible only through the subjective
form of sensibility, consequently that their objects can be represented only
as appearances, is to be recognized from the fact that they are accompanied
with necessity, but not from concepts by means of analysis... [R6355,
18:681, Guyer's translation, p. 358].
This passage shows that Kant believes the subjectivity of space to be crucial to
the possibility of a priori knowledge fof Geometry]. It does not imply that he
denies the importance of other considerations, however. Neither does it emulate
Berkeley's dramatic gesture of offering to stake his case for immaterialism on
the inability to imagine an unperceived tree. None of the passages Guyer cites
implies that Kant rests his whole case on Geometry. The issue is difficult, because
here I am only arguing from the absence of evidence. In the text I offer positive
reasons for believing that other considerations stand behind the theory of space.
I discuss Guyer's views on Geometry further in Chapter 5, note 63. Hatfield,
op. cit., also lays great emphasis on the importance of Geometry in Kant's
theory of spatial perception.
60. "The Historical and Conceptual Relations between Kant's Metaphysics
of Space and Philosophy of Geometry," Journal of the History of Philosophy
11 (1973): 483-512. See especially pp. 485-87, 497-500.
61. See Chapter 1, "Transcendental Psychology," for the meaning ofaprioriL.
62. "Kant's Theory of Geometry," Philosophical Review 94 (1985): 455-506.
63. Kerford and Walford, op. cit., pp. 69-70, 71 (AA 11:402-403, 404).
64. "Infinity and Kant's Conception of the 'Possibility of Experience,' " Phil-
osophical Review 73 (1964): 182-97; reprinted in Kant, edited by Robert Paul
Wolff (Garden City, N.Y.: Anchor Books, 1967), pp. 37-52. I later criticize
Parson's interpretation of these matters in the Critique. I should note, however,
that many of Kant's remarks about Geometry and the form of outer perception
in the Dissertation accord very well with Parsons's reading. See especially the
passages cited in the previous note.
65. Besides Parsons's own critique, see James Hopkins, "Visual Geometry,"
Philosophical Review 82 (1973): 3-34, especially p. 4, and Kitcher, "Kant and
the Foundations of Mathematics," op. cit., pp. 31ff.
66. See above, "Kant's Empirical Assumptions."
67. See Earman, op. cit.
68. The case may be different for the Dissertation. See notes 63 and 64.
69. Pippin, op. cit., draws a similar conclusion, pp. 84-87.
70. Herrnstein and Boring, op. cit., pp. 125-31. The discovery of binocular
vision was sometimes used to support nativism in perception, which was often
taken to be Kant's position. See Hatfield's discussion, op. cit., p. 273.
 Notes 243
71. See David Marr's discussion of this problem in Vision (San Francisco:
W. H. Freeman, 1978), pp. 111-59.
72. Below I consider a theory that is an explicit defense of Kant's view. I
think the work of Marr and his associates at MIT was strongly Kantian in spirit.
Marr hypothesized that general facts about the structure of the world to be
perceived are built into the perceptual system. See my "Marr's Computational
Theory of Vision," Philosophy of Science 55 (1988): 1-24.
73. (Princeton, N.J.: Princeton University Press, 1947).
74. "The Geometry of the Visibles," Nous 8 (1974): 87-117.
75. Barbara Landau, Henry Gleitman, and Elizabeth Spelke, "Spatial Knowl-
edge and Geometric Representation in Child Blind from Birth," Science 213
(1981): 1275-78.
76. B. White, F. Saunders, L. Scadden, P. Bach-y-Rita, and P. Collins,
''Seeing with the Skin," Perception and Psychophysics 7 (1970): 23-27.
77. See Chapter 6, "Kant's Long Argument."
78. The Hippocampus as a Cognitive Map (Oxford: Oxford University Press,
1978).
79. As I understand O'Keefe and Nadel's description, a set of spatial trans-
formation rules is a set of rules that so defines a space that any point in that
space can be expressed as a function of any other point.
80. Israel Leiblich and Michael Arbib, "Multiple Representations of Space
Underlying Behavior," Behavioral and Brain Sciences 5 (1982): 627-59.
81. Thomas Kuhn, The Structure of Scientific Revolutions, 2nd ed. (Chicago:
University of Chicago Press, 1970), Chapter 10.

Chapter 3
1. Robert Paul Wolff, op. cit., is a notable exception to this generalization.
2. See Kemp Smith's discussion of this issue in his Commentary, op. cit.,
pp. xix-xxv.
3. Bennett, Kant's Analytic, op. cit., p. 100.
4. Guyer, op. cit., Part II.
5. Op. cit., p. 171. Allison goes on to suggest things are not quite as bad
as they seem, because Kant is able to establish a number of "analytic" claims
(pp. 171-72). As I argue in Chapter 1, however, Kant does not conceive of the
theses established by his transcendental method as analytic.
6. See Chapter 6, "Kant's Long Argument" and "11 26 as Completing the
Argument of the Metaphysical Deduction."
7. The view that Kant's arguments are analytic, even if not obviously so,
may be encouraged by a quick reading of the opening sentence of the A De-
duction: "It is entirely contradictory and impossible that a concept should orig-
inate a priori and relate to an object, unless it is either contained in the concept
of possible experience, or else consists in elements of a possible experience"
244 NOTES
(A95, my translation). Those who offer analytic interpretations assume that the
first option is the more relevant; but it is clear from Kant's explicit discussion
in the Methodology that his transcendental method employs the second. See
Chapter 1, "Transcendental Psychology."
8. See Beiser's account of the "Garve-Feder" review, which is often thought
to have been a spur to Kant's recasting of the chapter for the second edition.
Op. cit., pp. 172-77.
9. This reading is elaborated in Karl Ameriks, "Kant's Transcendental De-
duction as a Regressive Argument," Kant-Studien 69 (1978): 273-87.
10. See Chapter 1, "Kant Against 'Psychology.' "
11. For an account of the difference between epistemic analyses and analyses
of empirical capacities, see Chapter 1, " Transcendental Psychology."
12. "Mental state" is another possibility for the generic term. As Henry
Allison pointed out to me, however, this would include volitions, and pleasure
and pain, which are not dealt with until the later Critiques. Although "cognitive
state" is somewhat clumsy, it is true to the spirit of transcendental psychology
in being a dummy name for a state that performs a role in cognition, but whose
nature we do not understand.
13. Ak IX:91.
14. Although the passage in the Logic and other passages (see A19/B34,
B208, A320/B376-77) make it clear that Kant does not believe that all Vorstel-
lungen represent, in the Critique he says: "All representations have, as repre-
sentations, their objects, and can themselves in turn become objects of other
representations" (A108). I do not believe that this indicates substantive con-
fusion, but it does require terminological clarification. Still such remarks are
misleading. So, for example, Robert Howell (op. cit.) asserts that Kant believed
that all representations are representational.
15. Typically, when Leibniz discusses perceptions, he is content to note that
we must acknowledge their existence and their alleged properties, because the
soul presents us with clear examples. See, for example, Monadology 14-17 in
Loemker, op. cit., p. 644.
16. Remnant and Bennet, op. cit., p. 114. See also Max Dessoir, Geschichte
der Neueren Deutschen Psychologic (Amsterdam: E. J. Bonset, 1964), Zweite
Auflage, p. 39.
17. Tetens, op. cit., p. 8.
18. Monadology 16, in Loemker, op. cit., p. 644.
19. Dessoir, op. cit., p. 69; see also Blackwell, op. cit., pp. 344-47.
20. "Kant's Sensationism," Synthese 47 (1981): 229-55. It is possible that
sensationism was also part of the background to Kant's discussion of space. As
George notes, Condillac claims that the eye cannot grasp even simple shapes
directly (See Condillac's Treatise on the Sensations, translated by Geraldine Carr
(Los Angeles: University of Southern California Press 1930), p. 68. Condillac
goes on to discuss the problem of how sight learns distance from touch, and the
results of the Cheselden case (pp. 59-60). So his discussion takes place against
the general background I present in Chapter 2. Kant might have been especially
 Notes 245
influenced by this discussion of the issue, but I know of no reason to think so.
I discuss this passage again in note 23 of Chapter 6.
21. See for example, Tetens, op. cit., p. 264.
22. Manfred Kuehn argues for the influence of Reid and other Scottish phi-
losophers on German philosophy in Scottish Common Sense in Germany, 1768-
1800: A Contribution to the History of the Critical Philosophy, (Montreal:
McGill-Queen's University Press, 1987). I discuss Reid's views further in Chap-
ter 6.
23. In Chapter 6, "Differences Between the Editions," I suggest that parts
of the A Deduction may have been shaped by Reid's view that perception
involves a concept of the object. However, I disagree with George's assessment
that Kant always took perception to involve concepts. At least in the B De-
duction, he avoided this substantive and controversial assumption. (See Chapter
6,) "11 26 in the B Deduction."
24. George, op. cit., pp. 236, 244. Although I agree with much of George's
argument for the influence of Condillac, I do not understand his suggestion that
Kant does not think of the mind as actually gathering up the elements in cognitive
states and putting them altogether (p. 236). Presumably this is exactly what
Kant means when he says that the synthesis of apprehension must run through
and hold together the multiple contents of cognitive states (A99; see also A77/
B103 and A120a).
25. Warda, op. cit., p. 50.
26. Inquiry, op. cit., p. 62. Compare Treatise, op. cit., pp. 98ff.
27. Inquiry, op. cit., p. 63.
28. Ibid. pp. 62-63.
29. Ibid., p. 67.
30. Among other sources, Kant was aware of many of the Treatise's discussion
through James Beattie's lengthy citations in Essay on the Nature and Immuta-
bility of Truth, vol. IV, Beattie's Works, (Philadelphia: Hopkins and Earle,
1809).
31. Treatise, op. cit., pp. 197-218.
32. Ibid., p. 218.
33. George, op. cit., p. 231. See also Blackwell's discussion on pp. 344-46.
34. Reflexion 695 (AA XV:308-9). This passage was drawn to my attention
by George's discussion and I amend his translation (op. cit., p. 232). See also
F. E. England, Kant's Conception of God (with translation of the Nova Dilu-
cidatio) (New York: Humanities Press, 1968), pp. 246-47.
35. For a classic discussion, see William Porterfield, A Treatise on the Eye
and the Manner and Phenomena of Vision, vol. Ill (Edinburgh: G. Hamilton
and J. Balfour, 1759), pp. 221, 336. Porterfield's work was cited by Reid in the
Inquiry, which as noted earlier, was widely known in Germany. This assumption
is still standard in work in perception. See David Marr's discussion of zero-
crossings in Vision, op. cit., pp. 67-74.
36. See, for example, Remnant and Bennett, op. cit., pp. 53-58; Treatise,
op. cit., p. 252.
246 NOTES
37. Dessoir notes that this theory had been conclusively overthrown by the
time Kant was writing (op. cit., p. 400).
38. Op. cit., p. 26 (AA IV:282).
39. J. J. Gibson, The Senses Considered as Perceptual Systems (Boston:
Houghton Mifflin, 1966), p. 5. For Gibson's general position see also J. J. Gib-
son, The Ecological Approach to Visual Perception (Boston Houghton Mifflin,
1979).
40. For example, Marr, op. cit., pp. 29-30.
41. R. P. Wolff is also an exception to this generalization. I discuss his position
in the section "Robert Paul Wolff on Rules of Synthesis." After completing this
chapter, I became aware of Hansgeorg Hoppe's book-length study of synthesis,
Synthesis Bei Kant (Berlin: Walter de Gruyter, 1983). Unlike most contemporary
philosophers, Hoppe tries to deal with synthesis by considering its relation to
some psychological claims, specifically Piaget's work. Both Ralph C. S. Walker
and Paul Guyer have written about synthesis. However, they do not explore
Kant's account of synthesis—what it really asserts and whether it is reasonable—
but simply read the doctrine of a priori synthesis as the doctrine that the mind
imposes various properties on the objects of knowledge. (See Guyer's "Kant
on Apperception and A Priori Synthesis," American Philosophical Quarterly 17
(1980): 205-12, and Ralph C. S. Walker, "Synthesis and Transcendental Ide-
alism," Kant-Studien 76 (1985): 14-27.)
Robert Howell (op. cit.) also discusses synthesis but never considers it as a
psychological doctrine. Following Hintikka's analysis of intuitions as analogues
of singular demonstratives, he describes synthesis as a means of cross-possible-
world identification (p. 214). Henry Allison's depsychologized version of syn-
thesis is more typical. Allison starts by suggesting that Kant's views about
synthesis do not in any way rest on contemporary assumptions about what the
inputs to our minds are like. I do not see how this can be so. Here and in our
later discussion it is obvious that Kant's remarks only make sense if we see them
as rooted in various contemporary assumptions about the data of sense. (Cer-
tainly they presuppose the falsity of the simulacra theory.) Allison transforms
Kant's psychological doctrine into an acceptable logical form. "[Synthesis is a]
logical act.... The consciousness of this act, that is, the consciousness of syn-
thesis is therefore, the consciousness of the form of thinking" (op. cit., p. 144).
This strategy is vulnerable to Walsh's objection to Paton: How can something
merely logical be an act?
42. In the Paralogisms chapter, he maintains that arguments about the nec-
essary capacities for knowledge are incapable of yielding such information. See
Chapter 7, "Understanding the First Paralogism."
43. Although I describe the representation as "stable," I should reiterate that
Kant says nothing about the nature of representations. So, for example, these
representations need not endure for appreciable amounts of time, or be located
in any one place in the mind or brain. In fact, there is nothing in Kant's doctrine
that would rule out the sort of ephemeral, distributed representations recently
suggested by work in Parallel Distributed Processing (or Connectionism as it is
 Notes 247
also known). See James L. McClelland and David E. Rumelhardt, Parallel
Distributed Processing (Cambridge, Mass.: MIT Press, 1986).
44. Kemp Smith's translation blunts this point somewhat. Kant's original
German is "... Regel..., welche die Reproduction des Mannigfaltigen a priori
nothwendig und einen Begriff, in welchem dieses sich vereinigt, moglich macht."
Kemp Smith renders this: "a rule... as makes the reproduction of the manifold
a priori necessary and renders possible a concept in which it is united." That
is, he takes the scope of "macht" to include both clauses, but leaves out the
moglich, when rendering the clause about reproduction. Although Kant's com-
pound modalities are very hard to render, I think this is an error.
45. By not mentioning representations explicitly, Kant seems to suggest that
the mind produces the regularity in the contents of mental states that is necessary
for the law of association to work. Were he to claim that the mind creates its
own content, he would go against all the principles on which his theories of
cognition and transcendental idealism are based.
46. Tetens observes that since the time of Locke the so-called law of asso-
ciation has been regarded as a fundamental law in psychology. Like Kant, he
thinks that it is inadequate, because it merely involves the reproduction of past
sequences and not the production of new representations (op. cit., vol. I,
pp. 108ff).
47. Tetens, op. cit., p. 266.
48. I discuss this issue further in Chapter 6.
49. R. P. Wolff, op. cit., pp. 125ff.
50. Ibid., p. 131.
51. The Groundwork of the Metaphysics of Morals, translated by Lewis White
Beck (Indianapolis: Bobbs-Merrill, 1959), p. 29.
52. This confusion was probably fostered by Christian Wolff's view that we
observe all that transpires in our our minds. See Chapter 1, "Kant Against
'Psychology.' "
53. I discuss this passage further in Chapter 4, "Judgments."
54. O'Neill, op. cit.
55. Ibid., p. 158.
56. Anne Treisman, "Features and Objects in Visual Processing," Scientific
American 225 (November 1986): 114B-25.
57. See A. M. Treisman and H. Schmidt, "Illusory Conjunctions in the Per-
ception of Objects," Cognitive Psychology 14 (1982): 107-41; and A. M. Treis-
man and G. A. Gelade, "A Feature Integration Theory of Attention," Cognitive
Psychology 12 (1980): 97-136.
58. Patricia Churchland made this observation in conversation.
59. Immanuel Kant, Metaphysical Foundations of Natural Science, op. cit.,
pp. 13-14.
60. Paton suggests that A. G. Baumgarten and G. H. Meier were Kant's
specific targets (Paton, op. cit., vol. I, p. 522n).
61. Tetens, op. cit., vol. 1, p. 361.
62. Locke, op. cit., vol. II, p. 362.
248 NOTES
63. Condillac, essay, pp. 74-75.
64. Ibid, vol. 1, p. 365.
65. The topic of this passage is badly obscured by Kemp Smith's rendering
of Erkenntnisse as "modes of knowledge" here.
66. I present a more detailed account of this argument in Chapter 6, "Kant's
Long Argument."
67. See, for example, Bennett, op. cit., Chapter 6.
68. The only exception would be syntactic relations that were grounded in
our ways of processing information derived from sensory encounters with
objects.

Chapter 4
1. Rorty asked this question during his Kant seminar at Princeton in the fall
of 1970, which I attended. While he meant it ironically, I thought it was an
interesting question.
2. For example, see Roger Scruton, Kant (Oxford: Oxford University Press,
1982), p. 32, and T. E. Wilkerson, Kant's Critique of Pure Reason (Oxford:
Oxford University Press, 1976), p. 49.
3. This seems to be the case for Wilkerson, loc. cit.
4. It will be clear in the discussion of the Deduction chapter here and in the
discussion of the Paralogisms in Chapter 7 that Kant's objection to the cogito
is not only that it cannot sustain claims about the soul's simplicity or immortality.
Kant's critique cuts to the heart of the cogito. Like Hume and unlike Descartes,
Kant does not believe that we can introspect a self.
5. This reading has been so popular that I will not try to list all its adherents.
Strawson certainly holds it and it is presupposed by Jonathan Bennett and Paul
Guyer. Both have argued that Kant is simply wrong in believing that conscious-
ness requires self-consciousness, so the central tenet of the deduction must be
abandoned. See Bennett, op. cit., p. 105, and Guyer, "Kant on Apperception,"
op. cit. In an earlier paper ("Kant's Real Self," in Wood, op. cit.) I tried to
figure out why Kant held this view. I now believe that he does not offer any
good arguments in favor of this position because he does not hold it.
6. Strawson, op. cit., p. 98.
7. Wilkerson, op. cit., p. 52.
8. Ralph C. S. Walker, Kant (London: Routledge and Kegan Paul, 1978),
p. 80.
9. Scruton, op. cit., p. 32.
10. In the Bounds of Sense, op. cit., Strawson also suggests that there is an
argument about intuitions and concepts—about seeming and being—that is re-
quired for self-ascription (see pp. 100-101). This point has not been as popular
among subsequent interpreters, however.
11. Ibid., p. 104.
12. See Chapter 1, "In Defense of Transcendental Psychology."
 Notes 249

13. In criticizing Strawson's reconstruction on interpretive grounds, I may be

thought to be missing the point: A reconstruction is not meant to be an inter-
pretation. The problem with this move is that, as will be clear in the text,
Strawson often defends his reading as if it were meant as an interpretation and
it is often taken to be an interpretation. Thus, I think it is reasonable and
necessary to criticize it as an interpretation. For an opposing view, see Chris-
topher Janaway, "History of Philosophy: The Analytic Ideal I," in Proceedings
of the Aristotelian Society Supplement LXVIl (1988): 169-89.
14. For example, he uses gehoren at A116 and A117a, at B132 he uses ge-
horen, angehoren, and gehorig. At B134, he again uses gehoren.
15. See Prolegomena, op. cit., pp. 36-40. See also B69-70, and the note
to Bxl.
16. See his "Self-Ascription and Objectivity," Philosophia: Philosophical
Quarterly of Israel 10 (1981): 189-98.
17. See Chapter 5, "The Problem of Self-Consciousness." S. C. Patten also
denies that apperception is primarily about self-ascription in "Kant's Cogito,"
in Kant-Studien 66 (1975): 331-41. Richard Aguila registers some doubts about
the importance of self-ascription in Matter in Mind (Bloomington: Indiana Uni-
versity Press, 1989).
18. Wilkerson, op. cit., p. 52.
19. Allison, Kant's Transcendental Idealism, op. cit., p. 144.
20. Hector-Neri Castaneda has pursued a project that is sometimes described
as a formal study of the "I." I think it is more clearly described as an account
of the peculiarities of indexical reference. See, for example, "He: A Study in
the Logic of Self-Consciousness," Ratio 8 (1966).
21. "Formal" is sometimes contrasted with "empirical." Since empirical ap-
perception is understood as a doctrine in empirical psychology, this contrast is
also ultimately grounded in the contrast between "formal" and "psychological."
22. Among the commentators that hold this view are Strawson, op. cit.;
Bennett, Kant's Analytic, op. cit.; Henrich, Identitat und Objektivitat: Eine Un-
tersuchung uber Kants transzendentale Deduktion (Heidelberg: C. Winter);
Wolff, op. cit.; and Guyer, op. cit. Allison, op. cit., is a rare exception.
23. I explain this dependency more fully in "Synthesis and Apperception"
and "Arguing for the Synthetic Unity of Apperception."
24. Treatise, op. cit., Part IV, Chapter 2.
25. See Chapter 6, "How the Argument Fails."
26. As I argue in Chapter 7, "Leibniz and the Simplicity of the Soul," the
Second Paralogism appears to be a critique of Leibniz's argument for the sim-
plicity of the soul. See Margaret Wilson's "Leibniz and Materialism," Canadian
Journal of Philosophy 3 (1974): 495-513.
27. April 1776, as printed in Beattie Works, op. cit., Essays, vol. I. See Kemp
Smith's prefatory comments about the importance of this discovery in his Com-
mentary, op. cit., pp. xxviiiff.
28. Ibid., p. 207. See also the note, p. 207.
29. Robert Paul Wolff, "Kant's Debt to Hume via Beattie," Journal of the
250 NOTES
History of Ideas 21 (1960): 117-23.1 read Wolffs paper after discovering Kemp
Smith's error myself. Oddly, his recognition that Kant was aware of Hume's
devastating critique of Descartes' view of the self did not seen to affect his
reading of the Deduction. He offers four different reconstructions of the De-
duction argument (on pp. 116,119,132,161) and each one begins with a premise
that all my representations are bound up in a unity—a premise that begs the
question against Hume.
30. I checked the sixth edition against the second, and the only difference is
that "Hume" is used in the second and "our author" in the sixth. The citations
that follow are from the sixth edition.
31. Kuehn, op. cit., pp. 92-93, 92-93n.
32. Tetens, op. cit., pp. 392-94. As noted earlier, Kant read Tetens' book
with care (Chapter 2, note A). The citation is from p. 393. Lewis White Beck's
discussion in Early German Philosophy drew my attention to this book (Cam-
bridge, Mass.: Harvard University Press, 1969), p. 419. (However, his refer-
ences for Tetens' discussion of Hume on p. 419n are incorrect.)
33. Beattie, op. cit. These excerpts appear on pp. 79,249-50. The excerpted
passages appear on pp. 251, 252, and 253 of L. A. Selby-Bigge's edition of the
Treatise, op. cit.
34. See Treatise, op. cit., pp. 261, 633ff.
35. Inquiry, op. cit., p. 64.
36. Treatise, op. cit., p. 636.
37. Ibid., p. 634.
38. There is one sentence in the Inquiry that seems to contradict this inter-
pretation. Hume appears to analyze the causal relation in terms of existential
dependence:
we may define a cause to be an object followed by another, and where all
the objects, similar to the first, are followed by objects similar to the second.
Or, in other words, where, if the first object had not been, the second never
had existed [Inquiry, op. cit., p. 87].
If causal relations connect the states of one mind and involve existential de-
pendence, then my account must be rejected. I do not think that this passage
can be taken at face value, however. Either Hume did not take "in other words"
to imply material equivalence, or he did not understand that the two analyses
flanking this phrase are not materially equivalent, or somehow he intended the
second analysis to be a notational variant of the first. Since the first two options
make Hume either sloppy or obtuse, the third is preferable. I think that he
intended the first statement to present the constant conjunction analysis and
that he intended the disputed statement to be understood not subjunctively, but
as a summation of past experience that reinforces the constant conjunction
analysis: we say "cause" where we have never encountered a case where, if the
first object had not been present, the second object had been present.
39. Treatise, op. cit., p. 635. I have changed Hume's archaic spelling.
40. Barry Stroud (Hume (London: Routledge and Kegan Paul, 1977),
 Notes 251
pp. 131-40) notes that the text is ambiguous between two possible concerns.
Hume believes either that he cannot explain some actual tie that binds the states
of one mind together, or that he cannot explain what leads us to think that
perceptions are connected in individual selves. While the text allows either, the
systematic considerations I raise in the text show that the former hypothesis is
correct. Although I disagree with Stroud's ultimate position, my discussion is
indebted to his. See also Don Garrett's interesting analysis in "Hume's Self-
Doubts About Personal Identity," Philosophical Review XCV (1981): 337-58.
41. My treatment of this issue is somewhat different here than in "Kant on
Self-Identity," Philosophical Review XCl (1982): 41-72. In the paper I assumed
that Hume's explanations of mental activity themselves implied the need for
existential dependence among cognitive states. Here my claim is only that his
account requires the soundness of reference to individual selves.
42. Inquiry, op. cit., p. 44.
43. At A121, Kant claims that without a productive imagination cognitive
states would only be reproduced as they came together, and this would produce
only "regellosse haufen derselben" [lawless heaps of them]. Kemp Smith's trans-
lation of "accidental collocations" is misleadingly genteel.
44. Chapter 1, "Transcendental Psychology."
45. Allison, op. cit., p. 140. I should note, however, that his discussion of
this whole passage differs somewhat from mine.
46. See the discussion of these various senses of "a priori" in Chapter 1,
"Transcendental Psychology."
47. Robert Pippin, "Kant on the Spontaneity of Mind," Canadian Journal
of Philosophy 77(1987): 449-76,459. Specifically, Pippin objects that the process
of synthesis could yield a further state, which I claim to be a representation of
an object, even though the subject does not "take" a presented object as m,
for example (p. 468). Since I read Kant as claiming that what it is for a subject
to represent something in a certain way, either in perception or by concepts,
to "take" it in a certain way, is to construct a representation of it through acts
of synthesis, I don't see how there can be a resulting state without a "taking."
48. I return to this issue briefly in Chapter 5.
49. See Richards, op. cit., p. 237, note 10, and Blackwell, op. cit., p. 340.
50. Loemker, op. cit., pp. 637, 638, 645-46.
51. Ibid., p. 638.
52. Gregor, op. cit., p. 9 (AA VII: 127).
53. Pippin, op. cit., p. 460.
54. Pippin suggests that to remember or assert it must be possible to be aware
that that one is asserting or remembering. I assume that this means that one
could consciously attend to the state and recognize what it is. Although this
seems plausible, I do not agree with his further claims that engaging in such
tasks requires one to be aware of one's entire mental history. As he notes, this
is a rather controversial issue and requires more defense than he offers.
55. See the discussion in Chapter 3, "Representing Objects."
56. Gregor, op. cit., p. 16 (AA VII:135).
252 NOTES
57. In light of this retraction, and others (e.g., B134), it is puzzling that Dieter
Henrich makes A108 (which asserts that we observe acts of synthesis, and that
it is only because we observe such acts that we know the identity of the self)
as the foundation of his analysis of the deduction. See Henrich, op. cit.,pp. 81ff.
58. Richard Aquila offers the following objection to my functionalist reading
of Kant:
"functionalism appears to get things backward. This is because, for Kant,
causes and effects, at least insofar as they are possible objects of knowledge,
are essentially governed by systems of causal laws. So according to the
functionalist approach, the very idea of conceptual content would have to
presuppose the idea of a system of causal laws.... But in Kant's own think-
ing, the very idea that a system of laws naturally obtains seem to be deriv-
ative from the idea of a being who is capable of representing determinate
sorts of objects" [Matter in Mind: A Study of Kant's Transcendental De-
duction (Bloomington: Indiana University Press, 1989), p. 31].
My claim is not that the idea of content presupposes the idea of a system of
causal laws. I take Kant to be offering a substantive claim that cognitive states
cannot have content in the absence of synthetic connections. As I note in the
text, these discussions of synthesis do not presuppose the strong notion of
"cause" defended in the Second Analogy. Indeed Kant uses a wealth of syn-
onyms—produce, bring forth, generate—to avoid hanging anything on the
expression "cause." In "Kant's Functionalism," op. cit., Ralf Meerbote sup-
ports a functionalist reading of many of Kant's discussions.
59. See Chapter 3, "Associationism."
60. Tetens, Phllosophische Versuche, vol. I, first two passages, p. 17, third,
p. 20.
61. See Dennett, op. cit., p. 122.
62. See Chapter 3, "Representing Objects."
63. Remnant and Bennett, op. cit., p. 54.
64. Gregor, op. cit., p. 16 (AA VII:135).
65. Remnant and Bennett, op. cit., pp. 53-59.
66. This is Kant's estimation. Leibniz thought that all perceptions had some
effect on all subsequent perceptions. Ibid., p. 114. Given the myriad of petites
perceptions, however, this view seems somewhat farfetched.
67. Kant Philosophical Correspondence 1759-1799, edited and translated by
Arnulf Zweig (Chicago: University of Chicago Press, 1967), pp. 153—54 (AA
XI:52).

Chapter 5
1. This appears to be one occurrence of "apriori" where it indicates temporal
priority. See the discussion of Kant's more standard uses in Chapter 1, "Tran-
scendental Psychology."
2. An important exception to this trend has been Wilfrid Sellars. He dis-
 Notes 253
cussed Kant's views in "this I or he or it (the thing) which thinks," op. cit., and
in "Metaphysics and the Concept of a Person," op. cit.
3. At least this assumption was not obvious to me. See "Kant on Self-
identity," op. cit., pp. 54ff and "Kant's Real Self," op. cit., pp. 117-18.
4. Here I do not mean to suggest that Hume was the thoroughgoing skeptic
he is often made out to be, but only that he is a skeptic about personal identity.
5. Henry E. Allison ("Kant's Refutation of Materialism," The Monist 72
(1989): 190-208), Robert B. Pippin "Kant on the Spontaneity of Mind,"
op. cit.), and Richard Aquila (op. cit., p. 32) all object that earlier versions of
my account of apperception do not do justice to the link between apperception
and spontaneity. I considered Aquila's basic objection to my functionalist read-
ing in note 58 of Chapter 4.
As Allison observes (p. 192), Kant links spontaneity and apperception quite
strongly, when he claims that the cognitive state with content "I think" is an
act of spontaneity, and, as such, cannot be regarded as belonging to the senses
(B132). At the very beginning of the B Deduction, however, Kant explains that
combination is an act of spontaneity, and so does not belong to the senses, but
to the understanding (B130). In characterizing the unity of apperception as the
outcome of acts of synthesis (or combination), I present it as involving spon-
taneity. The real disagreement is about the kind of spontaneity. I take it to be
a "relative spontaneity," whereas both Allison and Pippin think it is absolute.
There is a decisive objection to interpreting apperception in terms of absolute
spontaneity, however. If the doctrine of apperception, or synthesis, implied
absolute spontaneity, then Kant would have a proof of transcendental freedom
in the First Critique, something he manifestly does not believe that he has.
Although he might have been tempted by this sort of move at various times,
his position in the Critique is clear. The Critique allows for the possibility of
freedom by establishing the phenomenal—noumenal distinction (Bxxvi). Pippin
sees the problem and offers a reply: "Proving that reason must be assumed
spontaneous in one context [the epistemic context], does not prove that in other
contexts... it must be spontaneous too" (p. 473). This rebuttal is too weak,
however. Absolute freedom in thinking does not entail absolute freedom in
acting, but it would be a giant step in the right direction. If the universal sway
of determinism is once breached, then the plausibility of other exceptions would
increase dramatically, as Kant would fully appreciate.
6. See Content and Consciousness (London: Routledge and Kegan Paul,
1969), pp. 93-96.
7. See Chapter 7, "Understanding the First Paralogism."
8. See Chapter 5 of Perry's dissertation, Identity (Ann Arbor, Mich.: Uni-
versity Microfilms) and the Introduction to his anthology, Personal Identity (Los
Angeles: University of California Press, 1975), pp. 7-12.
9. In a well-known article, Williams argued that this would occur if two
people had good claims to remember all the events of Guy Fawkes life. See
Williams, "Personal Identity and Individuation," Proceedings of the Aristotelian
Society LVIl (1956-57).
10. See Essay, op. cit., p. 454.
254 NOTES

11. As noted in Chapter 2, Kant probably read this work in 1769. For two
interesting discussions of many issues in this debate that I do not touch on, see
Edwin Curley, "Leibniz and Locke on Personal Identity," in Michael Hooker,
ed., Leibniz: Critical and Interpretive Essays (Minneapolis: University of Min-
nesota Press, 1982), pp. 302-26), and Nicholas Jolley, Leibniz and Locke: A
Study of the "New Essays on Human Understanding" (Oxford: Clarendon Press
1984), Chapter 7.
12. Remnant and Bennett, op. cit., p. 114.
13. See Chapter 3, "The Law of Association."
14. As Kant introduces the relation of synthesis, it covers any case where
elements of earlier states are comprehended in a later state. Thus, memory
would also be an example of a synthetic connection between states, since the
contents of the later depend of those of the earlier state. So would the law of
association. By contrast transcendental syntheses are those syntheses governed
by nonassociative rules that are necessary for knowledge. Is the unity of ap-
perception produced by syntheses or only by transcendental syntheses? This
distinction collapses for Kant, because he makes the substantive assumption
that the categories are the only possible rules for combining elements of cognitive
states. Since I not think that his arguments warrant this assumption, I believe
that the sound part of the doctrine of apperception is that apperception is
produced by the synthetic connections yielded by syntheses necessary for knowl-
edge. I am grateful to Henry Allison for drawing my attention to this possible
source of confusion.
15. "Discourse on Metaphysics," para. 34. Loemker, op. cit., p. 325.
16. This is clear in the following citations from the Essay: "...per-
son ... stands for a thinking, intelligent being, that has reason and reflection,
and can consider itself as itself, the same thinking thing, in different times and
places; which it does only by that consciousness which is inseparable from
thinking...""... for since consciousness always accompanies thinking, and it
is that which makes everyone to be what he calls self, and thereby distinguishes
himself from all other thinking things, in this alone consists personal identity,
i.e., the sameness of a rational being: and as far as this consciousness can be
extended backwards to an past action or thought, so far reaches the identity of
the person...," op. cit., pp. 448-49.
17. See "Discourse on Metaphysics," para. 34, in Loemker, op. cit.,
p. 325; Jolley, op. cit., pp. 134—35; and Robert McRae, Leibniz: Perception,
Apperception, and Thought (Toronto: University of Toronto Press, 1976),
pp. 44-45.
18. See Margaret Wilson, "Leibniz, Self-Consciousness, and Immortality: In
the Paris Notes and After," Archiv fur Geschichte der Philosophic, Sonderheft
58 (1976): 335-52.
19. See Remnant and Bennett, op. cit., pp. 236-37 and 114 and the discus-
sions of Jolley and McRae cited in note 17.
20. Tetens, op. cit., for example, pp. 194-98.
21. For example, at A104 and B134.
 Notes 255
22. I discuss this trend at greater length in "Kant's Patchy Epistemology,"
op. cit.
23. Kant makes some remarks about children becoming self-conscious in the
Anthropology (op. cit., pp. 9-10; AA VII: 127-28), but nothing of philosophical
interest. In the second edition of the Paralogisms chapter, he notes that during
life others can use the body to ascribe experiences to a continuing person (B415).
24. In his introduction to John Perry, ed., Personal Identity (Los Angeles:
University of California Press, 1975), see pp. 11-12.
25. Although he has been a major figure in this debate, I do not discuss the
views of Bernard Williams in the text. One reason is that his most important
contribution—the argument that mental continuity criteria must be rejected
because they conflict with the law of the transitivity of identity—has been re-
futed, as noted in "The Issue" section. Williams has offered other arguments,
but these turn on what people would say or feel about various situations. As I
suggest in the text, this does not seem to me a very fruitful approach. I have
offered specific criticisms of Williams's more recent views in "Being Selfish
About Your Future," Philosophical Studies 32 (1977): 425-31.
26. See H. P. Grice, "Personal Identity," reprinted in John Perry, ed.,
op. cit., pp. 73-95; Anthony Quinton, "The Soul," also reprinted in Perry;
Terence Penelhum, "Personal Identity, Memory, and Survival," Journal of
Philosophy (1959); Sydney Shoemaker, Self-Knowledge and Self-Identity, (Ith-
aca: Cornell University Press, 1963) and "Persons and Their Pasts," American
Philosophical Quarterly (October, 1970); John Perry, "Can the Self Divide?"
Journal of Philosophy (September, 1972): 463-88; and David Lewis, "Survival
and Identity," in Amelie Rorty, ed., The Identities of Persons (Los Angeles:
University of California Press, 1969), pp. 17-40.
27. Shoemaker, op. cit., pp. 22-35.
28. As noted in Chapter 4, synthetic connection is a stronger relation than
causation, if causation is not held to entail existential dependence. I do not
think this difference is significant, since few people would maintain a Humean
analysis of causation.
29. For example, David Lewis, op. cit. See also the discussion of Parfit in
"Parfit's Denial of Personal Identity."
30. One reason to think that there is very little similarity of belief is that
small children appear to use quite different concepts from adults. See Susan
Carey, Conceptual Change in Childhood (Cambridge, Mass.: MIT Press 1985).
31. Perry, Personal Identity, op. cit., p. 7.
32. Curley notes that Locke was not concerned with the proper usage of the
expression "same person" either, despite the tendency to read his discussion in
the light of contemporary linguistic philosophy. Curley, op. cit., pp. 310-314.
33. I defend this point at greater length in "Natural Kinds and Unnatural
Persons," Philosophy 54 (1979): 541-47. David Wiggins expresses a view very
like the one I defend in the text, although he casts it in terms of the "concept"
'person'. See "Locke, Butler, and the Stream of Consciousness: And Men as a
Natural Kind," in Rorty, ed., op. cit., pp. 139-73.
256 NOTES
34. Both Perry and Wiggins make this point. See Perry, "The Importance of
Being Identical," in Rorty, ed., op. cit., pp. 67-90, and Wiggins, op. cit.
35. See Sameness and Substance (Cambridge, Mass.: Harvard University
Press, 1980), pp. 163ff.
36. In "Locke, Butler, and the Stream of Consciousness: And Men as a
Natural Kind," op. cit., p. 158.
37. "Personal Identity," Philosophical Review 80 (January 1971): 3-27;
"Later Selves and Moral Principles," in A. Montefiore, ed., Philosophy and
Personal Relations (London: Routledge and Kegan Paul, 1973); and Reasons
and Persons (Oxford: Clarendon Press, 1984).
38. See Reasons and Persons, op. cit., p. 206.
39. Ibid., Chapter 15.
40. See Curley's useful discussion of this matter, op. cit., pp. 318-19.
41. I should note that this criterion does not tell us whether what we think
of as continuing minds, namely, continuing human beings, actually are continu-
ing minds. It simply states the facts that are relevant.
42. Parfit suggests briefly that a position like Kant's might undermine his
critique of absolutism. However, he reads Kant through Strawson's eyes, and
so believes that the issue concerns the necessity of self-ascription of mental
states for knowledge. He says that he is unsure how to respond to this type of
abstract argument. Ibid., p. 225. Since I do not think this position is Kant's, I
do not consider it in the text.
43. Christine Korsgaard argues that Parfit's claim that identity does not matter
can be undermined by looking at Kant's theory of agency. See "Personal Identity
and the Unity of Agency: A Kantian Response to Parfit," Philosophy and Public
Affairs (in press).
44. See A363-64n and B415-18n and Chapter 7.
45. See Chapter 4, "Intuitions."
46. This does not violate the condition noted in "Refining the Account of
Synthetic Connection," that synthetic progenitors cannot contribute to the con-
tents of a product state through an additional outer sensation.
47. See " Leibniz Versus Locke" section.
48. In the note to the Third Paralogism where he raises the possibility of a
series of substances, Kant describes the situation in terms of one substance
communicating its motion to later ones (A363-64a). So he may think of a "trace"
as some sort of permanent motion induced in a material or immaterial medium
by cognitive states.
49. For a discussion of this view, see Alan Baddeley, Working Memory (Ox-
ford: Clarendon Press, 1986), pp. 19-29.
50. See Jerry A. Fodor, The Modularity of Mind (Cambridge, Mass.: MIT
Press, 1983); David Marr, op. cit.; Zenon Pylyshyn, "Cognition and Compu-
tation: Issues in the Foundations of Cognitive Science," Behavioral and Brain
Sciences 3 (1978): 111-32; and Michael Posner, Chronometric Explorations of
Mind (Hillsdale, N.J.: Erlbaum, 1978).
51. See Pylyshyn, op. cit.
52. Fodor, op. cit., Part 4.
 Notes 257
53. See Chapter 1, "In Defense of Transcendental Psychology."
54. Although this is sometimes put in terms of a two-self theory, Kant's view
is better understood as claiming two perspectives on the self. This point is made
forcefully by, among others, Erich Adickes in Kants Lehre von der Doppelten
Affektion unseres Ich as Schlussel zu seiner Erkenntnistheorie (Tubingen, 1929),
p. 3, and Karl Ameriks, op. cit., p. 266.
55. Sellars suggests something very like this at the end of his Presidential
Address (" . . . the I, He, or I t . . . , " op. cit.).
56. See Strawson, op. cit., pp. 235-39, 247-49.
57. Ibid, p. 249. Strawson casts this point in terms of noumenal objects ap-
pearing to the noumenal self. Do they appear in time or out of time? Since the
I of apperception must be phenomenal, I recast his point in terms of the synthetic
processes themselves.
58. Strawson, op. cit., p. 249.
59. The obvious way to avoid the inconsistency is to claim that the doctrine
that time is phenomenally real and transcendentally ideal does not amount to
a denial that time is real. I believe that the inconsistency is hopeless, because
I do not see how this strategy can produce a doctrine that is both coherent and
Kantian.
60. See Strawson, op. cit., pp. 15, 20-22, 32, 38-42, 247-50.
61. The principal evidence for this claim is Kant's paper on "Regions of
Space." He notes first that the endeavors of philosophers to settle the question
of whether space can exist independently of all matter have been "futile." He
then observes that the one attempt to settle this issue by an a posteriori proof
(Leonhard Euler's Reflexions sur I'espace et le temps) also ended in failure. Kant
goes on to present his own a posteriori proof, the argument from incongruent
counterparts. He is not very sanguine about his own results, however, and notes
that there are many difficulties with this concept, if we really try to understand
it. See Handyside, op. cit., pp. 20-21, 29.
62. This point was recognized in Trendelenberg's famous objection about the
neglected alternative. Since space and time might be both forms of intuition and
independently real, it is not sufficient to argue from the fact that they are forms
to the fact that they are not real. See Kemp Smith's discussion in his Commen-
tary, op. cit., pp. 113-14.
63. In his recent book (op. cit., pp. 354ff), Paul Guyer argues that this con-
clusion does follow. He regards the argument for space and time as forms of
intuition as ruling out the possibility that they are real. According to Guyer,
this is Kant's reasoning. We have a priori knowledge of space (i.e., geometry).
This is possible because space is the form of intuition. *But if something can
be known a priori, then it cannot be a real determination of things. Therefore,
space and time are not real. The crucial claim is the one I have starred. As
Guyer recognizes, his argument goes through only if what we know a priori has
a certain logical form. Specifically, if what we know is not:
(A) n K we are to perceive an object, then it is spatial and Euclidean,
but:
258 NOTES
(B) If we perceive an object, then Q it is spatial and Euclidean.
What (A) means is that in any possible world in which we perceive something,
that thing is spatial; conversely, (B) means that if we perceive something (in
the actual world), then in all possible worlds that thing is spatial. In the standard
terminology, (B) asserts a de re modality. Although I disagree with other as-
sumptions of Guyer's analysis (see later), the simplest way to put my objection
is that de re modalities do not fit naturally into Kant's philosophy. For Kant,
something is necessary (and universal) if it is true of all those worlds of which
we can have experience, constituted as we are. Unlike Leibniz, Kant does not
operate with an absolute notion of necessity. (See Philip Kitcher, "Kant on the
Foundations of Mathematics," op. cit.) So for Kant, Guyer's second interpre-
tation (B) should really be: If we perceive an object, then, in all those worlds
in which we can have experience, it is spatial and Euclidean. This is not quite
the same as (A), since there might be objects in the actual world which we do
not happen to perceive in other possible worlds, even though we can experience
objects in those world constituted as we are. Still this difference does not seem
significant, since we could experience the object.
Guyer's analysis also presupposes that geometry is the central concern of the
Aesthetic. This difference is far-reaching. While I recognize that Kant was
concerned with the epistemic status of geometry, I take the driving force behind
the ideality doctrine to be the scientific and metaphysical grounds I note in the
text. See Chapter 2, "What the Transcendental Aesthetic Is About" and "The
Transcendental Exposition" for further discussion. Despite this fundamental
disagreement, I do agree with Guyer on one important point—Kant did not
simply neglect the possibility that space and time might be both real and some-
thing else; he was antecedently convinced that they could not be real.
64. Strawson raises a crisply formulated attack on transcendental idealism
and transcendental psychology. However, there is also a common, but vaguer
worry about coherence that might be provoked by my suggestion that the doc-
trine of apperception be regarded as phenomenal. According to transcendental
idealism, the world appears to us as it does partly because of the ways in which
our senses and higher faculties are constituted. Apperception concerns our
higher faculties. So, if it is a phenomenal doctrine, then it provides an account
of how our faculties influence the way we think that is itself influenced by the
way we think. (In another context, the objection continues by noting that it
cannot be a noumenal doctrine.) Although this seems circular, it is intrinsically
no more problematic than writing an English grammar in grammatical English.

Chapter 6
1. See Chapter 3, "Representing Objects" and "Making Judgment About
Objects."
2. I discuss how this type of argument is supposed to establish a special role
for the categories further in "Universal Applicability and Objective Validity."
 Notes 259
3. See Chapter 3, "Representations and Concepts."
4. Kant vacillates somewhat in the passage. Perhaps without the categories
we could have intuitions without thought [gedankenlose Anschauung], but in
this circumstance we would lose all reference to objects. Since intuitions are
supposed to have reference to objects (see Chapter 3, "The Problem"), he
seems to be groping in the passage.
5. See Chapter 3, "The Synthesis of Intuitions."
6. Interestingly, Paton thought that it was a reasonable interpretation of the
text to see Kant as concerned with what it is for different states to be states of
one self. He rejected this reading only because he believed that taking self-
identity to consist in the interrelation of cognitive states destroys knowledge,
because it does not permit the unity of the subject. Op. cit., vol. I, p. 406.
7. As Kant notes, even though General Logic is general, it does not create
the contents of any cognitive states but only relates contents that have already
been created (A56/B80).
8. Despite the parenthetical aside, Kant clearly does not equate objective
validity with truth. As he says, there is no criterion of truth, and only a fool
looks for one (A58-59/B82-84). His point is that we cannot have true cognitions,
without having cognitions. Certain conditions are required for cognition, how-
ever, and those conditions are objectively valid.
9. William Porterfield, op. cit., vol. Ill, p. 329.
10. (Leipzig, 1787-1796), 6 vols, cited in Hatfield, op. cit. See Hatfield for
a fuller discussion of the standard view.
11. Gregor, op. cit., p. 35 (AA VII: 156-57).
12. Thomas Reid, An Inquiry into the Human Mind, edited by Timothy
Duggan (Chicago: The University of Chicago Press, 1970), for example, p. 236.
13. Kuehn cites Eberhard's views, op. cit., p. 106. Among psychologists,
versions of this position have been held by Helmholtz, Jerome Bruner, and
Richard Gregory. Among recent philosophers, D. M. Armstrong and George
Pitcher have both proposed belief theories of perception.
14. Essay, op. cit., p. 191.
15. Etienne Bonnot de Condillac, An Essay on the Origin of Human Knowl-
edge (1758), a facsimile reproduction of the translation of Thomas Nugent, with
a introduction by Robert G. Weyant (Gainesville, Fl.: Scholars' Facsimiles &
Reprints, 1971), p. 27. (I have updated the spelling.)
16. See Chapter 3, "Representations and Concepts."
17. Paul Guyer makes A99 central to his interpretation of the argument of
the deduction. See Guyer, op. cit., pp. 89,109,121-22,148,151,157,171,178,
207, 211, 255, 256, 289, 299, 301, 302, 347, 456. However, he never considers
that this passage is intended as preparation for the discussion at A119-20.
18. See Reid, op. cit., pp. 46-47; Condillac, op. cit., p. 67; Tetens, op. cit.,
for example, pp. 263ff.
19. Gregor, op. cit., p. 19 (AA VII: 138).
20. These notes were from metaphysics lectures given some time between
1773 and 1785, but probably closer to the latter end of the period. They were
260 NOTES
published under the title, Immanuel Kants Vorlesungen fiber die Metaphysik,
by K. H. L. Politz in 1821. For a discussion of the history of these notes and
their reception, see Satura, op. cit., pp. 7-20.
21. Cited in Satura, op. cit., pp. 114-15, my translation.
22. Satura cites a letter in which Kant worries that even capable students
select what to write down and what to omit and so do not present a very accurate
picture. Op. cit., p. 5.
23. Kant's position here is very close to Condillac's. In explaining how we
are able to attain the idea of an extended thing, Condillac writes:
[The eye] cannot grasp the whole of the simplest shape until it has analyzed
it, that is to say until it has noticed successively all its parts. It must make
a judgment on each individual part, and another judgment on the whole
of them together. It must say: here is one side, bounded by three sides,
and from it results this triangle.
Condillac, op. cit., p. 68. Kant may also have derived some support for this
view from his reading of Burke. In A Philosophical Enquiry into the Origin of
Our Ideas of the Sublime and the Beautiful (Oxford: Basil Blackwell, reprinted
1967), Edmund Burke raises the possibility that:
there is but one point of any object painted on the eye in such a manner
as to be perceived at once; but by moving the eye, we gather up with great
celerity, the several parts of the object, so as to form one uniform piece
(p. 137).
Rolf George's article (op. cit.) led me to these interesting discussions in Con-
dillac and Burke.
24. For a recent summary of musical perception, see Diana Deutsch, "Au-
ditory Pattern Recognition," in K. R. Boff, L. Kaufman, and J. P Thomas,
eds., Handbook of Perception and Human Performance, vol. II (New York:
Wiley, 1986), Chapter 32. Jay Rosenberg discusses this issue in The Thinking
Self (Philadelphia: Temple University Press, 1986), pp. 226, 236.
25. See Chapter 3, "Representations and Concepts."
26. In Kant's Analytic, op. cit., Jonathan Bennett dismisses the idea of pro-
viding an account of concept application as a fool's errand. This view was
probably inspired by the antipsychological direction set in philosophy of mind
by Ryle and the later Wittgenstein, and I doubt that Bennett still holds it. There
can be silly accounts of concept application that lead to infinite regress, but
there is nothing wrong with the project and nothing wrong with Kant's view.
A regress threatens only if the proposal is that we consciously apply concepts
by consciously applying others. On Kant's view, however, these syntheses are
carried out unconsciously, by the imagination, a "blind but indispensable func-
tion of the soul." For further discussion see Chapter 8. In the text I am delib-
erately noncommittal about how we represent concepts—by list of features,
prototypes, or whatever.
27. In 11 24, Kant describes the figurative synthesis of the imagination. This
 Notes 261
synthesis applies to sensory representations but is directed by the categories of
the understanding. This discussion seems to anticipate the results of f 26, which
will be described later. Roughly, it anticipates Kant's argument that the functions
which enable us to perceive spaces and times are identical with the syntheses
associated with the categories. The idea that figurative syntheses are directed
by the categories suggests that the imagination is subordinate to the understand-
ing. On the other hand, a coordination model seems to be presupposed in a
discussion of following the thread of conversation in the Anthropology, op. tit.,
p. 52 (AA VIL177):
Whether in silent thought or in conversation, there must always be a theme
on which the manifold is strung, so that understanding too must be operative
in it. In such a case the play of imagination still follows the laws of sensibility,
which provides the material, and this is associated without consciousness
of the rule but still in keeping with it. So the association is carried out in
conformity with understanding, though it is not derived from the under-
standing.
28. "The Proof-Structure of Kant's Transcendental Deduction," Review of
Metaphysics 22 (1969): 640-59.
29. Gregor, op. cit., p. 16 (AA VII:135).
30. In his most formal account of the relations among the faculties, Kant
claims that "perception" is the genus for "intuition" and "concept" (A320/
B376-77). Yet elsewhere he describes "perceptions" as "representations ac-
companied by sensation" (B147), which would mean that concepts were accom-
panied by sensation. Although the terminology is somewhat fluid, I think he
intends "intuition" to cover both perceptions and intuitions that we do not
notice but have theoretical reasons for assuming.
31. "Apprehension" was changed to "apperception" in the 4th edition.
32. Although this is a minimal sense of "perceive," it is somewhat more
restrictive than Kant's claim that the categories apply to "anything that can be
presented to our senses" (B160). Still, I take the opening claim to provide only
a rough idea of what Kant has in mind. He must intend the argument to be
about objects we perceive; otherwise there would be no point in considering
the synthesis of apprehension required for perception.
33. Is the mention of geometrical proof meant to establish that we have
intuitions of spaces and times, or merely to illustrate the fact? Some, Allison,
for example (op. cit., pp. 97-98), have taken the focus of this section to be a
special intuition involved in mathematical proof. That interpretation is incon-
sistent with two points in Kant's discussion, however. He illustrates the abstract
point of the section with examples of perceiving a house and the freezing of
water. In conclusion, he refers to "everything that is represented as determined
in space and time" (B161). Both these points indicate that Kant's topic is not
the perception of pure intuitions, but of particular spaces and times. Further,
this reading appears to be inconsistent with the passage at B137 just cited.
34. I am grateful to Henry Allison for pointing out that even though it is not
262 NOTES
a Critical doctrine, the distinction between judgments of experience and judg-
ments of perception provides some confirmation of my account of what is going
on in B 1 26.
35. Op. cit., p. 41.1 change the singular "perception" to "perceptions," since
Kant uses the plural (AA IV:298).
36. See Chapter 1, "Transcendental Psychology," where I consider the dif-
ferences between Kant's position and simple nativism.
37. In Kant's Analytic, Bennett suggests that the way to understand Kant's
claim that the categories apply to all experience is that at least one category
applies to every experience (or cognition). Op. cit., pp. 76-83.
38. See Chapter 3, "The 'One-Step' Deduction."
39. See the passage to which note 7 is appended.
40. See the discussion of indirect synthetic connection in Chapter 5, "Is It
Too Strong?"
41. I am grateful to Philip Kitcher for several helpful discussions about this
issue.
42. Even David Marr, who offers a "bottom-up" account of perception has
concepts involved in the final stages of perception. See, Vision, op. cit., Chap-
ter 5.
43. See "The Centrality of 11 26."
44. Henrich, op. cit., p. 646.
45. Ibid., p. 652.
46. See Allison, Kant's Transcendental Idealism op. cit., pp. 160-72.
47. Ibid; see especially the discussions on pp. 162, 166.
48. Ibid., pp. 170-71.
49. Ibid., p. 167-70.
50. Walsh, op. cit., p. 70.
51. Guyer, Kant and Knowledge, op. cit., pp. 208, 215.
52. DeVleeschauwer, op. cit., p. 82.
53. Guyer, op. cit., pp. 237-49.
54. I should note that this argument shows only that it is necessary to use a
concept of causation that includes necessity. In the Second Analogy, Kant does
not consider either the degree of universality involved in causal laws or the
sources from which they derive whatever degree of nomic necessity they possess.
These issues are not taken up until the Dialectic and the Critique of Judgment.
55. Guyer, op. cit., p. 258. Allison tries to avoid this difficulty by suggesting
that it is not particular causal rules, but the schema of causality itself that enables
us to determine temporal position (Kant's Transcendental Idealism, op. cit.,
pp. 229-32). The problem with this suggestion is that ordering states of affairs
in time must be a matter of ordering particular states of affairs. How could the
schema of causality help in this project? Seemingly, we can only order particular
states of affairs by appealing to particular causal laws. The laws need not be
ultimate, in the sense that they correctly and exhaustively capture the causal
structure of the world. They could be quite crude.
56. Loc. cit.
 Notes 263
57. It is not clear that the argument Guyer reconstructs from the Refutation
is sound. He assumes that there are no laws governing cognitive states (op. cit.,
p. 307) However, Kant may not have been committed to this view (see Chapter
1, "Kant Against 'Psychology"). Further, even if Kant believed it, it is almost
certainly false. In Chapter 4, "Strawson and the 'Self-Ascription' Reading," I
offered further reasons for denying the Refutation of Idealism a central role in
interpreting the deduction of the categories.
58. Although Locke, Hume, and Reid discuss the perception of time, they
never see this problem. See Locke, op. cit., vol. I, pp. 238-46; Hume, Treatise,
op. cit., pp. 34-35; and Thomas Reid's Essays on the Intellectual Powers of
Man, edited and abridged by A. D. Woozley, (London: Macmillan, 1941), pp.
206-12.
59. William Harper discusses motor detectors in a recent paper on the Second
Analogy. However, he sees them as offering evidence in favor of Kant's position.
I disagree, because, like Guyer, I believe that the basic thrust of Kant's argument
is that information about our states themselves cannot determine order. That
is, the subjective order of our states tells us nothing about the order of states
of affairs. I do not see how this can be squared with motor detectors. They
detect motion precisely by picking up on the sequence of our inner states. See
"Kant's Empirical Realism and the Difference Between Subjective and Objec-
tive Succession," in William Harper and Ralf Meerbote, eds., Kant on Causality,
Freedom, and Objectivity (Minneapolis: University of Minnesota Press, 1984),
pp. 108-37.
60. J. A. Fodor, T. G. Bever, and M. F. Garrett, The Psychology of Lan-
guage (New York: McGraw-Hill, 1974), pp. 249-54.
61. Arthur Melnick, Kant's Analogies of Experience (Chicago: University of
Chicago Press, 1973), p. 84.
62. Probably the best-known early discussion of this phenomenon is William
James. See, for example, Psychology: The Briefer Course (New York: Harper
Torchbooks, 1961), pp. 147-48 (originally published in 1892).

Chapter 7
1. Richards, op. cit., p. 229.
2. Margaret Wilson argues that Leibniz as well as Descartes is a target of
Kant's criticisms in "Leibniz and Materialism," op. cit.
3. See Richards, op. cit., and Blackwell, op. cit., for more detailed dis-
cussion.
4. Sellars discusses this chapter in two papers, "Metaphysics and the Concept
of a Person," op. cit., and "... this I or he or it (the thing) which thinks,"
op. cit. I am happy to acknowledge my indebtedness to these papers, particularly
for the suggestion that the Paralogisms chapter extends the discussion of the
self in the Transcendental Deduction. Although Sellars makes this suggestion,
he does not try to develop it, so his own account of the Paralogisms chapter is
264 NOTES
not really illuminated by his understanding of the connection between these two
texts. I discuss several aspects of his accounts in notes 10 and 20.
5. Jonathan Bennett devotes a paper, "The Simplicity of the Soul," Journal
of Philosophy LXIV (Oct. 26, 1967): 648-60, and three chapters of Kant's
Dialectic (Cambridge, 1974) to the Paralogisms, but I do not think that he is
able to unravel this part of the Critique with his usual dexterity. I raise several
objections to Bennett's analysis in notes 11 and 29.1 also discuss Jay Rosenberg's
analysis in note 10 and touch on Strawson's brief account of the Paralogisms in
The Bounds of Sense (Methuen, 1966) in note 28. Karl Ameriks has offered a
book-length study of the Paralogisms, Kant's Theory of Mind: An Analysis of
the Paralogisms of Pure Reason (Oxford: Clarendon Press, 1982). This offers
some interesting discussions but is very much rooted in the antipsychological
tradition I oppose. I discuss one part of Ameriks' analysis in note 42.
6. I could simply follow Kemp Smith in translating "Vorstellung" as "rep-
resentation" here. Although this would be somewhat less cumbersome, I think
it is helpful to stress the ambiguity of Vorstellung in Kant's philosophy, even
though in this context it is quite clear that he means to indicate the contents of
a cognitive state. In the B edition, he characterizes the thinking self as that
which is "thought" as the subject (B410-11).
7. See Chapter 1, "Transcendental Psychology."
8. See Loemker, op. cit., p. 307.
9. Kemp Smith places the "I's" in quotation marks, even though this is not
indicated in the German. I remove the quotation marks here and throughout,
because they amount to an interpretive addition, by implying that Kant intends
to talk about the representation or symbol "I" rather than about I's.
10. In his Presidential Address "... this I or he or it (the thing) which thinks
...," op. cit., Wilfrid Sellars argues that the central and common mistake of
all the Paralogisms is the attempt to infer properties of I's from properties of
the representation "I." This interpretation is also suggested by Kemp Smith,
op. cit., pp. 457-58. Jay Rosenberg offers a recent version of the Sellarsian
interpretation in " 'I Think': Some Reflections on Kant's Paralogisms," op. cit.
As will be clear below, I think that part of Kant's diagnosis of the errors of the
Paralogisms involves illicit inferences from properties of representations to prop-
erties of things. But I am not convinced that this is the key to the Paralogisms
for two reasons. First, in Rosenberg's version, for example, he thinks that Kant
is scouting the following modal fallacy:
X is [] represented as phi.
Whatever is phi is Q psi.
X is D psi-
The problem with this analysis is that it is not clear that Kant would accept the
first claim. Although there may be necessary properties of certain kinds of
representations, I do not think he believes that any real thing is necessarily
represented in a particular way. I base this on his interesting discussion of the
impossibility of proving the law of contradiction in the Nova Dilucidatio. There
 Notes 265
he claims that we cannot prove the law through symbols, because we presuppose
it in deciding to use symbols. (See F. E. England, Kant's Conception of God
(New York: Humanities Press, 1968), p. 217.) I take the general moral to be
that our decision to employ a certain representation must be based on consid-
erations of its suitability to an object. Thus, whether a certain representation
should be used of an object depends on contingent facts about the properties
of the object. The second objection is related to the first. If this interpretation
is to work, it still needs to provide an account of why Kant believes that the
representation "I" will have certain properties, and, even more fundamentally,
why he believes that we need to use this representation 3t all. My own analysis
concentrates on answering this two questions.
11. Kant's Dialectic, op. cit., p. 75. Bennett thinks that the interesting part
of the minor premise comes in the final clause, "this representation of myself
cannot be employed as predicate of any other thing." By Bennett's lights, this
clause means that the term "I" is "irreducibly substantival;" further, Kant
endorses this part of the "Rationalist" position because he has grasped the
genuine insights of the "Cartesian basis." Both these interpretive notions, the
"irreducibly substantival" and the "Cartesian basis," are elusive, but I think
Bennett has something like the following in mind. The "Cartesian basis" is the
position of methodological solipsism, that is, the view that when contemplating
the body of knowledge we have, we (or I) must recognize that all the knowledge
we (or I) have is founded upon knowledge or our own mental states. According
to Bennett, Kant grasped the essential correctness of the Cartesian basis and
thus asserted that "I" is "irreducibly substantival," meaning that any complete
and accurate description of the world must include a referential use of this term.
Kant's point against the Rational Psychologists is that they construe the fact
that I and my mental states play a special epistemological role, so that I am
something like an epistemological substratum, to mean that I am a special sort
of substance. I resist Bennett's interpretation of the First Paralogism, because
I do not think that Kant accepts the Cartesian basis.
12. Bennett notes this point in Kant's Dialectic, op. cit., p. 74, but dismisses
it as unimportant.
13. There is one piece of the minor premise that I have not tackled. Why
does Kant describe the I as the "absolute" subject of judgments? One possibility
is that "absolute" is simply a synonym for "necessary." A more likely expla-
nation is that this usage is an another attempt to connect the Paralogism to the
general discussion of the "transcendental ideas," where he talks about the "ab-
solute (unconditional) unity of the thinking subject" (A334/B391). Laying stress
on this piece of the Paralogism suggests that it is a criticism of the Monadology.
Leibniz was certainly a main target in Kant's opening description of the errors
to be exposed in the Dialectic. In the text of the First Paralogism, however,
there is no elaboration of this type of metaphysical error.
14. Loemker, op. cit., p. 646.
15. E. S. Haldane and G. R. T. Ross, translators, The Philosophical Works
of Descartes, vol. I, (Cambridge: Cambridge University Press, 1972), p. 196.
266 NOTES
16. See Blackwell, op. cit., p. 340.
17. "etwas von ihm zu kennen, oder zu wissen." Kemp Smith renders this
phrase "without knowing anything of it either by direct acquaintance or oth-
erwise," p. 337. I do not find this translation very satisfactory because "direct
acquaintance" is a twentieth-century term of art and because nothing in the
translation captures the force of "wissen."
18. This position may seem inconsistent with the premise of contemporary
cognitive science that psychology, neurophysiology, and so forth, can mutually
inform each other. Kant's point is that an abstract analysis of the faculties
required for a particular cognitive task cannot by itself determine the constitution
of the faculty that performs the task. Contemporary cognitive scientists would
agree. Psychology can still inform neurophysiology (and vice vera), however,
because given a functional decomposition of a task and further information
about what kinds of mechanisms could perform the tasks, and what kinds of
mechanisms are available in the brain, we can move from abstract functional
description to claims about particular mechanisms.
19. See A230-41/B300 and the Logic, op. cit., pp. 144-46 (AAIX: 143-145).
As I note in Chapter 8, Kant did not believe that any but arbitrary concepts
could be defined.
20. Ibid., p. 144.
21. In "... this I or he or it (the thing) which thinks...," Sellars characterizes
one of the great insights of the chapter as Kant's recognition that thinking is a
"functional concept." Unfortunately, he does not address the question of how
Kant came to this view and I think his suggestion that Kant simply recognized
that mental terms like "thought" are functional is implausible. See para. 21.
My contention is that Kant's own attempts to characterize the necessary prop-
erties of a thinking self led him to this recognition.
22. B422-23a. See Blackwell's discussion of Wolff's version, op. cit., p. 341.
23. I discuss the problems this passage raises for Transcendental Idealism in
Chapter 5, "Too Many Selves."
24. Zeno Vendler, The Matter of Minds (Oxford: Clarendon Press, 1984),
pp. 110-11.
25. Thomas Nagel, The View from Nowhere (New York: Oxford University
Press, 1986), p. 33.
26. Ibid., p. 40.
27. See Locke, op. cit., p. 448; Remnant and Bennett, op. cit., pp. 236ff;
and Loemker, op. cit., p. 325, "Discourse on Metaphysics," para. 34.
28. In the absence of this interpretive handle, even gifted readers of Kant
have had difficulty making any sense of this section of the Critique. So, for
example, in a his brief account of the Paralogisms, Strawson credits Kant with
the insight that we need physical criteria for reidentifying persons. While he
regards this as the overall message of the chapter, presumably it should be most
sharply etched in the Third Paralogism where the explicit topic is self-identity.
There is no mention of physical criteria for reidentification in these passages,
however. Strawson rests his interpretation on one slender text, namely, Kant's
remark that during life the permanence of the soul is evident, "since the thinking
 Notes 267
being (as man) is itself likewise an object of outer sense" (B415). Earlier Kant
had noted that in common parlance we speak of man as thinking (A359-60). I
take Kant's point at B415 to be that for ordinary purposes of reidentification
we do not need Rational Psychology to inform us about the continuity of thinking
beings. Rational Psychology is supposed to offer us philosophical proof that
souls must continue to exist even after death (B415). Putting this point in Straw-
son's own preferred idiom, I take Kant's observation to be simply that during
life bodily continuity is the usual way to determine continuity of the self, not
that bodily continuity is "criterial evidence" for self-identity. Also see the dis-
cussion of Bennett in the next note.
29. Bennett has trouble with these passages, because he cannot explain why
Kant believes that from my own point of view I must attribute identity and
permanence to myself. He takes the minor premise to assert that when a person
is conscious of, that is, remembers, a past mental state, he automatically attri-
butes that state to himself. According to Bennett, Kant's great insight, which
precisely anticipates an important argument given by Sydney Shoemaker in
"Persons and their Pasts," (American Philosophical Quarterly, October, 1970)
is that this attribution is too hasty because for all the person knows "he" may
have "branched" since the remembered event, so that his apparent memory is
only a "quasi-memory." The case for this interpretation rests on Bennett's
reading of the outside observer passage. Allegedly, Kant appeals to the outside
observer to demonstrate the point about quasi-memory. An outside observer
might have seen the individual branch and so realize that the inference from a
quasi-memory to identity is false. (I give a different account of this passage in
the text.) But this interpretation has serious problems. There is not a single
reference to memory or its potential fallibility in the entire chapter. And, as
Bennett admits, he has no idea why Kant believes that a subject must attribute
identity to himself at different times (Kant's Dialectic, op. cit., pp. 93-102).
30. See Chapter 2, "The Isolation Argument," and Chapter 4, "Transcen-
dental Synthesis." In this section of Chapter 4 I note that Kant's argument for
the a priori0 status of apperception is insufficient.
31. On this point I agree with the interpretation offered by R. I. G. Hughes
in "Kant's Third Paralogism," Kant-Studien 74 (1983): 405-11. I note some
disagreements in the text and in note 32.
32. R. I. G. Hughes criticizes this interpretation (which I offered originally
in "Kant's Paralogisms," Philosophical Review XCI (1982): 515-47) on the
grounds he can find no warrant in the text for the claim that this is the problem
with the outside observer: there will be times when he does not or cannot
attribute a mental state to me (Ibid., see his note 5). I take the warrant to be
the following: "For just as the time in which the observer sets me is not the
time of my own but of his sensibility, so the identity which is necessarily bound
up with my consciousness is not therefore bound up with his, that is, with the
consciousness which contains the outer intuition of my subject" (A363). Here
I take Kant to be stressing that the time in which the observer sets me is the
time of his sensibility. But his sensibility will obviously include many states that
have nothing to do with me and my states. So Kant concludes that the identity
268 NOTES
that must be present in all moments in which I am conscious will not be a feature
of his consciousness. Further, the outside observer is intended as a contrast with
my own case. And the peculiar feature of our own cases, that Kant repeats four
times between pages A363 and A365, is that since time is the form of inner
sense, at all times of which I am conscious, I am conscious of a state that I must
attribute to myself.
33. Ibid., p. 406.
34. Wilson, "Leibniz and Materialism," op. cit.
35. See AA XX:308. This passage was drawn to my attention by Allison's
paper "Kant's Refutation of Materialism," op. cit.
36. Loemker, op. cit., p. 644, "Monadology," para. 17.
37. She bases this interpolation on a number of passages. See the references
cited in "Leibniz and Materialism," op. cit., pp. 506-508.
38. This direction of argument is suggested by the passages she cites from
"A New System of the Nature and Communication of Substances" and from
the "Correspondence with Arnauld" (Loemker, op. cit., p. 456, 339).
39. This is the argument that is repeated in On the Progress of Metaphysics
(AA XX:308).
40. Wilson, "Leibniz and Materialism," op. cit., p. 509.
41. Ibid., p. 513.
42. I believe that this is all that Kant means at B420 (and in the discussion
of this issue in On the Progress of Metaphysics) when he claims that this argument
shows the falsity of materialism. Both Karl Ameriks, op. cit., Chapter 2, and
Henry Allison ("Kant's Refutation of Materialism," op. cit.) fasten on these
texts to suggest that Kant believes that he can argue against materialism. That
is true only in a very limited sense. Materialism has no ready explanation for
the needed synthetic connection. Kant recognizes fully in the Second Paralogism,
especially in the note about to be discussed, that this is equally true of imma-
terialisrn, however.
43. John Searle, "Minds, Brains, and Programs," reprinted in John Hauge-
land, ed., Mind Design (Cambridge, Mass.: Bradford/MIT Press, 1981),
pp. 282-306, 299.
44. While Searle's discussion is a blatant example of the Rational Psychol-
ogist's fallacy, Ned Block also falls into this error, despite his efforts to be
cautious. In "Troubles with Functionalism" (in C. Wade Savage, ed., Minnesota
Studies in the Philosophy of Science, vol. IX (Minneapolis: University of Min-
nesota Press, 1978), he offers the following argument. A homunculi-headed
robot does not seem able to have qualitative states. Since we do not know how
a brain can have qualitative states, one might think that the same doubt arises
in this case. But we know that brains do have qualitative states, so in this case
our empirical knowledge allows us to overcome the doubt. Again, the fallacy
is the same. Since we do not know what ordinary or remarkable properties of
brains enable them to have qualitative states, we cannot argue that other systems
cannot have them, and that only brains can, until we know how brains can have
such states. Pace Block it is not enough to know that brains can have qualitative
states. (See pp. 293-94.)
 Notes 269
Chapter 8

1. Logic, op. cit., p. 13 (AA IX:11).

2. Op. cit., pp. 7-8 (AA IV:471).
3. So, for example, Kant notes t h a t " . . . the empirical unity of apperception,
upon which we are not here dwelling, and and which besides is merely derived
from the former under given conditions in concrete, has only subjective validity"
(B140, my emphasis). Further, he regards his analyses as providing the basis
for all the activities of the understanding: "The first pure knowledge of under-
standing, then, upon which all the rest of its employment is based..." (B137,
my emphasis).
4. Rosch's original paper was "On the Internal Structure of Perceptual and
Semantic Categories," in T. E. Moore, ed., Cognitive Development and the
Acquisition of Language (New York: Academic Press, 1973). For a standard
overview of the subsequent literature (to 1981), see Edward E. Smith and
Douglas L. Medin, Categories and Concepts (Cambridge, Mass.: Harvard Uni-
versity Press, 1981).
5. See, for example, Susan Carey, Conceptual Change in Childhood (Cam-
bridge, Mass.: MIT Press, 1985); Frank Keil, Concepts, Kinds, and Cognitive
Development (Cambridge Mass.: MIT Press, 1989); and Ellen Markman, Cat-
egorization and Naming in Children (Cambridge, Mass.: MIT Press, 1989).
6. Smith and Medin, op. cit., p. 1, introduce this useful terminology.
7. Here I use "concept" to indicate both a classification and the mental
representation that enables individuals to use a classification. Although this
usage is somewhat awkward, I think this ambiguity is present in both the phil-
osophical and the psychological literature.
8. Ludwig Wittgenstein, The Philosophical Investigations, translated by
G. E. M. Anscombe (Oxford: Blackwell), 1953, pp. 31ff.
9. Smith and Medin, op. cit. p. 182.
10. Ibid. See the discussion of Frank Keil's work, pp. 175-79.
11. Here I am relying on Douglas Medin's later discussion in "Concepts and
Conceptual Structure," American Psychologist 44 (1989) 1469-81.
12. Ibid., pp. 7-9.
13. Smith and Medin, op. cit., p. 182.
14. T. P. McNamara and R. J. Sternberg, "Mental Models of Word Mean-
ing," Journal of Verbal Learning and Verbal Behavior 22 (1983): 449-74.
15. See Chapter 6, "Concept Application."
16. Similar proposals have recently been developed and have met with some
success. So, for example, the "motor theory" of phone perception claims that
we classify acoustic signals as different phones on the basis of considerations
about how we as speakers, would articulate the sounds. We would classify a
range of acoustic signals as "ba's," for instance, not on the basis of their intrinsic
acoustic similarity, but on the basis of whether all of the sounds could be
produced by the same articulatory movements. (Cited in Sally P. Springer and
and Georg Deutsch, Left Brain, Right Brain (San Francisco: W. H. Freeman,
1981), pp. 201-202).
270 NOTES
17. I am grateful to Margery Lucas for drawing this point to my attention
when I presented parts of this chapter at Wellesley College in April of 1989.
18. Logic, op. cit., p. 142 (AA IX:141-42). See also A727-28/B755-56.
19. See, e.g., ibid., p. 97 (AA IX:92-93).
20. As will be clear later, Kant has further requirements for empirically
adequate concepts beyond the Empiricist criterion of generation by sensory
contact with objects.
21. The only exception I am aware of is Jerry Fodor in The Language of
Thought (New York: Thomas Y. Crowell, 1975).
22. I borrow the example from Daniel Dennett, "Intentional Systems," in
Brainstorms, op. cit., p. 17.
23. W. V. Quine, Ontological Relativity and Other Essays (New York: Co-
lumbia University Press, 1969), pp. 114-39.
24. Ibid., p. 123.
25. Ibid., p. 128.
26. Quine notes that his account of how children are able to learn "yellow"—
namely, through having quality spaces like those of adults—cannot be extended
to more complicated terms. But he does not offer any further psychological
mechanisms to underlie these more sophisticated conceptual acquistions. Ibid.,
pp. 123-24.
27. Ibid., p. 128.
28. Keil offers the label and his criticisms in Concepts, Kinds, and Cognitive
Development, op. cit. Quine's position is also criticized by Gregory L. Murphy
and Douglas L. Medin, in "The Role of Theories in Conceptual Coherence,"
Psychological Review, 92 (July, 1985): 289-316.
29. Susan A. Gelman and Ellen M. Markman, "Young Children's Inductions
from Natural Kinds: The Role of Categories and Appearances," Child Devel-
opment, 58 (1987): 1532-41.
30. It is possible that even younger children just rely on appearances, but it
is not clear how this could be tested.
31. Susan Carey, op. cit.
32. Frank Keil, op. cit.
33. Carey's view is inconsistent with this position. Her central point is that
the theories underlying children's uses of concepts are different from adult
theories.
34. Medin, op. cit., p. 1476.
35. See F. C. Keil, "On the Development of Biologically Specific Beliefs:
The Case of Inheritance," forthcoming in Cognitive Development, and Elizabeth
Spelke, "On the origins of physical knowledge," in L. Weiskrantz (ed.), Thought
without Language (New York: Oxford University Press, 1988), pp. 168-83.
36. Murphy and Medin, op. cit., passim.
37. Medin, op. cit., p. 1479.
38. Loc. cit., p. 1477.
39. See especially Gelman and Markman, op. cit., and John H. Holland,
Keith J. Holyoak, Richard E. Nisbett, and Paul R. Thagard, Induction (Cam-
bridge, Mass.: MIT Press, 1986).
 Notes 271
40. See Fodor's discussion of this paradigm in "On The Present Status of
the Innateness Controversy," in Representations, op. cit., pp. 257-316, esp.
266-268.
41. See Dreyfus's discussion of Patrick Winston's concept learning program
in Hubert L. Dreyfus, What Computers Can't Do (New York: Harper & Row,
1979), pp. 21ff.
42. First Introduction to the Critique of Judgment, translated by James Haden
(Indianapolis: Bobbs-Merrill, 1965), p. 10, note 3 (AA XX:206).
43. See Chapter 3, "The Synthesis of Intuitions."
44. C. G. Hempel, "Aspects of Scientific Explanation," in Aspects of Sci-
entific Explanation and Other Essays (New York: The Free Press, 1965), p. 374.
45. Allison, op. cit., Part 1, Chapter 2, passim.
46. Logic, op. cit., p. 104 (AA IX:98).
47. Ibid., pp. 64-65 (AA IX:59).
48. Ibid., p. 66(AAIX:60).
49. Ibid., p. 64(AAIX:58).
50. Ibid., p. 65 (AA IX:59).
51. Ibid., p. 67(AAIX:61).
52. Several passages in this discussion strongly imply a different reading:
Necessary characteristics, finally, are those which must always be found in
the matter conceived [Ibid., p. 66 (AA IX:60)].
***
To the logical essence belongs nothing but the cognition of all predicates
in respect of which an object is determined by its concept.... And into this
we can easily have insight... we do not have to search out data for this in
nature; we only need to direct our reflection to those characteristics which
as essential elements (constitutiva, rationes) originally constitute its basic
concept [Ibid., p. 67 (AA IX:61)].
Contrary to the general interpretation I have been offering, these passages
suggest that Kant held something like a necessary and sufficient conditions view
of concepts. However, the natural reading of these remarks not only is incon-
sistent with my interpretation but also is flatly inconsistent with Kant's position
on defining empirical concepts in the Critique:
an empirical concept cannot be defined at all, but only made explicit. For
since we find in it only a few characteristics of a certain species of sensible
object, it is never certain that we are not using the word, in denoting one
and the same object, sometimes so as to stand for more, and sometimes
so as to stand for fewer characteristics.... We make use of certain char-
acteristics only so long as they are adequate for the purpose of making
distinctions; new observations remove some properties and add others; and
thus the limits of the concept are never assured [A727-28/B755-56, my
emphasis].
There are two ways to reconcile these views. Perhaps the discussion in the Logic
is not meant to apply to empirical concepts. Although this is possible, nothing
272 NOTES
in the text indicates that this is not a perfectly general discussion of concepts.
The second, and I believe better, strategy is to assume that the Logic's discussion
is about how logic, by invoking principles of coordination and subordination,
can capture the [current] mental representation of a concept. Since this section
is about the logical perfection of cognition, the first part of the interpretation
seems justified. I would justify the second—the appeal to a current mental
representation—on two grounds. Within the passage, Kant claims that finding
the logical essence is very easy. We look not at nature, but to our own repre-
sentation. Second, there is no inconsistency in holding that empirical concepts
cannot be fixed and that at any time, on the basis of previous experience, some
characteristics are regarded as necessary for everything in the extension of the
concept.
53. First Introduction to the Critique of Judgment, op. cit., p. 20n (AA
XX:216).
54. Metaphysical Foundations of Natural Science, op. cit., pp. 40-49 (AA
IV:496-503).
55. Although not, of course, a fact about our mental powers in themselves
(A648/B676).
56. Some experiments by D. Billman suggest that we acquire groups of in-
terrelated rules (or correlations) more easily than single correlations. See the
discussion of Billman's work in Holland et al., op. cit., pp. 200ff.
57. Barbara Landau, Linda B. Smith, and Susan S. Jones, "The Importance
of Shape in Early Lexical Learning," Cognitive Development 3 (1988): 299-321.
58. I owe this point to Jean Mandler.
59. Murphy and Medin, op. cit., pp. 297-98.
60. Keil made this point in a presentation at a Sloan conference on theories
and concepts in children, held at Stanford University on January 7-8, 1989.
61. See note 53.
62. Perhaps I am misinterpreting Medin's appeal to essences. I take an essence
to be the set of properties that make an individual, or a kind, the individual or
kind that it is. As such, once the essence has been determined, there is nothing
more to be said about these items. In particular, there is no sense in trying to
fathom deeper similarities that they might share with other objects. That is why
a commitment to essences blocks a wider system of inferences. This may be an
overly technical reading of something Medin intends in a much looser sense. If
the "essentialist" heuristic is just meant to indicate that we prefer concepts
where deeper attributes explain more superficial similarities, however, then it
would be reasonable to view it as one of the heuristics serving the systematizing
demands of reason. Medin offers a further discussion of this issue in Medin,
D. L., and Ortony, A., "Psychological Essentialism," in S. Bosiniadow and A.
Ortony, eds., Similarity and Analogical Reasoning (New York: Cambridge Uni-
versity Press, 1989), pp. 179-95.
63. I am happy to record my indebtedness to the participants in the Sloan
conference on theories and concepts in children at Stanford on January 7-8,
1989, which was organized by Ellen Markman. The papers and discussions
stimulated many of the ideas of this chapter.
 BIBLIOGRAPHY

Adickes, Erich. Kants Lehre von der Doppelten Affektion unseres Ich as Schlussel
zu seiner Erkenntnistheorie. Tubingen, 1929.
Allison, Henry E. "Kant's Refutation of Materialism," The Monist 72 (1989),
190-208.
Allison, Henry E. Kant's Transcendental Idealism: An Interpretation and De-
fense. New Haven: Yale University Press, 1983.
Allison, Henry E. The Kant-Eberhard Controversy. Baltimore: The Johns Hop-
kins University Press, 1973.
Ameriks, Karl. "Kant's Transcendental Deduction as a Regressive Argument,"
Kant-Studien 69 (1978), 273-87.
Ameriks, Karl. Kant's Theory of Mind: An Analysis of the Paralogisms of Pure
Reason. Oxford: Clarendon Press, 1982.
Angell, R. B. "The Geometry of the Visibles," Nous 8 (1974), 87-117.
Aquila, Richard. Matter in Mind: A Study of Kant's Transcendental Deduction.
Bloomington: Indiana University Press, 1989.
Arnauld, Antoine. The Art of Thinking. James Dickoff and Patricia James
(trans.). Indianapolis: Bobbs-Merrill, 1964.
Baddeley, Alan. Working Memory. Oxford: Clarendon Press, 1986.
Beattie, James. "Essay on the Nature of Immutability of Truth," in Beattie's
Works. Philadelphia: Hopkins and Earle, 1809.
Beck, Lewis W. Early German Philosophy. Cambridge, Mass.: Harvard Uni-
versity Press, 1969.
Beiser, Frederick. The Fate of Reason. Cambridge, Mass.: Harvard University
Press, 1987.
Bennett, Jonathan. "The Simplicity of the Soul," Journal of Philosophy LXIV
20 (Oct. 26, 1967), 648-60.
Bennett, Jonathan. Kant's Analytic. Cambridge: Cambridge University Press,
1966.
Bennett, Jonathan. Kant's Dialectic. Cambridge: Cambridge University Press,
1974.
Berkeley, George. The Works of George Berkeley Bishop of Cloyne. A. A.
Luce and T. E. Jessop (eds.). London and Edinburgh: Nelson, 1967.
Blackwell, Richard J. "Christian Wolff's Doctrine of the Soul," Journal of the
History of Ideas XXII (1961), 339-54.
273
274 BIBLIOGRAPHY
Block, Ned. "Troubles with Functionalism," in C. Wade Savage (ed.), Min-
nesota Studies in the Philosophy of Science, vol. IX. Minneapolis: Uni-
versity of Minnesota Press, 1978.
Bona Meyer, Jurgen. Kant's Psychologic. Berlin: Wilhelm Hertz, 1870.
Brett, George Sidney. Brett's History of Psychology. R. S. Peters (ed.). London:
George Allen and Unwin, 1962.
Brueckner, Anthony L. "Transcendental Arguments I," Nous 17 (1983),
551-75.
Brueckner, Anthony L. "Transcendental Arguments II" Nous 18 (1984),
197-225.
Burke, E. A Philosophical Enquiry into the Origin of Our Ideas of the Sublime
and the Beautiful. Oxford: Basil Blackwell, 1967.
Buroker, Jill. Space and Incongruence. Amsterdam: D. Reidel, 1981.
Carey, Susan. Conceptual Change in Childhood. Cambridge, Mass.: MIT Press,
1985.
Castaneda, Hector-Neri. "He: A Study in the Logic of Self-Consciousness,"
Ratio 8 (1966).
Condillac, E. B. An Essay on the Origin of Human Knowledge. A facsimile
reproduction of the translation of Thomas Nugent, with an introduction
by Robert G. Weyant. Gainesville, Fla.: Scholars' Facsimiles & Reprints,
1971.
Condillac, E. B. Condillac's Treatise on the Sensations. Geraldine Carr (trans.).
Los Angeles: University of Southern California, 1930.
Cummins, Robert. "Functional Analysis," Journal of Philosophy 72 (1975),
741-60.
Curley, E. "Leibniz and Locke on Personal Identity," in M. Hooker (ed.),
Leibniz: Critical and Interpretive Essays. Minneapolis: University of Min-
nesota Press, 1982, 302-26.
Davidson, Donald. "Mental Events," in Essays on Actions and Events. Oxford:
Clarendon Press, 1980, 207-25.
De Vleeschauwer, H. J. La Deduction Transcendentale dans I'Oeuvre de Kant.
Paris: Le Hague, 1934-1937.
Dennett, Daniel. "Artificial Intelligence as Philosophy and as Psychology," in
Brainstorms. Cambridge, Mass.: Bradford Books, 1978, 109-26.
Dennett, Daniel. "A Cure for the Common Code," in Brainstorms. Cambridge,
Mass.: Bradford Books, 1978, 90-108.
Dennett, Daniel. Content and Consciousness. London: Routledge and Kegan
Paul, 1969.
Descartes, Rene. The Philosophical Works of Descartes. E. S. Haldane and
G. R. T. Ross (trans.). Cambridge: Cambridge University Press, 1972.
Dessoir, Max. Geschichte de Neueren Deutschen Psychologic. Amsterdam: E. J.
Bonset, 1964.
Deutsch, D. "Auditory Pattern Recognition," in K. R. Boff, L. Kaufman, and
J. P. Thomas (eds.), Handbook of Perception and Human Performance,
vol. II. New York: Wiley, 1986.
 Bibliography 275
Dreyfus, Hubert L. What Computers Can't Do. New York: Harper, 1979.
Dummett, Michael. Frege: Philosophy of Language. New York: Harper & Row,
1973.
Barman, John. "Was Leibniz a Relationist?" Midwest Studies, vol. IV. Min-
neapolis, Minn.: University of Minnesota Press, 1979, 263-76.
England, F. E. Kant's Conception of God (with a translation of the Nova Di-
luddatio). New York: Humanities Press, 1968, 246-47.
Fauconnier, Giles. Mental Spaces. Cambridge, Mass.: MIT Press, 1985.
Fodor, J. A. "On The Present Status of the Innateness Controversy," in Rep-
resentations. Cambridge, Mass.: MIT Press, 1981, 257-316.
Fodor, J. A. The Language of Thought. Cambridge, Mass.: Bradford Books,
1975.
Fodor, J. A. The Modularity of Mind. Cambridge, Mass.: MIT Press,
1983.
Fodor, J. A., Bever, T. G., and Garrett, M. F. The Psychology of Language.
New York: McGraw-Hill, 1974.
Frege, G. The Foundations of Arithmetic. J. L. Austin (trans.). Oxford: Basil
Blackwell, 1968.
Friedman, Michael. "Kant's Theory of Geometry," Philosophical Review 94
(1985), 455-506.
Garrett, Don. "Hume's Self-Doubts About Personal Identity," Philosophical
Review, xc (1981), 337-58.
Gelman, Susan A., and Markman, Ellen M. "Young Children's Inductions from
Natural Kinds: The Role of Categories and Appearances," Child De-
velopment 58 (1987), 1532-41.
George, Rolf. "Kant's Sensationism," Synthese 47 (1981), 229-55.
Gibson, J. J. The Ecological Approach to Visual Perception. Boston: Houghton
Mifflin, 1979.
Gibson, J. J. The Senses Considered as Perceptual Systems. Boston: Houghton
Mifflin, 1966.
Gibson, E., and Spelke, E. "The Development of Perception," in John H.
Flavell and Ellen M. Markman (eds.), Cognitive Development, Volume
III of Handbook of Child Psychology (Paul H. Mussen, general ed.).
New York: John Wiley & Sons, 1983, 1-76.
Goldman, Alvin I. Epistemology and Cognition. Cambridge, Mass.: Harvard
University Press, 1986.
Grice, H. P. "Personal Identity," reprinted in John Perry (ed.), Personal Iden-
tity. Los Angeles: University of California Press, 1975, pp. 73-95.
Gueroult, Martial. "Space, Point, and Void in Leibniz's Philosophy," in Michael
Hooker (ed.), Leibniz, Critical and Interpretive Essays. Minneapolis:
University of Minnesota Press, 1982.
Guyer, Paul. "Kant on Apperception and A Priori Synthesis," American Phil-
osophical Quarterly 17 (1980), 205-12.
Guyer, Paul. Kant and the Claims of Knowledge. New York: Cambridge Uni-
versity Press, 1987.
276 BIBLIOGRAPHY

Handyside, John. Kant's Inaugural Dissertation and Early Writings on Space.

La Salle, 111.: Open Court, 1929.
Harman, Gilbert. "Quine on Meaning and Existence I," Review of Metaphysics
XXI (1967), 124-51.
Harper, W. "Kant's Empirical Realism and the Difference Between Subjective
and Objective Succession," in W. Harper and R. Meerbote (eds.), Kant
on Causality, Freedom, and Objectivity. Minneapolis: University of Min-
nesota Press, 1984, 108-37.
Hatfield, Gary, and Epstein, William. "The Sensory Core and the Medieval
Foundations of Early Modern Perceptual Theories," 75/5 70 (1979),
363-84.
Hatfield, Gary. The Natural and Normative: Theories of Perception from Kant
to Helmholtz. Cambridge, Mass.: MIT Press, in press.
Hempel, Carl G. "Implications of Carnap's Work for the Philosophy of Sci-
ence," in Paul Arthur Schlipp (ed.), The Philosophy of Rudolf Carnap.
La Salle, 111.: Open Court, 1963, 685-709.
Henrich, D. "The Proof-Structure of Kant's Transcendental Deduction," Re-
view of Metaphysics 22 (1969), 640-59.
Henrich, Dieter. Identitat und Objektivitat: Erne Untersuchung uber Kants Tran-
szendentale Deduktion. Heidelberg: C. Winter, 1976.
Herrnstein, Richard J., and Boring, Edwin G. A Source Book in the History
of Psychology. Cambridge, Mass.: Harvard University Press, 1968.
Hintikka, Jaakko. "On Kant's Notion of Intuition (Anschauung)," in T. Pe-
nelhum and J. Macintosh (eds.), The First Critique: Reflections on Kant's
Critique of Pure Reason. Belmont, Ca.: Wadsworth, 1969, 38-53.
Holland, John H., Holyoak, Keith J., Nisbett, Richard E., and Thagard, Paul
R. Induction. Cambridge, Mass.: MIT Press, 1986.
Hopkins, James. "Visual Geometry," Philosophical Review 82 (1973), 3-34.
Hoppe, Hansgeorg, Synthesis Bel Kant. Berlin: Walter de Gruyter, 1983.
Howell, Robert. "Intuition, Synthesis, and Individuation in the Critique of Pure
Reason," Nous XX (1973), 207-32.
Hughes, R. I. G. "Kant's Third Paralogism," Kant Studien 74 (1983), 405-11.
Hume, David. Inquiry Concerning Human Understanding. Eric Steinberg (ed.).
Indianapolis: Hackett, 1981.
Hume, David. A Treatise of Human Nature, L. A. Selby-Bigge (ed.). Oxford:
Oxford University Press, 1962.
Humphrey, Ted. "The Historical and Conceptual Relations Between Kant's
Metaphysics of Space and Philosophy of Geometry," Journal of the His-
tory of Philosophy 11 (1973), 483-512.
Husserl, Edmund. Logical Investigations, vol. I. J. N. Findlay (trans.). New
York: Humanities Press, 1970.
James, W. Psychology: The Briefer Course. New York: Harper Torchbooks,
1961.
Janaway, Christopher. "History of Philosophy: The Analytic Ideal I," Pro-
ceedings of the Aristotelian Society Supplement, LXVII (1988), 169-89.
 Bibliography 277
Jessop, T. E. A Bibliography of George Berkeley. The Hague: Martinus Nijhoff,
1973.
Jolley, Nicholas. Leibniz and Locke: A Study of the "New Essays on Human
Understanding." Oxford: Clarendon Press, 1984.
Kant, I. Anthropology from a Pragmatic Point of View. Mary Gregor (ed. and
trans.). The Hague: Martinus Nijhoff, 1974.
Kant, I. First Introduction to the Critique of Judgment. James Haden (trans.).
New York: Bobbs-Merrill, 1965.
Kant, I. Kant's Critique of Pure Reason, Norman Kemp Smith (trans.). New
York: St. Martin's, 1968.
Kant, I. Kant: Philosophical Correspondence 1759-99. Arnulf Zweig (ed. and
trans.). Chicago: University of Chicago Press, 1967.
Kant, I. Kant: Selected Pre-Critical Writings and Correspondence with Beck.
G. B. Kerford and D. E. Walford (eds.). New York: Barnes and Noble,
1968.
Kant, I. Kants gesammelte Schriften. Akademie edition, edited by the Konig-
lichen Preussischen Akademie der Wissenschaften, 29 volumes. Berlin and
Leipzig: Walter de Gruyter and predecessors, 1902-.
Kant, I. Logic. R. S. Hartman and W. Schwarz (trans.). New York: Bobbs-
Merrill, 1974.
Kant, I. Prolegomena to Any Future Metaphysics That Will Be Able to Come
Forward as Science. James Ellington (trans.). Indianapolis: Hackett Pub-
lishing Company, 1977.
Kant, I. The Groundwork of the Metaphysics of Morals. Lewis White Beck
(trans.). New York: Bobbs-Merrill, 1959.
Kant, I. The Metaphysical Foundations of Natural Science. J. Ellington (trans.).
New York: Bobbs-Merrill, 1970.
Keil, F. C. "On the Development of Biologically Specific Beliefs: The Case of
Inheritance," Cognitive Development (in press).
Keil, F. C. Concepts, Kinds, and Cognitive Development Cambridge, Mass.:
MIT Press, 1989.
Kemp Smith, Norman. A Commentary to Kant's "Critique of Pure Reason."
New York: Humanities Press, 1962.
Keynes, Geoffrey. A Bibliography of George Berkeley. Oxford: Oxford Uni-
versity Press, 1976.
Kitcher, Patricia. "Being Selfish About Your Future," Philosophical Studies 32
(1977), 425-31.
Kitcher, Patricia. "Discovering the Forms of Intuition," Philosophical Review
XCVI (1987), 205-48.
Kitcher, Patricia. "Kant on Self-Identity," Philosophical Review XCI (1982),
41-72.
Kitcher, Patricia. "Kant's Paralogisms," Philosophical Review XCI (1982), 515-
47.
Kitcher, Patricia. "Kant's Patchy Epistemology," Pacific Philosophical Quar-
terly XX (1987), 306-16.
278 BIBLIOGRAPHY
Kitcher, Patricia. "Kant's Real Self," in Allen W. Wood (ed.), Self and Nature
in Kant's Philosophy. Ithaca, N.Y.: Cornell University Press, 1984,
113-47.
Kitcher, Patricia. "Natural Kinds and Unnatural Persons," Philosophy 54
(1979), 541-47.
Kitcher, Philip. "How Kant Almost Wrote Two Dogmas of Empiricism' (and
Why He Didn't)," in J. N. Mohanty and Robert W. Shahan (eds.), Essays
on Kant's Critique of Pure Reason. Norman: University of Oklahoma
Press, 1982, 217-49.
Kitcher, Philip. "Kant and the Foundations of Mathematics," Philosophical
Review 84 (1975), 23-50.
Kitcher, Philip. "Kant's Philosophy of Science," in Allen W. Wood (ed.), Self
and Nature in Kant's Philosophy. Ithaca, N.Y.: Cornell University Press,
1984, 185-215.
Korsgaard, Christine. "Personal Identity and the Unity of Agency: A Kantian
Response to Parfit," Philosophy and Public Affairs (in press).
Kuehn, Manfred. Scottish Common Sense in Germany, 1768-1800: A Contri-
bution to the History of the Critical Philosophy. Montreal: McGill-
Queen's University Press, 1987.
Kuhn, Thomas. The Structure of Scientific Revolutions, Second Edition. Chicago:
University of Chicago Press, 1970.
Landau, B., Gleitman, H., and Spelke, E. "Spatial Knowledge and Geomet-
ric Representation in a Child Blind from Birth," Science 213 (1981),
1275-78.
Landau, B., Smith, L. B., and Jones, S. S. "The Importance of Shape in Early
Lexical Learning," Cognitive Development 3 (1988) 299-321.
Leary, David E. "Immanuel Kant and the Development of Modern Psychol-
ogy," in William R. Woodward and Mitchell G. Ash (eds.), The Prob-
lematic Science: Psychology in Nineteenth Century Thought. New York:
Praeger, 1982, 17-42.
Leary, David E. "The Philosophical Development of the Conception of Psy-
chology in Germany, 1780-1850," Journal of the History of the Behavioral
Sciences XIV (1978), 113-21.
Leary, David E. "The Psychology of Jakob Friedrich Fries (1773-1843): Its
Context, Nature, and Historical Significance," Storia e Critica delta Psis-
cologia III (1982), 217-48.
Leiblich, I., and Arbib, M. "Multiple Representations of Space Underlying
Behavior," Behavioral and Brain Sciences 5 (1982), 627-59.
Leibniz, Gottfried Wilhelm. New Essays on Human Understanding, Abridged
Edition. Peter Remnant and Jonathan Bennett (trans, and eds.). Cam-
bridge: Cambridge University Press, 1982.
Leibniz, Gottfried Wilhelm. Philosophical Papers and Letters. Leroy E. Lo-
emker (ed.). Amsterdam: D. Reidel, 1976.
Lewis, David. "Survival and Identity," in A. Rorty (ed.), The Identities of
Persons. Los Angeles: University of California Press, 1969,
17-40.
 Bibliography 279
Locke, John. An Essay Concerning Human Understanding, 2 vols. A. C. Fraser
(ed.). New York: Dover, 1959.
Luneberg, R. K. Mathematical Analysis of Binocular Vision. Princeton, N.J.:
Princeton University Press, 1947.
Markman, Ellen. Categorization and Naming in Children. Cambridge, Mass.:
MIT Press, 1989.
Marr, David. Vision. San Francisco: Freeman, 1982.
Martindale, Colin. "Can We Construct Kantian Mental Machines?" Journal of
Mind and Behavior 8 (1987), 261-68.
Mattey, G. J. "Kant's Conception of Berkeley's Idealism," Kant-Studien 74
(1983), 161-75.
McClelland, James L., and Rumelhardt, David E. Parallel Distributed Pro-
cessing. Cambridge, Mass.: MIT Press, 1986.
McNamara, T. P., and Sternberg, R. J. "Mental Models of Word Mean-
ing," Journal of Verbal Learning and Verbal Behavior 22 (1983), 449-
74.
McRae, Robert. Leibniz: Perception, Apperception and Thought. Toronto: Uni-
versity of Toronto Press, 1976.
Medin, Douglas. "Concepts and Conceptual Structure," American Psychologist
44 (1989) 1469-81.
Meerbote, Ralf. "Kant's Functionalism," in J. C. Smith (ed.), Historical Foun-
dations of Cognitive Science. Dordrecht, Holland: Reidel, 1989.
Melnick, A. Kant's Analogies of Experience. Chicago: University of Chicago
Press, 1973.
Murphy, G. L., and Medin, D. L. "The Role of Theories in Conceptual Co-
herence," Psychological Review 92 (1985), 289-316.
Nagel, Thomas. The View from Nowhere. New York: Oxford University Press,
1986.
Newell, Allen, and Simon, Herbert. Human Problem Solving. Englewood Cliffs,
N.J.: Prentice-Hall, 1972.
O'Keefe, John, and Nadel, Lynn. The Hippocampus as a Cognitive Map. Ox-
ford: Oxford University Press, 1978.
O'Neil, Onora. "Transcendental Synthesis and Developmental Psychology,"
Kant-Studien 75 (1984), 149-67.
Parfit, D. Reasons and Persons. Oxford: Clarendon Press, 1984.
Parfit, D., "Later Selves and Moral Principles," in A. Montefiore (ed.), Phi-
losophy and Personal Relations. London: Routledge and Kegan Paul,
1973.
Parfit, D., "Personal Identity," Philosophical Review 80 (1971), 3-27.
Parsons, Charles. "Infinity and Kant's Conception of the 'Possibility of Expe-
rience'," Philosophical Review 73 (1964), 182-97.
Pastore, Nicholas. Selective History of Theories of Visual Perception, 1650-1950.
Oxford: Oxford University Press, 1971.
Paton, H. J. Kant's Metaphysics of Experience, vols. I and II. New York: Hu-
manities Press, 1965.
Patten, S. C. "Kant's Cogito," Kant-Studien 66 (1975), 331-41.
280 BIBLIOGRAPHY
Penelhum, Terence. "Personal Identity, Memory, and Survival," Journal of
Philosophy XX (1959), 882-903.
Pereboom, Derk. "Kant's Notion of the A Priori and the Transcendental,"
ms.
Perry, J. "Can the Self Divide?" Journal of Philosophy XX (1972), 463-88.
Perry, J. "The Importance of Being Identical," in A. Rorty (ed.), The Identities
of Persons. Los Angeles: University of California Press, 1969, 67-90.
Perry, John. Identity. Ann Arbor, Mich.: University Microfilms, 19 (Doctoral
Dissertation).
Perry, John. Personal Identity. Los Angeles: University of California Press, 1975.
Pippin, Robert. "Kant on the Spontaneity of Mind," Canadian Journal of Phi-
losophy 17 (1987), 449-76.
Pippin, Robert. Kant's Theory of Form. New Haven, Conn.: Yale University
Press, 1982.
Porterfield, William. A Treatise on the Eye and the Manner and Phenomena of
Vision. Edinburgh: G. Hamilton and J. Balfour, 1759.
Posner, Michael. Chronometric Explorations of Mind. Hillsdale, N.J.: Erlbaum,
1978.
Pylyshyn, Zenon. "Cognition and Computation: Issues in the Foundations of
Cognitive Science," Behavioral and Brain Sciences 3 (1978), 111-32.
Quine, W. V. "Carnap on Logical Truth," in The Ways of Paradox. New York:
Random House, 1966.
Quine, W. V. "Truth by Convention," in The Ways of Paradox. New York:
Random House, 1966, 77-106.
Quine, W. V. "Two Dogmas of Empiricism," in From a Logical Point of View.
New York: Harper, 1961, 20-46.
Quine, W. V. Ontological Relativity and Other Essays. New York: Columbia
University Press, 1969.
Quine, W. V. Word and Object. Cambridge, Mass.: MIT Press, 1960.
Quinton, A. "The Soul," in John Perry (ed.), Personal Identity. Los Angeles:
University of California Press, 1975.
Reid, T. An Inquiry into the Human Mind. T. Duggan (ed.). Chicago: University
of Chicago Press, 1970.
Reid, T. Essays on the Intellectual Powers of Man, Abridged Edition. A. D.
Woozley (ed.). London: Macmillan, 1941.
Richards, Robert J. "Christian Wolffs Prolegomena to Empirical and Rational
Psychology: Translation and Commentary," Proceedings of the American
Philosophical Society 124 (1980), 227-39.
Ricketts, Thomas G. "Rationality, Translation, and Epistemology Naturalized,"
Journal of Philosophy 79 (1982), 117-36.
Rorty, Richard, "Verificationism and Transcendentaal Arguments," Nous 5
(1971), 1-14.
Rorty, Richard. Philosophy and the Mirror of Nature. Princeton, N.J.: Princeton
University Press, 1979.
Rosch, Eleanor. "On the Internal Structure of Perceptual and Semantic Cat-
 Bibliography 281
egories," in T. E. Moore (ed.), Cognitive Development and the Acqui-
sition of Language. New York: Academic Press, 1973.
Rosenberg, J. " 'I Think': Some Reflections on Kant's Paralogisms," ms.
Rosenberg, J. The Thinking Self. Philadelphia: Temple University Press, 1986.
Rostenreich, Nathan. "Self-Ascription and Objectivity," Philosophia: Philo-
sophical Quarterly of Israel 10 (1981), 189-98.
Russell, Bertrand. The Philosophy of Leibniz, 2nd Edition. London: Allen and
Unwin, 1937.
Ryle, Gilbert. The Concept of Mind. New York: Barnes and Noble, 1949.
Satura, Vladimir. "Kant's Erkenntnispsychologie," Kant-Studien XX (1971),
1-176.
Scruton, Roger. Kant. Oxford: Oxford University Press, 1982.
Searle, John. "Minds, Brains,and Programs," Behavioral and Brain Sciences 3
(1980), 417-424. (Reprinted in John Haugeland (ed.), Mind Design.
Cambridge, Mass.: MIT Press, 1981.)
Sellars, Wilfrid. " . . . this I or he or it (the thing) which thinks," Presidential
Address to the Eastern Division of the APA, 1970.
Sellars, Wilfrid. "Metaphysics and the Concept of a Person," in Karel Lambert
(ed.), The Logical Way of Doing Things. New Haven, Conn.: Yale Uni-
versity Press, 1969, 219-32.
Shoemaker, Sydney. "Persons and Their Pasts," American Philosophical Quar-
terly XX (1970), pp. 269-85.
Shoemaker, Sydney. Self-Knowledge and Self-Identity. Ithaca, N.Y.: Cornell
University Press, 1963.
Sluga, Hans. Gottlob Frege. Boston: Routledge & Kegan Paul, 1980.
Smith, E. E., and Medin, D. L. Categories and Concepts. Cambridge, Mass.:
Harvard University Press, 1981.
Spelke, Elizabeth. "On the Origins of Physical Knowledge," in L. Weiskrantz
(ed.), Thought Without Language. New York: Oxford, University Press,
1988, 168-83.
Springer, S. P., and Deutsch, G. Left Brain, Right Brain. San Francisco: W. H.
Freeman, 1981.
Strawson, Peter F. The Bounds of Sense. London: Methuen, 1966.
Strawson, Peter F., and Grice, H. P. "In Defense of a Dogma," Philosophical
Review LXV (1956), 141-58.
Stroud, Barry. "Transcendental Arguments," Journal of Philosophy 65 (1968),
241-56.
Stroud, Barry. Hume. London: Routledge and Kegan Paul, 1977.
Tetens, Johan Nicholas. Philosophische Versuche fiber die menschliche Natur
und ihre Entwicklung, 2 vols. Leibzig: M. G. Weidmans Erben und
Reich, 1777. (Reprinted by the Kantgesellschaft Verlag, 1911.)
Thomson, J. J. "Private Languages," American Philosophical Quarterly 1
(1964), 20-31.
Thomson, M. "Singular Terms and Intuitions in Kant's Epistemology," Review
of Metaphysics 26 (1972), 314-43.
282 BIBLIOGRAPHY
Treisman, A. M., and Gelade, G. A. "A Feature Integration Theory of At-
tention," Cognitive Psychology 12 (1980), 97-136.
Treisman, A. M., Schmidt, H. "Illusory Conjunctions in the Perception of
Objects," Cognitive Psychology 14 (1982), 107-41.
Vaihinger, H. Commentar zu Kants Kritik der reinen Vernunft, vol. II. New
York: Garland Publishing Company, 1970.
Vendler, Zeno. The Matter of Minds. Oxford: Clarendon Press, 1984.
Walker, Ralph C. S. Kant. London: Routledge and Kegan Paul, 1978.
Walker, Ralph C. S. "Synthesis and Transcendental Idealism," Kant-Studien
76 (1985), 14-27.
Walsh, W. H. "Philosophy and Psychology in Kant's Critique," Kant-Studien
56 (1966), 186-98.
Warda, A. Immanuel Kants Bucher. Berlin: Verlag von Martin Breslaner, 1922.
Watson, J. B. "Psychology as the Behaviorist Views It," reprinted in R. J.
Herrnstein and E. G. Boring (eds.), A Source Book in the History of
Psychology. Cambridge, Mass.: Harvard University Press, 1965, 513-14.
White, B., Saunders, F., Scadden, L., Bach-Y-Rita, P., and Collins, P., "Seeing
with the Skin," Perception and Psychophysics 1 (1970), 23-27.
Wiggins, David. "Locke, Butler and the Stream of Consciousness: And Men
as a Natural Kind," in A. Rorty (ed.), The Identities of Persons. Los
Angeles: University of California Press, 1969, 139-73.
Wiggins, David. Sameness and Substance. Cambridge, Mass.: Harvard Univer-
sity Press, 1980.
Wilkerson, T. E. Kant's Critique of Pure Reason. Oxford: Oxford University
Press, 1976.
Williams, Bernard. "Personal Identity and Individuation," Proceedings of the
Aristotelian Society LVH (1956), 302-26.
Wilson, Margaret. "Kant and 'The Dogmatic Idealism of Berkeley,'" Journal
of the History of Philosophy xii (1971), 459-75.
Wilson, Margaret. "Kant and the Refutations of Subjectivism," in L. W. Beck
(ed.), Kant's Theory of Knowledge. Dordrecht: Reidel, 1974, 208-17.
Wilson, Margaret. "Leibniz and Materialism," Canadian Journal of Philosophy
111 (1974), 495-513.
Wilson, Margaret. "Leibniz, Self-Consciousness, and Immortality: In the Paris
Notes and After," Archiv fiir Geschichte der Philosophic, Sonderhelf 58
(1976), 335-52.
Wittgenstein, Ludwig. The Philosophical Investigations. G. E. Anscombe
(trans.).
Wolff, R. P. "Kant's Debt to Hume Via Beattie," Journal of the History of
Ideas 21 (1960), 117-23.
Wolff, R. P. Kant's Theory of Mental Activity. Cambridge, Mass.: Harvard
University Press, 1963.
Wolff, R. P. (ed.). Kant. Garden City, N.Y.: Anchor Books, 1967.
Yolton, J. W. Perceptual Acquaintance from Descartes to Reid. Minneapolis:
University of Minnesota Press, 1984.
Young, J. M. "Kant's View of Imagination," Kant-Studien 79 (1988), 140-64.
 Index of Cited Passages

This index includes the passages quoted or cited from Kant. The location of the
citation precedes the semicolon; the location in this book follows it. Citations
from the Critique of Pure Reason are located by the standard A and B numbers.
Other citations are located by the volume and page numbers of the Akademie
edition.

Critique of Pure Reason

Aviii; 11 A40/B57; 48
Axii; 13, 62 A44/B61-62; 15
Axv; 16, 23 A50/B74; 13
Axvi; 65 A51/B75; 111
Axvii; 14, 224 A53/B77; 26, 206
A6/B10ff; 27 A54-55/B78-79; 11-12, 26, 206, 206, 269
A11-12/B25; 14 n. 3
A19/B33; 113, 237 n. 3, 244 n. 14 A55-56/B80-81; 13, 14, 234 n. 63, 259
A19-20/B32-34; 36 n. 1
A19-20/B34; 68 A56-57/B80-81; 25
A20/B34; 36, 37 A57/B82; 94
A20/B35; 39 A58-59/B82-84; 259 n. 8
A20-21/B35; 40 A68/B93; 74, 112
A21/B35; 30 A69/B94; 88
A21/B36; 36 A77/B103; 74, 103, 118, 245 n. 24
A22/B36; 39 A77-79/B103-4; 83, 209
A22/B37; 43, 44 A79/B105; 159
A23/B38; 46 A84/B116; 210
A24/B39; 48, 51 A86/B118-19; 12, 15
A26/B42; 39 A86-87/B119; 15
A26-27/B42-44; 19, 39, 51 A88/B121, B144; 173
A29; 37, 39 A90/B122; 142-43
A31-32/B47; 171 A90/B123; 144, 155
A36-37/B53-54; 141 A95; 243 n. 1
A39/B56; 141 A96; 73

283
284 INDEX OF CITED PASSAGES

A97;90 A248/B305; 184

A98-99; 134 A296/B352-53; 184
A99; 85-86, 149, 152, 201, 245 n. 24, 259 A298/B355; 185
n. 17 A299/B355; 224
A100; 78, 79 A301-2/B358; 219
A102; 201 A303/B359ff; 90
A103; 83, 107, 110, 153, 201 A320/B376; 66
A103-4; 209 A320/B376-77; 244 n. 14, 261 n. 30
A104; 254 n. 21 A334/B391; 265 n. 13
A104-5; 66, 71-74, 76, 77, 79, 94, 143, A339/B397; 185
247 n. 44 A340/B398; 190
A106; 81, 209 A346/B404; 190
A107; 100, 101, 102, 107, 187, 188 A348; 184, 187
A108; 107, 126, 201, 244 n. 14, 252 n. 57 A349; 184, 185, 186, 188, 195
A109; 73, 82 A350; 94, 100, 182, 185-86, 187, 188, 190
Alll; 143, 167 A351; 198
A112; 134 A352; 200
A113; 201 A353; 202
A116; 93, 108-9, 113, 119, 134, 135, 188,A354; 185, 201
249 n. 14 A354-55; 199, 201
A116-18; 118 A355; 189, 192, 200
A117a; 108, 113, 186, 249 n. 14 A356; 185, 199, 201-2
A118; 104, 119, 122 A359-60; 267 n. 28
A119-20; 149, 259 n. 17 A361; 195
A120; 81, 113, 151, 152 A361-62; 195
A120a; 151, 152, 158, 245 n. 24 A362; 195, 196
A120-21; 158 A362-64a; 124, 185, 196-98, 256 n. 44,
A121; 70, 79, 251 n. 43 48, 267 n. 32
A122; 93, 145, 167, 188 A364; 196
A123; 134 A365; 185, 192, 197, 268 n. 32
A123-24; 201 A381; 100, 187
A125; 145-46 A381-82; 189, 194
A128; 170 A398; 191
A133/B172; 209 A399; 191
A133-34; 122 A400; 182
A140/B179-80; 153 A402; 183-84, 188
A141/B180; 209 A403-4; 190
A143/B183; 195 A643/B670; 225
A165/B206; 44, 54 A645/B673; 224
A166/B208; 40, 239 n. 40 A647/B675; 224, 228-29
A179/B222; 102 A648/B676; 272 n. 55
A192/B237; 176 A649/B677; 223
A193/B238; 176, 177 A649/B677ff; 223
A195/B240; 176 A651/B679; 225
A224/B271; 151 A652/B680; 219, 221
A230-31/B300; 266 n. 19 A713/B741; 17
A241/B300; 185 A727-28/B755-56; 270 n. 19, 271 n. 52
A241-42; 184 A736-37/B764-65; 24
A242-43/B300-301; 184, 188 A783/B810; 17
 Index of Cited Passages 285
A783/B811; 14-15, 17 B144-45; 170
A848/B876; 11 B147; 261 n. 30
B149; 185
Bxl; 94, 249n. 15 B151; 151
Bxviii; 228 B151-52; 171
Bxxvi; 253 n. 5 B152; 11, 160
Bxxvii-xxix; 139 B153; 106
B2-3; 16
B157, 192
B3-4; 15
B157a; 193
B39; 171
B157-58a; 192
B40;50 B159; 122
B41; 51, 239 n. 42
B159-160; 155
B69-70; 249 n. 15
B160; 147, 156, 157, 160, 162, 163, 261 n.
B128-29; 89, 188
32
B130; 158, 209, 253 n. 5
B160a; 157-58, 162, 163
B131; 92
B160-61; 161
B131a; 201
B161; 147, 163, 164, 261 n. 33
B131-32; 188 B161a; 162
B132; 76, 93, 107, 113, 127, 135, 145,
B162; 156, 162
201, 249 n. 14, 253 n. 5
B167; 170
B132-34; 109, 119
B208; 244 n. 14
B133; 122, 126, 127, 134, 209
B234; 177
B133-34; 100, 187 B288; 185
B134; 72, 104, 107-8, 119, 249 n. 14, 252 B407; 186
n. 57, 254 n. 21
B409; 191, 199
B134a; 186 B410-11; 184, 264 n. 6
8135; 104, 119, 186
B411; 183-84,190
B137; 82, 86, 201, 261 n. 33
B411a; 184, 190, 192
B137-38; 157 B413; 100, 187
6138; 175
B415; 255 n. 23, 267 n. 28
B139; 82
B415-18n; 256 n. 44
B139-40; 94 B418a; 202
B140; 206 B420; 268 n. 42
B141; 87
B421-22; 188
B142; 87, 88
B422-23a; 139, 192, 193
B143; 89, 143

Gedanken von der wahren Shatzung der Lebendigen Krdfte

(Thoughts on the True Estimation of Living Forces)

AA 1:24-25; 31 AA 1:139-40; 31

De Mundi Sensibilis atque Intelligibilis Forma et Principiis

(Inaugural Dissertation)

AA 11:392-93; 35 A A 11:402-3; 52
AA 11:393; 36, 38 AA 11:404; 52
AA 11:402; 40, 47, 239 n. 37 AA 11:906; 38
286 INDEX OF CITED PASSAGES

Prolegomena zu einer jeden ktinftigen Metaphysik

(Prolegomena to Any Future Metaphysics)

AA IV:282; 71 AA IV:471; 11, 206

AA IV:298; 159 AA IV:496-503; 223-24

Kritik der Practischen Vernuft

(Critique of Practical Reason)

AA V:47; 20

Der Streit der Fakultaten

(The Conflict of the Faculties)

AA VII:34; 113-14

Anthropologie in pragmatischer Hinsicht

(Anthropology from a Pragmatic Point of View)

AA VII:121; 11 AA VII:140-41; 12
AA VII:127; 107 AA VII:155; 240 n. 50
AA VIM27-28; 255 n. 23 AA VII: 156-57; 36, 148
AA VII:135; 109, 156 AA VII:177; 261 n. 21
AA VII: 138; 150

Uber eine Entdeckung

(On a Discovery)

AA VIII:173; 35
AA VIII:221-23; 15-16, 38

Logik
(Logic)

AA IX: 10; 13 AA IX:91; 66

AA IX:11; 206 AA IX:92; 13
AA IX:33; 13 AA IX:92-93; 270 n. 19
AA IX:58; 221-22 AA IX:94; 13
AA IX:59; 221, 222 AA IX:98; 221
AA IX:60; 221, 271 n. 52 AA IX:141-42; 209, 210
AA IX:61; 222, 271 n. 52

Kant's Briefwechsel
(Correspondence)

AA XI:52; 114
 Index of Cited Passages 287
Kants handscriftlicher Nachlass
(Fragments)

AA XII:370; 27 AA XVHI:681; 242 n. 59

AA XV:308-9; 70 AA XX:206; 218
AA XVII:366; 40 AA XX:216; 223
AA XVIII:69; 42 AA XX:308; 268 n. 35, 268 n. 39
!"#$%&'()%#*+)*+#,*'--.%-)/+%0-'*1%
 General Index

Adickes, Erich, 61, 257 n.54 Apriority, 15-19

Allison, Henry, 4, 62, 94, 105, 169, 170- a priori^, 16—17
71, 221, 239 n.44, 244 n.12, 253 n.5, a priori^, 15
254 n.14, 257 n.54, 261-62 n.34, 262 a priorio, 15-16
n.55, 268 n.42 vs. innate, 37-38
Ameriks, Karl, 244 n.9, 257 n.54, 268 Aquila, Richard, 249 n.17, 252 n.58, 253
n.42 n.5
Amphiboly, 15 Arnauld, Antoine, 234 n.57
Analogies of Experience, 101-2, 168, Austin,!. L., 8-9
174-75 Axioms of Intuition, 54
Analytic of Concepts, 29, 61
Analyticity, 13, 25-27, 34, 54, 160, 166 Bach-Y-Rita, Paul, 57
and apperception, 166 Barrow, Isaac, 33
and Berkeley, 34 Beattie, James, 98, 100-01
and cognitive science, 212 Essay on Truth, 99
and intuition, 160 Beck, L. W., 241 n.57
and logic, 13 Behaviorism, 10
and perception, 54 Beiser, Frederick, 232 n.13, 16, 233 n.28,
and psychology, 25-26 244 n.8
Angell, R. B., 56 Beltrami, Eugenio, 7
Anthropology, 6, 11, 12, 107, 113, 148, Bennett, Jonathan, 4, 61, 186, 264 n.5,
150, 154 265 n.ll, 12, 267 n.29
Antinomies, 140 Berkeley, Bishop George, 33-35, 44, 47,
Apperception, 12, 21, 91-92, 106-7, 127, 93, 179, 240 n.45
133, 137, 139-42, 183, 186, 197, 200- on representation, 68
201 Theory of Vision, 33, 41
and the categories, 144, 166-67 Bever, T. G., 179
unity of, 96, 104-5, 108-9, 113, 115, Billman, D., 272 n.56
117-23, 133, 136, 138, 144-46, 164, Binding problem, 84-86
190, 193 Blackwell, Richard J., 266 n.22
Apprehension, 148-49, 151, 156 Block, Ned, 268 n.44
289
290 GENERAL INDEX

Bona Meyer, Jurgen, 6-7 and essence, 222

Brentano, Franz, 66 Locke on, 12
Brueckner, Anthony L., 236 n.87 necessary and sufficient conditions,
Bruner, Jerome, 169 206-7, 209-10, 213, 222
Buff on, Georges-Louis Leclerc, Comte origin of, 15
de, 34 and prototype, 208
Buroker, Jill, 241-42 n.58 relations among, 221
as rules, 82 ff., 209
Carey, Susan, 215, 255 n.30, 270 n.33 schemata, 109-10, 190, 153-54
Carnap, Rudolph, 8 of the self, 194
Carroll, Lewis, 131 spatiotemporal, 159-64, 168
Castaneda, Hector-Neri, 249 n.20 structure of, 221, 228
Categories, 89, 105, 127, 142, 146, 162- and synthesis, 80
64, 166, 171-73, 184 system of, 225-26
and apperception, 167 and understanding, 22
and cognition, 164 unity of, 224, 225
and experience, 172 Condillac, Etienne Bonnot de, 34, 68, 81,
and intuition, 155, 171 87, 148-50, 248 n.63, 260 n.23
and judgments, 144 Consciousness, 5, 107-9, 112-13, 115,
and knowledge, 143 122, 126-27
and perception, 157 personal identity, 196
and synthesis, 143 unity of, 166, 188-89, 195, 200-201
and understanding, 173 Content, 10-11, 113, 201
Categorization, 226-28 Cummins, Robert, 235 n.72
Causality, 176-77, 262 n.54, 55 Curley, Edwin, 254 n.ll, 255 n.32, 256
Cheselden, William, 33, 34 n.40
Churchland, Patricia, 85, 247 n.58
Clue to the Discovery of All Pure Con- D'Alembert, Jean Le Rond, 34
cepts of the Understanding. See Davidson, Donald, 233 n.33
Metaphysical Deduction Deduction, 64, 65
Cognition, 107, 115, 137-38, 141, 146-47, of the categories, 61, 62, 93, 96, 100,
163, 178-79, 184, 190-91, 193, 213, 122, 138, 140, 142, 144-47, 149, 155,
218, 221, 228 158, 163, 170, 172
a priori origins, 15 of consciousness of the self, 100
a priori sources, 18 and judgment, 165
features of, 19 and synthesis, 164
universal features of, 24, 25 and transcendental psychology, 63-65
Cognitive science, 28, 205-6, 212, 218, Dennett, Daniel, 112, 122, 235 n.72, 270
266 n. 18 n.22
Cognitive states, 117-19, 123, 129, 134, Descartes, Rene, 29, 32-33, 126, 127,
136, 144, 151, 187, 195, 196, 200, 244 181, 187, 189, 197
n.12 Cogito, 91-92, 95, 127
Concept, 66, 82, 213, 216, 217, 227, 229, common sense, 151
269 n.7, 271-72 n.52 on consciousness, 5
a priori, 15 La Dioptrique, 32-33
acquisition, 214-25, 218 natural geometry, 33, 35
classification, 214 DeVleeschauwer, H.-J., 174
and consciousness, 212 Dialectic, 184-85
empirical, 210-13, 221, 224-25, 229 Diderot, Denis, 34
 General Index 291
Dreyfus, Hubert L., 271 n.41 Gelman, Susan, 214-15, 226, 270 n.39
Dummett, Michael, 233 n.38 Geometry, 49, 242 n.59, 257-58 n.63
natural, 33
Barman, John, 241 n.58, 262 n.67 theories of, 7
Eberhard, Johann August, 15, 27, 38, 148 Geometry [Euclidean], 50-54, 56
Empirical Psychology, 12 and synthetic a priori, 50
Empiricism, 20 George, Rolf, 68, 81, 244 n.20, 245 n.23,
and judgment, 87-88 24, 33, 34, 260 n.23
Kant's view of, 79-80 Gibson, J.J., 72, 246/1.39
and representations, 67-68, 69-70, 77- Goldman, Alvin, 8, 9, 233 n.42
79 Gregory, Richard, 169
and spatial perception, 46-47 Grice, H. P., 128
Epistemology, and psychology, 8 Gueroult, Martial, 241 n.58
Eschenbach, Johann Christian, 34 Guyer, Paul, 4, 49, 62, 174, 234 n.65, 235
Essentialism, 272 n.62 n.70, 242 n.59, 246 n.41, 257-58
Experience, 6, 17 n.63, 259 n. 17, 263 n.57
and the a priori, 18 on the Second Analogy, 174-78
and cognition, 22
and skepticism, 28 Hamann, Johann Georg, 232 n. 14
and transcendental psychology, 23 Harper, W., 263 n.59
Hartley, David, 34
Falkenstein, Lome, 239 n.42 Hatfield, Gary, 13, 231 n.13, 232 n.20,
Feder, Johann Georg Heinrich, 40 25, 234 n.57, 235 n.75, 241 n.55, 242
Fichte, J. G.,27 n.70
First Introduction to the Critique of Judg- Helmholtz, Hermann von, 7, 169
ment, 223 Hempel, Carl Gustav, 220
Fischer, Kuno, 6 Henrich, Dieter, 4, 155-56, 169-71, 249
Fodor, J. A., 137, 179, 218, 270 n.21 n.22, 252 n.57
Form Herbart, Johann Friedrich, 8, 9, 40
as process, 36, 37, 44, 50-51, 52, 157- Herz, Marcus, 114
58, 160 Hintikka, Jaakko, 31, 237 n.3
as product, 36, 37 Hopkins, James, 232 n.25, 242 n.65
pure, 37-40 Hopkins, Jasper, 239 n.37
Forms of intuition, 7, 31, 39, 55, 60, 140- Hoppe, Hansgeorg, 246 n.41
41, 158, 168 Howell, Robert, 237 n.4, 244 n.14, 246
pure, 56 n.41
pure process, 49, 52, 160 Hughes, R. I. G., 197, 267 n.31, 32
pure product, 37, 43, 44-45 Hume, David, 20, 63, 95-96, 97, 99, 103,
space and time, 162-63, 165, 173 107-9, 113, 121, 124, 176, 178, 194,
space as, 47, 50 250-51 n. 38-41
Frege, Gottlob, 7-9 on beliefs in objects, 69-70
Freud, Sigmund, 137 on causation, 98, 101, 175-77
Friedman, Michael, 51, 54 on distinct existences, 100-02
Fries, Jakob Friedrich, 5, 7, 8, 22-23 on experience, 34
Functionalism, 111-12, 252 n.58 law of association, 69-70, 77-79
on mental unity, 50, 95
Garrett, Don, 251 n.40 on perception, 150
Garrett, Merrill, 179, 263 n.60 on personal identity, 97-99, 100-102,
Gehler, Johann, 148-49 128, 138
292 GENERAL INDEX

on representation, 69 Keil, Frank, 214-15, 228, 270 n.28, 35

Humphrey, Ted, 50, 242 n.60 Kemp Smith, Norman, 4, 21, 29, 30-31,
Husserl, Edmund, 8, 19 40, 42, 48, 61, 66, 75, 93, 97-98, 100,
118
Kitcher, Philip, 233 n.38, 41, 234 n.64,
Ideal of Reason, 217-28
235 n.61, 69, 236 n.83, 242 n.65, 267
Ideality thesis, 140-41
Identity of the self. See Personal identity
n.n
Knowledge, 13, 36, 66, 143, 217, 220-21
Imagination, 22, 151, 153-54
a priori, 15-18, 23, 51
and perception, 153
and experience, 16
and schemata, 154
and faculties, 19
and synthesis, 81, 83
and philosophy, 65
Inaugural Dissertation, 34, 35, 40, 42, 43,
requirements of, 26
47,50
of self, 194, 195
on sensory representations, 38
sources of, 121
on spatial perception, 35, 52
of space, 25
Inference, 219-20, 223, 228
transcendental, 15, 25
Innate, 15, 37-38
Korsgaard, Christine, 256 n.43
Introspection, 6, 10, 11, 12
Kowaleski, A., 34
Intuition, 36, 37, 38, 52-54, 66, 73, 104,
Kuehn, Manfred, 245 n.22, 259 «.13
109-12, 113-15, 141, 164-65, 170-71,
Kuhn, Thomas, 60
183, 190, 201-2. See also Forms of
intuition
Landau, Barbara, 226-27
and the categories, 158,171
Law of association, 69, 77-80, 81, 89,
construction of, 160
103, 152
objects of, 164
Leary, David, 232 n. 13
outer, 31, 48, 49, 54, 55, 158
Leibniz, Gottfried Wilhehn, 35, 43-44,
of the self, 191, 199, 203
45, 48-49, 53, 55, 57, 120, 126, 127,
as sensory representation, 36
138, 181, 185, 189, 191-92, 197, 241
and space, 32, 45
n.58
and space and time, 157
apperception, 106-7
synthesis of, 145
judgments, 218
on materialism, 198-99
James, William, 200 mental traces, 135-36
Janaway, Christopher, 249 n.13 monads, 204
Jessop, T. E., 34, 238 n. 16 on perception, 67, 149
Jolley, Nicholas, 254 n.ll, 17, 19 personal identity, 33-34, 124-25
Jones, Susan, 226-27 petite perceptions, 113-14, 134, 156
Judgment, 20, 86, 89-90, 110-11, 113, on representation, 70, 71
115, 145, 159, 164-65, 170, 186-88, on spatial perception, 33-34, 42
190-91, 217-18 unity of representation, 75
analytic, 27 Lewis, David, 128, 255 n.26, 29
and the categories, 144 Linnaeus, Carolus, 223
and law of association, 87 Locke, John, 12, 32, 120, 123-27, 132,
and reason, 20 138, 164, 197, 235 n.66
and representation, 88 common sense, 151
and synthesis, 165-66 on concepts, 18-19
temporal, 179 on judgment, 87
 General Index 293
on perception, 99, 149 Monads, 67
person, definition, 195 Murphy, G. L., 228
on personal identity, 128
on representation, 68, 71 Nadel, Lynn, 58
Logic, 8, 94, 233 «.39 Nagel, Thomas, 182, 194
in The Anthropology, 13 Nativism, 7, 37-38
and cognitive faculties, 23 Nature, Order of, 219, 221
Frege on, 7 Necessary and sufficient conditions, 211-
and judgment, 87 12
and psychologism, 9, 27 Necessity, 23-24
and psychology, 8, 11 Newell, Allen, 25
Quine on, 27 Newton, Isaac, 35, 48, 53, 55, 224
and understanding, 206 Noumenal, 139-40, 219
Logic, 66, 206, 210, 221
Lossius, J. C., 98
Lucas, Margery, 270 n. 17 O'Keefe, John, 58
Luneberg, R. K., 56 O'Neill, Onora, 83-84, 231 n.S, 247 n.54,
55
McNamara, T. P., 208
McRae, Robert, 241 n.58 Paralogisms, 25, 75, 100, 122, 126, 181-
Malebranche, Nicolas, 33, 68 83, 185-88, 190-91, 192-96, 197, 198,
Markman, Ellen, 214-15, 226, 270 n.39 199-203
Marr, David, 243 /i.72, 245 ».35, 262 n.42 Parfit, Derek, 128, 131-33
Materialism, 268 n.42 Parsons, Charles, 53, 54
Mattey, G. J.,238n.l6 Paton, H. J., 4, 61, 247 n.60, 259 n.6
Medin, Douglas, 208, 216, 226, 228 Patten, S.C., 249n. 17
Meerbote, Ralf, 231 n.S Penelhum, Terrence, 128
Melnick, Arthur, 179-80 Perception, 36, 113, 146, 149, 156, 160,
Memory, 14, 124-25, 126 163, 165, 166, 175, 179, 186, 241 n.50
Mental representations. See and cognition, 147
Representation construction of, 168
Mental states, 66 depth, 55
comental, 134-35, 136 and Euclidean space, 7
Leibniz on, 67 and images, 150
traces of, 135-36 J. J. Gibson on, 72
unity of, 203 of objects, 58
Mental unity, 98-102, 117-21, 123, 126- objects of, 152
27, 128, 129, 130, 133 and representation, 151
cognitive criterion of, 129, 131-32, 133- and sensibility, 22
35, 136 of space and time, 157
Metaphysical Deduction, 62, 63, 89, 143- and synthesis, 154
45, 173 synthesis of, 158
Metaphysical Exposition, 45 of the third dimension, 32-33
Metaphysical Foundations of Natural Sci- of time, 175
ence, 11, 86, 206, 223 unity of, 161, 171
Methodology, 24, 51, 54, 210 visual, 147-48, 151
Modularity, 137 Pereboom, Derk, 236 n.85
Molyneux, William, 8, 32, 34, 41, 42 Perry, John, 123, 128, 129-30, 253 n.S
294 GENERAL INDEX

Personal identity, 122, 127-28, 124-25, and judgment, 88

128, 131-33, 138, 195, 197 and law of association, 77-79
Petite perceptions. See Leibniz, Gottfried Leibniz on, 67
Wilhelm Lockeans on, 67-68
Phenomenal, 139-40, 219 object of, 70
Pippen, Robert, 105-6, 107, 234 n.57, 239 of objects, 80, 143
n.38, 242 n.69, 251 n.47, 53, 54, 253 and perception, 158
n.5 processes of, 84-85
Plato, 137 pure, 37
Politz, K. H. L., 150-51, 259-60 n.20 Reinhold on, 5
Porterfield, William, 147-48, 245 n.35 of space and time, 158, 170
Prauss, Gerold, 4 spatial, 42-44, 46-48, 55, 56, 57, 58
Principle of sufficient reason, 233 n.39 and synthesis, 74
Principles, 62, 63, 64, 142, 144, 146, 162, unit of, 76
170, 172-73, 174 unity of, 72-73, 82, 88
Prolegomena, 35, 38, 49, 71, 93 Richards, Robert J., 181
Psychologism, 4, 7, 8 Riemann, Georg Friedrich Bernhard, 7
definition of, 9 Rorty, Richard, 91, 248 n.\
Pylyshyn, Zenon, 137, 256 n.51 Rosch, Eleanor, 206-7
Rosenberg, Jay, 260 n.24, 264 n.10
Quine, W. V., 27, 214-15, 226, 229, 236 Rostenreich, Nathan, 94
n.82 Russell, Bertrand, 8
Quinton, Anthony, 128
Satura, Vladimir, 34, 238 n.23, 240 n.50,
Rational Psychology, 11, 12, 14, 181-83, 259-60 n. 20-22
186, 188-89, 198, 200, 203, 205 Schultz, J. G., 27
Reagan, Ronald, 46 Science, 222-24, 225
Reason, 51, 221, 223, 224-25, 228-29 and empirical psychology, 11
faculty of, 19-20 Scruton, Roger, 92, 248 n.2, 248 n.9
practical, 20 Searle, John, 182, 203
and systematicity, 225 Second Analogy, 64, 92, 103, 146, 174,
Reflexionen, 70 175-80
Refutation of Idealism, 92, 93, 94, 178 Self-ascription, 92-94, 126-27
Regulative Use of the Ideas of Reason, Self-knowledge. See Knowledge
217, 221 Self-observation, 6, 7
Reid, Thomas, 34, 98, 164, 169 Sellars, Wilfrid, 182, 235 n.77, 252 n.2,
distinction between sensation and per- 257 n.55, 263-64 n.4, 264 n.10, 266
ception, 68, 148 n.21
on perception, 150 Sensa, 36, 38, 40
Reinhold, Karl Leonhard, 5, 6, 7, 9, 22- Sensation, 36, 57, 148
23, 25 and reference, 70
Representation, 65, 70, 71, 90, 108-9, Sensationism, 68
112, 113, 115, 149, 154, 191, 217, 244 Sense impression, 57
n.14 Sensibility, 22, 151, 161
apriority of, 79-80 theory of, 50-52, 54
and concept, 82 Shoemaker, Sydney, 128, 255 n.26, 27
conceptual and sensory, 40 Simon, Herbert, 25
and form, 36 Simulacra theory, 71
and imagination, 81 Skepticism, 20, 121-22
 General Index 295
Sluga, Hans, 233 n.30, 32, 37 on perception, 149-50
Smith, Linda, 208, 226-27 on spatial perception, 34
Smith, Robert, 34 Third Paralogism, 123, 126
Solidity, 223-24 Thomson, J. J., 236 n.88
Soul, 191, 202 Thomson, Manley, 237 n.3
identity of, 196 Thoughts on the True Estimation of Liv-
and psychology, 12 ing Forces, 31
simplicity of, 199-200, 202 Time, 185, 267-68 n.32
and thought, 197-98 as the form of inner sense, 195-96
Space, 257 n.61 perception of, 175
theories of, 7 Transcendental, definition of, 184, 190
Spatial perception, 7, 33, 34, 40-46, 48- Transcendental Aesthetic, 25, 29, 30-31,
52, 54-60 35, 44, 45, 50, 54, 62-63, 140, 142,
transcendental psychology of, 31-32 146, 170, 170-71, 196, 203
Spelke, Elizabeth, 239 n.44, 243 n.75 Transcendental Deduction, 12, 103, 104,
Spontaneity, 122, 253 n.5 108-110, 115, 138, 159, 182, 186-87,
Starke, F. C., 34-35 190, 196, 200, 205
Sternberg, R. J., 208 Transcendental Exposition, 45, 50-54
Strawson, P. F., 3, 4, 9, 21, 23, 62, 92- Transcendental idealism, 21, 28, 49, 63,
93, 236/1.82, 249 n. 13, 266 n.28 140-41, 258 n.64
on transcendental idealism, 140 Transcendental Logic, 13, 23
Stroud, Barry, 236 n.88, 250-51 n.40 Transcendental psychology, 3-5, 11, 14,
Substance, 185 19, 21, 23-26, 28, 64-65
Synthesis, 73, 76-77, 82, 88, 90, 102-3, definition of, 21, 22
104, 105, 108, 109-12, 115, 120-21, and empiricism, 22, 44
124, 136, 137, 143, 146, 147, 153, and philosophy, 10
158, 165, 200-202, 210, 246 n.41, 254 and spatial perception, 44
n.14 Strawson on, 23
of apperception, 155 and twentieth-century psychology, 10
of apprehension, 151-52, 155, 158, 171 Treisman, Anne, 85, 247 n.56, 57
and concepts, 80, 159, 160, 161-64
definition of, 74-75 Understanding, 13, 22, 154, 158
figurative, 260-61 n.27 and logic, 8, 12, 206
and intuitions, 160 unity of, 94, 152
and perception, 158, 161-63, 168-69,
175 Vaihinger, Hans, 30, 40, 42, 61
and representation, 74, 82-83 Vendler, Zeno, 182, 194
of Reproduction in Imagination, 77-78 Vision. See Perception
rule governed, 82, 84 Voltaire, 34
synthetic connection, 117-22, 125, 126,
127, 129, 134, 141 Walker, Ralph C. S., 92, 236 n.87, 246
and time, 85-86 n.41
and unity, 96 Walsh, W. H., 4, 10, 232 n.12, 233 n.44,
and unity of apperception, 144-45 246 n.41
Synthetic a priori, 17 Watson, J. B., 10
Wheatstone, Charles, 55, 56
Tetens, Johann Nicolas, 67, 81, 87, 98, Wiggins, David, 128, 131, 255 n.33
105, 112, 126, 135, 148, 238 n.17, 247 Wilkerson, T. E., 92, 94, 248 n.2, 3, 7
n.46, 61 Williams, Bernard, 123, 253 n.9, 255 n.25
296 GENERAL INDEX

Wilson, Margaret, 181, 188, 200, 236 192, 247 n.52, 266 n.22
n.89, 240 n.45, 249 n.26, 254 n.18, Wolff, Robert Paul, 9, 14
263 n.2 on synthesis, 82-83
Wittgenstein, Ludwig, 8, 207
Wolff, Christian, 12, 67, 181, 186, 189, Young, J. M., 231 n.5